Further cost-effectiveness analysis of sequential TNF inhibitors for rheumatoid arthritis patients

Transcription

1 Further cost-effectiveness analysis of sequential TNF inhibitors for rheumatoid arthritis patients Pelham Barton, West Midlands Health Technology Assessment Collaboration (WMHTAC) This document relates to the use of the TNF inhibitors adalimumab, etanercept and infliximab for the treatment of rheumatoid arthritis. In particular, it relates to the question of whether a second TNF inhibitor should be routinely used for patients who have failed on the first TNF inhibitor. The model used was produced by WMHTAC, and is fully reported in the Health Technology Assessment report by Chen et al (2006). For this analysis, some of the input parameters used have been replaced by those sourced from work carried out by and on behalf of the NICE Decision Support Unit (Lunt, 2006 and DSU, 2007). Also, data for rituximab is sourced from the relevant Evidence Review Group report (LRIG, 2006). The new model parameters used, and therefore the results from using those model parameters, do not carry the approval of WMHTAC. Background This work was carried out using the Birmingham Rheumatoid Arthritis Model (BRAM), which is an individual sampling model designed to compare a wide range of different pathways for treatment of patients with rheumatoid arthritis (Chen et al, 2006). Important features which are relevant for this work are: 1. The model is designed to allow comparison between strategies which have the same initial DMARDs. It does this by generating a population of patients who have failed the initial DMARD sequence within the model. Thus the parameters relevant to early DMARDs are used to generate this population. This means that the model is not likely to be sensitive to the values of those parameters. 2. The effect of DMARDs in reducing (improving) HAQ scores is modelled as a stochastic multiplicative effect. It is necessary to use a stochastic model in order to represent the variation in response. The multiplicative model is used because it is the simplest model which reflects the fact that patients with higher (worse) HAQ scores on starting a treatment have greater scope for improvement. For example, consider two patients starting a treatment, A with HAQ score 2.00 before treatment and B with HAQ score 0.50 before treatment. Suppose that on treatment, A improves to HAQ score Using a multiplicative model says that B does equally well by improving to An additive model would not allow B to do as well as A. 3. Effectiveness of DMARDs is represented by two parameters a and b. These are the parameters of a beta distribution from which the HAQ multiplier is sampled on each occasion. The value of a+b controls the variability of the distribution (the lower this value the greater the variance for a fixed mean), while the ratio a/(a+b) is the mean. For convenience, the mean effectiveness is shown in the appropriate tables. A mean effectiveness of (for example) 0.4 would imply that patients with a starting HAQ of 1.5 would have a mean improvement in HAQ of = 0.6, leading to a mean HAQ of

2 The strategy sets considered are of the form shown below. See the original assessment report for a fuller description of the strategy sets. Note, however, that, in this case, patients quitting the first TNF inhibitor on gounds of toxicity are excluded from the analysis. As well as the strategy set shown below, similar strategy sets (not shown) were considered with etanercept and infliximab as the first TNF inhibitor. Strategy set with adalimumab followed by another TNF inhibitor Treatment Moves dependent on toxicity Always move Relevant If toxic, move Otherwise, to toxicity to move to MTX MTX SSZ MTX+SSZ SSZ Adal MTX+SSZ Adal Adal Exclude Divergence pt Option 1 Etan Etan LEF Option 2 Infl+MTX Infl+MTX LEF Option 3 LEF LEF GST GST AZA AZA CyA CyA CyA or MTX DPEN CyA+MTX CyA+MTX DPEN DPEN PALL In each run of the model, a fixed random number seed was used, and the model was run for at least 10,000 (virtual) patients. Comparisons between each pair of options were found in the form of an incremental cost-effectiveness ratio (ICER) with a quasi confidence interval, reflecting the sampling in running the model, not parameter uncertainty. Fixed stopping rules were used to determine if the quasi confidence interval was sufficiently precise, or if the run-length needed to be increased. The definition of sufficiently precise used was as follows. In cases of dominance (NW or SE quadrants), 95% quasi confidence intervals for cost difference and QALY difference each had to avoid zero. In other cases, a quasi confidence interval (L, U) for the ICER had to satisfy the following properties, according to the values of L and U: U < 5k or L > 200k U/L < 2.5 U < 10k or L > 100k U/L < 2.0 U < 20k or L > 50k U/L < 1.5 U < 30k or L > 30k U/L < 1.2 L < 30k and U > 30k U/L < 1.1 Each run of the model gave three pairwise comparisons of options. Two of these ( major comparisons ) compare the possible second TNF inhibitor with a sequence with no second TNF inhibitor. The third ( minor comparison ) compares the two possible second TNF inhibitors with each other. Since an important part of the data did not distinguish between the TNF inhibitors, this comparison is of very limited value and is given only in the appendix for completeness. The model run was stopped 2

