Roflumilast for Management of Chronic Obstructive Pulmonary Disease

Transcription

1 Roflumilast for Management of Chronic Obstructive Pulmonary Disease Submission to National Institute for Health and Clinical Excellence Single Technology Appraisal (STA) May 2011 Manufacturer MSD Ltd 1

2 Table Of Contents EXECUTIVE SUMMARY SECTION A DECISION PROBLEM DESCRIPTION OF TECHNOLOGY UNDER ASSESSMENT CONTEXT EQUITY AND EQUALITY STATEMENT OF THE DECISION PROBLEM SECTION B CLINICAL AND COST EFFECTIVENESS CLINICAL EVIDENCE Identification of studies Study selection Summary of methodology of relevant RCTs Critical appraisal of relevant RCTs Results of the relevant RCTs Meta-analysis Indirect and mixed treatment comparisons Non-RCT evidence Adverse events Interpretation of clinical evidence COST EFFECTIVENESS Published cost-effectiveness evaluations De novo analysis Clinical parameters and variables Measurement and valuation of health effects Resource identification, measurement and valuation Sensitivity analysis Results Validation Subgroup analysis Interpretation of economic evidence

3 SECTION C IMPLEMENTATION ASSESSMENT OF FACTORS RELEVANT TO THE NHS AND OTHER PARTIES (Indirect and mixed treatment comparisons)... Error! Bookma RCTs (i.e. criteria not provided as key eligibility criteria in Table B4)... Error! Bookma 8 REFERENCES APPENDICES... ERROR! BOOKM 9.1 Appendix 1: Roflumilast SPC... Error! Bookma 9.2 Appendix 2: Search strategy for section 5.1 (Identification of studies)... Error! Bookma 9.3 Appendix 3: Quality assessment of RCT(s) (section 5.4)... Error! Bookma 9.4 Appendix 4: Search strategy for section 5.7 (Indirect and mixed treatment comparisons)... Error! Bookma 9.5 Appendix 5: Quality assessment of comparator RCT(s) in section Appendix 6: Search Strategy for section 5.8 (Non-RCT evidence)... Error! Bookma 9.7 Appendix 7: Quality Assessment of non-rct(s) in section 5.8 (Non-RCT evidence)... Error! Bookma 9.8 Appendix 8: Search Strategy for section Error! Bookma 9.9 Appendix 9: Quality Assessment of adverse event data in section 5.9 (Adverse events)... Error! Bookma 9.10 Appendix 10: Search strategy for cost-effectiveness studies (section 6.1)... Error! Bookma 9.11 Appendix 11: Quality assessment of cost-effectiveness studies (section 6.1)... Error! Bookma 9.12 Appendix 12: Search strategy for section 6.4 (Measurement and valuation of health effects)... Error! Bookma 9.13 Appendix 13: Resource identification, measurement and valuation (section 6.5)... Error! Bookma 9.14 Appendix 14: M2-107 Study Details... Error! Bookma 9.15 Appendix 15: Additional Inclusion and Exclusion Criteria for Roflumilast 9.16 Appendix 16: Results of M2-124 and M Error! Bookma 9.17 Appendix 17: Results of the MTC Sensitivity Analyses... Error! Bookma 9.18 Appendix 18: Example calculation for average time in state... Error! Bookma 3

4 Tables Table A 1: Unit costs of technology being appraised 23 Table B1: Eligibility criteria used in search strategy 40 Table B2: List of relevant RCTs 43 Table B3: Comparative summary of methodology of the RCTs 46 Table B4: Eligibility criteria in the RCTs (see Appendix 15 for additional eligibility criteria) 50 Table B5: Characteristics of participants in the RCTs across randomised groups (Studies M2-124, M2-125, M2-127 and M2-128)* 53 Table B6: Characteristics of participants in the RCTs across randomised groups (Studies M2-111 and M2-112)* 55 Table B 7: Primary and secondary outcomes of the RCTs 57 Table B 8: Summary of statistical analyses in RCTs 61 Table B 9: Quality assessment results for RCTs 70 Table B 10: Patients Included in Efficacy Analyses for Roflumilast RCTs 73 Table B 11: Effects of Roflumilast on COPD Exacerbations in the RCTs 75 Table B 12: Proportion of patients experiencing an exacerbation 77 Table B 13: Change in FEV 1 from Baseline in Roflumilast RCTs 78 Table B 14: Effects of Roflumilast on HRQL, Symptom Control and Mortality 80 Table B 15: Baseline Characteristics of the Intention-to-Treat Population for the Pooled Studies (M2-124 and M2-125) 81 Table B 16: Mean Exacerbation Rates and Time to First and Second Exacerbation, With or Without LABA in the Pooled Data from Studies M2-124 and M Table B 17: Mean Changes in Pre- and Post-bronchodilator FEV 1 With or Without Concomitant LABA Therapy from the M2-124 and M2-125 Studies 84 Table B 18: Mean Exacerbation Rates and Time to First and Second Exacerbation, for Frequent and Infrequent Exacerbators from Pooled Data from Studies M2-124 and M Table B 19: Mean Changes in Pre- and Post-bronchodilator FEV 1, for Frequent and Infrequent Exacerbators from Pooled Data from Studies M2-124 and M Table B 20: Mean Rate of Exacerbations in patients with Chronic Bronchitis from Studies M2-111 and M

5 Table B 21: Change in pre-bronchodilator FEV 1 in patients with Chronic Bronchitis from Studies M2-111 and M Table B 22: Change in post-bronchodilator FEV 1 in patients with Chronic Bronchitis from Studies M2-111 and M Table B 23: Change in SGRQ Total Score from Baseline in Patients with Chronic Bronchitis from Studies M2-111 and M Table B 24: Eligibility criteria used in search strategy 92 Table B 25: Summary of Methodology - Trials Included in the MTC Analysis (i.e. those which included data on the total number of exacerbations and/or the mean annual rate of exacerbations) 95 Table B 26: Summary of Methodology - Additional Trials Included in the Sensitivity Analysis Examining the Binomial Outcome of Ever Having an Exacerbation Event (those trials which provided data on the proportion of patients with at least 1 exacerbation during the trial) 112 Table B 27: Summary of Methodology - Additional Trial Included in the Sensitivity Analysis Including Previously Unpublished Roflumilast Data (M2-111 Trial) 114 Table B 28: Summary of the trials used in the assessment of total number of exacerbations and/or the mean annual rate of exacerbations 116 Table B 29: Summary of the additional trials used in the binomial analysis (number of patients with 1 exacerbation) 117 Table B 30: Estimated Rate Ratios and 95% Confidence Intervals for the Relative Effects of Pairs of Treatments, Produced by the Random-Effects MTC Model Without Covariates 121 Table B 31: Outcomes from the Additive MTC Analysis 123 Table B 32: Absolute Treatment Effects Corresponding to the 10 Treatments for the Management of COPD, as Derived from the Primary MTC Analysis 125 Table B 33: Absolute Treatment Effects Corresponding to the 15 Treatments for the Management of COPD, as Derived from the Additive MTC Analysis 126 Table B 34: Adverse event frequency, organised by MedDRA system organ class and preferred term, as reported by Calverley et al Table B 35: Serious adverse event frequency ( 0.3%), as reported by Calverley et al Table B 36: Adverse events occurring in at least 2.5% of patients across M2-124 and M2-125, as reported by Calverley et al Table B 37: Adverse events occurring in at least 2% of patients across M2-127 and M2-128, as reported by Fabbri et al

6 Table B 38: Adverse events occurring in at least 2.5% of patients in M Table B 39: Adverse events occurring in at least 5% of patients across M2-112, as reported by Calverley et al Table B 40: Adverse Events Reported in the Post-hoc Analysis of M2-111 and M2-112 by Rennard et al 143 Table B 41: Overview of Identified Studies 153 Table B 42: Summary of data in abstracts identified in conference proceedings search 158 Table B 43: Severity of COPD, as defined by NICE CG101 (2010) and the GOLD classification (2009) 1; Table B 44: Key features of analysis 173 Table B 45: Therapy options for ICS-tolerant patients 176 Table B 46: Therapy options for ICS-intolerant/declining patients 176 Table B 47: Rate ratio of exacerbations for the common comparator (LAMA + LABA/ICS) vs. treatments, Mills et al., Table B 48: Inverted rate ratio of exacerbations from Table B 47 for treatments vs. the common comparator (LAMA + LABA/ICS), based on Mills et al., Table B 49: Summary of variables applied in the economic model (base case analyses) 181 Table B 50: Details of the studies in which HRQL is measured 195 Table B 51: Definition of exacerbations, from Rutten van Molken et al Table B 52: Utility Variables 203 Table B 53 : Cost of COPD drugs 206 Table B 54: Cost of maintenance - Disease state: Severe COPD 206 Table B 55: Cost of maintenance - Disease state: Very Severe COPD 207 Table B 56: Cost of community-treated exacerbation 208 Table B 57: National Schedule of Reference Costs Year : '2009/10' - NHS Trusts Non-Elective Inpatient (Long Stay) HRG Data 209 Table B 58: National Schedule of Reference Costs Year : '2009/10' - NHS Trusts Non-Elective Inpatient (Short Stay) HRG Data 210 Table B 59: Details of studies in which costs and resource use are measured 214 Table B 60: Monthly drug costs associated with the COPD therapies 227 Table B 61: List of health states and associated costs in the economic model 228 Table B 62: Parameters that were included in OWSA 231 Table B 63: List of distributions used in PSA for comparisons of interest 233 6

7 Table B 64: Results: Fully Incremental Analysis (ICS-tolerant patients, cohort starts in the Severe COPD state) 241 Table B 65: Results: Fully Incremental Analysis (ICS-intolerant/declining patients, cohort starts in the Severe COPD state) 242 Table B 66: Key economic comparisons summary 244 Table B 67: Summary of economic analysis: Comparison 1 (base case), LAMA + LABA/ICS + Roflumilast vs. LAMA + LABA/ICS 244 Table B 68: Summary of economic analysis: Comparison 2 (base case), LAMA + LABA + Roflumilast vs. LAMA + LABA 244 Table B 69: Markov trace: ICS-tolerant patients, LAMA + LABA/ICS + Roflumilast vs. LAMA + LABA/ICS 245 Table B 70: Markov trace: ICS-intolerant/declining patients, LAMA + LABA + Roflumilast vs. LAMA + LABA 247 Table B 71: QALY accrued over time: ICS-tolerant patients, LAMA + LABA/ICS + Roflumilast vs. LAMA + LABA/ICS 249 Table B 72: QALY accrued over time: ICS-intolerant/declining patients, LAMA + LABA + Roflumilast vs. LAMA + LABA 250 Table B 73: Number of exacerbations, accrued over time: ICS-tolerant patients, LAMA + LABA/ICS + Roflumilast vs. LAMA + LABA/ICS 251 Table B 74: Number of exacerbations, accrued over time: ICS-intolerant/declining patients, LAMA + LABA + Roflumilast vs. LAMA + LABA 252 Table B 75: Summary of QALY gain and number of exacerbations by health state: ICS-tolerant patients, LAMA + LABA/ICS + Roflumilast vs. LAMA + LABA/ICS 253 Table B 76: Summary of QALY gain and number of exacerbations by health state: ICS-intolerant/declining patients, LAMA + LABA + Roflumilast vs. LAMA + LABA 253 Table B 77: Summary of costs by health state: ICS-tolerant patients, LAMA + LABA/ICS + Roflumilast vs. LAMA + LABA/ICS 254 Table B 78: Summary of costs by health state: ICS-intolerant/declining patients, LAMA + LABA + Roflumilast vs. LAMA + LABA 254 Table B 79: Cost breakdown: ICS-tolerant patients, LAMA + LABA/ICS + Roflumilast vs. LAMA + LABA/ICS 254 Table B 80: Cost breakdown: ICS-intolerant/declining patients, LAMA + LABA + Roflumilast vs. LAMA + LABA 255 Table B 81: Summary base-case results 257 7

8 Table B 82: Fully incremental analysis, patients switch to 2nd line treatment upon disease progression: ICS-tolerant patients 267 Table B 83: Scenario analysis A: variation of number of exacerbations with the common comparator. ICS-tolerant patients 267 Table B 84: Scenario analysis B: Incorporation of a lung function benefit [+46ml]. ICS-tolerant patients 268 Table B 85: Scenario analysis C: variation of time on 1st line therapy. ICStolerant patients 268 Table B 86: Scenario analysis D: change of disutilities of exacerbations. ICStolerant patients 269 Table B 87: Scenario analysis E: 2nd line therapy is LAMA + LABA/ICS + Roflumilast for all patients. ICS-tolerant patients 269 Table B 88: Fully incremental analysis (patients switch to 2nd line only upon disease progression): ICS-intolerant/declining patients 270 Table B 89: Scenario analysis A: variation of number of exacerbations with the common comparator. ICS-intolerant/declining patients 270 Table B 90: Scenario analysis B: Incorporation of a lung function benefit [+46ml]. ICS-intolerant/declining patients 271 Table B 91: Scenario analysis C: variation of time on 1st line therapy. ICSintolerant/declining patients 271 Table B 92: Scenario analysis D: change of disutilities of exacerbations. ICSintolerant/declining patients 272 Table B 93: Scenario analysis E: 2nd line therapy is LAMA + LABA/ICS + Roflumilast for all patients. ICS-intolerant/declining patients 273 Table B 94: Subgroup analysis very severe COPD, ICS-tolerant patients 279 Table B 95: Subgroup analysis very severe COPD, ICS-tolerant patients, LAMA + LABA/ICS + roflumilast vs. LAMA + LABA/ICS 280 Table B 96: Subgroup analysis very severe COPD, ICS-intolerant/declining patients 280 Table B 97: Subgroup analysis very severe COPD, ICS-intolerant/declining patients, LAMA + LABA + roflumilast vs. LAMA + LABA 281 Table C 1: Eligible patient population 287 Table C 2: Current distribution of COPD therapies in UK clinical practice (presented for regimen of interest, as defined by Analysis 1 and 2) 288 Table C 3: Roflumilast market share uptake assumptions (within the population eligible to receive roflumilast) 288 8

9 Table C 4: Annual drug acquisition costs 289 Table C 5: Analysis 1: Estimated budget impact for roflumilast (year 1-5) 291 Table C 6: Analysis 2: Estimated budget impact for roflumilast (year 1-5) 291 Table C 7: Summary of Analysis 1 and Analysis Table A10 1: Search for Cost-effectiveness studies using Ovid MEDLINE (R ) and MEDLINE (R) In-Process Table A10 2: Embase search for cost-effectiveness studies Table A10 3: EconLIT search strategy for cost-effectiveness studies Table A10 4: NHS Economic Evaluation Database search strategy Error! Bookmark n Error! Bookmark n Error! Bookmark n Error! Bookmark n Table A11 1: Quality Assessment of Cost-Effectiveness Studies identified Table A11 2: Critical appraisal of methodology of reviewed cost-effectiveness studies Error! Bookmark n Error! Bookmark n Table A12 1: The search for measurement and valuation of health effects using Ovid MEDLINE (R ) and MEDLINE (R) In-Process Table A12 2: EMBASE search for measurement and valuation of health effects Table A12 3: NHS EED search for measurement and valuation of health effects Table A12 4: ECONLIT search for measurement and valuation of health effects Error! Bookmark n Error! Bookmark n Error! Bookmark n Error! Bookmark n Table A13 1: Search for Costs and Resource Use using Ovid MEDLINE (R ) and MEDLINE (R) In-Process Table A13 2:Search for Costs and Resource Use using Embase Table A13 3: Search for Costs and Resource Use using Econlit Table A13 4: Search for Costs and Resource Use using the NHS Economic Evaluation Database Error! Bookmark n Error! Bookmark n Error! Bookmark n Error! Bookmark n Table A14 1: Summary of methodology of M2-107 Table A14 2: Eligibility criteria in M2-107 Table A14 3: Characteristics of participants in M2-107 across randomised groups Table A14 4: Primary and secondary outcomes of M2-107 Table A14 5: Summary of statistical analyses in M2-107 Table A14 6: Quality assessment results for M2-107 Table A14 7: Patients Included in Efficacy Analyses for M2-107 Table A14 8: Effects of Roflumilast on Exacerbations in M2-107 Table A14 9: Change in FEV 1 from Baseline for M2-107 Error! Bookmark n Error! Bookmark n Error! Bookmark n Error! Bookmark n Error! Bookmark n Error! Bookmark n Error! Bookmark n Error! Bookmark n Error! Bookmark n 9

10 Table A14 10: HRQL, Symptom Control, Mortality Error! Bookmark n Table A14 11: Adverse events occurring in at least 5% of patients across M2-107, as reported by Rabe et al 20 Error! Bookmark n Table A15 1: Additional Inclusion and Exclusion Criteria for Roflumilast RCTs Error! Bookmark n Table A16 1: Patients Included in Efficacy Analyses for M2-124 and M2-125 Table A16 2: Effect of Roflumilast on exacerbations for M2-124 and M2-125 Table A16 3: Change in FEV 1 from Baseline in M2-124 and M2-125 Table A16 4: Effects of Roflumilast on HRQL, Symptom Control and Mortality Error! Bookmark n Error! Bookmark n Error! Bookmark n Error! Bookmark n Table A17 1: Probability each of the 10 treatments for management of COPD is best (in terms of reducing exacerbations), obtained on the basis of the random-effects MTC model without covariates. Table A17 2: Probability each of the 15 treatments for management of COPD is best (in terms of reducing exacerbations), obtained on the basis of the additive main effects MTC model without covariates. Table A17 3: Estimated rate ratios and associated 95% confidence intervals for the relative effects of pairs of treatments for the management of COPD, produced by the fixed-effect MTC model without covariates. Table A17 4: Absolute treatment effects corresponding to the 10 treatments for the management of COPD, as derived from the fixed-effect MTC model without covariates. Table A17 5: Probability that each of the 10 treatments for management of COPD is best, obtained on the basis of the fixed-effect MTC model without covariates. Table A17 6: Estimated relative risks and 95% confidence intervals for all pairs of treatments produced by the random-effects MTC model without covariates. (Binomial Model) Table A17 7: Absolute Treatment Effects Obtained from the Random-Effects MTC Model. (Binomial Model) Table A17 8: Probability each of the 10 treatments for management of COPD is best, obtained on the basis of the random-effects MTC model without covariates. (Binomial Model) Table A17 9: Estimated rate ratios and associated 95% CI for the relative effects of pairs of treatments for the management of COPD, produced by Error! Bookmark n Error! Bookmark n Error! Bookmark n Error! Bookmark n Error! Bookmark n Error! Bookmark n Error! Bookmark n Error! Bookmark n 10

11 random effects MTC involving previously unpublished roflumilast data (M2-111 trial) Table A17 10: Absolute treatment effects obtained from the random-effects MTC model including previously unpublished roflumilast data (M2-111 trial) Table A17 11: Probability each of the 15 treatments for management of COPD is best (in terms of reducing exacerbations), obtained on the basis of the additive main effects MTC model without covariates. This table includes previously unpublished roflumilast data (M2-111 trial) Table A17 12: Estimated rate ratios and 95% CI for the relative effects of specific pairs of treatment combinations, produced by the additive main effects model, including previously unpublished roflumilast data (M2-111 trial) Table A17 13: Random effects, direct (head to head) evaluation using DerSimmonian Laird random-effects model. Exacerbation rates Table A17 14: Random effects, direct (head to head) evaluation using DerSimmonian Laird random-effects model. Relative Risk of ever having at least one exacerbation event (Binomial Model). Table A17 15: Model Fit (random effects primary analysis) Table A17 16: Parameter estimates for the random-effects MTC primary analysis model Error! Bookmark n Error! Bookmark n Error! Bookmark n Error! Bookmark n Error! Bookmark n Error! Bookmark n Error! Bookmark n Error! Bookmark n Table A18 1: Equations for time in the Severe COPD state Error! Bookmark n Figures Figure 1: Asthma and COPD. Adapted from GOLD Figure 2: NICE Recommended Maintenance Therapy for Severe / Very Severe COPD (Adapted from NICE CG101, Algorithm 2a) 29 Figure 3: Potential Addition of Roflumilast to NICE Recommended Maintenance Therapy for Severe / Very Severe COPD 30 Figure 4: Flow Diagram of Included and Excluded Publications 41 Figure 5: Participant Flow Diagram for M2-124 Figure 6: Participant Flow Diagram for M Figure 7: Participant Flow Diagram for M2-127 Figure 8: Participant Flow Diagram for M Figure 9: Pariticpant Flow Diagram for M2-111 Figure 10: Pariticpant Flow Diagram for M Figure 11: Flow Diagram of Included and Excluded Studies in the MTC 93 11

12 Figure 12: The network of 10 treatments included in the MTC analysis of exacerbations 115 Figure 13: Flow diagram of the search for cost-effectiveness studies 151 Figure 14: Markov Model Structure 164 Figure 15: Detailed model scheme 165 Figure 16: Time to disease progression in patients starting in the Severe COPD state (illustrative example; no lung function benefit considered) 171 Figure 17: Time to disease progression in patients starting in the Severe COPD state (illustrative example; considering a lung function benefit due to the treatment) 172 Figure 18: Flow diagram for searches of HRQL/utility studies 193 Figure 19: Flow diagram for searches of cost and resource use studies 212 Figure 20: Cost-efficiency-frontier analysis (ICS-tolerant patients, cohort starts in the Severe COPD state) 242 Figure 21: Cost-efficiency-frontier analysis (ICS-intolerant/declining patients, cohort starts in the Severe COPD state) 243 Figure 22: Markov trace: ICS-tolerant patients, LAMA + LABA/ICS + Roflumilast 246 Figure 23: Markov trace: ICS-tolerant, LAMA + LABA/ICS 246 Figure 24: Markov trace: ICS-intolerant/declining patients, LAMA + LABA + Roflumilast 248 Figure 25: Markov trace: ICS-intolerant/declining patients, LAMA + LABA 248 Figure 26: Cost breakdown: ICS-tolerant patients, LAMA + LABA/ICS + Roflumilast vs. LAMA + LABA/ICS 255 Figure 27: Cost breakdown: ICS-intolerant/declining patients, LAMA + LABA + Roflumilast vs. LAMA + LABA 256 Figure 28: Tornado diagram: LAMA + LABA/ICS + Roflumilast vs. LAMA + LABA/ICS 259 Figure 29: Tornado diagram: LAMA + LABA + Roflumilast vs. LAMA + LABA 261 Figure 30: Cost-QALY scatterplot: ICS-tolerant patients, LAMA + LABA/ICS + Roflumilast vs. LAMA + LABA/ICS, 1000 runs 262 Figure 31: Incremental CE scatterplot: ICS-tolerant patients, LAMA + LABA/ICS + Roflumilast vs. LAMA + LABA/ICS, 1,000 runs 262 Figure 32: Cost-effectiveness acceptability curve (CEAC): ICS-tolerant patients, LAMA + LABA/ICS + Roflumilast vs. LAMA + LABA/ICS, 1,000 runs 263 Figure 33: Cost and QALY scatterplot: ICS-intolerant/declining patients, LAMA + LABA + Roflumilast vs. LAMA + LABA, 1,000 runs

13 Figure 34:Incremental CE scatterplot: ICS-intolerant/declining patients, LAMA + LABA + Roflumilast vs. LAMA + LABA, 1,000 runs 264 Figure 35: Cost-effectiveness acceptability curve: ICS-intolerant/declining patients, LAMA + LABA + Roflumilast vs. LAMA + LABA, 1,000 runs 264 Figure 36: Markov Model Structure for patients starting treatment in the Very Severe state 278 Figure 37: Cost-efficiency-frontier analysis, Subgroup analysis very severe COPD, ICS-tolerant patients 279 Figure 38: Cost-efficiency-frontier analysis, Subgroup analysis very severe COPD, ICS-intolerant/declining patients

14 Executive summary Background and unmet need There are approximately 900,000 people diagnosed with COPD in the UK and a further 2 million are estimated to have the disease but remain undiagnosed 1. COPD is a major health issue in the UK and is the fifth leading cause of death, currently accounting for 26,000 to 30,000 deaths per year 2-4. It is the second most common reason for emergency hospital admissions in the UK, with 1 in 8 emergency admissions each year being due to COPD 5. COPD places a significant financial burden on the NHS, accounting for over 1 million hospital bed days and 1.4 million general practice consultations each year 5;6. The Department of Health in 2004 estimated that COPD accounts for more than 800 million in direct healthcare costs each year 6. COPD is a chronic, progressive, inflammatory disease characterised by airflow obstruction which is not fully reversible 1. COPD is primarily caused by cigarette smoking, although other factors, particularly occupational exposures, may also contribute to its development. The term COPD encompasses patients presenting with chronic bronchitis, emphysema or a combination of both 2;5 with sufferers typically experiencing a prolonged decline in lung function punctuated by acute exacerbations of symptoms. The subset of patients with chronic bronchitis are readily characterised by the presenting signs of chronic cough and sputum production. Whilst COPD is frequently characterised by breathlessness, exacerbations are important events in the natural history of the disease, with both their frequency and severity being associated with a significant worsening of prognosis 1;7. Studies have demonstrated that patients experiencing frequent exacerbations have a more rapid decline in lung function 8 and significantly worse health related quality of life 9;10. An increase in the frequency of severe exacerbations has also been observed to lead to an increase in mortality 7. In addition, there is evidence that the subset of COPD patients with associated chronic bronchitis are at a higher risk of frequent exacerbations, including severe exacerbations requiring hospitalisation 11. In COPD, exacerbations account for the greatest burden on healthcare systems, therefore reducing their severity and frequency is an important therapeutic goal 4. Exacerbations of COPD are estimated to account for 15.9% of hospital admissions, at a cost to the NHS of more than 253 million per year

15 The chronic, systemic inflammation which underlies COPD is principally caused by an increase in CD8+ T-lymphocytes, macrophages and neutrophils and is different from the localised inflammation underlying asthma, which is predominately caused by an increased number of CD4+ T-lymphocytes and eosinophils 13. This difference in fundamental pathology requires that the treatment of COPD and asthma have clearly distinct approaches to clinical management. Currently used bronchodilators do not specifically influence the inflammatory process underlying COPD and while ICS have anti-inflammatory action, their efficacy in COPD is reduced compared to that demonstrated in asthma patients 13. This has driven the need for scientific research to identify novel therapies which specifically target COPD inflammatory processes. Current treatment Current guidelines for the treatment of patients with COPD recommend the addition of therapies in a stepwise manner according to patients' needs, with the aim of controlling symptoms and preventing exacerbations 1;13. NICE has recently updated the COPD Clinical Guideline (CG101) which stressed the importance of multidimensional assessment of the disease, as opposed to simple categorisation of disease severity through measure of lung function (Section 6.9 of CG101). Although recently published, the guidelines were produced prior to the availability of roflumilast in the UK. The guideline recommends that patients with severe COPD (FEV 1 < 50% predicted) receive a long-acting bronchodilator, either LABA (with ICS) or LAMA. If patients continue to exacerbate or remain breathless, it is recommended that patients progress to triple therapy with LAMA plus LABA / ICS. In patients who cannot tolerate or decline ICS, a combination of the two bronchodilators is recommended (LABA plus LAMA). The Guideline Development Group (GDG) has been prepared to recommend virtually all combinations of therapy in the face of an acknowledged poor quality of evidence 1. Despite the availability of these treatment options a significant unmet need still exists in COPD patients with chronic bronchitis, many of whom still experience frequent exacerbations despite treatment with currently recommended therapies. 11;14;15 A recent analysis found that 33% and 47% of patients with severe and very severe 15

16 COPD, respectively, suffered from two or more exacerbations within the first year of follow-up 14. The Technology Roflumilast (Daxas ) is a highly selective phosphodiesterase-4 (PDE-4) inhibitor designed to target both the systemic and pulmonary inflammation specifically associated with COPD 16. Roflumilast is the first in this new class of COPD treatments to be granted a licence that specifically includes data on reduction in the frequency of COPD exacerbations, and thus represents a step-change in the clinical management of the disease. Roflumilast is a once daily, oral therapy which is licensed for use as maintenance therapy within a well defined subgroup of COPD patients, specifically those whose severe COPD (FEV 1 post-bronchodilator less than 50% predicted) is associated with chronic bronchitis and who have a history of frequent exacerbations despite maintenance bronchodilator treatment 16. As described above, such patients are easily identifiable in clinical practice. The licensed dose of roflumilast for maintenance treatment of COPD is 500 µg once daily, administered as one film-coated tablet. Roflumilast is available in packs of 30 or 90 tablets, with acquisition costs of and , respectively. Clinical Effectiveness of Roflumilast The randomised controlled trials for roflumilast demonstrate that its main benefit is a significant reduction in COPD exacerbations within the licensed patient population, compared to placebo, both alone and in combination with bronchodilator therapy. In addition, the randomised clinical trials also demonstrated significant improvement in lung function. Analyses have also demonstrated that the benefits observed with roflumilast occur independently of concomitant ICS use. 17 The two pivotal, one year studies (M2-124 and M ), which assessed the efficacy of roflumilast in the patient population for which it is now licensed, showed that roflumilast significantly reduced the rate of moderate or severe exacerbations by 16.9% compared to placebo (rate ratio 0.83 [95% CI 0.75 to 0.92]; p=0.0003). In patients who received concomitant LABA therapy during the trials, the addition of roflumilast reduced the rate of moderate or severe exacerbations by 20.7% compared to LABA alone (RR 0.79 [95% CI 0.69 to 0.91]; p=0.0011), which 16

17 demonstrates that the efficacy of roflumilast is consistent with or without concomitant bronchodilator therapy. Other studies in the phase III clinical trial programme provide further evidence regarding the mechanism of action of roflumilast as a novel and distinct treatment for COPD The change from baseline in pre-bronchodilator FEV 1 was a co-primary endpoint in the pivotal trials, and a mean improvement of 48 ml (95% CI 35 to 62 ml; p<0.0001) was observed with roflumilast compared to placebo in the pooled data set. A similar improvement of 46 ml (95% CI 29 to 64 ml; p<0.0001) was also observed in patients who received concomitant LABA therapy compared to LABA alone. Significant lung function benefit has been demonstrated with roflumilast compared to placebo. As would be expected with an anti-inflammatory agent, due to the differing mode of action of roflumilast, improvements in lung function were not as great as those seen with a bronchodilator. Nevertheless, the additional improvement of roflumilast on top of a bronchodilator is expected to provide clinically meaningful benefit to patients with severe COPD. This additional improvement in lung function on top of bronchodilators has also been demonstrated in two 6 month trials 19, where roflumilast improved the change in pre-bronchodilator FEV 1 from baseline by 49 ml (95% CI 27 to 71; p<0.0001) on top of salmeterol (LABA) and by 80 ml (51 to 110; p<0.0001) on top of tiotropium (LAMA). An independently performed mixed treatment comparison meta-analysis 22 was conducted using published literature to assess the efficacy of COPD combination therapies in reducing exacerbations and provides comprehensive information regarding the efficacy of roflumilast in treatment combinations which were not studied in the clinical trial programme. This analysis demonstrated that COPD treatments are consistently effective in reducing exacerbation rates. The analyses also indicate that adding one treatment to one or more other treatments in combination will provide additional benefit. With regard to treatment combinations including roflumilast, the analysis suggests that adding roflumilast to other therapies, compared to adding nothing, provides benefit in terms of reducing exacerbations (approximately 15% reduction in exacerbation rate). Cost Effectiveness of Roflumilast 17

18 A Markov state-transition, cost-utility, economic model was developed to assess the long-term costs and health outcomes associated with roflumilast for the maintenance treatment of severe or very severe COPD (FEV 1 < 50% predicted), associated with chronic bronchitis in adult patients with a history of frequent exacerbations as add on to bronchodilator treatment. It was designed with sufficient flexibility to allow the consideration of the use of roflumilast in several positions in the treatment pathway, and against comparators used in routine clinical practice, via a fully incremental analysis. Functionality also included the ability to assess progression to subsequent lines of treatment. Specifically, the model was designed to capture the clinical benefits of roflumilast which have been demonstrated in clinical trials: a lower relative rate of exacerbations, and; functionality to explore improvement in lung function, as measured by post-bronchodilator FEV 1. The economic analyses were based on the cohort of patients treated with LABA in the subset of the M2-124 and M2-125 clinical trials. 18 Results from these studies support the efficacy of roflumilast when added to a bronchodilator in improving lung function and reducing exacerbations. The economic analyses are supplemented through the results of the independently conducted MTC (described above) in order to account for relative treatment effects where data from head-to-head trials were not available. 22 Analyses were conducted from an NHS perspective, and were based on a 30 year time horizon. Two fully incremental analyses were performed for patients with severe COPD; one which included ten possible treatment combinations for patients able to receive ICS, and a second assessing the six possible treatment combinations for patients who do not tolerate or decline ICS. A range of sensitivity analyses were also performed on the major comparisons within the analyses. The results of the first fully incremental cost-effectiveness analysis (conducted in patients who are able to receive ICS) demonstrate the cost-effectiveness of roflumilast when used as an add-on to the current recommended standard of care, as described in NICE CG 101, of triple therapy with LAMA + LABA / ICS. The use of roflumilast as an add-on to triple therapy is associated with an incremental costeffectiveness ratio (ICER) of 16, For those patients who do not tolerate or decline ICS, the second fully incremental analysis demonstrates that the addition of roflumilast to the recommended standard 18

19 of care of LAMA + LABA is a cost-effective strategy, associated with an ICER of 13, The clinical data, along with the cost-effectiveness analyses, identify other positions in the treatment pathway in which roflumilast confers significant benefit. This submission focuses on those areas where roflumilast is expected to be used in clinical practice: namely, as an add-on to the standard of care. Conclusion There are approximately 900,000 people diagnosed with COPD in the UK, with a further 2 million estimated to remain undiagnosed 1. COPD is a major health issue in the UK and is the fifth leading cause of death, currently accounting for 26,000 to 30,000 deaths per year 2-4. The Department of Health in 2004 estimated that COPD accounts for more than 800 million in direct healthcare costs each year 6. COPD is a systemic inflammatory condition with an underlying pathophysiology that differs from asthma. In COPD patients with chronic bronchitis, the disease is characterised by breathlessness, cough and sputum production and exacerbations. Roflumilast treats the underlying cause of the disease through acting on a COPD specific inflammatory pathway, mediated by PDE-4. The evidence presented within this submission demonstrates that roflumilast is effective at reducing moderate or severe exacerbations and provides additional lung function benefits when used in addition to a bronchodilator in severe and very severe COPD patients (FEV 1 < 50% predicted). The results of an independently conducted mixed treatment comparison suggest that the addition of roflumilast to any baseline standard of care (including triple therapy) is effective in reducing the exacerbation rate. Roflumilast, when added on to the recommended standard of care of triple therapy (or dual bronchodilator therapy in patients who do not tolerate or decline ICS), is a clinically- and cost-effective novel treatment that will help meet the unmet need for effective treatment options in this difficult to treat subgroup of COPD patients in the UK. Effectively reducing exacerbations in this defined patient population will have a significant impact on reducing healthcare costs in England and Wales in the future. 19

20 Section A Decision problem Manufacturers and sponsors will be requested to submit section A in advance of the full submission (for details on timelines, see the NICE document Guide to the single technology appraisal (STA) process A (draft) summary of product characteristics (SPC) for pharmaceuticals or information for use (IFU) for devices, a (draft) assessment report produced by the regulatory authorities (for example, the European Public Assessment Report (EPAR)), and a (draft) technical manual for devices should be provided (see section 9.1, appendix 1). 1 Description of technology under assessment 1.1 Give the brand name, approved name and, when appropriate, therapeutic class. For devices, provide details of any different versions of the same device. Daxas (roflumilast) 500 microgram tablets. Therapeutic class: Drugs for obstructive airway diseases, Other systemic drugs for obstructive airway diseases. 1.2 What is the principal mechanism of action of the technology? Roflumilast is a highly selective phosphodiesterase 4 (PDE4) inhibitor, a non-steroid, anti-inflammatory agent designed to target both the systemic and pulmonary inflammation associated with COPD. The mechanism of action is the inhibition of PDE4, a major cyclic adenosine monophosphate (camp)-metabolizing enzyme found in structural and inflammatory cells important to the pathogenesis of COPD. Roflumilast targets the PDE4A, 4B and 4D splicing variants with similar potency in the nanomolar range. The affinity to the PDE4C splicing variants is 5 to 10-fold lower. This mechanism of action and the selectivity also apply to roflumilast N oxide, which is the major active metabolite of roflumilast. 1.3 Does the technology have a UK marketing authorisation/ce marking for the indications detailed in this submission? If so, give the date on which authorisation was received. If not, state current UK regulatory status, with relevant dates (for example, date of application and/or expected approval dates). EMEA marketing authorisation for roflumilast was received on 5 th July

21 1.4 Describe the main issues discussed by the regulatory organisation (preferably by referring to the [draft] assessment report [for example, the EPAR]). If appropriate, state any special conditions attached to the marketing authorisation (for example, exceptional circumstances/conditions to the licence). Issues discussed by the regulators in the risk-benefit assessment of roflumilast for management of COPD were noted in the CHMP assessment report and included 23 : Current first line therapy (LABA + ICS) was not included in the clinical trials with roflumilast. The marketing authorisation holder has committed to conduct a controlled study to evaluate the use of roflumilast as add-on to LABA and ICS. The effect of roflumilast on primary outcomes was significant yet modest, however it was noted that this modest benefit may be of interest for severe to very severe COPD patients and that the effect was reached on top of bronchodilators. This resulted in the licence for roflumilast in a restricted group of COPD patients, ie those with severe COPD, chronic bronchitis and frequent exacerbations. The side effect of weight loss and the possible risk of triggering suicide were particularly noted by the CHMP, as was the fact that the rate of AEs leading to withdrawal was greater in the roflumilast group than placebo. The marketing authorisation holder will conduct a long-term comparative observational safety study to further assess the risks associated with use of roflumilast. [The SPC for Daxas advises that patients check their body weight regularly, and that Daxas should be discontinued in the event of an unexplained and clinically concerning weight decrease. The SPC also states that Daxas is not recommended in patients with a history of depression associated with suicidal ideation or behaviour, and patients should notify their prescriber of any changes in behaviour or mood and of any suicidal ideation.] The CHMP noted the medical need for new treatments for COPD, and considered this in their assessment of the risk-benefit balance of roflumilast. Based on the reduction in exacerbation rates and improvement in lung function in patients with severe or very severe COPD, the use of roflumilast was justified in the specific subgroup of patients outlined within the licensed indication. 21

22 The conditions to the licence for roflumilast are that all healthcare professionals expected to prescribe roflumilast must be provided with an educational pack, including a SPC and patient information leaflet, educational material and patient cards. Further details of these materials can be found in the annex to the EPAR (Conditions imposed on member states for safe and effective use) What are the (anticipated) indication(s) in the UK? For devices, provide the (anticipated) CE marking, including the indication for use. Roflumilast is indicated for maintenance treatment of severe chronic obstructive pulmonary disease (COPD) (FEV1 post-bronchodilator less than 50% predicted) associated with chronic bronchitis in adult patients with a history of frequent exacerbations as add on to bronchodilator treatment. 1.6 Please provide details of all completed and ongoing studies from which additional evidence is likely to be available in the next 12 months for the indication being appraised. No additional evidence is expected to be available in the next 12 months. 1.7 If the technology has not been launched, please supply the anticipated date of availability in the UK. Roflumilast has been available in the UK since 1st September Does the technology have regulatory approval outside the UK? If so, please provide details. Yes, roflumilast has a marketing authorisation valid throughout the EU, and also has regulatory approval in Norway, Iceland and Liechtenstein. 1.9 Is the technology subject to any other form of health technology assessment in the UK? If so, what is the timescale for completion? The Scottish Medicines Consortium has completed a review of roflumilast for maintenance treatment of severe COPD. MSD intend to resubmit to the Scottish Medicines Consortium in For pharmaceuticals, please complete the table below. If the unit cost of the pharmaceutical is not yet known, provide details of the anticipated unit cost, including the range of possible unit costs. 22

23 Table A 1: Unit costs of technology being appraised Pharmaceutical formulation Film coated tablet Acquisition cost (excluding VAT) 30 day pack day pack Method of administration Doses Dosing frequency Average length of a course of treatment Average cost of a course of treatment Anticipated average interval between courses of treatments Anticipated number of repeat courses of treatments Dose adjustments Oral 500 micrograms Once daily Continuous treatment One month (30 days) of treatment costs Not applicable Not applicable None 1.11 For devices, please provide the list price and average selling price. If the unit cost of the device is not yet known, provide details of the anticipated unit cost, including the range of possible unit costs. Not applicable 1.12 Are there additional tests or investigations needed for selection, or particular administration requirements for this technology? None. (Roflumilast requires post-bronchodilator FEV 1 testing to be conducted prior to its use, however this test is recommended for all patients with COPD in NICE Clinical Guideline 101.) Is there a need for monitoring of patients over and above usual clinical practice for this technology? No, although patients are advised to monitor for weight loss and changes in mood, sleep or suicidal ideation What other therapies, if any, are likely to be administered at the same time as the intervention as part of a course of treatment? Roflumilast is licensed for use in addition to current maintenance bronchodilator therapy in severe COPD patients with concurrent chronic bronchitis and frequent 23

24 exacerbations. In practice, a variety of interventions are used as standard of care for patients with severe COPD. Roflumilast may be administered in combination with: Long-acting β 2 agonists (LABA) salmeterol; formoterol LABA/corticosteroid combination inhaler budesonide/formoterol; fluticasone/salmeterol; Long-acting muscarinic antagonists (LAMA) tiotropium LAMA + LABA/corticosteroid combination inhaler LAMA + LABA (where ICS is not tolerated) Short-acting bronchodilators (SABA and SAMA) are also likely to be used by patients as required. 24

25 2 Context In this background section the manufacturer or sponsor should contextualise the evidence relating to the decision problem. 2.1 Please provide a brief overview of the disease or condition for which the technology is being used. Include details of the underlying course of the disease. An estimated 3 million people have COPD in the UK. About 900,000 have diagnosed COPD, and an estimated 2 million people have COPD which remains undiagnosed. 1 COPD is the fifth leading cause of death in the UK, accounting for around 26-30,000 deaths each year, 2-4;25 and is the second largest cause of emergency admissions to hospital. 2 In total, COPD accounts for more than 800 million in direct health care costs, with more than half of these costs relating to the services provided in secondary care. 6 COPD is a chronic, progressive, inflammatory disease characterised by airflow obstruction which is not fully reversible. 1 The term COPD encompasses patients presenting with chronic bronchitis and/or emphysema; 2;4-6;25 roflumilast is licensed for use only in those who have concurrent chronic bronchitis. COPD is primarily caused by cigarette smoking, although other factors, particularly occupational exposures, may also contribute to its development. Although COPD can co-exist with asthma, the underlying chronic airway inflammation characterising the two diseases is very different and thus, unlike asthma, COPD is not fully reversible (Figure 1). 25

26 Figure 1: Asthma and COPD. Adapted from GOLD 13 ASTHMA Sensitising agent (e.g. allergens) COPD Noxious agent (e.g. tobacco smoke) Asthmatic airway inflammation CD4+ T-lymphocytes and Eosinophils COPD airway inflammation CD8+ T-lymphocytes, Macrophages and Neutrophils Reversible Airflow limitation Not fully reversible Common symptoms of COPD are breathlessness, cough, sputum, wheeze and chest tightness. 3 Exacerbations often occur, when there is a rapid and sustained worsening of the patient's symptoms beyond normal day-to-day variations. 1 Studies have demonstrated that patients experiencing frequent exacerbations have a more rapid decline in lung function 8 and significantly worse health related quality of life, 9;10;26;27 and mortality has been observed to increase with the frequency of severe exacerbations. 7 The risk of death during a hospital-treated exacerbation was reported as 7.7% in 2008, based on a sample of 9,716 admissions. 28 Classification of the severity of COPD has classically been conducted based on assessment of lung function alone, however the recent NICE Clinical Guideline for COPD states "the severity of COPD from a clinical patient perspective depended upon far more than the degree of impairment of spirometry (e.g. symptoms of breathlessness, exercise limitation, frequency of exacerbations) and that more attention should be paid to the multidimensional assessment of impairment in COPD than to purely categorising disease severity in terms of lung function impairment". 1 There are substantial geographical variations in COPD incidence reflecting socioeconomic issues, 2 with deprived urban populations having the highest prevalence, incidence and highest under diagnosis of COPD. 29 Social class gradients were also seen in respiratory disease mortality with men aged years 26

27 employed in unskilled manual occupations being around 14 times more likely to die from COPD than men employed in professional roles How many patients are assumed to be eligible? How is this figure derived? In 2009/10, the Quality and Outcomes Framework identified that there were 861,341 and 62,744 patients with diagnosed COPD in England and Wales, respectively (924,085 in total). Data on the proportion of COPD patients with FEV1 < 50% of predicted, associated with chronic bronchitis and a history of frequent exacerbations is not readily available in the literature. However, the recent ECLIPSE study (which included a cohort of patients from the UK) aimed to assess COPD in clinical practice, and included an evaluation of the proportion of patients with two or more exacerbations in the last 12 months. 14 The study identified that 42.1% of enrolled patients had severe COPD, as defined by the NICE 2010 guidelines (30% FEV 1 49%). Among these patients, 33% experienced two or more exacerbations within the first year of follow-up. This implies a patient population of approximately 128,383 severe COPD patients with frequent exacerbations. A further 13.7% had very severe COPD (FEV 1 < 30%), in whom 47% had two or more exacerbations, implying an additional 59,502 patients, and a total of 187,885. However, patients with mild COPD (FEV 1 80%), were not included in the ECLIPSE study (only moderate, severe and very severe). Assuming approximately 10% of all COPD patients have FEV 1 80%, it is estimated that there are 169,096 patients with severe or very severe COPD in the UK, who have a history of exacerbations. Included within this population will be those patients with COPD associated with emphysema (along with those with chronic bronchitis). This should, therefore, be regarded as an upper-bound estimate of the eligible patient population. 2.3 Please give details of any relevant NICE guidance or protocols for the condition for which the technology is being used. Specify whether any specific subgroups were addressed. 27

28 NICE Clinical Guideline 101 Chronic Obstructive Pulmonary Disease. Issued June Specifically for the subgroup of patients with severe COPD (FEV 1 < 50% predicted) CG101 recommends: " In people with stable COPD who remain breathless or have exacerbations despite using short-acting bronchodilators as required, offer the following as maintenance therapy: if FEV1 < 50% predicted: either LABA with an inhaled corticosteroid (ICS) in a combination inhaler, or LAMA. [new 2010] Offer LAMA in addition to LABA+ICS to people with COPD who remain breathless or have exacerbations despite taking LABA+ICS, irrespective of their FEV1. [new 2010] Consider LABA+ICS in a combination inhaler in addition to LAMA for people with stable COPD who remain breathless or have exacerbations despite maintenance therapy with LAMA irrespective of their FEV1. [new 2010]" Algorithm 2a details recommended therapy for patients with severe COPD and, in addition to the above, advises to consider LAMA plus LABA where ICS is declined or not tolerated. 2.4 Please present the clinical pathway of care that depicts the context of the proposed use of the technology. Explain how the new technology may change the existing pathway. If a relevant NICE clinical guideline has been published, the response to this question should be consistent with the guideline and any differences should be explained. Figure 2 is adapted from NICE Clinical Guideline 101 (Algorithm 2a), to only include treatment options for patients with severe or very severe COPD (FEV 1 < 50% predicted) and exacerbations, in line with the licence for roflumilast. 28

