Article Text

Extended Report
Effectiveness of TNF inhibitor switch in RA: results from the national Swedish register
  1. Katerina Chatzidionysiou1,
  2. Johan Askling2,
  3. Jonas Eriksson2,
  4. Lars Erik Kristensen3,
  5. Ronald van Vollenhoven1,
  6. for the ARTIS group
  1. 1Unit for Clinical Research Therapy, Inflammatory Diseases (ClinTRID), Karolinska Institutet, Stockholm, Sweden
  2. 2Clinical Epidemiology Unit, Karolinska Institutet, Stockholm, Sweden
  3. 3Department of Rheumatology, Skåne University Hospital, Lund, Sweden
  1. Correspondence to Dr Katerina Chatzidionysiou, Unit for Clinical Research Therapy, Inflammatory Diseases (ClinTRID), Karolinska Institutet, Rheumatology Department, D1:00, Karolinska University Hospital, Stockholm 171 76, Sweden; aikaterini.chatzidionysiou{at}


Background Switching to a second tumour necrosis factor inhibitor (TNFi) after discontinuation of a first in rheumatoid arthritis (RA) is a common strategy. The reason for the switch from the first TNFi could potentially influence the response to therapy. Data on direct comparisons between TNFi after switching are limited.

Methods The national Swedish register was used. RA patients who switched to a second TNFi (infliximab, etanercept or adalimumab) after failure of a TNFi as first-ever biologic were identified. Effectiveness of treatment was compared across the three drugs according to the first TNFi used, the reason for discontinuing and the drug survival. Drug survival across TNFi used as second biologic was compared.

Results Half of all patients starting infliximab, adalimumab or etanercept during the period 2005–2012 discontinued treatment for various reasons. Of these patients, a third switched within 2 months to a second TNFi (infliximab, etanercept or adalimumab). Around 35% of all patients achieved low disease activity or remission at 6 months. Regarding the switching strategy, best results were observed among patients who switched from infliximab to etanercept because of (secondary) inefficacy. Etanercept as second TNFi was associated with longer drug survival compared with infliximab.

Conclusions Switching to a second TNFi after the failure of the first may lead to good clinical results. The inter-drug differences in drug survival on the second TNFi mirror those reported previously for the first TNFi, suggesting that these differences are not solely due to channelling bias.

  • Rheumatoid Arthritis
  • Anti-TNF
  • Treatment
  • DMARDs (biologic)
  • Epidemiology

Statistics from


The first biologic agents to be approved for the treatment of active rheumatoid arthritis (RA), refractory to conventional antirheumatic therapy, were the tumour necrosis factor inhibitors (TNFis) etanercept (ETA), infliximab (INF) and adalimumab (ADA).1–3 ETA is a soluble TNF receptor, INF a chimeric anti-TNF monoclonal antibody and ADA a fully human anti-TNF monoclonal antibody. In the last 3 years, two new anti-TNF monoclonal antibodies have become available, namely, certolizumab pegol and golimumab.4–6

The significant efficacy and acceptable safety profile of these drugs have been demonstrated in large randomised controlled clinical trials.1–11 However, it has also been shown that a significant number of patients discontinue treatment for various reasons, mainly due to inefficacy or intolerance. Indeed, a number of studies from clinical practice indicate that as many as 50% of all patients discontinue their TNFi treatment during the first 3 years.12–14

In clinical practice, switching to a second TNFi is common, the rationale being that the different TNFis differ in their molecular structure, immunological action, immunogenicity and pharmacokinetics. Both clinical trials and epidemiological studies have demonstrated the efficacy of TNF inhibition after TNF failure.15–20 To optimise switching between TNFis in clinical practice, more information is needed on whether factors such as reason for switching (inefficacy or intolerance of the first TNFi) or type of the first TNFi influence the response to the second TNFi. Previous studies have examined the effectiveness of switching from the first TNFi.13 ,14 ,20 ,21 However, no study to our knowledge has hitherto compared specific switching strategies (switching between individual TNFis).

In many observational studies, slightly better retention rates and effectiveness have been reported for ETA than for ADA and INF, but there is some uncertainty whether this superiority reflects channelling bias or a true difference.14 ,22 One way to ‘disconnect’ such channelling from drug-related effects is to make inter-drug comparisons of drug survival and effectiveness of individual TNFis used as second TNFi (ie, when the initial channelling was to another TNFi).

