Key words: rheumatoid arthritis, anti-tnf, methotrexate, outcomes research, randomized controlled trials

Transcription

1 RHEUMATOLOGY Rheumatology 2017;56: doi: /rheumatology/kew467 Advance Access publication 7 January 2017 Original article Efficacy and effectiveness of tumour necrosis factor inhibitors in the treatment of rheumatoid arthritis in randomized controlled trials and routine clinical practice Kalle J. Aaltonen 1, Suvi Ylikylä 1, Jaana Tuulikki Joensuu 1, Pia Isomäki 2,3, Laura Pirilä 4, Markku Kauppi 3,5, Tuomas Rannio 6,7, Kari Eklund 8,9, Marja Blom 1 and Dan Nordström 8,9 Abstract Objective. Efficacy of TNF inhibitors in the treatment of RA assessed in randomized controlled trials (RCTs) may not be fully comparable to routine care owing to the stringent inclusion criteria. The objective of this study was to observe the effectiveness of TNF inhibitors in real-world patients and assess the patients potential eligibility for the RCTs. Methods. RA patients starting a TNF-inhibitor treatment between 2004 and 2014 were identified from the National Register for Biologic Treatment in Finland, which is a longitudinal observational cohort study. Effectiveness was measured using the ACR and EULAR response criteria and by studying the proportion of patients reaching DAS28 remission. The patients baseline characteristics were compared against the inclusion criteria of 27 RCTs. Results. EULAR moderate and good treatment responses at 6 months were achieved by 69 and 40% of the users of the first TNF inhibitor, respectively. ACR20, ACR50 and ACR70 responses were reached by 48, 27 and 13%, respectively. DAS28 remission was reached by 47%. Only % of the patients would have been potentially eligible for the RCTs. The eligible patients had better treatment responses compared with the non-eligible patients. Different TNF inhibitors were mostly equipotent, but the usage of MTX cotherapy had a major influence on treatment response. Conclusion. Only a small proportion of patients would have been eligible for RCTs, and the efficacy of TNF inhibitors assessed in them cannot be generalized directly into Finnish routine health care. Key words: rheumatoid arthritis, anti-tnf, methotrexate, outcomes research, randomized controlled trials Rheumatology key messages. Different TNF inhibitors in the treatment of RA are equipotent in routine clinical practice.. Concomitant MTX improves the treatment effectiveness in RA.. The results of the randomized controlled trials on TNF inhibitors in RA have limited generalizability into routine health care. CLINICAL SCIENCE 1 Faculty of Pharmacy, University of Helsinki, Helsinki, 2 Department of Internal Medicine, Centre for Rheumatic Diseases, Tampere University Hospital, 3 School of Medicine, University of Tampere, Tampere, 4 Department of Rheumatology, Turku University Hospital, Turku, 5 Department of Rheumatology, Päijät-Häme Central Hospital, Lahti, 6 Department of Rheumatology, Jyväskylä Central Hospital, Jyväskylä, 7 Faculty of Health Sciences, University of Eastern Finland, Kuopio, 8 Department of Medicine, Helsinki University Central Hospital and 9 Faculty of Medicine, University of Helsinki, Helsinki, Finland Submitted 15 March 2016; revised version accepted 18 November 2016 Correspondence to: Kalle J. Aaltonen, University of Helsinki, Faculty of Pharmacy, Biocenter 1, PO Box 56 (Viikinkaari 5), University of Helsinki, Finland. kalle.jm.aaltonen@helsinki.fi! The Author Published by Oxford University Press on behalf of the British Society for Rheumatology. All rights reserved. For Permissions, please journals.permissions@oup.com

2 Kalle J. Aaltonen et al. Introduction According to the Finnish current care guidelines updated in 2015, the treatment of RA is initiated with a combination of MTX, SSZ, HCQ and low-dose glucocorticoid [1]. In case of insufficient treatment response or intolerance to conventional synthetic DMARDs (csdmards), biologic DMARDs (bdmards) may be added to the treatment. Nevertheless, it is still considered beneficial to continue using MTX concomitantly with a bdmard owing to better effectiveness [2]. Most of the information on the efficacy and safety of biologic treatments has been derived from randomized controlled trials (RCTs). Although RCTs can provide high-quality information, their stringent inclusion criteria for patients and often brief follow-up times limit the generalizability of the results to routine care [3]. Observational trials based on either retrospective, administrative healthcare data or purpose-collected prospective data can provide information based on the true use of medicines among real patients. However, observational studies are prone to various types of biases, which reflect the lack of randomization and the quality and completeness of the data. The proportion of TNF-inhibitor users achieving ACR50 and moderate EULAR responses and DAS28 remission at 6 12 months after treatment onset has ranged from 21 to 66, 55 to 90 and 7 to 66%, depending on the patients baseline disease activity, concomitant csdmard use and the trial protocol [4, 5]. Previous studies have shown that 22 80% of RA patients treated with TNF inhibitors in routine clinical practice would have been eligible for RCTs and that the observed treatment effectiveness was better among the eligible patients compared with