29 Figure 2: NICE Recommended Maintenance Therapy for Severe / Very Severe COPD (Adapted from NICE CG101, Algorithm 2a) FEV1 < 50% predicted and exacerbations or persistent breathlessness LABA + ICS LAMA Patients in whom ICS is not tolerated or is declined: LABA + LAMA Persistent exacerbations or breathlessness LABA + ICS + LAMA It is anticipated that roflumilast could be administered to patients in addition to their existing bronchodilator therapy, either because the physician deems that the patient would benefit from roflumilast as an additional therapy, or in patients for whom the physician believes other therapies (usually ICS) are inappropriate. Figure 3 outlines potential positions where roflumilast could be added in. Based on discussion with experienced and respected respiratory physicians, it is believed that roflumilast would most typically be used as an add-on to triple therapy (LAMA + LABA / ICS), as depicted by the red circles in Figure 3. Dr Michael Rudolf (Respiratory Physician, Ealing Hospital) and Professor Wisia Wedzicha (Consultant Respiratory Physician, Royal Free Hospital) both advised use of roflumilast would be in addition to this standard of care, and in addition of LAMA + LABA for patients who do not tolerate, or decline, ICS. There remains a significant unmet need in these patient cohorts. 15;30 29

30 Figure 3: Potential Addition of Roflumilast to NICE Recommended Maintenance Therapy for Severe / Very Severe COPD FEV1 < 50% predicted and exacerbations or persistent breathlessness LABA + ICS + LAMA + Roflumilast Patients in whom ICS is not tolerated or is declined: LABA + LAMA Roflumilast Persistent exacerbations or breathlessness LABA + ICS + LAMA + Roflumilast LABA + LAMA + Roflumilast 2.5 Please describe any issues relating to current clinical practice, including any variations or uncertainty about best practice. A variety of treatment options are recommended by NICE for patients with severe COPD, as described in Figure 2 above. The lack of a specific and single treatment pathway correlates with an observed variation in prescribing of these agents for patients with severe COPD in the UK. 31 As noted in the Summaries of Product Characteristics for all of the agents recommended by NICE for maintenance 30

31 treatment of severe COPD (see Figure 2), each should be used in caution in a different group of patients, which may partly explain the observed variation. In addition, currently available therapies for the management of COPD are focused on treating the immediate symptoms of COPD, whereas roflumilast targets the underlying inflammation and long term sequelae. 2.6 Please identify the main comparator(s) and justify their selection. As described in Section 2.4, it is expected that roflumilast will most commonly be used as an add-on treatment, in patients currently receiving LAMA + LABA / ICS or LAMA + LABA (in those who do not tolerate or decline ICS), who continue to exacerbate or remain breathless. Such use is in line with NICE CG 101, and with clinical practice. Through a fully incremental economic analysis, and in line with NICE CG 101, the use of roflumilast is assessed against the following comparators, and combinations of these: LABA ICS LABA / ICS LAMA As a step-change in the management of COPD, roflumilast also offers a new therapeutic intervention where no treatment alternative may otherwise exist. In these situations, the comparator would be: Placebo Theophylline has not been included in the decision problem as it is not thought to be an appropriate comparator for roflumilast. The NICE Clinical Guideline for management of COPD positions theophylline very late in the treatment pathway. 1 This is largely due to the side effect profile and the need for additional monitoring with use of theophylline. In clinical practice, only a small proportion of patients are prescribed theophylline, including amongst those currently receiving triple therapy with LAMA + LABA / ICS, 31

32 or LAMA + LABA (in those who do not tolerate or decline ICS). There is uncertainty about the mechanism of action of theophylline and how this relates to therapeutic benefit. The efficacy data for theophylline has not demonstrated a reduction in exacerbations. Given the better tolerability profile and proven efficacy of roflumilast in clinical trials, as well as the lack of requirement for additional monitoring, we would expect roflumilast to be used earlier in the treatment pathway, therefore theophylline is not an appropriate comparator. 2.7 Please list therapies that may be prescribed to manage adverse reactions associated with the technology being appraised. The most commonly reported adverse effects reported with use of roflumilast were diarrhoea (5.9%), weight loss (3.4%), nausea (2.9%), abdominal pain (1.9%) and headache (1.7%). The majority of reported side effects were mild to moderate in nature. The GI symptoms mostly occurred within 4 weeks of the start of therapy and resolved with continued treatment with roflumilast. 2.8 Please identify the main resource use to the NHS associated with the technology being appraised. Describe the location of care, staff usage, administration costs, monitoring and tests. Provide details of data sources used to inform resource estimates and values. Roflumilast is expected to be prescribed in the primary care setting, or in hospital, in patients who fall within the licensed indication and in whom it is clinically relevant. Roflumilast will not incur any administration costs or staff usage beyond dispensing of the tablets from the pharmacy. No monitoring or tests are required for use of roflumilast beyond those recommended for all COPD patients. 2.9 Does the technology require additional infrastructure to be put in place? No 32

33 2.10 What do you consider to be the relevant clinical outcomes and other potential health related benefits of the [technology] in the treatment of [disease/disorder], particularly when compared with currently used treatment options? Roflumilast is a once-daily, oral therapy which offers a novel mechanism of action as a first in class PDE4 inhibitor. In patients with severe and very severe COPD and associated chronic bronchitis, roflumilast has been shown to improve lung function and reduce the rate of moderate or severe exacerbations. 18;19 Exacerbations are associated with an accelerated decline in the general health of COPD patients and the frequency of exacerbations is correlated with a decline in lung function and health status. 8-10;32;33 In COPD, exacerbations account for the greatest burden on healthcare systems, therefore reducing their severity and frequency is an important therapeutic goal. 5;6;13 A significant unmet need still exists in COPD patients with chronic bronchitis, many of whom still experience frequent exacerbations despite treatment with currently recommended therapies. 11;14;15;30 Roflumilast represents an additional therapy option for patients already receiving maintenance bronchodilator therapy, which has been proven to improve lung function and reduce exacerbations Please identify the nature of the data which you understand to be available to enable the Appraisal Committee to take account of these benefits. The data to account for the benefits of roflumilast comes from the clinical trial programme for roflumilast and is referenced above. A mixed treatment comparison has also been conducted, to determine the efficacy of COPD treatments, and combinations of these treatments, including roflumilast, 22 and will form part of the full submission. 33

34 3 Equity and equality NICE considers equity in terms of how the effects of a health technology may deliver differential benefits across the population. Evidence relevant to equity considerations may also take a variety of forms and come from different sources. These may include general-population-generated utility weightings applied in health economic analyses, societal values elicited through social survey and other methods, research into technology uptake in different population groups, evidence on differential treatment effects in different population groups, and epidemiological evidence on risks or incidence of the condition in different population groups. Identification of equity and equalities issues 3.1 Please specify any issues relating to equity or equalities in NICE guidance, or protocols for the condition for which the technology is being used. As described in section 2.1, there are longstanding geographical and socioeconomic variations in COPD incidence and prevalence Are there any equity or equalities issues anticipated for the appraisal of No this technology (consider issues relating to current legislation and any issues identified in the scope for the appraisal)? 3.3 How have the clinical and cost-effectiveness analyses addressed these Not applicable issues? 34

35 4 Statement of the decision problem In this section the manufacturer or sponsor should specify the decision problem that the submission addresses. The decision problem should be derived from the final scope issued by NICE and should state the key parameters that the information in the evidence submission will address. 35

36 Final scope issued by NICE Decision problem addressed in the submission Rationale if different from the scope Population Adults with severe COPD (FEV 1 post bronchodilator less than 50% predicted) associated with chronic bronchitis in adult patients with a history of frequent exacerbations. Adults with severe COPD (FEV 1 post bronchodilator less than 50% predicted) associated with chronic bronchitis in adult patients with a history of frequent exacerbations. Not applicable Intervention Roflumilast in combination with maintenance bronchodilator treatment. Roflumilast in combination with maintenance bronchodilator treatment. Not applicable Comparator (s) ICS (in combination with LABA) LABA (in combination with LAMA) LAMA (in combination with LABA/ICS combination inhaler [or with LABA alone if ICS not tolerated]) Theophylline (in combination with maintenance bronchodilator treatment [LAMA plus LABA/ICS combination inhaler, or LAMA plus LABA, if ICS not tolerated]) ICS (in combination with LABA) LABA (in combination with LAMA) LABA / ICS (in combination with LAMA) LAMA (in combination with LABA / ICS, or LABA alone) Placebo (in combination with LAMA and / or LABA [+ / - ICS]) Theophylline is not thought to be an appropriate comparator for roflumilast. The NICE Clinical Guideline for management of COPD 1 positions theophylline very late in the treatment pathway. This is largely due to the side effect profile and the need for additional monitoring with use of theophylline. In clinical practice, only a small proportion of patients are prescribed theophylline, including amongst those currently receiving triple therapy with LAMA + LABA / ICS, or LAMA + LABA (in those who do not tolerate or decline ICS). There is uncertainty about the mechanism of action of theophylline and how this relates to therapeutic benefit. The efficacy data for theophylline has not demonstrated a reduction in exacerbations. Given the better tolerability profile and proven efficacy of roflumilast in clinical trials, as well as the lack of requirement for additional monitoring, we would expect roflumilast to be used earlier in the treatment pathway, therefore theophylline is not an appropriate 36

37 comparator. Outcomes Lung function Lung function Not applicable Incidence and severity of acute exacerbations, including hospitalisation Incidence and severity of acute exacerbations, including hospitalisation Symptom control (eg shortness of breath) Symptom control (eg shortness of breath) Mortality Mortality Adverse effects of treatment Adverse effects of treatment Health-related quality of life Health-related quality of life Economic analysis The reference case stipulates that the cost effectiveness of treatments should be expressed in terms of incremental cost per quality-adjusted life year. The reference case stipulates that the cost effectiveness of treatments should be expressed in terms of incremental cost per quality-adjusted life year. Not applicable The reference case stipulates that the time horizon for estimating clinical and cost effectiveness should be sufficiently long to reflect any differences in costs or outcomes between the technologies being compared. The reference case stipulates that the time horizon for estimating clinical and cost effectiveness should be sufficiently long to reflect any differences in costs or outcomes between the technologies being compared. Subgroups to be considered Special consideratio ns, including issues related to equity or equality Costs will be considered from an NHS and Personal Social Services perspective. Severity at baseline Smoking status BMI Costs will be considered from an NHS and Personal Social Services perspective. Severity at baseline subgroups will be based on baseline FEV 1 as defined by the NICE Clinical Guideline (severe vs very severe). Insufficient data on smoking status and BMI subgroups. Not applicable Not applicable Not applicable 37

38 Section B Clinical and cost effectiveness 5 Clinical Evidence Summary of the Roflumilast Clinical Trial Programme The clinical programme for roflumilast included seven phase III, randomised controlled trials, which are summarised in the following sections. The two pivotal, one year studies, M2-124 and M2-125, were conducted in patients with severe COPD associated with chronic bronchitis and a history of exacerbations, and support the efficacy of roflumilast in line with product licence. In these two trials, approximately 50% of the study population received concomitant LABA therapy and additional efficacy benefits were observed with roflumilast when used concomitantly with LABA, thus supporting the efficacy of roflumilast as add-on to bronchodilator therapy. Two additional studies investigated the use of roflumilast as add-on to bronchodilator therapy. M2-127 was conducted in a broad population of moderate to severe COPD patients, all of whom were taking concomitant salmeterol. M2-128 was conducted among patients with moderate to severe COPD and associated chronic bronchitis who were receiving concomitant tiotropium. Both studies demonstrated significant additional benefit with roflumilast when added on to the bronchodilator for the primary outcome measure of change in FEV 1. M2-111 and M2-112 were similarly designed studies conducted to assess the efficacy and safety of roflumilast over one year, in a broad population of patients with severe COPD (ICS was permitted as background therapy). A post-hoc subgroup analysis of studies M2-111 and M2-112 demonstrated the efficacy of roflumilast with or without concomitant ICS therapy in the specific group of patients with severe COPD and associated chronic bronchitis. M2-107 was the initial phase III trial and investigated the efficacy and safety of two doses of roflumilast, 250µg and 500µg, compared to placebo over a 6 month treatment period. The study included a broad population of moderate and severe COPD patients (i.e. patients were not required to have chronic bronchitis or a history of exacerbations to enter to study). 38

39 5.1 Identification of studies Describe the strategies used to retrieve relevant clinical data, both from the published literature and from unpublished data that may be held by the manufacturer or sponsor. The methods used should be justified with reference to the decision problem. Sufficient detail should be provided to enable the methods to be reproduced, and the rationale for any inclusion and exclusion criteria used should be provided. Exact details of the search strategy used should be provided in section 9.2, appendix 2. A systematic literature search was conducted to identify randomised controlled trials, which included roflumilast, in patients with COPD. The following databases were searched from inception to May 2011: Medline, Medline In-Process and Embase (via Dialog Datastar), and the Cochrane Library. Full details of the search strategy used are provided in section 9.2, appendix 2. The inclusion and exclusion criteria used to select studies are given in section In addition, MSD Ltd is aware of one additional randomised controlled trial with roflumilast which was not published (M2-111). As this study meets the inclusion and exclusion criteria of the systematic search it is included in the summary of RCTs in this dossier. 39

40 5.2 Study selection Describe the inclusion and exclusion selection criteria, language restrictions and the study selection process. A justification should be provided to ensure that the rationale is transparent. A suggested format is provided below. Table B1: Eligibility criteria used in search strategy Clinical effectiveness Inclusion criteria Population patients with COPD Interventions - roflumilast Outcomes exacerbations and FEV 1 Study design randomised controlled trial Language restrictions publication in English Exclusion criteria Population patients who did not have COPD; healthy subjects Pharmacokinetics or cost effectiveness studies References relating solely to an entry on a clinical trials registry were excluded A flow diagram of the numbers of studies included and excluded at each stage should be provided using a validated statement for reporting systematic reviews and meta-analyses such as the QUOROM statement flow diagram ( The total number of studies in the statement should equal the total number of studies listed in section A total of 315 publications were identified (Medline, Medline In-Process and Embase via Datastar 264; the Cochrane Library 51). Abstracts were reviewed for 197 studies, and full publications or posters were reviewed for

41 Figure 4: Flow Diagram of Included and Excluded Publications 315 Publications Identified 118 publications excluded due to duplication or based on title 197 Abstracts Reviewed Publications Excluded: 26 Meeting abstracts where data has since been fully published 143 Reviews or general articles 2 Pre-clinical studies 11 Publication not in English 1 Comments or letters 3 Clinical trials registry entries 11 Full Publications/Posters Reviewed 1 publication excluded (post-hoc analysis of 2 RCTs, which used a different analysis plan to RCT protocols) 5 Full Publications Included which Report on 7 RCTs. 5 Meeting Abstracts Included which Report on 2 RCTs When data from a single RCT have been drawn from more than one source (for example, a poster and a published report) and/or when trials are linked (for example, an open-label extension to an RCT), this should be made clear. Not applicable Complete list of relevant RCTs Provide details of all RCTs that compare the intervention with other therapies (including placebo) in the relevant patient group. The list must be complete and will be validated by independent searches conducted by the Evidence 41

42 Review Group. This should be presented in tabular form. A suggested format is presented below. 42

43 Trial no. (acronym) M2-124 & M2-125 Table B2: List of relevant RCTs Intervention Comparator Population Primary study ref. Roflumilast 500 µg Placebo Patients with severe or very severe COPD (FEV 1 50% predicted) with associated chronic bronchitis and a history of moderate or severe exacerbations (defined as having at least one recorded COPD exacerbation requiring systemic glucocorticosteroids or treatment in hospital, or both, in the previous year). M2-127 Roflumilast 500 µg o.d. and salmeterol (50 µg b.i.d.) M2-128 Roflumilast 500 µg o.d. and tiotropium (18 µg) Placebo and salmeterol (50 µg b.i.d.) Placebo and tiotropium (18 µg) Patients with moderate to severe COPD (post-bronchodilator FEV % predicted). Patients were not required to have chronic bronchitis or a history of exacerbations. Patients with moderate to severe COPD (post-bronchodilator FEV % predicted) with associated chronic bronchitis and frequent use of as-needed SABA (at least 28 puffs per week) during run-in. M2-111 Roflumilast 500 µg Placebo Patients with severe or very severe COPD (post-bronchodilator FEV 1 50% predicted). Patients were not required to have chronic bronchitis or a history of frequent exacerbations M2-112 Roflumilast 500 µg Placebo Patients with severe or very severe COPD (post-bronchodilator FEV 1 50% predicted). Patients were not required to have chronic bronchitis or a history of frequent exacerbations. M2-107 (All further details are provided in Appendix 14) Roflumilast 250 µg and 500 µg FK1-101 Roflumilast 250 µg and 500 µg Placebo Placebo Patients with moderate to severe COPD (post-bronchodilator FEV % predicted). Patients were not required to have chronic bronchitis or a history of frequent exacerbations. Patients with moderate to severe COPD (post-bronchodilator FEV % predicted). Patients were not required to have chronic bronchitis or a history of frequent exacerbations. FK1-103 Roflumilast 500 µg Placebo Patients with moderate to severe COPD (post-bronchodilator FEV % predicted). Patients were not required to have chronic bronchitis or a history of frequent exacerbations. Grootendo rst et al Roflumilast 500 µg Placebo Patients with moderate to severe COPD (post-bronchodilator FEV % predicted). Patients were not required to have chronic bronchitis or a history of frequent exacerbations. Calverley et al, Fabbri et al, Fabbri et al, Clinical Study Report for trial M Calverley et al, Rabe et al, Bredenbroker et al, ; Bredenbroker at al, ; Leichtl et al, ; Leichtl et al, Boszormenyi- Nagy et al, Grootendorst et al,

44 5.2.5 Please highlight which of the RCTs identified above compares the intervention directly with the appropriate comparator(s) with reference to the decision problem. If there are none, please state this. None of the above trials directly compare roflumilast with the comparators listed in the decision problem. A mixed treatment comparison has been conducted, and is presented in section 5.7, which assesses the efficacy of various combinations of COPD treatments including roflumilast, that have not been studied in RCTs When studies identified above have been excluded from further discussion, a justification should be provided to ensure that the rationale for doing so is transparent. For example, when studies have been identified but there is no access to the level of trial data required, this should be indicated. Only the key, phase III roflumilast trials (M2-124, M2-125, M2-127, M2-128, M2-111 and M2-112) are included in the following sections. Details of study M2-107 are not included in the following sections as the patient population does not reflect the licensed indication for roflumilast and 250 µg roflumilast is not the labelled dose, but full details are provided in Appendix 14. Studies FK1-101 and FK1-103 were part of the early development programme for roflumilast. FK-101 was a dose-finding study of roflumilast 250µg and 500µg, and FK1-103 studied the effect of withdrawal following 12 weeks of therapy with roflumilast 500µg. The study conducted by Grootendorst et al 40 was a crossover study of two 4 week treatment periods, primarily investigating the reduction in sputum neutrophils and included only 38 participants. These three trials are not discussed further in the following sections; details can be provided on request. List of relevant non-rcts Please provide details of any non-rcts (for example experimental and observational data) that are considered relevant to the decision problem and a justification for their inclusion. Full details should be provided in section 5.8 and key details should be presented in a table; the following is a suggested format. None. 44

45 5.3 Summary of methodology of relevant RCTs As a minimum, the summary should include information on the RCT(s) under the subheadings listed in this section. Items 2 to 14 of the CONSORT checklist should be provided, as well as a CONSORT flow diagram of patient numbers ( It is expected that all key aspects of methodology will be in the public domain; if a manufacturer or sponsor wishes to submit aspects of the methodology in confidence, prior agreement must be requested from NICE. When there is more than one RCT, the information should be tabulated. Methods Describe the RCT(s) design (for example, duration, degree and method of blinding, and randomisation) and interventions. Include details of length of follow-up and timing of assessments. The following tables provide a suggested format for when there is more than one RCT. 45

46 Study Number Table B3: Comparative summary of methodology of the RCTs Study Design M2-124 Double-blind, multicentre, placebo controlled, phase III studies in Australia, Austria, France, Germany, Hungary, New Zealand, Romania, Russia, UK, USA. M2-125 Double-blind, multicentre, placebo controlled, phase III studies in Canada, Germany, India, Italy, Poland, Spain, South Africa, USA. Study Duration 1 year Study Intervention(s) and Comparator(s) Interventions = roflumilast 500 µg o.d Comparator = placebo o.d. Number of Patients 1523 (765 received roflumilast, 758 received placebo) 1568 (772 received roflumilast, 796 received placebo) Permitted Concomitant Respiratory Therapies Approximately 50% patients received concomitant LABA. Patients could also use SABA and SAMA. ICS and LAMA prohibited. Primary Outcomes (Also see Table B6) Rate of moderate or severe COPD exacerbations. Change in prebronchdilator FEV 1. Key Secondary Outcomes (Also see Table B6) Total number of exacerbations, using the wider definition (treated with antibiotics alone, in addition to primary outcome definition). Change in post-bronchodilator FEV 1. Time to mortality due to any reason Change in natural log-transformed C-reactive protein Mean TDI focal score during treatment. EQ-5D. 46

47 M2-127 Double-blind, multicentre, placebo controlled, phase III studies in Austria, Belgium, Canada, France, Germany, Hungary, Italy, Netherlands, South Africa, Spain, UK. Interventions = roflumilast 500 µg o.d. + salmeterol 50 µg b.i.d Comparator = placebo o.d. + salmeterol 50 6 months µg b.i.d. M2-128 Interventions = M2-111 Double-blind, multicentre, randomised, placebo controlled, phase III study in Canada, France, Germany, Poland, South Africa, USA. 1 year roflumilast 500 µg o.d. + tiotropium 18 µg Comparators = placebo o.d. + tiotropium 18 µg Interventions = roflumilast 500 µg o.d Comparator = placebo o.d. 933 (466 received roflumilast, 467 received placebo) 743 (371 received roflumilast, 372 received placebo) 1173 (567 received roflumilast, 606 received placebo) All patients received concomitant salmeterol. All other respiratory medications were prohibited. All patients received concomitant tiotropium. All other respiratory medications were prohibited. Salbutamol as rescue medication was permitted. A constant daily dose of ICS (2000 µg beclometasone diproprionate or equivalent) or SAMA were allowed if used before study entry. All other COPD medications were prohibited. Change in prebronchodilator FEV 1 from baseline to each postrandomisation visit. Rate of moderate or severe COPD exacerbations Change in prebronchodilator FEV 1 Rate of COPD exacerbations. Post-bronchodilator FEV1 and FVC. TDI score. Mean change in Shortness of Breath Questionnaire. Use of rescue medication. Change in post-bronchodilator FEV1 Rate of moderate or severe COPD exacerbations in patients with different disease characters: o o o o Baseline post bronchodilator FEV1 < 30% predicted (very severe) measured at the randomization visit V0. History of chronic bronchitis and with or without history of emphysema Cough score > 2 in the week before randomisation Cough score > 1 in the week 47

48 o before randomisation History of at least one moderate or severe COPD exacerbations in the year prior to screening Rate of moderate COPD exacerbations treated with oral or parenteral glucocorticoids and/or antibiotics (here a different definition for moderate exacerbation was used) or severe COPD exacerbations Rate of mild or moderate or severe COPD exacerbations. M2-112 Double-blind, multicentre, randomised, placebo controlled, phase III study in Australia, Austria, Canada, France, Hungary, Italy, Netherlands, Poland, Portugal, Russia, Spain, Switzerland, UK, South Africa, 1513 (760 received roflumilast, 753 received placebo) Rate of moderate or severe COPD exacerbations Change in postbronchodilator FEV 1 Rate of moderate or severe COPD exacerbations requiring systemic corticosteroid treatment. Change from baseline in SGRQ total score. Change from baseline in: pre-bronchodilator FEV 1, post-bronchodilator FEV 6, post-bronchodilator FVC, post-bronchodilator FEF

49 Participants Provide details of the eligibility criteria (inclusion and exclusion) for the trial. The following table provides a suggested format for the eligibility criteria for when there is more than one RCT. Highlight any differences between the trials. Table B4 provides the key inclusion and exclusion criteria for each study. In some publications, additional inclusion and exclusion criteria were provided in online supplements and details of these are provided in appendix 15. Differences In Eligible Patient Populations Between Trials In M2-124, M2-125 and M2-128 only patients with chronic bronchitis were eligible, whereas M2-127 and M2-112 and M2-111 included a broader population of COPD patients. Studies M2-124, M2-125, M2-111 and M2-112 included only patients with severe COPD (FEV 1 50% predicted), whereas M2-127 and M2-128 included less severe COPD patients with FEV 1 between 40% and 70% predicted. Studies M2-124 and M2-125 included only patients with a history of exacerbations ( 1 exacerbation requiring systemic glucocorticosteroids or hospital treatment in the previous year). This was not part of the inclusion criteria for the other roflumilast RCTs. 49

50 Trial no. (acronym) M2-124 & M2-125 Table B4: Eligibility criteria in the RCTs (see Appendix 15 for additional eligibility criteria) Key inclusion criteria Clinical diagnosis of COPD (confirmed with postbronchodilator FEV 1 /FVC ratio 70%), Current or former smokers with at least a 20 pack-year history, Age 40 years, Chronic cough and sputum production, Postbronchodilator FEV 1 50% predicted, At least one recorded COPD exacerbation requiring systemic glucocorticosteroids or hospital treatment, or both, in the previous year. M2-127 History of COPD for at least 12 months prior to baseline visit (defined by ATS/ERS 2005), Current smoker or former smoker (defined as smoking cessation at least one year ago) with a smoking history of at least 10 pack-years, Age 40 years, Postbronchodilator FEV 1 between 40 % and 70 % of predicted, Postbrochodilator FEV 1 /FVC ratio 70%, Fixed airway obstruction (defined as an FEV1 increase of 12 % or 200 ml after receiving 400 µg salbutamol) Key exclusion criteria COPD exacerbation (indicated by corticosteroid treatment of hospitalisation) unresolved at initial visit (4 weeks pre-baseline) Asthma or other relevant lung disease Clinically significant cardiopulmonary abnormalities not related to COPD COPD exacerbation requiring treatment with systemic glucocorticosteroids and/or antibiotics not stopped at least 4 weeks prior to the baseline visit, Lower respiratory tract infection not resolved 4 weeks prior to the baseline visit, Diagnosis of asthma and/or other relevant lung disease (e.g. history of bronchiectasis, cystic fibrosis, bronchiolitis, lung resection, lung cancer, interstitial lung disease [e.g. fibrosis, silicosis, sarcoidosis], and active tuberculosis) Current participation in a pulmonary rehabilitation programme or completion of a pulmonary rehabilitation programme within 3 months preceding the baseline visit, Known alpha-1-antitrypsin deficiency M2-128 History of COPD for at least 12 months prior to baseline visit (defined by ATS/ERS 2005) and chronic productive cough for 3 months in each of the 2 years prior to baseline visit (if other causes of productive cough have been excluded), Current smoker or former smoker (defined as smoking cessation at least one year ago) with a smoking history of at least 10 pack-years, COPD exacerbation requiring treatment with systemic glucocorticosteroids and/or antibiotics not stopped at least 4 weeks prior to the baseline visit, Lower respiratory tract infection not resolved 4 weeks prior to the baseline visit, Diagnosis of asthma and/or other relevant lung disease (e.g. history 50

51 Age 40 years, Postbronchodilator FEV 1 between 40 % and 70 % of predicted, Postbrochodilator FEV 1 /FVC ratio 70%, Fixed airway obstruction (defined as an FEV1 increase of 12 % or 200 ml after receiving 400 µg salbutamol), Patients must be pre-treated with tiotropium for at least 3 months prior to baseline visit. M2-111 Current or ex-smoker (no tobacco 1 year) with a smoking history of at least 10 pack-years. Age 40 years, Postbronchodilator FEV 1 of 50% of predicted value, Postbronchodilator FEV 1 /FVC ratio 0.70, Clinically stable COPD indicated by no exacerbation and no change in COPD treatment within 4 weeks prior to baseline, M2-112 Diagnosis of COPD, as defined by GOLD criteria, Current or ex-smoker (no tobacco 1 year) with a smoking history of at least 10 pack-years. Age 40 years, Postbronchodilator FEV 1 of 50% of predicted value, Postbronchodilator FEV 1 /FVC ratio 0.70, FEV 1 reversibility of 15% and/or 200 ml after 200 µg inhaled salbutamol, Clinically stable with unchanged COPD treatment and no exacerbations or respiratory tract infection for 4 weeks before run-in. of bronchiectasis, cystic fibrosis, bronchiolitis, lung resection, lung cancer, interstitial lung disease [e.g. fibrosis, silicosis, sarcoidosis], and active tuberculosis) Current participation in a pulmonary rehabilitation programme or completion of a pulmonary rehabilitation programme within 3 months preceding the baseline visit, Known alpha-1-antitrypsin deficiency. History of asthma or other relevant lung disease (e.g. lung cancer, bronchiectasis), Requirement for long-term oxygen treatment, Known alpha-1-antitrysin deficiency, Exacerbation or respiratory tract infection within 4 weeks before run-in period, Any significant cardiopulmonary comorbidity. History of asthma or other relevant lung disease (e.g. lung cancer, bronchiectasis), Requirement for long-term oxygen treatment, Known alpha-1-antitrysin deficiency, Exacerbation or respiratory tract infection within 4 weeks before run-in period, Any significant cardiopulmonary comorbidity. 51

52 5.3.4 Describe the patient characteristics at baseline. Highlight any differences between study groups. The following table provides a suggested format for the presentation of baseline patient characteristics for when there is more than one RCT. Tables B5 and B6 summarise the baseline characteristics of participants in the RCTs. The patient characteristics were comparable between treatment groups for all studies. 52

53 Table B5: Characteristics of participants in the RCTs across randomised groups (Studies M2-124, M2-125, M2-127 and M2-128)* Baseline Characteristic M2-124 (1523) M2-125 (n = 1568) M2-127 (n = 933) M2-128 (n = 743) Roflumilast 500 µg Placebo (n = 758) Roflumilast 500 µg Placebo (n = 796) Salmeterol + Roflumilast Salmeterol + Placebo Tiotropium + Roflumilast Placebo (n = 372) (n = 765) (n = 772) 500 µg (n =466 ) (n = 467) 500 µg (n = 371) Median Age, years 64 (10) 63 (9) 64 (9) 64 (9) 65 (9) 65 (9) 64 (9) 64 (9) (range) Gender, n (%) Male 540 (71) 538 (71%) 610 (79%) 648 (81%) 319 (68%) 299 (64%) 262 (71%) 267 (72%) Race, n (%) White 737 (96%) 732 (97%) 559 (72) 568 (71) Not reported Not reported Not reported Not reported Body-mass Index 26.4 (5.5) 26.0 (5.5) 25.2 (6.2) 25.4 (5.9) Not reported Not reported Not reported Not reported C-reactive Protein (mg/l) 8.1 (14.0) 7.2 (12.5) 8.3 (14.6) 9.2 (17.6) Not reported Not reported Not reported Not reported Smoking Status Current smokers, n (%) Ex-smokers, n (%) Pack-years 365 (48) 400 (52) 48 (24) 361 (48) 397 (52) 46 (23) 270 (35%) 502 (65%) 49 (26) 282 (35%) 514 (65%) 47 (24) 184 (39%) 282 (61%) 43 (22) 184 (39%) 283 (61%) 43 (22) 147 (40%) 224 (60%) 43 (22) 146 (39%) 226 (61%) 42 (22) Prebronchodilator FEV 1, 1.07 (0.4) 1.06 (0.4) 0.95 (0.3) 0.98 (0.4) 1.43 (0.4) 1.41 (0.4) 1.47 (0.5) 1.49 (0.5) L Postbronchodilator FEV 1, L 1.16 (0.04) 1.15 (0.4) 1.05 (0.4) 1.07 (0.4) 1.51 (0.4) 1.49 (0.4) 1.55 (0.5) 1.56 (0.5) Prebronchodilator FEV 1, 34.7 (10.2) 34.6 (10.3) 31.4 (10.1) 32.2 (10.8) 51.9 (9.6) 52.4 (9.8) 53.3 (11.7) 53.4 (11.6) % predicted Postbronchodilator 37.6 (10.7) 37.5 (10.4) 34.6 (10.3) 35.3 (10.9) 54.7 (9.1) 55.3 (9.2) 56.0 (11.6) 56.2 (11.2) FEV 1, % predicted FEV 1 /FVC, % 43.3 (11.6) 42.7 (11.10) 41.2 (10.7) 41.3 (10.8) 49.8 (9.4) 50.0 (9.7) 52.7 ( (9.9) Postbronchodilator Not reported Not reported Not reported Not reported Not reported Not reported Not reported Not reported FEV 6, L Concomitant SAMA 240 (31%) 245 (32%) 297 (38%) 324 (41%) Not permitted Not permitted Not permitted Not permitted Concomitant ICS Not permitted Not permitted Not permitted Not permitted Not permitted Not permitted Not permitted Not permitted Concomitant SABA 761 (99%) 753 (99%) 769 (100%) 791 (99%) Not reported Not reported Not reported Not reported Concomitant LABA 378 (49%) 385 (51%) 371 (48%) 408 (51%) 466 (100%) 467 (100%) Not permitted Not permitted Pretreatment with ICS 338 (44%) 335 (44%) 312 (40%) 322 (40%) Not reported Not reported Not reported Not reported 53

54 (during run-in period) Use of as-needed Not reported Not reported Not reported Not reported 1.4 (0-17.1) 1.7 (0-28.7) 4.7 (0-20.0) 4.6 ( ) relievers during run-in (mdian, range). (puffs per day in M2-127; puffs per week in M2-128) COPD Severity Moderate Severe Very Severe 486 (64%) 199 (26%) 510 (67%) 184 (24%) 457 (59%) 264 (34%) 479 (60%) 256 (32%) 303 (65%) 162 (35%) Not reported 324 (69%) 141 (30%) Not reported 235 (63%) 125 (34%) Not reported 240 (65%) 119 (32%) Not reported Chronic cough and sputum during run-in Not reported Not reported Not reported Not reported 367 (79%) 362 (78%) 371 (100%) 372 (100%) * Characteristics only listed where reported in trial. 54

55 Table B6: Characteristics of participants in the RCTs across randomised groups (Studies M2-111 and M2-112)* M2-111 (n=1173) M2-112 (n = 1513) Roflumilast 500 µg (n = 567) Placebo (n = 606) Roflumilast 500 µg (n = 760) Placebo (n = 753) Baseline Characteristic Median Age, years (range) 65 (40-87) 64 (41-86) 65 (9.6) 64 (9.1) Gender, n (%) Male 387 (68) 400 (66) 571 (75) 574 (76) Body-mass Index 26 (5.7) 26 (5.7) 25 (5.0) 26 (5.1) C-reactive Protein (mg/l) Not reported Not reported Not reported Not reported Smoking Status Current smokers, n (%) Ex-smokers, n (%) Pack-years 240 (42.3%) 327 (57.7%) 50 (28.2) 265 (43.7%) 341 (56.3%) 51(26.7) 289 (38) 471 (62) 42 (22.9) 265 (35) 488 (65) 45 (26.2) Prebronchodilator FEV 1, ml 960 (400) 930 (300) 1029 (35) 1047 (34) Postbronchodilator FEV 1, ml 1120 (400) 1090 (400) 1126 (36) 1145 (36) Postbronchodilator FEV 1, % predicted 36.8 (10.7) 36 (9.8) 41 (11) 41 (11) Reversibility: change in FEV 1, ml % (143.9) 18.9 (16.7) (155.3) 18.8 (17.7) 98 (124) 11 (14) 100 (137) 11 (14) Concomitant short-acting 316 (55.7) 334 (55.1) 432 (57) 459 (61) anticholinergics, Concomitant ICS 314 (55.4) 320 (52.8) 472 (62) 471 (63) COPD Severity Moderate 64 (11.3) 45 (7.4) Not reported Not reported Severe 366 (64.6) 396 (65.3) Very Severe 137 (24.2) 165 (27.2) * Characteristics only listed where reported in trial. 55

56 Outcomes Provide details of the outcomes investigated and the measures used to assess those outcomes. Indicate which outcomes were specified in the trial protocol as primary or secondary, and whether they are relevant with reference to the decision problem. This should include therapeutic outcomes, as well as patient-related outcomes such as assessment of health-related quality of life, and any arrangements to measure compliance. Data provided should be from pre-specified outcomes rather than post-hoc analyses. When appropriate, also provide evidence of reliability or validity, and current status of the measure (such as use within UK clinical practice). The following table provides a suggested format for presenting primary and secondary outcomes when there is more than one RCT. Key Outcome Measures Mean rate of moderate or severe COPD exacerbations This was a coprimary endpoint in the pivotal studies (M2-124 and M2-125) and studies M2-111 and M2-112, and a secondary endpoint in studies M2-127 and M Change in FEV 1 this was chosen as a co-primary endpoint (M2-124, M2-125, M2-111 and M2-112) or primary endpoint (M2-127 and M2-128) due to the precedent of clinical trials for other agents in this therapy area. Roflumilast is an anti-inflammatory agent, not a bronchodilator, therefore it's effect on FEV 1 is expected to be more in line with that of ICS, rather than LABA or LAMA. ICS are known to have a limited effect on FEV 1 in COPD patients, however their use as add on to bronchodilator therapy is clinically recognised as an effective treatment for severe COPD patients 1. 56

57 Table B 7: Primary and secondary outcomes of the RCTs Trial no. (acronym) M2-124 & M2-125 Primary outcome(s) and measures Rate of moderate or severe COPD exacerbations. Moderate exacerbations were defined as those requiring oral or parenteral corticosteroids, and severe exacerbations were defined as being associated with hospital admission or death. Change in pre-bronchdilator FEV 1. FEV 1 tests were conducted every 4 weeks after randomisation up to week 12 and every 8 weeks thereafter Reliability/validity/ current use in clinical practice Exacerbations are important events in COPD, associated with faster disease progression, worsening of lung function, increased morbidity and mortality as well as decreased quality of life. FEV1 is an established measure of lung function and is widely used in clinical practice. Key Secondary outcome(s) and measures Change in post-bronchodilator FEV 1. FEV 1 tests were conducted every 4 weeks after randomisation up to week 12 and every 8 weeks thereafter. Time to mortality due to any reason Change in natural log-transformed C-reactive protein Mean TDI focal score during treatment. This was measured every 4 weeks after randomisation up to week 12 and every 8 weeks thereafter. A change of one unit in the TDI focal score was considered clinically significant. Total number of exacerbations, using the wider definition (included treated with antibiotics alone) in addition to primary outcome definition Other spirometric outcomes, including change from baseline in FVC [L], FEF25-75% [L/s], FEV3 [L], FEV6 [L], FEV1/FVC [%], FEV1/FEV6 [%], PEF [L/min]. EQ-5D. The EQ-5D questionnaire was completed by patients at every visit (every 4 weeks after randomisation up to week 12 and every 8 weeks thereafter.) Reliability/validity/ current use in clinical practice FEV 1 is an established measure of lung function and is widely used in clinical practice. Not applicable C-Reactive protein levels are widely acknowledged as non-specific markers of inflammation in the body TDI score is a commonly used index to assess breathlessness and the impact of intervention. Exacerbations are important events in COPD, associated with faster disease progression, worsening of lung function, increased morbidity and mortality as well as decreased quality of life. Pulmonary function tests are a key standard in assessing lung function EQ-5D is a standardised instrument for use as a measure of health outcome. 57

58 M2-127 & M2-128 Change in prebronchodilator FEV 1 from baseline to each postrandomisation visit. FEV 1 tests were conducted at every visit (every 4 weeks after randomisation up to week 12 and every 6 weeks thereafter). FEV1 - as described above Rate of COPD exacerbations per patient per year. COPD exacerbations were defined as mild if the patient needed an increase in rescue medication of at least 3 puffs per day on at lease 2 consecutive days, moderate if the patient needed oral corticosteroids and severe if hospitalisation was required or if the patient died. Post-bronchodilator FEV 1 and FVC. Pulmonary function tests were conducted at every visit (every 4 weeks after randomisation up to week 12 and every 6 weeks thereafter). Mean TDI focal score during the treatment period. Investigator-administered TDI was conducted at every visit. Mean change in Shortness of Breath Questionnaire (SOBQ) from baseline to each post-randomisation visit. This was completed by the patient at every visit. Use of rescue medication. Patients' use of SABA was recorded at every visit. Exacerbations as described above Pulmonary function tests as described above TDI score as described above SOBQ is a tool to assess patient symptoms based on reports of activity/restrictions to activity due to dyspnoea. Use of rescue medication gives an indication of how much additional intervention is required to manage patient symptoms. 58

59 M2-111 Rate of moderate COPD exacerbations (treated with oral or parenteral glucocorticoids) or severe COPD exacerbations. Change from baseline during the treatment period in pre-bronchodilator FEV 1. Exacerbations as described above FEV1 - as described above Change in post-bronchodilator FEV 1 Rate of moderate COPD exacerbations (treated with oral or parenteral glucocorticoids) or severe COPD exacerbations in patients with different disease characters: o Baseline post bronchodilator FEV1 < 30% predicted (very severe) measured at the randomization visit V0. o History of chronic bronchitis and with or without history of emphysema o Cough score > 2 in the week precede V0 o o Cough score > 1 in the week precede V0 History of at least one moderate or severe COPD exacerbations in the year prior to screening (B0 is equivalent of V0) Rate of moderate COPD exacerbations treated with oral or parenteral glucocorticoids and/or antibiotics (here a different definition for moderate exacerbation was used) or severe COPD exacerbations Rate of mild or moderate or severe COPD exacerbations. FEV 1 is an established measure of lung function and is widely used in clinical practice. Exacerbations - as described above. 59

60 M2-112 Rate of moderate or severe COPD exacerbations. Moderate exacerbations were defined as symptomatic deteriorations requiring systemic corticosteroids and/or antibiotics. Severe exacerbations were those which required hospitalisation. Change in postbronchdilator FEV 1. FEV 1 tests were conducted according to American Thoracic Society standards at recruitment, at week 2 of run-in, at randomisation (baseline) and at weeks 4, 8, 12, 20, 28, 36, 44 and 52. Exacerbations - as described above. FEV1 - as described above Rate of moderate or severe COPD exacerbations requiring systemic corticosteroid treatment Change from baseline in SGRQ total score. The SGRQ was completed by patients at week 2 of run-in, at randomisation (baseline) and at weeks 12, 28, 36, 44 and 52. Change from baseline in: - prebronchodilator FEV 1, - postbronchodilator FEV 6, - postbronchodilator FVC, - postbronchodilator FEF Pulmonary function tests were conducted according to American Thoracic Society standards at recruitment, at week 2 of run-in, at randomisation (baseline) and at weeks 4, 8, 12, 20, 28, 36, 44 and 52. Exacerbations as described above The St George's Respiratory Questionnaire is a commonly used standardized selfcompleted questionnaire for measuring impaired health and perceived well-being ('quality of life') in airways disease. Pulmonary function tests as described above Statistical analysis and definition of study groups State the primary hypothesis or hypotheses under consideration and the statistical analysis used for testing hypotheses. Also provide details of the power of the study and a description of sample size calculation, including rationale and assumptions. Provide details of how the analysis took account of patients who withdrew (for example, a description of the intention-to-treat analysis undertaken, including censoring methods; whether a per-protocol analysis was undertaken). The following table provides a suggested format for presenting the statistical analyses in the trials when there is more than one RCT. 60

61 Table B 8: Summary of statistical analyses in RCTs Trial no. (acronym) M2-124 & M2-125 Hypothesis objective To assess whether roflumilast reduces the rate of exacerbations which require systemic corticosteroids. Statistical analysis Sample size, power calculation Data management, patient withdrawals Efficacy analyses were conducted in the intention-totreat population. Data were analysed in the two studies separately and in a pooled analysis. Changes from baseline in pre- and postbronchodilator FEV 1 were analysed using a repeatedmeasures analysis of covariance with all data available for patients during the 52-week treatment. A Cox proportional hazard model was used to test for differences in time-to-event data. For the regression models (analysis of covariance, Cox, and Poisson), the covariates included treatment age, sex, smoking status (current or former smoker), country, and treatment with LABAs. In the Cox analysis, country was included as a stratum. In the Poisson regression analysis, baseline postbronchodilator FEV 1 was also included as a covariate. To address the issue of multiple comparisons, a hierarchical hypothesis-testing approach was adopted. If the primary outcomes were positive, the key secondary outcomes were tested in the order above. If a significant difference between treatments was not obtained for the primary or key secondary outcomes, all subsequent analyses were considered exploratory. Adverse events were analysed with descriptive statistics and 95% CIs for the differences between treatments. On the basis of an assumption of a mean exacerbation rate of 1 25 per patient per year in the placebo group and 1 00 in the roflumilast group, and using a Poisson regression model. We estimated that 750 patients per treatment group in each trial would provide 90% power to detect a significant difference between treatments with a two-sided α level of A negative binominal regression analysis was done to assess the robustness of the results against the distributional assumptions. Efficacy analyses were conducted in the intention-to-treat population. 61

62 M2-127 To investigate the effect of roflumilast on lung function in patients with moderate to severe COPD who are already being treated with salmeterol. Data for efficacy were evaluated in the intention-totreat population. The primary endpoint, change in pre-bronchodilator FEV 1 from baseline, was analysed by repeated-measures analysis of covariance. The repeated-measures analysis of covariance model included the factors and covariates of treatment, value at baseline, age, sex, smoking status, country, time, and treatment-by-time interaction. Adverse events were analysed with descriptive statistics and 95% CIs for the differences between treatment groups. The study was powered for the primary endpoint of change in pre-bronchodilator FEV 1 from baseline. The estimate of the treatment effect (50 ml) was based on clinical considerations and was in agreement with previous studies of ICS added to LABA. The sample size was calculated for the repeated-measures analysis of covariance model. On the basis of use of a one-sided significance level of 2.5%, the power was 97% with a sample size of 469 patients per treatment group. This trial was originally powered for a traditional analysis of covariance model and not for a repeated-measures analysis of covariance model. After completion of recruitment, but before unmasking the studies, the statistical analysis model was changed to the more powerful repeatedmeasures analysis of covariance model, accounting for the larger number of patients recruited and the higher statistical power in trial M2-127 than in the linked trial M The intention-to-treat population was defined as patients given at least one dose of study medication. A conservative approach was taken for the main analysis of the repeated measurements of expiratory lung function variables, patient diary variables, SOBQ, and TDI scores, and no missing values were replaced in the trial. M2-128 To investigate the effect of roflumilast on lung function in patients with moderate to severe COPD who are already being treated with tiotropium. As reported above for M As reported above for M On the basis of use of a one-sided significance level of 2.5%, the power was 91% with a sample size of 350 patients per treatment group. As reported above for M

63 M2-111 To determine the improvement in lung function and reduction in exacerbations with roflumilast over 1 year of treatment in patients with severe or very severe COPD. Efficacy data were analysed in the intention-to-treat population. Differences in lung function variables and SGRQ were tested using analysis of covariance with the factors and covariates of treatment, sex, country, value at baseline, smoking status, pretreatment with inhaled corticosteroids, and age included in the model. For within- and between-group comparisons, the two-sided tests are reported at an level of The co-primary variable of frequency of moderate or severe exacerbations per patient per year was analysed non-parametrically using Poisson regression. To account for the effects of premature study withdrawal for any reason, the number of exacerbations for those patients was annualized considering their study duration. In addition, a predefined analysis of exacerbation frequency was performed using a Poisson regression model, with the covariates of treatment, age, sex, smoking status, country, and pretreatment with inhaled corticosteroids to estimate the rate ratio. The study was powered so that 550 patients per group gave a 91% chance of detecting a difference of 50 ml in FEV1 between treatments (assuming a common SD of 250 ml). It was estimated that the study was also powered with a 91% chance of detecting a 20% reduction in exacerbation rate, assuming a mean of one exacerbation per patient per year in the placebo group. Assuming independence of the two primary endpoint variables, the study had a power of at least 82% to identify a treatment difference in both variables. The intention-to-treat population was defined as all randomised patients who received at least one dose of study medication. For missing values, the last observation carried forward imputation technique was applied. 63

