Introduction
The chronic nature of rheumatoid arthritis (RA) and its progression over time in spite of a variety of treatment options implies that long-term treatment will most often involve a sequence of therapies. The optimal therapeutic sequence strategy will be determined largely by the patient's response to therapy and by disease progression, as well as detailed knowledge of the role of different therapies along treatment pathways. Thus, understanding the effectiveness of different therapeutic sequences is of particular importance in the evaluation of long-term RA treatment strategies.
There are three main drug classes commonly used in the treatment of RA: nonsteroidal anti-inflammatory drugs (NSAIDs), corticosteroids and disease-modifying antirheumatic drugs (DMARDs). Several studies [
1‐
3] have provided evidence that early treatment with DMARDs results in superior clinical and radiological outcomes. Two main classes of DMARDs are available for the treatment of RA: synthetic DMARDs and biologic DMARDs. Oral administration, lower cost and greater prescriber familiarity support the use of synthetic DMARDs as a first-line strategy. Biologic DMARDs, most often in combination with synthetic DMARDs, are generally reserved for the treatment of patients with moderate to severe RA who have had an inadequate response or have developed toxicities to synthetic DMARDs [
4].
A review of 16 clinical practice guidelines and 20 consensus statements on RA treatment revealed that while tumor necrosis factor (TNF)-α inhibitors were consistently recommended for patients with active RA and a history of inadequate response to synthetic DMARDs [
5], the management of patients who stopped an initial TNF-α treatment because of lack of initial response, loss of initial response or side effects continues to be the subject of much debate, and guidelines for patient management are nearly absent. Despite the lack of guidelines, it is estimated that upon encountering an inadequate response or side effects with a TNF-α inhibitor, over 90% of rheumatologists in the United States switch patients to a different TNF-α inhibitor [
6].
Estimates of efficacy rates of TNF-α inhibitors may depend on a number of factors, including patient characteristics, such as disease duration, prognostic factors, number of previously failed DMARDs and disease activity, as well as the dose of TNF-α inhibitor and the designs of the studies from which they were obtained. Despite some variation attributable to these factors, estimates derived from randomized, controlled trials (RCTs) suggest that between 40% and 50% [
7] of RA patients treated for at least 6 months with one of the three first-generation TNF-α inhibitors (etanercept, adalimumab and infliximab) failed to achieve the American College of Rheumatology 50% (ACR50) improvement criteria [
8], while the results from a large, registry-based study [
9] indicated that over 70% of these patients fail to achieve Disease Activity Score 28 joint count (DAS28)-defined "remission" (DAS28 <2.6).
Although the efficacy of TNF-α inhibitors in patients who are naïve to biologic treatment has been evaluated in multiple studies [
10‐
12], evaluating the efficacy of these drugs in patients who have already experienced an inadequate response to a TNF-α inhibitor poses greater methodological challenges. One key aspect of evaluating the efficacy of sequential TNF-α therapy is to determine whether the probability of responding to a TNF-α inhibitor depends on the results of prior treatment with these drugs. Early evidence from small observational studies suggested that a significant proportion of patients who had an inadequate response to an initial TNF-α inhibitor benefited from subsequent treatment with an alternative TNF-α inhibitor [
13‐
15]. Recent data derived from registries, however, have suggested that the response rates of patients switching to a second or third TNF-α inhibitor are often lower than the response rates of patients to the first TNF-α inhibitor [
16,
17]. Moreover, the broader question whether it is more effective to switch to another mechanism of action or to use a second TNF-α inhibitor after the patient has had an inadequate response to a first one has not been formally addressed.
In the present study, several biologic treatment options currently available to RA patients with an inadequate response to an initial TNF-α inhibitor were evaluated using evidence gathered from published reports. We undertook a systematic review of published, peer-reviewed studies that reported clinical outcomes of biologic treatment among this group of patients. Our study expands on previously published reviews in two ways: first, information on efficacy rates of newer biologics with different mechanisms of action among patients with an inadequate response to TNF-α inhibitors was also included and results were examined separately for TNF-α inhibitors and other biologic DMARDs; second, a quantitatively based evaluation of the relationship between response to biologic treatment and the number of failed TNF-α inhibitors was undertaken by summarizing the results of published studies. Within the limitations of the existing data, potential effect-modifying factors, such as study design and treatment duration, were also examined. A secondary objective of this study was to determine whether clinical response to a subsequent TNF-α differed by reason for discontinuation.
Materials and methods
Search strategy
A search was carried out in the PubMed database using each of the following search terms as keywords or text words: "golimumab," "adalimumab," "infliximab," "etanercept," "abatacept," "rituximab," "anakinra," "tocilizumab," "certolizumab pegol," "anti-TNF," "TNF-antagonist," "TNF-inhibitor," "biologic*" in combination with "switch*" or "sequential therapy" or "therapy interchange," and "rheumatoid arthritis." Brand names of biologics were also used for each of the drugs cited above. The search was restricted to the English language and had an end date of 31 December 2009. The reference lists of selected review publications were further examined to identify any studies that were not captured by our search.