those not eligible [3, 6]. Results based on previous observational studies are very heterogeneous, with the fraction of EULAR good and moderate responders on the first course of bdmard therapy ranging from 11 to 53 and 50 to 93%, respectively [3, 6 11]. Even greater variation, 4 50%, is present in the fraction of patients achieving DAS28 remission. The objective of this study was to assess the effectiveness of TNF inhibitors in routine clinical practice and study the effect of concomitant csdmard therapy on it. As a secondary outcome, we compared the baseline characteristics of TNF-inhibitor users with the inclusion criteria of RCTs to find out how large a proportion of Finnish patients treated in routine clinical practice would have been eligible for the RCTs and whether there is a difference in the effectiveness of TNF inhibitors between eligible and noneligible patients. Methods Patients were identified from National Register for Biologic Treatment in Finland (ROB-FIN), which is a longitudinal cohort study comprising data on patients diagnosed with rheumatic diseases treated with bdmards in routine clinical practice. Data are collected on routine care visits to rheumatologists, which usually take place 3 and 6 months after the onset of TNF therapy and biannually thereafter. RA patients having started TNF-inhibitor therapy as their first, second or third biologic treatment between 2004 and 2014 and from whom a valid baseline visit was available were included in this study. Additional data on the patients co-morbidities and hospital treatment were retrieved from the Finnish Hospital Discharge Register and Finnish Cancer Register. The results are reported as medians with their respective interquartile ranges (IQRs) or percentages. Differences in baseline patient characteristics and disease activity between different TNF inhibitors were tested using the 2 test, Mann Whitney U-test or Kruskal Wallis test where appropriate. The effectiveness of TNF inhibitors was measured using either ACR or EULAR response criteria, which were estimated at time points of 6, 12 and 24 months after the treatment onset using linear interpolation. LUNDEX adjustment, in which the fraction of responders is multiplied by the fraction of patients remaining in the study, was used to account for the fraction of patients discontinuing the treatment or lost to follow-up [12]. Ordinal logistic regression on reaching moderate or good EULAR response was used to compare different TNF inhibitors with one another while accounting for the differences in baseline characteristics, as well as to evaluate the effect of concomitant csdmards on the treatment effectiveness. The best models in terms of Akaike information criteria were used to adjust the ordinal logistic regression analyses. Effect modification was explored by including interaction terms in the model. The effect of patient attrition was explored in ordinal logistic regression by treating the patients having discontinued the treatment as nonresponders, which is equivalent to LUNDEX adjustment at group level. RCTs of at least 6 months of duration, in which the efficacy of TNF inhibitors had been compared with placebo, csdmards or another bdmard, were identified from the two systematic reviews informing the EULAR guidelines for management of RA [4, 5]. RCTs that excluded patients with prior MTX treatment were not included in this study. We assumed that all TNF-inhibitor users included in ROB- FIN had previously used MTX in accordance with Finnish treatment guidelines, although we did not have data on patients medication history prior to TNF-inhibitor treatment onset to verify this. Information on the inclusion criteria and efficacy results was retrieved from the original publications of the selected RCTs and clinicaltrials.gov. The inclusion criteria were then used to stratify the patients identified from ROB-FIN using the same TNF inhibitor into two groups based on their eligibility for the given RCT. Not all criteria could be adapted fully, either because we lacked the data do so or because they were too vaguely reported. Also, we chose not to use the patients concomitant csdmard therapy at baseline as inclusion criteria. Furthermore, the number of swollen and tender joints in our data was measured on a 54/53 index as opposed to 66/68 index in most RCTs (supplementary Fig. S1, available at Rheumatology Online). Hence, we 726

3 TNF inhibitors in treatment of RA? multiplied the number of tender and swollen joints in our data by 1.22 and 1.28, respectively, and rounded them to the nearest integer. However, this correction was used only in the context of the patients compatibility with the inclusion criteria, whereas joint indices were left unaltered for all other purposes. The treatment responses of eligible and non-eligible patients were compared using Fisher s exact test with Bonferroni correction for multiple comparisons and pooled using a random-effects model with Mantel Haenszel method. Baseline differences in patient characteristics, disease activity and the use of csdmards were compared between eligible and non-eligible patients. Results are reported as medians with their respective IQR, percentages and proportional odds ratios (pors) with their respective 95% CI. Missing data were imputed by multiple imputation with predictive