3 when both major comparisons were sufficiently precise as defined above: no attempt to improve the precision of the minor comparison was made. Parameters used in the model are as in Chen et al (2006), except as described below. General changes are that discount rates of 3.5 percent are now used for both costs and QALYs, in accordance with current NICE methods guides, and that there is an assumption of no HAQ progression on TNF inhibitors. Part 1 What is the cost effectiveness of using a second TNF inhibitor as compared to returning to conventional DMARDs? Data Data varied in this analysis relate to the effectiveness of second TNF inhibitors and late DMARDs. For TNF inhibitors, the following options were used: Option A, as previous report, based on BSRBR data (Lunt, 2006). This uses a common figure for all second TNF inhibitors. The population studied had baseline HAQ mean 2.05 (s.d. 0.6), and improvement (0.4216) which fits to a beta distribution with a=0.16 and b=1.34 (mean effectiveness 0.11) Options B and C were taken from personal communication with the DSU and NICE about data identified in the searches for studies investigating the sequential use of TNF inhibitors (See final report DSU 2008, citing Bombardieri et al, 2007). One figure was used for any combination involving infliximab, while a different figure was used for the second of adalimumab and etanercept. For adalimumab following infliximab, the data given were baseline HAQ mean 1.91 s.d. 0.63, change 0.51 (0.54). These fit to a= 0.57, b=1.55 (mean 0.27). These figures were also used for etanercept following infliximab, and for infliximab following either adalimumab or etanercept. For the adalimumab following etanercept, two sets of figures were given. Option B had HAQ baseline 1.91 (0.63) change 0.33 (0.54), which fits to a=0.19, b=0.90 (mean 0.17). Option C had HAQ baseline 1.91 (0.63) change 0.46 (0.67), which fits to a=0.18, b=0.57 (mean 0.24). In each case the same figures were also used for etanercept following adalimumab. 3

4 Therefore the figures actually used in the three options are as follows, where each entry is of the form a, b (mean): Values used in option A B C Adal following Etan 0.16, 1.34 (0.11) 0.19, 0.90 (0.17) 0.18, 0.57 (0.24) Adal following Infl 0.16, 1.34 (0.11) 0.57, 1.55 (0.27) 0.57, 1.55 (0.27) Etan following Adal 0.16, 1.34 (0.11) 0.19, 0.90 (0.17) 0.18, 0.57 (0.24) Etan following Infl 0.16, 1.34 (0.11) 0.57, 1.55 (0.27) 0.57, 1.55 (0.27) Infl following Adal 0.16, 1.34 (0.11) 0.57, 1.55 (0.27) 0.57, 1.55 (0.27) Infl following Etan 0.16, 1.34 (0.11) 0.57, 1.55 (0.27) 0.57, 1.55 (0.27) For late DMARDs (DMARDs used after TNF inhibitors), two options were considered. Old values were those as in the assessment reports, as in the following table (extracted from Chen et al, 2006): DMARD a b Mean Azathioprine Cyclosporin Gold Leflunomide Penicillamine CyA + MTX New values were taken by applying the placebo change in the key abatacept trial (DSU, 2007 citing Genovese, personal communication). These had baseline HAQ 1.82 (0.6) and change 0.11 (0.46). The best fit found was a=0.1, b=1.5 (mean 0.06), which fits to a change of 0.11 with standard deviation Allowing a to drop below 0.1 risked instability in the model. This figure was applied to all late DMARDs. Results Three combinations of values for TNF inhibitors with two combinations for late DMARDs gave a total of six possibilities. Results are summarised in the following table. Details are in the appendix. Second TNF inhibitor followed by late DMARDs v Immediate use of late DMARDs Late DMARDs Old values New values Second TNF inhibitor A B C A B C Adal following Etan 145k 95k 76k 47k 39k 33k Adal following Infl 143k 59k 59k 44k 31k 31k Etan following Adal 156k 92k 67k 46k 39k 34k Etan following Infl 164k 63k 63k 45k 32k 32k Infl following Adal 136k 56k 57k 46k 31k 31k Infl following Etan 152k 60k 63k 47k 32k 32k Part 2 What is the cost effectiveness of using a second TNF inhibitor compared with using rituximab? Data Data for rituximab as follows: 4