Articles were included in the analyses if the publications reported any quantitative clinical and/or health-related quality of life outcomes for RA patients previously failing one or more TNF-α inhibitors. Studies with fewer than 20 participants were excluded.
Database development
The characteristics of each study were recorded, including the study design and major findings. Disease duration, age, sex distribution, duration of treatment, duration of washout period (if reported), concomitant use of methotrexate (percentage of patients within the group), dose of biologic drug and all clinical and quality-of-life measures were recorded for each group of patients on the basis of the total number of biologics that had been tried at the time the outcome was measured. Studies differed with respect to the way in which washout periods were reported. For this study, washout periods were noted in the following manner: (1) if the mean or median was reported (the median was preferred if both were reported) for the time elapsed between the last dose of prior treatment until the first dose of subsequent treatment, this value was recorded; and (2) if no summary statistic for the washout period was reported, the minimum washout period required per study protocol was recorded. For RCTs in which different doses of biologic DMARDs were administered, efficacy estimates based on all study arms were included. The sensitivity of the results to this parameter was assessed in the analyses.
Some studies reported outcomes of multiple switches for the same group of patients, so the same group of patients might have contributed to more than one combination of outcome measures and number of biologics tried. A few studies did not report results disaggregated by the actual number of prior TNF-α inhibitors tried and reported only the results of the biologic under study for subjects with an inadequate response to at least one TNF-α inhibitor (in these cases, the number of biologics under study was recorded as 2+).
Several studies did not allow for within-study evaluation of differences in clinical or health-related quality-of-life outcomes across groups differing in the number of previous TNF-α inhibitors used. For these studies, the results of various outcome measures were reported for a single comparison group.
Efficacy estimates
All estimates were evaluated for each combination of measure and biologic number (that is, first, second, and so on), as well as for relevant subgroups. All estimates were evaluated as weighted averages using sample size as the weight in the following formula:
where Ratejrepresents the average response rate for measure j, i indexes the group, niis the sample size for the ith group and N is the combined sample size of all groups.
The main focus was on estimating the efficacy rate on the basis of each of the main response criteria reported in the studies identified by our review across the number of previously failed biologics. Nevertheless, some of the publications included in the current study also reported efficacy rates associated with a first trial of TNF-α treatment. Weighted estimates were also evaluated for this group of patients and served as a further check of how the values obtained in the current study compared to published rates. Estimates were also evaluated within the following subgroups: type of study (observational study versus RCT), duration of follow-up (<6 months versus 6 months or longer), type of biologic (TNF-α inhibitor versus other) and reason for discontinuation (lack of response, loss of response or intolerance).
Discussion
In the current study, the association of response to subsequent biologic treatment with number of previous TNF-α inhibitor treatments was evaluated on the basis of data reported in peer-reviewed publications. After combining these data, the results indicated that an association does in fact appear to exist and that response is likely to decline with increasing number of previous TNF-α treatments. Our results also suggest that the pattern of decreasing response for increasing number of failed TNF-α inhibitors was maintained even when the analyses were restricted to more homogeneous groups of studies. Importantly, we found that the relationship was also maintained across a number of important RA response measures, which further contributes to the validity of the findings. Furthermore, the ACR response rates derived in the current study for patients with no prior exposure to biologic drugs are in line with those reported in previous studies that examined 6-month outcomes from RCTs using patients previously not exposed to biologic DMARDs [
10].
One limitation of the current study is that the degree of heterogeneity across studies was not summarized quantitatively. Nevertheless, it must be clear from our exposition regarding the difficulties in combining results across studies that a large degree of variation in study design and patient characteristics was present. In addition, the results that were central to our research question were frequently incidental to the primary objectives of the studies that were reviewed, which meant that information specific to groups of patients differing in the number of prior TNF-α inhibitors used was often lacking. Consequently, no formal statistical inference was undertaken, which is another limitation of the study.
Despite having found several results supporting a trend of lower efficacy rates with increased number of previous TNF-α inhibitors used, we believe that more research into this topic is needed before a conclusion can be reached. In particular, the safety of biologic treatment specifically in patients with an inadequate response to TNF-α inhibitors must also be addressed before the strategy of switching can be fully evaluated vis-à-vis alternative therapies.
