Patients with rheumatoid arthritis (RA) who experience treatment failure with one anti–tumor necrosis factor (anti-TNF) agent, due to either inefficacy or toxicity, are frequently switched to a second anti-TNF agent, although the benefits of switching are unknown. The present study was undertaken to compare drug continuation rates between the first course and second course of anti-TNF therapy.
The study involved a prospective cohort of RA patients from a UK national register of new anti-TNF treatment starts (n = 6,739; 876 starting adalimumab, 2,826 starting etanercept, and starting 3,037 infliximab). Over a mean 15 months of followup, 841 patients stopped taking the first drug due to inefficacy and 1,023 stopped the first drug due to toxicity, of whom 503 and 353, respectively, were switched to a second anti-TNF agent. Kaplan-Meier survival curves were plotted to determine continuation rates for each course, and Cox regression was used to compare each course for the risk of stopping and the reason for stopping (inefficacy or toxicity).
Overall, 73% of patients who switched to a second anti-TNF agent remained on the new therapy by the end of followup. First drug discontinuation due to inefficacy was associated with an increased rate of second drug discontinuation due to inefficacy (hazard ratio [HR] 2.7, 95% confidence interval [95% CI] 2.1–3.4) but not toxicity (HR 1.1, 95% CI 0.9–1.5). Similarly, first drug discontinuation due to toxicity was associated with an increased rate of second drug discontinuation due to toxicity (HR 2.3, 95% CI 1.9–2.9) but not inefficacy (HR 1.2, 95% CI 0.8–1.6).
RA patients who are switched to a second anti-TNF drug have high rates of continuation, although among those who must discontinue treatment, the reasons for stopping a second drug are related to the reasons for stopping the first drug. This large data set from the UK provides the first estimates of the magnitude of these effects in patients with long-standing severe RA.
Tumor necrosis factor α (TNFα) is a key cytokine driving synovial inflammation in rheumatoid arthritis (RA) (1). The introduction of anti-TNFα therapies has dramatically improved the treatment of severe RA. Currently, 3 such agents are available: adalimumab, etanercept, and infliximab. These agents have been shown to be very effective in randomized clinical trials of patients with RA who have active disease despite treatment with conventional disease-modifying antirheumatic drugs (DMARDs), including methotrexate (MTX). Although there have been no head-to-head trials, the 3 anti-TNF agents appear to have similar efficacy, with up to 70% of patients achieving at least a 20% improvement in disease activity (2–6). In addition, prospective cohort studies comparing these agents have not shown any clinically important differences between them (7–11).
Despite these impressive findings, approximately one-third of patients show a lack of response to anti-TNFα therapies, either due to lack of drug efficacy or following the development of adverse events (AEs). Although all 3 drugs exert their beneficial effects through blockade of TNFα, there are differences in their site of action and molecular structure. In individual patients, these differences may explain the differential response to the 3 agents, although there is no direct evidence to support this. Indeed, there are now numerous reports in which it has been suggested that initiating therapy with a second anti-TNFα agent in patients who have failed therapy with a first agent (due to lack of efficacy or because of an AE) may be beneficial and not associated with an increased rate of AEs with the second agent (12–20). However, all of the studies to date have been based on small case series (the largest included only 37 patients), which makes it difficult to establish a substantive estimate of recurrent AEs and treatment inefficacy. Therefore, using data from a large UK national biologics register, we estimated the risk of recurrent nonresponse and recurrent AEs in patients who are switched to a second anti-TNFα agent.
PATIENTS AND METHODS
The subjects for this analysis were participants in a large prospective observational study in the UK, the British Society for Rheumatology Biologics Register (BSRBR). The methods of this register have been described in detail elsewhere (21). Briefly, as part of UK national guidelines, all patients with RA starting a new anti-TNFα therapy are registered with the BSRBR, with the goal of recruiting a total of 12,000 biologically naive patients (4,000 in each treatment start group [etanercept, adalimumab, or infliximab]) (22). The present analysis was restricted to patients registered with the BSRBR with a diagnosis of RA (as determined by the treating rheumatologist) who had reached a minimum of 6 months of followup by the end of April 2005.
UK national guidelines (22) recommend that anti-TNFα drugs be administered to patients with active RA (defined as a Disease Activity Score in 28 joints [DAS28]  >5.1 despite previous therapy with at least 2 DMARDs, one of which should be MTX). During the period of recruitment, etanercept was administered as a subcutaneous injection of 25 mg twice weekly. Adalimumab was administered as a subcutaneous injection of 40 mg every 2 weeks. The suggested starting dose of infliximab was 3 mg/kg, administered at weeks 0, 2, 6, and 8 and every 8 weeks thereafter, although the treating physician could have increased this dose or schedule when warranted. It is recommended that infliximab be administered in combination with MTX.