64 M2-112 To determine the improvement in lung function and reduction in exacerbations with roflumilast over 1 year of treatment in patients with severe or very severe COPD. Efficacy data were analysed in the intention-to-treat population. Differences in lung function variables and SGRQ were tested using analysis of covariance with the factors and covariates of treatment, sex, country, value at baseline, smoking status, pretreatment with inhaled corticosteroids, and age included in the model. For within- and between-group comparisons, the two-sided tests are reported at an level of The co-primary variable of frequency of moderate or severe exacerbations per patient per year was analysed non-parametrically using the Wilcoxon rank sum test. To account for the effects of premature study withdrawal for any reason, the number of exacerbations for those patients was annualized considering their study duration. In addition, a predefined analysis of exacerbation frequency was performed using a Poisson regression model, with the covariates of treatment, age, sex, smoking status, country, and pretreatment with inhaled corticosteroids to estimate the rate ratio. The study was powered so that 550 patients per group gave a 91% chance of detecting a difference of 50 ml in FEV1 between treatments (assuming a common SD of 250 ml). It was estimated that the study was also powered with a 91% chance of detecting a 20% reduction in exacerbation rate, assuming a mean of one exacerbation per patient per year in the placebo group. Assuming independence of the two primary endpoint variables, the study had a power of at least 82% to identify a treatment difference in both variables. The intention-to-treat population was defined as all randomised patients who received at least one dose of study medication. For missing values, the last observation carried forward imputation technique was applied. 64

65 5.3.7 Provide details of any subgroup analyses that were undertaken and specify the rationale and whether they were pre-planned or post-hoc. A pre-specified subgroup analysis of the pivotal studies M2-124 and M2-125 was conducted to assess the effects of roflumilast in patients who were receiving concomitant LABA therapy, compared to those who were not. In both studies eligible patients were stratified according to their use of LABA, therefore approximately 50% of patients in each treatment arm were receiving concomitant LABA therapy during the study period (see Table B5). A subgroup analysis of these studies also investigated whether the reduction in exacerbations seen with roflumilast occurs in those at greatest risk, ie those with a history of frequent exacerbations ( 2 exacerbations in previous year) in the previous year 41. In addition a post-hoc, pooled analysis of studies M2-111 and M2-112 was conducted to determine the effects of roflumilast in a precisely defined subset - patients with severe COPD and associated chronic bronchitis with or without concomitant ICS therapy 17. Participant flow Provide details of the numbers of patients who were eligible to enter the RCT(s), randomised, and allocated to each treatment. Provide details of, and the rationale for, patients who crossed over treatment groups and/or were lost to follow-up or withdrew from the RCT. This information should be presented as a CONSORT flow chart. 65

66 Figure 5: Participant Flow Diagram for M2-124 Figure 6: Participant Flow Diagram for M

67 Figure 7: Participant Flow Diagram for M2-127 Figure 8: Participant Flow Diagram for M

68 Figure 9: Pariticpant Flow Diagram for M2-111 Figure 10: Pariticpant Flow Diagram for M

69 5.4 Critical appraisal of relevant RCTs The validity of the results of an individual study will depend on the robustness of its overall design and execution, and its relevance to the decision problem. Each study that meets the criteria for inclusion should therefore be critically appraised. Whenever possible, the criteria for assessing published studies should be used to assess the validity of unpublished and part-published studies. The critical appraisal will be validated by the ERG. The following are the minimum criteria for assessment of risk of bias in RCTs, but the list is not exhaustive. Was the method used to generate random allocations adequate? Was the allocation adequately concealed? Were the groups similar at the outset of the study in terms of prognostic factors, for example, severity of disease? Were the care providers, participants and outcome assessors blind to treatment allocation? If any of these people were not blinded, what might be the likely impact on the risk of bias (for each outcome)? Were there any unexpected imbalances in drop-outs between groups? If so, were they explained or adjusted for? Is there any evidence to suggest that the authors measured more outcomes than they reported? Did the analysis include an intention-to-treat analysis? If so, was this appropriate and were appropriate methods used to account for missing data? Please provide as an appendix a complete quality assessment for each RCT. See section 9.3, appendix 3 for a suggested format. The complete quality assessments for each of the roflumilast RCTs can be found in section 9.3, appendix If there is more than one RCT, tabulate a summary of the responses applied to each of the critical appraisal criteria. A suggested format for the quality assessment results is shown below. 69

70 Table B 9: Quality assessment results for RCTs Trial no. (acronym) M2-124 & M2-125 M2-127 & M2-128 M2-111 M2-112 Was randomisation carried out appropriately? Yes. Yes. Yes. Yes. Was the concealment of treatment allocation adequate? Yes. Yes. Yes. Yes. Were the groups similar at the outset of the study in terms of prognostic factors? Were the care providers, participants and outcome assessors blind to treatment allocation? Were there any unexpected imbalances in drop-outs between groups? Is there any evidence to suggest that the authors measured more outcomes than they reported? Did the analysis include an intention-to-treat analysis? If so, was this appropriate and were appropriate methods used to account for missing data? Yes. Yes. Yes. Yes. Yes. Yes. Yes. Yes. No. No. No. No. No. No. No. No. Yes. Yes. Yes. Yes. 70

71 5.5 Results of the relevant RCTs Provide the results for all relevant outcome measure(s) pertinent to the decision problem. Data from intention-to-treat analyses should be presented whenever possible and a definition of the included patients provided. If patients have been excluded from the analysis, the rationale for this should be given. If there is more than one RCT, tabulate the responses The information may be presented graphically to supplement text and tabulated data. If appropriate, please present graphs such as Kaplan-Meier plots For each outcome for each included RCT, the following information should be provided. The unit of measurement. The size of the effect; for dichotomous outcomes, the results ideally should be expressed as both relative risks (or odds ratios) and risk (or rate) differences. For time-to-event analysis, the hazard ratio is an equivalent statistic. Both absolute and relative data should be presented. A 95% confidence interval. Number of participants in each group included in each analysis and whether the analysis was by intention to treat. State the results in absolute numbers when feasible. When interim RCT data are quoted, this should be clearly stated, along with the point at which data were taken and the time remaining until completion of that RCT. Analytical adjustments should be described to cater for the interim nature of the data. Other relevant data that may assist in interpretation of the results may be included, such as adherence to medication and/or study protocol. Discuss and justify definitions of any clinically important differences. Report any other analyses performed, including subgroup analysis and adjusted analyses, indicating those pre-specified and those exploratory. For the identically designed trials M2-124 and M2-125 the results of the pooled analysis are presented. The results of the individual studies are provided in Appendix

72 A post-hoc pooled analysis of trials M2-111 and M2-112 has been conducted and published 17. The results of this pooled analysis are not presented here but are available on request. In this pooled analysis all data from M2-111 and M2-112 were re-analysed based on the analytical methods used in the trials M2-124 and M2-125, therefore the statistical analysis used in the pooled analysis differed from the original analysis plan for the two studies. For this reason the results of the individual studies M2-111 and M2-112, based on the original analysis plans, are presented in this submission. 72

73 Table B 10: Patients Included in Efficacy Analyses for Roflumilast RCTs M-124 & M2-125 Definition of Patients Included in Analysis Efficacy analyses were conducted in the intention-to-treat population. No further details of the intention-to-treat population are provided. M2-127 Data for efficacy were evaluated in the intention-to-treat population, defined as patients given at least one dose of study medication. M2-128 Data for efficacy were evaluated in the intention-to-treat population, defined as patients given at least one dose of study medication. M2-111 Efficacy data were analysed in the intention-to-treat population, defined as all randomised patients who received at least one dose of study medication. M2-112 Efficacy data were analysed in the intention-to-treat population, defined as all randomised patients who received at least one dose of study medication. Number of Patients Randomised Number of Patients in ITT Analysis Roflumilast 1537 Placebo Roflumilast 466 Placebo Roflumilast 371 Placebo Roflumilast 567 Placebo Roflumilast 760 Placebo 753 Patients Excluded from Analysis Three patients randomly assigned to placebo and two to roflumilast did not receive study medication, therefore were excluded from the intention-to-treat analysis. One patient randomly assigned to placebo and one to roflumilast did not receive study medication, therefore were excluded from the intention-to-treat analysis. One patient randomly assigned to roflumilast did not receive study medication, therefore was excluded from the intentionto-treat analysis. One patient assigned to receive roflumilast and two assigned to receive placebo did not receive study medication, therefore was excluded from the intentionto-treat analysis. One patient randomised to roflumilast did not take any study medication therefore was excluded from the intention-to-treat analysis. Exacerbations In studies M2-124 and M2-125 a significant improvement in the rate of moderate or severe exacerbations was observed with roflumilast compared to placebo (rate ratio [RR, 95% CI] = 0.83 [ ]; p=0.0003). The patients in this study represent the sub-population of severe COPD patients for whom roflumilast is licensed, ie those with chronic bronchitis and a history of frequent exacerbations. These studies were not powered to detect a difference in severe exacerbations. Treatment with roflumilast also resulted in a significantly longer time to 1 st and 2 nd moderate or severe exacerbation compared to placebo. Studies M2-127 and M2-128 were not powered to detect a difference in exacerbations and the six month duration was too short to allow reliable 73

74 measurement of the effect of the treatments on exacerbation rates. Therefore, despite numerical improvements with roflumilast, no significant difference was observed in the rate of all exacerbations between roflumilast and placebo (M2-127: RR = 0.79 [ ]; p=0.1408: M2-128: RR = 0.84 [ ]; p=0.3573). Roflumilast did significantly reduce the proportion of patients who experienced an exacerbation (mild, moderate or severe) in both studies (M2-127: risk ratio = 0.82 [ ]; p=0.0419: M2-128: risk ratio = 0.75 [ ]; p=0.0169). M2-111 and M2-112 were conducted in a broad patient population and concomitant chronic bronchitis was not required for participation in the studies, therefore no significant difference between placebo and roflumilast in rate of moderate or severe exacerbations was seen. The sub-group analysis of patients with chronic bronchitis in these trials is presented later in this section and shows that roflumilast reduces the rate of moderate or severe exacerbations 26.2% (p=0.001) in this patient group

75 Table B 11: Effects of Roflumilast on COPD Exacerbations in the RCTs M2-124 & M2-125 Rate of exacerbations (mean rate per patient per year[95% CI]) Overall exacerbation Moderate/ severe exacerbation * Roflumilast Not reported 1.14 (1.05 to 1.24) (n=717) Placebo Not reported 1.37 (1.28 to 1.48) Difference: roflumilast vs placebo (n=821) Not reported RR 0.83 (0.75 to 0.92); p = Moderate exacerbation (wider definition) ** 1.13 (1.04 to 1.23) (n=700) 1.35 (1.26 to 1.46) (n=798) RR 0.84 (0.76 to 0.92); p = Moderate exacerbation * 0.99 (0.91 to 1.08) (n=624) 1.19 (1.10 to 1.29) (n=723) RR 0.83 (0.75 to 0.92); p = Severe exacerbation 0.12 (0.10 to 0.16) (n=157) 0.15 (0.12 to 0.19) (n=198) RR 0.82 (0.63 to 1.06); p = Median time to moderate/ severe exacerbation (days, IQR) First moderate/ Second severe moderate/ exacerbation * severe 80.0 (28.0 to 190.0) 71.0 (28.0 to 160.0) HR 0.89 (0.80 to 0.98); p = exacerbation * (92.0 to 262.0) (85.0 to 236.0) HR 0.79 (0.69 to 0.91); p = M2-127 Roflumilast + Salmeterol 1.9 (1.5 to 2.5) (n=131) Not applicable Not applicable Not applicable Not applicable 83.0 (41.0 to 102.0) Not applicable Placebo + Salmeterol 2.4 (1.9 to 3.1) (n=159) Not applicable Not applicable Not applicable Not applicable 71.0 (33.0 to 109.0) Not applicable Difference: roflumilast + salmeterol vs placebo + salmeterol RR 0.79 (0.58 to 1.08); p = Not applicable Not applicable Not applicable Not applicable HR 0.6 (0.4 to 0.9); p = Not applicable M2-128 Roflumilast + Tiotropium 1.8 (1.3 to2.5) (n=82) Not reported Not applicable Not reported Not reported 80.5 (49.0 to 124.0) Not applicable Placebo +Tiotropium 2.2 (1.7 to 2.9) (n=112) Not reported Not applicable Not reported Not reported 74.5 (35.0 to 123.0) Not applicable Difference: RR 0.84 (0.57 Not reported Not applicable Not reported Not reported HR 0.8 (0.5 to Not applicable 75

76 roflumilast + tiotropium vs placebo + tiotropium M2-111 Roflumilast (n=567) Placebo (n=606) Difference: roflumilast vs placebo M2-112 Roflumilast (n=760) Placebo (n=753) Difference: roflumilast vs placebo to 1.23); p = ); p = Not applicable Not applicable Not applicable Not applicable Not applicable Not applicable Not applicable Not applicable RR (0.715, 1.064); p = RR (0.713, 1.021)); p = RR (0.760, 1.064); p = Not applicable Not applicable Not applicable Not applicable Not applicable Not applicable Not applicable Not applicable Not applicable Not applicable Not applicable Not applicable Not applicable RR (SE 0.090); p = RR (SE 0.075); p = RR (0.090); p = Not applicable Not applicable Not applicable n= number of patients with at least one exacerbation for M2-124, M2-125, M2-127 and M2-128; n= number of patients in ITT analysis for M2-111 and M2-112 * Moderate exacerbations defined as those requiring systemic corticosteroids ** Moderate exacerbations defined as those treated with systemic corticosteroids and/or antibiotics (wider definition) 76

77 Table B 12: Proportion of patients experiencing an exacerbation M2-124 & M2-125 M2-127 Proportion of patients with a moderate* or severe Roflumilast Placebo Difference Roflumilast + Salmeterol Placebo + Salmeterol Difference exacerbation (n, %) 717 (47%) 821 (53%) RiR (11%) 83 (18%) RiR 0.60 (0.43 to 0.82); p = M2-128 M2-111 Roflumilast + Tiotropium Placebo +Tiotropium Difference Roflumilast Placebo Difference 42 (11%) 58 (16%) RiR 0.73 (0.51 to 1.05); p = (33%) 240 (40%) OR 0.74 (0.58 to 0.95); p = Data not reported in M2-112 * Moderate exacerbations defined as those requiring systemic corticosteroids Lung Function In all RCTs roflumilast significantly improved pre- and post-bronchodilator FEV 1 compared to placebo. This improvement was observed both when the effect of roflumilast alone was measured, and when in combination with LABA (salmeterol; M2-127) or LAMA (tiotropium; M2-128), suggesting that the efficacy of roflumilast in improving FEV 1 occurs independently of concomitant treatment with LABA or LAMA. As mentioned previously, the magnitude of the effect of roflumilast on FEV 1 is not expected to be as large as that of a LABA or LAMA, due to it's mode of action. However, the additional improvement in FEV 1 induced by roflumilast in patients already receiving bronchodilators is expected to provide clinically meaningful benefit to patients with severe COPD. 77

78 Table B 13: Change in FEV 1 from Baseline in Roflumilast RCTs M2-124 & M2-125 Change in prebronchodilator FEV 1 (ml; least square mean, SE) Roflumilast 40 (6) (n=1475) Placebo -9 (5) Difference: roflumilast vs placebo (mean difference, 95% CI) Change in postbronchodilator FEV 1 (ml; least square mean, SE) 50 (6) (n=1453) -4 (6) (n=1500) (n=1511) 48 (35 to 62); p < (41 to 69); p < M2-127 Roflumilast + Salmeterol 39 (9) (n = 456) 68 (9) (n=452) Placebo + Salmeterol -10 (9) (n=463) 8 (9) (n=460) Difference: roflumilast + salmeterol vs placebo + salmeterol (difference, 95% CI) 49 (27 to 71); p < (38 to 82); p < M2-128 Roflumilast + Tiotropium 65 (12) (n=365) 74 (12) (n=364) Placebo + Tiotropium -16 (12) (n=364) -7 (11) (n=363) Difference: roflumilast + tiotropium vs placebo + tiotropium (difference, 95% CI) 80 (51 to 110); p < (51 to 110); p < M2-111 Roflumilast 29 (8) (n=488) Placebo -7 (7) (n=541) Difference: roflumilast vs 36 (10); placebo (difference in least p = square means, SE) 21 (8) (n = 500) -17 (7) (n = 534) 38 (10); p = M2-112 Roflumilast (n=760) 9 (11) 12 (11) Placebo (n=753) -27 (11) -26 (11) Difference: roflumilast vs placebo (difference in least square means, SE) 36 (12); p < (12); p = n=number of patients with data available for M2-124, M2-125, M2-127 and M n=number of patients in ITT analysis for M2-111 and M Results of secondary lung function variables for these studies can be made available on request. 78

79 HRQL, Symptom Control, Mortality HRQL No significant differences were observed in the HRQL measures (SRGQ and EQ-5D) with roflumilast compared to placebo. A retrospective analysis of study M showed a greater improvement in SGRQ in patients who were GOLD stage IV at baseline (improvement of units with roflumilast vs placebo; p = 0.086), and no difference was found in those patients who were GOLD stage III at baseline between roflumilast and placebo (+1.1 units roflumilast vs placebo; p = 0.140). Symptom control A significant improvement in TDI focal score was observed with roflumilast compared to placebo in studies M2-124 and M A significant improvement in TDI focal score and SOBQ was also seen with roflumilast compared to placebo in patients receiving concomitant tiotropium (study M2-128), but not in patients receiving concomitant salmeterol (M2-127). The variation may be due to the differences in patient inclusion criteria between the studies; M2-127 included a broader population of COPD patients whereas M2-124, M2-125 and M2-128 included only patients with chronic bronchitis, which led to the recruitment of more symptomatic patients, in line with the licence for roflumilast. Mortality No difference was observed in mortality from any cause between roflumilast and placebo (M2-124 and M2-125). 79

80 Table B 14: Effects of Roflumilast on HRQL, Symptom Control and Mortality M2-124 & M2-125 HRQL Symptom Control Mortality Change from baseline in SGRQ total score (Units, SD) EQ-5D Total Score (Least square means, SE) TDI Focal Score (Least square means, SE) SOBQ Time to mortality due to any cause (Days; mean, SD) Roflumilast Not applicable (0.0043) 0.7 (0.1) Not applicable (116.4) (n=1470) (n=1470) (n=1537) Placebo Not applicable (0.0042) 0.4 (0.1) Not applicable (125.1) Difference: roflumilast vs placebo (mean difference or HR, 95% CI) Not applicable (n=1504) ( to ); p = (n=1514) 0.3 (0.1 to 0.4); p = (n=1554) Not applicable HR 1.1 (0.7 to 1.8); p = M2-127 Roflumilast + Salmeterol Not applicable Not applicable 1.2 (0.1) (n=454) -0.6 (0.7) (n=454) Not applicable Placebo + Salmeterol Not applicable Not applicable 1.1(0.1) (n=460) -1.1 (0.7) (n=461) Not applicable Difference: roflumilast + salmeterol vs placebo + salmeterol (difference, 95% CI) Not applicable Not applicable 0.1 (-0.2 to 0.4); p = (-1.2 to 2.2); p = Not applicable M2-128 Roflumilast + Tiotropium Not applicable Not applicable 1.4 (0.1) (n=364) -3.4 (0.7) (n=359) Not applicable Placebo + Tiotropium Not applicable Not applicable 0.9 (0.1) (n=364) -0.7 (0.7) (n=359) Not applicable Difference: roflumilast + tiotropium vs placebo + tiotropium (difference, 95% CI) Not applicable Not applicable 0.4 (0.1 to 0.7); p = (-4.5 to -0.8); p = Not applicable M2-111 Roflumilast (0.49) (n=444) Not applicable 0.84 (0.09) (n=541) Not applicable Not applicable Placebo (0.44) (n=533) Not applicable 0.46 (0.09) (n=595) Not applicable Not applicable Difference: roflumilast vs (-2.69, -0.27); p= Not applicable 0.38 (0.14, 0.61); p Not applicable Not applicable placebo = M2-112 Roflumilast -1.7 Not applicable Not applicable Not applicable Not applicable Placebo -2.0 Not applicable Not applicable Not applicable Not applicable Difference: roflumilast vs placebo p = Not applicable Not applicable Not applicable Not applicable n=number of patients with data available for M2-124, M2-125, M2-127 and M2-128; n=number of patients in ITT analysis for M2-111 and M

81 Subgroup Analyses of Roflumilast RCT Data M2-124 and 125 Subgroup analysis of patients receiving concomitant LABA A pre-specified subgroup analysis was conducted on the pooled data from studies M2-124 and M2-125, which investigated the effects of roflumilast with or without concomitant LABA therapy 41;42. A total of 3091 patients were randomised and treated in the two studies, 1542 of whom received concomitant therapy with LABA and 1549 who did not. In both studies eligible patients were stratified according to their use of LABA, therefore approximately 50% of patients in each treatment arm were receiving concomitant LABA therapy during the study period (see Table B5). Table B15 provides baseline characteristics of patients in the LABA and no LABA subgroups. In the LABA subgroup there were slightly fewer smokers and also a higher proportion of patients pre-treated with ICS compared to the total trial population and no LABA subgroup. In addition fewer patients in the LABA subgroup were receiving concomitant SAMA treatment. Table B 15: Baseline Characteristics of the Intention-to-Treat Population for the Pooled Studies (M2-124 and M2-125) Baseline characteristic Concomitant treatment With LABA Without LABA Roflumilast (n=749) Placebo (n=793) Roflumilast (n=788) Placebo (n=761) Age (years) 64.2 (9.3) 64.7 (9.0) 63.3 (9.4) 63.0 (9.2) Men, n (%) 558 (74.5) 591 (74.5) 592 (75.1) 595 (78.2) Cigarette pack-years 48.9 (24.9) 47.6 (23.9) 47.8 (25.5) 46.1 (22.6) Current smoker, n (%) 286 (38.2) 294 (37.1) 349 (44.3) 349 (45.9) Pre-bronchodilator FEV 1 (l) 1.0 (0.4) 1.0 (0.3) 1.0 (0.4) 1.0 (0.4) Post-bronchodilator FEV 1 (l) 1.1 (0.4) 1.1 (0.3) 1.1 (0.4) 1.1 (0.4) Pre-bronchodilator FEV 1 (% 33.2 (10.1) 33.4 (9.5) 32.9 (10.5) 33.3 (11.6) predicted) Post-bronchodilator FEV 1 (% 35.7 (10.5) 35.9 (9.7) 36.5 (10.7) 36.8 (11.7) predicted) FEV 1 reversibility increase 8.5 (12.5) 8.6 (15.0) 12.8 (16.3) 12.6 (16.3) (%) Post-bronchodilator 42.1 (11.0) 41.4 (10.5) 42.5 (11.4) 42.6 (11.3) FEV 1 /FVC (%) Body mass index (kg/m 2 ) 26.8 (5.5) 26.3 (5.4) 24.9 (6.1) 25.0 (5.8) 81

82 COPD severity, n (%) Severe 452 (60.3) 529 (66.7) 491 (62.3) 460 (60.4) Very severe 239 (31.9) 218 (27.5) 224 (28.4) 222 (29.2) Concomitant LABA, n (%) Concomitant SAMA, n (%) 162 (21.6) 178 (22.4) 426 (54.1) 438 (57.6) Prior ICS, n (%) 469 (62.6) 476 (60.0) 181 (23.0) 181 (23.8) Data shown are mean (SD) values; Efficacy Results Rate of COPD exacerbations o In the total pooled population roflumilast reduced mean rate of moderate or severe exacerbations by 16.9% (RR 0.83 [95% CI 0.75 to 0.92], p=0.0003). o In patients receiving concomitant LABA roflumilast significantly reduced the mean rate of moderate or severe exacerbations by 20.7% compared with LABA alone (RR 0.79 [95% CI 0.69 to 0.91], p=0.0011). o The time to first or second exacerbation was significantly delayed by roflumilast compared to placebo in those receiving concomitant LABA therapy. Table B16 provides further data on COPD exacerbations during therapy with roflumilast, with or without concomitant LABA. 82

83 Table B 16: Mean Exacerbation Rates and Time to First and Second Exacerbation, With or Without LABA in the Pooled Data from Studies M2-124 and M2-125 Mean moderate or severe exacerbation rate (per patient, per year)* Median time to first exacerbation (moderate or severe, days) Median time to second exacerbation (moderate or severe, days) Median time to third exacerbation (moderate or severe, days) LABA Subgroup No LABA Subgroup Full analysis set (All patients) Roflumilast + LABA vs. LABA Difference HR/RR (95% CI) Roflumilast vs. placebo Difference HR/RR (95% CI) Roflumilast vs. placebo Difference (95% CI) 1.23 vs vs vs vs % (-31 to -9) RR 0.79 (0.69 to 0.91); p= HR 0.86 (0.75 to 0.99); p= HR 0.79 (0.653 to 0.961); p= HR 0.68 (0.52 to 0.90) p= vs vs vs vs % (-26 to -1) RR 0.85 (0.736 to 0.992); p= HR (0.796 to 1.073); p= HR (0.646 to 0.999); p= HR 0.79 (0.58 to 1.09) p= vs vs vs vs % (-25 to -8) RR 0.83 (0.75 to 0.92); p= HR 0.89 (0.80 to 0.98); p= HR 0.79 (0.69 to 0.91); p= HR 0.73 (0.59 to 0.90) p = RR=rate or risk ratio, HR= Hazard ratio. *Exacerbations were defined as moderate if they required oral or parenteral glucocorticosteroids or severe if they required hospitalisation or led to death. Lung Function o In patients already receiving concomitant LABA, roflumilast significantly increased pre-bronchodilator FEV 1 by 46 ml compared to LABA alone (95% CI 29 to 64 ml, p<0.0001) (Table B17). In the total population roflumilast increased pre-bronchodilator FEV 1 by 48 ml compared to placebo (95% CI 35 to 62, p<0.0001). o Post-bronchodilator FEV 1 was significantly improved by roflumilast compared to placebo, irrespective of concomitant LABA use. 83

84 Table B 17: Mean Changes in Pre- and Post-bronchodilator FEV 1 With or Without Concomitant LABA Therapy from the M2-124 and M2-125 Studies LABA Subgroup No LABA Subgroup Full analysis set (All patients Roflumilast + LABA vs. LABA Difference (95% CI) Roflumilast vs. placebo Difference (95% CI) Roflumilast vs. placebo Difference (95% CI) Mean changes in prebronchodilator, FEV 1 (ml) 28 vs ml (29 to 64) p< vs.1 50 ml (29 to 71) p< vs ml (35 to 62) p< Mean changes in postbronchodilator, FEV 1 (ml) 36 vs ml (28 to 64) p< vs ml (42 to 86) p< vs ml (41 to 69) p< This subgroup analysis demonstrates that roflumilast provides additional efficacy benefits in terms of improving lung function and reducing exacerbations in patients with severe COPD associated with chronic bronchitis and a history of exacerbations, independently of concomitant LABA treatment. M2-124 and M2-125 Subgroup analysis of patients with a history of frequent exacerbations An analysis was also conducted to assess the efficacy of roflumilast in reducing COPD exacerbations in patients with a history of frequent exacerbations 41;42. Frequency of COPD exacerbations was determined based on patient recall. Frequent exacerbators were defined as those who had experienced 2 exacerbations in the previous year and infrequent exacerbators as those with < 2 exacerbations in the previous year. Efficacy Results Rate of COPD exacerbations o Roflumilast significantly reduced the mean rate of moderate or severe exacerbations in both frequent and infrequent exacerbators, with the greatest reduction observed in the frequent exacerbator subgroup. o The time to onset of second and third moderate or severe exacerbation was significantly delayed with roflumilast in frequent exacerbators 84

85 compared to placebo, and time to second exacerbation was significantly delayed in infrequent exacerbators. Table B 18: Mean Exacerbation Rates and Time to First and Second Exacerbation, for Frequent and Infrequent Exacerbators from Pooled Data from Studies M2-124 and M2-125 Frequent Exacerbators Infrequent Exacerbators Full analysis set (All patients) Roflumilast vs. placebo Roflumilast vs. placebo Difference HR/RR (95% CI) Difference HR/RR (95% CI) Roflumilast vs. placebo Difference HR/RR (95% CI) Mean moderate or severe exacerbation rate (per patient, per year)* 1.5 vs % (-33 to -9) RR 0.78 (0.66 to 0.91) p = vs % (-26 to -5) RR 0.84 (0.73 to 0.95) p = vs % (-25 to -8%) RR 0.83 (0.75 to 0.92); p= Median time to first exacerbation (moderate or severe, days) Median time to second exacerbation (moderate or severe, days) Median time to third exacerbation (moderate or severe, days) 69.0 vs vs vs HR 0.90 (0.75 to 1.07) p = HR 0.74 (0.59 to 0.93) p = HR 0.65 (0.48 to 0.89) p = vs vs vs HR 0.89 (0.78 to 1.00) p = HR 0.80 (0.67 to 0.97) p = HR 0.78 (0.59 to 1.03) p = vs vs vs HR 0.89 (0.80 to 0.98); p= HR 0.79 (0.69 to 0.91); p= HR 0.73 (0.59 to 0.90) p = RR=rate or risk ratio, HR= Hazard ratio. *Exacerbations were defined as moderate if they required oral or parenteral glucocorticosteroids or severe if they required hospitalisation or led to death Lung Function o Significant improvements in pre- and post-bronchodilator FEV 1 were observed with roflumilast compared to placebo, irrespective of exacerbation history. 85

86 Table B 19: Mean Changes in Pre- and Post-bronchodilator FEV 1, for Frequent and Infrequent Exacerbators from Pooled Data from Studies M2-124 and M2-125 Frequent Exacerbators Infrequent Exacerbators Full analysis set (All patients) Roflumilast vs. placebo Difference (95% CI) Roflumilast vs. placebo Difference (95% CI) Roflumilast vs. placebo Difference (95% CI) Mean changes in prebronchodilator, FEV 1 (ml) 36 vs ml (14 to 69) p= vs ml (35 to 66) p< vs ml (35 to 62) p< Mean changes in postbronchodilator, FEV 1 (ml) 47 vs ml (20 to 78) p= vs ml (41 to 73) p< vs ml (41 to 69) p< This analysis demonstrates that the greatest efficacy benefit in terms of reducing exacerbations occurs in patients with the greatest risk of experiencing exacerbations, ie those with a history of two or more exacerbations in the previous year. M2-111 and M2-112 subgroup analysis of patients with COPD and associated chronic bronchitis 17 A post-hoc analysis of the pooled data from studies M2-111 and M2-112 was conducted to assess the efficacy of roflumilast in subgroups of patients, which included those with chronic bronchitis or emphysema, and those receiving or not receiving ICS therapy 17. The results of this analysis led to the subsequent focus on COPD patients with chronic bronchitis in the roflumilast clinical programme. The pooled analysis included a total of 2686 patients. In the roflumilast arm, 817 patients (62%) had COPD with associated chronic bronchitis, 492 of whom received concomitant ICS. In the placebo arm, 847 patients (62%) had chronic bronchitis, of which 493 received concomitant ICS therapy. Efficacy Results Rate of COPD exacerbations o Significant reductions in moderate or severe exacerbations were observed with roflumilast compared to placebo for patients with chronic bronchitis, with or without co-existing emphysema (26.2% 86

87 o reduction vs placebo; p=0.001), but no effect was seen with roflumilast in patients with emphysema only (-1.1%; p=0.93) Table B20 presents the results for roflumilast in patients with chronic bronchitis, with or without concomitant ICS. In this patient group, a greater effect was seen with roflumilast in those patients who received concomitant ICS (30.2% reduction vs placebo; p=0.001) compared to those who did not receive ICS (15.5% reduction; p=0.31). Table B 20: Mean Rate of Exacerbations in patients with Chronic Bronchitis from Studies M2-111 and M2-112 All patients n 1664 Concomitant ICS Treatment 985 No ICS Treatment 679 Mean rate of moderate/severe* exacerbations per patient per year Roflumilast Placebo 0.49 (n=817) 0.61 (n=492) 0.39 (n=325) 0.66 (n=847) 0.87 (n=493) 0.46 (n=354) Difference roflumilast vs. placebo * % (CI -38 to -11) p= % (CI -44 to -13) p= % (CI -39 to 17) p= *Moderate exacerbations defined as those requiring oral or parenteral glucosteroids and severe exacerbations those requiring hospitalisation Lung Function o Roflumilast significantly improved pre- and post-bronchodilator FEV 1 in all subgroups, independent of disease history or concomitant medication. o Tables B21 and B22 present the results for patients with chronic bronchitis, with or without concomitant ICS. A greater improvement was seen in patients who were taking an ICS compared with those who were not. 87

88 Table B 21: Change in pre-bronchodilator FEV 1 in patients with Chronic Bronchitis from Studies M2-111 and M2-112 All patients 1639 Concomitant ICS Treatment n 969 No ICS Treatment 670 Mean change in prebronchodilator FEV 1 (ml) Roflumilast Placebo 45 (n=800) 48 (n=478) 34 (n=322) -2 (n=839) -7 (n=491) 0 (n=348) Difference roflumilast vs. placebo 46 (95% CI 28 to 65) p< (95% CI 33 to77) p< (95% CI 2 to 65) p= Table B 22: Change in post-bronchodilator FEV 1 in patients with Chronic Bronchitis from Studies M2-111 and M2-112 All patients 1638 Concomitant ICS Treatment n 969 No ICS Treatment 669 Mean change in postbronchodilator FEV 1 (ml) Roflumilast Placebo 48 (n=799) 44 (n=478) 44 (n=321) -3 (n=839) -16 (n=491) 8 (n=348) Difference roflumilast vs. placebo 51 (95% CI 31 to 1.0) p< (95% CI 36 to 83) p< (95% CI 4 to 68) p= Health Related Quality of Life o Significant improvements in St George's Respiratory Questionnaire (SGRQ) total score were observed with roflumilast compared to placebo in the subgroup of patients with chronic bronchitis, and in those who received concomitant ICS (Table B23). 88

89 Table B 23: Change in SGRQ Total Score from Baseline in Patients with Chronic Bronchitis from Studies M2-111 and M2-112 All patients n 1537 Concomitant ICS Treatment 910 No ICS Treatment 627 Change in SGRQ from Baseline Roflumilast (n=734) (n=438) (n=296) Placebo (n=803) (n=472) (n=331) Difference roflumilast - placebo (95% CI to ) p= (95% CI to ) p= (95% CI to 0.700) p= This subgroup analysis shows that roflumilast effectively improves FEV 1, reduces the rate of moderate or severe exacerbations and improves SGRQ total score in patients with severe COPD and associated chronic bronchitis, and that these effects are not blunted by concomitant ICS use. The findings from this analysis (that is, roflumilast confers particular therapeutic benefits in patients with COPD associated with chronic bronchitis) informed the design of the pivotal 1 year trials for roflumilast (M2-124 and M2-125). 89

90 5.6 Meta-analysis When more than one study is available and the methodology is comparable, a meta-analysis should be undertaken. This section should be read in conjunction with NICE s Guide to the methods of technology appraisal, sections to A systematic review of PDE4 inhibitors for COPD was published by the Cochrane Collaboration on 11 th May 43. The authors concluded that treatment with PDE4 inhibitors (roflumilast and cilomilast) was associated with a significant improvement in FEV 1 and a significant reduction in COPD exacerbations compared to placebo, and a small improvement was observed in quality of life (measured by SGRQ) and COPD symptoms. It was noted that non-serious adverse events were more common in the treatment group compared to control. 90

91 5.7 Indirect and mixed treatment comparisons Data from head-to-head RCTs should be presented in the reference-case analysis, if available. If data from head-to-head RCTs are not available, indirect treatment comparison methods should be used. This section should be read in conjunction with NICE s Guide to the methods of technology appraisal, sections to An independently performed mixed treatment comparison (MTC) meta-analysis and an additive MTC analysis were conducted using published literature to assess the effects of various combinations of COPD treatments, with regard to their efficacy in reducing COPD exacerbations 22. The analyses provide information on the effects of roflumilast in combination with other therapies recommended by NICE, for which direct RCT evidence is not available, thus allowing the comparisons stated in the decision problem to be assessed. The primary outcome was evaluation of the event rate of exacerbations. A sensitivity analysis which evaluated binomial outcomes (i.e. ever having an exacerbation during the treatment period) was conducted to confirm findings from the primary analysis and to satisfy methodological considerations regarding the assessment and comparison of exacerbations in COPD Describe the strategies used to retrieve relevant clinical data on the comparators and common references both from the published literature and from unpublished data. The methods used should be justified with reference to the decision problem. Sufficient detail should be provided to enable the methods to be reproduced, and the rationale for any inclusion and exclusion criteria used should be provided. Exact details of the search strategy used should be provided in section 9.4, appendix 4. A systematic literature search was conducted to retrieve relevant clinical data for the analyses. The databases Medline, Embase, Medline (R) In-Process and the Cochrane Library were searched independently and in duplicate. The search strategy used is provided is Section 9.4, Appendix 4. Any published RCT which evaluated patients with moderate to severe COPD (defined by GOLD/NICE criteria) for a minimum duration of 24 weeks (6 months) was included. RCTs evaluating the following interventions were considered: LABA (formoterol or salmeterol), LAMA (tiotropium), ICS (fluticasone or budesonide), 91

92 roflumilast, and any combinations of these interventions. Control interventions included these active interventions or placebo. The primary analysis evaluated the event rate of exacerbations, therefore studies which provided the rates of exacerbations or the total number of exacerbations in each patient arm were included. The sensitivity analysis which evaluated the binary outcome of ever having an exacerbation event included studies which reported the number of patients in each treatment arm who had experienced at least one exacerbation during the study period. Inclusion and exclusion criteria are provided in the table below. Table B 24: Eligibility criteria used in search strategy Clinical effectiveness Inclusion criteria Exclusion criteria Study design randomised controlled trials Study duration at least 24 weeks (6 months) Population - moderate to severe COPD patients, defined by GOLD/NICE criteria (FEV 1 80% predicted) Interventions LABA (formoterol or salmeterol), LAMA (tiotropium), ICS (fluticasone or budesonide), roflumilast, and any combinations of these interventions Outcome exacerbations of COPD Language restrictions publication in English Study duration less than 24 weeks (6 months) Pharmacokinetic studies and proof of concept studies Language restrictions publication not in English The eligibility criteria permitted the inclusion of studies of patients taking concomitant medications with moderate to severe COPD; if the inclusion criteria was limited to include only studies of patients with severe COPD (FEV 1 50% predicted), only 8 published RCTs would have been eligible. An MTC derived from this data set would not have been informative. The following figure shows the study flow diagram. 92

93 Figure 11: Flow Diagram of Included and Excluded Studies in the MTC Please follow the instructions specified in sections 5.1 to 5.5 for the identification, selection and methodology of the trials, quality assessment and the presentation of results. Provide in section 9.5, appendix 5, a complete quality assessment for each comparator RCT identified. The identification and selection of trials is described above. Twenty six published studies met the inclusion criteria, reporting on 29 clinical trials. All of these trials reported on COPD exacerbations following treatment, and all recruited patients with a FEV 1 < 80% predicted value. The duration of treatment in these studies ranged from 24 weeks to 4 years. All trials permitted the use of background therapy and/or rescue therapy. All 26 of these published studies provided data on the total number of exacerbations and/or the mean annual rate of exacerbations and were included in the primary analysis of exacerbation event rates. An additional 3 publications provided data on the proportion of patients with at least one exacerbation during the study and were included in the sensitivity analysis of the binomial outcome. 93

94 A summary of the key aspects of the methodology of the 26 published studies, the additional 3 publications included in the binomial sensitivity analysis, and the additional study included in the sensitivity analysis which included recently published roflumilast data, are provided in the tables below (Tables B25, B26 and B27). A quality assessment for the trials included in the mixed treatment comparison and the additive analysis is provided in Section 9.5, Appendix 5. 94

95 Table B 25: Summary of Methodology - Trials Included in the MTC Analysis (i.e. those which included data on the total number of exacerbations and/or the mean annual rate of exacerbations) Trial Ref Trial Design (blinding, phase, randomisation, location) # of Patients Treatment Comparisons Durati on of Treatm ent Baseline FEV 1 % Predicted (Inclusion criteria) Baseline FEV 1 Measured Pre- or Post- Bronchod ilator Permitted Concomitant Medication Primary Outcomes Secondary Outcomes Definition of Exacerbation Paggiaro et al, Multicentre, double-blind, randomised, placebocontrolled study in 13 European countries, New Zealand and South Africa. 281 Fluticasone proprionate 500 µg bid (142) vs Placebo bid (139) 6 months FEV % Not reported in publication. Patients could take SABA for the relief of symptoms as required. Other COPD medications, such as anticholinergics and xanthine derivatives could be continued without dose changes. Number of patients who had had at least one exacerbation at the end of treatment period. Number and severity of exacerbation s, clinic lung function, diary card symptoms and peak expiratory flow and 6 min walking distance.) Defined as worsening of COPD symptoms, requiring changes to normal treatment, including antimicrobial therapy, short courses of oral steroids, and other bronchodilator therapy. Mild - selfmanaged by the patient at home; moderate -, requiring treatment by a family physician or as a hospital outpatient; severe - resulting in admission to hospital. 95

96 Burge et al, Casaburi et al., Rossi et al., Double-blind, randomised, placebocontrolled study in 18 hospitals in the UK Two identical randomized, double-blind, placebocontrolled in 50 clinical centres. Randomised, parallel group, 742 Fluticasone Proprionate 500 µg bid (372) vs Placebo (370) 921 Tiotropium 18 µg od (550) vs Placebo (371) 625 Formoterol 12 or 24 µg (425) 3 years FEV 1 < 85% Postbronchodil ator 1 year FEV 1 65% Not reported in publication 1 year FEV 1 <70% Not reported in Nasal and ophthalmic corticosteroids, theophyllines, and all other bronchodilators were allowed during the study. Albuterol metered-dose inhaler, as needed, stable doses of theophylline (i.e unchanging doses that had been used for 6 weeks prior to entry), inhaled glucocorticoster oids and the equivalent of 10mg/day oral prednisone throughout the study. Inhaled salbutamol (up Decline in postbronchodil ator FEV 1 (ml/year) Proportions of patients experiencing at least one COPD exacerbation and hospitalization associated with such exacerbations were compared across treatment groups using logistic regression analysis. Standardised AUC fro FEV 1 Frequency of exacerbation, changes in health status, withdrawals because of respiratory disease, morning serum cortisol concentration s, and adverse events. Not reported Standardised AUC for FVC, Defined as worsening of respiratory symptoms that required treatment with oral corticosteroids or antibiotics, or both, as judged by the general practitioner; specific symptom criteria were not used. Defined as a complex of respiratory events (cough, wheezing, dyspnoea or sputum production) lasting >3 days and generally requiring treatment with antibiotics and/or oral steroids. Defined as undergoing a 96

97 double-blind study in 81 centres worldwide. Brusasco et al., Two 6-month randomised, placebo controlled, double blind, double dummy studies in 18 countries. 1,20 7 vs Placebo (200) Tiotropium 18 µg od (402) vs Salmeterol 50 µg bid (405) vs Placebo (400) 6 months FEV 1 65% publication Not reported in publication to 8 puffs per day) was allowed as the rescue medication. Short courses of antibiotics, oral corticosteroids, and/or oxygen were permitted in case of exacerbations or respiratory infection up to two times during the study. Oral steroids bursts or theophylline for exacerbations measured over a period of 12 hours following the morning dose of study treatment, after 3 and 12 months. Health outcomes consisting of exacerbations, health resource use, dyspnoea (assessed by the transitional absolute FEV 1 values at all individual time points during each 12 hour spirometry period, predose FEV 1 at all 3 monthly visits, daily morning premedicatio n PEF, daily total symptom score, daily number of puffs of rescue salbutamol, frequency of COPD exacerbation s, and QOL. Not reported course of additional therapy (ie, corticosteroids, antibiotics or oxygen); and (severe), COPDrelated hospitalisations. Defined as a complex of respiratory symptoms (new onset or an increase in at least one of cough, sputum, dyspnoea, 97

98 Calverley et al., Calverley et al., Randomised, double-blind, placebocontrolled, parallel-group in 109 centres in 15 countries or regions. Randomised, double-blind, parallel-group, placebocontrolled study ,02 2 1,46 5 Budesonide/for moterol 2x160 µg/4.5 µg bid (254) vs Formoterol 2x4.5 µg bid (255) vs Budesonide 2x200 µg bid (257) vs Placebo (256) Salmeterol 50 µg bid (372) vs Fluticasone 500 µg bid (374) vs Salmeterol/fluti casone 50/500 1 year FEV 1 50% Prebronchodil ator. 1 year FEV % Prebronchodil ator. Oral corticosteroids (maximum 3 weeks per course) and antibiotics in the event of exacerbations. Parenteral steroids and/or nebulised treatment (single injections/inhal ations) at emergency visits. Inhaled salbutamol as relief medication throughout the study, and dyspnoea index, TDI), health related quality of life (assessed by St George s Respiratory Questionnaire, SGRQ), and Spirometry (see Methods on the first page) Time to first exacerbation and change in FEV1. Pretreatment forced expiratory volume in 1 s (FEV1) after 12 months Number of exacerbation s, time to and number of oral corticosteroid treated episodes, morning and evening PEF, slow VC, HRQL, symptoms, use of reliever medication and AEs. Other lung function measurement s, symptoms and rescue treatment wheeze, chest discomfort) lasting at least 3 days and usually associated with therapeutic intervention. Defined as symptoms requiring medical intervention (oral antibiotics and /or corticosteroids or hospitalization) Defined a priori as a worsening of symptoms that required treatment with antibiotics, oral corticosteroids, or 98

99 Szafrans ki et al., hospitals in 25 countries. Randomised, double-blind, placebocontrolled, parallel-group study. 89 centres in 11 countries. µg bid (358) vs Placebo (361) 812 Formoterol 4.5 µg bid (201) vs Budesonide/for moterol 160/4.5 µg bid (208) vs Budesonide 200 µg bid (198) vs Placebo (205) 1 year FEV 1 50% Not reported in publication regular treatment with anticholinergics, mucolytics, and theophylline as needed. All non-copd medications could be continued if the dose remained constant whenever possible, and if their use would not be expected to affect lung function. Terbutaline 0.5 mg was allowed (no preventive medications allowed) treatment and after patients had abstained from all bronchodilators for at least 6 h and from study medication for at least 12 h. Number of severe exacerbations and FEV1. use, the number of exacerbation s, patient withdrawals, and diseasespecific health status. Vital capacity (VC), HRQL and diary card data: COPD symptoms, peak expiratory flow (PEF), use of reliever medication and mild exacerbation s. both. Episodes that required corticosteroid treatment or hospital admission were noted separately. Severe exacerbations were defined as the use of oral steroids and/or antibiotics and/or hospitalization due to respiratory symptoms. Mild exacerbations were defined as a day with 4 inhalations of reliever medication above the mean run-in use. 99

100 Niewoeh ner et al., Rabe et al., 2005 (M2-107) 20 Randomized, double-blind study in 26 Veterans Affairs medical centres in the United States. Phase III, multicentre, double-blind, randomised, placebocontrolled study in 159 centres in 11 countries 1,82 9 1,41 1 Tiotropium 18 µg od (914) vs Placebo (915) Roflumilast 250 µg or 500 µg od (1,131) vs Placebo (280) 6 months 6 months FEV 1 < 60% FEV % Not reported in publication Postbronchodil ator. All other respiratory medications (including inhaled corticosteoids and long-acting beta-agonists). Also, antibiotics and systemic steroid prescriptions for exacerbations. Concomitant respiratory medications permitted included salbutamol as rescue therapy, and shortacting anticholinergics at a constant daily dose. Oral corticosteroids Percentage of patients with a COPD exacerbation and the percentage of patients with a hospitalization due to COPD exacerbation Postbronchodil ator FEV1 and St. George s respiratory questionnaire (SGRQ) total score and were calculated as the change from baseline to the endpoint (last observation Time to first COPD exacerbation, time to first hospitalisatio n due to COPD exacerbation, frequency of exacerbation and exacerbation related health-care utilisation, frequencies of all-cause hospitalisatio ns and hospitalisatio n days, and results of spirometry. Change from baseline in prebronchodil ator FEV1, postbronchod ilator FVC, postbronchod ilator forced expiratory volume in the first 6 s (FEV), postbronchod A complex of respiratory symptoms (increase or newonset) of more than 1 of the following: cough, sputum, wheezing, dyspnea, or chest tightness with a duration of at least 3 days requiring treatment with antibiotics or systemic steroids, hospitalization, or both. Mild exacerbations were defined as an increase in bronchodilator use on 2 or more consecutive days (>4 puffs of salbutamol) without the need for additional healthcare contact. Moderate exacerbations were 100