One potential key factor in predicting response after a switch to an alternative biologic drug is the reason for discontinuation of the prior drug. The current study examined this aspect of RA treatment within the scope of the available data on this topic, which did not enable a comparison of patients who switched to a second biologic DMARD that was not within the TNF-α inhibitor class. Although several studies did report reasons for discontinuation of prior TNF-α inhibitor treatment, only a few reported response rates for groups of patients who differed in the reason for discontinuation. Our results indicate that patients who discontinue treatment as a result of adverse events are more likely to achieve a clinical response to a second TNF-α inhibitor than are patients who discontinue the first TNF-α inhibitor for efficacy-related reasons. Higher response rates [
30] and greater declines in DAS28 values [
49] among safety failures have been reported in other studies. Nevertheless, some caution should be taken when interpreting these findings, since the assessment of these rates could be sensitive to the length of follow-up in a particular study. Two large studies based on registry cohorts [
16,
45] reported that the reason for discontinuation of a first TNF-α inhibitor was likely to explain the reason for discontinuation of a second TNF-α inhibitor, while the findings from another large registry cohort could not confirm this relationship [
50]. Further, response rates in a RCT of the newer biologic agent tocilizumab [
52] were nearly identical for patients who discontinued treatment because of inefficacy or for unrelated reasons. Overall, we observed consistently higher response rates among safety failures than among primary and secondary efficacy failures. Data obtained over longer follow-up periods, however, are needed to confirm or disprove the findings of the current study.
The exact form of the relationship between response to treatment and number of failed TNF-α inhibitors is likely to play an increasingly important role in defining treatment strategies for RA patients who have had an inadequate response to treatment with TNF-α inhibitors. Given the high cost of biologic agents and evolving knowledge of their safety profiles, there is a growing need to compare and evaluate the relative benefit of strategies involving alternative sequences of therapies. The recent availability of a number of newer biologic drugs with different mechanisms of action makes this need even more salient. Establishing patterns of response along the treatment pathway is a key element of these evaluations, as is the identification of subgroups of patients who may differ in these response patterns. In addition to providing insight into the existence of an association between treatment response and increasing number of failed TNF-α inhibitors, the current study has also offered an overview of the multiple difficulties that are faced when synthesizing evidence pertaining to this question. Differences in reporting, study design and the overall availability of data have made this task a difficult one, which is apparent from the results presented herein.
Given that a substantial proportion of patients will fail an initial biologic treatment, establishing when and how to initiate treatment with these agents is just as important as establishing the relative value of long-term treatment strategies. In many instances, these long-term strategies involve a sequence of treatments, so the question arises at each step which therapy to use as a replacement when a particular therapy must be stopped because of inefficacy or intolerance. Multiple studies have addressed this question in a number of different ways. To date, the large majority consist of observational studies. Some of the results of earlier observational studies were based on small samples and relatively short follow-up periods [
29,
44,
46,
51], but more recently the accumulation of data from several registries of RA patients [
61] treated with biologic drugs has provided larger sample sizes and longer duration of follow-up. In addition, while the older biologic DMARDs were compared primarily with nonbiologic DMARDs, RCTs of newer biologic drugs have used samples of patients with an inadequate response to one or more biologic DMARDs, primarily TNF-α inhibitors [
63]. Several reviews [
64‐
68] have attempted to summarize data from these studies, but no clear guidance has emerged from these publications. A recent meta-analysis [
69] presented a quantitative evaluation of the effectiveness of switching treatments, specifically between TNF-α inhibitors. These results suggested that the probability of achieving a clinical response declines after the first TNF-α inhibitor, a trend that was observed in our study as well, even when other biologic DMARDs were considered. Although data were limited, we found that the magnitude of this decline may depend on the type of biologic drug administered. For example, DAS28-based results suggested that when the third biologic is a TNF-α inhibitor, response rates may be lower than those of an alternative biologic. Although readers are cautioned to consider the limitations of the available data and the preliminary nature of the findings, our study strengthens and extends the findings presented in the meta-analysis of Lloyd
et al.
[
69].
Establishing the existence and magnitude of a relationship between responses to biologic drug treatment and the number of previously failed TNF-α inhibitors has a number of important implications. From a clinical practice perspective, understanding such a relationship can help clinicians to decide on a treatment to use from among an increasing number of alternative strategies. From a research perspective, the current investigation may help inform future studies that involve the treatment of RA patients over the course of their lifetimes, such as economic assessments of RA therapies based on treatment sequence models. One of the key objectives of this work was to provide a more quantitative analysis of the data reported in studies of patients refractory to TNF-α inhibitors, something that was lacking in the current literature. In addition, the current study has exposed some of the difficulties associated with combining results across studies with different designs, patient populations and reported outcome measures. Exposing these limitations may improve the design of future research and foster greater harmonization of RA clinical studies that aim to investigate the effects of sequential biologic drug therapies in RA.
This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/. The Creative Commons Public Domain Dedication waiver (http://creativecommons.org/publicdomain/zero/1.0/) applies to the data made available in this article, unless otherwise stated in a credit line to the data.
Competing interests
GVW, TK, JB and SHZ are employees of Pfizer Inc., the sponsor of this study. RRB, MY and MK have served as consultants for Pfizer.
Authors' contributions
RRB and GVW performed the literature search and data analysis and wrote and edited the manuscript. TK, JB and SHZ assisted with the editing of the manuscript, the subject matter content and the conclusions. MY and MK assisted with the development of the manuscript outline, the conceptualization of the study and data interpretation. All authors were involved in discussion of the findings as well as in the drafting and final approval of the manuscript.