Tumor necrosis factor antagonists and infections: The small print on the price tag
Tumor necrosis factor (TNF) antagonists are biologic response modifiers that have significantly impacted the outcomes of patients with chronic inflammatory disorders such as rheumatoid arthritis (RA) or ankylosing spondylitis (1, 2). Since their introduction in 1998, the indications and use of these agents have markedly expanded with more than half a million individuals worldwide receiving anti–TNF therapy.
Although infectious complications were seriously considered as a possible side effect of TNF antagonists, prelicensure, randomized, controlled trials failed to reveal a statistically significant increase among patients receiving TNF antagonists (1–7). The safety data were compelling and led to the perception that other than the tremendous direct drug costs there would be little else added to the price tag. Concerns based on work in animal models indicating an essential role of TNF in combating infection (8–11) seemed to be proven wrong.
Some indications of possible concerns regarding increased infection risk in humans were present in prelicensure trials. For example, in the continuing second 6-month extension of the Anti–Tumor Necrosis Factor Trial in Rheumatoid Arthritis with Concomitant Therapy (ATTRACT) study of infliximab, 2 patients (out of 340) in the anti–TNF group died because of atypical infections (2).
Shortly after adoption for routine use in patients with RA, case reports and postmarketing surveillance data raised awareness that infections may pose a significant threat to patients undergoing anti–TNF therapy. The reported infections were exactly those anticipated from the animal model studies: serious granulomatous infections such as tuberculosis (TB) (12–15), endemic fungal infections (16–19), listeriosis (20, 21), Pneumocystis carinii pneumonia (22), and aspergillosis (23).
In their review of the Food and Drug Administration Adverse Event Reporting System (FDA-AERS) data, Wallis et al reported 374 cases of TB, 42 cases of histoplasmosis, and 223 cases of other granulomatous infections among 346,000 US patients treated with BRMs over a 4.5-year period (24). The annualized risk of TB (0.27%) was 8 times higher than the estimated annual risk of new TB infection in the US (0.03%). Lending plausibility to causality, the median time to onset of infection was 236 days for etanercept and 40 days for infliximab, and 38% of these cases occurred in the absence of other immunosuppressive agents. A further 25 cases of TB reported to the FDA-AERS have recently been reviewed elsewhere (25).
A principal weakness of these spontaneous reports is the lack of a control group that would allow assessment of whether these events were due to disease or drug-related factors in the face of an already well-established increased risk of infection for patients with RA (26). The strength of the surveillance data and the anecdotal reports is their consistent reporting of particular types of infections, including those caused by intracellular organisms, and the biologic plausibility of increased susceptibility due to anti–TNF treatment based on the available knowledge about the importance of using TNF to control them.
The concerns raised from the bench have reached the bedside.
During the last year, data from controlled studies have emerged that challenge the previously presumed safety profile of anti–TNF therapy and impose a perhaps greater degree of uncertainty in weighting potential benefits and risks of anti–TNF treatment in the setting of routine clinical practice. Kroesen et al compared the incidence of severe infections (defined as hospitalization and/or need for intravenous antibiotic therapy) in 60 patients 2 years before and after receiving anti–TNF therapy (27). The incidence of serious infections increased from 0.008 to 0.183 per patient-year, and 18.3% of the patients had a severe infection during the observation period under anti–TNF therapy (27).
A large, randomized, controlled trial of the anti–TNF antibody adalimumab including ∼200 patients per treatment group demonstrated a 6-fold increase of severe infections in patients with RA receiving anti–TNF therapy compared with those taking placebo (28). In addition, the largest randomized, controlled trial of an anti–TNF agent, including 1,049 patients with early RA, revealed that 2.1% of 291 patients receiving methotrexate (MTX) plus placebo, 5.6% of 372 patients receiving infliximab 3 mg/kg plus MTX, and 5.3% of 377 patients receiving infliximab 10 mg/kg plus MTX developed at least 1 serious infection (29). Pneumonia was the most common serious infection in this study, and TB occurred in 5 of the 749 patients who received infliximab.