101 Dusser et al., Randomized, double-blind, parallel-group. 177 centres in France 1,01 0 Tiotropium 18 µg od (500) vs Placebo (510) 1 year FEV % Prebronchodil ator were also permitted for the treatment of exacerbations. Any additional medication deemed necessary to treat exacerbations (excluding anticholinergics and long-acting b2-agonist) carried forward) Endpoints included PEF and spriometric tests (FEV 1, FVC, SVC and inspiratory capacity), exacerbation rates, healthcare resource (not reported as primary or secondary) ilator forced expiratory flow between 25% and 75% of the vital capacity (FEF), and number of COPD exacerbation s Not reported, see primary outcomes. defined as home management with administration of oral glucocorticosteroid treatment or unscheduled healthcare contact, or both. Severe exacerbations were defined as the need for hospital admission (or emergency room treatment) Defined as the onset of at least one clinical descriptor (worsening of dyspnoea, cough, or sputum production; appearance of purulent sputum; fever (>38 C); appearance of new chest radiograph abnormality) lasting 2 days and requiring a new prescription or an increase in dose of beta-2 agonists, antibiotics, corticosteroids, or 101

102 Stockley et al., Aaron et al., Randomized double blind study in 19 European countries. Randomized, double-blind, placebocontrolled trial in 27 medical centres in Canada. 634 Salmeterol 50 µg bid (316) vs Placebo (318) 449 Tiotropium 18 µg od plus salmeterol 2x25 µg bid (148) vs Tiotropium 18 µg od plus fluticasone/sal meterol 2x250/25 µg bid (145) vs Tiotropium 18 µg od plus Placebo (156) 1 year FEV 1 <70% Prebronchodil ator 1 year FEV 1 <65% Postbronchodil ator Doses up to 2000 µg/day beclomethason e dipropionate or equivalent or 1000 µg/day fluticasone propionate were allowed during the study, as were usual treatments such as methylxanthine s, short-acting anti-cholinergic agents, and mucolytics Inhaled abuterol for use when necessary. Use of any treatment with inhaled corticosteroids, long-acting beta 2- agonists, and anticholinergics not permitted. Therapy with other The number of exacerbations in the intent-totreat (ITT) population Proportion of patients in each treatment group who experienced a COPD exacerbation within 52 weeks of randomization Mild exacerbation s, lung function, diary card parameters, and health status Mean number of COPD exacerbation s per patientyear; the total number of exacerbation s that resulted in urgent visits to a health care provider or emergency department; the number bronchodilators. Defined as worsening of symptoms requiring a change in medications. Defined according to the 2000 Aspen Lung Conference Consensus definition. An exacerbation is "a sustained worsening of he patient's respiratory condition, from the stable state and beyond normal day-to-day variations, necessitating a change in regular 102

103 Calverley et al., Calverley et al., 2007 (M2-112) 34 Randomized, double-blind trial in 444 centres in 42 countries. Randomized, double-blind trial in 159 centres in 14 countries. 6,11 2 1,51 3 Salmeterol 50µg bid (1,521) vs Fluticasone 500µg bid (1,534) vs Salmeterol/fluti casone 50 µg/500 µg bid (1,533) vs Placebo (1,524) Roflumilast 500 µg od (760) vs Placebo (753) 3 years FEV 1 <60% Prebronchodil ator 1 year FEV 1 50% Postbronchodil ator. respiratory medications, such as oxygen, antileukotrienes, and methylxanthine s permitted. Continuation of co-medications, except for corticosteroids and inhaled long-acting bronchodilators Inhaled corticosteroids of 2000 µg or less, beclomethason e dipropionate or equivalent and shortacting anticholinergics allowed. Salbutamol Time to death from any cause by 3 years. Change from baseline to endpoint in postbronchodilator FEV1 and the number of moderate or severe exacerbations per patient per year. of hospitalizatio ns for COPD; the total number of hospitalizatio ns for all causes; and changes in health-related quality of life, dyspnea, and lung function. Frequency of exacerbation s, and health status, as assessed according to scores on the St. George s Respiratory Questionnair e Change from baseline in St. George s Respiratory Questionnair e (SGRQ) total score was the main secondary variable. Other efficacy medication in a patient", Defined as a symptomatic deterioration requiring treatment with antibiotics agents, systemic corticosteroids, hospitalization, or a combination. Moderate exacerbations were defined as symptomatic deteriorations treated with systemic corticosteroids and/or antibiotics. Severe exacerbations were defined as 103

104 Chan et al., Multicentre, randomized, double-blind, placebocontrolled, 913 Tiotropium 18 µg od (608) vs Placebo (305) 48 weeks FEV 1 <65% Not reported in publication available to patients as rescue therapy. Permitted oral corticosteroids (at a stable dose of 10 mg or less of Morning predose (trough) FEV1 at study end (48 weeks). variables included change from baseline in prebronchodilato r FEV1, postbronchodilato r forced expiratory volume in 6 seconds (FEV6), FVC, forced expiratory flow between 25 and 75% of the vital capacity (FEF25 75), and number of moderate or severe COPD exacerbation s requiring systemic corticosteroid treatment per patient per year FVC and FEV6, change from baseline in St. symptoms requiring hospitalization. Defined as a complex of respiratory symptoms (new onset or increase in 104

105 parallel-group study in 101 centres in Canada prednisone daily or equivalent), stable doses of inhaled corticosteroids, theophylline preparations, mucolytic preparations (not containing bronchodilators ), LABA and, for acute symptom relief, salbutamol as needed. George s Respiratory Questionnair e (SGRQ) total score. Other efficacy variables included number of exacerbation s and associated hospitalizatio ns, and the number of courses of both oral steroids and antibiotics administered for treatment of exacerbation s. Health care resource use, specifically related to COPD. at least one of cough, sputum, sputum purulence, dyspnea, wheeze, chest discomfort) lasting at least 3 days and requiring treatment with antibiotics and/or systemic steroids. Kardos et al., Randomized, double-blind, parallel-group study in 95 respiratory 994 Salmeterol 50 µg bid (487) vs Salmeterol/fluti casone 50/500 µg bid (507) 44 weeks FEV 1 <50% Postbronchodil ator. All non-copd medications, including preexisting selective beta- Number of moderate and severe exacerbations in each Time to first exacerbation, prebronchodilato r peak flow Defined as an increase in or the new onset of respiratory symptoms that may 105

106 Powrie et al., centres in Germany Single-centre, double-blind, randomised, placebocontrolled trial in a single centre. 142 Tiotropium 18 µg od (69) vs Placebo (73) 1 year FEV 1 <80% Not reported in publication blocker therapy, could be continued if the dose remained constant. Acute moderate and severe exacerbations treated with systemic corticosteroids and/or antibiotics. Except for other anticholinergics, pulmonary drugs were permitted treatment group. Concentration of interleukin (IL)-6 in sputum. (PEF), postbronchodilato r FEV1, and diseasespecific quality of life as evaluated by the St. George s Respiratory Questionnair e (SGRQ), which investigated three different domains consisting of activity, symptom, and impact scores. Sputum IL-8 and myeloperoxid ase (MPO) levels, serum IL-6 and C- reactive protein (CRP) levels, sputum bacterial colonisation, FEV1 and exacerbation require oral steroids and/or antibiotics and/or hospitalisation. Defined as the presence, for 2 days consecutively, of an increase in any 2 major symptoms (dyspnoea, sputum purulence, and sputum volume) or in 1 major and 1 minor symptom (wheeze, sore throat, cough, and symptoms of a 106

107 Ferguson et al., Tashkin et al., Randomized, double-blind, parallel-group study in 94 research sites in the United States and Canada Randomized, double-blind trial in 487 centres in 37 countries. 776 Salmeterol 50 µg bid (385) vs Fluticasone propionate/sal meterol 250/50 µg bid (391) 5,99 3 Tiotropium 18µg od (2,987) vs Placebo (3,006) 1 year FEV 1 50% Not reported 4 years FEV1<70% Postbronchodil ator Oral corticosteroids and antibiotics were allowed for the acute treatment of exacerbations All respiratory medication allowed, except inhaled anticholinergic drugs. Annual rate of moderate to severe exacerbations. Yearly rate of decline in the mean FEV1 before the use of a study drug and shortacting bronchodilators in the morning (prebronchodilator) and after the use of a study drug (postbronchodilator) from day 30 (steady state) until completion of double-blind treatment frequency. Time to first moderate to severe exacerbation, the annual rate of exacerbation s requiring oral corticosteroid s, and predose FEV1. Rate of decline in the mean forced vital capacity (FVC) and slow vital capacity (SVC); health-related quality of life, as measured by the total score on St. George s Respiratory Questionnair e (SGRQ), in which scores range from 0 to 100, with lower scores indicating improvement common cold) Defined as the worsening of 2 or more major symptoms or 1 major and 1 minor symptom for 2 or more consecutive days Defined as an increase in or the new onset of >1 respiratory symptom (cough, sputum, sputum purulence, wheezing, or dyspnea) lasting 3 days and requiring an antibiotic or corticosteroid. 107

108 Tonnel et al., Randomized, double-blind, multicenter study in 123 centres in France 554 Tiotropium 18 µg od (266) vs Placebo (288) 9 months FEV % Pre- and postbronchodil ator Use of theophylline preparations, mucolytics, inhaled corticosteroids, and oral steroids were allowed if dosage was stabilized for 6 or more weeks before study entry. One 10- day course of oral steroids was permitted Proportion of patients achieving a reduction of at least 4 units in the SGRQ total score at study end (Month 9) and a change of 4 units or more considered to be clinically meaningful; exacerbation s of COPD (as defined below) and related hospitalizatio ns; and the rate of death from any cause and from lower respiratory conditions FEV 1, FVC, IC, SVC, FIV 1, and exacerbation frequency, adverse events. Defined as a sustained worsening of the patient's COPD (from the stable state and beyond normal day-to-day variation) that was acute in onset and necessitated a change in regular medication. 108

109 Wedzich a et al., Anzueto et al., Calverley et al., Randomized, double-blind, double-dummy parallel study in 179 centres in 20 countries. Randomized, double-blind, parallel-group, multicenter study in 98 research centres in the United States and Canada. Two placebocontrolled, 1,32 3 Salmeterol/fluti casone 50/500µg bid (658) vs tiotropium 18 µg od (665) 778 Salmeterol 50 µg bid (393) vs fluticasone propionate/sal meterol 250/50µg bid (385) 3,09 1 Roflumilast 500 µg od (1,537) 2 years FEV 1 <50% Postbronchodil ator 1 year FEV 1 50% Not reported in publication 1 year FEV 1 50% Postbronchodil for treatment exacerbations. Antibiotics could also be used for treatment exacerbations. Short acting inhaled betaagonists for relief therapy and standardized short courses of oral systemic corticosteroids and/or antibiotics where indicated for treatment of exacerbations. Albuterol was provided asneeded. Ipratropium was permitted throughout the study. Oral corticosteroids and antibiotics were allowed for the treatment of exacerbations. Short-acting 2- agonist as Rate of health care utilization (HCU) exacerbations, defined as those that required treatment with oral corticosteroids and/or antibiotics or required hospitalization Annual rate of moderate/ severe exacerbations Change in prebronchodilator Health status measured by the SGRQ, post-dose FEV1 (measured 2 h after inhalation of study medication), and study withdrawal rate Time to first moderate/sev ere exacerbation, the annual rate of exacerbation s requiring oral corticosteroid s, and predose FEV1 The postbronchodilato Defined as symptoms that required treatment with oral corticosteroids and/or antibiotics or required hospitalization. Defined as the worsening of 2 or more major symptoms or 1 major and 1 minor symptom for 2 or more consecutive days Moderate exacerbations were 109

110 2009 (M2-124 and M2-125) 18 Fabbri et al., 2009 (M2-127) 19 double-blind, multicentre trials (M2-124 and M2-125) in 246 centres in 10 countries (M2-124) and 221 centres in 8 countries (M2-125) Double-blind, multicentre study in 85 centres in 7 countries vs Placebo (1,554) 933 Salmeterol 50 µg bid plus Roflumilast 500 µg od (466) vs Salmeterol 50 µg bid plus Placebo bid (467) 6 months FEV % ator Postbronchodil ator needed. Longacting and short-acting anti-cholinergic drugs at stable doses. Besides salmeterol, no inhaled corticosteroids, short-acting anti-cholinergic drugs, other longacting bronchodilator drugs, theophylline, or other FEV1 during treatment and the rate of COPD exacerbations, defined as moderate if they required oral or parenteral corticosteroids, or severe if they were associated with admission or death Change in mean prebronchodilator FEV1 from baseline to each postrandomisation visit. r FEV1 (change from baseline during treatment), time to death from any cause, natural logtransformed C-reactive protein concentration (a possible marker of systemic inflammation in COPD; change from baseline to study end) and TDI focal score (during treatment) Postbroncho dilator FEV1 and FVC, TDI score, SOBQ, rate of COPD exacerbation s, and use of rescue medications defined as symptoms requiring corticosteroids. Severe exacerbations were defined as those resulting in hospitalization or death. Additionally, data for the total number of COPD exacerbations (defined as above together with those treated with antibiotics alone) were gathered. Mild exacerbations were defined as the need for an increase in rescue medication of at least 3 puffs per day on at least 2 consecutive days. Moderate exacerbations were defined as the need for oral 110

111 Fabbri et al (M2-128) 19 Double-blind, multicentre study in 85 centres in 7 countries 743 Tiotropium 18 µg od plus roflumilast 500 µg od (371) vs Tiotropium 18 µg od plus Placebo (372) 6 months FEV % Postbronchodil ator respiratory drugs were allowed Besides tiotropium, no inhaled corticosteroids, short-acting anti-cholinergic drugs, other long-acting bronchodilator drugs, theophylline, or other respiratory drugs were allowed Change in mean prebronchodilator FEV1 from baseline to each postrandomisation visit. Postbronchodilato r FEV1 and FVC, TDI score, SOBQ, rate of COPD exacerbation s, and use of rescue medications corticosteroids (not antibiotics). Severe exacerbations were defined as the need for treatment in hospital or death. Mild exacerbations were defined as the need for an increase in rescue medication of at least 3 puffs per day on at least 2 consecutive days. Moderate exacerbations were defined as the need for oral corticosteroids (not antibiotics). Severe exacerbations were defined as the need for treatment in hospital or death. 111

112 Table B 26: Summary of Methodology - Additional Trials Included in the Sensitivity Analysis Examining the Binomial Outcome of Ever Having an Exacerbation Event (those trials which provided data on the proportion of patients with at least 1 exacerbation during the trial) Trial Ref Bourbea u et al., Chapma n et al., Trial Design (blinding, phase, randomisation, location) Phase I, single blind study, followed by phase II, double blind, randomised, parallel group trial in 1 outpatient clinic in Canada Double-blind, randomized, parallel-group study in 28 centres in Canada, 6 in Denmark, 3 in the Netherlands, 4 in Russia, 4 in Sweden, and 7 in the United Kingdom # of Patients Treatment Comparisons 79 Budesonide 400 µg bid (39) vs Placebo (40) 408 Salmeterol 50 µg bid (201) vs Placebo (207) Duration of Treatment Baseline FEV 1 % Predicted (Inclusion criteria) 1 year FEV 1 <65% (prebronchodilat or); FEV 1 <80% (postbronchodilat or) 6 months FEV 1 85% Baseline FEV 1 Measured Preor Post- Bronchodilator Postbronchodilator and prebronchodilator Not reported in publication Permitted Concomitant Medication Regular treatment with at least one bronchodilator. Methylxanthines, anticholinergic agents and inhaled steroids were allowed. Primary Outcomes Forced expiratory volume in one second (FEV1) Morning and evening PEFR, and daytime and nighttime symptom scores. Secondary Outcomes Exercise capacity, dyspnoea with exertion, quality of life, peak expiration flow rate, and respiratory symptoms Use of rescue medication, total score on SGRQ, exacerbations of COPD and related hospitalization s; and the rate of death from any cause and from lower respiratory conditions. Definition of Exacerbation Defined as worsening of symptoms leading to treatment Defined as a worsening of respiratory disease requiring a change in medication and/or hospital care, emergency room care, or an unscheduled outpatient visit. 112

113 Vogelmei er et al., Randomised trial. Doubleblind for treatment comparisons formoterol 10 mg twice daily (b.i.d.) vs. Placebo b.i.d, and tiotropium 18 mg o.d. + formoterol 10 mg b.i.d. vs. tiotropium 18 mg o.d. + placebo b.i.d. but not for other comparisons as tiotropium was administered open-label. In 30 centres in Germany, 19 in Italy, 9 in the Netherlands, 9 in Russia, 7 in Poland, 4 in Czech Republic, 4 in Spain, and 4 in Hungary 812 Formoterol 10 µg bid (204) vs Tiotropium 18 µg od (209) vs Formoterol 10 µg bid plus Tiotropium 18 µg od (196) vs Placebo (203) 6 months FEV 1 <70% Not reported in publication Salbutamol was permitted as rescue medication. Inhaled corticosteroids at stable doses were also allowed. FEV1 measured 2 h postdose after 24 weeks of treatment FEV1 and FVC at other time points during the study (5 min, 2 and 3 h post-dose following the first dose of treatment, and after 12 and 24 weeks of treatment); COPD exacerbations; symptom scores, rescue medication use and PEF; quality of life, and 6-minute walking distance. Defined as at least two COPD symptoms, possibly requiring additional treatment or hospitalizations. 113

114 Table B 27: Summary of Methodology - Additional Trial Included in the Sensitivity Analysis Including Previously Unpublished Roflumilast Data (M2-111 Trial) Trial Ref Rennard et al., Trial Design (blinding, phase, randomisation, location) Post-hoc, pooled analysis of two phase IIIrandomised, double blind, placebocontrolled studies (M2-111 and M2-112), conducted in 6 countries (M2-111) and 14 countries (M2-112) # of Patients Treatment Comparisons 2686 Roflumilast 500 µg (1327) vs Placebo (1359) Duration of Treatment Baseline FEV 1 % Predicted (Inclusion criteria) Baseline FEV 1 Measured Preor Post- Bronchodilator Permitted Concomitant Medication Salbutamol was permitted as rescue medication. A constant daily dose of ICS (2000µg beclometasone diproprionate or equivalent) or SAMA were allowed if used before study entry. All other COPD medications were prohibited. Primary Outcomes 1 year FEV 1 50% Postbrochodilator Prebronchodil ator FEV 1 and rate of moderate or severe exacerbati ons per patient year. Secondary Outcomes Postbronchodilator FEV 1. Definition of Exacerbation Need for oral or parenteral corticosteroid treatment, hospitalisation or death. 114

115 5.7.3 Provide a summary of the trials used to conduct the indirect comparison. A suggested format is presented below. Network diagrams may be an additional valuable form of presentation. The following diagram (Figure 12) shows the network of treatments compared in the MTC and additive meta-analyses for exacerbations. The network forms a closed loop which is amenable to MTC analyses Figure 12: The network of 10 treatments included in the MTC analysis of exacerbations Each node represents a treatment in the network. The links between the nodes represent a direct comparison between pairs of treatments. The numbers shown along the links indicate the number of trials comparing pairs of treatments head to head. 115

116 Table B 28: Summary of the trials used in the assessment of total number of exacerbations and/or the mean annual rate of exacerbations No. trials References of trials PBO FLUT BUD SAL FOR TIO ROF FLUT + SAL 2 7 Paggiaro et al, 1998;Burge et al, 2000 Casaburi et al, 2002; Niewoehner et al, 2005; Dusser et al, 2006; Chan et al, 2007; Powrie et al, 2007; Tashkin et al, 2008; Tonnel et al, Rossi et al, Brusasco et al, 2003; BUD + FOR Calverley et al, 2003; Szafranski et al, 2003 Calverley et al, 2003; Calverley et al, 2007 Rabe et al, 2005; Calverley et al, 2007; Calverley et al, 2009 (M2-124 and M2-125) 1 Stockley et al, 2006 TIO + SAL SAL + ROF TIO + ROF FLUT + SAL + TIO 1 Aaron et al, Kardos et al, 2007; Ferguson et al, 2008; Anzueto et al, Wedzicha et al, Fabbri et al, 2009 (M2-127) Fabbri et al, 2009 (M2-128) (Included above) 116

117 Table B 29: Summary of the additional trials used in the binomial analysis (number of patients with 1 exacerbation) No. trials References of trials PBO FLUT BUD SAL FOR TIO ROF FLUT + SAL 1 Bourbeau et al, Chapman et al, Vogelmeier et al, 2008 BUD + FOR TIO + SAL FOR + TIO SAL + ROF TIO + ROF FLUT + SAL + TIO For the selected trials, provide a summary of the data used in the analysis. The primary outcome of interest for the analyses was exacerbation rates and the following data on exacerbations was extracted from the 26 published studies. The total number of exacerbations experienced in each treatment arm The total number of patient-years at risk, as reported in the article. If this was unavailable, total patient-years at risk were calculated by the intention to treat principle by multiplying the number of participants by the planned study duration. Where reported data on exacerbations were extracted as moderate and severe. 117

118 A sensitivity analysis was conducted to confirm findings using the binomial endpoint of a patient having had at least one exacerbation event during the course of the trial. For this analysis, data were extracted from the 26 published studies used in the primary analysis and in 3 additional trials as follows: The total number of patients in each arm having experienced at least one exacerbation event over the course of the trial, regardless of how many additional exacerbations that may have been experienced Please provide a clear description of the indirect/mixed treatment comparison methodology. Supply any programming language in a separate appendix. The MTC and additive analyses were based on the rates of exacerbation in COPD, reported as rate ratios, and this was confirmed using the sensitivity analysis of the binomial secondary outcome, which was reported as relative risks. Mixed Treatment Comparison Meta-analysis The MTC meta-analysis was used to compare the efficacy of different treatment regimens, primarily with regard to rates of exacerbations, and focused on the 10 treatments or treatment combinations included in the 26 identified studies (See Figure 12). The expected rates of exacerbation per person-years were modelled to allow expression of the relative effects of pairs of treatments in terms of rate ratios of exacerbations in COPD. This reflects the fact that individual patients could experience multiple exacerbations in COPD over the duration of follow up and that different trials had different durations of follow up. The primary analysis was a standard random-effects Poisson regression, that did not include any covariate. The analysis assumed that the study-specific relative effects were different yet similar enough to combine from a common population, and that the potential heterogeneity in study-specific relative treatment effects was consistent across pairwise treatment comparisons. The following sensitivity analyses were also examined: 118

119 1. A fixed-effects analysis 2. A random-effects model assessing the binomial event rates using relative risk as the pooled effect size 3. Assumptions on covariates examining severity of COPD exacerbations, baseline FEV 1, year of publication, and various Bayesian priors. 4. The analysis was re-run including data from a recently published trial of roflumilast vs placebo (study M2-111). In this sensitivity analysis data from Rennard et al, was used, which includes pooled data from roflumilast trials M2-111 and M2-112, therefore data from Calverley et al, (M2-112) was excluded. In addition to the assumptions made for the primary random-effects model, the sensitivity analysis models also assumed that potential heterogeneity in studyspecific relative treatment effects could not be explained by chance alone and investigated to what extent a study-specific covariate would help explain the excess between-study variation. All of the models assumed that the effect of the covariate of interest on the relative effects of pairs of treatments was common across all pair-wise treatment comparisons. All the models took into account the correlation structure induced by the multi-arm trials, except for the random-effects logistic regression model used in the sensitivity analysis relying on binomial event rates. For each MTC model, the estimated rate ratios of exacerbations per patient-years, with corresponding 95% confidence intervals, were produced for each pair-wise treatment comparison. Estimates of the absolute effect of each treatment expressed as mean exacerbations per patient-years were also produced, as well as estimated probabilities that each treatment is best (in the sense of being associated with the lowest rate of exacerbations in COPD per patient-years). The goodness of fit of each of the models to the data was measured by calculating the residual deviance and comparing it against the number of unconstrained data points, where the number of unconstrained data points was obtained by summing up the number of study arms across all studies included in the analyses. Under the null hypothesis that the model provides an adequate fit to the data, the residual deviance is expected to have a mean equal to the number of unconstrained data points. The fits of the models were compared using the Deviance Information Criterion (DIC). A model with its DIC being at least 3 points lower than a second model is considered to have a better fit

120 All models were fitted via a Bayesian Markov Chain Monte Carlo (MCMC) method, as implemented in the software WinBUGS Version 1.4. For each model, noninformative normal priors were used for all model parameters except for the betweenstudy standard deviation, for which a non-informative uniform prior (range 0-10) was used. Two MCMC chains were run for 100,000 iterations with a thin parameter of 10 after a burn-in of 20,000 in order to ensure convergence of the MCMC sampler. Posterior inference was conducted after discarding the burn-in iterations, thereby relying on 20,000 samples. The results of the model fits are also presented in Appendix 17. The MTC rate outcomes were compared to standard pairwise random-effects meta-analysis for consistency. Additive MTC Analysis To address the issue of the lack of information on a number of clinically important combinations including roflumilast (LABA + ICS + roflumilast; LABA + LAMA + ICS + roflumilast; LABA + LAMA + roflumilast) an additive MTC analysis was conducted. This analysis used an additive main effects model, in which each of the single treatments (placebo, LABA, LAMA, ICS and roflumilast) were assigned a mean treatment effect. The effect of each combination of 2 or more agents was expressed as the sum of the relevant active components. These combined treatment effects subsequently informed a MTC meta-analysis. This additive MTC analysis is based on multiplicative methodology described previously in analyses investigating the efficacy of combination therapies in reducing cardiovascular disease 69, and optimal combinations of blood pressure lowering medications 70. This additive methodology has been used previously to estimate the combined effect of psychological interventions in coronary heart disease 71. The additive analysis combines treatment effects by addition on a log scale, rather that by multiplication, which allows a more precise estimation of combined confidence intervals. The applicability of using an additive or multiplicative approach is reliant on an underlying assumption that the benefits of the individual treatments are independent of one another. i.e. adding one treatment to another provides additional benefit for the outcome being analysed. 120

121 Each class of agents included in the current analysis acts in a distinct way, therefore it was believed reasonable to assume that this methodological approach is applicable. The evidence for this for LABA, LAMA, ICS and roflumilast is provided within the studies that form the MTC. The analysis does not take into account any potential drug-drug interactions and therefore could overestimate treatment effects. In the current analysis the addition of one bronchodilator to another, or the addition of one anti-inflammatory drug to another, could be expected to provide less additional benefit than the addition of a bronchodilator to an anti-inflammatory agent. Individual trial results and findings from the MTC suggest that this applies to a degree for the addition of one bronchodilator to another but not for the other scenarios. The overall consistency of results between direct trial data, the MTC and the additive analysis support the use of the additive approach to provide the effects of the missing combinations. The WinBUGS code for the MTC and additive analyses are provided in Appendices 19 and 20 respectively Please present the results of the analysis. Overall, the MTC and additive analyses demonstrated that COPD treatments are consistently effective in reducing exacerbation rates. The analyses also indicate that adding one treatment to one or more of the other treatments in combination is likely to provide additional benefit. Estimated Rate Ratios for Pairs of Treatments The following tables present the estimated rate ratios and 95% confidence intervals for the relative effects of pairs of treatments produced by the random effects MTC model without covariates, and the additive MTC analysis. Table B 30: Estimated Rate Ratios and 95% Confidence Intervals for the Relative Effects of Pairs of Treatments, Produced by the Random-Effects MTC Model Without Covariates Treatment vs. Comparator Random-Effects MTC Model Rate Ratio 95% CI Roflumilast vs. Placebo 0.85 (0.72, 0.97) 121

122 LABA vs. Placebo 0.84 (0.75, 0.93) LAMA vs. Placebo 0.74 (0.66, 0.81) ICS vs. Placebo 0.80 (0.71, 0.90) Roflumilast + LABA vs. Placebo 0.67 (0.48, 0.91) Roflumilast + LAMA vs. Placebo 0.62 (0.44, 0.85) LABA + LAMA vs. Placebo 0.80 (0.56, 1.12) ICS + LABA vs. Placebo 0.69 (0.61, 0.77) ICS + LAMA + LABA vs. Placebo 0.68 (0.47, 0.95) LABA vs. Roflumilast 0.98 (0.80, 1.19) LAMA vs. Roflumilast 0.87 (0.71, 1.05) ICS vs. Roflumilast 0.94 (0.76, 1.15) Roflumilast + LABA vs. Roflumilast 0.79 (0.54, 1.12) Roflumilast + LAMA vs. Roflumilast 0.73 (0.50, 1.04) LABA + LAMA vs. Roflumilast 0.94 (0.63, 1.36) ICS + LABA vs. Roflumilast 0.81 (0.66, 0.99) ICS + LAMA + LABA vs. Roflumilast 0.80 (0.53, 1.16) LAMA vs. LABA (0.77, 1.01) ICS vs. LABA 0.96 (0.84, 1.09) Roflumilast + LABA vs. LABA 0.80 (0.59, 1.08) Roflumilast + LAMA vs. LABA 0.75 (0.52, 1.03) LABA + LAMA vs. LABA 0.96 (0.66, 1.35) ICS + LABA vs. LABA 0.82 (0.74, 0.92) ICS + LAMA + LABA vs. LABA 0.82 (0.56, 1.15) ICS vs. LAMA 1.09 (0.93, 1.26) Roflumilast + LABA vs. LAMA 0.91 (0.64, 1.25) Roflumilast + LAMA vs. LAMA 0.84 (0.61, 1.14) LABA + LAMA vs. LAMA 1.09 (0.77, 1.49) ICS + LABA vs. LAMA 0.94 (0.81, 1.07) ICS + LAMA + LABA vs. LAMA 0.92 (0.66, 1.28) Roflumilast + LABA vs. ICS 0.84 (0.60, 1.15) Roflumilast + LAMA vs. ICS 0.78 (0.54, 1.08) LABA + LAMA vs. ICS 1.01 (0.69, 1.42) ICS + LABA vs. ICS 0.86 (0.76, 0.98) ICS + LAMA + LABA vs. ICS 0.86 (0.58, 1.21) Roflumilast + LAMA vs. Roflumilast + LABA 0.95 (0.58, 1.46) LABA + LAMA vs. Roflumilast + LABA 1.23 (0.74, 1.90) ICS + LABA vs. Roflumilast + LABA 1.05 (0.75, 1.43) LAMA + ICS + LABA vs. Roflumilast + LABA 1.04 (0.63, 1.63) 122

123 LABA + LAMA vs. Roflumilast + LAMA 1.32 (0.82, 2.03) ICS + LABA vs. Roflumilast + LAMA 1.14 (0.80, 1.58) ICS + LAMA + LABA vs. Roflumilast + LAMA 1.12 (0.70, 1.72) ICS + LABA vs. LABA + LAMA 0.88 (0.61, 1.24) ICS + LAMA + LABA vs. LABA + LAMA 0.86 (0.61, 1.18) ICS + LAMA + LABA vs. ICS + LABA 0.99 (0.68, 1.40) A rate ratio smaller than 1 indicates that the treatment is associated with a reduction in the rate of exacerbations in COPD relative to the comparator. This reduction is statistically significant at the 5% level only if the upper end of the associated 95% confidence interval is less than 1. Table B 31: Outcomes from the Additive MTC Analysis Treatment vs. Comparator Additive Main Effects Model Rate Ratio 95% CI LABA + ICS + Roflumilast vs LABA + ICS 0.84 (0.74, 0.95) LABA + ICS + LAMA vs LABA + ICS + Roflumilast 0.89 (0.75, 1.03) LABA + ICS + LAMA + Roflumilast vs LABA + ICS + LAMA 0.84 (0.74, 0.95) LABA + LAMA + Roflumilast vs LABA + LAMA 0.84 (0.74, 0.95) LAMA + LABA + Roflumilast vs LAMA + LABA + ICS 1.00 (0.85, 1.13) A rate ratio smaller than 1 indicates that the treatment is associated with a reduction in the rate of exacerbations in COPD relative to the comparator. This reduction is statistically significant at the 5% level only if the upper end of the associated 95% confidence interval is less than 1. Key findings from the MTC and additive analysis Addition of roflumilast vs. placebo: The results of the primary analysis suggest that adding roflumilast to other therapies, compared to adding nothing, provides benefit in terms of reducing exacerbations. For example: LAMA + Roflumilast vs LAMA-alone; RR = 0.84 (0.61 to 1.14) LAMA + LABA + Roflumilast vs LAMA + LABA; RR = 0.84 (0.74 to 0.95) LAMA + LABA/ICS + Roflumilast vs LAMA + LABA/ICS; RR = 0.84 (0.74 to 0.95) Addition of roflumilast to a bronchodilator vs. an additional bronchodilator: The primary analysis also indicates that adding roflumilast to LABA or LAMA, compared to using the two bronchodilators in combination, reduces the exacerbation rate. For example: LABA + LAMA vs LABA + Roflumilast; RR = 1.23 (0.74 to 1.90) 123

124 LAMA + LABA vs LAMA + Roflumilast; RR = 1.32 (0.82 to 2.03) Limited benefit provided by addition of a bronchodilator to an existing bronchodilator: The MTC analysis suggests that the combination of LABA plus LAMA is not as effective in reducing the rate of exacerbations as may be expected, given the efficacy of the agents individually. For example, the addition of LAMA to LABA (vs. LABA alone) is associated with a RR of 0.96 (0.66, 1.35), whilst the addition of LABA to LAMA (vs. LAMA-alone) has an RR of 1.09 (0.77 to 1.49). A discrepancy is noted between results of the MTC and additive analyses with regard to the efficacy of LABA and LAMA in combination (See absolute treatment effects, Tables B32 and B33). The lack of clinical data to support this treatment combination, and thus to inform the treatment comparisons including this combination, may explain the lack of additional efficacy of LABA plus LAMA observed in the MTC analysis. The additive MTC analysis allows an estimate of the efficacy of LABA plus LAMA to be calculated based on the effects of the individual agents, thus suggesting an efficacy benefit of the two bronchodilators combined. There is, however, an argument that the use of two bronchodilators may not be as efficacious as anticipated, in terms of reducing exacerbations. Two randomised clinical trials have found no significant efficacy improvement, in terms of COPD exacerbations, with LABA plus LAMA combination therapy, compared to LABA 55;67 or LAMA 67 alone. Of note, these trials also found no significant difference in FEV 1 with LABA plus LAMA combination therapy, compared to LAMA monotherapy 55;67, but a benefit in FEV 1 was found with the combination compared to LABA alone 67. A possible explanation for this observation is that LABA and LAMA are both bronchodilators, therefore target a similar physiological outcome, which may reduce the additive effect when they are used in combination. For optimal treatment of severe COPD, a bronchodilator is required to achieve maximal improvement in lung function, in combination with an anti-inflammatory agent to manage other COPD parameters, such as reducing exacerbations. 124

125 From the results of this primary analysis it can be seen that for all treatment comparisons based on direct trial evidence a significant difference is observed, however the confidence intervals widen as soon as the comparisons rely on indirect evidence. This may be due to the power of the analysis, as a number of the comparisons did not have as much data as would ideally be required to show statistical differences, therefore it may be that significant effects may not have been identified where they do exist. Despite the lack of significant treatment difference observed in this analysis, a clinically meaningful difference in efficacy is known to exist for several of the comparisons. For example, LABA/ICS plus LAMA was not found to be significantly different to LABA/ICS alone, however the addition of LAMA in patients who remain symptomatic despite taking LABA/ICS is clinically recognised as an effective treatment option 1. Absolute Treatment Effects Table B32 presents the absolute treatment effects of 10 treatments, derived from the primary MTC analysis. The absolute treatment effects are expressed as mean exacerbations per patient-years. The findings observed here, although not significant, favour intensive current therapy (absolute treatment effect of LAMA + LABA/ICS is 0.82 exacerbations per patient-year [95% CI 0.57 to 1.15]) and the addition of roflumilast (absolute treatment effect of roflumilast + LAMA is 0.75 exacerbations per patient-year [0.53 to 1.02]; roflumilast + LABA = 0.81 [0.58 to 1.10]). Table B 32: Absolute Treatment Effects Corresponding to the 10 Treatments for the Management of COPD, as Derived from the Primary MTC Analysis Treatment Absolute treatment effect 95% CI Placebo 1.21 (1.17, 1.24) Roflumilast 1.03 (0.87, 1.21) LABA 1.01 (0.90, 1.11) LAMA 0.89 (0.80, 0.98) ICS 0.96 (0.85, 1.08) Roflumilast + LABA 0.81 (0.58, 1.10) Roflumilast + LAMA 0.75 (0.53, 1.02) LABA + LAMA 0.97 (0.67, 1.34) ICS + LABA 0.83 (0.73, 0.93) 125

126 ICS + LAMA + LABA 0.82 (0.57, 1.15) Table B33 shows the absolute treatment effects of the 10 interventions included in the published trials and the 5 additional treatment combinations, derived from the additive MTC analysis. These findings favour both current intensive therapy (absolute treatment effect of LAMA + LABA/ICS is exacerbations per patient-year [95% CI]) and the addition of roflumilast to this triple therapy (absolute treatment effect of LAMA + LABA/ICS + roflumilast is exacerbations per patient-year [95% CI]). Table B 33: Absolute Treatment Effects Corresponding to the 15 Treatments for the Management of COPD, as Derived from the Additive MTC Analysis Treatment Absolute treatment effect 95% CI Placebo 1.20 (0.17, 1.23) Roflumilast 1.01 (0.89, 1.14) LABA 1.04 (0.96, 1.12) LAMA 0.89 (0.81, 0.98) ICS 0.98 (0.91, 1.06) Roflumilast + LABA 0.87 (0.75, 1.01) Roflumilast + LAMA 0.75 (0.64, 0.87) Roflumilast + ICS 0.82 (0.71, 0.95) LABA + LAMA 0.77 (0.67, 0.87) LABA + ICS 0.85 (0.77, 0.94) LAMA + ICS 0.73 (0.64, 0.82) Roflumilast + LABA + LAMA 0.65 (0.54, 0.77) Roflumilast + LABA + ICS 0.71 (0.61, 0.83) LAMA + LABA + ICS 0.63 (0.54, 0.73) Roflumilast + LABA + LAMA + ICS 0.53 (0.43, 0.64) Optimal Treatment Combinations The results from the MTC and additive analyses were also used to determine the probability that each treatment is the most effective in terms of reducing the rate of exacerbations in patients with COPD. The findings show that, when considering only the 10 treatment combinations included in the MTC analysis, roflumilast + LAMA was associated with the highest probability of being the most effective treatment. When 126

127 considering all possible treatment combinations in the additive analysis, quadruple therapy with LAMA, LABA/ICS and roflumilast had the highest probability of being most effective for reducing COPD exacerbations. (See Appendix 17 for details). Sensitivity analyses A fixed effects analysis was also conducted and the results are shown in Appendix 17, along with other sensitivity analyses. The findings from the fixed effects model were consistent with those from the random effects analysis, with less conservative 95% CIs. The residual deviance of the fixed effects model was considerably higher (189) than the random effects model. The results of the senstivity analysis examining the binomial event of ever having had an exacerbation are also provided in Appendix 17. The results obtained when using binomial outcomes were consistent with those found when using rates of exacerbations. This strengthens the inference about the credibility of the analysis and the relative effects of the interventions. When probability of each treatment combination being the most effective best was assessed using the binomial outcome, this analysis also found roflumilast + LAMA as having the highest probability of being the best. The results of the other sensitivity analyses found that the choice of prior for the between-study standards deviation, disease severity and publication year did not influence the outcome of the analysis. Change in FEV 1 An MTC analysis using the endpoint of change in FEV 1 from baseline has been considered. Few studies were found which reported change in FEV 1 and among these the data reported was commonly of poor quality. In addition, the reported measure of variance differed between studies, and in some cases no measure of variance was reported. The length of the studies reporting FEV 1 also varies, which influences the absolute values of change in FEV 1 from baseline. Various analytical approaches have been explored with none considered robust enough to justify proceeding with exploratory work. A review of these analytical approaches is ongoing with experts in this area. 127

128 5.7.7 Please provide the statistical assessment of heterogeneity undertaken. The degree of, and the reasons for, heterogeneity should be explored as fully as possible. Heterogeneity was investigated in all analyses using the I 2 statistic 72. At least 2 pooled trials are required to calculate I 2. It is noted, however, that most evaluations of heterogeneity are underpowered and in this particular analysis most comparisons were sparsely populated. A large degree of heterogeneity in the pairwise comparisons was interpreted as an I 2 value above 40%, as recommended by Higgins and Thompson 72. Heterogeneity was examined by removing trials of predominantly severe COPD and then calculating the residual heterogeneity. Given that most tests for heterogeneity have low power, a random-effects analysis was applied, which places greater weight on smaller studies. The random-effects analysis was applied for both the pairwise and MTC analysis. When conducting the MTC analysis, several assessments of the impact of co-variates on the overall analysis were performed. The goodness of fit of the model was examined by evaluating the deviance information criteria. In general, the model fit was very good If there is doubt about the relevance of a particular trial, please present Not applicable. separate sensitivity analyses in which these trials are excluded Please discuss any heterogeneity between results of pairwise comparisons and inconsistencies between the direct and indirect evidence on the technologies. Heterogeneity was found in the pooled pairwise estimates examining all treatments versus placebo and also ICS plus LABA vs LABA alone. Whether trials including only severe patients reduced residual heterogeneity was examined and it was found that they did not do so importantly. The authors were unable to explain residual heterogeneity and thus applied randomeffects analysis throughout and corrected for underlying patient characteristics using a regression analysis. 128

129 5.8 Non-RCT evidence Non-RCT, both experimental and observational, evidence will be required, not just for those situations in which RCTs are unavailable, but also to supplement information from RCTs when they are available. This section should be read in conjunction with NICE s Guide to the methods of technology appraisal, sections to If non-rct evidence is considered (see section 5.2.7), please repeat the instructions specified in sections 5.1 to 5.5 for the identification, selection and methodology of the trials, and the presentation of results. For the quality assessments of non-rcts, use an appropriate and validated quality assessment instrument. Key aspects of quality to be considered can be found in Systematic reviews: CRD s guidance for undertaking reviews in health care ( Exact details of the search strategy used and a complete quality assessment for each trial should be provided in sections 9.6 and 9.7, appendices 6 and 7. Not applicable. 129

130 5.9 Adverse events This section should provide information on the adverse events experienced with the technology in relation to the decision problem. Evidence from comparative RCTs and regulatory summaries is preferred; however, findings from non-comparative trials may sometimes be relevant. For example, postmarketing surveillance data may demonstrate that the technology shows a relative lack of adverse events commonly associated with the comparator, or the occurrence of adverse events is not significantly associated with other treatments If any of the main trials are designed primarily to assess safety outcomes (for example, they are powered to detect significant differences between treatments with respect to the incidence of an adverse event), please repeat the instructions specified in sections 5.1 to 5.5 for the identification, selection, methodology and quality of the trials, and the presentation of results. Examples for search strategies for specific adverse effects and/or generic adverse-effect terms and key aspects of quality criteria for adverse-effects data can found in Systematic reviews: CRD s guidance for undertaking reviews in health care ( Exact details of the search strategy used and a complete quality assessment for each trial should be provided in sections 9.8 and 9.9, appendices 8 and 9. As described in Table B7 (Section 5.3.5), four trials in the roflumilast clinical trial programme included a safety outcome within their primary efficacy assessments; the rate of moderate or severe exacerbations (M2-111, M2-112, M2-124 and M2-125). 18;34 These trials, along with the rate of exacerbation which was identified by them, were described in detail above. Other (non-exacerbation) safety results of these trials are described below Please provide details of all important adverse events for each intervention group. For each group, give the number with the adverse event, the number in the group and the percentage with the event. Then present the relative risk and risk difference and associated 95% confidence intervals for each adverse event. A suggested format is shown below. Please note that, as part of the conditions to the licence for roflumilast, all healthcare professionals expected to prescribe roflumilast are to be provided with an educational pack, including a SPC and patient information leaflet, educational material and patient cards

131 Pooled safety analysis A recently presented analysis pooled data from 14 randomised, double-blind, placebo-controlled, parallel-group studies performed with roflumilast 250 μg or 500 μg once daily in patients with moderate-to-very-severe COPD. 73 The pooled data comprised 12,054 patients (roflumilast 500 μg, 5,766 patients; roflumilast 250 μg, 797 patients; placebo, 5,491 patients). The incidence of AEs (62.8% and 67.2%, RR = 1.07; Table B34) and serious AEs (SAEs; 14.2% and 13.5%, RR = 0.95; Table B35) was similar with placebo and roflumilast 500 μg, respectively. Apart from decreased weight, the majority ( %) of the most frequently reported AEs lasted < 4 weeks with no difference in duration seen between groups 74. Table B 34: Adverse event frequency, organised by MedDRA system organ class and preferred term, as reported by Calverley et al 73 Roflumilast 250 µg, % of patients (n=797) Roflumilast 500 µg, % of patients (n=5,766) Placebo, % of patients (n=5,491) Relative risk, 250 µg vs. placebo** Relative risk, 500 µg vs. placebo** Patients with AEs 60.7% 67.2% 62.8% Patients with SAEs 7.2% 13.5% 14.2% Deaths 0.9% 1.5% 1.6% AEs with suggested causality 6.5% 17.4% 5.4% Discontinuation due to AE 8.9% 14.3% 9.2% Most common AEs ( 2%): Infections and infestations 23.6% 25.9% 27.5% Nasopharyngitis 6.3% 6.3% 6.3% URTI 4.0% 3.8% 4.3% Bronchitis 3.1% 3.1% 3.5% Influenza 2.0% 2.5% 2.4% Pneumonia 0.6% 1.8% 2.0% Respiratory, thoracic and mediastinal disorders 24.7% 25.6% 29.3% COPD (exacerbation)* 21.2% 19.8% 23.1% Dyspnoea 2.3% 1.5% 2.2% Gastrointestinal disorder 13.0% 22.0% 10.7% Diarrhoea 4.9% 10.1% 2.6% Nausea 2.3% 5.2% 1.4% Investigations 6.9% 14.1% 10.6% Weight decreased 0.8% 6.8% 1.8% Nervous system disorders 5.6% 10.7% 5.5%