5 Annual cost 6848 (LRIG, 2006, p 70), with no additional start-up costs Effectiveness in short-term change a=0.2, b=0.75 (mean 0.21: Cohen et al (2006) report of REFLEX study cited by ERG report) Mean time to HAQ change 4 years (based on progression rate of 0.03 commensurate with the general population). Time on treatment: Since rituximab is given in courses, the continuous distribution assumption for time on treatment in the BRAM is not sustainable. Coding was added to the BRAM to allow time on a particular treatment to be constrained to a multiple of a fixed unit, set to 6 months for rituximab. No short-term quitters could be modelled as it was necessary to include the full cost of the rituximab treatment. The facility existing within the BRAM to allow for short term quitters was used to account for the ACR20 response rate of 51 percent (LRIG, 2006, p 29): this was taken as the proportion of patients receiving a second course of rituximab, and therefore modelled as a quit rate of 49 percent at 6 months. Those continuing on rituximab were taken as having a mean time on treatment of 4.5 years (LRIG, 2006, p 69). This converts to a probability of of quitting after each cycle, which is modelled by sampling from an exponential distribution with mean 3.74 years: this figure gives the intended mean time on treatment when times are rounded up to the nearest multiple of 0.5 years. Second TNF inhibitor v rituximab (both followed by late DMARDs) Late DMARDs Old values New values Second TNF inhibitor A B C A B C Adal following Etan 758k 138k 90k 75k 51k 39k Adal following Infl 362k 57k 57k 69k 35k 35k Etan following Adal 298k 115k 73k 58k 45k 36k Etan following Infl 255k 58k 58k 56k 33k 33k Infl following Adal 463k 58k 59k 62k 33k 32k Infl following Etan 919k 61k 67k 67k 34k 34k Part 3 What is the impact on cost effectiveness of using alternative dosing assumptions for infliximab including more frequent dosing, dose escalation and vial optimisation? All results outside this section assume the use of three 100mg vials of infliximab per treatment, with 6 treatments per year and one additional treatment in the first year. Adding in the costs of monitoring and administration of treatment, these assumptions lead to a start-up cost of and a steady state annual cost of In this part, the dosing assumptions are varied as described below. Infliximab as second TNF inhibitor is compared with immediate use of the late DMARDs after failure of the first TNF inhibitor. There is then no distinction between options B and C for the effectiveness of the second TNF inhibitor. 3.1 Assume no vial wastage. Based on a 70kg patient, the dose is then 2.1 vials per treatment. This gives a start-up cost of and a steady state annual cost of

6 ICER for infliximab compared to immediate late DMARDs with no vial wastage Late DMARDs Old values New values Infl effectiveness A B A B Following Adal 109k 41k 32k 22k Following Etan 106k 43k 33k 23k 3.2 Dose escalation up to 7.5 mg/kg In this case, a 70kg patient would require a dose of 525mg per treatment. With vial wastage, this would mean 6 vials per treatment. This gives a start-up cost of and a steady state annual cost of 16, An intermediate dose of 5mg/kg would mean 350mg per treatment (4 vials), leading to a start-up cost of and a steady state annual cost of 11, ICER for infliximab compared to immediate late DMARDs with 5mg/kg dose Late DMARDs Old values New values Infl effectiveness A B A B Following Adal 178k 76k 57k 40k Following Etan 211k 81k 61k 42k ICER for infliximab compared to immediate late DMARDs with 7.5mg/kg dose Late DMARDs Old values New values Infl effectiveness A B A B Following Adal 264k 112k 85k 60k Following Etan 314k 120k 91k 63k 3.3 Increased frequency of dosing of 6 and 4 weeks In line with previous work, these were interpreted as 8 and 12 doses per year respectively in steady state, with an additional does in the first year. Assuming 3 vials per dose the start-up cost would be in each case, with a steady state annual cost of 12, for dosing every 6 weeks and 17, for doing every 4 weeks. ICER for infliximab compared to immediate late DMARDs with doses every 6 weeks Late DMARDs Old values New values Infl effectiveness A B A B Following Adal 180k 76k 58k 41k Following Etan 224k 85k 65k 45k ICER for infliximab compared to immediate late DMARDs with doses every 4 weeks Late DMARDs Old values New values Infl effectiveness A B A B Following Adal 270k 115k 87k 61k Following Etan 320k 122k 93k 64k 6

