Decline In lung-function Among Patients with chronic obstructive Lung disease On maintenance therapy (DIAPLO)

Transcription

1 Version V1.4 Decline In lung-function Among Patients with chronic obstructive Lung disease On maintenance therapy (DIAPLO) An observational study evaluating the benefits of early intervention with maintenance therapies to prevent or slow down rapid lung function decline in patients who are at high risk at the time of COPD diagnosis in the combined Optimum Patient Care Research Database and Clinical Practice Research Datalink databases

2 TABLE OF CONTENTS PAGE TITLE PAGE... 1 TABLE OF CONTENTS... 2 LIST OF ABBREVIATIONS... 5 RESPONSIBLE PARTIES... 6 PROTOCOL SYNOPSIS DIAPLO STUDY... 7 AMENDMENT HISTORY MILESTONES BACKGROUND AND RATIONALE Background Rationale OBJECTIVES AND HYPOTHESES First Objective & Hypothesis Second Objective & Hypothesis Third Objective & Hypothesis Fourth Objective & Hypothesis METHODOLOGY Study Design General Aspects Objective 1: Validation of Zafari prediction model Objective 2: Characterisation of patients at high risk of rapid lung function decline Objective 3: Comparison of observed FEV1 decline under maintenance therapy with predicted decline under minimal therapy Objective 4: Effectiveness of triple therapy on lung function decline Data Source(s) Study Population Inclusion Criteria Overall inclusion criteria Additional inclusion criteria objective Additional inclusion criteria objective Additional inclusion criteria objective Exclusion Criteria VARIABLES AND EPIDEMIOLOGICAL MEASUREMENTS

3 4.1 Predictors of Zafari model Therapy exposure Period of minimal inhalation therapy Period of maintenance inhalation therapy Period of triple therapy Period of poor adherence to maintenance inhalation therapy High risk of rapid FEV1 decline Outcomes Primary end-point objective 4: Rate of FEV1 decline Secondary endpoints objective 4: Matched comparison of patients on triple therapy with patients on minimal therapy Time to first COPD exacerbation / Exacerbation rate Time to first hospitalisation for COPD Treatment success Other Variables and Covariates Baseline characteristics Comorbidities Clinical characteristics STATISTICAL ANALYSIS PLAN Statistical Methods General Aspects First objective(s): Validation of Zafari prediction model Validation of predicted FEV1 values during follow up under minimal therapy Second objective: Characterisation of patients at high risk of rapid lung function decline Third Objective: Comparison of individual observed lung function decline under maintenance therapy with predicted rate under minimal therapy Fourth Objective: effectiveness of triple therapy on lung function decline and secondary COPD outcomes Matching Outcome analyses objective Bias Methods to Minimize Bias Strengths and Limitations Sample Size and Power Calculations Communication Plan Publication Plan LIST OF REFERENCES APPENDICES Diagnostic Read codes for COPD

4 7.2 Read codes for exclusion of patients with overlapping asthma Read codes other chronic respiratory conditions Zafari prediction model MRC Dyspnoea Scale Co-morbidities Read codes Diagnosis of Asthma Allergic or non-allergic rhinitis diagnosis Allergic rhinitis drugs Prodcodes for CPRD Allergic rhinitis drugs Readcode for OPCRD Eczema diagnosis Nasal polyps Chronic sinusitis diagnosis and treatment procedures Gastro-oesophageal reflux Diabetes Mellitus Osteoporosis diagnosis Hypertension Ischaemic heart disease Heart Failure Depression Anxiety Obstructive Sleep Apnea Sleep disorders Chronic Kidney disease

5 LIST OF ABBREVIATIONS Abbreviation or special term COPD CPRD FEV 1 FVC HES ICS ID LABA LAMA MPR OPCRD OPRI SABA SAMA RMSE Explanation Chronic Obstructive Pulmonary Disease Clinical Practice Research Datalink Forced Expiratory Volume in one second Forced Vital Capacity Hospital Episode Statistics Inhaled Corticosteroids Index Date Long-Acting Beta Agonist Long-Acting Muscarinic Antagonist Medication Possession Ratio Optimum Patient Care Research Database Observational and Pragmatic Research Institute Short-Acting Beta Agonist Short-Acting Muscarinic Antagonist Root mean squared error 5

6 RESPONSIBLE PARTIES Name Professional Title Role in Study Affiliation Address Claudia Cabrera, MPH PhD Epidemiology Lead Researcher AZ Paul Dorinsky, MD Global Clinical Lead Researcher PEARL David Price Professor Chief Scientist University of Aberdeen Observational and Pragmatic Research Institute Pte Ltd (OPRI) Sen Yang, MD Vice president OPRI affiliates and project co-ordinator Senior Medical Lead OPRI Chris Price Commercial Director Contracting OPRI Marjan Kerkhof, MD, PhD Senior Researcher, Epidemiologist Researcher OPRI Marcus Ngantcha. Msc Statistician Statistician OPRI 6