mean matching and 20 imputed data sets. The level of statistical significance was set at Data management and statistical analyses were carried out using R software version 3.11 (R foundation for statistical computing, Vienna, Austria). This study was approved by the Helsinki and Uusimaa Hospital District coordinating ethical committee and the National Institute for Health and Welfare. Patient consent was not required because patient data were taken from existing registries and hospital records. Results Patients and follow-up Of the 4165 TNF-inhibitor treatment periods identified from ROB-FIN, 739 and 934 were excluded owing to the treatment not being started between 2004 and 2014 and the lack of reliable baseline measurement, respectively. In total, 2053 patients were included, of whom 1682, 633 and 177 used TNF inhibitors as their first, second or third bdmard, respectively. Included patients were predominantly female (74%), and the median age at the baseline of the first and second or third bdmard was 55 (IQR: 45 62) and 56 years (IQR: 47 63), respectively (Table 1). Adalimumab and etanercept were the most frequently used TNF inhibitors, covering 80% of all TNF-inhibitor treatments in this study. MTX, SSZ, HCQ and glucocorticoid were used as a co-therapy by 56, 26, 33 and 72% of the patients at the baseline of the TNF-inhibitor treatment period, respectively. Thirteen per cent of the TNF-inhibitor treatments were initiated as monotherapy without concomitant csdmards. At the baseline of the first bdmard, there were numerous statistically significant differences in baseline characteristics of the included patients between the users of different TNF inhibitors and concomitant csdmards (data not shown). Patients excluded from the study owing to the lack of reliable baseline measurement, that is, whose treatment had been initiated before the first visit recorded in our study, had significantly lower DAS28 score (3.0 vs 4.1, P < 0.001) at their first visit compared with the included patients. Of the patients using TNF inhibitors as their first bdmard, 65% remained on the treatment until 6 months, 22% discontinued the treatment and possibly switched to using another bdmard, whereas the remaining 14% of the patients were lost to follow-up. Similar percentages for the patients second or third bdmard were 58, 25 and 17%, respectively. The respective proportions of patients remaining on treatment at 12 and 24 months were 51 and 36% among the users of first bdmard and 47 and 36% among the users of second or third bdmard. The amount of missing data varied from 0 to 27% across the variables in the data set. Treatment effectiveness EULAR moderate and good treatment responses at 6 months were achieved by 69 and 40% of the users of the first bdmard, respectively, whereas similar percentages after LUNDEX adjustment were 44 and 26%, respectively (Fig. 1). ACR20, ACR50 and ACR70 responses were reached by 49, 27 and 13% of the patients, respectively, and after LUNDEX adjustment by 32, 17 and 8.5%, respectively. DAS28 remission criteria were fulfilled by 47 and 30% of the patients with and without LUNDEX adjustment, respectively. The highest EULAR and ACR response rates on the patients first bdmard at 6 months were observed with certolizumab pegol, although this was no longer the case after LUNDEX adjustment (supplementary Fig. S2, available at Rheumatology Online). Golimumab, in contrast, was associated with the largest proportion of patients in remission at 6 months. However, there were no statistically significant differences between individual TNF inhibitors in unadjusted response or remission rates. The adjusted ordinal logistic regression model revealed no statistically significant differences between individual TNF inhibitors (Fig. 2). The variables included in the final regression model with their respective pors are reported in supplementary Tables S1 and S2, available at Rheumatology Online. The patients not using any concomitant csdmards at baseline of the first TNF-inhibitor treatment had the lowest EULAR and ACR treatment responses at 6 months with and without LUNDEX adjustment (supplementary Fig. S3, available at Rheumatology Online). Six months after the TNF-inhibitor treatment onset, 41% of the patients on biologic monotherapy were in DAS28 remission (24% with LUNDEX adjustment), whereas remission was reached by 51% (LUNDEX 35%) and 39% (LUNDEX 24%) of the patients using concomitant csdmards including or excluding MTX, respectively. The highest ACR and EULAR responses were also reached by patients using concomitant MTX. In adjusted ordinal logistic regression analysis of reaching EULAR moderate or good response, co-therapy with MTX had a por of 1.5 (95% CI: 1.0, 2.3). After accounting for the patients having discontinued the treatment or lost to follow-up, the corresponding por was 1.6 (95% CI: 1.1, 2.2). In the unadjusted model, glucocorticoid use was associated with reaching EULAR response (por 1.4, 95% CI: 1.1, 1.8), but this was not the 727