At the time of initiation of the biologic drug in these patients, the rheumatologist or rheumatology nurse specialist completes a standardized form to record each patient's demographic data, including age, sex, diagnosis, and disease duration, and to record items on current disease activity, including the swollen and tender joint counts (based on the 28-joint count), the erythrocyte sedimentation rate, and/or C-reactive protein level, and the patient's global assessment of health. Details of past and present antirheumatic therapies and current comorbidities are also recorded. Each patient completes a separate questionnaire that includes smoking history, and information is also obtained from the Health Assessment Questionnaire (HAQ) adapted for British use (24).
Rheumatologists and patients each complete a followup questionnaire every 6 months. Of relevance to this analysis, the rheumatologists record whether their patient's registered anti-TNFα drug has been continued, switched to another anti-TNFα agent, or stopped. The reasons for stopping or switching treatment are based on the physician's opinion, as recorded in the medical charts, and classified into lack of efficacy, AEs, or other reasons. Details of all AEs, regardless of whether the physician believes they were directly related to the anti-TNFα therapy, are also recorded.
Three separate analyses of anti-TNFα treatment exposures were performed. The first analysis assessed the experience of all study patients with their first anti-TNFα agent. The other 2 analyses focused on the patients' experience with a second agent, and hence comprised only those patients who were switched. The latter was subdivided into 2 groups based on whether the patients 1) had not shown an adequate response to the first agent (inefficacy) or 2) had developed an AE while receiving their first agent. Kaplan-Meier survival curves were plotted to determine the rates of continuation of the first treatment as compared with the rates of continuation of the second treatment. Cox regression analyses were used to compare drug discontinuation, in terms of the risk of stopping the treatment due to 1) inefficacy or 2) development of an AE, with the second agent as compared with the first agent.
Separate analyses stratified according to the reason for stopping the first agent were undertaken in subgroups of patients who were switched to a second agent. Briefly, the risks of stopping an anti-TNFα agent for either reason (inefficacy or toxicity) were compared between patients taking their first agent and patients taking their second agent. Subsequently, the patients who were switched to a second agent were compared by their subgroups, comprising those stopping their first agent due to inefficacy and those stopping their first agent due to an AE. Use of this approach enabled us to determine the relationship between the reasons for stopping the first drug and the reasons for stopping the second drug. The analysis was restricted to the first and second courses of anti-TNFα therapy. Due to small patient numbers in some of the subgroups, a separate analysis based on which anti-TNFα therapy was taken as either the first or second course was not included.
All analyses were adjusted for age, sex, disease duration, baseline DAS28 score, and baseline HAQ score. Further analyses were also undertaken, with adjustment for use of concomitant DMARDs and oral corticosteroids.
A total of 6,739 biologically naive patients with RA had been registered with the BSRBR and had completed at least 6 months of followup by the end of April 2005 (Table 1). Of these, 876 (13%) had started adalimumab, 2,826 (42%) had started etanercept, and 3,037 (45%) had started infliximab. The mean length of followup per patient was 15 months (maximum 61 months). Due to the difference in availability of the 3 agents, the mean followup of patients starting infliximab (18 months) was longer than that of patients starting either adalimumab (11 months) or etanercept (13 months). The baseline characteristics of the cohort are shown in Table 2.
Table 1. Details of treatment with the first anti–tumor necrosis factor α agent
Table 2. Baseline characteristics of the entire cohort and the groups having stopped the first biologic agent, according to reason for stopping and need to switch to a second agent*
Total cohort (all starts)
Stopped first agent for inefficacy
Stopped first agent for adverse event
DMARDs = disease-modifying antirheumatic drugs; IQR = interquartile range; DAS28 = Disease Activity Score in 28 joints; HAQ = Health Assessment Questionnaire; MTX = methotrexate.