The aims of this study were therefore to (1) assess switching from first to second TNFi under different circumstances (according to the reason for discontinuation and the type of the first TNFi) trying to identify an optimal switching strategy and (2) examine drug survival of ADA, ETA and INF when used as second TNFi after switching from a first TNFi in patients with RA identified in clinical practice.


Study population

Data from the nationwide Swedish Biologics Register (Anti-Rheumatic Therapy in Sweden (ARTIS)) were used. To this register, data on adult patients prescribed biologic agents for the treatment of rheumatic diseases in Sweden have been collected since 1999. The coverage of the ARTIS database has been estimated to be nearly 90% of all eligible patients with RA.23 ,24 Most RA patients in Sweden start an anti-TNF after failure of at least one synthetic disease modifying antirheumatic drug.

Patients included in this observational cohort study had a diagnosis of RA and had started treatment with a ‘first-ever’ used TNFi (INF, ETA or ADA) during the period 01 January 2005–01 September 2012. We chose to include patients who started TNFi treatment after 2005 since all three TNFis were available and switching from one TNFi to another was common in clinical practice. Among these, we further identified those patients who switched to INF, ETA or ADA as second TNFi. Patients switching to another biologic (rituximab, tocilizumab or abatacept), to one of the newest TNFis (golimumab, certolizumab pegol), or those who stopped and restarted the same TNFi were excluded from the analysis. Although many patients who discontinue a first biologic may eventually start a second, for the purpose of this study, we defined ‘switching’ as starting a second and different TNFi within 2 months from the date of discontinuation of the first TNFi. By setting this time frame, we eliminated the risk that a patient would start another non-biologic disease modifying antirheumatic drug before switching to a second TNFi, as the assessment of treatment efficacy takes place after 3 months. Patients who did not have enough follow-up time (started the second TNFi within 8 months from the last observation date (01 September 2012)) were also excluded from this analysis. Four different reasons for discontinuation of the first TNFi were considered: primary inefficacy (lack of efficacy, including partial efficacy), secondary inefficacy (loss of efficacy), intolerance and other (pregnancy, patient's or physician's decision, inactive disease, death, unknown).

The following information was collected: demographic data (age, sex); RA disease duration (from the time of diagnosis); rheumatoid factor; type of first TNFi, date and reason for discontinuation of the first TNFi; date of initiating therapy with the second TNFi and, if discontinued, date and reason; disease activity score based on 28-joint status (DAS28) and its components (swollen joint count, tender joint count, visual analogue scale general health and erythrocyte sedimentation rate); and functional ability based on health assessment questionnaire. The reason for discontinuation was based on the rheumatologist's opinion and was recorded according to a predefined list of different options.

Effectiveness of treatment was assessed by DAS28 change (ΔDAS28) between baseline and 6 months (150–240 days from baseline), the percentage of patients achieving low disease activity or remission (according to DAS28 status: low disease activity=DAS28≤3.2, remission=DAS28≤2.6) as well as the percentage of patients achieving DAS28 improvement ≥1.2 at 6 months. Baseline was defined as start of second TNFi. Drug survival of the three TNFis as second biologic was also assessed.

Statistical analysis

Three exposure categories of interest were defined: patients who switched to INF, ETA or ADA after the failure of either of these drugs used as the first-ever TNFi. Baseline characteristics across the three groups were summarised and compared. The normality of all continuous variables was tested by skewness. Variables which were similar to normal distribution were presented as mean±SD, while non-normally distributed variables were presented as median with IQR. One-way analysis of variance was used to compare continuous variables (with a distribution similar to normal) followed by Bonferroni test for post hoc comparisons between the groups, while χ2 test was used for nominal variables. Kruskal–Wallis test was used to compare non-normally distributed variables. The level of statistical significance was set to 5%.

We used two time frames for follow-up, 3 and 6 months. To comply with clinical practice, the time window was 30–150 days from baseline for the 3-month visit and 150–240 days from baseline for the 6-month visit. The median time to first visit in ARTIS is 4.5 months after the start date. We prioritised 6 months data over 3 months data, when available, but when no such data were available, we used data from the 3-month visit. Non-responder imputation (defined as high or moderate disease activity at 6 months) was used for patients switching to another biologic during the follow-up period and for patients who did not have a follow-up visit but for whom the stated reason for stopping treatment was primary or secondary inefficacy. Three main analyses were performed. (1) The effectiveness of INF, ETA and ADA as second TNFi was directly compared in a first analysis. Stratification for baseline DAS28 (high baseline disease activity defined as DAS28≥5.1 vs non-high with DAS28<5.1) was performed to investigate effect modification by disease activity, that is, any inter-drug difference would differ for different levels of baseline disease activity. (2) The effectiveness of the three TNFis was assessed as a function of the first TNFi. In total, six different switching strategies were thus compared (first ADA then ETA, first ADA then INF, first ETA then ADA, etc). (3) Effectiveness of the second TNFi was assessed as a function of the reason for discontinuation of the first TNFi. Four different reasons for discontinuation of the first TNFi were considered: intolerance, primary inefficacy (lack of efficacy, including partial efficacy), secondary inefficacy (loss of efficacy) and other (pregnancy, patient's or physician's decision, inactive disease, death, unknown).