4 Kalle J. Aaltonen et al. TABLE 1 Baseline characteristics and anti-rheumatic medication of the included patients Variable First biologic, n = 1594 Second or third biologic, n = 665 Missing data, % Age, median (IQR), years 55 (45 62) 56 (47 63) 0 Female, n (%) 1214 (72) 629 (78) 0 Disease duration, median (IQR), years 9.3 (3.3 18) 13 (6.8 21) 8.6 Treatment onset, median (IQR) 2008 ( ) 2007 ( ) 0 RF positive, n (%) 1330 (79) 630 (78) 10 Erosions in hands and feet, n (%) 1341 (80) 574 (71) 27 TJC28, median (IQR) 4.0 ( ) 2.0 (0 6.0) 11 SJC28, median (IQR) 3.0 ( ) 2.0 (0 5.0) 12 TJC, median (IQR) 6.0 (2.0 11) 3.0 ( ) 7.0 SJC, median (IQR) 5.0 (2.0 12) 4.0 ( ) 7.6 HAQ, median (IQR) 1.0 ( ) 1.0 ( ) 11 Rheumatic activity, patient, VAS 0 100, median (IQR) 52 (32 70) 46 (24 66) 7.1 Rheumatic activity, physician, VAS 0 100, median (IQR) 40 (23 60) 30 (15 50) 12 Pain, VAS 0 100, median (IQR) 55 (32 71) 48 (25 68) 6.3 ESR, median (IQR) 19 (9.0 36) 18 (9.0 36) 6.0 CRP, median (IQR), mg/ml 11 (5.0 27) 8.0 (5.0 24) 5.0 DAS28, median (IQR) 4.3 ( ) 3.8 ( ) 26 Etanercept, n (%) 676 (40) 320 (40) 0 Adalimumab, n (%) 643 (38) 350 (43) 0 Infliximab, n (%) 200 (12) 46 (5.7) 0 Certolizumab pegol, n (%) 82 (4.9) 44 (5.4) 0 Golimumab, n (%) 81 (4.8) 50 (6.2) 0 MTX, n (%) 983 (58) 408 (50) 0 HCQ, n (%) 628 (37) 190 (23) 0 SSZ, n (%) 514 (31) 142 (18) 0 LEF, n (%) 294 (17) 111 (14) 0 IM gold, n (%) 74 (4.4) 34 (4.2) 0 Any csdmard, n (%) 1483 (88) 678 (84) 0 Oral glucocorticoids, n (%) 1247 (74) 548 (68) 0 csdmard: conventional synthetic DMARD; IQR: interquartile range; SJC: swollen joint count; TJC: tender joint count; VAS: visual analog scale. case after adjustment for confounding (por 1.1, 95% CI: 0.79, 1.4). No statistically significant effect modification between the medication modalities was found. The treatment response and remission rates for the second and third bdmard users were lower compared with the first bdmard therapy, because only 32, 21 and 26% of the patients reached LUNDEX-adjusted moderate EULAR and ACR20 responses and DAS28 remission criteria, respectively, at 6 months (supplementary Fig. S4, available at Rheumatology Online). We were unable to find statistically significant associations in logistic regression between the TNF-inhibitor, csdmard or glucocorticoid therapy and EULAR treatment response at 6 months. Eligibility for RCTs We identified 27 RCTs fulfilling the inclusion criteria for our study, of which 13 and 2 limited inclusion to biologic-naïve patients and prior TNF-inhibitor users, respectively [13 39]. The remaining 12 RCTs imposed some restrictions to prior biologics or did not state whether prior biologics were allowed. Prior biologic therapy was not, however, used as an inclusion criterion but rather to select the appropriate subset of patients from ROB-FIN for each of the ensuing comparisons. The inclusion criteria of the included RCTs are listed in supplementary Fig. S3, available at Rheumatology Online. The fraction of patients identified from ROB-FIN eligible for the included RCTs ranged from 7.6 to 44% (Table 2). Inclusion was most often rejected because of insufficient disease activity at baseline, but only rarely because of co-morbidities or a history with certain medical conditions. Efficacy was reported as ACR50 and EULAR responses at 6 or 12 months after treatment onset in 24 and 6 trials, respectively. Additionally, the fraction of patients in DAS28 remission was reported in 12 trials. Figure 3 shows the ACR50 responses measured in RCTs at 6 or 12 months after TNF-inhibitor treatment onset and the corresponding responses based on ROB- FIN for patients using the same TNF inhibitor either eligible or not eligible for the given RCT. In 8 out of 24 comparisons between eligible and non-eligible patients, the former had a statistically significantly better ACR50 treatment response. Pooled results for reaching the ACR50 and moderate EULAR responses between eligible and non-eligible patients showed risk ratios of 1.7 (95% CI: 1.5, 1.9) and 1.3 (95% CI: 1.2, 1.4), respectively, in 728

5 TNF inhibitors in treatment of RA? FIG. 1Treatment responses and remission rates of TNF inhibitors as patients first biologic DMARD Data given at 3, 6, 12 and 24 months after treatment onset. bdmard: biologic DMARD; ROB-FIN: National Register for Biologic Treatment in Finland. favour of the eligible ones. However, the non-eligible patients reached DAS28 remission more often compared with the eligible patients, with a risk ratio of 0.58 (95% CI: 0.42, 0.79). This implies that with the treatment aim of remission, TNF-inhibitor therapy should not be limited to RCT-eligible patients only, although the treatment costs should also be kept in mind. Pooled risk ratios for achieving ACR50 between RCTs and eligible patients, and between RCTs and non-eligible patients were 0.91 (95% CI: 0.78, 1.1) and 1.6 (95% CI: 1.4, 1.9), respectively. Baseline differences in disease activity and concomitant csdmard therapy were compared between eligible and 729