No. of patients
Age, mean ± SD years
55 ± 12
58 ± 12
54 ± 12
59 ± 11
54 ± 12
Female sex, no. (%)
Disease duration, mean ± SD years
14 ± 9
14 ± 10
13 ± 9
15 ± 10
14 ± 9
Previous no. of DMARDs, median (IQR)
DAS28, mean ± SD
6.6 ± 1.0
6.7 ± 0.9
6.8 ± 1.0
6.7 ± 1.0
6.8 ± 1.0
HAQ, mean ± SD score
2.1 ± 0.6
2.2 ± 0.5
2.2 ± 0.5
2.2 ± 0.5
2.2 ± 0.5
Receiving concomitant corticosteroids, no. (%)
Receiving concomitant DMARDs
Receiving concomitant MTX
Slightly more than one-third of the patients discontinued their primary anti-TNFα therapy during the followup period (Table 1). Twelve percent of the patients discontinued the first drug due to lack of efficacy, and 15% discontinued the first drug because of an AE. A total of 856 of these patients (46%) switched to a second anti-TNFα agent. More patients who received infliximab as their first anti-TNFα drug were subsequently switched to a second agent (Table 3), which is a reflection of the differential time of introduction of these drugs to the UK market. It was more common for patients to switch to a second agent following treatment inefficacy than following development of an AE. Sixty percent of the patients (503 of 841) started a second agent after discontinuing their first agent due to inefficacy, compared with only 35% of patients (353 of 1,023) who switched after discontinuing their first agent due to an AE. Details of the adverse events that led to treatment discontinuation are presented in Table 4.
Table 3. Pattern of anti–tumor necrosis factor α (anti-TNFα) switches, based on the reason for discontinuation of the first anti-TNFα agent*
First anti-TNFα agent
Second anti-TNFα agent
Values are the number of patients. NA = not applicable.
Reason for stopping
Reason for stopping
Reason for stopping
Table 4. Adverse events (AEs) leading to discontinuation of the first anti–tumor necrosis factor α agent*
Total with AE, no. (% of AEs leading to discontinuation)
Switched due to AE, no. (% of total)
Events leading to drug discontinuation in ≥1% of patients are reported.
Systemic drug reaction
Shortness of breath
Abnormal liver enzyme levels
Ischemic heart disease
Nonhealing skin ulcer
Compared with those who switched to a second agent, patients who discontinued their first anti-TNFα therapy and did not switch to a second agent were older, more likely to be receiving corticosteroids, and less likely to be receiving a concurrent DMARD. When all switched patients were compared according to their reason for switching, no differences in baseline characteristics were seen. Moreover, there were no differences between patients who switched and those who did not switch in terms of disease duration, the DAS28 score, or the HAQ score at the start of therapy with the first anti-TNFα agent. There were also no substantial differences in baseline characteristics between the entire cohort and those patients who switched to a second anti-TNFα agent (Table 2).
As of the end of April 2005, 73% of patients were still receiving therapy with their second anti-TNFα agent (mean length of therapy 6 months, maximum 32 months). Interestingly, the overall rates of discontinuation of the second anti-TNFα drug were similar to those of the first anti-TNFα drug, with 13% of patients stopping the second drug due to inefficacy and 14% stopping the second drug due to an AE (Table 5). However, when switched patients were compared according to the reason for switching, the rates of discontinuation differed significantly.
Table 5. Outcomes of treatment with the second biologic agent*
Reason for switch
Outcome with second biologic agent
Still taking agent at end of April 2005
Stopped for inefficacy
Stopped for adverse event
Values are the number (%) of patients.
Inefficacy (n = 503)
Adverse event (n = 353)
Total switches (n = 856)
The cumulative rate of discontinuation due to inefficacy of the second agent was 16% among patients who also discontinued their first agent because of inefficacy, compared with a rate of only 9% among patients who discontinued their first agent because of an AE. Thus, after adjusting for potential confounders (age, sex, disease duration, baseline DAS28 scores, baseline HAQ scores, DMARD cotherapy, and treatment with oral corticosteroids), the risk of stopping a second anti-TNF agent because of inefficacy among patients who discontinued the first agent because of inefficacy was 2.7 (95% CI 2.1–3.4), whereas the risk of stopping a second agent because of inefficacy was 1.2 (95% CI 0.8–1.6) among patients who discontinued the first agent because of an AE (Figure 1). In other words, if the reason for switching to a second agent was development of an AE following treatment with the first agent, the risk of drug discontinuation due to inefficacy was almost the same for the first and second agents. In contrast, if the first anti-TNFα agent was discontinued due to inefficacy, then the likelihood of recurrent inefficacy of the second anti-TNFα agent was increased ∼3-fold.