Kaplan–Meyer curves were plotted to determine continuation rates for the second TNFi during the first 2 years following switch. Curves were compared with the Log-Rank test. Discontinuation of treatment (for the reasons described above) was considered an event. Continuation of treatment at the time of data collection was treated as censored observations during the analysis.


We identified 7052 patients with a diagnosis of RA who started treatment with a first-ever TNFi (INF, ETA or ADA) during the period 01 January 2005–01 September 2012: 2174 with INF, 3076 with ETA and 1802 with ADA. During the same time-period, 50% of these patients discontinued their TNFi treatment. Of these, 2649 (38%) patients started a second biologic, 1457 to a second TNFi. In the final study population, 952 patients who switched (according to our definition) to a second TNFi and had long enough follow-up time were included (74 switched to INF, 448 to ETA and 430 to ADA within 2 months after discontinuing the first TNFi). In figure 1, the flow chart of patients is shown.

Figure 1

Flow chart of patients (selection as described in the methods). 7052 Swedish RA patients started treatment with infliximab (INF), etanercept (ETA) or adalimumab (ADA) as 1st TNFi during 2005-2012. Around 50% discontinued treatment during the observational period (number and percentage of patients discontinued indicated in the boxes). 1457 patients continued with a 2nd TNFi, of which 1068 ‘switched’ to INF, ETA or ADA within 2 months after discontinuing the 1st TNFi. (Grey boxes: patients excluded).

Baseline for our analyses was defined as the time of start of the second TNFi. Patients in the three groups were quite well balanced regarding baseline characteristics, except for baseline DAS28 which was significantly lower in the ADA group compared with the other two groups (table 1). Regarding baseline DAS28 status, the completeness of available data was 80% for INF, 83% for ETA and 78% for ADA. Regarding available DAS28 status/ΔDAS28 at 6 months, completeness of available data was 81/64% for INF, 86/74% for ETA and 86/73% for ADA. The percentage of missing data at 3 and 6 months was similar across the three TNFi groups.

Table 1

Characteristics for Swedish patients with RA starting a second TNFi (INF, ETA or ADA) during 2006–2012 at the time-point of their second TNFi start

Effectiveness of second TNFi, per drug

At 6 months patients in all three groups achieved significant reductions in DAS28: ΔDAS28 0–6 months=−1.1±1.5 [n=38] for INF, −1.4±1.6 [n=275] for ETA and −0.8±1.5 [n=244] for ADA. The inter-drug difference was statistically significant between ETA and ADA (p<0.0001). After adjustment for baseline DAS28, this difference remained significant (p=0.04). The percentage of patients who achieved ΔDAS28≥1.2 was 47% (18/38) for INF, 55% (151/275) for ETA and 36% (87/244) for ADA (p<0.0001 ETA vs ADA). The percentage of patients with low disease activity or remission at 6 months, however, was similar for all groups: 37% (18/49) for INF, 34% (113/332) for ETA and 35% (103/294) for ADA. When stratified by baseline DAS28, the ETA group achieved numerically higher but statistically not significantly different DAS28 reductions than ADA (table 2).

Table 2

Effectiveness of second TNF inhibitor (INF, ETA and ADA) based on ΔDAS28, % of patients with low disease activity or remission at 6 months and % of patients achieving ΔDAS28 of at least 1.2 at 6 months, stratified by disease activity at baseline (high vs no high)

Effectiveness of second TNFi as a function of the type of first TNFi

In the second analysis, we examined the effectiveness of the second TNFi taking into consideration the TNFi switched from. The six possible switching groups were distributed as follows: from ADA to ETA (N=206), from ADA to INF (N=16), from ETA to ADA (N=329), from ETA to INF (N=58), from INF to ADA (N=101) and from INF to ETA (N=242). The effectiveness observed for each switching strategy is shown in table 3. For the two largest groups, statistically greater ΔDAS28 was observed in the group switching from INF to ETA than for the group switching from ETA to ADA (p<0.0001, after adjustment for baseline DAS28, p=0.004). Overall better results were observed for the INF→ETA and the ADA→ETA groups compared with ETA→ADA (the two reciprocal groups, ETA→INF and ADA→INF, were too small to allow any comparisons).