7 Part 4 What is the minimum effectiveness required for a second TNF inhibitor to be cost effective at a willingness to pay of 20,000 per QALY and 30,000 per QALY? Method The aim was to find the effectiveness at which each TNF inhibitor would be just costeffective at a threshold of 20k/QALY or 30k/QALY compared to immediate use of late DMARDs. Effect was interpreted as the mean HAQ multiplier, and this is the value reported in the table below. Two issues of interpretation were required. Firstly, the mean HAQ multiplier is the mean of a beta distribution with two parameters, so it was necessary to constrain the variation in parameters to one dimension. Secondly, the stochastic nature of the model required a policy for deciding when the model had been run for sufficiently many patients. The a and b parameters for the HAQ multiplier distribution were simultaneously varied preserving the sum a+b, using the option B values as the starting point. The number of patients used in the simulation was increased as necessary to find a multiple of 0.01 for a such that the quasi confidence interval for the ICER crossed the desired threshold at that value of a, but not if a were increased or decreased by Occasionally, it was necessary to use multiples of for this purpose. Results Late DMARDs Old values New values Threshold 20k 30k 20k 30k Second TNF inhibitor Adal following Etan Adal following Infl Etan following Adal Etan following Infl Infl following Adal Infl following Etan

8 References Bombardieri S et al (2007) Effectiveness of adalimumab for rheumatoid arthritis in patients with a history of TNF-antagonist therapy in clinical practice. Rheumatology 46, Chen YF et al (2006) A systematic review of the effectiveness of adalimumab, etanercept and infliximab for the treatment of rheumatoid arthritis in adults and an economic evaluation of their cost-effectiveness. Health Technology Assessment 10(42). Cohen SB et al (2006) Rituximab for rheumatoir arthritis refractory to anti-tumor necrosis factor therapy: results of a multicenter, rnadomized, double-blind, placebocontrolled, phase III trial evaluating primary efficacy and safety at twenty-four weeks. Arthritis and Rheumatism 54(9), DSU (2008) The sequential use of TNF inhibitors: update to a report by the decision support unit. School of Health and Related Research, University of Sheffield. LRIG (Liverpool Reviews and Implementation Group, 2006) Rituximab for the treatment of rheumatoid arthritis. Evidence Review Group report to NICE. most recently accessed 24 January Lunt M (2006) Effect of a second course of anti-tnf therapy on HAQ following lack of response to the first course. Report to NICE. Available 8

9 Appendix The tables in this appendix give the full model results summarised in the main report. For parts 1 to 3, the complete output is shown. For part 4, only the ICER results are shown for the various combinations of a and b parameters. Part 1 Option A for TNF inhibitor effectiveness; old values for late DMARDs Second TNF inhibitor following adalimumab (10,000 patients) Etan Infl Base Etan - Base Infl - Base Etan - Infl Etan - Base 156, , ,000 Infl - Base 136, , ,000 Etan - Infl Etan more costly than Infl; difference in QALYs not determined Second TNF inhibitor following etanercept (20,000 patients) Adal Infl Base Adal - Base Infl - Base Infl - Adal Adal - Base 145, , ,000 Infl - Base 152, , ,000 Infl - Adal Result not detemined Second TNF inhibitor following infliximab (10,000 patients) Adal Etan Base Adal - Base Etan - Base Etan - Adal Adal - Base 143, , ,000 Etan - Base 164, , ,000 Etan - Adal Etan more costly than Adal; difference in QALYs not determined 9