132 Headache 3.5% 4.6% 2.0% Dizziness 1.1% 2.4% 1.2% Musculoskeletal and connective tissue disorders 7.8% 10.2% 8.1% Back pain 2.8% 3.1% 2.1% General disorders 3.5% 6.4% 5.8% Psychiatric disorders 2.8% 6.0% 3.0% Insomnia 1.4% 2.6% 0.9% Cardiac disorders 3.5% 5.7% 5.9% Metabolism and nutrition disorders 2.4% 5.4% 3.4% Decreased appetite 0.5% 2.2% 0.4% Injury, poisoning and procedural complications 3.0% 3.6% 4.0% Skin and subcutaneous tissue disorders 2.9% 3.6% 2.8% Vascual disorders 2.5% 3.4% 4.2% Hypertension 1.5% 1.6% 2.5% Neoplasms (benign, malignant and unspecified) 0.5% 2.0% 1.7% URTI = Upper respiratory tract infection; CI's and p values not reported * defined as moderate to severe exacerbations ** data on file 75 Table B 35: Serious adverse event frequency ( 0.3%), as reported by Calverley et al 73 Roflumilast 250 µg, % of patients (n=797) Roflumilast 500 µg, % of patients (n=5,766) Placebo, % of patients (n=5,491) Relative risk, 250 µg vs. placebo* Relative risk, 500 µg vs. placebo* All SAEs 7.2% 13.5% 14.2% Respiratory, thoracic and mediastinal disorders 2.5% 6.5% 7.9% COPD (exacerbation) 1.9% 5.8% 7.1% Dyspnoea 0.3% 0.1% 0.1% Infections and infestations 0.5% 2.0% 2.3% Pneumonia 0.3% 1.1% 1.1% Cardiac disorders 1.1% 1.8% 2.1% Atrial fibrilation 0.3% 0.4% 0.2% Myocardial infarction 0.0% 0.2% 0.4% Angina pectoris 0.3% 0.2% 0.2% Gastrointestinal disorder 0.6% 1.4% 0.9% Diarrhoea 0.3% 0.2% 0.0% Neoplasms (benign, malignant and unspecified) 0.5% 1.3% 1.1% Nervous system disorders 0.1% 0.6% 0.8% General disorders 0.4% 0.5% 0.7% Chest pain 0.3% 0.2% 0.3% Injury, poisoning and procedural complications 0.9% 0.5% 0.5% Vascual disorders 0.3% 0.4% 0.5%

133 Renal and urinary disorders 0.4% 0.4% 0.3% Musculoskeletal and connective tissue disorders 0.1% 0.3% 0.3% Hepatobiliary disorders 0.1% 0.26% 0.16% Investigations 0.3% 0.24% 0.16% Surgical and medical procedures 0.4% 0.19% 0.02% * data on file 75 Body weight: In the two 1-year trials in which body weight was assessed at each study visit, between-treatment weight loss for roflumilast 500 μg and placebo was 2.17 kg (p<0.0001); weight loss was most pronounced in the first 8 weeks, with little weight change from month 6 until study end. 18 Weight loss was greatest in the obese subgroup (body mass index [BMI] 30 kg/m 2 ), which also had the highest proportion of patients with weight decrease classified as clinically relevant (16.5% of patients, vs. 12.6% in those with BMI < 18.5 kg/m 2 ). Only 4 patients (0.3%) receiving roflumilast 500 μg and 1 patient (<0.1%) receiving placebo in the two one-year studies withdrew due to weight loss as an AE. Weight loss was largely reversible after treatment discontinuation. Bioimpedance measurements taken in M2-128 found that almost two-thirds of the total weight decrease of 2.1 kg could be attributed to a loss of fat mass. Gastrointestinal effects: Across the 14 trials, more patients receiving roflumilast 500 µg (22%) vs. placebo (10.7%) experienced a gastrointestinal AE (RR = 2.06). This was largely accounted for by a difference in diarrhoea (10.2% vs. 2.6%, RR = 3.90) and nausea (5.2% vs. 1.4%, RR = 3.58). Diarrhoea tended to be experienced within the first 4 weeks of treatment, and was mild to moderate in severity in patients receiving roflumilast. Diarrhoea was reported as an SAE in 0.17% and 0.02% of all patients 74. As with all other AEs not related to weight loss, GI AEs most frequently lasted < 4 weeks with no difference in duration seen between groups. Cardiovascular events: Across all 14 trials, the incidence of cardiac AEs was lower in the roflumilast 500 μg group than with placebo (5.7% vs 5.9%, RR = 0.95) as was the incidence of cardiac SAEs (1.8% vs 2.1%, RR = 0.84), cardiac AEs leading to death (0.4% vs 0.5%, data not presented), and cardiac AEs leading to study discontinuation (0.9% vs 1.0%, data not presented). 133

134 Psychiatric events: Psychiatric disorders were reported in 6% of roflumilast 500 μg patients and 3% of placebo patients (RR = 2.00). Insomnia was the most frequent psychiatric event (2.6% vs. 0.9%, RR = 2.82) but depression was also more common during roflumilast treatment (1.2% vs 0.8%, RR = 1.48). However, the incidence of psychiatric SAEs was similar in the roflumilast 500 μg and placebo groups (0.4% vs 0.3%). Completed suicide/suicidal ideation/attempt was infrequent but seen in three roflumilast-treated patients (two more completed suicides occurred 3 weeks after treatment) and one placebo-treated patient (rate: vs per 1000 patientyears of exposure; p=0.6327). Headache: There were a greater proportion of patients with roflumilast 500 µg vs. placebo reporting headache (4.6% vs. 2%) and dizziness (2.4% vs. 1.2%). Infections: The occurrence of infection was similar between all treatment groups in the pooled analysis, including the incidence of pneumonia. Elsewhere, the incidence of pneumonia was reported more frequently with ICS treatment. 76 This increased frequency suggests that pneumonia might relate to the local effects of ICS, rather than representing a general outcome of treatment with anti-inflammatory drugs in patients with COPD. M2-124 and M2-125 Adverse events occurring in at least 2.5% of patients in one of the treatment groups from studies M2-124 and M2-125 are presented in Table B36. A statistically significant difference with roflumilast vs. placebo was observed for weight loss, diarrhoea, nausea and headache (Table B36)

135 Table B 36: Adverse events occurring in at least 2.5% of patients across M2-124 and M2-125, as reported by Calverley et al 18 COPD Diarrhoea Weight loss Nasophary- M2-124 M2-125 Roflumilast, % of patients (n=769) Placebo, % of patients (n=755) 9% 11% 8% 3% 12% 3% ngitis 7% 7% URTI Headache Pneumonia Back pain Acute bronchitis Nausea Hypertension Insomnia Decreased appetite Influenza * data on file 77 2% 3% 3% 2% 2% 2% 4% 3% 5% 5% 5% 2% 3% 4% 2% 1% 3% 0% 4% 2% Roflumilast vs. placebo (difference, 95%CI) Relative risk* Roflumilast, % of patients (n=778) Placebo, % of patients (n=790) -1.76% ( ) % 15% 4.75% ( ) % 3% 8.78% ( ) % 3% 0.79% ( ) % 6% -0.70% ( ) % 5% 1.13% ( ) % 1% 0.22% ( ) % 2% 0.60% ( ) % 2% -0.75% ( ) % 3% 3.34% ( ) % 2% -1.11% ( ) % 3% 1.41% ( ) % 2% 2.47% ( ) % 1% 1.13% ( ) % 3% Roflumilast vs. placebo (difference, 95%CI) Relative risk* -4.26% ( ) % ( ) % ( ) % ( ) % ( ) % ( ) % ( ) % ( ) % ( ) % ( ) % ( ) % ( ) % ( ) % ( )

136 Mortality rates did not differ in the roflumilast and placebo groups in the M2 124 study (2.2% vs. 2.3%), or M2 125 study (3.2% vs. 3.2%). AEs in the pooled study population were reported in 67.2% of patients in the roflumilast group and 62.3% in the placebo group; SAEs were reported by 19.5% and 21.7% patients, respectively. The most frequent AEs leading to discontinuation were diarrhoea, nausea, and headache in the pooled analysis (data not shown). Discontinuations due to these AEs were most evident in the first 4-12 weeks of treatment; thereafter, no difference was noted between treatment groups in the occurrence of these AEs and the withdrawals associated with them. Body weight: After 1 year of treatment, the mean weight change was a reduction of 2 09 kg (SD 3 98) in the pooled roflumilast group and an increase of 0 08 kg (3 48) with placebo, a mean difference of 2.17 kg (p < ). The largest absolute weight loss with roflumilast occurred in obese patients (BMI > 30; kg vs kg in those with BMI 18). 78 The change in weight in the pooled roflumilast group occurred in the first 6 months, and attenuated thereafter; weight change from week 36 to study end was kg and 0.38 kg for roflumilast and placebo, respectively. 79 Patients in the roflumilast group reporting diarrhoea, nausea, vomiting, or headache had greater weight loss than did those not reporting these symptoms (2 60 kg [3 72] vs 2 02 kg [4 01]). 78 Five patients discontinued treatment due to weight loss (4 in the pooled roflumilast group and 1 in the placebo group). 79 Gastrointestinal effects: Diarrhoea was reported in 8% and 9% of patients receiving roflumilast in studies M2-124 and M2-125 respectively, and in 3% patients who received placebo in both M2-124 and M Vomiting was reported by 1% patients in the pooled roflumilast groups and <1% in the placebo groups. Cardiovascular events: There was no difference noted in the proportion of reported cardiovascular AEs in the roflumilast and placebo groups (7% and 8%, respectively). Atrial fibrillation was an infrequent complication reported by 1% patients in the pooled roflumilast group and <1% of those receiving placebo. Infections: The incidence of pneumonia or other pulmonary infections did not increase during treatment with roflumilast (data not shown). 136

137 M2-127 and M2-128 Adverse events occurring in at least 2% of patients in one of the treatment groups of M2-127 and M2-128 are presented in Table B

138 Table B 37: Adverse events occurring in at least 2% of patients across M2-127 and M2-128, as reported by Fabbri et al 19 COPD Weight loss M2-127 M2-128 Salmeterol + roflumilast, % of patients (n=466) Salmeterol + placebo, % of patients (n=467) 16% 24% 9% 1% Diarrhoea 8% 3% Nasopharyngitis 7% 7% Nausea Headache Back pain Bronchitis 5% 0% 3% 1% 3% 2% 2% 3% Tremor 2% 0% Decreased appetite 2% 0% Insomnia URTI Influenza Dyspnoea * data on file 80 2% 0% 2% 4% 2% 2% 0% 3% Salmeterol + roflumilast vs. salmeterol + placebo (difference, 95%CI) Relative risk* Tiotropium + roflumilast, % of patients (n=374) Tiotropium + placebo, % of patients (n=369) -7.89% ( ) % 18% 7.51% ( ) % 1% 4.73% ( ) % 1% -0.41% ( ) % 5% 5.15% ( ) % 1% 1.93% ( ) % 0% 0.86% ( ) % 1% -0.85% ( ) % 3% 1.72% ( ) % 1% 1.93% ( ) % 0% 1.93% ( ) % 0% -2.14% ( ) % 1% -0.42% ( ) % 0% -2.57% ( ) % 1% Tiotropium + roflumilast vs. tiotropium + placebo (difference, 95%CI) Relative risk* -2.65% ( ) % ( ) % ( ) % ( ) % ( ) % ( ) % ( ) % ( ) % ( ) % ( ) % ( ) % ( ) % ( ) % ( )

139 In M2-127, 63% patients assigned to salmeterol plus roflumilast reported AEs, compared to 59% of the patients assigned to receive salmeterol plus placebo. In M2-128, 46% of all patients in the tiotropium plus roflumilast group reported AEs, vs. 41% in the tiotropium plus placebo group. The proportion of patients with AEs adjudged by the investigator to be related to treatment was 18% with salmeterol and roflumilast, 3% with salmeterol and placebo, 12% with tiotropium and roflumilast, and 2% with tiotropium and placebo. Diarrhoea, nausea, and weight loss were the most common treatment related AEs, with no major difference between the two studies. Compared with placebo, roflumilast was associated with increased withdrawal from the study; this increase was significant in the salmeterol plus roflumilast trial (p=0 0019) but not in the tiotropium plus roflumilast trial (p=0 0864). Body weight: In both trials, similar gradual reductions were noted in mean body weight in the roflumilast groups during the 24 weeks of treatment (M2-127, -2 0 kg; M2-128, -1 8 kg), whereas there was little change in the placebo control groups (M2-127, +0 2 kg; M2-128, +0 3 kg). Weight loss was similar in the two trials and not significantly different between patients in different BMI categories, although patients with higher BMI experienced a larger absolute change in weight (M2-127: BMI > 30, kg vs. BMI 18, kg; M2-128: BMI > 30, kg vs. BMI 18, kg). In M2-127, weight loss associated with roflumilast was greater in patients with gastrointestinal AEs or headache, or both. 81 M2-111 Adverse events occurring in at least 2.5% of patients in either the roflumilast 500 µg or placebo arms from study M2-111 are presented in Table B38 below 82. Table B 38: Adverse events occurring in at least 2.5% of patients in M2-111 M2-111 Roflumilast 500 µg, % of patients (n=567) Placebo, % of patients (n=606) Relative risk Patients with AEs 86.24% 83.33% Patients with SAEs 22.22% 21.78% Deaths 1.94% 1.98%

140 AEs with suggested causality by investigator AEs with suggested causality by sponsor 26.46% 11.55% % 4.46% Discontinuation due to AE 19.58% 10.23% Most common AEs ( 2.5%) Infections and infestations 36.33% 42.24% Nasopharyngitis 6.53% 7.26% URTI 6.88% 9.90% Bronchitis 2.29% 3.47% Bronchitis acute 3.00% 3.47% Influenza 3.00% 3.30% Pneumonia 3.35% 3.47% Sinusitis 3.70% 4.62% Respiratory, thoracic and mediastinal disorders 46.38% 53.47% COPD 40.56% 47.69% Cough 2.47% 2.31% Dyspnoea Exacerbated 1.76% 2.97% Pharyngolaryngeal Pain 2.47% 0.50% Gastrointestinal disorder 30.86% 14.52% Abdominal Pain 2.65% 1.16% Diarrhoea 14.11% 3.30% Nausea 7.23% 1.65% Vomiting 3.00% 1.32% Investigations 30.34% 27.23% Electrocardiogram QT Corrected Interval Prolonged 7.76% 6.44% Heart Rate Increased 2.47% 2.97% Occult Blood Postive 4.23% 3.47% Weight decreased 12.35% 5.12% Weight Increased 0.35% 2.48% Nervous system disorders 13.93% 11.06% Headache 7.41% 3.63% Tremor 2.47% 0.17% Musculoskeletal and connective tissue disorders 17.64% 13.20% Back pain 5.64% 4.46% Muscle Spasms 2.65% 1.32% General disorders and 10.23% 10.23%

141 Administraion Site Conditions Chest Pain 2.29% 3.14% Oedema Peripheral 1.41% 2.64% Psychiatric disorders 9.52% 4.95% Depression 2.82% 1.32% Insomnia 3.53% 1.65% Metabolism and nutrition disorders 9.35% 5.94% Decreased appetite 4.06% 0.33% Vascual disorders 4.59% 5.78% Hypertension 2.47% 3.30% The incidence of AEs judged by the investigator to be likely or definitely related to the study medication was 26.5% in the roflumilast 500 µg group and 11.6% in the placebo group. Most patients with AEs experienced mild to moderate events in both treatment groups. AEs judged to be severe occurred in 24% of patients taking roflumilast 500 µg and 22.9% of patients taking placebo. The incidence of most of the AE terms was low and the great majority of individual AEs reported occurred with an incidence < 5%. Differences in the incidence of individual AEs between the two treatment groups of > 3% were observed for the following: weight decreased, diarrhoea, nausea, headache and decreased appetite were higher in the roflumilast 500 µg group; whereas COPD and upper respiratory tract infection occurred more frequently in the placebo group. M2-112 Adverse events occurring in at least 5% of patients in either the roflumilast 500 µg or placebo arms from study M2-112 are presented in Table B39. A statistically significant difference with roflumilast vs. placebo was observed for weight loss, diarrhoea, nausea and headache

142 Table B 39: Adverse events occurring in at least 5% of patients across M2-112, as reported by Calverley et al 34 M2-112 Roflumilast 500 µg, % of patients (n=760) Placebo, % of patients (n=753) Relative risk* Patients with 1 adverse event 77.9% 77.6% 1.00 COPD 44.6% 48.1% 0.93 Diarrhoea 9.3% 2.7% 3.52 Nasopharyngitis 7.0% 7.4% 0.94 Headache 6.2% 2.4% 2.59 Influenza 5.3% 4.5% 1.17 Nausea 5.0% 1.3% 3.77 URTI = Upper respiratory tract infection; CI's and p values not reported * data on file 83 The incidence of AEs judged by the investigator to be treatment-related was 17.8% with roflumilast and 5.6% with placebo. Among these events, diarrhoea, nausea, and headache were the most common AEs. Consistent with the pooled safety analysis and other trials, the authors reported that drug-related AEs were observed mainly in the first 4 weeks of treatment with roflumilast. Although most AEs lasted less than 4 weeks and resolved with continued treatment, they were considered to be the most likely explanation for the excess of AE related dropouts in the early phase of the study. The median duration of AEs was 11 and 12 days in the roflumilast and placebo groups, respectively. SAEs occurred in 18.0% of roflumilast- and 17.5% of placebo-treated patients (data not reported). COPD exacerbation was the most frequent SAE, followed by pneumonia, both having a similar incidence rate in each treatment group. More patients died during the study while receiving placebo than roflumilast (20 vs. 12 patients, respectively). The most frequent causes of death were respiratory disorders (1.1 vs. 0.5%, respectively), infections (0.9 vs. 0.5%, respectively), and cardiac disorders (1.1 vs. 0.1%, respectively). Physical examinations, routine laboratory tests, and electrocardiograms did not show any clinically significant changes due to roflumilast administration. 142

143 Subgroup Treatment Post-hoc Analysis of M2-111 and M2-112 Safety in Patients with Severe COPD and Associated Chronic Bronchitis, With or Without Concomitant ICS Therapy In this post-hoc, pooled analysis of studies M2-111 and M adverse events were found to be similar to those reported in other roflumilast clinical trials. As shown in Table B40, the analysis did not identify any group of subjects that appeared to be at higher risk of adverse events from roflumilast. Importantly, patients with chronic bronchitis, in whom a greater benefit of roflumilast on exacerbations was observed, did not experience an increase in adverse events with roflumilast. Furthermore, this subgroup of patients with chronic bronchitis reported numerically fewer adverse events, in both the roflumilast and placebo groups. Concomitant ICS therapy did not affect the adverse event profile of roflumilast. Table B 40: Adverse Events Reported in the Post-hoc Analysis of M2-111 and M2-112 by Rennard et al Post-hoc Pooled Analysis of M2-111 and M2-112 All Patients, % of patients Roflumilast 500 µg (n=1327) Placebo (n=1359) Roflumilast 500 µg (n=817) Chronic Bronchitis Patients (+/- emphysema), % of patients All With ICS Without ICS Placebo (n=847) Roflumilast 500 µg (n=492) Placebo (n=493) Roflumilast 500 µg (n=325) Placebo (n=354) Patients with AEs 81.5% 80.1% 78.6% 79.5% 81.7% 80.9% 73.8% 77.4% Patients with SAEs AEs related to study medication AEs leading to discontinuation of study medication Most common AEs ( 5% patients) COPD exacerbation 19.8% 19.4% 18.8% 17.9% 22.8% 22.1% 12.9% 12.1% 21.5% 8.3% 16.4% 7.9% 15.7% 7.1% 17.5% 9.0% 17.7% 10.0% 11.5% 6.6% 13.2% 8.1% 8.9% 4.5% 42.9% 48.0% 43.0% 48.5% 49.8% 54.4% 32.6% 40.4% Diarrhoea 12.1% 2.9% 7.1% 3.1% 8.3% 3.2% 5.2% 2.8% Nausea 6.0% 1.5% 4.4% 1.3% 4.7% 1.0% 4.0% 1.7%% Weight loss 7.5% 2.8% 6.1% 2.5% 5.3% 1.6% 7.4% 3.7% Nasopharyngitis 6.8% 7.4% 7.5% 7.7% 6.5% 7.3% 8.9% 8.2% Pneumonia 2.8% 4.0% 3.5% 4.1% 4.3% 5.7% 2.5% 2.0% Upper respiratory tract infection 5.4% 6.3% 5.4% 5.5% 4.5% 5.1% 6.8% 6.2% Headache 6.9% 3.0% 5.6% 2.1% 6.1% 2.2% 4.9% 2.0% Influenza 4.4% 4.0% 4.4% 3.7% 4.5% 2.2% 4.3% 5.6% 143

144 5.9.3 Give a brief overview of the safety of the technology in relation to the decision problem. Across a pooled analysis of 14 clinical trials, the use of roflumilast was generally well tolerated. The most commonly reported AEs with roflumilast 500 µg were diarrhoea (10.1%), weight loss (6.8%), nausea (5.2%) and headache (4.6%). The majority of reported side effects were mild to moderate in nature, and the GI symptoms mostly occurred within 4 weeks of the start of therapy and resolved with continued treatment with roflumilast. Weight loss was predominantly due to loss of body fat with the largest absolute weight loss seen in patients with a BMI > 30, and was reversible upon treatment discontinuation. The occurrence of infection was similar between all treatment groups in the pooled analysis, including the incidence of pneumonia. Clinical trials involving ICS suggest the incidence of pneumonia was reported more frequently with this form of treatment. 76 In turn, this increased frequency suggests that pneumonia might relate to the local effects of ICS, rather than representing a general outcome of treatment with anti-inflammatory drugs in patients with COPD. As described in Section 1.4, the conditions to the licence for roflumilast are that all healthcare professionals expected to prescribe roflumilast must be provided with an educational pack, including a SPC and patient information leaflet, educational material and patient cards. Further details of these materials can be found in the annex to the EPAR (Conditions imposed on member states for safe and effective use)

145 Roflumilast is a first-in-class treatment and a new way of treating COPD, so it is important that clinicians who prescribe roflumilast are as educated as possible about the most appropriate ways to use the treatment. This will help to ensure that it is prescribed to those COPD patients who can most benefit from roflumilast and will optimise patient outcomes. The clinical education pack contains the Summary of Product Characteristics and Patient Information Leaflet for roflumilast. It also includes other educational material for the clinician that supports the appropriate use of the treatment including which patients may not be suitable for roflumilast. There is also a patient card that helps patients to monitor their body weight. 145

146 5.10 Interpretation of clinical evidence Please provide a statement of principal findings from the clinical evidence highlighting the clinical benefit and harms from the technology. Roflumilast significantly reduces the rate of moderate or severe exacerbations and improves lung function in severe COPD patients (FEV 1 <50%) with chronic bronchitis and a history of frequent exacerbations. It has been demonstrated that these benefits occur independently of concomitant bronchodilator and ICS therapy. The MTC and multiplicative analysis demonstrate that COPD treatments are consistently effective in reducing exacerbation rates and indicate that adding treatments in combination is likely to provide additional benefit. The analysis also suggests that adding roflumilast to other therapies, provides benefit in terms of reducing exacerbations. Using roflumilast in place of either LABA or LAMA, as opposed to using two bronchodilators in combination, also appears to reduce exacerbation rates Roflumilast has an acceptable tolerability profile with the majority of reported adverse events being mild to moderate and transient in nature. An educational pack is also provided to all healthcare professionals expected to prescribe roflumilast, which includes a SPC and patient information leaflet, educational material and patient cards Please provide a summary of the strengths and limitations of the clinicalevidence base of the intervention. The clinical studies which support the efficacy of roflumilast were large, randomised, double-blind, placebo-controlled, multicentre trials with clinically recognised primary and secondary endpoints. The clinical trial programme for roflumilast initially included all patients with COPD. Based on the findings of early trials, where particular benefit was demonstrated within a specific subgroup of COPD patients, the programme was refined to focus on those with chronic bronchitis. Data from two large, multicentre RCTs (M2-124 and M2-125) provide evidence for the efficacy of roflumilast compared to placebo in the specific patient population for whom it is now licensed. These twin studies were of similar design which allowed a pooled analysis to be conducted, with a total patient population of In these studies, patients were stratified according to their use of LABA during the trials (50% of total trial population received LABA), thus allowing a subgroup analysis to be 146

147 conducted. This demonstrated that the benefits of roflumilast occur independently of LABA therapy. Two additional studies were conducted in which all patients received concomitant salmeterol (M2-127) or tiotropium (M2-128) and demonstrate the efficacy of roflumilast on lung function in addition to bronchodilator therapy. Although a reduction in the secondary endpoint of exacerbations was observed, the short duration of the studies (6 months) and size of study population reduced the likelihood of detecting a statistically significant difference. This is supported by the availability of the subgroup analysis of patients receiving concomitant LABA in M2-124 and M A post-hoc subgroup analysis of the pooled data from studies M2-111 and M2-112 assessed the efficacy of roflumilast in patients with severe COPD and chronic bronchitis, with or without concomitant ICS use. The analysis demonstrated that roflumilast effectively reduced the rate of moderate or severe exacerbations, and improved lung function and SGRQ total score, and that these effects were not blunted by concomitant ICS use. A limitation of the clinical trial programme for roflumilast is that no studies were conducted to evaluate the efficacy of roflumilast compared to, or in addition to LABA/ICS, LABA plus LAMA or a combination of all three. To overcome this an independently performed MTC was conducted, which provides information on the effects of roflumilast in combination with therapies that were not studied in clinical trials. The analyses allow the comparisons stated in the decision problem to be assessed. The decision to use an MTC was based on advice in the NICE Guide to Methods Manual (Section ). MTCs are becoming more frequently used within NICE appraisals, whether developed by an assessment group in an MTA or a manufacturer in an STA. As with the majority of MTCs, the main limitation of the analysis was the power of the analysis. As several treatment comparisons relied on only one trial; it may have been possible that significant effects were missed where they do exist Please provide a brief statement of the relevance of the evidence base to the decision problem. Include a discussion of the relevance of the outcomes assessed in clinical trials to the clinical benefits experienced by patients in practice. 147

148 The pivotal trials M2-124 and M2-125, and study M2-128 provide evidence for the efficacy of roflumilast in the specific patient population described in the decision problem (ie those with severe COPD, chronic bronchitis and a history of exacerbations) and demonstrate that roflumilast effectively improves lung function and reduces exacerbations in this patient group. The clinical trial programme did not investigate the use of roflumilast with, or compared to, all treatment combinations which are used in clinical practice. Therefore the trial evidence does not directly relate to the comparisons in the decision problem. However, the MTC provides evidence which relates to each of the treatment comparisons outlined. The outcomes assessed in the clinical trials are relevant to the benefits experienced by patients in clinical practice. FEV 1 is a widely recognised measure of lung function, and poor lung function is a key feature of COPD. Exacerbations have a considerable negative impact on patients' health, as well as being a burden to the NHS, and are associated with worsening prognosis, thus represent an important therapeutic target in COPD. NICE CG101 also stressed the importance of the multidimensional assessment of COPD, as opposed to simple categorisation of disease severity through measure of lung function (Section 6.9 of CG101; analysis not restricted to the subset of patients with chronic bronchitis) Identify any factors that may influence the external validity of study results to patients in routine clinical practice; for example, how the technology was used in the trial, issues relating to the conduct of the trial compared with clinical practice, or the choice of eligible patients. State any criteria that would be used in clinical practice to select patients for whom treatment would be suitable based on the evidence submitted. What proportion of the evidence base is for the dose(s) given in the SPC? Although some of the earlier trials (M2-111 and M2-112) included a broad population of patients, the pivotal trials included only patients with severe COPD, chronic bronchitis and a history of exacerbations, therefore accurately reflect the population in whom roflumilast will be used in clinical practice. No issues have been identified regarding how roflumilast was used or the conduct of the trial, apart from the limited clinical trial data comparing roflumilast to some treatment combinations which are used in clinical practice, as discussed previously. 148

149 Patients eligible for treatment with roflumilast are easily identifiable in clinical practice. NICE clinical guidelines recommend that the severity of all patients' COPD is assessed by spirometry, therefore severe COPD patients will be able to be identified. Patients with chronic bronchitis are also easily identifiable in practice, as those patients presenting with chronic cough and sputum production. In clinical trials the history of frequent exacerbations was based on patient recall and it is expected that this will be assessed in the same way in clinical practice. No additional tests are required to indentify patients suitable for roflumilast. All trials discussed in this submission assessed the 500 µg dose of roflumilast. 149

150 6 Cost effectiveness 6.1 Published cost-effectiveness evaluations Identification of studies Describe the strategies used to retrieve relevant cost-effectiveness studies from the published literature and from unpublished data held by the manufacturer or sponsor. The methods used should be justified with reference to the decision problem. Sufficient detail should be provided to enable the methods to be reproduced, and the rationale for any inclusion and exclusion criteria used should be provided. The search strategy used should be provided as in section 9.10, appendix 10. Economic analyses of pharmacological treatment of COPD specified in the decision problem were identified from the published literature by searching Medline, Medline In Process, Embase, EconLIT and NHSEED. Abstracts from conference proceedings (from the International Society for Pharmacoeconomics and Outcomes Research (ISPOR) and the European Respiratory Society (ERS)) were searched separately. The titles and abstracts (if available) of the identified articles were then assessed according to the inclusion/exclusion criteria described below and were eliminated if they were not relevant. Appendix 9.10 provides details of search strategies for cost-effectiveness. Due to the COPD population in the decision problem being adults with severe COPD that is associated with chronic bronchitis, search terms for COPD disease severity and chronic bronchitis were included in the search strategy (see Appendix). This was expected to identify studies where patients with severe COPD associated with chronic bronchitis were included within a broader COPD patient population. The inclusion criteria for studies were as follows: Studies must be an economic evaluation. The economic evaluation should relate directly to the management of COPD. 150

151 The economic evaluation should relate to pharmacological therapy with roflumilast or comparators included in the decision problem being addressed (LABA, LAMA, ICS). The exclusion criteria were: Economic evaluations related to non-pharmacological management of COPD e.g. studies of the cost-effectiveness of pulmonary rehabilitation. Patients with asthma or COPD mixed with asthma. Figure 13 shows a flow diagram of the search for cost-effectiveness studies related to the decision problem. Figure 13: Flow diagram of the search for cost-effectiveness studies Records from databases: Medline: n = 890 Embase: n = 910 EconLIT: n =23 NHSEED: n=141 N = 1964 Additional records identified from conference proceedings n = 309 Duplicates removed n = 315 Records screened based on title/abstract n = 1649 (databases) (conferences) = 1958 Studies excluded due to non-english language or being outside the scope of the decision problem n = 1915 Potentially relevant articles and abstracts assessed for eligibility n = 43 Articles meeting inclusion criteria n = 37* 151

152 Description of identified studies Provide a brief overview of each study, stating the aims, methods, results and relevance to decision-making in England and Wales. Each study s results should be interpreted in light of a critical appraisal of its methodology. When studies have been identified and not included, justification for this should be provided. If more than one study is identified, please present in a table as suggested below. The identified studies are presented in Table B 41 below. Eight are abstracts were identified in the search, and are summarised in Table B 42 below. Four of the identified articles are reviews and were not included in this table. One study demonstrates how the data synthesis technique used can affect resulting model parameters, which may in turn impact estimates of cost effectiveness and the uncertainty around them (Oppe et.al., 2011)

153 Table B 41: Overview of Identified Studies Study Pub. year Country Summary of model Patient population (average age in years) QALYs (intervention, comparator) Costs (currency) (intervention, comparator) ICER (per QALY gained) Ayres JG, Price MJ, Efthimiou J Multi country RCT analysis UK NHS Costeffectiveness perspective Bootstrap replications of cost and effect data from the ITT population of the sixmonth clinical trial on which the analysis was based Placebo: 64 Fluticasone Propionate: 62 Symptomatic patients with moderate to severe COPD Effectiveness was measured using FEV1 improvement and occurrence of exacerbations Direct costs per patient per day UK ( ) 1998 Placebo: 2.72 Fluticasone Propionate: 2.75 UK ( ) 1998 Direct costs ICER Cost per day to achieve; Patient with >=10% FEV1 improvement: 0.25 Exacerbation free patient: 0.54 Moderate/severe exacerbation free patient: 0.25 Briggs AH, Glick HA, Lozano- Ortega G, et al Multi country RCT (42 countires) Multivariate 64.7 regression models on Recruited patients had a patient-level data diagnosis of COPD with a where Eq-5D data FEV1 of <60% of the were collected (i.e. predicted value, a smoking not Asia) value of >10 pack-years and were aged between years QALYs for three year trial length Placebo: Salmeterol: Fluticasone Propionate: Salm/FP combination: All in USD ($) 2007 three year trial length Placebo: 9456 Salmeterol: Fluticasone Propionate: Salm/FP combination: USD ($) Salm/FP vs Placebo 43,600 per QALY Combination vs Salm 26,500 per QALY Briggs AH, Lozano- Ortega G, Spencer S et al UK perspective, analysis of RCT Analysis of PLD from ISOLDE trial Patients with moderate to severe COPD. Age not stated Fluticasone propionate: 1.88 Placebo: 1.74 UK Pounds ( ) Fluticasone propionate: 2,530 Placebo: 1,509 UK ( ) Fluticasone propionate vs placebo 9,500 per QALY Chuck A, Jacobs P, Mayers I, Marciniuk D Canadian cost data. Markov model with 3- month cycles. Assumed FEV decline. Exacerbation rate based on 4 clinical trials COPD patients (93% stage 1) representative of those who had COPD in a US population-based sample survey: 79% male, 87% white, mean (± SD) age of 61±7.7 years. 97% smoking history. CT = Combination therapy (LABA + ICS) LABA only at all stages: CT for stage 3 only: CT for stages 2 and 3: CT for all stages: Canadian Dollars 2006 CT = Combination therapy (LABA + ICS) LABA only at all stages: 9,636 CT for stage 3 only: 9,788 CT for stages 2 and 3: 10,142 CT for all stages: 13,282 Life time model $/QALY ICER CT for stage 3 only: $25,333 CT for stages 2 and 3: $50,571 CT for all stages: $448,571 Dal Negro et al Italy Markov Model with 5 health states representing 4 COPD stages and death. 1- year cycle length Range used 20 -over 65 QALYs not used. Exacerbations per patient Salmeterol/fluticasone vs salmeterol: 0.98 Formoterol/budesonide vs Salmeterol: 0.41 Direct costs per patient (Euro) Salmeterol/fluticasone vs salmeterol: Formoterol/budesonide vs Salmeterol: Not reported Cost per exacerbation used 153

154 Study Pub. year Country Summary of model Patient population (average age in years) QALYs (intervention, comparator) Costs (currency) (intervention, comparator) ICER (per QALY gained) Dalal et al Summary of Two RCTs Analysis of pooled RCT data 65.1 Not reported Mean annual exacerbation rates Moderate/severe exacerbation Fluticasone propionate/salmeterol: 1.1 Salmeterol: 1.58 Total annualized cost USD ($) MEAN Fluticasone propionate/salmeterol: 4291 Salmeterol: 4596 QALYs Not reported Earnshaw et 2009 US Markov model with 3 al. 91 COPD health states and death Placebo: 6.74 Fluticasone propionate/salmeterol: 7.42 Salmeterol: 7.24 Fluticasone propionate: 6.58 Total costs USD ($) Placebo: Fluticasone propionate/salmeterol: Salmeterol: Fluticasone propionate: Incremental cost per QALY gained USD ($) vs Placebo Fluticasone propionate/salmeterol: Salmeterol: Fluticasone propionate: Dominated Hogan et al Not stated EU and US data used Analysis of RCT data (12 week) Not stated QALYs not used. QoL measured using the St Georges respiratory questionnaire (SGRQ) Total costs per patient (12 weeks) USD ($) Placebo: Ipratropium bromide: Formoterol 12 ug: Formoterol 24 ug: $/FEV1 vs placebo Ipratropium bromide: Formoterol 12 ug: Formoterol 24 ug: Dominated Jones et al Analysis of Multi country RCT. UK payer perspective Analysis of RCT data Placebo: 62 Salmeterol: 63 QALYs not used FEV1 recorded and QoL using the SGRQ Total direct medical costs (16 weeks) Mean UK ( ) per patient Placebo: Salmeterol: CE ratios were estimated per arm instead of in ICER format Lee et al Singapore Unspecified model type Not stated QALYs not used Hospitalisation and exacerbation rates used from 3 different studies for tiotropium bromide vs placebo Singapore ($) based on data from the ministry of health Incremental cost per year of tiotropium bromide treatment per patient vs placebo Estimates based on 3 different studies No ICER stated 154

155 Study Pub. year Country Summary of model Patient population (average age in years) QALYs (intervention, comparator) Costs (currency) (intervention, comparator) ICER (per QALY gained) Lofdahl et al Swedish Payer Analysis of RCT data 64 QALYs not used Ratio used for comparisons of: Budesonide, formoterol, placebo and Budesonide/formoterol Outcomes: Exacerbations, FEV1, SGRQ Annual health care costs (Euro, 2001) Budesonide: 3194 Formoterol: 3653 Placebo: 3213 Budesonide/formoterol: ICER not reported. Cost comparison with CEAC Maniadakis et 2006 Greek national al. 96 health service Markov model with 1- year time horizon. First 8 days of treatment followed by monthly cycles. Performed microsimulations Not Stated Total exacerbations over 12 months (mean estimates) Tiotropium: 0.85 Salmeterol: 1.02 Tiotropium vs. salmeterol Difference in QALYs = Euros Total costs Tiotropium: 2504 Salmeterol: 2655 Tiotropium arm is dominant Naik et al Third party payer perspective only using direct costs Markov model with 1- year time horizon. 6- month cycle length. Two health states ( on treatment and maintenance therapy ) no death 65 Exacerbations avoided per patient per year. Placebo: Salmeterol: Tiotropium: Mean direct costs of treatment US ($ ) 1 Year model duration Placebo: Salmeterol: Tiotropium: ICER ($ ) Cost per exacerbation avoided per patient per year versus placebo Salmeterol: Tiotropium: Najafzadeh et al Canadian health care system perspective Analysis of 52 week RCT data Not stated QALYs over 52 weeks Tiotropium + Fluticasone/ Salmeterol: Tiotropium + Placebo: Tiotropium + Salmeterol: Costs are in 2006 Canadian dollars Tiotropium + Fluticasone/ Salmeterol: 4042 Tiotropium + Placebo: 2678 Tiotropium + Salmeterol: 2801 Incremental cost per QALY vs Tiotropium + Placebo Tiotropium + Salmeterol: Dominated Tiotropium + Fluticasone/ Salmeterol: Neyt Belgium health care payer Decision tree with one-year data. Mortality not captured. Based on observational data Mean incremental QALYs of COPD treatment for Tiotropium use versus non- Tiotropium use (Euro) 1 year per patient Mean incremental costs of COPD treatment for Tiotropium use versus non- Tiotropium use (Euro) 1 year 373 per patient ICER Tiotropium use versus non- Tiotropium use (Euro) 1 year 1,244, 023 Euro per QALY Oba US third party payer Markov model with 3 COPD states and death. 3-year time horizon, 3-month cycles. Efficacy data from TORCH study 65 Total QALYs gained (3 years) Placebo: Salmeterol propionate: Fluticasone: Total treatment costs (2006 US $) Placebo: 2780 Salmeterol propionate: 5832 Fluticasone: 6927 salmeterol fluticasone propionate: salmeterol fluticasone propionate: ICER $/QALY vs Placebo Salmeterol propionate: 56,345 Fluticasone: 87,600 salmeterol fluticasone propionate: 41,

156 Study Pub. year Country Summary of model Patient population (average age in years) QALYs (intervention, comparator) Costs (currency) (intervention, comparator) ICER (per QALY gained) Oba US third party payer perspective Retrospective analysis of pooled systematic review data Mean age per group Tiotropium: 65.7 Tiotropium Placebo: 66.2 Salmeterol: 65.7 Salmeterol Placebo: 66.2 Mean incremental QALYs per patient year Tiotropium: Salmeterol: Total costs (2005 US $) Tiotropium: 835 Salmeterol: 1066 ICER $/QALY (2005 US $) vs placebo Tiotropium: 26,094 Salmeterol: 41,000 Onukwugha Perspective of the Veterans Affairs Health Care System, USA Retrospective analysis of electronic medical records. One-year analysis. Applied data from trials for tiotropium treatment effects 68.9 Mean emergency room visits per year per patient Ipratropium: 0.3 Tiotropium: 0.24 Total costs not stated ICER for the sum of avoided Veterans Affair hospitalizations and emergency room visits due to a COPD exacerbation per year vs Tiotropium US ($) 2006 Ipratropium: 2360 Oostenbrink The Netherlands and Canada Markov model with three COPD health states (moderate, severe and very severe). No death state. 1-year time horizon. 8-day followed by 1-month cycle length. Not stated Quality adjusted life months, 1 year time horizon Tiotropium: 8.42 Salmeterol: 8.17 Ipratropium: 8.11 Total treatment costs Euros (2001) The Netherlands Tiotropium: 1760 Salmeterol: 1802 Ipratropium: 1503 Canada Tiotropium: 1309 Salmeterol: 1306 Ipratropium: 1307 Euros (2001) Tiotropium vs. salmeterol difference The Netherlands Costs: -42 Exacerbations avoided: 0.17 Quality-adjusted life months: 0.25 Salmeterol vs. ipratropium difference The Netherlands Costs: -128 Exacerbations avoided: 0.12 Quality-adjusted life months: 0.06 Tiotropium vs. salmeterol difference Canada Costs: 3 Salmeterol vs. ipratropium Difference Canada Costs: -1 Oostenbrink et 2004 The Prospective, trial- Tiotropium: 64 Mean number of exacerbations Total health care costs Euro 2001 ICER tiotropium vs ipratropium 156

157 Study Pub. year Country Summary of model Patient population (average age in years) QALYs (intervention, comparator) Costs (currency) (intervention, comparator) ICER (per QALY gained) al. 103 Netherlands and Belgium, societal perspective based costeffectiveness analysis of one-year study Ipratropium: 65 per patient over 1 year Tiotropium: 0.74 Ipratropium: 1.01 Percentage of patients with an improvement of at least four units on the SGRQ after 1 yr Tiotropium: 51.2% Ipratropium: 34.6% (Dutch prices) Tiotropium: 1721 Ipratropium: 1541 costs Euro 2001 (Dutch prices) Cost per exacerbation Avoided: 667 Cost per patient improved on the SGRQ: 1084 Rutten-van Molken et al UK Payer Cost-effectiveness analysis performed alongside 1-year RCT Roflumilast: 65 Placebo: 64 Mean 1 year health outcomes Roflumilast Moderate or severe exacerbations: 0.96 Net SGRQ result: 0.19 Placebo Moderate or severe exacerbations: 1.06 Net SGRQ result: 0.14 Total treatment costs (Euros 2004) Roflumilast: Placebo: Cost per moderate or severe exacerbation avoided Roflumilast vs placebo (Euros 2004) Rutten-van Molken et al b Spanish National Health System Markov model with 5- year time horizon with 3 COPD health states and death. First cycle length 8 days followed by monthly-cycles Not stated Cumulative 5 year QALYs Tiotropium: 3.15 Salmeterol: 3.02 Ipratropium: 3.00 Total treatment costs (Euro 2005) Tiotropium: 6424 Salmeterol: 5869 Ipratropium: 5181 Cost per QALY (Euro 2005) Tiotropium vs Salmeterol: 4118 Salmeterol vs Ipratropium: Sin US, UK and other Western Europe Health state transition model using Markov techniques. 3-year time horizon with 3- month cycles. 3 FEVbased health states. Death captured Inhaled corticosteroids: Stage 3 disease: 2.72 Stage 2 or 3: 2.72 All patients: 2.75 No ICS: 2.71 Marginal costs over 3 yrs (CAN D) for inhaled corticosteroids: Stage 3 disease: $774 Stage 2 or 3: $922 All patients: $3612 Incremental cost per QALY gained (US D) for inhaled corticosteroids: Stage 3 disease: $11,100 Stage 2 or 3 disease $17,000 All patients: $46,200 Spencer Not stated. Pooled data from trials. Markov model with 3 COPD states plus death. 25-year time horizon. 3-month cycle Salmeterol/fluticasone 4.21 Usual care (placebo) 4.08 Total discounted cost (CAN D) Salmeterol/fluticasone $25,780 Usual care (placebo) C$16,415 Incremental cost per QALY gained (CAN D) Salmeterol/fluticasone vs usual care C$74,

158 Study Pub. year Country Summary of model Patient population (average age in years) QALYs (intervention, comparator) Costs (currency) (intervention, comparator) ICER (per QALY gained) ICER: Incremental cost-effectiveness ratio; QALY(s): Quality-adjusted life year(s); RCT: Randomised controlled trial; SGRQ: St George s Respiratory Questionnaire; CEAC: Cost-effectiveness acceptability curves Table B 42: Summary of data in abstracts identified in conference proceedings search Ref Author(s) Title Country Summary of Model ISPOR 13th Hoogendoorn Developing and applying a The Incidence-based Annual M, Feenstra T, stochastic dynamic population Netherlands model including European hoogenveen model for chronic obstructive age, gender, 4 Congress RT, Al M, pulmonary disease severity stages and PRS6, 2010 Rutten-van smoking status. Mölken MP 108 FEV decline tracked. ISPOR 13th Annual European Congress PRS25, 2010 ISPOR 13th Annual European Congress PRS27, 2010 ISPOR 13th Annual European Congress PRS32, 2010 ISPOR 15th Annual International Meeting PRS26, Zanlolo O, Iannazzo S, Carsi M 109 Vemer P, Goossens LM, Rutten-Van Mölken MP 110 Demko IV, Tolkushin A, Yagudina RI, Kulikov A, Ogorodova LM, Chuchalin AG 111 Igarashi A, Kato Y, Makita H, Fukuda T, Nishimura M, Tsutani K 112 A cost-utility analysis for tiotropium bromide in the long term treatment of specific subgroups of Italian COPD patients Country adaptation of a health economic model: the case for roflumilast in The Netherlands Regional differences as a basis for sensitivity analysis of cost-effectiveness of salmeterol + fluticasone propionate (SAL/FP) vs. Inhaled corticosteroids (MONO-ICS) Cost-utility analysis of tiotropium medicine for chronic obstractive pulmonary diseases (COPD) in Japan Italy The Netherlands Russia Probabilitstic, patient-level simulation Markov model with lifetime horizon Microsimulationbased Markov model Societal perspective Previously published OPTIMA model Patient population COPD patients COPD patients Adult patients with severe COPD and a history of frequent exacerbations COPD patients Japan Markov model COPD patients Outcomes Costs (currency) ICER(s) QALYs Euros Compared to a reference of minimal treatment, CE of the four interventions (2 x pharmacotherapy) ranged from EUR 6,100 to EUR 12,200 per QALY gained Patients treated with tiotropium gained an average of 0.5 life years and 0.42 QALYs compared to routine care The incremental lifetime cost for patients treated with tiotropium vs. to routine care was EUR 3,357 The ICER for tiotropium versus routine care was EUR 7,916 per QALY gained, ranging from EUR 6,627 per QALY gained in females at GOLD III to EUR 13,187 per QALY gained in all patients <65 years at GOLD IV QALYs Euros The ICER for LABA + roflumilast was EUR 7,900 / QALY gained vs LABA alone. The ICER for LABA + roflumilast was EUR 10,000 / QALY gained vs. LABA + ICS QALYs Roubles Salmeterol + fluticasone was cost-saving vs. ICS in 18 regions, cost-effective in 62 regions (ICERs ranged from from 3210RUB to 639,480RUB per QALY), and not cost-effective in 4 regions. Tiotropium compared to ipratropium would increase Tiotropium compared to ipratropium would save JPY 966,000 per patient Tiotropium would be dominant compared to ipratropium. 158

159 Ref Author(s) Title Country Summary of Patient Outcomes Costs (currency) ICER(s) Model population QALY Brosa M, Diaz Cost-effectiveness analysis of Spain Not stated - 13 COPD Life years Euros Tiotropium bromide was costeffective S, Miravitlles tiotropium in the treatment of year time horizon patients in vs. standard therapy M, Gonzalez- chronic obstructive pulmonary was used Spain with an ICER of EUR2873 / LYG Rojas N, disease (COPD) patients in and vs. ipratropium bromide with Nieves D 113 Spain an ICER of EUR4208 / LYG ISPOR 12th Annual European Congress PRS22, 2009 ISPOR 12th Annual European Congress PRS27, 2009 ISPOR 13th Annual International Meeting PRS12, 2008 Menn P 114 Oba Y 115 Markov model for healtheconomic evaluations of COPD in Germany Incremental cost-effectiveness of combination inhaler therapy in moderate to severe COPD Germany US Seven-stage Markov model including 4 GOLD stages, postsurgery, posttransplatation and death. 60 year time horizon. Exacerbations categorized into mild, moderate and severe. Markov model with 3 year time horizon COPD patients with smoking- and ex-smokingspecific probabilities Moderate to severe COPD Smoking cessation resulted in an increase of 0.5 QALYs Smoking cessation resulted in cost savings of EUR 1000 The smoking cessation program was dominant. QALYs USD Tiotropium + salmeterol was associated with an ICER of $152,743 / QALY gained vs. tiotropium alone. Tiotropium + salmeterol-fluticasone was associated with an ICER of $51,610 / QALY gained vs. tiotropium alone. 159