Table 3

Effectiveness of six different switching strategies (first TNFi→second TNFi) based on ΔDAS28 after 6 months, % of patients with low disease activity or remission at 6 months and % of patients achieving ΔDAS28 of at least 1.2 at 6 months

Effectiveness of second TNFi by reason for discontinuation of first TNFi

In the third analysis, we examined the effectiveness of the second TNFi according to the reason for switch from the first TNFi. The majority of patients switched to the second TNFi after lack or loss of efficacy of the first TNFi (66%), while 17.4% of patients switched due to intolerance. The number of patients as well as effectiveness data according to the reason for switch is shown in table 4. Overall numerically higher ΔDAS28 at 6 months was observed after loss of efficacy (mean ΔDAS28 at 6 months=−1.4±1.6) than after lack of efficacy (mean ΔDAS28=-1.2±1.6) or intolerance (mean ΔDAS28=−1.1±1.5) to the first TNFi. Significantly higher rates of low disease activity/remission at 6 months were achieved when the reason for switch was secondary inefficacy (40%) or intolerance (39%) than primary inefficacy (26%) (p<0.0001). The best responses were observed when switching to ETA after losing efficacy of ADA or INF as the first TNFi.

Table 4

Effectiveness of the second TNFi (INF, ETA, ADA) at 6 months by reason for switching (primary and secondary inefficacy, intolerance, other)

Drug survival on the second TNFi, per drug

During the first 24 months after switching to the second TNFi, 567 patients (60%) discontinued their second TNFi: 46 out of 74 in the INF group (62%), 257 out of 448 (57%) in the ETA group and 264 out of 430 (61%) in the ADA group (figure 2). The median (95% CI) survival time for INF, ETA and ADA was 14 (7–21), 24 (16–32) and 16 (9–23) months, respectively. Since the survival curves crossed at approximately 8 months (figure 2), two separate analyses were performed; up to 8 months following switch, no significant differences in drug survival among ADA, ETA and INF were observed, but for the second time-period significant differences were observed (figure 2).

Figure 2

Drug survival during the first 2 years of treatment for INF, ETA and ADA as 2nd TNFi. Significant difference was observed between INF and ETA as well as INF and ADA during the period 8–24 months (p=0.02 ETA vs. INF, p=0.05 ADA vs. INF by Log–rank test).


The results of our study support the findings from previous studies that switching to a second TNFi may lead to significant clinical improvements, with almost 40% of patients achieving low disease activity or remission, regardless of the specific TNFi.

Because of the real-world setting, our findings must be interpreted in light of a number of potential or real differences among the three groups of switchers. As shown in table 1, most baseline demographic and disease variables were quite similar across groups. However, baseline disease activity was significantly higher for those starting ETA as their second TNFi than ADA, which (under the assumption that the capacity for DAS28 reduction is non-linear, or simply through regression to the mean) might explain the greater improvement in DAS28 for ETA as second TNFi compared with ADA. Indeed, when stratified by baseline disease activity, we observed overall greater reductions in DAS28 among those patients with a disease activity score above 5.1 at the time point of the second TNFi start than in the non-high disease activity group, but, as expected, a higher percentage of patients achieving low disease activity/remission in the latter. ETA achieved numerically but not significantly better improvements than ADA, which might represent a true effect but also reflect limited statistical precision. The difference was not large enough to assure a true clinical difference.

Overall better results were achieved with the second TNFi after loss of efficacy or intolerance to first TNFi than after lack of efficacy of the first TNFi, supporting the results from previous studies.18 ,20 This observation is rational and supports the hypothesis that for patients who do not respond to TNFi (primary inefficacy), TNF might not play that important role in their disease and a second TNFi would yield only modest results. Best responses were observed when switching to ETA after losing efficacy of ADA or INF as the first TNFi. The production of antidrug antibodies, drug immunogenicity, has been proposed as one possible mechanism behind inefficacy.25–28 Secondary inefficacy to a first TNFi might be due to development of antidrug antibodies, and non-responders in that case might benefit from a switch to a less immunogenic drug, such as ETA.