6 Kalle J. Aaltonen et al. FIG. 2Proportional odds ratio for reaching EULAR response on TNF inhibitors as the patients first biologic DMARD The results of ordinal logistic regression analyses on reaching moderate or good EULAR treatment response on TNF inhibitors within 6 months after the first biologic DMARD onset. bdmard: biologic DMARD; csdmards: conventional synthetic DMARDs; por: proportional odds ratio. non-eligible patients. In 25 out of 27 comparisons, the eligible patients had a statistically significantly higher DAS28 score at baseline compared with the non-eligible patients. In 15 comparisons, erosions in hands and feet were more prevalent among the eligible patients. The eligible patients were older and had a longer time from diagnosis in eight and seven comparisons, respectively. In the majority of the comparisons, there were no statistically significant differences in the csdmard used; however, in seven comparisons a larger fraction of non-eligible patients were using MTX, whereas in another two comparisons the situation was reversed. The percentage of patients on biologic monotherapy was similar across all comparisons. Discussion According to our results, TNF inhibitors are equipotent in routine clinical practice, but concomitant treatment with MTX improves the treatment response. Our study also showed that % of the patients included in the ROB-FIN register and treated with TNF inhibitors in routine clinical practice in Finland would have been potentially eligible for the RCTs conducted during the past 15 years and that the eligibility predicted better treatment responses. Previous studies have also shown that TNF inhibitors are more effective in combination with MTX than as monotherapy [2]. The present study confirmed the benefit of using concomitant MTX therapy alongside TNF-inhibitor treatment in routine clinical practice. Despite bdmard monotherapy being inferior to combination therapy, 61 and 36% of the patients starting on biologic monotherapy were still able to reach moderate EULAR response and DAS28 remission, respectively, without accounting for the patient attrition. This study did not include tocilizumab, which may be more effective as monotherapy in comparison with TNF inhibitors [16]. Co-therapy with csdmards not including MTX did not appear to provide any additional effectiveness in comparison with bdmard monotherapy. Certolizumab pegol appeared possibly more effective than infliximab without adjusting for discontinuations, whereas golimumab was possibly more effective than infliximab after the aforementioned adjustment, which implies differences in treatment survival. The treatment response to TNF inhibitors used as the second and 730

7 TNF inhibitors in treatment of RA? TABLE 2 Randomized controlled trials included and the percentage of eligible routine care patients Study ID TNF inhibitor Percentage eligible considering inclusion criteria related to RA and age Percentage eligible considering exclusion criteria related to current or previous medical conditions Total of patients eligible, n (%) Gabay et al. [16] Adalimumab (26) Keystone et al. [18] Adalimumab (44) Kim et al. [21] Adalimumab (32) Van de Putte et al. [32] Adalimumab (20) Weinblatt et al. [36] Adalimumab (36) Weinblatt et al. [37] Adalimumab (19) Choy et al. [13] Certolizumab pegol (10) Fleischmann et al. [15] Certolizumab pegol (11) Keystone et al. [20] Certolizumab pegol (7.6) Smolen et al. [29] Certolizumab pegol (9.5) Weinblatt et al. [35] Certolizumab pegol (36) Combe et al. [14] Etanercept (16) Klareskog et al. [22] Etanercept (14) Moreland et al. [26] Etanercept (20) Moreland et al. [25] Etanercept (18) O Dell et al. [27] Etanercept (42) van der Heijde et al. [33] Etanercept (14) Weinblatt et al. [34] Etanercept (37) Keystone et al. [17] Golimumab (37) Keystone et al. [19] Golimumab (37) Smolen et al. [29] Golimumab (18) Tanaka et al. [31] Golimumab (31) Lipsky et al. [23] Infliximab (19) Maini et al. [24] Infliximab (19) Schiff et al. [28] Infliximab (17) Westhovens et al. [38] Infliximab (40) Yoo et al. [39] Infliximab (40) third bdmards was inferior compared with when they were used as the first bdmard treatment. Oral glucocorticoid use seemed to improve the treatment response, but this effect diminished after adjusting for confounding factors, probably because glucocorticoids are prescribed for patients with higher baseline disease activity. Ineligibility for the RCTs in our study was most often attributable to insufficient disease activity at baseline. Relatively few patients would have been excluded owing to their co-morbidities because many of the exclusion criteria used in RCTs related to hepatitis, for example, apply in routine care as well. The eligible patients generally had more severe disease in comparison with non-eligible patients on most of the disease activity measures used in our study. It could be argued that the inclusion criteria related to the activity of the rheumatic disease for RCTs are too stringent, considering that their results are directly generalizable to only a fraction of real-world patients. In Finland, the threshold for initiating bdmard therapy has lowered during the past 15 years, and a greater proportion patients treated in the early 2000s would have been eligible for the RCTs than in Regretfully, we were not able to use all of the inclusion criteria of the RCTs owing to the excessive amount of missing data in morning stiffness and anti-ccp measurements. Lacking the 66/68 joint counts, we approximated them from 54/53 joint counts, which we believe to be a reasonably good estimation, albeit one that has not been validated. We did not exclude patients using the laboratory criteria on liver function or cell concentrations criteria from RCTs. Furthermore, the exclusion criteria for some RCTs were not reported at all or were too vague to be considered in our analyses. We also did not consider concomitant or prior csdmard therapies as an inclusion criterion. Had we done so, the proportion of potentially eligible patients would probably have been even lower, because the use of csdmards and their combinations is strongly endorsed in Finland. Although the inclusion criteria of several early RCTs do not mention prior bdmard treatment, it is very unlikely that patients enrolled in an RCT in 1999 would have had any previous experience with bdmards. The eligible patients achieved an ACR50 response 1.6 times more often than the non-eligible patients. A similar trend has been observed in previous studies [3, 6], but our study was the first to include certolizumab pegol and golimumab in such a comparison. According to the pooled results, the efficacy measured in RCTs was comparable to the effectiveness observed among the real-world patients 731