160 6.1.3 Please provide a complete quality assessment for each costeffectiveness study identified. Use an appropriate and validated instrument, such as those of Drummond and Jefferson (1996) or Philips et al. (2004). For a suggested format based on Drummond and Jefferson (1996), please see section 9.11, appendix 11. A quality assessment of the studies in Table B 41 and Table B 42 above has been presented in Error! Reference source not found. in Appendix 11. The critical appraisal of these studies is presented inerror! Reference source not found., also in Appendix

161 6.2 De novo analysis The initial approach used in the model, and described below, was tested with Professor Andrew Briggs in June The objectives of the initial validation meeting were to: Discuss and validate the modelling approach and the overall model structure. Gain insight into the appropriate use of evidence from clinical trials and literature. Gain insights into the model assumptions that are likely to be acceptable by HTA agencies in the EU. Identify any additional re-analysis of the clinical trial data to populate the model and provide supportive evidence. Taking into consideration the validation meeting with Professor Andrew Briggs, the cost-effectiveness model was further developed. A final model validation meeting with Professor Andrew Briggs took place in August The objectives of this meeting were to: Present and discuss the initial CE model results. Discuss the methods to model COPD progression (Sections and 6.3.2). Validate key assumptions in the model, e.g. the use of a hospital case fatality rate post severe (hospital-treated) exacerbations (Section 6.3.4). Discuss the most appropriate utilities to be used in the model in light of recently published data (Section 6.4.9). Discuss the implementation of a lung function benefit. Discuss the implementation of the probabilistic sensitivity analysis (PSA) for the cost-effectiveness model. At the same stage of model development, Dr Maureen Rutten-van Mölken collaborated in the validation of the model to ensure it meets the requirements of the wider European HTA bodies. The recommendations from the validation meeting were implemented in the cost-effectiveness model described in this submission. A multiple treatment comparison (MTC) was conducted after these consultations, and developed independently by Mills et al

162 As recommended by the ERG 1, the scope of the economic analysis was further extended to allow for a full incremental analysis, i.e. simultaneously comparing all treatment combinations. To meet these requirements, the results of an additive main effects model, also conducted by Mills et.al., was used within the modelling. 22 The underlying methods remained the same as validated at the model development stage. Consultation has also been held with experienced and respected specialists in the area of COPD: Dr Michael Rudolf (Respiratory Physician, Ealing Hospital) and Professor Wisia Wedzicha (Consultant Respiratory Physician, Royal Free Hospital). Both were consulted for their views on where in the treatment pathway they expected roflumilast to be used in clinical practice, the appropriateness of assumptions used to the define the base case analyses (Section 6.3.8). Patients What patient group(s) is(are) included in the economic evaluation? Do they reflect the licensed indication/ce marking or the population from the trials in sections 1.4 and 5.3.3, respectively? If not, how and why are there differences? What are the implications of this for the relevance of the evidence base to the specification of the decision problem? For example, the population in the economic model is more restrictive than that described in the (draft) SPC/IFU and included in the trials. The economic evaluation includes a cohort of patients with COPD. Results from the MTC study (Mills et al, 2011), that included the results from M2-124 and M2-125 clinical trials, support the efficacy of roflumilast when added to a bronchodilator in reducing exacerbations and improving lung function. The modelled population is consistent with the patient population anticipated to receive roflumilast in line with its licensed indication (Appendix 1), i.e. in the maintenance treatment (as an add-on to bronchodilator treatments) of severe chronic obstructive pulmonary disease (FEV 1 post-bronchodilator less than 50% predicted) associated with chronic bronchitis in adult patients who continue to exacerbate. COPD patients with associated chronic bronchitis are easily identifiable in clinical practice, through chronic cough and sputum production and represent a COPD 1 Referring to the from 13th January 2011, sent by Bijal Joshi. 162

163 population at higher risk of exacerbations compared with COPD patients without chronic bronchitis. Exacerbations are associated with a worsening prognosis including an increase in mortality. 8;11;116 These patients, who continue to exacerbate despite receiving currently recommended treatment regimens, therefore represent a high unmet need. Model structure Please provide a diagrammatical representation of the model you have chosen. The economic model was developed to assess the long-term costs and health outcomes of roflumilast as an add-on to bronchodilator therapies vs. standard of care. The model developed is a Markov state-transition cohort model with five health states: 1. Severe COPD, first line regimen. 2. Severe COPD, second line regimen. 3. Very severe COPD, first line regimen. 4. Very severe COPD, second line regimen. 5. Dead A simplified scheme of the Markov structure is shown in Figure 14 below. This approach is broadly consistent with published economic analyses assessing treatments for COPD. Figure 14 illustrates the patients pathways within the model. In Severe and Very severe COPD, patients can experience exacerbations, which are modelled as events occurring within a given health state. These exacerbations may be community treated (moderate), or hospital treated (severe), with an associated risk of in-hospital mortality. 163

164 Figure 14: Markov Model Structure. Severe COPD, first line regimen (S 1 ): The S 1 health state corresponds to patients with severe COPD, who are continuing to exacerbate, despite their current treatment regimen (and so are eligible to receive roflumilast). The model compares 10 appropriate treatment regimens for this population, refered to as "first line treatment" in the model. Severe COPD, second line regimen (S 2 ): The S 2 health state corresponds to patients with severe COPD who initially receive first line treatment (as described above), but continue to exacerbate, or remain breathless on the model first-line therapy. In line with recommendations made in NICE CG101 1, and clinical practice, they are switched to a subsequent regimen, which is referred to as "second line treatment" in the model. Very severe COPD, first line regimen (VS 1 ): The VS 1 health state describes patients who were on a first line treatment regimen and progress to Very severe COPD, while remaining on first line treatment. This is not inconsistent with NICE CG101, as the guideline does not specifically recommend a change in the treatment regimen when FEV 1 falls below 30% of the predicted value. 164

165 Very severe COPD, second line regimen (VS 2 ): The VS 2 health state represents: a) patients who were on a first line treatment regimen in the Very severe COPD state (VS 1 ), and who are then switched to a subsequent second line treatment regimen; b) patients who were on a second line regimen in the Severe COPD state (S 2 ), and then progressed to very severe COPD, continuing the second line regimen, and; c) patients initially in the S 1 health state, who were switched to a second line treatment regimen and progressed to very severe COPD. The base case analysis assumes that the cohort starts in the Severe COPD, first-line regimen, health state (S 1 ). The details of possible transitions are shown in Figure 15 below. Figure 15: Detailed model scheme 165

166 The primary analysis estimates the incremental cost per QALY gained from the NHS perspective, using a 30-year time horizon. Other health outcomes estimated in the model include life years (LY) and the number of exacerbations Please justify the chosen structure in line with the clinical pathway of care identified in section 2.4. The use of Markov models is common in the economic modelling of chronic diseases (such as COPD) which have recurring symptoms over a prolonged period of time, with these symptoms being characterised by recurrent costs and health outcomes. This model was designed to capture the benefits of treatment resulting from both a lower rate of exacerbations, and the impact of treatment on lung function (FEV 1 ). Such an approach is broadly consistent with that used in published economic analyses in COPD. The severity of COPD in the defined model health states is determined by lung function, specifically as predicted post-bronchodilator FEV 1 %, as described by NICE CG101, and the GOLD classification system (Table B 43). 1;13 The progression from Severe COPD to Very severe COPD is determined by the decline of lung function in the COPD population, relative to the natural decline of lung function in the general population. The presence of first line and second line treatment regimens in the model structure is to allow the possibility of changing therapies over a patient s lifetime, reflecting long term clinical pathways, and recommendations made within the COPD guidelines. The base case assumes that changes in the treatment regimen occur independently of disease progression: this is not inconsistent with NICE CG The model incorporates the functionality to vary a patient's time on first line therapy, in addition to the functionality to enforce a therapy switch upon disease progression. Table B 43: Severity of COPD, as defined by NICE CG101 (2010) and the GOLD classification (2009) 1;13 COPD State NICE/GOLD classification of COPD severity (based on post-bronchodilator FEV 1 % predicted) Severe COPD 30% FEV 1 < 50% Very severe COPD FEV 1 < 30% 166

167 6.2.4 Please define what the health states in the model are meant to capture. The health states in the model capture the worsening of underlying COPD over time in a cohort of patients (i.e. disease progression), and a possibility to modify treatment. The transitions from the states with severe COPD (i.e. S 1 and S 2 ) to the states with very severe COPD (i.e. VS 1 and VS 2 ) represent progression of the disease. Compared to the severe COPD state, very severe COPD is associated with increased patient management costs, lower quality of life, and an increased risk of COPD exacerbations and/or death. Progression between health states with the same severity of COPD (S 1 to S 2 and VS 1 to VS 2 ) reflects the possible modification of the therapy regimen in clinical practice. Each model state is characterised by a cost per cycle and health-related utility. The duration of the model cycle is one month. COPD maintenance costs, transition probabilities and exacerbation rates are thus calculated to reflect a one month cycle duration. The cost of an exacerbation is calculated based on resource use per event. The full loss in QALYs due to an exacerbation is applied within the model cycle where the exacerbation occurred. Within each model cycle the cohort is at risk of an exacerbation, which can be moderate (community-treated) or severe (hospital treated). In the analysis, exacerbations are modelled as events. After having an exacerbation in the severe COPD state (S 1 ), patients can: remain in the original state, on first line treatment (S 1 ); remain in the original state, and switch to second line treatment (S 2 ); progress to the more severe state (very severe COPD) and remain on first line treatment (VS 1 ); progress to the more severe state (very severe COPD) and switch to second line treatment (VS 2 ), or; die (hospital mortality). 167

168 Exacerbations are associated with a management cost and a utility decrement. Mild exacerbations are not included in the model, as they do not require the same level of healthcare intervention, and were not included in the primary endpoints of the roflumilast clinical trial programme. However, a therapeutic benefit resulting in a reduction of rates of moderate and severe exacerbations can be expected to also decrease the number of mild exacerbations, with resultant benefit to QoL and cost of managing the exacerbation. Within any comparison of treatment options, the exclusion of mild exacerbations is therefore a conservative estimation of the total benefit for the superior treatments. In summary, within each model cycle a patient can (options are not mutually exclusive): Die due to background mortality. The background mortality rate is determined by the mortality rate in the general population, adjusted by a standardised mortality ratio (SMR) which excludes deaths due to hospital-treated exacerbations (SMR values are assigned according to the degree of severity of COPD). Experience an exacerbation. The model considers both community-treated and hospital-treated exacerbations, with a fixed proportion of all exacerbations being treated in hospital. This proportion is assumed to differ between patients with severe COPD, and those with very severe COPD. The proportions are defined by data from the roflumilast clinical trial programme. 41 Patients who experience a hospital-treated (severe) exacerbation are also at risk of death (defined by a hospital case fatality rate; see Section 6.3.4). Background mortality and hospital mortality combined represent all-cause mortality in the modelled COPD population. Switch to a second line treatment regimen. Patients can switch from a first line treatment regimen to a second line regimen with the average time T 1 on the first line regimen. The base case assumes patients remain on first line treatment for a period of 12 months (see Section 6.3.8). The continuous monthly probability of switching regimens is estimated from the average time on first line regimen as 1/ T 1. The time in state is varied in a sensitivity analysis.. 168

169 Progress to very severe COPD. Patients in the states with severe COPD on first and second line regimens (S 1 and S 2 ) are at risk of progression to very severe COPD. In the base case analysis, the risk of progression from "S" to "VS" health states does not depend on the treatment regimen and is determined by the constant annual rate of decline of lung volume (as measured by FEV 1 ) in patients with COPD, in relation to the natural decline of lung volume in the general population. 117 Due to the irreversible nature of COPD, it is assumed that there are no back transitions from very severe to severe COPD. It is also assumed that patients who switch to a second line regimen in the model do not return to first line regimens How does the model structure capture the main aspects of the condition for patients and clinicians as identified in section 2 (Context)? What was the underlying disease progression implemented in the model? Or what treatment was assumed to reflect underlying disease progression? Please cross-reference to section 2.1. The model structure captures potential therapeutic benefits through a difference in event rates. The events of interest are exacerbations (classified as either moderate (community-treated) or severe (hospital-treated)), and deaths (due to a reduction in hospital-treated exacerbations and the CFR; see Section 6.3.4). In addition the model incorporates functionality to explore the effect of a one-off benefit in lung function. As reported in section 5.7.6, few studies were found which reported change in FEV 1 between the treatments of interest. Among those studies identified which did report such data, the data were commonly of poor quality. The length of the studies reporting FEV 1 was also found to vary and had influence on the absolute values of change in FEV 1 from baseline. Various analytical approaches were explored to estimate the one-off lung function benefit between therapies; however, none were considered robust enough to justify the inclusion of a lung function benefit in the base case. The effects of inclusion of a lung function benefit on base case results are reported via a Scenario analysis (section and 6.7.9). 169

170 It should also be noted that the primary aim of COPD treatment is the prevention of exacerbations, as opposed to improvement in the surrogate outcome of lung function, per se. Calculation of time in state The method used to calculate the time a patient spends in a particular health state, as described below, is analogous to that used by Spencer et al. 107 The cohort is assumed to start in the severe COPD health state S 1, with a lung function which lies in the middle of the Severe COPD FEV 1 % range, i.e. 40% predicted (Table B 43). The transition of patients from the Severe COPD state to the Very severe COPD state was determined by continuous per-cycle probabilities, defined as 1/T, based upon the estimated average time (T) in the Severe COPD states. The base case analysis assumes that the probability of COPD progression does not depend on the line of treatment, and is therefore equal for the S 1 and S 2 states. The average time (T) in Severe COPD is calculated from the rate of decline of lung volume in patients with COPD (0.052 litres / year; Scanlon et al) 118, and the rate of decline in a general (non-copd) population (Crapo et al) 117, to estimate the time at which lung volume in patients with severe COPD reaches 30% of lung volume in a general adult population. Figure 16 illustrates the concept used to calculate the time to COPD progression, and so the probability of transition to the very severe disease states (VS 1 and VS 2 ). 170

171 Figure 16: Time to disease progression in patients starting in the Severe COPD state (illustrative example; no lung function benefit considered) In Figure 16, the dotted black line represents the threshold over time for a general adult population for transition to 30% predicted FEV 1. The solid blue line represents the declining lung volume in the modelled cohort of patients who start with Severe COPD (40% of predicted FEV 1 ). The predicted intersection of these lines determines the average time in the Severe COPD state. That time in state is used to calculate the continuous monthly probability of transition from the Severe COPD states to the Very severe COPD states. The impact of incorporation of a lung function benefit, relative to the common comparator treatment, was tested in a scenario analysis (see Sections 6.6 and 6.7.9). Duration of lung function benefit was assumed to be one year, while a patient is on first line treatment. A lung function benefit while a patient is on second line treatment was not considered. Figure 17 illustrates the impact of improvement in lung function due to a particular treatment (vs. a comparator) on the progression of patients from severe to very severe COPD. An improvement in lung function results in a prolonged estimated time in the severe COPD states, and therefore, a lower per cycle probability of progression to very severe COPD. A lung function improvement would be expected to increase the time spent in a particular state, with resultant cost savings and QOL benefit. 171

172 Figure 17: Time to disease progression in patients starting in the Severe COPD state (illustrative example; considering a lung function benefit due to the treatment) The transition from first line regimen to second line regimen was determined from an estimated average time on first line regimen, assumed to be 12 months in the base case. In two separate scenario analyses, the average time on first line therapy, and the assumption that patients switch treatment regimens on progression from severe to very severe COPD, were explored. 172

173 6.2.6 Please provide a table containing the following information and any additional features of the model not previously reported. A suggested format is presented below. Table B 44: Key features of analysis Factor Time horizon Cycle length Half-cycle correction Were health effects measured in QALYs? Discount of 3.5% per annum for utilities and costs Perspective (NHS/PSS) Chosen values 30 years (COPD patient lifetime) 1 month Yes Yes Yes NHS NHS, National Health Service; PSS, Personal Social Services; QALYs, quality-adjusted life years The choice of model cycle length of one month is based on the assumption that no more than one exacerbation is likely to occur within one model cycle and that one month is sufficient to capture the disutility of an exacerbation. A cycle length of one month is consistent with previous economic analyses of COPD treatment. Technology Are the intervention and comparator(s) implemented in the model as per their marketing authorisations/ce marking and doses as stated in sections 1.3 and 1.5? If not, how and why are there differences? What are the implications of this for the relevance of the evidence base to the specified decision problem? Yes. 173

174 6.2.8 Please note that the following question refers to clinical continuation rules and not patient access schemes. Has a treatment continuation rule been assumed? If the rule is not stated in the (draft) SPC/IFU, this should be presented as a separate scenario by considering it as an additional treatment strategy alongside the base-case interventions and comparators. Consideration should be given to the following. The costs and health consequences of factors as a result of implementing the continuation rule (for example, any additional monitoring required). The robustness and plausibility of the endpoint on which the rule is based. Whether the response criteria defined in the rule can be reasonably achieved. The appropriateness and robustness of the time at which response is measured. Whether the rule can be incorporated into routine clinical practice. Whether the rule is likely to predict those patients for whom the technology is particularly cost effective. Issues with respect to withdrawal of treatment from nonresponders and other equity considerations. Not applicable. 174

175 6.3 Clinical parameters and variables When relevant, answers to the following questions should be derived from, and be consistent with, the clinical-evidence section of the submission (section 5). Cross-references should be provided. If alternative sources of evidence have been used, the method of identification, selection and synthesis should be provided as well as a justification for the approach Please demonstrate how the clinical data were implemented into the model. Clinical management of COPD patients The patient population being assessed in the base case of the model was assumed to have severe COPD, with an average FEV 1 of 40% predicted, who continued to exacerbate, despite prior treatment. In the base case, different therapeutic options were compared in a fully incremental manner, in response to the suggestion from the ERG 2. In order to explore the effect on a second line treatment regimen, the model allows patients to switch from a first line to second line treatment regimen. The transition was determined from an estimated average time on first line regimen, assumed to be 12 months in the base case (see section and 6.3.8). It was assumed that ICStolerant patients, and patients who decline/are ICS intolerant, receive LAMA + LABA/ICS and LAMA + LABA, respectively, if they continue to exacerbate. Such an approach is consistent with NICE CG101, and is assumed to be common across all treatment options. 1 As per the clinical indication for roflumilast, patients were assumed to continue to exacerbate despite current therapy. A common first line comparator was selected in order to assign the baseline rate of exacerbation for each potential treatment regimen within the fully incremental analysis. LAMA + LABA/ICS was chosen as the common comparator, and was assigned a background exacerbation rate, assumed to be 2 exacerbations per year. The reduction in the rate of exacerbations for different treatments versus LAMA + LABA/ICS were derived from the independently-performed additive main effects model in the MTC (Table B 47 and Table B 48) and applied to the respective regimens relevant for ICS-tolerant patients and ICS-intolerant/declining patients Referring to the from 13th January 2011, sent by Bijal Joshi. 175

176 For a common treatment regimen (e.g. LAMA + LABA), no difference was assumed in the exacerbation rate between a patient who is able to receive ICS, vs. a patient who is unable to receive ICS. This assumption was confirmed as being acceptable after consultation with Dr Michael Rudolf and Professor Wisia Wedzicha. The different therapeutic regimens for ICS-tolerant patients are listed below. Table B 45: Therapy options for ICS-tolerant patients Therapy option no. 1st line treatment ICS-tolerant patients 2nd line treatment if patients continue to exacerbate 1 LABA 2 LAMA 3 LABA + roflumilast 4 LAMA + roflumilast 5 LAMA + LABA LAMA + LABA/ICS 6 LABA/ICS 7 LAMA + LABA + roflumilast 8 LABA/ICS + roflumilast 9 LAMA + LABA/ICS* 10 LAMA + LABA/ICS + roflumilast * Selected common 1 st line comparator (the assigned background exacerbation rate was assumed to be 2 exacerbations per year) For patients in whom ICS is not tolerated or declined, a fully incremental analysis was performed on the treatment options specified in the following table: Table B 46: Therapy options for ICS-intolerant/declining patients Therapy option no. 1st line treatment ICS-intolerant/declining patients 2nd line treatment if patients continue to exacerbate 1 LABA 2 LAMA 3 LABA + roflumilast LAMA + LABA 4 LAMA + roflumilast 5 LAMA + LABA 6 LAMA + LABA + roflumilast 176

177 Exacerbations The exacerbation rates are important parameters in the economic evaluation as they determine the patient groups who are projected to benefit from the treatment, through a reduction in the number of exacerbations. 14 Sections and described the methodology and results from the additive MTC analysis, as published by Mills et al, The fully incremental analysis requires comparisons to be performed against common comparator, which was selected to be LAMA + LABA/ICS. Where rate ratios of exacerbations between a regimen interest and LAMA + LABA/ICS were not reported in the publication, they have been calculated from results of the additive analysis. Relative rate ratios (RRR) of exacerbation are presented in Table B 47 below. 22 To model the comparison of the treatment effect, the background exacerbation rate (2 exacerbations per year in first line LAMA + LABA/ICS) was multiplied by the relative rate ratio associated with the respective treatment (Table B 47). The relative rate ratio of exacerbations was assumed to be the same in both severe and very severe COPD states. Table B 47: Rate ratio of exacerbations for the common comparator (LAMA + LABA/ICS) vs. treatments, Mills et al., # Comparisons Relative rate ratio (RRR) of exacerbations 95% CI 1 LAMA + LABA/ICS vs. LABA (0.5327, ) 2 LAMA + LABA/ICS vs. LAMA (0.6359, ) 3 LAMA + LABA/ICS vs. LABA + roflumilast (0.6038, ) 4 LAMA + LABA/ICS vs. LAMA + roflumilast (0.7183, ) 5 LAMA + LABA/ICS vs. LAMA + LABA (0.7594, ) 6 LAMA + LABA/ICS vs. LABA /ICS (0.6704, ) 7 LAMA + LABA/ICS vs. LAMA + LABA + roflumilast (0.8449, ) 8 LAMA + LABA/ICS vs. LABA/ICS + roflumilast (0.7549, ) 9 LAMA + LABA/ICS vs. LAMA + LABA/ICS LAMA + LABA/ICS + roflumilast vs. LAMA + LABA/ICS (0.7372, ) With the exception of comparison 10 in Table B 47, the analysis required the relative rate ratios (and associated credible intervals) to be inverted from those presented within the publication (Table B 48). Standard errors (S.E.) for distributions used in probabilistic sensitivity analyses were calculated subsequently. 177

178 Table B 48: Inverted rate ratio of exacerbations from Table B 47 for treatments vs. the common comparator (LAMA + LABA/ICS), based on Mills et al., # Treatment 1 LABA 2 LAMA 3 LABA + roflumilast 4 LAMA + roflumilast 5 LAMA + LABA 6 LABA/ICS 7 LAMA + LABA + roflumilast 8 LABA/ICS + roflumilast 9 LAMA + LABA/ICS 10 LAMA + LABA/ICS + roflumilast Common comparator vs. LAMA + LABA/ICS vs. LAMA + LABA/ICS vs. LAMA + LABA/ICS vs. LAMA + LABA/ICS vs. LAMA + LABA/ICS vs. LAMA + LABA/ICS vs. LAMA + LABA/ICS vs. LAMA + LABA/ICS vs. LAMA + LABA/ICS vs. LAMA + LABA/ICS 1) Assumed to be the average across all S.E. Relative rate ratio (RRR) of exacerbations 95% CI S.E (1.4567, ) (1.2750, ) (1.1583, ) (1.0052, ) (1.1361, ) (1.2265, ) (0.8842, ) (0.9662, ) (0.70, 1.30) ) (0.7372, ) Demonstrate how the transition probabilities were calculated from the clinical data. If appropriate, provide the transition matrix, details of the transformation of clinical outcomes or other details here. The clinical impact of therapy is primarily differentiated through a reduction in exacerbation rates. As described in Section 6.2.5, continuous probabilities for transition from the Severe COPD state to the Very severe COPD state were derived from the relationships between the decline of the lung function in COPD population and the natural decline of the lung function in the non-copd population, by estimating the time required for patients in the Severe COPD state to progress to the Very severe COPD state. An example calculation for the average time in state, and probability for a patient transitioning from Severe to Very severe COPD, is provided in Appendix

179 6.3.3 Is there evidence that (transition) probabilities should vary over time for the condition or disease? If so, has this been included in the evaluation? If there is evidence that this is the case, but it has not been included, provide an explanation of why it has been excluded. As described above, and shown in Figure 16 and Figure 17, transition probabilities are based on age rather than on a specific time function Were intermediate outcome measures linked to final outcomes (for example, was a change in a surrogate outcome linked to a final clinical outcome)? If so, how was this relationship estimated, what sources of evidence were used, and what other evidence is there to support it? The intermediate health outcomes in the model are COPD exacerbations. Moderate (community-treated) and severe (hospital-treated) COPD exacerbations affect final clinical outcomes in two ways: 1. An exacerbation (either community- or hospital-treated), affects health-related quality of life. In the model, this is represented by a reduction of health utility for one cycle, the loss of utility depending upon the severity of the exacerbation. 2. A severe (hospital-treated) exacerbation can result in death, due to a case fatality rate for hospital-treated COPD exacerbations. The hospital case fatality rate (CFR) was estimated using data from the National COPD Audit (2008) which was conducted as a collaboration consisting of the Royal College of Physicians, The British Thoracic Society and the British Lung Foundation. 28 The 2008 audit reports data on 9,716 COPD exacerbation admissions, to 232 NHS acute hospitals, within 177 NHS Trusts. The results indicate that the 90- day mortality rate post hospital-treated exacerbation is 13.9%, with COPD being the cause of death in 65% of mortality cases. The inpatient mortality rate was reported as 7.7%. This rate (7.7%) is used in the model as the hospital CFR, assuming that all COPD deaths are within the context of a hospital stay, and occur within the cycle of exacerbation occurrence. This is below the in-hospital mortality rate following an acute exacerbation of 10% (with one-year mortality approaching 25%), as reported by FitzGerald, A reduction of the rate of exacerbation due to treatment is thus implicitly linked to a lower rate of hospital-treated exacerbation, and consequently to a lower CFR. Given 179

180 the immediate risks of a severe exacerbation to in-hospital mortality and posthospitalisation 1-year mortality rates, a reduction in exacerbation rate can have a considerable impact on the humanistic burden of COPD, implicitly reducing its lifethreatening consequences. 30 The inclusion of the implicit link between an effect on the exacerbation rate and resultant effect on mortality was discussed with Dr Michael Rudolf and Professor Wisia Wedzicha, who both agreed that such inclusion was appropriate. Lung function improvement is another surrogate outcome that can be captured within the model (Figure 17), although these effects are not considered within the base case analysis. Change of lung function due to treatment affects the time of progression of COPD patients from the Severe COPD state to the Very severe COPD state, and therefore keeps patients in the health state associated with lower maintenance cost, higher utility, lower risk of exacerbations, and lower background risk of death. Sensitivity analysis for the incorporation of a lung function benefit for roflumilast, relative to the comparator treatment, was assessed in a scenario analysis (B) (Sections and 6.7.9) If clinical experts assessed the applicability of values available or estimated any values, please provide the following details: See Section 6.2. Summary of selected values Please provide a list of all variables included in the cost-effectiveness analysis, detailing the values used, range (distribution) and source. Provide cross-references to other parts of the submission. Please present in a table, as suggested below. 180

181 Table B 49: Summary of variables applied in the economic model (base case analyses) Variable Variable Name Value CI Description Cost associated with AEs AE_costs 0 in the base case and in sensitivity analysis N/A Used in sensitivity analysis only. An assumption for the cost of mild or moderate transient AE, assumed as 50% of the cost of treatment of moderate exacerbation Utility loss associated with an AE (definition for root) AE_disutil_root 0 in the base case and 0.06 in sensitivity analysis N/A (dur_of_comm_exac*u_loss_comm_ex)*50% Used in sensitivity analysis only an assumption of the loss of utility corresponding to 50% loss of utility due to a moderate exacerbation Current average age of cohort in the model age 64 N/A age_start + Int(_stage/12) To account for patient ageing that is used to estimate the background mortality from standard life tables Age at start age_start 64 N/A Absolute reduction of utility due to a community exacerbation that AR_util_dec_comm applies for one year Based on the mean age in LABA subset of the trials M2-124 and M2-125, can be user defined. 41 Used to estimate the rate of decline of lung volume in the non-copd population Accounted for within the one-month model cycle by recalculation as 0.01 * 12 = 0.12 to deduct from baseline utility value in stable Severe and Very severe COPD 26 Absolute reduction of utility due to a hospital exacerbation that applies for one year Monthly COPD drug cost of the comparator arms AR_util_dec_hosp N/A cost_drugs_sev_1 to cost_drugs_sev_10 Accounted for within the one-month model cycle by recalculation as * 12 = to deduct from baseline utility value in stable Severe and Very severe 26 Dependent upon Scenario analysed (cost of comparator arms see below) Monthly cost of LABA N/A Monthly cost of LAMA N/A Serevent (most commonly prescribed LABA for COPD) BNF 60 price of 30-days pack, further recalculated to adjust to the average days per month as 365 days /12 months (Section ) Spiriva As above 181

182 Variable Variable Name Value CI Description Monthly cost of LAMA + LABA N/A Serevent + Spiriva As above Monthly cost of LABA/ICS N/A Seretide (most common LABA/ICS) As above Monthly cost of LAMA + LABA/ICS N/A Spiriva + Seretide As above Monthly cost of roflumilast cost_dx N/A Based on the cost of 30-day pack of adjusted to the average days per month of 365/12 Cost of community treated exacerbations Cost of hospital treated exacerbation cost_ex_comm Section cost_ex_hosp 1, Section Monthly cost of maintenance of patients with Severe COPD cost_maint_sev_base SE = 4.76 Cost of maintenance, means and standard errors are estimated from the National Schedule of Reference Costs (see Section 6.5.1). Monthly cost of maintenance when in the Very severe COPD state cost_maint_vsev_base SE = 9.54 Cost of maintenance, means and standard errors are estimated from the National Schedule of Reference Costs (see Section 6.5.1). Monthly cost of being in the severe state (maintenance cost) cost_sev Calculated from components N/A Represents a sum of monthly COPD treatment costs and monthly cost of management, cost of COPD treatment. Dependent upon Scenario and model arm Monthly cost of being in the very severe state (maintenance cost) cost_vsev Calculated from components N/A Represents a sum of monthly COPD treatment costs and monthly cost of management, cost of COPD treatment. Dependent upon Scenario and model arm Annual discount rate for costs disc_cost N/A NICE guide to the methods of technology appraisal Annual discount rate for effect disc_effect N/A NICE guide to the methods of technology appraisal discount factor cost disc_fact_cost Time dependent N/A 1/((1+ disc_cost)^int(_stage/12)). Factor determining present 182

183 Variable Variable Name Value CI Description value of future costs, calculated from the discount rate discount factor for effects disc_fact_effect Time dependent N/A 1/((1+ disc_effect)^int(_stage/12)). Factor determining present value of future outcomes, calculated from the discount rate Decline in FEV 1 for patients with COPD FEV1_decline Calculated from annual decline of FEV 1 in patients with COPD N/A Yearly FEV 1 decline FEV1_decline_pa_base Litres SE = Scanlon The threshold FEV 1 % value for the progression from Severe to Very severe COPD FEV1_thresh_sev2vsev 30% N/A Definition of severity of COPD from the NICE and GOLD guidelines 1;13 Life years in one cycle LY_per_Cycle N/A Duration of one month expressed as fraction of a year Markov termination (cycles) Markov_Term - N/A The number of cycles determining the time horizon within the Treeage model; can be controlled by the user from the Excel front end Mean height in cm of females in the model Mean height (cm) of males in the model Proportion of males in the model cohort Proportion of patients in Severe COPD state at start of model Proportion of patients in Very severe COPD state at start of model mean_height_females LABA subgroup Pooled analysis M2-124 & M mean_height_males LABA subgroup Pooled analysis M2-124 & M prop_males 0.75 LABA subgroup in pooled analysis M2-124 and M prop_start_sev 1 Defined in base case analysis specification prop_start_vsev 0 Defined in base case analysis specification Proportion of all exacerbations pr_ex_hosp Calculated from Proportion of hospital treated exacerbation depend on the 183

184 Variable Variable Name Value CI Description that require hospitalisation components model arm and the severity of COPD Proportion of all exacerbations requiring hospitalisation when in the Severe state pr_ex_hosp_sev SE = LABA subgroup in pooled analysis M2-124 and M Standard error for proportion of hospital exacerbations in Severe COPD state pr_ex_hosp_sev_se SE = LABA subgroup in pooled analysis M2-124 and M Proportion of all exacerbations requiring hospitalisation when in the Very severe state Standard error for proportion of hospital exacerbations in Very severe COPD state Standard error for the cost of community treated exacerbation Standard error for the cost of hospital treated exacerbation Standard error for the monthly cost of maintenance of patients with severe COPD Standard error for the monthly cost of maintenance of patients with very severe COPD Standard error for the hospital CFR Standard error for the value of annual FEV 1 decline in patients with COPD pr_ex_hosp_vsev LABA subgroup in pooled analysis M2-124 and M pr_ex_hosp_vsev_se LABA subgroup in pooled analysis M2-124 and M psa_cost_ex_comm_se NHS Reference Costs 2009/2010 psa_cost_ex_hosp_se NHS Reference Costs 2009/2010 psa_cost_maint_sev_se 4.76 NHS Reference Costs 2009/2010 psa_cost_maint_vsev_se NHS Reference Costs 2009/2010 psa_death_hosp_cfr_se Distribution from UK COPD audit. 28 See Section psa_fev1_decline_pa_se Scanlon Standard error for the value of psa_u_sev_se LABA subgroup in the pooled analysis of M2-124 and M

185 Variable Variable Name Value CI Description the value of utility in the very severe COPD state trials 41 Hospital CFR p_death_hosp_cfr UK COPD audit report 28 Relative rate ratio of exacerbations in LABA vs. LAMA rr_ex_tx1_comp_sev LABA/ICS Relative rate ratio of exacerbations in LAMA vs. LAMA + LABA/ICS rr_ex_tx2_comp_sev Relative rate ratio of exacerbations in LABA + rr_ex_tx3_comp_sev roflumilast vs. LAMA + LABA/ICS Relative rate ratio of exacerbations in LAMA + rr_ex_tx4_comp_sev roflumilast vs. LAMA + LABA/ICS Relative rate ratio of exacerbations in LAMA + LABA vs. LAMA + LABA/ICS Relative rate ratio of exacerbations in LABA/ICS vs. LAMA + LABA/ICS Relative rate ratio of exacerbations in LAMA + LABA + roflumilast vs. LAMA + LABA/ICS rr_ex_tx5_comp_sev rr_ex_tx6_comp_sev rr_ex_tx7_comp_sev Relative rate ratio of exacerbations in LABA/ICS + rr_ex_tx8_comp_sev roflumilast vs. LAMA + LABA/ICS Relative rate ratio of exacerbations in LAMA + rr_ex_tx9_comp_sev (1.4567, ) (1.2750, ) (1.1583, ) (1.0052, ) (1.1361, ) (1.2265, ) (0.8842, ) (0.9662, ) (0.70, 1.30) Additive main effects model from the MTC 22 Additive main effects model from the MTC Additive main effects model from the MTC Additive main effects model from the MTC Additive main effects model from the MTC Additive main effects model from the MTC Additive main effects model from the MTC Additive main effects model from the MTC Additive main effects model from the MTC 185

186 Variable Variable Name Value CI Description LABA/ICS vs. LAMA + LABA/ICS Relative rate ratio of exacerbations in LAMA + rr_ex_tx10_comp_sev LABA/ICS + roflumilast vs. LAMA + LABA/ICS (0.7372, ) Additive main effects model from the MTC Standardised mortality ratio in the severe state Standardised mortality ratio for very severe COPD SMR_sev 1.5 SMR_vsev 2 Standardised mortality ratio (to apply to risk of death in the general population), does not include mortality due to hospital exacerbation (assumption) Standardised mortality ratio (to apply to risk of death in the general population), does not include mortality due to hospital exacerbation (assumption) Time in state for females starting with severe COPD, in monthly cycles time_in_state_sev_female Depends on the parameters in the underlying formulas ((0.0342*mean_height_females *age_start )*FEV1_thresh_sev2vsev-((0.0342*mean_height_females *age_start-1.578)* fev1_percent_start_se v) - FEV1_Improvement)/ (0.0255* (1/12)*FEV1_thresh_sev2vsev- FEV1_decline_pa*(1/12)) Time in state for males starting in severe COPD, in monthly cycles time_in_state_sev_male Depends on the parameters in the underlying formulas ((0.0414*mean_height_males *(age_start)- 2.19)*FEV1_thresh_sev2vsev-((0.0414*mean_height_males *(age_start)-2.19)*fev1_percent_start_sev) - FEV1_Improvement)/ (0.0244* (1/12)*FEV1_thresh_sev2vsev - FEV1_decline_pa* (1/12)) Utility in the current stable state u_current Calculated utility decrement (absolute reduction) while having a community based exacerbation AR_util_dec_comm 0.01 SE = Rutten-van Mölken 2009 Utility decrement (absolute reduction) while in hospital for an AR_util_dec_hosp exacerbation SE = Rutten-van Mölken 2009 utility in stable severe COPD u_sev SE = % CI LABA subgroup in pooled analysis M2-124 and M

187 Variable Variable Name Value CI Description (0.7378, ) utility in stable very severe COPD u_vsev Not applicable This utility is sampled from the distribution for utility in severe COPD multiplied by the sampled reduction of utility in very severe COPD relative to severe COPD. To avoid the scenario of the very severe health state being associated with a higher utility than the severe health state when sampling from separate distributions, the ratio is also used in PSA. 187

188 6.3.7 Are costs and clinical outcomes extrapolated beyond the trial follow-up period(s)? If so, what are the assumptions that underpin this extrapolation and how are they justified? In particular, what assumption was used about the longer term difference in effectiveness between the intervention and its comparator? For the extrapolation of clinical outcomes, please present graphs of any curve fittings to Kaplan-Meier plots. Costs and outcomes are extrapolated beyond time horizon of the trials, with the following assumptions: Background rates of exacerbations and the proportion of severe (hospitaltreated) exacerbations in the Severe and the Very severe COPD states remain constant. Rate ratios for exacerbations remains constant over duration of the analysis. Where used in scenario analyses, the one-off change in lung function has a limited time horizon of 1 year Provide a list of all assumptions in the de novo economic model and a justification for each assumption. The following key assumptions are used within the base case analysis: The patient cohort starts in the Severe COPD state. Patients have 2 exacerbations per year on the common first line comparator (LAMA + LABA/ICS) in the Severe COPD state. Progression to second line therapy is rapid, with a mean time of 1 year on first line therapy. No lung function benefit due to first (or second) line therapy. Switch to second line therapy is independent of progression of disease (i.e. from severe to very severe COPD). All assumptions listed above were discussed for appropriateness through consultation with experienced and respected respiratory physicians; Dr Michael Rudolf (Respiratory Physician, Ealing Hospital) and Professor Wisia Wedzicha (Consultant Respiratory Physician, Royal Free Hospital). The advisors felt that there may be variation in the time to switch to second line therapy, with patients progressing faster than one year (for example, if symptoms continue, or if another 188

189 [single] exacerbation occurs). The effect of variation in this assumption on model results is investigated in the Scenario analyses (see Section and 6.7.9). In the base case, the model differentiates treatment effect through the rate of resulting events, namely exacerbations. The relative reduction in exacerbation rates for roflumilast treatment vs. the common comparator is provided by the additive MTC and is assumed to be independent of patients COPD stage, and independent of exacerbation severity. 22 These assumptions are supported by the post-hoc analysis of the pooled data set for studies M2-124 and M2-125, within subgroup of patients treated with LABA. 41 Based on these assumptions, the following are implied: A certain proportion of exacerbations are severe (hospital treated) exacerbations. Reduction of the rate of all exacerbations consequently results in a reduction of hospital treated exacerbations. A certain proportion of hospital-treated exacerbations can result in death, due to hospital case fatality rate (CFR). The estimates of case fatality rates (7.7%) are based on the results of the COPD audit report (see Section 6.3.4). Further assumptions made within the economic analysis are as follows: Exacerbations can be classified as events occurring within a COPD health state, and can be further grouped into moderate or severe events, with differential associated cost and disutility. A maximum exacerbation rate of one per cycle (one month) is assumed. Any AEs of therapy are mild or moderate and transient in nature, and are assumed not to impact on HRQL. 189

190 6.4 Measurement and valuation of health effects This section should be read in conjunction with NICE s Guide to the methods of technology appraisal, section 5.4. The HRQL impact of adverse events should still be explored regardless of whether they are included in cost-effectiveness analysis. All parameters used to estimate cost effectiveness should be presented clearly in tabular form and include details of data sources. For continuous variables, mean values should be presented and used in the analyses. For all variables, measures of precision should be detailed. Patient experience Please outline the aspects of the condition that most affect patients quality of life. COPD causes disability in patients, including physical impairment and activity limitation. In addition to this, exacerbations requiring hospital admission are shown to place a significant burden on patients health status, with high levels of difficulty in mobility, self care and usual activities. Consistent with the various effects of COPD, NICE CG101 stressed the importance of multidimensional assessment of the disease, as opposed to simple categorisation of disease severity through measure of lung function (Section 6.9 of CG101). 1 The utility decrement associated with exacerbations has been estimated to consistently increase with COPD severity, with estimated decrements ranging from for having one non-serious exacerbation to for having one serious exacerbation per year Please describe how a patient s HRQL is likely to change over the course of the condition. HRQL in COPD deteriorates with disease severity and with age. 119 The increased disability associated with COPD leads to a reduction in quality of life due to a loss of, or decline in, lung function, and an increase in the risk of exacerbation. Exacerbations lead to further disutility, with a greater disutility with each exacerbation experienced. There is thus a HRQL benefit associated with delaying patient s progression to a more severe stage of disease. 190

191 HRQL data derived from clinical trials If HRQL data were collected in the clinical trials identified in section 5 (Clinical evidence), please comment on whether the HRQL data are consistent with the reference case. The following are suggested elements for consideration, but the list is not exhaustive. Method of elicitation. Method of valuation. Point when measurements were made. Consistency with reference case. Appropriateness for cost-effectiveness analysis. Results with confidence intervals. It is appropriate to use health state utilities as measured from the trials, and the EQ- 5D instrument was used during the roflumilast clinical trials. 18 The health utilities measured during the pivotal clinical trials were used to populate the health state utilities in the model. It is not clinically appropriate to administer a health utility instrument during an exacerbation so the disutility of an exacerbation was not elicited during trials. Published literature was used to source model inputs for these events. 26 Sensitivity analyses to the methods for handling disutility for the exacerbations were applied to the cost-effectiveness analysis, in addition to sensitivity analysis of health state values. Mapping If mapping was used to transform any of the utilities or quality-of-life data in clinical trials, please provide the following information. Which tool was mapped from and onto what other tool? For example, SF-36 to EQ-5D. Details of the methodology used. Details of validation of the mapping technique. There was no mapping undertaken for the de novo analysis. 191

192 HRQL studies Please provide a systematic search of HRQL data. Consider published and unpublished studies, including any original research commissioned for this technology. Provide the rationale for terms used in the search strategy and any inclusion and exclusion criteria used. The search strategy used should be provided in section 9.12, appendix 12. The flowchart in Figure 18 illustrates the strategy for the search of studies reporting HRQL/utility values associated with COPD. The search was carried out in each of Medline, Medline In Process, Embase, EconLIT and NHSEED databases. The searches were limited to articles published in the last 10 years. The inclusion criteria for studies were as follows: Studies must be related to the COPD population. Studies must report HRQOL/utility values related to the severity of the disease. Studies must report utilities/utility decrements related to experiencing an exacerbation. The exclusion criteria were: Study is not a utility/hrqol study. Studies related to non-pharmacological management of COPD, e.g. studies of the QOL associated with self-management. Studies relating to patients with asthma or to COPD mixed with asthma. 192

193 Figure 18: Flow diagram for searches of HRQL/utility studies Records from databases: Medline: n = 1458 Embase: n = 1285 EconLIT: n =23 NHSEED: n=294 N = 3060 Duplicates removed n = 1084 Records screened based on title/abstract n = 1976 Studies excluded due to non-english language or being outside the scope of the decision problem n = 1931 Potentially relevant articles and abstracts assessed for eligibility n = 45 Articles meeting inclusion criteria n = Provide details of the studies in which HRQL is measured. Include the following, but note that the list is not exhaustive. Population in which health effects were measured. Information on recruitment. Interventions and comparators. Sample size. Response rates. Description of health states. Adverse events. Appropriateness of health states given condition and treatment pathway. 193

194 Method of elicitation. Method of valuation. Mapping. Uncertainty around values. 194

195 Table B 50: Details of the studies in which HRQL is measured Author Pub. Year Country Population Information on recruitment Comparators Sample size Health states Method of elicitation Method of valuation Mapping Results (CI/SD) UK - specific Andenæs et al Norway COPD Prospective longitudinal study on hospitalization and 2 subsequent phases. COPD N=51 Health status was assessed with the SGRQ in 3 different times. Interview/ protocol WHOQOL NA Scores are presented by domain (Physical, Psychological, Social and Environmental) in Time 1, 2 and 3. O Reilly et al UK COPD patients admitted to hospital for an exacerbation 12-month, prospective, observational COPD (admission, discharge and follow-up) N=39 GOLD, but scores not presented by severity stage. Admission and discharge: N=149 Followup: Selfadministered questionnaire EQ-5D NA Utility mean: -Admission: (0.366) -Discharge: (0.243) Follow-up: (0.313) VAS mean: -Admission: 27.9 (17.6) -Discharge:62.3 (18.8) Follow-up:51.0 (20.6) Parshall et 2008 UK COPD Exploratory, al. 122 correlational study COPD N=299 (total sample) N=127 (Exacerbations cohort) N=198 (any diary cards) GOLD, but scores not presented by severity stage. Interview SF-36 NA General Health domain: -No symptom defined exacerbations *No exacerbations: 44.1 (19.3) *At least 1 exacerbation: 31.6 (21.1) -Health care defined exacerbations: *No exacerbation:41.5 (15.9) *Any exacerbation: 32.1 (21.7) Rutten-van Mölken et al UK and US COPD Patients from the UPLIFT trial COPD N=1,235 GOLD Interviewing and selfadministered questionnaire EQ-5D; SGRQ NA Baseline: VAS: 64.84(16.41) EQ-5D utility score: 0.76(0.21) SGRQ total score: 195