7 PROTOCOL SYNOPSIS DIAPLO STUDY An observational study evaluating the benefits of early intervention with maintenance therapies to prevent or slow down rapid lung function decline in patients who are at high risk at the time of COPD diagnosis in the combined Optimum Patient Care Research Database and Clinical Practice Research Datalink databases Background/Rationale: Chronic obstructive pulmonary disease (COPD) is a progressive disease characterised by nonreversible airflow limitation. A few clinical trials demonstrated modest effects of maintenance inhalation therapy on reduction of lung function decline in patients with COPD. 1,2 To date, there is a lack of evidence on the effect of triple therapy (ICS/LABA/LAMA) on lung function decline, although several RCTs have found adding a LAMA to ICS/LABA to improve lung function. 3-6 As lung function data are captured in electronic health records in the UK, there is a unique opportunity to study the benefits of early intervention with maintenance therapies to prevent or slow down rapid lung function decline in patients who are at high risk and most likely to benefit from preventive therapies at the time of COPD diagnosis. Objectives and Hypotheses: The hypothesis of this study is that early treatment with maintenance therapy, particularly triple therapy, will prevent or slow down rapid lung function loss in patients diagnosed with mild or moderate COPD (FEV1 %predicted 50%) who are at high risk of rapid lung function decline. Four sub-studies will be performed with the following objectives: 1. In later objectives, the Zafari model a prediction model for lung function decline under minimal therapy- will be used to identify patients at high risk of rapid FEV1 decline. Thus, our first objective concerns the validation of this model: To study whether the observed FEV1 decline during minimal inhalation therapy in patients diagnosed with mild or moderate COPD corresponds with the rate of decline predicted by the model of Zafari et al To characterise patients at high risk of rapid FEV1 decline at the time they were identified when diagnosed with mild or moderate COPD and at the time of initiation of maintenance therapy (LAMA, LABA and/or ICS) and to compare these characteristics with those of patients not being at high risk of rapid FEV1 decline. 3. To compare observed FEV1 values recorded during maintenance therapy with predicted values for minimal inhalation therapy within patients identified as being at high risk of rapid lung function decline by the prediction model. 4. To study the effectiveness of triple therapy on slowing down the rate of FEV1 decline in patients who were identified as being at high risk of rapid FEV1 decline by comparing the rate of FEV1 decline in patients during triple therapy with that of similar patients during minimal therapy or poor adherence to maintenance therapy. 7

8 Methods: Study design: Patients who are newly diagnosed with mild or moderate COPD and have a positive history of tobacco smoking will be identified at the time of their first FEV1 recording, that is within 3 months prior to and 12 months after their COPD diagnosis. This point will be called Index Date 1 (ID1). Patients will be followed from ID1 until the date at which maintenance therapy is introduced, which will be called Index date 2 (ID2). All FEV1 values recorded under minimal therapy (no therapy, or short-acting bronchodilators only) during this period between ID1 and ID2 will be used for analyses in patients. The recorded FEV1 values will be compared with the individual s predicted values at that time as calculated from the model published by Zafari et al. If the observed FEV1 values during minimal inhalation therapy are found to correspond well with the predicted values, the Zafari model will be used to identify patients at high risk of rapid FEV1 decline, defined as a probability of 80% of 40 ml/year decline in FEV1 over the next 11 years. These patients will be selected for the subsequent analyses. For objective 2, the patients will be characterised at the date of identification (ID1=date of first FEV1 recording at or around time of first diagnosis of mild to moderate COPD) and at the date of initiation of maintenance therapy (ID2). For objective 3, FEV1 values recorded during maintenance therapy (after ID2) will be compared with predicted values for minimal therapy within patients at high risk of rapid lung function decline at the time of COPD diagnosis. Analyses will be performed for patients with first maintenance therapy being a single inhaler, dual therapy and triple therapy separately. For objective 4, patients who are initiated on triple therapy will be matched to similar patients from a combined group of patients on minimal therapy or patients who are persistently non-adherent to maintenance therapy (Medication Possession Ratio for maintenance therapies < 40% in 3 consecutive years prior to index date (ID3)). The mean observed rates of FEV1 decline after index date will be compared between these two study arms. 8

9 Data Source(s): A combined dataset of patients with COPD using data extracted from OPCRD and CPRD will be used for analyses. A subpopulation of patients from CPRD with linkage to Hospital Episode Statistics (HES) available will be selected to study the association between therapy and the secondary outcome of COPD-related hospital admission. Interim analyses will be delivered based on OPCRD data only for objectives 1 and 2. Study Population: The study population will consist of patients with COPD who are registered at general practices providing data to OPCRD and CPRD across England, Scotland, Wales, and Northern Ireland Overall inclusion Criteria: 1. A diagnostic Read code for COPD as from 1 January 2005 (date of first diagnostic Read code is ID0) 2. Valid spirometry recorded in period between 3 months before and 12 months 1 after ID0 (date of spirometry is ID1) 3. Evidence of obstruction, defined as FEV1/FVC<0.7 and mild or moderate COPD, i.e. i.e. FEV1 % predicted 50 and 90% at ID1 4. Age 35 years at ID year of medical records prior to ID1 for baseline characterisation 6. A positive history of tobacco smoking at the time of first COPD diagnosis (ID0) at any recording of smoking status prior to ID0 or at first recording after ID0 7. Recorded height at adult age ( 21 years) and weight within 5 years of ID1 Additional inclusion criteria objective 1 1. Follow-up period covering 1 spirometry records after ID1 2. Minimal inhalation therapy during follow-up period after ID1, defined as no inhalation therapy or only short-acting bronchodilator prescriptions Additional inclusion criteria objective 3 1. Identified as being at high risk of rapid FEV1 decline at ID1, defined as 80% probability of 40ml FEV1 decline per year over 11 years, using prediction model 2. Initiation of maintenance inhalation therapy at any time during follow-up after ID1. Date of initiation is index date 2 (ID2) 3. Follow-up period covering 1 spirometry records after ID2 Additional inclusion criteria objective 4 1. Identified as being at high risk of rapid FEV1 decline at ID1, defined as 80% probability of 40ml FEV1 decline per year over 11 years, using prediction model 2. Period of treatment at any time during follow-up after ID1 (definitions at paragraph 4.2) with: a. Triple therapy b. Control therapy: 1 Time period following QOF indicator for spirometry confirmed COPD diagnosis 9