When different switching strategies were compared, overall better results were observed for patients who switched from INF or ADA (monoclonal antibodies) to ETA than the other way round. We acknowledge, however, that some of the switching strategies included a too limited number of patients to allow for meaningful interpretation. The observed difference was present mainly when the reason for switching was inefficacy rather than intolerance (see online supplementary file). A clinical interpretation of this observation might be that after the failure of a monoclonal anti-TNF antibody one could consider switching to ETA, but when a patient has failed ETA as first TNFi, the choice of a monoclonal antibody as second TNFi might not have the same potential to lead to a clinically significant improvement. Indeed, other studies have shown that in the latter case it might be more beneficial to change to a biologic of a different mechanism of action.29 ,30 This might imply an underlying mechanism that could explain this finding: as it is known, ETA competitively inhibits the binding of both TNF and lymphotoxin-α to cell surface TNF receptors, rendering TNF biologically inactive,31 while INF and ADA bind and neutralise both soluble and membrane-bound TNF but not lymphotoxin. Lymphotoxin plays a crucial role in chronic inflammation.32 Thus, failing a monoclonal antibody (especially because of inefficacy) a patient might respond well to ETA, and the reason could be the binding of lymphotoxin.33 This remains, however, a hypothesis.

A limitation of this study was the observational cohort study design, as patients were not randomly placed in the treatment groups. The risk of selection bias is therefore present. The treatment groups under comparison were not entirely balanced for baseline characteristics and some significant differences were observed, which might introduce the risk for confounding. Other and unknown baseline factors may also differ between the drugs. Indeed and as mentioned in the Introduction section, confounding by indication may be responsible for some of the previously observed differences in drug survival between ETA and ADA/INF as first TNFi. In this study, we could limit some of such confounding by indication by assessing drug survival on these drugs when used as the second TNFi (hence, when patients starting ETA were recruited from a pool of patients who had initially all be channelled to treatment with INF or ADA, and vice versa). We found some differences in favour of ETA as second TNFi, similar to what has been shown in previous analyses of first TNFi, suggesting that it is more possible that these results represent true difference and are not entirely due to confounding by indication related to the choice of the first TNFi. The clinical relevance of the observed difference is, however, unclear. Other differences between TNFis, such as route of administration, have to be taken into account. INF is given intravenously, which might influence negatively its retention rate. Such factors limit the possibility of comparing these agents using retention rates as surrogates for effectiveness. We tried to partially overcome the problem with selection bias by first adjusting for baseline DAS28 and then by stratifying according to disease activity (high vs low).

The limited number of patients in some of the comparisons was a challenge. Some numerical differences did not reach statistical significance, and this might be due to lack of power. Missingness is a common problem in register-based observational studies. In our study, missingness did not differ between drugs. Due to the limitations described above, however, the results of this study have to be interpreted with caution.

In summary, after the failure of the first TNFi, up to 40% of patients switching to a second TNFi may achieve low disease activity or remission. After the failure of a monoclonal antibody as first TNFi because of inefficacy, switching to ETA yielded good clinical results, but not the other way round. ETA was associated with longer drug survival compared with INF as second TNFi. The inter-drug differences in drug survival on the second TNFi mirror those reported previously for the first TNFi, suggesting that these differences are not solely due to channelling bias.


Supplementary materials

  • Supplementary Data

    This web only file has been produced by the BMJ Publishing Group from an electronic file supplied by the author(s) and has not been edited for content.

    Files in this Data Supplement:


  • Handling editor Tore K Kvien

  • Collaborators The ARTIS (Anti-Rheumatic Therapy in Sweden) group: Johan Askling, Eva Baecklund, Lars Cöster, Lars Erik Kristensen, Nils Feltelius, Helena Forsblad d'Elia, Pierre Geborek, Lennart Jacobsson, Lars Klareskog, Staffan Lindblad, Martin Neovius, Solbritt Rantapää-Dahlqvist and Ronald van Vollenhoven.

  • Contributors KC, JA, JE, LEK and RvV contributed to the conception and design of this study. All authors contributed to the analysis and interpretation of data, drafting the article and revising it critically for important intellectual content. They all gave their approval of the version to be published.

  • Funding Supported in part by an unrestricted grant from Wyeth (now Pfizer).

  • Competing interests LEK has received consultancy fees and fees for speaking from AbbVie, BMS, MSD and Pfizer. JA has participated in an unrelated advisory board organised by Pfizer. RvV has received research support and/or honoraria from AbbVie, Biotest, BMS, GSK, Lilly, Merck, Pfizer, Roche, UCB and Vertex.

  • Ethics approval Stockholm ethical review board.

  • Provenance and peer review Not commissioned; externally peer reviewed.

Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.