The cumulative rate of discontinuation of the second agent due to development of an AE was 20% among patients who also discontinued the first agent because of an AE. In contrast, the cumulative rate of drug discontinuation due to an AE was only 10% among patients who discontinued use of the first agent because of inefficacy. These rates are equivalent to adjusted hazard ratios of 2.3 (95% CI 1.9–2.9) and 1.1 (95% CI 0.9–1.5), respectively (Figure 2). In other words, if the reason for switching to a second agent was inefficacy of the first agent, the risk of developing an AE with a second biologic agent was roughly the same as the risk of developing an AE with the first agent. In contrast, if the reason for the switch was an AE with the first agent, then the likelihood of a recurrent AE with the second agent was increased more than 2-fold. However, of the 71 patients who had recurrent AEs, only 19 experienced the same AE during therapy with the second anti-TNFα agent (rash, n = 12; general malaise, n = 5; infection, n = 1; nasal congestion, n = 1).
The findings from this large UK national cohort of patients with established RA suggest that following failure of a first anti-TNFα agent, the majority of patients who start a second anti-TNFα agent will continue this therapy beyond 6 months. In those patients who discontinued the second agent, the reasons for discontinuation were recurrent; that is, a patient was more likely to stop taking a second anti-TNFα agent because of inefficacy if the first anti-TNFα agent was also stopped due to inefficacy. Similarly, a patient was more likely to experience an AE from a second agent if the first agent was also stopped due to an AE, although recurrence of the same AE was rare. However, the rate of AEs with a second anti-TNFα agent was not increased if a patient had not experienced an AE with the first agent.
This study was based on a large UK cohort of more than 6,000 subjects, among whom more than 850 anti-TNFα switches occurred. However, there are potential limitations to the interpretation of our data. One limitation is attributable to the nature of our study design, since this was a prospective observational cohort study of patients from rheumatology practices in the UK, in which treatment decisions are not randomized but rather left to the treating physician. However, patients cannot be randomized to receive a second agent on the basis of their outcome (inefficacy or toxicity), which was the key factor that divided the patients in our analysis, and any decision to start a second or subsequent course of therapy must be made on an informed basis by the individual patient. Therefore, the present results reflect the treatment outcomes as they have occurred in the “real world.”
In extrapolating these results more widely, it is important to consider the limitations inherent in this population register approach, in terms of the data on both inefficacy and safety of the treatments. Because this was a real-world study, the treating physician determined, based on clinical opinion, whether a patient's treatment had failed, either due to nonresponse or due to toxicity, and thus the initial anti-TNFα therapy should have been stopped. Although it is unlikely that nonresponders would have continued to receive their initial therapy, no strict guidelines to define failure were provided. UK national guidelines do stipulate that treatment should be discontinued after 6 months if improvement of ≥1.2 in the DAS28 has not been achieved. However, results of a UK national survey suggest that this recommendation is not always observed (BSR Biologics Register: personal communication).
Furthermore, the BSRBR records DAS values at 6-month time points following the start of the first anti-TNFα agent, but these time points may not coincide with the decision to change therapy. For this reason, the current analysis focused on drug continuation rather than change in the DAS as the main efficacy outcome.
In addition, due to the nature of the data collection, this analysis could not differentiate between those patients who had a complete lack of response compared with those with a secondary loss of treatment efficacy. The other major caution is that the threshold for stopping therapy may have differed between the first and second agents. In patients with resistant disease, physicians may be less likely to stop a second agent, even if a significant improvement has not been achieved, reflecting the belief that no other options are available. However, the fact that the risk of inefficacy was increased ∼3- fold if the first anti-TNFα had been stopped for this same reason suggests that this is likely to be a true phenomenon and that the estimates of risk may even have been underestimated.
Similarly, any decision regarding the severity of an AE and its attribution is also subject to physician opinion. Furthermore, the nature of the AEs that led to discontinuation also influenced which patients went on to receive subsequent anti-TNFα treatment. Very few patients started a second anti-TNFα agent if the first was discontinued because of an infection. However, there was a very high rate of switching among patients who discontinued their primary anti-TNFα following a presumed drug reaction, either systemic or at an injection site.
Treatment decisions and thresholds for stopping do change both over time and between physicians, which will act to limit the external validity of the results. The data presented herein represent the pooled current experience in the UK. Results of subgroup analyses (data not shown) do not suggest that there have been any important temporal shifts in these thresholds during the course of data collection since 2002. Nevertheless, these data suggest that a substantial proportion of patients may subsequently exhibit a response to one anti-TNFα agent after having shown a lack of response to another. The biologic plausibility of this possibility requires examination.