An observational study in Spain used a national registry to calculate the incidence rate of TB in a cohort of patients treated with these agents and compared it with an age- and sex-matched cohort of patients with RA assembled prior to the era of anti–TNF treatment (30). There was a 19-fold increase of TB in patients treated with TNF inhibitors compared with patients with RA not exposed to this therapy. The high incidence rate dropped dramatically after the introduction of recommendations concerning screening for latent TB prior to initiation of anti–TNF therapy (30).
Why was there no signal for an increased rate of infection in prelicensure randomized controlled trials?
At issue is whether there was a discrepancy in infection reporting between early controlled licensing studies and the postmarketing trials and surveillance reports. In fact, there was no discrepancy: most of the prelicensure randomized controlled trials were designed to prove effectiveness, not to disclose specific and rare side effects. A sample size of 81–87 patients per dose group such as in the ATTRACT study (2) may provide good estimates of efficacy, but has only a 58% chance to reveal even a 3-fold increase in the incidence of severe infections, given the seemingly high baseline rate of 6% infections in the placebo group (power calculation for dichotomous data with n = 85, m = 1, P0 = 0.06, P1 = 0.18, α = 0.05).
If we presume a baseline rate of severe infections in patients with RA of 1% per year as reported in a recent population-based study (26), a clinical trial designed to have an 80% chance of detecting a 3-fold increase would require 865 patients per group (sample size calculation for dichotomous data with β = 0.8, m = 1, P0 = 0.01, P1 = 0.03, α = 0.05), a sample size not even approached in any controlled clinical trial of one of these agents to date.
Which patients should be considered for anti–TNF therapy?
There is no doubt that anti–TNF agents have added a very effective therapeutic option in rheumatic disease management, most notably because of their ability to improve measures of disease activity and outcomes in patients who are not candidates for or who fail to respond to conventional disease-modifying drug treatment. This may be especially true for diseases such as ankylosing spondylitis, for which few adequate treatment options were available until the striking effectiveness of TNF inhibition redefined therapy of this disabling disease (1).
Although these agents are efficacious in the majority of patients, clinicians are left with many unanswered questions when weighing the efficacy against the hefty cost and legitimate safety concerns. Patients treated with anti–TNF drugs in randomized controlled trials are highly selected and represent only a small proportion of patients seen in routine care. Consequently, controlled data concerning efficacy and side effects are not available for a majority of patients (31).
Most prelicensure, randomized, controlled trials compared anti–TNF therapy with a disease-modifying antirheumatic drug (DMARD) that already failed to control disease activity, a phenomenon that is encountered with all available traditional DMARDs and anti–TNF agents. There are several effective approaches to the treatment of patients with suboptimal response to MTX (32–34), but to date, there is no head-to-head comparison with anti–TNF drugs.
Although not true head-to-head comparisons of standard care versus anti–TNF therapy, the results of 2 published trials evaluating TNF inhibition and traditional DMARD therapy in early arthritis may provide further insight into this problem (3, 29). The first compared etanercept with MTX, finding both agents able to completely halt radiographic progression according to the Sharp score in more than half of patients (MTX in 52%, etanercept in 63%), with no statistically significant difference in the clinically meaningful American College of Rheumatology 50% and 70% response after 2 years (3). The second trial (29), comparing MTX monotherapy to infliximab/MTX combination therapy, confirmed the results of previous trials (35–37) by showing that a combined-therapy regimen offers additional control of disease activity and radiographic progression over a single DMARD regimen.
These studies did not use combination therapies of traditional DMARDs or addition of low-dose steroids for comparators, strategies that are commonly used in routine clinical practice. Additionally, patients in the infliximab/MTX trial had high disease activity but were not allowed to receive a steroid bolus to bridge the period until onset of MTX effect. As a result, patients in the monotherapy group were without effective treatment for at least 8–12 weeks despite their high disease activity.
The point is that the tradeoff between clinical benefits and the risks of these therapies, including that of infection, may not be as favorable as it appears from the clinical trial situation. A degree of poor disease control in carefully screened patients, which would not be acceptable in daily practice, creates an unrealistic magnitude of benefit in the comparison of patients treated with anti–TNF drugs with those receiving conventional DMARD therapy.