10 i. Minimal inhalation therapy for 1 year OR ii. Poor adherence to maintenance inhalation therapy for 3 consecutive years, defined as a Medication Possession Ratio (MPR) <40% Date of triple therapy initiation or matched control date is index date 3 (ID3) 3. 2 follow-up spirometry records during treatment period after ID3 Exclusion criteria 1. Active asthma after date of first COPD diagnosis (ID0), defined as 1 diagnostic Read code for asthma or 1 asthma monitoring or review Read code recorded 2. Diagnostic Read code for Asthma-COPD overlap syndrome as first COPD diagnostic Read code or recorded during follow-up after ID0 3. Diagnostic Read code for other chronic lower respiratory conditions recorded ever prior to ID0 Exposures: A period of minimal inhalation therapy is defined as the absence of any inhalation therapy or prescriptions for short-acting bronchodilators starting at the time of spirometry recording closest to first COPD diagnosis (ID1). This period will end at the time of initiation of maintenance inhalation therapy, loss to follow-up or death. A period of maintenance therapy starts with the date of first prescription of LAMA and/or LABA and/or ICS after ID1 and ends with the absence of prescriptions for 120 days, at loss to follow-up or death. A period of triple therapy is defined as the presence of overlapping periods covered by prescriptions for LAMA, LABA and ICS, i.e. prescription for a third drug at index date while the first two are still active (based on number of days prescribed at most recent prescription prior to ID2) or prescribed at the same date. Drugs need to be followed by a repeat prescription within 30 days after end of prescription duration during followup. The period of triple therapy ends with absence of prescriptions for any of the three drugs for >120 days. A period of poor adherence is defined as a Medication Possession Ratio <40% for LABA, LAMA or ICS in the last 3 consecutive years prior to index date. Outcomes: Recorded FEV1 values will be compared with predicted values calculated from the Zafari prediction model. For objective 4, the change in FEV1 over time, expressed as ml/year of follow up, will be estimated using a multilevel model for change. Feasibility: A feasibility check in the OPCRD of the number of patients with a COPD diagnosis, confirmed by spirometry found more than 6,000 patients who fulfilled the inclusion and exclusion criteria for objective 1. 10

11 Statistical Analysis: In mild to moderate COPD patients, all FEV1 values captured in the combined databases when the patient was on minimal therapy will be compared to those predicted by the model developed by Zafari et al. Scatterplots will be produced and the overall root mean squared error (RMSE) of the predicted versus observed FEV1 value will be calculated. The coverage probability, defined as the proportion of the observed FEV1 values falling within the 95% prediction interval for that observation will be determined. For objective 3, the absolute difference between the individual s observed FEV1 values after initiation of maintenance therapy and his/her predicted values at that same date (calculated from the Zafari model) will be calculated and distributions will be described. Analyses will be performed for patients with first maintenance therapy being a single inhaler, dual therapy or triple therapy, separately. FEV1 values recorded after a switch from the first maintenance therapy category to another will not be considered. A linear mixed model will be used with the absolute difference between observed and predicted FEV1 value as the outcome variable to estimate the influence of duration of therapy, taking repeated measurements within patients into account. For objective 4, direct matching will be applied to compare mean observed rates of FEV1 decline between the two study arms. Potential confounders of the association between triple therapy and the rate of FEV1 decline will be identified in a baseline year and the choice of the most relevant confounders will be based on a combination of baseline imbalance, bias potential and expert judgement. For the primary outcome analyses, the change in FEV1 over time will be analysed using a multilevel model for change. The level-1 component of this model will be the individual growth model, representing the change in FEV1 over time within patients. The level-2 component codifies the mean difference in FEV1 decline within matched pairs and the level-3 component will represent the mean population difference in lung function decline between the two study arms. Mixed linear regression will be used to analyse the data. Intercept and slope will be modelled as random effects (to vary across individuals). A matched-pair specific random effect will be added to the model next to covariates representing potential confounders. The time difference between the date at which triple therapy was initiated and the time of first diagnosis will be included into the model as a potential confounder or effect modifier (testing of interaction term with therapy). We will also study whether there is a potential non-linear component of decline during follow-up by including polynomial functions of time into the model. For analysing time-to-event outcomes, unadjusted Kaplan-Meier estimates of the survivor functions are plotted for the matched study arms. A stratified log-rank test will be performed to statistically test significance of observed differences. A stratified Cox proportional hazard model will be used to estimate the hazard ratio with 95% confidence interval for the association between therapy and time to the first event. Conditional negative binomial regression will be used to estimate the adjusted rate ratio with 95% confidence interval for the association between therapy and the number of COPD exacerbations in the outcome year. Conditional logistic regression analyses will be used to estimate the adjusted odds ratio with 95% confidence interval for the association between therapy and treatment success. 11

12 AMENDMENT HISTORY Date Brief description of change Administrative Change / Amendment / New Protocol Version. 25 April First draft protocol V1.0 5 May Amendments after review V May Amendments after review V June Update Medcodes/Read codes and amendments following ISAC protocol V July Update following feedback ISAC V1.4 12