8 Kalle J. Aaltonen et al. FIG. 3ACR50 treatment response on TNF inhibitors in randomized controlled trials and routine clinical practice Percentage and number of patients reaching ACR50 treatment response on TNF inhibitors measured in RCTs and clinical practice among patients either eligible or ineligible for the given RCTs at 6 or 12 months after treatment onset. RCT: randomized controlled trial; ROB-FIN: National Register for Biologic Treatment in Finland

9 TNF inhibitors in treatment of RA? potentially eligible for them. More modest effectiveness can be expected for the non-eligible patients, although many of them still benefited from the treatment. The difference in treatment responses between patients eligible and not eligible for RCTs was most probably mediated by the differences in disease activity at baseline. Consequently, the non-eligible patients were more likely to reach DAS28 remission compared with their eligible counterparts by having a lower DAS28 score initially. The patients eligible for RCTs were older and had a longer time from diagnosis and more erosions in comparison with the non-eligible patients. We hypothesized that the treatment response and remission criteria comprising swollen and tender joint counts could be less sensitive for the patients with prolonged and erosive disease. If this were to be the case, the difference in treatment responses between the eligible and non-eligible patients could be even greater than our results indicate. The seven statistically significant comparisons in ACR50 response between eligible and non-eligible patients all featured either etanercept or adalimumab as the intervention, but a similar trend was observed across all TNF inhibitors. Similar treatment responses were observed between RCTs and routine care patients eligible for them. However, comparing treatment responses measured in RCTs and observational studies is not straightforward because there might be many unreported differences in patient characteristics, not to mention the dissimilar follow-up procedures. At 6 months, 14% of the users of the first bdmard had been lost to follow-up. In addition, 22% of the patients discontinued the treatment within the first 6 months for various reasons, including lack of treatment response and adverse events. Regardless of the reason for discontinuation, we treated all discontinuers and patients lost to follow-up as non-responders in LUNDEX-adjusted sensitivity analyses. In truth, however, not all patients lost to follow-up discontinued the treatment and, as a result, the true effectiveness lies somewhere between the base case and LUNDEX-adjusted results. Missing data are often an issue in observational studies, but with multiple imputation we took the uncertainty caused by it into account. We excluded nearly one-quarter of the medication periods owing to the lack of reliable baseline visit, but this was necessary to ensure the validity of the effectiveness measurements. ROB-FIN did not collect data on patients prior medication data until recently; hence, these data were not available for the present study. Regardless, a recent survey study found that 95% of Finnish RA patients treated with self-injectable bdmards are being or have previously been treated with MTX [40]. To summarize, different TNF inhibitors were mostly equipotent in routine clinical practice, but MTX co-therapy improved the treatment response over TNF-inhibitor monotherapy, perhaps owing to a lower presence of anti-drug antibodies to TNF inhibitors in MTX-treated patients. The efficacy of TNF inhibitors measured in RCTs cannot be generalized directly into Finnish routine health care because only % of the patients were potentially eligible for RCTs, and the eligible patients had better treatment responses compared with the non-eligible patients. Acknowledgements We thank all the Finnish rheumatologists who have participated in the data collection. ROB-FIN has received financial support from the following sources: Abbvie, Hospira, Merck Sharp and Dohme (MSD), Pfizer, Roche, Union Chimique Belge (UCB) and Finska Läkaresällskapet. The funders did not participate in study design, data analysis or writing of the manuscript in any way. Funding: No specific funding was received from any bodies in the public, commercial or not-for-profit sectors to carry out the work described in this manuscript. Disclosure statement: J.T.J. owns shares in Orion Pharma. K.J.A. has received fees for consultancy and speaking from AbbVie, Bristol-Myers Squibb (BMS), Janssen, MSD, Roche