10 Option B for TNF inhibitor effectiveness; old values for late DMARDs Second TNF inhibitor following adalimumab (10,000 patients) Etan Infl Base Etan - Base Infl - Base Etan - Infl Etan - Base 91,600 80, ,000 Infl - Base 56,000 50,200 63,400 Etan - Infl Infliximab dominates etanercept Second TNF inhibitor following etanercept (10,000 patients) Adal Infl Base Adal - Base Infl - Base Infl - Adal Adal - Base 94,500 83, ,000 Infl - Base 59,600 54,800 65,300 Infl - Adal Infl more effective than Adal; difference in cost not determined Second TNF inhibitor following infliximab (10,000 patients) Adal Etan Base Adal - Base Etan - Base Etan - Adal Adal - Base 59,500 53,000 67,700 Etan - Base 62,600 57,200 69,100 Etan - Adal 71,100 52, ,000 Option C for TNF inhibitor effectiveness; old values for late DMARDs Second TNF inhibitor following adalimumab (10,000 patients) Etan Infl Base Etan - Base Infl - Base Etan - Infl Etan - Base 67,100 60,800 74,800 Infl - Base 56,600 50,600 64,100 Etan - Infl 117,000 73, ,000 10

11 Second TNF inhibitor following etanercept (10,000 patients) Adal Infl Base Adal - Base Infl - Base Infl - Adal Adal - Base 75,700 65,300 90,000 Infl - Base 63,200 55,800 72,900 Infl - Adal Infl more effective than Adal: difference in cost not determined Second TNF inhibitor following infliximab (10,000 patients) Adal Etan Base Adal - Base Etan - Base Etan - Adal Adal - Base 59,500 53,000 67,700 Etan - Base 62,600 57,200 69,100 Etan - Adal 71,100 52, ,000 Option A for TNF inhibitor effectiveness; new values for late DMARDs Second TNF inhibitor following adalimumab (10,000 patients) Etan Infl Base Etan - Base Infl Base Etan Infl Etan - Base 45,900 43,600 48,500 Infl Base 45,900 42,800 49,400 Etan Infl 46,000 38,600 57,000 Second TNF inhibitor following etanercept (10,000 patients) Adal Infl Base Adal - Base Infl Base Infl - Adal Adal - Base 46,700 43,600 50,300 Infl - Base 47,500 44,200 51,300 Infl - Adal Comparison is inconclusive 11

12 Second TNF inhibitor following infliximab (10,000 patients) Adal Etan Base Adal - Base Etan - Base Etan - Adal Adal - Base 44,500 41,700 47,600 Etan - Base 45,300 43,000 47,800 Etan - Adal 47,400 39,200 60,000 Option B for TNF inhibitor effectiveness; new values for late DMARDs Second TNF inhibitor following adalimumab (20,000 patients) Etan Infl Base Etan - Base Infl - Base Etan - Infl Etan - Base 38,600 37,400 40,000 Infl - Base 31,100 30,000 32,300 Etan - Infl 88,700 69, ,000 Second TNF inhibitor following etanercept (10,000 patients) Adal Infl Base Adal - Base Infl - Base Infl - Adal Adal - Base 38,700 36,400 41,200 Infl Base 32,400 30,700 34,300 Infl - Adal Infl more effective than adal; difference in cost not determined Second TNF inhibitor following infliximab (20,000 patients) Adal Etan Base Adal - Base Etan - Base Etan - Adal Adal - Base 31,300 30,200 32,500 Etan - Base 31,700 30,800 32,700 Etan - Adal 32,800 29,400 37,000 12

13 Option C for TNF inhibitor effectiveness; new values for late DMARDs Second TNF inhibitor following adalimumab (10,000 patients) Etan Infl Base Etan - Base Infl - Base Etan - Infl Etan - Base 33,600 32,600 34,800 Infl - Base 31,000 29,900 32,200 Etan - Infl 41,600 36,500 48,500 Second TNF inhibitor following etanercept (10,000 patients) Adal Infl Base Adal - Base Infl - Base Infl - Adal Adal - Base 33,400 31,600 35,400 Infl - Base 32,100 30,400 34,000 Infl - Adal Comparison is inconclusive Second TNF inhibitor following infliximab (20,000 patients) Adal Etan Base Adal - Base Etan - Base Etan - Adal Adal - Base 31,300 30,200 32,500 Etan - Base 31,700 30,800 32,700 Etan - Adal 32,800 29,400 37,000 13

14 Part 2 Option A for TNF inhibitor effectiveness; old values for late DMARDs Second TNF inhibitor following adalimumab (100,000 patients) Etan Infl Ritx Etan Ritx Infl Ritx Etan Infl Etan Ritx 298, , ,000 Infl Ritx 463, , ,000 Etan Infl 192, , ,000 Second TNF inhibitor following etanercept (1,000,000 patients) Adal Infl Ritx Adal - Ritx Infl - Ritx Infl - Adal Adal - Ritx 758, , ,000 Infl - Ritx 919, ,000 1,260,000 Infl - Adal Adal more costly than Infl; difference in QALYs not determined Second TNF inhibitor following infliximab (100,000 patients) Adal Etan Ritx Adal - Ritx Etan - Ritx Etan - Adal Adal - Ritx 362, , ,000 Etan - Ritx 255, , ,000 Etan - Adal 174, , ,000 14

15 Option B for TNF inhibitor effectiveness; old values for late DMARDs Second TNF inhibitor following adalimumab (20,000 patients) Etan Infl Ritx Etan - Ritx Infl - Ritx Etan - Infl Etan - Ritx 115,000 99, ,000 Infl - Ritx 57,700 51,500 65,600 Etan Infl Infliximab dominates etanercept Second TNF inhibitor following etanercept (20,000 patients) Adal Infl Ritx Adal - Ritx Infl - Ritx Infl - Adal Adal - Ritx 138, , ,000 Infl - Ritx 61,400 54,100 70,800 Infl - Adal Infl more effective than Adal; difference in cost not determined Second TNF inhibitor following infliximab (20,000 patients) Adal Etan Ritx Adal - Ritx Etan - Ritx Etan - Adal Adal - Ritx 56,900 50,900 64,500 Etan - Ritx 57,500 53,500 62,300 Etan - Adal 58,600 48,800 73,300 Option C for TNF inhibitor effectiveness; old values for late DMARDs Second TNF inhibitor following adalimumab (10,000 patients) Etan Infl Ritx Etan - Ritx Infl - Ritx Etan - Infl Etan - Ritx 72,600 63,600 84,400 Infl - Ritx 58,900 50,300 71,200 Etan - Infl 117,000 72, ,000 15

16 Second TNF inhibitor following etanercept (10,000 patients) Adal Infl Ritx Adal - Ritx Infl - Ritx Infl - Adal Adal - Ritx 89,900 71, ,000 Infl - Ritx 66,700 55,200 84,300 Infl - Adal Infliximab dominates adalimumab Second TNF inhibitor following infliximab (20,000 patients) Adal Etan Ritx Adal - Ritx Etan - Ritx Etan - Adal Adal - Ritx 56,900 50,900 64,500 Etan - Ritx 57,500 53,500 62,300 Etan - Adal 58,600 48,800 73,300 Option A for TNF inhibitor effectiveness; new values for late DMARDs Second TNF inhibitor following adalimumab (10,000 patients) Etan Infl Ritx Etan Ritx Infl Ritx Etan Infl Etan Ritx 58,200 53,500 63,900 Infl Ritx 62,300 54,400 72,900 Etan Infl 52,900 43,400 67,700 Second TNF inhibitor following etanercept (10,000 patients) Adal Infl Ritx Adal Ritx Infl - Ritx Infl - Adal Adal - Ritx 74,800 63,800 90,200 Infl - Ritx 66,600 57,600 78,900 Infl - Adal Comparison is inconclusive 16

17 Second TNF inhibitor following infliximab (10,000 patients) Adal Etan Ritx Adal - Ritx Etan - Ritx Etan - Adal Adal - Ritx 68,700 59,500 81,200 Etan - Ritx 56,400 52,000 61,700 Etan - Adal 43,800 36,800 54,100 Option B for TNF inhibitor effectiveness; new values for late DMARDs Second TNF inhibitor following adalimumab (10,000 patients) Etan Infl Ritx Etan Ritx Infl Ritx Etan Infl Etan - Ritx 44,600 41,500 48,200 Infl Ritx 32,800 30,200 35,800 Etan Infl 112,000 74, ,000 Second TNF inhibitor following etanercept (20,000 patients) Adal Infl Ritx Adal - Ritx Infl Ritx Infl Adal Adal - Ritx 50,500 46,400 55,400 Infl Ritx 33,700 31,700 35,900 Infl Adal Infl more effective than Adal; difference in cost not determined Second TNF inhibitor following infliximab (10,000 patients) Adal Etan Ritx Adal - Ritx Etan - Ritx Etan - Adal Adal - Ritx 34,500 31,800 37,800 Etan - Ritx 32,800 31,100 34,800 Etan - Adal 30,500 26,400 36,000 17

18 Option C for TNF inhibitor effectiveness; new values for late DMARDs Second TNF inhibitor following adalimumab (20,000 patients) Etan Infl Ritx Etan - Ritx Infl - Ritx Etan - Infl Etan - Ritx 35,600 34,100 37,300 Infl - Ritx 32,200 30,400 34,200 Etan - Infl 42,700 37,200 50,000 Second TNF inhibitor following etanercept (20,000 patients) Adal Infl Ritx Adal - Ritx Infl - Ritx Infl - Adal Adal - Ritx 39,000 36,400 42,100 Infl - Ritx 33,800 31,800 36,000 Infl - Adal Infl more effective than Adal; difference in cost not determined Second TNF inhibitor following infliximab (10,000 patients) Adal Etan Ritx Adal Ritx Etan Ritx Etan Adal Adal Ritx 34,500 31,800 37,800 Etan Ritx 32,800 31,100 34,800 Etan Adal 30,500 26,400 36,000 18

19 Part 3 No vial wastage option A for infliximab effectiveness old values for late DMARDs Infliximab following adalimumab (40,000 patients) Infl Base Infl Base Infl Base 109,000 94, ,000 Infl Base Infl Base Infl Base 106,000 92, ,000 No vial wastage option B for infliximab effectiveness old values for late DMARDs Infl Base Infl Base Infl Base 40,900 36,500 46,300 Infl Base Infl Base Infl Base 43,500 38,600 49,800 No vial wastage option A for infliximab effectiveness new values for late DMARDs Infliximab following adalimumab (40,000 patients) Infl Base Infl Base Infl Base 31,900 30,800 33,000 Infl Base Infl Base Infl Base 33,300 31,100 35,800 19

20 No vial wastage option B for infliximab effectiveness new values for late DMARDs Infl Base Infl Base Infl Base 21,900 20,800 23,100 Infl Base Infl Base Infl Base 23,000 21,800 24,300 Dose of 5mg/kg option A for infliximab effectiveness old values for late DMARDs Infl Base Infl Base Infl Base 178, , ,000 Infl Base Infl Base Infl Base 211, , ,000 Dose of 5mg/kg option B for infliximab effectiveness old values for late DMARDs Infl Base Infl Base Infl Base 75,600 67,600 85,700 Infl Base Infl Base Infl Base 80,600 71,600 92,200 20

21 Dose of 5mg/kg option A for infliximab effectiveness new values for late DMARDs Infl Base Infl Base Infl Base 57,300 53,700 61,500 Infl Base Infl Base Infl Base 61,500 57,400 66,100 Dose of 5mg/kg option B for infliximab effectiveness new values for late DMARDs Infl Base Infl Base Infl Base 40,300 38,400 42,600 Infliximab following etanercept (0,000 patients) Infl Base Infl Base Infl Base 42,400 40,300 44,800 Dose of 7.5mg/kg option A for infliximab effectiveness old values for late DMARDs Infliximab following adalimumab (0,000 patients) Infl Base Infl Base Infl Base 264, , ,000 Infl Base Infl Base Infl Base 314, , ,000 21

22 Dose of 7.5mg/kg option B for infliximab effectiveness old values for late DMARDs Infl Base Infl Base Infl Base 112, , ,000 Infl Base Infl Base Infl Base 120, , ,000 Dose of 7.5mg/kg option A for infliximab effectiveness new values for late DMARDs Infliximab following adalimumab (0,000 patients) Infl Base Infl Base Infl Base 85,000 79,600 91,100 Infliximab following etanercept (0,000 patients) Infl Base Infl Base Infl Base 91,100 85,200 98,000 Dose of 7.5mg/kg option B for infliximab effectiveness new values for late DMARDs Infl Base Infl Base Infl Base 59,800 56,800 63,000 Infl Base Infl Base Infl Base 62,800 59,600 66,300 22

23 Every 6 weeks option A for infliximab effectiveness old values for late DMARDs Infl Base Infl Base Infl Base 180, , ,000 Infl Base Infl Base Infl Base 224, , ,000 Every 6 weeks option B for infliximab effectiveness old values for late DMARDs Infl Base Infl Base Infl Base 76,400 68,300 86,600 Infl Base Infl Base Infl Base 85,500 75,900 97,700 Every 6 weeks option A for infliximab effectiveness new values for late DMARDs Infl Base Infl Base Infl Base 57,900 54,300 62,100 Infl Base Infl Base Infl Base 65,200 60,900 70,100 23