How can we prevent infectious complications in patients treated with anti–TNF agents?
The increased frequency, severity, and atypical presentation of infections associated with TNF inhibitors mandate the need for effective preventive strategies. Certain patients may be at greater risk due to concomitant immunosuppression, their underlying diseases, and travel or infectious disease exposures. Some of these infections may be preventable by personal protective measures, some by immunization, and some by drug prophylaxis, but the risks and benefits of each of these strategies have yet to be defined for this patient population. Patients must be taught before and during therapy about signs and symptoms of possible infectious complications, and they should have instructions about what to do in case such symptoms appear. This includes ready access to competent professional advice 24 hours a day. A biologic therapy information card with telephone numbers and contact information must be given to every patient.
Clearly, TB has been identified as a significant risk for patients receiving anti–TNF treatment. The drop in occurrence of TB cases in Spain after introduction of recommendations concerning screening for latent TB prior to initiation of anti–TNF therapy (30) should underline the importance of pretreatment TB screening. However, TB cases occurred also in controlled studies where patients had a negative purified protein derivative skin test prior to inclusion (28). A major problem with this screening strategy is that relatively high numbers of false-negative test results can be observed in immunocompromised patients (38). This fact should sharpen awareness of the major limitations of purified protein derivative testing. The significance of an abnormal chest radiograph prior to treatment is also unclear in this population, and a high number of false-positive results can be expected considering the frequency of lung abnormalities in patients with RA.
Because of a lack of controlled studies, recommendations for pretreatment TB screening have been extrapolated from guidelines followed in other immunocompromised hosts (39). Pneumococcal vaccination is recommended routinely prior to initiation of treatment, with reimmunization every 5 years; yet the immunogenicity, safety, and timing of immunization in this group is unclear (40). TNF inhibitors decrease the primary IgG response to Streptococcus pneumoniae, and patients with RA receiving TNF inhibitors are less likely to respond to the 23-valent vaccine (41). Ideally, immunizations should be administered prior to receipt of any immunosuppressive therapy.
Registration of patients undergoing anti–TNF therapy whenever surveillance programs are available and reporting of any severe adverse event to regulatory bodies should be seen as integral components of treatment. The significant drop in newly diagnosed cases of TB following the introduction of recommendations for routine screening for latent TB (30) demonstrates the importance and success of an ongoing pharmacovigilance program.
Similar to the introduction of glucocorticoids in the treatment of RA in 1948, the use of TNF-blocking agents since 1998 has created great excitement among physicians about the dramatic treatment responses to the agents. But just as toxic glucocorticoid effects soon appeared in the 1950s, the severe side effects of anti–TNF therapy became clearer after their introduction into widespread clinical use.
As we pursue an ambitious agenda of clinical trials with targeted therapies designed to halt disease progression early in the disease course, concerns continue to be raised by the appearance of side effects that, unlike glucocorticosteroids at the time of their introduction, can in many cases already be anticipated by available knowledge about the physiologic role of the targeted molecules. It stands to reason that the targeted, often highly conserved molecules would not have been maintained during evolution (42) if the benefits of their function did not outweigh their adverse effects. The anticipated beneficial effects of blocking certain molecules with a biologic drug will inevitably be accompanied by side effects, which themselves can be anticipated to some extent. To disclose the small print on the price tag of these promising therapies as early as possible, the design of future clinical trials should not only aim at the targeted effect, but should also consider targeted side effects and the appropriate power to detect them.
The large, recently-published randomized, controlled trials of anti–TNF agents including approximately 600 and 1,000 patients, respectively, provide valuable data to draw conclusions about their short-term safety profile in selected patient populations (28, 29). However, even large sample sizes will not be sufficient to detect some severe but rare or late-occurring side effects such as lymphomas or solid malignancies. Long-term pharmacovigilance by industry, investigator, and/or publicly initiated databases is an essential element for a complete appraisal of these agents.