and Pfizer. P.I. has received compensation for lectures from MSD, Pfizer, Roche and UCB and for consultancy/advisory board meetings with Abbvie, BMS, Pfizer, Roche and UCB and for participation in international congresses from BMS, Pfizerand Roche outside the submitted work. L.P. has received fees for attending advisory board meetings or speaker s fees from Bristol- Myers Squibb, Roche Pharmaholding, Novartis, Jansen- Cilag, Labquality, Pfizer, UCB Pharma and Sandoz. D.N. has received compensation for lectures, consultancy and advisory board meetings from AbbVie, BMS, MSD, Pfizer, Roche and UCB and research grants from Pfizer outside the submitted work. E.K. has received compensation for lectures from AbbVie, BMS, Pfizer, Roche and UCB, for consultancy/advisory board meetings from MSD and Pfizer, for participation in international congresses from UCB and a research grant from Pfizer outside the submitted work. M.J.K. has given consultations in advisory boards of Abbvie, BMS, Pfizer, MSD, Roche and UCB. All other authors have declared no conflicts of interest. Supplementary data Supplementary data are available at Rheumatology Online. References 1 Rheumatoid Arthritis (online). Current Care Guidelines. Working group set up by the Finnish Medical Society Duodecim and the Finnish Society for Rheumatology. [Internet] [cited ]. web/kh/suositukset/suositus?id=hoi21010 (18 October 2015, date last accessed). 2 Aaltonen KJ, Virkki LM, Malmivaara A et al. Systematic review and meta-analysis of the efficacy and safety of existing TNF blocking agents in treatment of rheumatoid arthritis. PLoS One 2012;177:e Zink A, Strangfeld A, Schneider M et al. Effectiveness of tumor necrosis factor inhibitors in rheumatoid arthritis in an observational cohort study: comparison of patients 733

10 Kalle J. Aaltonen et al. according to their eligibility for major randomized clinical trials. Arthritis Rheum 2006;54: Nam JL, Ramiro S, Gaujoux-Viala C et al. Efficacy of biological disease-modifying antirheumatic drugs: a systematic literature review informing the 2013 update of the EULAR recommendations for the management of rheumatoid arthritis. Ann Rheum Dis 2014;73: Nam JL, Winthrop KL, van Vollenhoven RF et al. Current evidence for the management of rheumatoid arthritis with biological disease-modifying antirheumatic drugs: a systematic literature review informing the EULAR recommendations for the management of RA. Ann Rheum Dis 2010;69: Kievit W, Fransen J, Oerlemans A et al. The efficacy of anti-tnf in rheumatoid arthritis, a comparison between randomised controlled trials and clinical practice. Ann Rheum Dis 2007;66: Escudero-Vilaplana V, Ramírez-Herraiz E, Trovato-López N et al. Influence on effectiveness of early treatment with anti-tnf therapy in rheumatoid arthritis. J Pharm Pharm Sci 2012;15: Fernández-Nebro A, Irigoyen MV, UreñaIet al. Effectiveness, predictive response factors, and safety of anti-tumor necrosis factor (TNF) therapies in anti-tnfnaive rheumatoid arthritis. J Rheumatol 2007;34: Canhão H, Rodrigues AM, Mourão AFet al. Comparative effectiveness and predictors of response to tumour necrosis factor inhibitor therapies in rheumatoid arthritis. Rheumatology 2012;51: Flouri I, Markatseli TE, Voulgari PV et al. Comparative effectiveness and survival of infliximab, adalimumab, and etanercept for rheumatoid arthritis patients in the Hellenic Registry of Biologics: low rates of remission and 5-year drug survival. Semin Arthritis Rheum 2014;43: Hetland ML, Christensen IJ, Tarp U et al. Direct comparison of treatment responses, remission rates, and drug adherence in patients with rheumatoid arthritis treated with adalimumab, etanercept, or infliximab: results from eight years of surveillance of clinical practice in the nationwide Danish. Arthritis Rheum 2010;62: Kristensen LE, Saxne T, Geborek P. The LUNDEX, a new index of drug efficacy in clinical practice: results of a fiveyear observational study of treatment with infliximab and etanercept among rheumatoid arthritis patients in Southern Sweden. Arthritis Rheum 2006;54: Choy E, McKenna F, Vencovsky J et al. Certolizumab pegol plus MTX administered every 4 weeks is effective in patients with RA who are partial responders to MTX. Rheumatology 2012;51: Combe B. Etanercept and sulfasalazine, alone and combined, in patients with active rheumatoid arthritis despite receiving sulfasalazine: a double-blind comparison. Ann Rheum Dis 2006;65: Fleischmann R, Vencovsky J, Van Vollenhoven RF et al. Efficacy and safety of certolizumab pegol monotherapy every 4 weeks in patients with rheumatoid arthritis failing previous disease-modifying antirheumatic therapy: The FAST4WARD study. Ann Rheum Dis 2009;68: Gabay C, Emery P, Van Vollenhoven R et al. Tocilizumab monotherapy versus adalimumab monotherapy for treatment of rheumatoid arthritis (ADACTA): A randomised, double-blind, controlled phase 4 trial. Lancet 2013;381: Keystone EC, Genovese MC, Klareskog L et al. Golimumab, a human antibody to tumour necrosis factor given by monthly subcutaneous injections, in active rheumatoid arthritis despite methotrexate therapy: the GO-FORWARD Study. Ann Rheum Dis 2009;68: Keystone EC, Kavanaugh AF, Sharp JT et al. Radiographic, clinical, and functional outcomes of treatment with adalimumab (a human anti-tumor necrosis factor monoclonal