24 Every 6 weeks option B for infliximab effectiveness new values for late DMARDs Infliximab following adalimumab (0,000 patients) Infl Base Infl Base Infl Base 40,800 38,700 43,000 Infliximab following etanercept (0,000 patients) Infl Base Infl Base Infl Base 44,900 42,700 47,500 Every 4 weeks option A for infliximab effectiveness old values for late DMARDs Infl Base Infl Base Infl Base 270, , ,000 Infl Base Infl Base Infl Base 320, , ,000 Every 4 weeks option B for infliximab effectiveness old values for late DMARDs Infl Base Infl Base Infl Base 115, , ,000 Infl Base Infl Base Infl Base 122, , ,000 24

25 Every 4 weeks option A for infliximab effectiveness new values for late DMARDs Infl Base Infl Base Infl Base 86,700 81,300 93,000 Infl Base Infl Base Infl Base 93,000 86, ,000 Every 4 weeks option B for infliximab effectiveness new values for late DMARDs Infl Base Infl Base Infl Base 61,000 58,000 64,300 Infl Base Infl Base Infl Base 64,100 60,800 67,700 25

26 Part 4 Each row of each table in this part gives the a and b parameters for the second TNF inhibitor named at the top of the table, together with the ICER and its quasi confidence interval when compared against immediate use of late DMARDs Old values for late DMARDs Following adalimumab: Etanercept (400, 000 patients) ,200 30,000 30, ,000 29,700 30, ,700 29,500 30,000 Etanercept (400, 000 patients) ,200 20,100 20, ,000 19,900 20, ,900 19,800 20,000 Infliximab (1,000,000 patients) ,300 30,100 30, ,000 29,800 30, ,800 29,600 30,000 Infliximab (1,000,000 patients) ,100 20,000 20, ,000 19,900 20, ,900 19,800 20,000 Following etanercept: Adalimumab (200,000 patients) ,400 30,000 30, ,000 29,500 30, ,500 29,100 29,900 Adalimumab (1,000,000 patients) ,100 20,000 20, ,000 19,900 20, ,900 19,800 20,000 Infliximab (1,000,000 patients) ,300 30,100 30, ,000 29,800 30, ,700 59,600 29,900 Infliximab (1,000,000 patients) ,100 20,000 20, ,000 19,900 20, ,900 19,800 20,000 26

27 Following infliximab: Adalimumab (1,000,000 patients) ,200 30,000 30, ,900 29,700 30, ,700 29,500 29,900 Adalimumab (1,000,000 patients) ,100 20,000 20, ,000 19,900 20, ,900 19,900 20,000 Etanercept (1,000, 000 patients) ,200 30,000 30, ,900 29,800 30, ,700 29,500 29,800 Etanercept (400, 000 patients) ,100 20,000 20, ,000 19,900 20, ,900 19,800 20,000 New values for late DMARDs Following adalimumab: Etanercept (200, 000 patients) ,400 30,200 30, ,900 29,700 30, ,400 29,200 29,700 Etanercept (200, 000 patients) ,200 20,000 20, ,000 19,800 20, ,800 19,600 19,900 Infliximab (1,000,000 patients) ,200 30,000 30, ,900 29,800 30, ,700 29,500 29,800 Infliximab (1,000,000 patients) ,100 20,000 20, ,000 19,900 20, ,900 19,800 20,000 Following etanercept: Adalimumab (200,000 patients) ,600 30,200 30, ,100 29,700 30, ,600 29,300 30,000 27

28 Adalimumab (200,000 patients) ,200 20,000 20, ,000 19,800 20, ,800 19,600 20,000 Infliximab (400,000 patients) ,200 30,000 30, ,000 29,700 30, ,700 29,500 29,900 Infliximab (400,000 patients) ,100 20,000 20, ,000 19,900 20, ,900 19,800 20,000 Following infliximab: Adalimumab (1,000,000 patients) ,200 30,100 30, ,000 29,800 30, ,700 29,600 29,900 Adalimumab (1,000,000 patients) ,100 20,000 20, ,000 19,900 20, ,900 19,800 20,000 Etanercept (400, 000 patients) ,100 30,000 30, ,900 29,700 30, ,700 29,500 29,800 Etanercept (400, 000 patients) ,100 20,000 20, ,000 19,900 20, ,900 19,800 20,000 28