13 MILESTONES Date Deliverables 3 July 2017 Report OPCRD-based Zafari model validation (Objective 1-interim) 1 August 2017 Report OPCRD-based characterisation fast-decliners (Objective 2-interim) 4 September 2017 Report validation of Zafari model in combined dataset or of new model in CPRD dataset (Objective 1) 8 September 2017 Report characterisation fast-decliners in combined dataset (Objective 2) 27 October 2017 Report effects maintenance therapy on FEV 1-observed vs predicted (Objective 3) 15 December 2017 Effects triple therapy vs minimal therapy-matched design (Objective 4) 12 March 2018 Delivery first manuscript draft 9 April 2018 Manuscript journal submission 13

14 1. BACKGROUND AND RATIONALE 1.1 Background Chronic obstructive pulmonary disease (COPD) is a progressive respiratory condition characterised by non-reversible airflow limitation. 8 According to the WHO around 65 million people have a moderate to severe form of the disease worldwide. 9 Current WHO estimates predict that by 2030, COPD will become the third leading cause of death globally, 10 The estimated direct cost of COPD is substantial. In the European Union, the total direct cost of respiratory disease are estimated to be about 6% of the total health care budget, with COPD accounting to 56% (38.6 billion Euros) of the combined cost of respiratory disease. The diagnosis of COPD rests on demonstration of chronic airflow obstruction, which is not fully reversible. 11 After diagnosis, disease progression is monitored by repeated spirometry measurements to track the decline in Forced Expiratory Volume in one second (FEV1), which, in the UK, should be part of the standard management of COPD in general practice. Concepts relating to the natural history of tobacco-related COPD arise from the landmark study of Fletcher and colleagues. 12 They suggested that lung function (FEV1) declines continuously and smoothly over an individual s life with a slightly accelerated decline with aging. 12 However, susceptible smokers demonstrate a much faster decline and eventually develop clinically severe airflow obstruction, leading to a diagnosis of COPD, mostly when they have mild to moderate disease. 12 Figure 1. The decline in lung function with age, smoking, and smoking cessation by Fletcher and Peto 12 Tantucci et al. reviewed spirometric data of patients with COPD in the placebo arms of clinical trials and found that lung function decline was most accelerated in patients with moderate COPD (GOLD stage II) (Figure 2). 13 The authors highlighted a lack of knowledge about lung function decline in mild disease (FEV1 80%) and emphasised the need for further studies to determine whether intervention during the initial stages of COPD can ameliorate the progression of the disease. 14

15 Figure 2. Range of average rate of decline in the forced expiratory volume in the first second (FEV1) of chronic obstructive pulmonary disease patients according to initial severity of airflow reduction by Tantucci et al. 13 Until recently, the only effective intervention was smoking cessation. However, newly reported large clinical trials have demonstrated modest effects of maintenance inhalation therapy on reduction of lung function decline in patients with COPD. 1,2 Post-hoc analyses of the Toward a Revolution in COPD Health (TORCH) study found that the average rate of FEV1 decline among 5,343 former smokers with moderate to severe COPD was 55 ml/year in the placebo group, 42 ml/year in patients treated with long-acting beta-agonists (LABA), 42 ml/year in patients treated with inhaled corticosteroids (ICS) only, and 39 ml/year in the combined ICS/LABA group measured over 3 years. 14 The reduction in rate of decline in the ICS/LABA group versus placebo was 16 ml/year (95% confidence interval, 7 25 ml; P<0.001), suggesting that combination therapy (ICS/LABA) slowed the rate of disease progression. 2 A subgroup analysis of 2,739 patients with mild to moderate COPD from the Understanding Potential Long-Term Impacts on Function with Tiotropium (UPLIFT) study found that the average rate of FEV1 decline was lower in patients treated with long-acting muscarinic antagonists (LAMA) (43 ml/year) compared with the control group (49 ml/year), providing evidence that LAMA also slows the rate of decline. 15 However, these observed effects are relatively small and to be of clinical benefit, these therapies likely need to start early in the course of disease and be prolonged. 16 To date, there is a lack of evidence on the effect of triple therapy (ICS/LABA/LAMA) on lung function decline, although several RCTs have found adding a LAMA to ICS/LABA to improve lung function. 3-6 Current COPD management guidelines recommend dual and triple therapy, including inhaled corticosteroids (ICS) to be reserved for patients with more severe airflow limitation (FEV1<50% predicted). 11,17 However, triple therapy is often prescribed earlier than GOLD guidelines recommend. A previous study analysing medical records data from the Optimum Patient Care Research Database (OPCRD) suggested that of all patients prescribed triple 15

16 therapy, 25% had this prescribed within 1 year of diagnosis, irrespective of GOLD classification. 18 As lung function data are captured in electronic health records in the UK, there is a unique opportunity to study whether early treatment with triple therapy has a beneficial effect on lung function decline. The Lung Health Study (LHS) was a multicentre clinical trial that recruited 5,887 heavy smokers, aged years with mild to moderate COPD (FEV1% predicted 55-90%) in when recommended pharmacological treatment for COPD was limited to short-acting bronchodilators. 19 Patients were randomly assigned to 3 arms of usual care and intervention (smoking cessation) with or without ipratropium. Based on annual follow-up spirometry data during the first 5 years and an extended follow-up measurement at about the 11 th year collected in this study, Zafari et al. recently published an externally validated model to predict the natural history of FEV1 decline over time in patients on minimal treatment. 7 This framework will allow to enable risk stratification of patients with mild to moderate COPD in terms of their future lung function decline under minimal therapy and to identify patients with rapid disease progression, who can be targeted for intervention. 7 16

17 1.2 Rationale COPD manifests itself with symptoms of dyspnoea, chronic cough or sputum, mostly at the time of mild to moderate airflow limitation. To date, there is no conclusive clinical trial evidence that any existing medication for COPD modifies the long-term decline in lungfunction. 11 Recently reported clinical trials have demonstrated modest effects to slow lung function decline in patients with COPD. However, to be of clinical benefit, these therapies likely need to start early in the course of disease and be prolonged. To our knowledge, the effectiveness of triple therapy as an intervention in patients who were identified as being at high risk of rapid future lung function decline at early stages of disease and who are most likely to benefit from preventive therapies, has never been studied.. This study aims to study the benefits of early intervention with maintenance therapies to prevent or slow down rapid lung function decline in patients who are at high risk at the time of COPD diagnosis in the combined Optimum Patient Care Research Database and Clinical Practice Research Datalink databases. First, the study will validate the accuracy of the model developed by Zafari et al. to predict the future rate of FEV1 decline under minimal therapy using lung function test results extracted from General Practice (GP) s electronic medical records in the UK. Second, the study will characterise patients identified by the prediction model as being at high risk of rapid FEV1 decline under minimal therapy at two moments in time: at first diagnosis of mild to moderate COPD and at initiation of maintenance therapy. Characteristics will be compared with those from patients not at high risk. The third and fourth sub-studies will focus on patients who were identified as being at high risk of rapid lung function decline by the prediction model applied at the time of first diagnosis of mild to moderate COPD. For sub-study 3, FEV1 values recorded during maintenance therapy will be compared with predicted values for minimal inhalation therapy within patients identified as being at high risk of rapid lung function decline by the prediction model. Analyses will be performed for patients with first maintenance therapy being a single inhaler, dual therapy and triple therapy separately. In the fourth sub-study, the effectiveness of triple therapy on FEV1 decline will be studied in a matched study design with similar patients on minimal therapy or patients who are persistently non-adherent to maintenance therapy included in the control arm. 17

18 2. OBJECTIVES AND HYPOTHESES 2.1 First Objective & Hypothesis Hypothesis: In patients diagnosed with mild to moderate COPD, the rate of FEV1 decline under minimal inhalation therapy corresponds well with the rate of decline predicted by the model published by Zafari et al. using GP s electronic medical records. Objective: In later objectives, the Zafari model a prediction model for lung function decline under minimal therapy- will be used to identify patients at high risk of rapid FEV1 decline. Thus, our first objective concerns the validation of this model: To study whether the observed FEV1 decline during minimal inhalation therapy in patients diagnosed with mild or moderate COPD corresponds with the rate of decline predicted by the model of Zafari et al. 2.2 Second Objective & Hypothesis Objective: To characterise patients at high risk of rapid FEV1 decline at the time they were identified when diagnosed with mild or moderate COPD and at the time of initiation of maintenance therapy and to compare these characteristics with those of patients not being at high risk of rapid FEV1 decline. 2.3 Third Objective & Hypothesis Hypothesis: Treatment with maintenance therapy, particularly triple therapy, will reduce the rate of lung function decline in patients diagnosed with mild or moderate COPD who are at high risk of rapid lung function decline. Objective: To compare observed FEV1 values recorded during maintenance therapy with predicted values for minimal inhalation therapy within patients identified as being at high risk of rapid lung function decline by the prediction model. 2.4 Fourth Objective & Hypothesis Hypothesis: Treatment with maintenance therapy, particularly triple therapy, will reduce the rate of lung function decline in patients diagnosed with mild or moderate COPD who are at high risk of rapid lung function decline. Objective: To study the effectiveness of triple therapy on slowing down the rate of FEV1 decline in patients who were identified as being at high risk of rapid FEV1 decline by comparing the rate of FEV1 decline in patients during triple therapy with that of similar patients during minimal therapy or poor adherence to maintenance therapy. 18

19 3. METHODOLOGY 3.1 Study Design General Aspects Objective 1: Validation of Zafari prediction model Patients who are newly diagnosed with mild or moderate COPD and have a positive history of tobacco smoking will be identified at the time of their first FEV1 recording, that is within 3 months prior to and 12 months after their COPD diagnosis. This point will be called Index Date 1 (ID1). Patients will be followed from ID1 until the date at which maintenance therapy is introduced, which will be called Index Date 2 (ID2). All FEV1 values recorded under minimal therapy (no therapy, or short-acting bronchodilators only) during this period between ID1 and ID2 will be used for analyses in patients. The recorded FEV1 values will be compared with the individual s predicted values at that time as calculated from the model published by Zafari et al. 7 (Figure 3) The prediction model will be considered valid when a coverage probability of at least 80% will be achieved (% of observed FEV1 values that fall within the 95% prediction interval for that observation). Figure 3. Study design objectives

20 3.1.2 Objective 2: Characterisation of patients at high risk of rapid lung function decline If the Zafari model was found to accurately predict the rate of decline in patients with mild to moderate COPD, this model will be used to select patients with a high probability of rapid FEV1 decline under minimal therapy. These patients will be characterised at two moments in time: at time of identification (first FEV1 recording at or around time of first diagnosis of mild to moderate COPD (ID1)) and at the date of initiation of maintenance therapy (ID2). Characteristics will be compared with those from patients not at high risk (Figure 3). If the Zafari model was found not to provide accurate estimates of the predicted rate of decline, a similar new prediction model will be developed from OPCRD data and validated in CPRD data Objective 3: Comparison of observed FEV1 decline under maintenance therapy with predicted decline under minimal therapy Patients identified as being at high risk of rapid lung function decline under minimal therapy will be selected using the prediction model (either Zafari or newly-created model). The index date (ID2) will be the date of first initiation of maintenance therapy (Figure 3). All FEV1 values recorded during the period of maintenance therapy will be compared with the patient s own predicted values under minimal therapy at the same date. Analyses will be performed for patients with first maintenance therapy being a single inhaler, dual therapy and triple therapy separately. FEV1 values recorded after a switch from the first maintenance therapy option to another will not be considered Objective 4: Effectiveness of triple therapy on lung function decline Patients identified as being at high risk of rapid lung function decline under minimal therapy, who are initiated on triple therapy will be matched to similar patients from a combined group of patients on minimal therapy or patients who are persistently non-adherent to maintenance therapy (Medication Possession Ratio for maintenance therapies < 40% in 3 consecutive years prior to index date (ID3)). Individual growth models will be used to compare the observed rate of FEV1 decline after ID3 between these two study arms Data Source(s) Patient data will be extracted from two large UK primary care databases that will be combined to a single dataset: the Optimum Patient Care Research Database (OPCRD) 20 and the Clinical Practice Research Datalink (CPRD). 21 The OPCRD is developed, maintained, and owned by Optimum Patient Care (OPC), a social enterprise company that aims to improve patient outcomes through medical research and services. OPC provides evidence-based recommendations to UK general practices through bespoke software and practice reports. The OPCRD currently comprises longitudinal medical records for over 3 million patients from over 601 primary care practices across the UK. It has received a favorable opinion from the Health Research Authority for clinical research use (REC reference: 15/EM/0150). Governance is provided by The Anonymous Data Ethics Protocols and Transparency 20

21 (ADEPT) committee, an independent body of experts and regulators commissioned by the Respiratory Effectiveness Group (REG, to govern the standard of research conducted on internationally recognized databases. The CPRD contains de-identified, longitudinal data from 5 million active medical records from more than 600 subscribing practices throughout the UK. It has been widely used for medical and health research. The study will be performed in compliance with all applicable local and international laws and regulations, including without limitation ICH E6 guidelines for Good Clinical Practices. All research conducted by OPRI will be registered on established study databases such as the European Network of Centres for Pharmacoepidemiology and Pharmacovigilance (ENCePP, OPCRD and CPRD data have been combined in multiple previous studies conducted by OPRI. A unique dataset is provided for each study with only the variables necessary to inform the study objectives included. From pooling of data from these two databases for previous studies conducted by OPRI, we observed a 2 4% overlap and we have no reason to expect the percentage overlap to be different from this. We remove duplicates as standard procedure. Throughout construction of the data set for analysis, patient records remained unidentifiable. Initially, the study will use the OPCRD dataset for interim results. This will be followed with the use of a combined CPRD/OPCRD dataset for final results. Hospital Episode Statistics (HES) linkage with CPRD is required to identify COPD-related hospitalisations (HES APC), A&E attendances (HES A&E) and outpatient visits (HES outpatient) for COPD. These analyses will be adjusted for socio-economic status (quintiles of practice-level index of Multiple Deprivation), a potential confounder. 3.2 Study Population The study population will consist of patients with COPD who are registered at general practices providing data to OPCRD and CPRD across England, Scotland, Wales, and Northern Ireland 3.3 Inclusion Criteria Overall inclusion criteria 1. A diagnostic Read code for COPD as from January (date of first diagnostic Read code / Medcode is ID0) (Appendix 7.1) 2. Valid spirometry recorded in period between 3 months before and 12 months 2 after ID0 (date of spirometry is ID1) 3. Evidence of obstruction, defined as FEV1/FVC< Mild or moderate COPD, i.e. FEV1 % predicted 50% and 90% at ID1 5. Age 35 years at ID year of of up-to-standard medical records prior to ID1 for baseline characterisation 7. A positive history of tobacco smoking at the time of first COPD diagnosis (ID0) at any recording of smoking status prior to ID0 or at first recording after ID0 2 Time period following QOF indicator for spirometry confirmed COPD diagnosis 21

22 8. Recorded height at adult age ( 21 years) and weight within 5 years of ID Additional inclusion criteria objective 1 1. Up-to-standard follow-up period covering 1 spirometry records 3 after ID1 2. Period of minimal inhalation therapy immediately following ID1, defined as no inhalation therapy or only short-acting bronchodilators prescriptions (definition at paragraph 4.2.1) Additional inclusion criteria objective 3 1. Identified as being at high risk of rapid FEV1 decline at ID1, defined as 80% probability of >40ml FEV1 decline per year over 11 years, using prediction model 2. Initiation of maintenance inhalation therapy at any time during follow-up after ID1 (definition at paragraph 4.2.2) (Date of initiation is index date 2 (ID2) 3. Up-to-standard follow-up period covering 1 spirometry records after ID Additional inclusion criteria objective 4 1. Identified as being at high risk of rapid FEV1 decline at ID1, defined as 80% probability of >40ml FEV1 decline per year, using prediction model 2. Period of treatment at any time during follow-up after ID1 with (definitions at paragraph of 4.2): a. Triple therapy b. Control therapy: iii. Minimal inhalation therapy for 1 year iv. Poor adherence to maintenance inhalation therapy other than triple therapy for 3 consecutive years Date of triple therapy initiation or matched control date is index date 3 (ID3) 3. Up-to-standard follow-up period covering 2 follow-up spirometry records during period after ID3 3.4 Exclusion Criteria 1. Active asthma at or after date of first COPD diagnosis, defined as 1 diagnostic Read code for asthma or 1 asthma monitoring or review Read code recorded (Appendix 7.2) 2. Diagnostic medcode (109958) or Read code (H3B.. or Xac33) for Asthma-COPD overlap syndrome as first COPD diagnostic code or recorded during follow-up 3. Diagnostic medcode or Read code for other chronic lower respiratory 4. Diagnostic medcode or Read code for other chronic lower respiratory conditions recorded ever (Appendix 7.3) 3 The NICE guideline for COPD recommends follow-up reviews, which includes measurement of FEV1. Reviews should be performed at least annually for people with mild/moderate/severe COPD (stages 1 3). The % of patients with a record of FEV1 in the preceding 12 months is a QOF indicator. 22

23 4. VARIABLES AND EPIDEMIOLOGICAL MEASUREMENTS 4.1 Predictors of Zafari model The following variables will be assessed to predict the rate of decline (ml/year) using the Zafari model: 1. Baseline FEV1 assessed within 3 months before and 12 months 4 after first COPD diagnostic Read code measured under minimal therapy (paragraph 4.2.1). (ID1) 2. Age at baseline FEV1 recording 3. Sex 4. Weight (kg), closest to ID1, recorded within 5 years 5. Height (m), closest to ID1, recorded after age of Smoking history, with the following categories, assessed during the first 5 years of follow-up after ID1 5 a. Continuous smokers, defined as being recorded as a current smoker at all recordings b. Sustained quitters, defined as being recorded as an ex-smoker at all recordings c. Intermittent quitters, varying recording of being current and ex-smoker 7. Follow-up time from ID1 8. Smoking cessation intervention within 3 months (advice or drugs) after ID1 9. Short-acting bronchodilator prescription within 3 months after ID1 4.2 Therapy exposure Period of minimal inhalation therapy A period of minimal inhalation therapy is defined as the absence of any inhalation therapy or prescriptions for short-acting bronchodilators from the time of spirometry recording closest to first COPD diagnosis (ID1). This period will end at the time of initiation of maintenance inhalation therapy (ID2), loss to follow-up or death Period of maintenance inhalation therapy Several objectives specify patients should have a period of maintenance inhalation therapy. This period should include 2 prescriptions for the drugs and starts with the date of first prescription of LAMA and/or LABA and/or ICS after ID1 and ends with the absence of prescriptions for >120 days for the same drug, at loss to follow-up or death. Three separate periods will be distinguished: 1. Monotherapy, therapy with one single inhaler (LAMA or LABA or ICS). The period starts at the date of first prescription (ID2), which should be followed by 1 repeat prescription within 30 days after prescription duration, and ends when there is no 4 Time period following QOF indicator for spirometry confirmed COPD diagnosis 5 If smoking status is unrecorded in the first 5 years of follow-up, the most recent recording prior to ID1 will be used, assuming that there was no change in smoking behaviour 23

24 consecutive prescription of any of the drugs within 120 days 6 of the first repeat prescription. A switch to another type of monotherapy is allowed during follow-up as long as the second drug is prescribed within 120 days of the last prescription of the first drug and there is no consecutive prescription of the first drug. 2. Dual therapy, the presence of overlapping periods covered by 2 prescriptions for each of two inhalers, i.e. prescription for a second drug at index date while the first drug is still active (based on number of days prescribed at most recent prescription prior to index date) or prescribed at the same date. Drugs need to be followed by a repeat prescription within 30 days after end of prescription duration during follow-up. 3. Triple therapy (paragraph 4.2.3) Period of triple therapy A period of triple therapy is defined as the presence of overlapping periods covered by prescriptions for LAMA, LABA and ICS. The period starts at the date when the third drug was dispensed and the first two were still active (based on number of days prescribed at most recent prescription prior to index date) or prescribed at the same date. Drugs need to be followed by a repeat prescription within 30 days after end of prescription duration during follow-up. The period of triple therapy ends with absence of prescriptions for any of the three drugs for >120 days Period of poor adherence to maintenance inhalation therapy Objective 4 investigates patients with a period of poor adherence. This period will be defined as a Medication Possession Ratio <40% for LABA, LAMA and ICS in the last 3 consecutive years prior to ID High risk of rapid FEV1 decline The primary definition of rapid FEV1 decline will be a predicted decline of 40 ml/year over the next 11 years at ID1 following Zafari et al 7. The primary definition of high risk of rapid decline will be 80%. For objective 2, characteristics of patients at high risk of rapid decline will be described for the following combinations of probability of rate of decline: Probability of Cut-point definition rapid FEV1 decline (ml/year) rapid decline 25% 40 50% 40 75% 40 80% Prescription data in OPCRD and CPRD do not allow accurate estimation of the duration of a prescription for all patients. Therefore a standard duration of 30 days is assumed. Two consecutive prescriptions will be considered part of the same period if the gap between the end of this period of 30 days and the next prescription will be 90 days ( 120 days between two prescriptions) 24