antibody) in patients with active rheumatoid arthritis receiving concomitant methotrexate therapy: a randomized, placebo-controlled, 52-week trial. Arthritis Rheum 2004;50: Keystone E, Genovese MC, Klareskog L et al. Golimumab in patients with active rheumatoid arthritis despite methotrexate therapy: 52-week results of the GO- FORWARD study. Ann Rheum Dis 2010;69: Keystone E, Heijde DVD, Mason D et al. Certolizumab pegol plus methotrexate is significantly more effective than placebo plus methotrexate in active rheumatoid arthritis: findings of a fifty-two-week, phase III, multicenter, randomized, double-blind, placebo-controlled, parallelgroup study. Arthritis Rheum 2008;58: Kim H-Y, Lee S-K, Song YW et al. A randomized, doubleblind, placebo-controlled, phase III study of the human anti-tumor necrosis factor antibody adalimumab administered as subcutaneous injections in Korean rheumatoid arthritis patients treated with methotrexate. APLAR J Rheumatol 2007;10: Klareskog L, van der Heijde D, de Jager JP et al. Therapeutic effect of the combination of etanercept and methotrexate compared with each treatment alone in patients with rheumatoid arthritis: double-blind randomised controlled trial. Lancet 2004;363: Lipsky PE, Van Der Heijde DMFM, St Clair EW et al. Infliximab and methotrexate in the treatment of rheumatoid arthritis. N Engl J Med 2000;343: Maini R, St Clair EW, Breedveld F et al. Infliximab (chimeric anti-tumour necrosis factor a monoclonal antibody) versus placebo in rheumatoid arthritis patients receiving concomitant methotrexate: a randomised phase III trial. Lancet 1999;354: Moreland LW, O Dell JR, Paulus HE et al. A randomized comparative effectiveness study of oral triple therapy versus etanercept plus methotrexate in early aggressive rheumatoid arthritis: the treatment of early aggressive rheumatoid arthritis trial. Arthritis Rheum 2012;64: Moreland LW, Schiff MH, Baumgartner SW et al. Etanercept therapy in rheumatoid arthritis. A randomized, controlled trial. Ann Intern Med 1999;130: O Dell JR, Mikuls TR, Taylor TH et al. Therapies for active rheumatoid arthritis after methotrexate failure. N Engl J Med 2013;369: Schiff M, Keiserman M, Codding C et al. Efficacy and safety of abatacept or infliximab vs placebo in ATTEST: a phase III, multi-centre, randomised, double-blind, placebo-controlled study in patients with rheumatoid arthritis and an inadequate response to methotrexate. Ann Rheum Dis 2008;67:

11 TNF inhibitors in treatment of RA? 29 Smolen J, Landewé RB, Mease P et al. Efficacyand safety of certolizumab pegol plus methotrexate in active rheumatoid arthritis: the RAPID 2 study. A randomised controlled trial. Ann Rheum Dis 2009;68: Smolen JS, Kay J, Doyle MK et al. Golimumab in patients with active rheumatoid arthritis after treatment with tumour necrosis factor a inhibitors (GO-AFTER study): a multicentre, randomised, double-blind, placebo-controlled, phase III trial. Lancet 2009;374: Tanaka Y, Harigai M, Takeuchi T et al. Golimumab in combination with methotrexate in Japanese patients with active rheumatoid arthritis: results of the GO-FORTH study. Ann Rheum Dis 2012;71: van de Putte LB, Atkins C, Malaise M et al. Efficacy and safety of adalimumab as monotherapy in patients with rheumatoid arthritis for whom previous disease modifying antirheumatic drug treatment has failed. Ann Rheum Dis 2004;63: van der Heijde D, Klareskog L, Rodriguez-Valverde V et al. Comparison of etanercept and methotrexate, alone and combined, in the treatment of rheumatoid arthritis: twoyear clinical and radiographic results from the TEMPO study, a double-blind, randomized trial. Arthritis Rheum 2006;54: Weinblatt ME, Kremer JM, Bankhurst AD et al. A trial of etanercept, a recombinant tumor necrosis factor receptor:fc fusion protein, in patients with rheumatoid arthritis receiving methotrexate. N Engl J Med 1999;340: Weinblatt ME, Fleischmann R, Huizinga TWJ et al. Efficacy and safety of certolizumab pegol in a broad population of patients with active rheumatoid arthritis: results from the REALISTIC phase IIIb study. Rheumatology 2012;51: Weinblatt ME, Keystone EC, Furst DE et al. Adalimumab, a fully human anti-tumor necrosis factor a monoclonal antibody, for the treatment of rheumatoid arthritis in patients taking concomitant methotrexate: the ARMADA trial. Arthritis Rheum 2003;48: Weinblatt ME, Schiff M, Valente R et al. Head-to-head comparison of subcutaneous abatacept versus adalimumab for rheumatoid arthritis: findings of a phase IIIb, multinational, prospective, randomized study. Arthritis Rheum 2013;65: Westhovens R, Yocum D, Han J et al. The safety of infliximab, combined with background treatments, among patients with rheumatoid arthritis and various comorbidities: a large, randomized, placebo-controlled trial. Arthritis Rheum 2006;54: Yoo DH, Hrycaj P, Miranda P et al. A randomised, doubleblind, parallel-group study to demonstrate equivalence in efficacy and safety of CT-P13 compared with innovator infliximab when coadministered with methotrexate in patients with active rheumatoid arthritis: the PLANETRA study. Ann Rheum Dis 2013;72: Aaltonen KJ, Turunen J, Sokka T, Puolakka K, Valleala H. A survey on the medication adherence to methotrexate among rheumatoid arthritis patients treated with selfadministered biologic drugs. Clin Exp Rheumatol 2016;34: