Three anti-tumor necrosis factor-alpha (anti-TNFα) therapies are approved for rheumatoid arthritis (RA) by the US Food and Drug Administration: infliximab (Remicade^®), adalimumab (Humira^®), and etanercept (Enbrel^®). Two more will come soon (certolizumab pegol and golimumab). Although similarities clearly predominate when comparing the three available anti-TNFα agents, a number of clinical differences in efficacy or safety have been noted 12. First, the half-lives - 3 days for etanercept, 10 days for infliximab, and 13 days for adalimumab - may translate into differences in the duration of TNFα neutralization 2. Also, the two monoclonal antibodies, infliximab and adalimumab, have very strong affinity for TNFα, increasing the percentage of neutralized TNFα molecules. In addition, the complexes formed when monomeric and trimeric soluble and membrane-associated TNFα molecules bind to the anti-TNFα agent are far more stable with infliximab and adalimumab than with etanercept. Finally, the monoclonal antibodies are highly specific for TNFα, whether soluble or at the membrane level, whereas etanercept binds to lymphotoxin-α in addition to soluble TNFα, leading to the control of another possible pathogenetic pathway. Soluble TNFα binds to the fusion protein, becoming unable to act on its cellular receptor. Thus, etanercept has a buffering effect on TNFα, and this effect is probably reversible and does not result in permanent elimination of TNFα molecules. Furthermore, binding of etanercept to membrane-associated TNFα does not cause cell lysis. Infliximab and adalimumab can bind two soluble or membrane-associated TNFα molecules, forming a stable and long-lasting complex and causing cell lysis (for example, macrophages and some T-cell subsets) or cell function impairments 2. These differences may influence the risk of immune response impairment and the ability to ward off infections, explaining the greater risk of tuberculosis with infliximab and adalimumab than with etanercept. Immunogenicity seems extremely weak for etanercept and adalimumab but higher for infliximab, inducing antibodies to its murine component (human anti-murine antibodies, or HAMA) and leading to allergic reactions and the often-seen escape phenomenon 2. All of these data have led physicians to treat RA patients who experience treatment failure with one anti-TNFα agent (due to either inefficacy or toxicity) by switching to a second anti-TNFα agent, although the clear-cut benefits of switching are unknown because no controlled trial has ever been conducted.

Rituximab, or anti-CD20, is an antibody used in RA, whereas abatacept is a dimeric fusion soluble protein made of the extracellular part of CTLA-4 present on T cells and Fc of IgG1. It links CD80/86 on antigen-presenting cells with a higher affinity than CD28, thus preventing the costimulation. Tocilizumab is a humanized antibody that links both soluble and membranous interleukin-6 receptor. The differences in the mechanism of action should allow clinicians to rescue patients not fully responding to a TNFα blocker since a different pathway is targeted; however, a definite analysis of the gain of effect in terms of disease activity score (DAS) remission or of an American College of Rheumatology 70% improvement criteria (ACR70) response - that clearly allows clinicians to identify the crucial pathway alternative to TNFα - has not been provided. The aim of this study was to investigate the evidence in the literature about the efficacy of switching between different biologics in RA patients.

We performed a search on MEDLINE, EMBASE, and the Cochrane Library from inception to December 2008 to identify all of the available articles. The terms we used were 'arthritis', 'rheumatoid', 'biological agents (infliximab, etanercept, adalimumab, rituximab, anakinra, abatacept, tocilizumab)', 'switch or switching', 'randomized controlled trials', 'multicenter studies', 'clinical trials phase II', 'clinical trials phase III', and 'clinical trials phase IV'. We searched even in the abstract databases of both the European League Against Rheumatism (EULAR) and the ACR from 1996 to the present in order to identify unpublished studies. Articles were selected by applying predefined inclusion and exclusion criteria, and their methodological quality was graded according to the levels of evidence of the Centre for Evidence-Based Medicine (Oxford, UK) 3.

The language of the paper was not restricted. References of the studies were analyzed to find any study that was not included in the electronic databases. A study was included in the systematic extraction of the data if (a) it was published before December 2008, (b) it was about patients with RA, (c) a biological agent was used and failed according to the clinical response (DAS using 28 joint counts [DAS28] of greater than 2.6 or EULAR criteria for poor responders), and (d) a second biological agent was used and clinical outcomes were evaluated. The ACR responses (ACR20, ACR50, and ACR70) were used as efficacy parameters. During the writing process, only articles concerning major response, as ACR70 or the DAS showing disease remission, were included. The extraction of the data was conducted independently by two investigators.

After the systematic literature research, nine articles fulfilled the inclusion criteria and were included in the review. All of the articles were divided according to the biological agent failed and switched (Figure 1 and Table 1).

Several studies analyzed the efficacy and the gaining of major responses in RA patients treated with different biological drugs. Table 1 summarizes the results obtained from our research. Few studies had strong evidence, mainly evaluating efficacy in switching from anti-TNFα to new targeted therapies (rituximab, abatacept, and tocilizumab). The data available suggest that many of the clinical considerations and decisions that we adopt rely on less than strong scientific evidence. Clinicians should consider the data we present as a practical approach in making the everyday decisions to reach the lowest possible level of clinical activity 14.

The data show that switching from infliximab to etanercept is effective, with an ACR20 response of between 33% and 64% according to different studies. Moreover, etanercept seems to be able to maintain the ACR20 response obtained with infliximab when the latter biological agent is stopped because of side effects 15. Data show that a previous failure with etanercept does not influence the response to infliximab with a reach of clinical response. Indeed, there is no doubt that a clinical effect can be reached by switching.

Switching to another biological agent in patients treated with infliximab is due mainly to a loss of response during treatment, whereas in patients treated with etanercept, switching is due to non-response. The effectiveness of adalimumab seems not to be influenced by previous treatment with infliximab.

In regard to patients treated with a second TNFα inhibitor who failed the first, data show a continuation rate of a second TNFα blocker of 73% of switchers after 15 months of follow-up 16. Switchers to a second anti-TNFα drug seem to have high treatment rates of continuation, and first drug withdrawal due to inefficacy is associated with an increased rate of a second drug withdrawal due to inefficacy but not toxicity; similarly, first drug withdrawal due to toxicity is associated with an increased rate of a second drug withdrawal due to toxicity but not inefficacy. All of these data suggest that a clinical response can be obtained from switching, but given the high cost of the drug, it seems questionable to rely on this response value as a way of treating all poor responders (not reaching low disease activity or DAS remission). In fact, in terms of major outcomes such as DAS remission and ACR70, the results are less convincing. Not more than 5% to 15.4% of the patients overall can be led to remission. This was confirmed in the last trial with golimumab recently published, that showed the 10% rescue to the major response 17. This is clearly disappointing when facing the first failure.

In clinical practice, it is reasonable to switch from one anti-TNFα agent to a second one, possibly with a different mechanism of action. It does not appear useful to switch to a third one 18, but it seems worthwhile to treat patients who are resistant to this therapeutic approach with other biological agents, such as anti-CD20, abatacept, or tocilizumab.

Rituximab has been said to be more effective than switching to an alternative anti-TNFα agent. Moreover, it was demonstrated that treatment with rituximab may be more effective than switching to an alternative anti-TNFα agent in RA patients in whom active disease persisted despite anti-TNFα therapy 10, even though the difference seems to be really more statistically than clinically significant. Similar data have been seen in clinical trials that have confirmed the efficacy of switching to abatacept from an anti-TNFα agent, thus suggesting that using a drug with a different mechanism of action may help.

In RA patients who do not respond to the anti-TNFα strategy, another therapeutic option is tocilizumab with a dose-dependent clinical response with a strong discrepancy between DAS remission and ACR70. Data suggest that a strong effect on pain and acute-phase reactants leads to this dissociation. However, comparing the changes in disease activity and response to tocilizumab using the Clinical Disease Activity Index (CDAI) and Simplified Disease Activity Index (SDAI) in three clinical trials (LITHE, OPTION, and TOWARD), Smolen and colleagues 19 recently found significant differences in changes of CDAI and SDAI between treatment with tocilizumab and placebo, with a minimal impact of C-reactive protein on clinical disease activity measures.

The introduction of anti-TNFα marked the beginning of a new era in the treatment of RA. Nonetheless, the efficacy of each biological agent is not similar in all patients, underlying the need for a drug switch between the agents. Moreover, the switch to different biological target therapies could be necessary because of the presence of driving cytokines other than TNFα. In general, an ACR20 can be obtained in 60% or more of the poor responders to the TNFα blockers, and an ACR70-DAS remission value can be obtained in roughly 10% to 15% of the cohorts after the TNFα blockade, with either CD20 depletion or CTLA4 stimulation. The rule of a 10% to 15% gain of effect after TNFα, the major target, seems to be a good, but not ideal, perspective for the future of RA. These data strongly argue for a clear-cut need for biomarkers capable of identifying, at baseline, which RA patients having different pathways will be the best responders to some biologics or poor responders to other biologics. Yet these biomarkers are not at hand. Furthermore, prescribing a biological agent is an important decision that can greatly impact the quality of life of a patient and can be associated with varying medical costs, owing to differences in routes of administration (depending on the agent). Etanercept and adalimumab are administered subcutaneously. Both can be self-administered at home, but they have some disadvantages, including the requirement of more frequent administration and the risk of pain and local site reactions. Intravenous drugs can be inconvenient for patients with respect to traveling and taking time off from work for infusions and also for patients with difficulties regarding intravenous access. Intravenous administration may use more health care resources and require infusion rooms and equipment as well as medical and nursing supervision 20. Anti-TNFα agents require a short time of administration to reach their effect, and anti-CD20 and abatacept have to compete with this very practical issue that needs to be considered when a choice has to be made to reach a major response. In addition, the more frequent administration with some TNFα blockers, required to better control the disease activity status, should lead to a careful analysis of cost in all of these situations.

The efficacy of different biological agents in selected randomized controlled trial populations is not similar in all patients, underlying the presence of different pathways driving the inflammatory process. The rescue to a major response seems to occur at a rate of 10% to 15% among the various biologics. Annual costs and administration modalities need to be taken into account when making therapeutic decisions in non-responders.

ACR: American College of Rheumatology; ACR20: American College of Rheumatology 20% improvement criteria; ACR50: American College of Rheumatology 50% improvement criteria; ACR70: American College of Rheumatology 70% improvement criteria; CDAI: Clinical Disease Activity Index; DAS: disease activity score; DAS28: disease activity score using 28 joint counts; EULAR: European League Against Rheumatism; HAQ: Health Assessment Questionnaire; RA: rheumatoid arthritis; SDAI: Simplified Disease Activity Index; TNFα: tumor necrosis factor-alpha.

GF has received consulting and speaking fees from Abbott Laboratories (Abbott Park, IL, USA), Wyeth (Madison, NJ, USA), Roche (Basel, Switzerland), and Bristol-Myers Squibb Company (Princeton, NJ, USA) (less than €10,000 each). The other authors declare that they have no competing interests.

SA, AL, and GF conceived of the study and participated in its design and coordination, in data acquisition and analysis, and in manuscript preparation. EG and AZ participated in data acquisition and in manuscript preparation. All authors read and approved the final manuscript.