196 Author Pub. Year Country Population Information on recruitment Comparators Sample size Health states Method of elicitation Method of valuation Mapping Results (CI/SD) 0.76(0.21) By COPD severity -Severe VAS: 62.45( ) EQ-5D utility: 0.75( ) SGRQ total score: 46.51( ) -Very severe: VAS:57.84 ( ) EQ-5D utility: 0.647( ) SGRQ total score:57.31( ) Rutten-van Mölken et al Poland, France, South Africa, Spain, Hungary, Russia, UK, Canada, Austria, CH, Australia, Netherlands, Italy, Portugal COPD 1-year; randomised, double-blind; placebo controlled, multinational trial Roflumilast vs. placebo N=1505 GOLD Interview SGRQ NA SGRQ: Roflumilast: ( 17.61) Placebo: ( 18.15) Results also presented by specific SGRQ s scores Suresh et al France, Germany, Italy, Spain, UK and USA COPD Patients attending primary care physicians, patients treated by resp. specialist and general population COPD vs. General population N=2703 GOLD, but scores are not presented by severity stage. Selfadministered questionnaire EQ-5D NA Primary Care Physician (95% CI): *Overall mean: 0.70( ) *No exacerbations: 0.78( ) 1-2 exacerbations: 0.74( ) 3 or more exacerbations: 0.61( ) Resp. specialist (95% CI): *Overall mean: 0.68(

197 Author Pub. Year Country Population Information on recruitment Comparators Sample size Health states Method of elicitation Method of valuation Mapping Results (CI/SD) 0.69) *No exacerbations: 0.75( ) 1-2 exacerbations: 0.73( ) 3 or more exacerbations: 0.57( ) Non-UK Berit Bentsen et al Norway COPD Cross-sectional COPD N=100 SGRQ; HADS; ISWT Interview/ protocol SF-36 NA PCS: 38.4 ( 9.9) MCS: 48.6 ( 10.4) Bridevaux 2008 Switzerland COPD NR COPD patients et al. 125 with and without symptoms N=5000 GOLD (Normal lung function, Stage 1 and 2-4 stages) Interview/ protocol SF-36 NA Scores are presented by GOLD stage for PCS and MCS. Carrasco et al Spain COPD Multicenter, epidemiological, observational and descriptive in primary care. COPD N=10711 SEPAR (Mild, moderate, severe) Interview / protocol SF-12 NA PCS: 36.0 ( 9.9) MCS: 48.3 ( 10.9) Conte et al Italy Patients with bronchial obstruction attributable to COPD Multicenter study COPD, two groups based on SGRQ and MDA. N=531 Geriatric MDA, MMSE and 15-item GDS Interview/ protocol SGRQ NA Results presented in tertiles of the SGRQ. Delgado et al Brazil COPD patients Patients consecutively treated at the outpatient services of a resp. clinic. COPD N=42 GOLD, but scores are not presented by severity stage. Interview SF36; SGRQ NA SF-36: -PCS: ( 11.19) -MCS: ( 15.65) SGRQ:54.00 ( 22.7) Results for both instruments are also presented by each of the components / domains Esteban et 2006 Spain COPD Prospective al. 128 longitudinal clinical COPD N=611 HADO (Mild, moderate, Interview SF-36; SGRQ; NA Scores presented by component and by HADO 197

198 Author Pub. Year Country Population Information on recruitment Comparators Sample size Health states Method of elicitation Method of valuation Mapping Results (CI/SD) study severe) CRQ stage. Esteban et al. Jan 2009 Spain COPD Prospective cohort study COPD N=611 GOLD; BTS; ATS Interview SF-36; SGRQ NA Scores presented for the SF-36 and the SRGQ by severity scale and by specific domains of the QOL questionnaires Esteban et al. 129 Feb 2009 Spain COPD Prospective COPD, 3 groups based on the no. of hospitalisations (5 years) N=611 GOLD Interview SF-36; SGRQ NA Scores are presented for the SF-36 and the SRGQ by no. of hospitalisations (0, 1-2, 3) Esteban et al Spain COPD Prospective COPD, 3 groups based on level of physical activity (Low, moderate and high) N=611 GOLD Interview SF-36; SGRQ; CRQ NA Scores are presented for the SF-36; SGRQ and CRQ by level of physical activity. Garcia-Rio 2010 Spain COPD vs. et al. 131 Non-COPD Multicentre, crosssectional, population based, observational COPD vs. Non- COPD COPD=324 Non-COPD=110 GOLD; BODE index Interview/ protocol EQ-5D; SGRQ NA EQ-5D: -COPD: 75 (60-85) -Non-COPD: 85 ( ) SGRQ: -COPD: 16.7 ( ) -Non-COPD: 1.3 ( ) Hajiro et al Japan COPD Cross-sectional study COPD, level of dyspnea and disease severity (FEV1) N=194 FEV1 Interview SF-36; SGRQ NA Scores are presented for SF-36 and SGRQ by level of dyspnea and disease severity and also and by specific domains of the QOL questionnaires Katsura et 2007 Japan COPD NR COPD, males vs. al. 133 females groups N=156 FEV1; FVC Interview SF-36; SGRQ NA Scores presented for SF-36 and SGRQ by gender. Martin et 2008 Spain COPD Cross-sectional al. 134 survey COPD N=9405 Mild, moderate and severe Interview SF-12 NA PCS: 36.8 ( 10.4) MCS: 47.2 ( 11.2) Menn et 2010 Germany COPD stages al. 135 III and IV Prospective, observational COPD, admission vs. discharge N=117 GOLD Selfadministered questionnaire EQ-5D; SF- 12; SGRQ NA Scores presented for the 3 different methods of valuation by time of admission and discharge; and by stage 198

199 Author Pub. Year Country Population Information on recruitment Comparators Sample size Health states Method of elicitation Method of valuation Mapping Results (CI/SD) Miravitlles 2006 Spain COPD patients Observational, crosssectional et al. 136 survey Groups acc. to no. of exacerbations during past year N=1057 GOLD (stages II, III and IV) Self administered questionnaire SGRQ NA Results presented by domain of SGRQ & by no. of exacerbations. Miravitlles 2009 Spain COPD Cross-sectional et al. 137 multicentre COPD N=827 GOLD Interview EuroQOL NA EuroQOL score: 0.64 ( 0.23) Monninkhof 2004 Netherlands COPD Randomized Self-management et al. 138 vs. usual care N=248 Moderate and severe patients Interview EQ-5D; SGRQ NA Results presented for each questionnaire in three different times (baseline, 6 month and 12 months) Oba Y USA COPD Utilities used in CEA Placebo Salmeterol Fluticasone Salmeterolfluticasone in combination NA / literature review Moderate to severe patients Literature review SGRQ mapping EQ5D Index = ( ) SGRQ Score (correlation coefficient, ) Utilities are reported by publication and by FEV1 values. Omata et al Japan COPD Consecutive patients who initially consulted O/P resp. clinic COPD N=63 Levels of FEV1 Interview SGRQ; SF- 36; VAS-8 NA PCS: 44.8 ( 1.14) MCS 47.1 ( 1.44) Results by domain are also presented in the paper. Rodriguez et al Spain COPD Epidemiological, observational, crosssectional and descriptive COPD N=1596 (moderate) N=2012 (very severe) GOLD Interviewing and self administered questionnaire LCADL; Item 4 of SF-36; EQ- 5D, VAS NA Results presented for the different instruments by severity stage and by domains. Rutten-van Mölken et al Netherlands COPD Empirical valuation of health profiles of chronic diseases 16 COPD health profiles, combining Severity with exacerbation profile N=239 GOLD Interviewing and self administered questionnaire SGRQ; EQ-5D NA TTO: Absolute utility reduction for a communitytreated exacerbation: (0.007); Absolute utility reduction for a hospitaltreated exacerbation: (0.009) Stahl et 2005 Sweden COPD Survey COPD by severity al. 119 stage N=168 GOLD; BTS Survey SF-36; SGRQ; EQ-5D NA Results presented by severity criteria, severity stage and by instrument Voll- Aanerud et al Norway COPD General population sample COPD by resp. symptoms N=2306 ATS Interview SF-12 NA Results presented by score (PCS; MCS); resp. symptoms; no. of resp. symptoms; non-copd 199

200 Author Pub. Year Country Population Information on recruitment Comparators Sample size Health states Method of elicitation Method of valuation Mapping Results (CI/SD) patients; and COPD stage Abbreviations: SGRQ, St George s Resp. Questionnaire; WHOQOL, World Health Organization Quality of Life; ISWT, Incremental Shuttle Walking Test; HADS-A, Hospital Anxiety and Depression Scale; SF-36, Short Form 36 Health Survey; PCS, Physical component summary; MCS, Mental component summary; VAS-8, Visual Analogue Scale-8; GOLD, Global Initiative for Chronic Obstructive Lung Disease; BTS, British Thoracic Society; ATS, American Thoracic Society; SEPAR, Spanish Society of Pneumology and Chest Surgery; MDA, Multidimensional Assessment; MMSE, Mini Mental Status Examination; GDS, Geriatric Depression Scale; HADO, Health Activity Dyspnea Obstruction; CRQ, Chronic Resp. Questionnaire; LCADL, London Chest Activity of Daily Living Scale; NR, Not reported; NA, Not Applicable. 200

201 6.4.7 Please highlight any key differences between the values derived from the literature search and those reported in or mapped from the clinical trials. Please refer to Table B 52 in Section Adverse events Please describe how adverse events have an impact on HRQL. Any AEs experienced during treatment according to the labelled indication are expected to be mild to moderate and transient in nature, with no impact on HRQL. Quality-of-life data used in cost-effectiveness analysis Please summarise the values you have chosen for your costeffectiveness analysis in the following table, referencing values obtained in sections to Justify the choice of utility values, giving consideration to the reference case. Two types of utility inputs exist within the model: Utility associated with each COPD Markov health state. Utility decrements for community-treated and hospital-treated exacerbations, used to estimate QALY loss due to exacerbations. The utilities for the disease states (S 1 and S 2, VS 1 and VS 2 ) are taken from the LABA subgroup of the pooled analysis of studies M2-124 and M Within studies M2-124 and M2-125, the EQ-5D was distributed at each study visit and patients asked to assess their health status on that specific day. As exacerbations may have occurred between visits, utility decrements for these exacerbations were not collected. Utility decrements for COPD exacerbations (community-treated and hospital-treated) were obtained from a Dutch preference study reported by Rutten-van Mölken et al. 26 This study evaluated the health related quality of life values for COPD using health profiles that included both COPD severity and COPD exacerbations. The severity of exacerbation was classified according to four domains covering respiratory symptoms, non-respiratory symptoms, response to treatment and impact on daily activities. In addition to these domains the duration of a non-serious exacerbation 201

202 was defined as 10 days with a serious exacerbation lasting 28 days (see Table B 51). The utility valuations were undertaken by adults from the general public using a time trade off method. For exacerbations the absolute utility decrement lasting for one year was for one non-serious exacerbation occurring within one year and for one serious exacerbation. Table B 51: Definition of exacerbations, from Rutten van Molken et al. 26 Respiratory symptoms Other symptoms Treatment Impact Non-serious exacerbation Patients experience a mild to moderate worsening of their symptoms, especially breathlessness and cough. This may or may not be accompanied by a change in sputum production. Some patients have fever and general malaise and suffer from increased fatigue and insomnia. Patients can sometimes treat the exacerbation themselves, for instance by increasing the use of their existing respiratory medication. However, their symptoms may be so severe that they will need to see their GP or a specialist, who will often prescribe extra medication. The symptoms interfere with daily activities. Serious exacerbation Patients experience a severe to very severe worsening of their symptoms, especially breathlessness and cough. This may or may not be accompanied by a change in sputum production. Many patients have fever and suffer from severe fatigue and insomnia. Anxiety, confusion and impaired consciousness may also be present. Patients feel very ill. Despite treatment, the symptoms worsen to the point where hospitalization is required. Oxygen therapy and ventilatory support may be necessary during hospitalization. Symptoms may be life-threatening in some cases. Daily activities are completely disrupted by the symptoms In the model the absolute utility decrements are applied at the time of the event, assuming the patients live long enough to experience the full decrement. The absolute utility decrement applied for a community-treated exacerbation and for a hospital-treated exacerbation are presented in Table B 52. The use of different data sources and methods of health state utilities and utility decrements for exacerbations are explored in the sensitivity analysis. Oostenbrink et al. reported utilities of and for the Severe and Very severe states respectively. 27 This alternative approach applies the reduction of utility relative to the baseline utility for one month to estimate the QALY loss due to an exacerbation (15% for a community exacerbation and 50% for a hospital-treated exacerbation). 202

203 Table B 52: Utility Variables Utility Variable Value Source Utility for the severe COPD state Studies M2-124 and M2-125 pooled analysis of the Utility for the very severe COPD state LABA subgroup 41 Absolute utility reduction for a communitytreated exacerbation Absolute utility reduction for a hospitaltreated exacerbation Rutten-van Mölken et al These reductions of utility due to exacerbations are estimated to last for one year, and in application to the economic model, the loss of utility was accounted for within model cycle (see Table B44) If clinical experts assessed the applicability of values available or See Section 6.2. estimated any values, please provide the following details: Please define what a patient experiences in the health states in terms of HRQL. Is it constant or does it cover potential variances? HRQL worsens with disease progression (i.e. from the Severe COPD state to Very severe COPD). Baseline HRQOL is affected by sporadic periods of exacerbation which cause a short term losses in HRQOL Were any health effects identified in the literature or clinical trials excluded from the analysis? If so, why were they excluded? None were excluded If appropriate, what was the baseline quality of life assumed in the analysis if different from health states? Were quality-of-life events taken from this baseline? The quality of life values in the Severe COPD and Very Severe states were derived from the studies M2-124 and M It was assumed that values presented in Table B47 apply to both S 1 and S 2, and the VS 1 and VS 2 health states equally. Events in the model which affect quality of life are community-treated exacerbations and hospital-treated exacerbations. It is assumed that these quality-of-life events affect health related utility of life for one model cycle, after which heath related utility returns to the level characterising stable COPD state. 203

204 Please clarify whether HRQL is assumed to be constant over time. If not, provide details of how HRQL changes with time. Within stable COPD states (with no exacerbations), health related utility of life is assumed constant Have the values in sections to been amended? If so, please describe how and why they have been altered and the methodology. Health state utilities were taken from the pivotal trials and thus not amended. The base case analysis did not include health utilities associated with adverse events as these were assumed to be mild to moderate and transient in nature.. 204

205 6.5 Resource identification, measurement and valuation This section should be read in conjunction with NICE s Guide to the methods of technology appraisal, section 5.5. All parameters used to estimate cost effectiveness should be presented clearly in a table and include details of data sources. For continuous variables, mean values should be presented and used in the analyses. For all variables, measures of precision should be detailed. NHS costs Please describe how the clinical management of the condition is currently costed in the NHS in terms of reference costs and the payment by results (PbR) tariff. Provide the relevant Healthcare Resource Groups (HRG) and PbR codes and justify their selection. Please consider in reference to section 2. Costs used within the model reflect the UK NHS perspective and are represented by the following categories: 1) cost of COPD regimens; 2) cost of maintenance of patients with COPD; and 3) cost of exacerbations (community-treated and hospital-treated). Cost of COPD regimens The cost of a COPD regimen is defined by the monthly cost of roflumilast and COPD drugs that are used in combination with roflumilast, or those used in the comparator regimen. Roflumilast is prescribed as one tablet per day, with a 30-tablet pack having an acquisition cost of This implies a daily cost of 37.71/30 and one-cycle (one month) cost of (daily cost x (365/12)). The 90-tablet pack of roflumilast is associated with the same daily and monthly cost as the 30-tablet pack. A similar approach is used to define the per cycle cost of all other COPD treatments assessed within the model, as shown in Table B 53. Please refer to Section for a presentation of the cost associated with each regimen. 205

206 Table B 53 : Cost of COPD drugs Treatment class LABA LAMA LABA/ICS Drug Dose Formulation Cost per model cycle (month) Salmeterol 50 Evohaler: 25 mcg per dose, ( 29.26/30)*(365/12) (Serevent micrograms 120-dose unit = = (1) Evohaler, twice daily Accuhaler: 50 mcg per Accuhaler, dose, 60-dose unit = and Diskhaler ) Diskhaler: 50 mcg per disk, 60-disk pack, with device = 35.79; 60-disk refill pack = Tiotropium (Spiriva ) Salmeterol/ fluticasone (Seretide 500 Accuhaler) 18 micrograms once daily 1 dose of Seretide 500, twice daily cap refill pack (18 mcg per cap) = cap pack (18 mcg per cap) with HandiHaler device = mcg salmeterol (as xinafoate) mcg fluticasone per puff, 60- dose Accuhaler = ( 31.89/30)*(365/12) = (2) ( 40.92/30)*(365/12) = Source of drug costs: BNF 61, 2011 (1) Calculated cost for LABA assumes all items are prescribed as the Accuhaler or Evohaler, which are associated with a lower daily treatment cost than the Diskhaler (or refills) (2) Calculated cost for LAMA assumes all items are prescribed as refill packs, which are associated with a lower daily treatment cost than when combined with the Handihaler device Maintenance cost for patients with COPD The cost of maintaining a patient with COPD is assumed to vary by COPD severity, and is based on the following assumptions of unit costs (Table B 54 and Table B 55). Table B 54: Cost of maintenance - Disease state: Severe COPD Cost category Out patient visit respiratory physician Unit cost Resource use Cost per model cycle (month) Twice a year Spirometry Twice a year 8.33 Influenza vaccination % of patients annual vaccination 0.26 Reference National schedule of reference costs 2009/10. Consultant-led face to face attendance/respiratory medicine. Service code 340. National Schedule of Reference Costs Year : ' ' - NHS Trusts and PCTs combined Direct Access: Diagnostic Services/ Spirometry Test and Broncho Dilator Response Test. DA07. BNF: Injection, suspension of propiolactone-inactivated influenza virus (surface antigen, grown in fertilised hens' eggs), net price 0.5-mL prefilled syringe = Oxygen days Cost per day of Euro has been 206

207 Cost category therapy Unit cost Resource use Cost per model cycle (month) Monthly cost Reference inflated from 2001 to 2010 values and converted to GBP. * Estimates of resource use are based on the publication by Oostenbrink et al. 27 This study was conducted in the Netherlands and Canada, and resource utilisation was assumed to be applicable to the UK. Table B 55: Cost of maintenance - Disease state: Very Severe COPD Cost category Out patient visit respiratory physician Unit cost Spirometry Influenza vaccination Oxygen therapy 4.15 Resource use Four times a year Four times a year 75% of patients annual vaccination Cost per model cycle (month) days Monthly cost Reference National schedule of reference costs 2009/10. Consultant-led face to face attendance/respiratory medicine. Service code 340. National Schedule of Reference Costs Year : ' ' - NHS Trusts and PCTs combined Direct Access: Diagnostic Services/ Spirometry Test and Broncho Dilator Response Test. DA07. BNF: Injection, suspension of propiolactone-inactivated influenza virus (surface antigen, grown in fertilised hens' eggs), net price 0.5-mL prefilled syringe = Cost per day of Euro has been inflated from 2001 to 2010 values and converted to GBP. * Estimates of resource use are based on the publication by Oostenbrink et al. 27 This study was conducted in the Netherlands and Canada, and resource utilisation was assumed to be applicable to the UK. 207

208 Cost of exacerbations Exacerbation costs are stratified into moderate (community-treated) and severe (hospital-treated) exacerbations (Table B 56 and Tables B57-B58). Table B 56: Cost of community-treated exacerbation Cost category GP (consultation lasting 11.7 minutes) Unit cost Resource use Proportion Cost per model cycle (month) Reference / PSSRU 2010 A&E (no admission) / National Schedule of Reference Costs 2009/10 Prednisolone 30 mg / BNF Prednisolone 30 mg days 1/ BNF Co-amoxiclav (500 mg) days 1/ BNF Co-amoxiclav (500 mg) days 1/ BNF Total cost per exacerbation

209 Cost of hospital-treated exacerbation Average cost per hospital-treated exacerbation was estimated based on relevant National Reference Costs for COPD non-elective admissions, weighted by the activity volume, with the assumption that 90% of patients would be delivered to hospital by an ambulance. Table B 57: National Schedule of Reference Costs Year : '2009/10' - NHS Trusts Non-Elective Inpatient (Long Stay) HRG Data Currency Code Currency Description Activity National Average Unit Cost DZ21A Chronic Obstructive Pulmonary Disease or Bronchitis with length of stay 1 day or less discharged 8, home DZ21B Chronic Obstructive Pulmonary Disease or Bronchitis with Intubation with Major CC 101 2,306 DZ21C Chronic Obstructive Pulmonary Disease or Bronchitis with Intubation with CC 19 2,171 DZ21D Chronic Obstructive Pulmonary Disease or Bronchitis with Intubation without CC 4 1,725 DZ21E Chronic Obstructive Pulmonary Disease or Bronchitis with NIV without Intubation with Major CC 3,360 2,280 DZ21F Chronic Obstructive Pulmonary Disease or Bronchitis with NIV without Intubation with CC 993 2,099 DZ21G Chronic Obstructive Pulmonary Disease or Bronchitis with NIV without Intubation without CC 218 1,724 DZ21H Chronic Obstructive Pulmonary Disease or Bronchitis without NIV without Intubation with Major 34,530 2,029 CC DZ21J Chronic Obstructive Pulmonary Disease or Bronchitis without NIV without Intubation with CC 46,986 1,643 DZ21K Chronic Obstructive Pulmonary Disease or Bronchitis without NIV without Intubation without CC 11,721 1,279 Weighted average 1,

210 Table B 58: National Schedule of Reference Costs Year : '2009/10' - NHS Trusts Non-Elective Inpatient (Short Stay) HRG Data Currency Code Currency Description Activity National Average Unit Cost DZ21A Chronic Obstructive Pulmonary Disease or Bronchitis with length of stay 1 day or less discharged 61, home DZ21B Chronic Obstructive Pulmonary Disease or Bronchitis with Intubation with Major CC DZ21C Chronic Obstructive Pulmonary Disease or Bronchitis with Intubation with CC DZ21D Chronic Obstructive Pulmonary Disease or Bronchitis with Intubation without CC DZ21E Chronic Obstructive Pulmonary Disease or Bronchitis with NIV without Intubation with Major CC DZ21F Chronic Obstructive Pulmonary Disease or Bronchitis with NIV without Intubation with CC DZ21G Chronic Obstructive Pulmonary Disease or Bronchitis with NIV without Intubation without CC DZ21H Chronic Obstructive Pulmonary Disease or Bronchitis without NIV without Intubation with Major 4, CC DZ21J Chronic Obstructive Pulmonary Disease or Bronchitis without NIV without Intubation with CC 4, DZ21K Chronic Obstructive Pulmonary Disease or Bronchitis without NIV without Intubation without CC 1, Weighted average

211 The weighted average estimates from the Table B57 and Table B58 is 1, per hospitalisation (calculated according to the formula: [(sum_activity units x sum_national average unit cost) LONG STAY + (sum_activity units x sum_national average unit cost) SHORT STAY ] / [(sum_activity units) LONG STAY + (sum_activity units) SHORT STAY]. The cost of ambulance transportation was estimated at based on the National Schedule of Reference Costs Year: '2009/10' - NHS Trusts Paramedic Services: Category B / Amber, service code PS06B (06 Breathing Problems; Breathing Difficulty). Assuming that 90% of patients would be delivered to hospital by ambulance the average cost of hospital treated exacerbation is estimated at 1, This estimate is conservative compared to the 2,403 applied in the model described in NICE CG Please describe whether NHS reference costs or PbR tariffs are appropriate for costing the intervention being appraised. This model used NHS reference costs, as they provide relevant costs and volume that enable the estimation of a conservative weighed average, based on what the disease is costing the service rather than simply what is reimbursed. When compared to Tariff values, the Reference costs also allow for a greater level of granularity to be assessed. Resource identification, measurement and valuation studies Please provide a systematic search of relevant resource data for the UK. Include a search strategy and inclusion criteria, and consider published and unpublished studies. The search strategy used should be provided as in section 9.13, appendix 13. If the systematic search yields limited UK-specific data, the search strategy may be extended to capture data from non-uk sources. Please give the following details of included studies: country of study 211

212 date of study applicability to UK clinical practice cost valuations used in study costs for use in economic analysis technology costs. The flow diagram for the search of resource use data in the UK is shown in Figure 19 below. The search was carried out in each of Medline, Medline(R) In Process, Embase, EconLIT and NHS EED databases. The search strategies for the EconLIT database was the same as for the HRQL search as only search terms for COPD were used. The searches were limited to articles published in the last 10 years. Figure 19: Flow diagram for searches of cost and resource use studies Records from databases: Medline: n = 1016 Embase: n = 1588 EconLIT: n = 23 NHSEED: n= 97 N = 2724 Duplicates removed n = 727 Records screened based on title/abstract n = 1997 Studies excluded due to non-english language or being outside the scope of the decision problem n = 1925 Potentially relevant articles and abstracts assessed for eligibility n = 72 Articles meeting inclusion criteria n =

213 The inclusion criteria for studies were as follows: Studies must be related to the COPD population Studies related to resource use in a UK setting The exclusion criteria for studies were as follows: Studies related to patients with asthma or to COPD mixed with asthma Studies published prior to 1999 Studies related to non-pharmacological management of COPD e.g. studies of the cost-effectiveness of pulmonary rehabilitation. Table B 59 below includes details of included studies. Two included studies are not captured in the table as they were in abstract format and did not contain enough detail. 213

214 Table B 59: Details of studies in which costs and resource use are measured Author Title Citation Year Country Date* Cost valuation Resource use Measure Cost Resource Reported utilization UK specific O'Reilly et al. 121 Health status impairment and costs associated with COPD exacerbation managed in hospital Int J Clin Pract, July 2007, 61, 7, UK Not specified Mean direct cost of exacerbation (SD ) Mean bed cost NR - Price et al. 142 UK National COPD Audit 2003: Impact of hospital resources and organisation of care on patient outcome following admission for acute COPD exacerbation Thorax 2006;61: United 842. doi: Kingdom /thx NA NR - Data is presented as percentages of population attending the different types of consultation; by length of stay; by inpatient death. Wouters EFM 143 Economic analysis of the confronting COPD survey: an overview of results Respiratory Medicine (2003) 97, Supplement S3-S Canada, France, 2002 NR - UK Italy, Hospital visits 0.43 (SD 1.82) Netherlands, Spain, United Kingdom ER visits 0.33 (SD 1.81) Canada Hospital visits 0.32 (SD 1.12) ER visits 0.38 (SD 1.22) France Hospital visits 0.34 (SD 1.79) ER visits 0.15 (SD 0.55) Italy Hospital visits 0.33 (SD 2.16) ER visits 0.10 (SD 0.41) Netherlands 214

215 Author Title Citation Year Country Date* Cost valuation Resource use Measure Cost Resource Reported utilization Hospitalisations 0.27 (SD 1.65) ER visits 0.45 (SD 5.60) Spain Hospital visits (SD 5.28) ER visits 0.78 (SD 3.03) USA Hospital visits 0.38 (SD 1.78) ER visits 0.52 (SD 1.81) Non - UK specific Andersson et al. 144 The costs of exacerbations Respiratory Medicine in chronic obstructive Vol.96 (2002) pulmonary disease (COPD) 2002 Sweden 1999 Visit/contact physician SEK 1170 Health care visits for severe exacerbations Hospitalisation (respiratory) SEK Hospitalisation (general) SEK 3650 Health care visits for moderate exacerbation 1.56 Hospitalisation (intensive care unit) Hospitalisation (primary care centre) Transportation, per 10km Ambulance transport, per transport SEK 5440 SEK 3030 SEK 15 SEK 2479 Average length of stay in hospital for severe exacerbation 6.6 days Mild exacerbation SEK 36 (SD 264) 215

216 Author Title Citation Year Country Date* Cost valuation Resource use Measure Cost Resource Reported utilization Mild/Moderate exacerbation SEK 375 (SD194) Moderate exacerbation SEK 1483 (SD1235) Severe exacerbation SEK (SD16645) Borg et al. 145 A Computer Simulation Model of the Natural History and Economic Impact of Chronic Obstructive Pulmonary Disease Value in Health Vol. 7, No. 2, Sweden 1999 Severe exacerbation (health care) Severe exacerbation (societal) SEK (95% CI ) SEK (95% CI ) N/R - Chiang 146 Cost analysis of chronic obstructive pulmonary disease in a tertiary care setting in Taiwan Respirology (2008) 13, Taiwan 2002/ 2003 Mean total direct c. COPD management severe COPD New Taiwan $ (NTD) ± /8252±9320 ER visits/year severe COPD 1.73±2.44 Outpatient visit NTD 369 / No of patient hospitalizations/y severe 45 mean Hospitalization non ICU/day mean Hospitalisation (ICU/day) Hospitalisation severe NTD 4638±4818 / ± NTD 10601±2717 / ±77.63 NTD ±32464/ 7839±9276 mean ER/day NTD 5044±3666 ER visits/y severe spirometry NTD11417±15666 / 326±448 NTD360/10.29 Dalal et al. 147 Direct costs of chronic International Journal of 2010 USA 2008 Outpatient visit $305 (SD 310) Outpatient visit 2.7 (SD 2.9) 216

217 Author Title Citation Year Country Date* Cost valuation Resource use Measure Cost Resource Reported utilization obstructive pulmonary disease among managed patients Chronic Obstructive Pulmonary Disease 2010: Urgent outpatient visit $274 (SD336) Urgent outpatient visit 0.7(SD1.2) ED visit $327 (SD65) ED visit 0.08 (SD 0.6) ICU stay $9745 (SD2968) Standard admission 0.05 (SD0.3) Dal Negro 148 Costs of chronic obstructive pulmonary disease (COPD) in Italy: The SIRIO study (Social Impact of Respiratory Integrated Outcomes) Respiratory Medicine (2008) 102, Italy 2004 /1996 Mean annual c./patient baseline, stage 2 COPD Mean annual c./patient baseline, stage 3 COPD ± ± H. admission mean/patient at baseline H. admission mean/patient at 1y follow-up Day hospital mean/patient at baseline Day H. mean/patient at 1y follow-up Emergency mean/patient at baseline 7.62 Emergency mean/patient at 1y follow-up Visits to GP and specialist mean /patient baseline Visits to GP and specialist Mean /patient 1y follow-up

218 Author Title Citation Year Country Date* Cost valuation Resource use Measure Cost Resource Reported utilization De Miguel et. Al. 149 Determinants and predictors of the cost of COPD in primary care: A Spanish perspective International Journal of COPD 2008:3(4) Spain 2003 Total annual/patient Hospital admission ± ± 1, Mean no GP visits/year severe COPD Pneumologist visits mean/y severe COPD 9.54 (95%CI ) 2.32 (95%CI ) ER ER visits mean/y severe 2.88 (95%CI ) Primary care visit male/ female ( ) / ( ) H. admissions severe COPD 1.33 (95%CI ) Pneumologist visits M/F ( ) / ( ) Influenza vaccine M/F 0.20 ( ) / 0.23 (95%CI ) Earnshaw et al. 91 Cost-effectiveness of fluticasone propionate/salmeterol (500/50 mug) in the treatment of COPD Respiratory Medicine (2009) 103, United States 2006 Outpatient visit $95.50 NR - X-ray $35.24 ER visit $ Severe Exacerbation $8,858 (95%CI:2814, 14901) Moderate Exacerbation $ FitzGerald et al 150 Resource Use Study In COPD (RUSIC): A prospective study to quantify the effects of COPD exacerbations on health care resource use among COPD patients Can Respir J Vol 14 No 3 April Canada 2002 N/R Hospitalizations/patient/y in total pop. H. in the exacerbating group

219 Author Title Citation Year Country Date* Cost valuation Resource use Measure Cost Resource Reported utilization Detournay 151 The SCOPE Study: Health- Care Consumption Related to Patients with Chronic Obstructive Pulmonary Disease in France Value in Health, Vol 7, number 2, France 2001 Physician visits/year moderately severe COPD 138 Physician visits/year severe COPD 165 Av. No of exacerbations/y moderately severe COPD Av. No of exacerbations/y severe COPD Hospitalizations/ y. moderately severe COPD Hospitalizations/ y. severe COPD transportation / y. moderately severe COPD transportation / y. severe COPD 82 Friedman 152 Economic Burden of Chronic Obstructive Pulmonary Disease Impact of New Treatment Options Pharmacoeconomics 2001; 19 (3): USA 1999 Clinic visit $28 Clinic visits stage III COPD ER/visit $125 ER visit stage III ~4 Hospitalization/ day ICU $650 H admissions stage III ~3 ~5.9 H./day non-icu $375 Annual cost /patient Stage III COPD $10812 Friedman et al 153 Healthcare Costs with Tiotropium Plus Usual Care versus Usual Care Alone Following 1 Year of Treatment in Patients with Chronic Obstructive Pharmacoeconomics 2004; 22 (11): USA 1999 ER visit $ 217 Exacerbations/y tiotropium 0.76 Outpatient visit $ 42 Exacerbations/y placebo 0.95 Hospital day $ 1286 Mean ER annual c./patient tiotropium $

220 Author Title Citation Year Country Date* Cost valuation Resource use Measure Cost Resource Reported utilization Pulmonary Disorder (COPD) Mean ER annual c./patient placebo Mean annual c. outpatient physician /patient tiotropium Mean annual c. outpatient physician /patient placebo $ 18 $ 69 $ 67 Hilleman et al. 154 Pharmacoeconomic Evaluation of COPD Chest 2000;118; USA Not specified Clinic visit $ 28 Clinic visits Data presented for each of the 5 years follow up of patients in each of the COPD severity stages (I, II and III). ED visit $ 125 ER visits Hospitalisation (ICU/d) $ 650 Hospital visits Hospitalisation (Non- ICU/d) $ 375 Izquierdo- Economic Impact of Alonso et al 155 Pulmonary Drugs on Direct Costs of Stable Chronic Obstructive Pulmonary Disease COPD: J OF CHRONIC OBSTRUCTIVE PULMONARY DISEASE Vol. 1, No. 2, pp , 2004 Spain 2000 Outpatient clinic visit ER visit Emergency visits stage III Outpatient clinic visits stage III 1.11± ±3.34 H./day general ward Hospitalizations stage III 0.72±1.11 H/day ICU exacerbation treatment without hospitalization Emergency visits stage III Outpatient clinic visits stage III

221 Author Title Citation Year Country Date* Cost valuation Resource use Measure Cost Resource Reported utilization Hospitalizations stage III Jansson et al. 156 Costs of COPD in Sweden According to Disease Severity Chest 2002;122; Sweden 1999 Visit to primary care physician SEK 1170 Hospitalisation (No. of days) FEV1<40: 8.88 FEV140-59: FEV160-79: FEV1 80: Visit to hospital-based physician SEK 1170 Hospitalisation 24hrs (pulmonary medicine) SEK 2810 Outpatient care (no. of visits) FEV1<40: 5.46 FEV140-59: 1.03 FEV160-79: 0.80 FEV1 80:0.91 Hospitalisation (medicine dep) SEK 3650 Hospitalisation (intensive care) SEK 5440 Oxygen therapy (% of patients) FEV1<40: 20 FEV140-59: 3 FEV160-79: 0 FEV1 80:0 Hospitalisation (primary care centre) SEK 3030 Joo et al 157 Patterns of Healthcare Utilization by COPD Severity: A Pilot Study Lung (2008) 186: USA N/A N/R outpatient visits mean 14.4 (30.47) (SD)/year GOLD stage 0 outpatient visits mean (SD) GOLD stage I,II (13.53) outpatient visits mean (SD) GOLD stage III, IV (14.24) Koleva et al. 158 Healthcare costs of COPD in Italian referral centres: Respiratory Medicine (2007) 101, Italy 2004 Mean total c./year/patient GOLD Stage III COPD Spirometry GOLD III stage 3.0 ( ) 221

222 Author Title Citation Year Country Date* Cost valuation Resource use Measure Cost Resource Reported utilization A prospective study Mean total c./year/patient GOLD Stage IV COPD Spirometry GOLD IV stage 2.2 ( ) Specialist consultations stage III ( ) Pneumologic consultation GOLD III stage 4.3 ( ) Specialist consultations stage IV ( ) Pneumologic consultation GOLD IV stage 4.2 ( ) H. admissions stage III ( ) H. admissions GOLD III stage 0.4 ( ) H. admissions stage IV ( ) H. admissions GOLD IV stage 0.8 ( ) Miravitlles et al. 159 Cost of chronic bronchitis and COPD. A 1-yar followup study Chest 2003;123; Spain 2000 GP clinic visit $ 7.11 GP clinic visit 5.14 (SD 3.89) Chest physician ambulatory visit $ 20.1 Chest physician ambulatory visit 0.46 (SD0.85) Chest physician hospital visit $ Chest physician hospital visit 0.39 (SD0.93) ED visit $ ED visit 0.53 (SD1.24) Hospitalisation (per day) $ Hospitalisation (per day) 0.23 (SD 1.15) ICU (per day) $ ICU (per day) 0.02 (SD 0.13) Spirometry $ Spirometry 0.54 (SD 0.85) Exacerbation $ 58.7 Mittmann et al. 136 The cost of moderate and severe COPD exacerbations to the Canadian healthcare system Respiratory Medicine (2008) 102, Canada 2006 General practitioner CAD Outpatient visits 618 (N=639) Respiratory specialist CAD ER visits 245 (N=639) Other specialist CAD Hospitalisations 151 (N=639) 222

223 Author Title Citation Year Country Date* Cost valuation Resource use Measure Cost Resource Reported utilization Emergency physician CAD Length of hospital stay 10.0 days (SD 14.5) Spirometry CAD Length of ICU stay 6.2 days (SD 6.5) Ambulance CAD Moderate exacerbation Severe exacerbation ITT analysis: CAD 641 CE analysis: CAD 756 ITT analysis: CAD 9557 CE analysis: CAD 9953 Nielsen et al. 160 Present and future costs of COPD in Iceland and Norway: Results from the BOLD study ERJ Express. Published on April 8, Norway and Iceland 2005 GP visit Iceland: 46.6 Norway: 52.5 Specialist visit Iceland: 77.5 Norway: N/R - COPD hospitalisation Iceland: 5966 Norway: 3920 Nowak et. Al. 161 Epidemiology and Health Economics of COPD Across Europe Treat Respir Med 2005; 4 (6): Europe, multiple countries 2002 Av. direct c./exacerbation severe Sweden 2378 N/R Av. c. exacerbation episode France 62 Oostenbrink 162 Resource use and risk factors in high-cost exacerbations of COPD Respiratory Medicine (2004) 98, The Netherlands 2001 GP unit c./day/visit 18 GP visits mean c./ exacerbation severe 12 (SD 18) GP visits /exacerbation all patients GP visits/ severe exacerbation 0.66 (SD 1.03) 0.70 (SD 0.94) GP visits/ exacerbation 9 (SD 17) Resp. physician visits 0.39 (SD 0.73) /exacerbation all patients 223

224 Author Title Citation Year Country Date* Cost valuation Resource use Measure Cost Resource Reported utilization Resp. physician c./day/visit 52 Resp. physician visits/ severe exacerbation 0.82 (SD 1.32) Resp. physician/ exacerbation severe 43 (SD 69) Resp. physician visits /moderate exacerbation 0.37 (SD 0.61) ER c./day/visit 98 ER/ severe exacerbation 0.25 (SD 0.44) ER/exacerbation severe 25 (SD 43) Ambulance/exacerbation severe 77 (189) ER/moderate exacerbation ambulance / severe exacerbation 0.05 (SD 0.22) 0.31 (0.75) Piperno et al. 163 The burden of COPD in France: results from the Confronting COPD survey Respiratory Medicine Vol.97 (2003) (SUPPLEMENTC ), S France N/R Primary care visit 18 Primary care practitioner visits Scheduled: 3.79 (SD7.81) Unscheduled: 1.06 (SD3.92) Specialist visit 23 Specialist visits Scheduled: 0.69 (SD2.21) Hospitalisation 351 Emergency room 23 Unscheduled: 0.27 (SD2.12) Rutten-Van et al. 105 Modelling the 5-year cost effectiveness of tiotropium, salmeterol and ipratropium for the treatment of chronic obstructive pulmonary disease in Spain Pharmacoeconomics 2007; 25 (8): Spain 2004 Hospitalisation (medical ward) Hospitalisation (ICU ward) Family practitioner office visit NR Pulmonologist ER visit w/hospitalisation 145 ER preceding an admission

225 Author Title Citation Year Country Date* Cost valuation Resource use Measure Cost Resource Reported utilization Patient transport (ambulance) Paramedic unit (ambulance) Shaya et al. 164 Utilization of Health Care Resources in a High-Risk Medicaid Population with Chronic Obstructive Pulmonary Disease P&T May 2006 Vol. 31 No USA NA NR - Inpatient claims (per year) Outpatient and office visit claims(per year) Emergency claims(per year) Schermer et al. 165 Exacerbations and associated healthcare cost in patients with COPD in general practice Monaldi Arch Chest Dis (2006), 65, 3, Netherlands 2001 ER visit 161 GP visits 381 (N=286) Hospitalisation 2237 Respiratory consultant 25 (N=286) ER visits 1 (N=286) Hospitalisations 1 (N=286) Ward et al. 166 Direct medical cost of chronic obstructive pulmonary disease in the U.S.A. Respiratory Medicine (2000) 94, USA 1994 Average per-visit cost for a COPD-related outpatient physician visit $ 25 Specialist visits (per year) Severe COPD: 6.7 * Indicates the date of the costs reported in the paper. Abbreviations: NR, Not reported; NA, Not applicable. Average per-visit cost for a COPD-related outpatient physician visit $

226 6.5.4 If clinical experts assessed the applicability of values available or estimated any values, please provide the following details: o o o o o o o o o the criteria for selecting the experts the number of experts approached the number of experts who participated declaration of potential conflict(s) of interest from each expert or medical speciality whose opinion was sought the background information provided and its consistency with the totality of the evidence provided in the submission the method used to collect the opinions the medium used to collect opinions (for example, was information gathered by direct interview, telephone interview or self-administered questionnaire?) the questions asked whether iteration was used in the collation of opinions and if so, how it was used (for example, the Delphi technique). Not applicable. Intervention and comparators costs Please summarise the cost of each treatment in the following table. Cross-reference to other sections of the submission; for example, drugs costs should be cross-referenced to sections 1.10 and Provide a rationale for the choice of values used in the cost-effectiveness model discussed in section Table B 60 shows a summary for costs for each treatment comparator strategy (see Table B 53 and Section 6.5.1). The cost of COPD regimens are based on the latest available BNF prices. It is assumed that the COPD regimen are prescribed at their recommended doses, therefore no confidence intervals are indicated in the table. According to product labels, the doses of COPD regimens are the same for patients in the Severe COPD and Very severe COPD health states. 226

227 The monthly cost of patient care in the roflumilast-based treatment strategies can be estimated as the monthly cost of COPD treatment regimens (Scenario-dependent), plus the monthly maintenance cost for a COPD health state. Contribution of the cost of exacerbations to the average cost of treatment depends on the rate of exacerbations, which is dependent on the Scenario assessed, the disease state, and the proportion of hospital-treated exacerbation among all exacerbations. Table B 60: Monthly drug costs associated with the COPD therapies # Treatment Costs 1 LABA LAMA LABA + roflumilast LAMA + roflumilast LAMA + LABA LABA/ICS LAMA + LABA + roflumilast LABA/ICS + roflumilast LAMA + LABA/ICS LAMA + LABA/ICS + roflumilast Health-state costs Please summarise, if appropriate, the costs included in each health state. Cross-reference to other sections of the submission for the resource costs. Provide a rationale for the choice of values used in the costeffectiveness model. The health states should refer to the states in section

228 Table B 61: List of health states and associated costs in the economic model Cost of maintenance in COPD health state Cost of patients monthly maintenance in the Severe COPD state Reference Section (Table B 54) Cost of patients monthly maintenance in the Very Severe COPD state Section (Table B 55) Cost of exacerbations Cost of moderate exacerbations (communitytreated) Cost of severe exacerbations (hospital-treated) , Section (Table B 56) Section (Table B 57 & Table B 58) Adverse-event costs Please summarise the costs for each adverse event listed in section 5.9 (Adverse events). These should include the costs of therapies identified in section 2.7. Cross-reference to other sections of the submission for the resource costs. Provide a rationale for the choice of values used in the cost-effectiveness model discussed in section It is assumed that adverse events associated with roflumilast treatment are mild or moderate and transient in nature and do not have a significant impact on costs or utility. Miscellaneous costs Please describe any additional costs that have not been covered anywhere else (for example, PSS costs). If none, please state. There are no additional costs in the model. 228

229 6.6 Sensitivity analysis This section should be read in conjunction with NICE s Guide to the methods of technology appraisal, sections , 5.8, and to Sensitivity analysis should be used to explore uncertainty around the structural assumptions used in the analysis. Analysis of a representative range of plausible scenarios should be presented and each alternative analysis should present separate results. The uncertainty around the appropriate selection of data sources should be dealt with through sensitivity analysis. This will include uncertainty about the choice of sources for parameter values. Such sources of uncertainty should be explored through sensitivity analyses, preferably using probabilistic methods of analysis. All inputs used in the analysis will be estimated with a degree of imprecision. Probabilistic sensitivity analysis (PSA) is preferred for translating the imprecision in all input variables into a measure of decision uncertainty in the cost effectiveness of the options being compared. For technologies whose final price/acquisition cost has not been confirmed, sensitivity analysis should be conducted over a plausible range of prices Has the uncertainty around structural assumptions been investigated? Provide details of how this was investigated, including a description of the alternative scenarios in the analysis. Fully incremental scenario analysis: patients switch to second line therapy upon disease progression In the base case, patients switch from a first line treatment regimen to a second line regimen, assuming an average time on the first line regimen of 12 months. In a oneway sensitivity analysis, a switch in treatment regimen following progression from severe to very severe COPD was explored. Comparisons of particular interest: scenario analyses A to F Several scenario analyses were performed on the comparisons of particular interest, i.e. LAMA + LABA/ICS + roflumilast vs. LAMA + LABA/ICS and LAMA + LABA + roflumilast vs. LAMA + LABA. The rationale for selecting these comparisons is explained in section The scenario analyses tested alternative assumptions for the model structure and base case assumptions, namely: 229

230 Scenario analysis A: The number of exacerbations per year in the Severe COPD state was set to 1 and 3 (vs. base case value of 2) for the common comparator (LAMA + LABA/ICS). Scenario analysis B: Incorporation of a lung function benefit, relative to the defined common comparator treatment, based on results from the LABA subgroup of the M2-124 & M2-125 clinical trials. 41 The duration of the lung function benefit was assumed to be 1 year. A lung function improvement would be expected to increase the time spent in a particular state, with resultant cost savings and QOL benefit. Scenario analysis C: Time on first line therapy was adjusted to 0.5 years, 1.5 years, 2 years and life time (base case = 1 year). Scenario analysis D: Estimates of disutilities of exacerbations based on the relative reduction of baseline utilities: 15% due to moderate exacerbation and by 50% due to severe exacerbation. 27 Scenario analysis E: Patients who start on first line LAMA + LABA/ICS + roflumilast (ICS-tolerant) and LAMA + LABA + roflumilast (ICS-intolerant) stay on this therapy for lifetime (base case: patients were assumed to receive a second-line treatment which was common to all comparator regimen). Scenario analysis F: Subgroup of patients start treatment in the Very severe COPD health state (see Section 6.9 for results of this analysis). 230

231 6.6.2 Which variables were subject to deterministic sensitivity analysis? How were they varied and what was the rationale for this? If any parameters or variables listed in section (Summary of selected values) were omitted from sensitivity analysis, please provide the rationale Table B 62: Parameters that were included in OWSA Variable Base case value Lower Upper Description/source Cost of community treated exacerbations % and -50% of the base case value Baseline annual exacerbation rates in comparator % and -20% of the base case value Age at start Arbitrary Cost of AE Half of the cost of a community treated exacerbation Cost of hospital treated exacerbations % and -50% of the base case value Cost of maintenance severe COPD % and -50% of the base case value Cost of maintenance very severe COPD % and -50% of the base case value Discount costs 3.50% 0 6 NICE Discount effect 3.50% 0 6 NICE Mean height (females) Assumption Mean height (males) Assumption Proportion of all exacerbations requring hospitalisation when in the severe state % and - 20% of the base case value, arbitrary Proportion of hospital exacerbations in very severe % and - 20% of the base case value, arbitrary CFR in hospital % and - 20% of the base case value, arbitrary RRR LAMA + LABA/ICS + Roflumilast vs LAMA + LABA/ICS % CI from Mills et al RRR LAMA + LABA vs LAMA + LABA/ICS % CI from Mills et al RRR LAMA + LABA + Roflumilast vs LAMA + LABA/ICS SMR in severe COPD % CI from Mills et al Low estimate corresponds to the risk of death in the general population, the upper estimate of 2.0 is arbitrary 231