25 4.4 Outcomes Primary end-point objective 4: Rate of FEV1 decline The change in FEV1 over time, expressed as ml/year of follow up, will be estimated using a multilevel model for change as described in paragraph Secondary endpoints objective 4: Matched comparison of patients on triple therapy with patients on minimal therapy The following secondary outcomes will be analysed: Time to first COPD exacerbation / Exacerbation rate Both the time to the first COPD exacerbation and the number of COPD exacerbations in the first year of follow-up after ID3 (for those who have complete follow-up) will be analysed as secondary outcomes of the analyses for objective 4. COPD exacerbations will be defined as the occurrence of at least one of the following events: COPD-related hospital attendance / admission AND/OR Respiratory-related A&E attendance AND/OR An acute oral corticosteroid (OCS) course AND/OR Antibiotics prescribed with evidence of a lower respiratory consultation Hospital admissions and A&E attendances will be identified from Read codes / Medcodes referring to a COPD specific admission / attendance or a general admission or attendance code with evidence of a lower respiratory consultation code on the same day (Appendix code lists for symptoms, COPD diagnostic codes and lower respiratory infections). A sensitivity analyses will be performed using HES APC and A&E data in the subpopulation of patients with HES linkage available. Using HES, COPD-related hospital admissions will be identified as hospital spells with ICD-10 codes J40-J44: any COPD-related code as primary diagnosis. Respiratory-related A&E attendances will be identified as events with respiratory conditions as diagnostic code (diag2=25) or ICD-10 codes J40-J44 (first 3 positions diag). Where 1 of these events occurred within 2 weeks of each other, they were considered to be the result of the same exacerbation (and only counted once). OCS or antibiotics prescribed at the same day as a COPD review performed within the framework of QOF will not be considered Time to first hospitalisation for COPD Hospital Episode Statistics (HES), linked to CPRD will be required to analyse the time to the first hospital admission for COPD as primary diagnosis, analysed as secondary outcome. Two definitions will be used to define occurrence of a hospital admission (spell) for COPD: a. ICD-10 codes J44.0 or J44.1: COPD exacerbations as primary diagnosis 25

26 b. ICD-10 codes J40-J44: any COPD-related code as primary diagnosis Treatment success Treatment success will be defined over the first year of follow-up as the absence of COPD exacerbations. 4.5 Other Variables and Covariates Baseline characteristics 1. Age group at index date (years): 2. Sex 3. Smoking status, Read code closest to and within 5 years prior to index date a. Ex-smoker b. Current smoker 4. BMI, calculated from height and weight data if available 7 and taken from practice recorded BMI value if not, within 10 years prior to index date a. Underweight: <18.5 b. Normal weight: 18.5 and <25 c. Overweight: 25 and <30 d. Obese: Socio-economic status (quintiles of practice-level index of Multiple Deprivation) for analyses on time to hospitalisation performed in CPRD data only Comorbidities 6. Comorbidities: a. Asthma ever (asthma or monitoring code recorded ever) b. Allergic / non-allergic rhinitis a. Never b. Active 8 c. Ever, not active c. Eczema diagnosis a. Never b. Active, defined a diagnostic Read code in baseline year c. Ever, not active d. Nasal polyps ever e. Chronic sinusitis diagnosis ever f. Gastro-oesophageal reflux disease (GERD) g. Diabetes Mellitus ever h. Osteoporosis ever i. Hypertension ever j. Ischaemic heart disease ever 7 Weight (kg) divided by height (metres) squared 8 Active refers to those for which a diagnosis was recorded within the baseline year and/or a prior diagnosis was accompanied by a prescription for the comorbidity within the baseline year. 26

27 k. Heart failure ever l. Depression ever m. Anxiety ever n. Obstructive Sleep apnoea ever o. Sleep disorders ever p. Chronic kidney disease ever 7. Charlson co-morbidity index: based on diagnoses ever a. 1 b. 2-4 c Clinical characteristics 8. Type of COPD diagnosed (Read code), Emphysema, Bronchitis or Unspecified/Other 9. Modified Medical Research Council (mmrc) dyspnoea score, within 5 years of the index date (see Appendix 7.5): a. 0-1 b. 2 c. Unknown 10. Degree of airflow limitation, FEV1% predicted, mean (SD) and categorical: a. Mild: FEV1% predicted 80% b. Moderate: FEV1% predicted 50 79% c. Severe: FEV1% predicted 30 49% d. Very severe: FEV1% predicted <30% The formula used to predict FEV1 will be 22 : For Males: Predicted FEV1= (4.30 x height (m)) (0.029 x age ) For Females: Predicted FEV1= (3.95 x height (m)) (0.025 x age ) 11. GOLD groups, defined based on a combined assessment of degree of dyspnoea and exacerbation risk following 2017 GOLD recommendations: a. GOLD A: 1 COPD exacerbation 9 and no hospitalization for COPD in baseline year and mmrc 0 1 b. GOLD B: 1 COPD exacerbation and no hospitalization for COPD in baseline year and mmrc 2 c. GOLD C: 2 COPD exacerbations or 1 hospitalization for COPD in baseline year) and mmrc COPD exacerbation, as the occurrence of one or more of the following events: a. COPD-related hospital attendance / admission AND/OR b. COPD-related A&E attendance AND/OR c. An acute oral corticosteroid course AND/OR d. Antibiotics prescribed with evidence of a lower respiratory consultation 27