Adalimumab, etanercept, and infliximab are all potent neutralizers of the bioactivity of TNFα, which is a key cytokine of importance in the pathogenesis of RA. However, there are fundamental differences between the 3 agents, which may explain a differential effect. Infliximab is a chimeric monoclonal antibody with a murine variable and a human IgG1. Adalimumab is a fully humanized monoclonal antibody. Etanercept is a genetic fusion of a recombinant soluble p75 TNFα receptor and the Fc portion of human IgG (IgG1). Although all 3 drugs bind to TNFα, etanercept also binds to lymphotoxin. This cytokine has been found in rheumatoid synovial tissue, although its role remains to be defined (25, 26). Recent work has also shown that complexes formed between etanercept and both soluble and transmembrane TNFα are less stable than those formed with infliximab (27). Both infliximab and etanercept have been shown to induce apoptosis of macrophages but not lymphocytes in rheumatoid synovium (28). However, only infliximab has been shown to induce apoptosis in lamina propria T lymphocytes in patients with Crohn's disease (29), which could be one possible explanation for the differential efficacy of infliximab and etanercept in this latter condition.
There are also differences in the half-lives of the 3 agents, with etanercept having the shortest half-life, only 4 days, and adalimumab having the longest, 14 days (30). The 3 drugs also differ in their dosing regimens. Infliximab is given as a weekly infusion of at least 3 mg/kg for 8 weeks, whereas etanercept is given twice per week at 25 mg, and adalimumab is administered every 2 weeks as a 40-mg subcutaneous injection. The larger but less frequently administered dose of infliximab may result in higher peak serum concentrations compared with the smaller but more frequently administered doses of etanercept and adalimumab, resulting in higher tissue concentrations.
Finally, anti-TNFα agents are associated with the development of antibodies directed against the drug (31). The rate may be higher in patients receiving infliximab, due to its chimeric component. The effects of these antibodies are not clear but may be associated with drug reactions and loss of efficacy over time. Studies of adalimumab and infliximab have shown that the addition of MTX may reduce formation of these antibodies (6, 32). The significance of these differences between the 3 agents in adult RA is unknown. Unlike the situation with Crohn's disease, in which a definite difference in efficacy between infliximab and etanercept has been shown, neither a meta-analysis (33) nor longitudinal followup studies (7, 8) have demonstrated any significant differences between the 3 drugs in patients with RA. However, given the complexity of the underlying inflammation in an individual patient with RA, these factors may become important.
In conclusion, these data from a large unselected population of RA patients suggest that, based on treatment continuation rates, there is a strong case for switching patients to a second anti-TNFα agent when failure to respond to the first agent occurs. In fact, our findings showed that more than 70% of patients continued on a second agent for at least 6 months. This was the case for patients stopping either because of inefficacy or because of certain AEs. However, our data also indicated that the reasons for anti-TNFα failure were recurrent. Moreover, our study identified a group of patients with multiple anti-TNFα failures. Thus, further studies will be needed to determine whether there is a role for yet a third anti-TNFα agent in these patients or whether it is better to move on to a different class of biologic agent.
Dr. Silman had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.
Study design. Drs. Hyrich, Lunt, Symmons, and Silman.
Acquisition of data. Drs. Hyrich, Watson, Symmons, and Silman.
Analysis and interpretation of data. Drs. Hyrich, Lunt, Watson, Symmons, and Silman.
Manuscript preparation. Drs. Hyrich, Symmons, and Silman.
Statistical analysis. Drs. Hyrich, Lunt, and Silman.
ROLE OF THE STUDY SPONSORS
The BSRBR was established primarily to investigate the safety of biologic agents in routine practice. This registry receives indirect financial support from the pharmaceutical companies that market biologic agents in the UK (Schering Plough, Wyeth, Abbott, and Amgen), but the independence of the BSRBR and its investigators is assured. The resources used to fund the BSRBR are received under contract with the BSR, which also funds a research grant under a separate contract to the University of Manchester, allowing the investigators normal academic freedom in relation to the use and analysis of the data. Under the terms of the contract between the BSR and the sponsoring pharmaceutical companies, all publications are sent in advance to the sponsors prior to submission, to review for factual errors only. The sponsors had no involvement in the study design, content, or interpretation of the data and had no influence on the publication of the data.
We appreciate the enthusiastic collaboration of all consultant rheumatologists and their specialist nurses in the UK in providing the data used in this report. The substantial contributions of Andy Tracey and Katie McGrother in database design and manipulation are acknowledged. We also thank Dr. Ian Griffiths (Chairman of the BSRBR Management Committee) and Professors Gabriel Panayi, David G. I. Scott, and David Isenberg (BSR presidents during the period of data collection) for their active role in enabling the BSRBR to undertake its tasks, and Mervyn Hogg and Samantha Peters (Chief Executive Officer of the BSR and BSR staff member, respectively) for considerable administrative support.