232 Variable Base case value SMR in very severe COPD Utility loss in community exacerbation Lower Upper Description/source Utility loss in hosp exacerbation Assumption Utility loss associated with adverse events when treated with DAXAS Assumption U severe U very severe Low estimate corresponds to the SMR in severe COPD, the upper estimate of 2.0 is arbitrary 50% of the base case and twice the base case value, assumption Upper boundary assumption, lower boundary base case value for the very severe COPD Lower boundary assumption, upper boundary base case value for the severe COPD state 232

233 6.6.3 Was PSA undertaken? If not, why not? If it was, the distributions and their sources should be clearly stated if different from those in section 6.3.6, including the derivation and value of priors. If any parameters or variables were omitted from sensitivity analysis, please provide the rationale for the omission(s). An extensive set of PSA was undertaken around key treatment strategies (see Section 6.7.1). Parameters with high likely impact on the costeffectiveness and high uncertainty (e.g. relative rates of exacerbations; hospital case fatality rates) were included. Table B 63: List of distributions used in PSA for comparisons of interest Description Name in the model Distribution type Parameters/Info Coefficient to dist_rr_usev2versev Beta alpha = ((rr_usev2versev^2)*(1- estimate the rr_usev2versev)/(rr_usev2versev_se^2)), value of utility in beta = (rr_usev2versev*(1-rr_usev2versev)/(rr_usev2versev_se^2))- the very severe ((rr_usev2versev^2)*(1-rr_usev2versev)/(rr_usev2versev_se^2)); COPD state Expected value: relative to the sampled utility value in the severe COPD state Annual decline of lung volume, litres Monthly cost of maintenance of patients in the very severe COPD state Monthly cost of maintenance of patients in the very severe dist_fev1_decline_pa Gamma alpha = (FEV1_decline_pa_base^2)/(psa_FEV1_decline_pa_se^2), lambda = FEV1_decline_pa_base/(psa_FEV1_decline_pa_se^2); Expected value: dist_cost_maint_sev Gamma alpha = (cost_maint_sev_base^2)/(psa_cost_maint_sev_se^2), lambda = cost_maint_sev_base/(psa_cost_maint_sev_se^2); Expected value: dist_cost_maint_vsev Gamma alpha = (cost_maint_vsev_base^2)/(psa_cost_maint_vsev_se^2), lambda = cost_maint_vsev_base/(psa_cost_maint_vsev_se^2); Expected value: Comments Parameters are estimated from mean value (variable rr_usev2versev) and standard e (variable; rr_usev2versev_se) Determining the value of utility i the very severe state via the va of utility in the severe COPD sta ensures that the sampled utility the very sever state and the relative ratio with the value bounded between 1 and 0 is always lower than the utility in t severe COPD state. Based on Scanlon et al 118 Parameters are estimated from mean value (variable:cost_maint_sev_base and standard error (variable psa_cost_maint_sev_se estima from NHS Reference costs). Parameters are estimated from mean value (variable: cost_maint_vsev_base) and standard error (variable 233

234 Description Name in the model Distribution type COPD state Absolute reduction of health related utility (per year) due to communitytreated exacerbation Absolute reduction of health related utility (per year) due to hospitaltreated exacerbation Standardised mortality ratio in the severe state (excluding CFR) Ratio used to calculate SMR in very severe state Annual rate of exacerbation in patients in the severe COPD state Parameters/Info dist_ar_util_dec_comm Beta alpha = ((bilink(57)^2)*(1-bilink(57))/(psa_ar_util_dec_comm_se^2)), beta = (bilink(57)*(1-bilink(57))/(psa_ar_util_dec_comm_se^2))- ((bilink(57)^2)*(1-bilink(57))/(psa_ar_util_dec_comm_se^2)); Expected value: 0.01 dist_ar_util_dec_hosp Beta alpha = ((bilink(64)^2)*(1-bilink(64))/(psa_ar_util_dec_hosp_se^2)), beta = (bilink(64)*(1-bilink(64))/(psa_ar_util_dec_hosp_se^2))- ((bilink(64)^2)*(1-bilink(64))/(psa_ar_util_dec_hosp_se^2)); Expected value: dist_smr_sev Uniform Low Value = SMR_sev_base - 0.5, High Value = SMR_sev_base+0.5; Expected value: 1.5 dist_rr_smr_sev_2_vsev Uniform Low Value = 1, High Value = 3; Expected value: 2 dist_n_exac_sev_comp_pa Gamma alpha = (n_exac_sev_comp_pa_base^2)/(psa_n_exac_sev_comp_pa_se^2), lambda = n_exac_sev_comp_pa_base/(psa_n_exac_sev_comp_pa_se^2); Expected value: (LAMA + LABA/ICS) Comments psa_cost_maint_vsev_se estimated from NHS Reference costs). Rutten van Mölken 26 The value of (bilink(57) refers to the link between the Treeage model engine and the Excel fro end, and represents the mean value of the parameter of Variable psa_ar_util_dec_comm_se ref to the standard error. Rutten van Mölken 26 The value of (bilink(64) refers to the link between the Treeage model engine and the Excel fro end and equals to the mean val of the parameter of Variable psa_ar_util_dec_hosp_se refe to the standard error. Assumption that moderate COP in the ATS classification is comparable to severe COPD in GOLD classification system. Uniform distributions varying fro to 2 (assumption). Distribution of the RR is assum to be uniformly distributed betw 1 and 3. Source: SE take as the average across standard errors in other treatments within analysis 1 (IC tolerant) Variable n_exac_sev_comp_pa_base re to the deterministic value of this parameter and variable 234

235 Description Name in the model Distribution type Parameters/Info Comments the standard error. Annual rate of exacerbation in patients in the very severe COPD state dist_n_exac_vsev_comp_pa Gamma alpha = (n_exac_vsev_comp_pa_base^2)/(psa_n_exac_vsev_comp_pa_se^2), lambda = n_exac_vsev_comp_pa_base/(psa_n_exac_vsev_comp_pa_se^2); Expected value: (LAMA + LABA/ICS) Source: ratio of exacerbation fo very severe to severe is taken f pooled analysis for LABA subse the studies M2-124 and M2-125 Variable n_exac_vsev_comp_pa_base this parameter and variable the standard error. Cost of a community treated exacerbation dist_cost_ex_comm Gamma alpha = (bilink(16)^2)/(psa_cost_ex_comm_se^2), lambda = bilink(16)/(psa_cost_ex_comm_se^2); Expected value: UK NHS National reference cos Values of parameters of the distribution are derived from the mean value (dynamic link to th Cost of a hospital-treated exacerbation dist_cost_ex_hosp Gamma alpha = (bilink(17)^2)/(psa_cost_ex_hosp_se^2), lambda = bilink(17)/(psa_cost_ex_hosp_se^2); Expected value: UK NHS National reference cos Values of parameters of the distribution are derived from the Utility in the severe COPD state dist_u_sev Beta alpha = ((u_sev_base^2)*(1-u_sev_base)/(psa_u_sev_se^2)), beta = (u_sev_base*(1-u_sev_base)/(psa_u_sev_se^2))- ((u_sev_base^2)*(1-u_sev_base)/(psa_u_sev_se^2)); Expected value: Note that the utility in the severe COPD state is sampled independently and the utility in t very severe COPD state is Proportion of dist_pr_ex_hosp_sev Beta alpha = ((pr_ex_hosp_sev_base^2)*(1- Pooled analysis of the LABA 235

236 Description Name in the model Distribution type severe (hospital treated) exacerbations among all exacerbations in the severe COPD state Proportion of severe (hospital treated) exacerbations among all exacerbations in the very severe COPD state Probability of exacerbations in Tx10 (LAMA + LABA/ICS + roflumilast) in the severe COPD state Probability of exacerbations in Tx6 (LAMA + LABA + roflumilast) in the severe COPD state Parameters/Info pr_ex_hosp_sev_base)/(pr_ex_hosp_sev_se^2)), beta = (pr_ex_hosp_sev_base*(1- pr_ex_hosp_sev_base)/(pr_ex_hosp_sev_se^2))- ((pr_ex_hosp_sev_base^2)*(1- pr_ex_hosp_sev_base)/(pr_ex_hosp_sev_se^2)); Expected value: dist_pr_ex_hosp_vsev Beta alpha = ((pr_ex_hosp_vsev_base^2)*(1- pr_ex_hosp_vsev_base)/(pr_ex_hosp_vsev_se^2)), beta = (pr_ex_hosp_vsev_base*(1- pr_ex_hosp_vsev_base)/(pr_ex_hosp_vsev_se^2))- ((pr_ex_hosp_vsev_base^2)*(1- pr_ex_hosp_vsev_base)/(pr_ex_hosp_vsev_se^2)); Expected value: dist_p_ex_sev_comp_10 Beta alpha = (((n_ex_sev_fullincr_10/12)^2)*(1- (n_ex_sev_fullincr_10/12))/((n_ex_sev_fullincr_10_se/12)^2)), beta = ((n_ex_sev_fullincr_10/12)*(1- (n_ex_sev_fullincr_10/12))/((n_ex_sev_fullincr_10_se/12)^2))- (((n_ex_sev_fullincr_10/12)^2)*(1- (n_ex_sev_fullincr_10/12))/((n_ex_sev_fullincr_10_se/12)^2)); Expected value: dist_p_ex_sev_comp_6 Beta alpha = (((n_ex_sev_fullincr_06/12)^2)*(1- (n_ex_sev_fullincr_06/12))/((n_ex_sev_fullincr_06_se/12)^2)), beta = ((n_ex_sev_fullincr_06/12)*(1- (n_ex_sev_fullincr_06/12))/((n_ex_sev_fullincr_06_se/12)^2))- (((n_ex_sev_fullincr_06/12)^2)*(1- (n_ex_sev_fullincr_06/12))/((n_ex_sev_fullincr_06_se/12)^2)); Expected value: Comments subset of the studies M2-124 an M Parameters of the log-normal distribution are derived from the mean deterministic value of the relative ration of reduction of exacerbation rate (variable pr_ex_hosp_sev_base) and the standard error (variable pr_ex_hosp_sev_se). Pooled analysis of the LABA subset of the studies M2-124 an M Parameters of the log-normal distribution are derived from the mean deterministic value of the relative ration of reduction of exacerbation rate (variable pr_ex_hosp_vsev_base) and th standard error (variable pr_ex_hosp_vsev_se). Parameters are estimated from mean value (variable: n_ex_sev_fullincr_10) and standard error (variable: n_ex_sev_fullincr_10_se). This distribution is only active w comparison 2 (LAMA + LABA + roflumilast vs. LAMA + LABA) is assessed in ICS-intolerant patie Parameters are estimated from mean value (variable: n_ex_sev_fullincr_06) and standard error (variable: n_ex_sev_fullincr_06_se). 236

237 Description Name in the model Distribution type Parameters/Info Case fatality rate dist_death_hosp_cfr_symm Beta alpha = ((p_death_hosp_cfr^2)*(1-p_death_hosp_cfr)/(0.044^2)), due to hospitaltreated beta = (p_death_hosp_cfr*(1-p_death_hosp_cfr)/(0.044^2))- ((p_death_hosp_cfr^2)*(1-p_death_hosp_cfr)/(0.044^2)); exacerbation Expected value: Relative risk reduction of exacerbations in (LAMA + LABA/ICS + roflumilast) in the severe COPD state versus (LAMA + LABA/ICS) (Tx10 vs. Tx9) Relative risk reduction of exacerbations (LAMA + LABA + roflumilast) in the severe COPD state versus (LAMA + LABA) (Tx6 vs Tx5) distr_rr_ex_tx10_comp_sev Gamma alpha = (bilink(299)^2)/(bilink(300)^2), lambda = bilink(299)/(bilink(300)^2); Expected value: dist_rr_ex_tx6_comp_sev Gamma alpha = (bilink(113)^2)/(rr_ex_tx6_comp_sev_se^2), lambda = bilink(113)/(rr_ex_tx6_comp_sev_se^2); Expected value: Comments The beta distribution or the risk hospital death uses the estimat confidence interval from the dat on mortality given in the UK CO audit report 28. Values of parameters of the distribution are derived from the mean value (dynamic link to the Excel model front-end bilink(299 and the estimated standard erro (bilink(300)). The mean value of the relative r reduction is with standa error and standard error (95% CI (0.7372, ) ) were provided by Mills et al This parameter has been create specifically to run the tornado & PSA for (LAMA + LABA/ICS + roflumilast) versus (LAMA + LABA/ICS) for the analytical; scenario 1 (ICS tolerant).and th distribution is only active within scenario This distribution is only active w comparison 2 (LAMA + LABA + roflumilast vs. LAMA + LABA) is assessed in ICS-intolerant patie Values of parameters of the distribution are derived from the mean value (dynamic link to the Excel model front-end bilink(314 and the estimated standard erro (bilink(325)). Mean value (0.8354) was provid by Mills et al Standard er (0.06) was based from the 95% confidence interval for risk 237

238 Description Name in the model Distribution type Probability of exacerbations in the 1 st line treatment in LAMA+LABA/ICS (Tx9 in the analytical scenario 1 ICS Tolerant) Probability of exacerbations in the 1 st line treatment in LAMA+LABA (Tx5 in the analytical scenario 2 ICS Intolerant) Probability of exacerbations in 2 nd line treatment in severe COPD Parameters/Info dist_p_ex_sev_comp_9 Beta Alpha = (((n_ex_sev_fullincr_09/12)^2)*(1- (n_ex_sev_fullincr_09/12))/((n_ex_sev_fullincr_09_se/12)^2)) Beta = ((n_ex_sev_fullincr_09/12)*(1- (n_ex_sev_fullincr_09/12))/((n_ex_sev_fullincr_09_se/12)^2))- (((n_ex_sev_fullincr_09/12)^2)*(1- (n_ex_sev_fullincr_09/12))/((n_ex_sev_fullincr_09_se/12)^2)) dist_p_ex_sev_comp_5 Beta Alpha = (((n_ex_sev_fullincr_05/12)^2)*(1- (n_ex_sev_fullincr_05/12))/((n_ex_sev_fullincr_05_se/12)^2)) Beta = ((n_ex_sev_fullincr_05/12)*(1- (n_ex_sev_fullincr_05/12))/((n_ex_sev_fullincr_05_se/12)^2))- (((n_ex_sev_fullincr_05/12)^2)*(1- (n_ex_sev_fullincr_05/12))/((n_ex_sev_fullincr_05_se/12)^2)) d N_2nd_line_ALL Gamma Alpha = (bilink(336)^2)/(bilink(337)^2) Lambda = bilink(336)/(bilink(337)^2) Comments reduction LABA + LAMA + roflumilast) relative to LAMA+LABA/ICS available from Mills et al Mean value (0.166) was calcula from estimated rates of exacerbations of 2 per year (baseline treatment assumption 1 st line treatment with LAMA+LABA/ICS with standard error (0.0138) assumed to be th average across standard errors other treatments within analysis (ICS tolerant) This distribution has been creat for the PSA comparing Tx10 wi Tx9 in the analysis 1 (ICS tolera and is active for this analysis on Mean value (0.203) was calcula from estimated rates of exacerbations of per year (based on the annual rate of exacerbations of 2.0 in LAMA+LABA/ICS and the relati risk increase ratio of exacerbati in LABA+LAMA of ) in 1 s line treatment with LAMA+LABA/ICS with standard error estimated based on data provided by Mills et al This distribution has been creat for the PSA comparing Tx6 with Tx5 in the analysis 2 (ICS tolera and is active for this analysis on This probability describes the uncertainty in 2 nd line treatment for the PSAs comparing Tx6 vs Tx6 and Tx19 vs. Tx9. This 238

239 Description Name in the model Distribution type Parameters/Info The notation for the parameters of distributions in the table follows the respective notation in the modelling platform Treeage Pro Comments distribution is the same within e comparison pair, as it is assume that patients switch to the same line treatment with the same ris exacerbations. The mean value and the standa error for this distribution is provi to the Treeage model tree from Excel front end and depends on the selected analytical scenario For ICS tolerant: mean =0.166, standard error For ICS intolerant: mean 0.204; standard error

240 6.7 Results Provide details of the results of the analysis. In particular, results should include, but are not limited to, the following. Link between clinical- and cost-effectiveness results. Costs, QALYs and incremental cost per QALY. Disaggregated results such as LYG, costs associated with treatment, costs associated with adverse events, and costs associated with followup/subsequent treatment. A statement as to whether the results are based on a PSA. Cost-effectiveness acceptability curves, including a representation of the cost-effectiveness acceptability frontier. Scatter plots on cost-effectiveness quadrants. A tabulation of the mean results (costs, QALYs, ICERs), the probability that the treatment is cost effective at thresholds of 20,000 30,000 per QALY gained and the error probability. Clinical outcomes from the model For the outcomes highlighted in the decision problem (see section 4), please provide the corresponding outcomes from the model and compare them with clinically important outcomes such as those reported in clinical trials. Discuss reasons for any differences between modelled and observed results (for example, adjustment for cross-over). Please use the following table format for each comparator with relevant outcomes included. The primary outcomes in the decision problem include the rate of COPD exacerbations and the quality-adjusted life year (QALY). In the model, these outcomes are estimated over a lifetime horizon. Exacerbation data are taken from results of the additive MTC which provided estimates of the ratio in exacerbation rates of the 10 treatment regimens presented in this submission. Relevant efficacy data from the pooled analysis of the LABA subgroups in trials M2-124 and M2-125 were also used. Due to the concurrent nature of the MTC, the outcomes reported in this economic analysis are not directly comparable to any previously performed economic analysis. Treatment comparators Roflumilast is licensed for the treatment of severe and very severe COPD. Since such patients are commonly treated with multiple therapies in order to obtain the best possible patient outcomes, the exploration of the cost-effectiveness of roflumilast in this clinical context required the examination of multiple treatment scenarios. Within 240

241 the fully incremental analysis for ICS-tolerant patients, ten treatment options (Table B 64) were analysed in order to evaluate the cost-efficiency frontier. For patients in whom ICS is not tolerated/declined, six therapeutic options (Table B 65) were assessed. ICS-tolerant patients The results of the fully incremental analyses are shown in Table B 64 for ICS-tolerant patients. Table B 64: Results: Fully Incremental Analysis (ICS-tolerant patients, cohort starts in the Severe COPD state) # Treatment Cost Exacerbations 1 LY QALYs 1 LABA 22, LAMA 22, LABA + roflumilast 22, LAMA + roflumilast 22, LAMA + LABA 22, LABA/ICS 22, LAMA + LABA + roflumilast 23, LABA/ICS + roflumilast 22, LAMA + LABA/ICS 22, LAMA + LABA/ICS + roflumilast 23, Exacerbation rates not discounted. The cost-efficiency frontier analysis (Figure 20) suggests starting patients on LABA or LAMA. If patients continue to exacerbate, LAMA + LABA/ICS is a cost-effective second line option, followed by LAMA + LABA/ICS + roflumilast. The remaining options are either excluded due to direct dominance (e.g. being more costly and less effective than a comparator) or by extended dominance. Such results are fully in line with the treatment algorithm recommended in NICE CG101, with the use of roflumilast as an add-on to current stand of care (LAMA + LABA/ICS). It is also in line with the anticipated use of roflumilast in clinical practice, as believed by respected respiratory physicians consulted during the development of this submission, and a recent article by FitzGerald. 15;30 241

242 Figure 20: Cost-efficiency-frontier analysis (ICS-tolerant patients, cohort starts in the Severe COPD state) ICS-intolerant/declining patients For ICS-intolerant/declining patients, the results of the fully incremental analyses are listed in Table B 65. Table B 65: Results: Fully Incremental Analysis (ICS-intolerant/declining patients, cohort starts in the Severe COPD state) # Treatment Cost Exacerbations 1 LY QALYs 1 LABA 21, LAMA 21, LABA + roflumilast 21, LAMA + roflumilast 21, LAMA + LABA 21, LAMA + LABA + roflumilast 22, Exacerbation rates not discounted. The cost-efficiency frontier analysis (Figure 21) suggests starting ICSintolerant/declining patients on LABA or LAMA. If patients continue to exacerbate, LAMA + LABA is a cost-effective second line option, followed by LAMA + LABA + Roflumilast. LABA + Roflumilast and LAMA + Roflumilast are either excluded due to direct dominance or by extended dominance. 242

243 As with the results for the analysis of patients who are able to use ICS, these results are in line with the treatment algorithm recommended in NICE CG101. The guideline advocates the use of LAMA + LABA dual therapy. This analysis suggests that the use of roflumilast in addition to this standard of care is a cost-effective use of resources. It is also in line with the anticipated use of roflumilast in clinical practice, as believed by respected respiratory physicians, Dr Michael Rudolf and Professor Wisia Wedzicha. Figure 21: Cost-efficiency-frontier analysis (ICS-intolerant/declining patients, cohort starts in the Severe COPD state) Comparisons of interest Two comparisons of particular interest have been defined by the incremental analyses, and these will be assessed in further depth below. The rationale for this selection is firstly to focus on the comparisons that are likely to be most important for use of roflumilast in clinical practice in each patient group, and secondly, to provide the most clarity when presenting the results of the sensitivity analysis. 243

244 Table B 66: Key economic comparisons summary Key economic comparisons 1: LABA/ICS + LAMA + Roflumilast vs. LABA/ICS + LAMA 2: LABA + LAMA + Roflumilast vs. LABA + LAMA Patient population ICER /QALY gained ICS-tolerant patients 16, ICS-intolerant/ declining patients 13, Comparisons 1 and 2 are assessed in depth within this results section. In order to examine the uncertainty surrounding the point estimates, a range of sensitivity analyses were performed on key model parameters; both one-way deterministic and probabilistic analyses. Results of the key comparisons are presented in the following tables (Table B 67 and Table B 68). Table B 67: Summary of economic analysis: Comparison 1 (base case), LAMA + LABA/ICS + Roflumilast vs. LAMA + LABA/ICS LAMA + LABA/ICS + roflumilast LAMA + LABA/ICS Incremental Cost, 23,230 22, Total Exacerbations Hospital treated exacerbations LY QALY ICER, /QALY 16, ICER, /Exacerbation avoided 1, LAMA + LABA/ICS + Roflumilast minus LAMA + LABA/ICS Table B 68: Summary of economic analysis: Comparison 2 (base case), LAMA + LABA + Roflumilast vs. LAMA + LABA LAMA + LABA + roflumilast LAMA + LABA Incremental Cost, 22,222 21, Total Exacerbations Hospital treated exacerbations LY

245 LAMA + LABA + roflumilast LAMA + LABA Incremental QALY ICER, /QALY 13, ICER, /Exacerbation avoided 1, LAMA + LABA + Roflumilast minus LAMA + LABA Please provide (if appropriate) the proportion of the cohort in the health state over time (Markov trace) for each state, supplying one for each comparator. Table B 69: Markov trace: ICS-tolerant patients, LAMA + LABA/ICS + Roflumilast vs. LAMA + LABA/ICS Months LAMA + LABA/ICS + roflumilast LAMA + LABA/ICS S1 S2 VS1 VS2 Dead S1 S2 VS1 VS2 Dead

246 Figure 22: Markov trace: ICS-tolerant patients, LAMA + LABA/ICS + Roflumilast Figure 23: Markov trace: ICS-tolerant, LAMA + LABA/ICS 246

247 Table B 70: Markov trace: ICS-intolerant/declining patients, LAMA + LABA + Roflumilast vs. LAMA + LABA Months LAMA + LABA + roflumilast LAMA + LABA S1 S2 VS1 VS2 Dead S1 S2 VS1 VS2 Dead

248 Figure 24: Markov trace: ICS-intolerant/declining patients, LAMA + LABA + Roflumilast Figure 25: Markov trace: ICS-intolerant/declining patients, LAMA + LABA 248

249 6.7.3 Please provide details of how the model assumes QALYs accrued over time. For example, Markov traces can be used to demonstrate QALYs accrued in each health state over time. The total QALY in a cohort of patients starting in the Severe COPD state over a lifetime can be estimated as the areas under the curves describing the proportion of patients in the Severe COPD and Very severe COPD state. The loss of QALYs due to severe and moderate exacerbations is deducted. Tables are provided below (Table B 71 and Table B 72). Table B 71: QALY accrued over time: ICS-tolerant patients, LAMA + LABA/ICS + Roflumilast vs. LAMA + LABA/ICS Months LAMA + LABA/ICS + roflumilast LAMA + LABA/ICS

250 Table B 72: QALY accrued over time: ICS-intolerant/declining patients, LAMA + LABA + Roflumilast vs. LAMA + LABA Months LAMA + LABA + roflumilast LAMA + LABA

251 6.7.4 Please indicate the life years and QALYs accrued for each clinical outcome listed for each comparator. For outcomes that are a combination of other states, please present disaggregated results. Table B 73: Number of exacerbations, accrued over time: ICS-tolerant patients, LAMA + LABA/ICS + Roflumilast vs. LAMA + LABA/ICS Months LAMA + LABA/ICS + roflumilast LAMA + LABA/ICS

252 Table B 74: Number of exacerbations, accrued over time: ICSintolerant/declining patients, LAMA + LABA + Roflumilast vs. LAMA + LABA Months LAMA + LABA + roflumilast LAMA + LABA

253 6.7.5 Please provide details of the disaggregated incremental QALYs and costs by health state, and of resource use predicted by the model by category of cost. Suggested formats are presented below. Table B 75: Summary of QALY gain and number of exacerbations by health state: ICS-tolerant patients, LAMA + LABA/ICS + Roflumilast vs. LAMA + LABA/ICS Health state LAMA + LABA/ICS + Roflumilast QALYs LAMA + LABA/ICS LAMA + LABA/ICS + Roflumilast Exacerbations LAMA + LABA/ICS Severe COPD states (1st and 2nd line) 1st line only nd line only Very severe COPD states (1st and 2nd line) 1st line only nd line only Cohort total Table B 76: Summary of QALY gain and number of exacerbations by health state: ICS-intolerant/declining patients, LAMA + LABA + Roflumilast vs. LAMA + LABA Health state LAMA + LABA + Roflumilast QALYs LAMA + LABA LAMA + LABA + Roflumilast Exacerbations LAMA + LABA Incremental Incremental Incremental Incremental Severe COPD states (1st and 2nd line) 1st line only nd line only Very severe COPD states (1st and 2nd line) 1st line only nd line only Cohort total

254 Table B 77: Summary of costs by health state: ICS-tolerant patients, LAMA + LABA/ICS + Roflumilast vs. LAMA + LABA/ICS Health state Cost LAMA + LABA/ICS + Roflumilast Cost LAMA + LABA/ICS Absolute incremental % absolute incremental Severe COPD states (1st and 2nd line) 8,792 8, % 1st line only 1,895 1, % 2nd line only 6,897 6, % Very severe COPD states (1st and 2nd line) 14,438 14, % 1st line only % 2nd line only 14,000 13, % Cohort total 23,230 22, % Table B 78: Summary of costs by health state: ICS-intolerant/declining patients, LAMA + LABA + Roflumilast vs. LAMA + LABA Health state Cost LAMA + LABA + Roflumilast Cost LAMA + LABA Absolute incremental % absolute incremental Severe COPD states (1st and 2nd line) 8,520 8, % 1st line only 1,858 1, % 2nd line only 6,662 6, % Very severe COPD states (1st and 2nd line) 13,702 13, % 1st line only % 2nd line only 13,266 13, % Cohort total 22,222 21, % Table B 79: Cost breakdown: ICS-tolerant patients, LAMA + LABA/ICS + Roflumilast vs. LAMA + LABA/ICS LAMA + LABA/ICS + roflumilast LAMA + LABA/ICS Incremental Roflumilast (Daxas ) cost Maintenance cost (COPD drugs (except rofluminlast) plus resources) 16,896 16, Hospital treated exacerbation cost 4,875 4, Community treated exacerbation cost 1,046 1,

255 Table B 80: Cost breakdown: ICS-intolerant/declining patients, LAMA + LABA + Roflumilast vs. LAMA + LABA LAMA + LABA + roflumilast LAMA + LABA Incremental Roflumilast (Daxas ) cost Maintenance cost (COPD drugs (except rofluminlast) plus resources) 14,939 14, Hospital treated exacerbation cost 5,650 5, Community treated exacerbation cost 1,221 1, Figure 26: Cost breakdown: ICS-tolerant patients, LAMA + LABA/ICS + Roflumilast vs. LAMA + LABA/ICS 255

256 Figure 27: Cost breakdown: ICS-intolerant/declining patients, LAMA + LABA + Roflumilast vs. LAMA + LABA 256

257 Base-case analysis Please present your results in the following table. List interventions and comparator(s) from least to most expensive and present ICERs in comparison with baseline (usually standard care) and then incremental analysis ranking technologies in terms of dominance and extended dominance. For the results of the incremental analyses, with regards to ranking technologies in terms of dominance and extended dominance, please refer to section The strategies in the main analysis are represented by two key treatment-comparator pairs, namely: Comparison 1: LAMA + LABA/ICS + roflumilast vs. LAMA + LABA/ICS Comparison 2: LAMA + LABA + roflumilast vs. LAMA + LABA Table B 81: Summary base-case results Technologies Total costs ( ) Total LYs Total QALYs ICER ( ) Incremental (QALYs) Comparison 1 LAMA + LABA/ICS + Roflumilast vs. LAMA + LABA/ICS LAMA + LABA/ICS + Roflumilast 23, LAMA + LABA/ICS 22, ICER ( ) incremental (LYs) Incremental , , Comparison 2 LAMA + LABA + Roflumilast vs. LAMA + LABA LAMA + LABA + Roflumilast 22, LAMA + LABA 21, Incremental , , Sensitivity analyses Please present results of deterministic sensitivity analysis. Consider the use of tornado diagrams. Deterministic sensitivity analysis was performed separately for each treatmentcomparator pair, with analyses presented in the following tornado diagrams. 257

258 ICS-tolerant patients In the base case analysis, treatment with LAMA + LABA/ICS + roflumilast had an ICER of approximately 16,500/QALY gained vs. LAMA + LABA/ICS. Results for the sensitivity analysis of comparison 1 demonstrate cost-effectiveness results to be primarily sensitive to the relative rate of exacerbation between the LAMA + LABA/ICS + Roflumilast and LAMA + LABA/ICS treatment regimens (Figure 28). The confidence interval around the ratio confirms the statistical significance of the expected reduction. Sensitivity analysis to the boundaries of this confidence interval demonstrates the exponential nature of the relationship between the ICER and the ratio, with the ICER rising as the ratio approaches 1 (i.e. a non statistically significant reduction in exacerbation rates. (The numbers of exacerbations in the Severe COPD state in the common comparator, i.e. LAMA + LABA/ICS, was further tested separately and for a broader range in scenario analysis A, section 6.7.9). 258

259 Figure 28: Tornado diagram: LAMA + LABA/ICS + Roflumilast vs. LAMA + LABA/ICS 259

260 ICS-intolerant/declining patients In the base case analysis, treatment with LAMA + LABA + roflumilast was associated with an ICER close to 14,000/QALY gained vs. LAMA + LABA. Results for the sensitivity analysis of comparison 2 suggest cost-effectiveness results are, again, primarily sensitive to the ratio of the relative rates of exacerbation between treatments of interest, and the common comparator (Figure 29). Sensitivity analysis to the boundaries of the confidence interval around the statistically significant ratio demonstrates an acceptable range for the ICER. (The numbers of exacerbations in the Severe COPD state in the common comparator, was further tested separately and for a broader range in scenario analysis A, section 6.7.9). 260

261 Figure 29: Tornado diagram: LAMA + LABA + Roflumilast vs. LAMA + LABA 261

262 6.7.8 Please present the results of a PSA, and include scatter plots and costeffectiveness acceptability curves. PSA results: ICS-tolerant patients In this section PSA results are presented for each of the two treatment-comparator pairs of interest (Figure 30 to Figure 35). Figure 30: Cost-QALY scatterplot: ICS-tolerant patients, LAMA + LABA/ICS + Roflumilast vs. LAMA + LABA/ICS, 1000 runs Figure 31: Incremental CE scatterplot: ICS-tolerant patients, LAMA + LABA/ICS + Roflumilast vs. LAMA + LABA/ICS, 1,000 runs 262

263 Figure 32: Cost-effectiveness acceptability curve (CEAC): ICS-tolerant patients, LAMA + LABA/ICS + Roflumilast vs. LAMA + LABA/ICS, 1,000 runs PSA results: ICS-intolerant/declining patients Figure 33: Cost and QALY scatterplot: ICS-intolerant/declining patients, LAMA + LABA + Roflumilast vs. LAMA + LABA, 1,000 runs 263

264 Figure 34:Incremental CE scatterplot: ICS-intolerant/declining patients, LAMA + LABA + Roflumilast vs. LAMA + LABA, 1,000 runs Figure 35: Cost-effectiveness acceptability curve: ICS-intolerant/declining patients, LAMA + LABA + Roflumilast vs. LAMA + LABA, 1,000 runs 264

265 6.7.9 Please present the results of scenario analysis. Include details of structural sensitivity analysis. Fully incremental scenario analysis: patients switch to second line therapy upon discease progression As described in Section 6.6.1, a fully incremental scenario analysis was performed, which assumed that patients switch to second line therapy upon disease progression. Comparisons of particular interest: scenario analyses A to F Scenario analyses were performed on comparisons 1 and 2 for the following situations: Scenario analysis A: The number of exacerbations per year in the Severe COPD state was set to 1 and 3 (base case = 2) for the common comparator (LAMA + LABA/ICS). Scenario analysis B: Incorporation of a lung function benefit, relative to the common comparator treatment, based on results from the M2-124 / M2-125 clinical trial. 41 The duration of the lung function benefit was assumed to be 1 year. A lung function improvement would be expected to increase the time spent in a particular state, with resultant cost savings and QOL benefit. Scenario analysis C: Time on 1st line therapy was adjusted to 0.5 years, 1.5 years, 2 years and life time (base case = 1 year). Scenario analysis D: Estimates of disutilities of exacerbations based on the relative reduction of baseline utilities: 15% due to moderate exacerbation and by 50% due to severe exacerbation. 27 Scenario analysis E: Patients who start on first line LAMA + LABA/ICS + roflumilast (ICS-tolerant) and LAMA + LABA + roflumilast (ICS-intolerant) stay on this therapy for lifetime (base case: patients were assumed to receive a second-line treatment which was common to all comparator regimen). Scenario analysis F: Subgroup of patients start treatment in the Very severe COPD health state (see Sections 6.9 for results of this analysis). The results of the fully incremental analysis and those for scenario analyses A to E, are summarised in the tables below. All analyses are presented for comparison 1 265

266 (LAMA + LABA/ICS + roflumilast vs. LAMA + LABA/ICS) and comparison 2 (LAMA + LABA + roflumilast vs. LAMA + LABA). Results for scenario analysis F are presented in section

267 ICS-tolerant patients (Comparison 1) Fully incremental analysis (switch to second line on disease progression) Table B 82: Fully incremental analysis, patients switch to 2nd line treatment upon disease progression: ICS-tolerant patients # Treatment Cost Exacerbations LY QALYs 1 LABA 20, LAMA 20, LABA + roflumilast 22, LAMA + roflumilast 22, LAMA + LABA 22, LABA/ICS 21, LAMA + LABA + roflumilast 24, LABA/ICS + roflumilast 23, LAMA + LABA/ICS 22, LAMA + LABA/ICS + roflumilast 24, Comparisons of particular interest: scenario analyses A to E Table B 83: Scenario analysis A: variation of number of exacerbations with the common comparator. ICS-tolerant patients Cost, Exacerbati ons 1 exacerbation in 1st line LAMA + LABA/ICS LABA/ICS + LAMA + Roflumilast LABA/ICS + LAMA 22, ,4 64 Hospital treated exacerbati ons exacerbations in 1st line LAMA + LABA/ICS (base case) LABA/ICS + LAMA + Roflumilast LABA/ICS + LAMA 23, , exacerbations in 1st line LAMA + LABA/ICS LABA/ICS + LAMA + Roflumilast LABA/ICS + LAMA 23, , LY QAL Y ICER ( /QALY) 29, , ,

268 Table B 84: Scenario analysis B: Incorporation of a lung function benefit [+46ml]. ICS-tolerant patients With lung function benefit LABA/ICS + LAMA + Roflumilast LABA/ICS + LAMA Cost, 23, ,8 16 Exacerbati ons Without lung function benefit (base case) LABA/ICS + LAMA + Roflumilast LABA/ICS + LAMA 23, ,8 16 Hospital treated exacerbati ons LY QAL Y ICER ( /QALY) 8, , Table B 85: Scenario analysis C: variation of time on 1st line therapy. ICStolerant patients Time on 1st line = 0.5 years LABA/ICS + LAMA + Roflumilast LABA/ICS + LAMA Cost, 23, ,8 16 Time on 1st line = 1 year (base case) LABA/ICS + LAMA + Roflumilast LABA/ICS + LAMA Time on 1st line = 1.5 years LABA/ICS + LAMA + Roflumilast LABA/ICS + LAMA Time on 1st line = 2 years LABA/ICS + LAMA + Roflumilast LABA/ICS + LAMA Time on 1st line = lifetime LABA/ICS + LAMA + Roflumilast 23, , , , , , ,6 02 Exacerbati ons Hospital treated exacerbati ons LY QAL Y ICER ( /QALY) 16, , , , ,

269 LABA/ICS + LAMA Cost, 22,8 16 Exacerbati ons Hospital treated exacerbati ons LY QAL Y ICER ( /QALY) Table B 86: Scenario analysis D: change of disutilities of exacerbations. ICStolerant patients Cost, Exacerbati ons Hospital treated exacerbati ons Utilities (exacerbations) derived from Oostenbrink et al LABA/ICS + LAMA + Roflumilast LABA/ICS + LAMA 23, , LY QAL Y Utilities (exacerbations) derived from Rutten-van Molken et al (base case) LABA/ICS + LAMA + Roflumilast LABA/ICS + LAMA 23, , ICER ( /QALY) 16, , Table B 87: Scenario analysis E: 2nd line therapy is LAMA + LABA/ICS + Roflumilast for all patients. ICS-tolerant patients Cost, Exacerbati ons Hospital treated exacerbati ons 2nd line therapy is LAMA + LABA/ICS + Roflumilast for all patients LABA/ICS + LAMA + Roflumilast LABA/ICS + LAMA 26, , Base case (2 nd line therapy is LAMA + LABA/ICS for all) LABA/ICS + LAMA + Roflumilast LABA/ICS + LAMA 23, , LY QAL Y ICER ( /QALY) 16, ,

270 ICS-intolerant/declining patients (Comparison 2) Fully incremental analysis (switch to second lone on disease progression) Table B 88: Fully incremental analysis (patients switch to 2nd line only upon disease progression): ICS-intolerant/declining patients # Treatment Cost Exacerbations LY QALYs 1 LABA 20, LAMA 20, LABA + roflumilast 22, LAMA + roflumilast 22, LAMA + LABA 21, LAMA + LABA + roflumilast 23, Comparisons of particular interest: scenario analyses A to E Table B 89: Scenario analysis A: variation of number of exacerbations with the common comparator. ICS-intolerant/declining patients Cost, Exacerbatio ns 1 exacerbation in 1st line LAMA + LABA LABA + LAMA + Roflumilast LABA + LAMA 21, ,2 96 Hospital treated exacerbatio ns exacerbations in 1st line LAMA + LABA (base case) LY QAL Y ICER ( /QALY) 24,

271 LABA + LAMA + Roflumilast LABA + LAMA Cost, 22, ,8 14 Exacerbatio ns 3 exacerbations in 1st line LAMA + LABA LABA + LAMA + Roflumilast LABA + LAMA 22, ,1 28 Hospital treated exacerbatio ns LY QAL Y ICER ( /QALY) 13, , Table B 90: Scenario analysis B: Incorporation of a lung function benefit [+46ml]. ICS-intolerant/declining patients With lung function benefit LABA + LAMA + Roflumilast LAMA + LABA Cost, 22, ,8 14 Exacerbatio ns Without lung function benefit (base case) LABA + LAMA + Roflumilast LAMA + LABA 22, ,8 14 Hospital treated exacerbatio ns LY QAL Y ICER ( /QALY) 7, , Table B 91: Scenario analysis C: variation of time on 1st line therapy. ICSintolerant/declining patients Time on 1st line = 0.5 years LABA + LAMA + Roflumilast LABA + LAMA Cost, 22, ,8 14 Time on 1st line = 1 year (base case) LABA + LAMA + Roflumilast LABA + LAMA 22, ,8 14 Exacerbatio ns Hospital treated exacerbatio ns LY QAL Y ICER ( /QALY) 13, ,

272 Time on 1st line = 1.5 years LABA + LAMA + Roflumilast LABA + LAMA Time on 1st line = 2 years LABA + LAMA + Roflumilast LABA + LAMA Time on 1st line = lifetime LABA + LAMA + Roflumilast LABA + LAMA Cost, 22, , , , , ,8 14 Exacerbatio ns Hospital treated exacerbatio ns LY QAL Y ICER ( /QALY) 13, , , Table B 92: Scenario analysis D: change of disutilities of exacerbations. ICSintolerant/declining patients Cost, Exacerbatio ns Hospital treated exacerbatio ns Utilities (exacerbations) derived from Oostenbrink et al LABA + LAMA + Roflumilast LAMA + LABA 22, , LY QAL Y Utilities (exacerbations) - derived from Rutten-van Molken et al (base case) LABA + LAMA + Roflumilast LAMA + LABA 22, , ICER ( /QALY) 14, ,

273 Table B 93: Scenario analysis E: 2nd line therapy is LAMA + LABA/ICS + Roflumilast for all patients. ICS-intolerant/declining patients Cost, Exacerbatio ns Hospital treated exacerbatio ns LY QAL Y ICER ( /QALY) 2 nd line therapy is LAMA + LABA + Roflumilast in patients who started on this therapy in 1 st line LABA + LAMA + Roflumilast LAMA + LABA 25, , Base case (2 nd line therapy is LAMA + LABA for all) LABA + LAMA + Roflumilast LAMA + LABA 22, , , ,

274 What were the main findings of each of the sensitivity analyses? The OWSA identified that cost-effectiveness results for the comparisons of interest were most sensitive to the relative rate of exacerbation between treatment regimens, and the baseline exacerbation rates. With regards to the relative rate of exacerbation, the confidence interval around the ratio confirms the statistical significance of the expected reduction. When considering the results of the analyses varying the baseline exacerbation rate for the common comparator, it should also be noted that the licence for roflumilast requires patients to have a history of frequent exacerbations. The base case assumption of patients moving to a second line treatment after 1 year was identified by Dr Michael Rudolf and Professor Wisia Wedzicha as one which may vary in clinical practice. However, Scenario analysis C shows that results are not sensitive to time on 1st line therapy, which was tested for alternative values of 0.5 years, 1.5 years, 2 years and lifetime. This is because switching patients to a common second line (at the same time) removes both the potential benefits and cost difference between first line treatment regimens. Alternative disutility values associated with exacerbations (Scenario analysis D), and assuming that all patients moved to a common second-line treatment which included roflumilast (Scenario analysis E) also had minimal impact on base case results. The sub-group analysis focusing on patients with very severe COPD (Scenario analysis F; see Section 6.9) identified a reduction in ICERs for both comparisons of interest, relative to their base case results. These results are to be expected, given the higher underlying baseline exacerbation rate. From the PSA, the cost-effectiveness acceptability curve (CEAC) indicates a probability of LAMA + LABA/ICS + roflumilast being cost-effective vs. LAMA + LABA/ICS of about 55% in ICS-tolerant patients, given a WTP of 20,000. This increases to approximately 70% and 80% at WTP ceiling of 25,000 and 30,000. In ICS-intolerant/declining patients, the probability of LAMA + LABA + roflumilast being cost-effective vs. LAMA + LABA amounts to approximately 73%, at a WTP ceiling of 20,000, and 89% at 30,

275 What are the key drivers of the cost-effectiveness results? The cost-effectiveness model is primarily driven by: the relative rate of exacerbation between the assessed treatment regimen vs. the common comparator (LAMA + LABA/ICS). the assumed number of background exacerbations per year in the common comparator. 275

276 6.8 Validation Please describe the methods used to validate and quality assure the model. Provide references to the results produced and cross-reference to evidence identified in the clinical, quality of life and resources sections. The cost-effectiveness model has been validated in the following manner: A review of key literature in the field of COPD economic modelling was conducted prior to formulating the model design concept. The initial design of the cost-effectiveness model was extensively discussed at an advisory board meeting with the involvement of key experts in clinical aspects of COPD and the economic modelling of COPD. The model design and the results of the economic evaluation were presented to leading health economic experts throughout the period of model development. Within the validated model framework, the model was populated with the results of the most recently available evidence synthesis for COPD treatments. 22 The assumptions of extrapolating key outcomes of interest beyond the time horizon of the clinical trials is the common practice in the economic modelling of chronic disease including COPD. The assumptions defining the base case analysis were discussed with experienced and respected respiratory physicians, and confirmed as being appropriate. Dr Michael Rudolf (Respiratory Physician, Ealing Hospital) and Professor Wisia Wedzicha (Consultant Respiratory Physician, Royal Free Hospital) both also advised that the most likely use of roflumilast in clinical practice would be as an add-on to current standard of care (LAMA + LABA/ICS, or LAMA + LABA for patients who can not tolerate or decline ICS). 276

277 6.9 Subgroup analysis For many technologies, the capacity to benefit from treatment will differ for patients with differing characteristics. This should be explored as part of the reference-case analysis by providing separate estimates of clinical and cost effectiveness for each relevant subgroup of patients. This section should be read in conjunction with NICE s Guide to the methods of technology appraisal, section Types of subgroups that are not considered relevant are those based solely on the following factors. Individual utilities for health states and patient preference. Subgroups based solely on differential treatment costs for individuals according to their social characteristics. Subgroups specified in relation to the costs of providing treatment in different geographical locations within the UK (for example, when the costs of facilities available for providing the technology vary according to location) Please specify whether analysis of subgroups was undertaken and how these subgroups were identified. Were they identified on the basis of an a priori expectation of differential clinical or cost effectiveness due to known, biologically plausible, mechanisms, social characteristics or other clearly justified factors? Cross-reference the response to section As described in Section 6.7.9, scenario analyses were run for several situations. A further pair of sub-group analyses were performed for comparison 1 and comparison 2, focussing only on those patients with very severe COPD Please clearly define the characteristics of patients in the subgroup. This subgroup consists of patients who start treatment in the Very severe COPD health state (Figure 36). The starting lung function in this subgroup is a predicted FEV 1 < 30%, relative to the natural decline of lung function in the general population (Table B 43). However, as there is no further progression of COPD once patients are in the Very severe state of the model, providing the starting lung function in this group is < 30% predicted FEV 1, there is no impact on results. 277

278 Figure 36: Markov Model Structure for patients starting treatment in the Very Severe state Please describe how the statistical analysis was undertaken. Key characteristics of patients in this subgroup were derived from the statistical analysis of the data from studies M2-124 and M What were the results of the subgroup analysis/analyses, if conducted? Please present results in a similar table as in section (Base-case analysis). The subgroup analysis of patients who start treatment in the Very severe COPD health state (Scenario analysis F) yielded lower ICERs compared to the base case ( 12,310 vs. 16,566 [ICS-tolerant patients] and 10,479 vs. 13,765 [ICSintolerant/declining patients]). 278