Is There an Increased Risk of Malignancy and Serious Infection Among RA patients Receiving High-dose Anti-TNF Therapy?
Whether TNF inhibitor therapy can be independently linked to an increased risk of malignancy or serious non-opportunistic infection has been difficult to discern since RA itself is associated with an increased risk for both of these outcomes. Thus far, safety registries and population-based epidemiologic studies have been used to address these issues, as data from clinical efficacy trials are limited by inadequate numbers of patients to discern differences for rare outcomes, populations that may not be generalizable and, in general, follow-up times are short. Here, Bongartz et al (JAMA 2006; 295: 2275) investigate the risk of serious infections and malignancies in TNF inhibitor treated patients using meta-analysis of results from efficacy trials of infliximab and adalimumab.
Methods: Published and unpublished phase II-IV randomized, placebo-controlled clinical trials of adalimumab or infliximab of at least 12 weeks duration were selected. Reported serious infections (defined as requiring antimicrobial therapy or hospitalization) and first malignancies (including non-melanoma skin cancers) were verified from safety data reported to the FDA. Sub-analyses were stratified by high-dose (infliximab < 3 mg/kg every 4 weeks or adalimumab 20 mg per week) and low-dose (infliximab > 6 mg/kg every 8 weeks or adalimumab 40 mg every other week) TNF inhibitor therapy. Fixed-effects meta-analysis was performed to estimate the risk of malignancy or serious infection in TNF inhibitor treated patients compared to placebo treated patients from the combined data set.
Results: Nine trials met inclusion criteria, comprising 5014 patients randomized to receive adalimumab or infliximab (n=3493) or placebo (n=1512). Trials were statistically heterogeneous in terms of included patients. Four of the included trials used infliximab, 3 of which used initial doses of 10 mg/kg (dosed as frequently as every 4 weeks). The remaining adalimumab trials included doses ranging from 20 mg every other week to 40 mg per week. The study duration of included trials ranged from 12 to 54 weeks.
29 verified malignancies occurred in the combined TNF inhibitor treated patients, compared to 3 in the placebo group. Included malignancies included solid tumors, hematologic malignancies, and 9 non-melanoma skin cancers. Time to diagnosis of included malignancies ranged from 2 to 45 weeks after randomization. Using all included malignancies, the risk of any malignancy was 7.91 times higher in the TNF inhibitor treated patients compared to placebo treated patients (95% CI 1.29 1652.0). The risk was decreased if malignancies occurring in the first 6 weeks of the trial were omitted (OR 4.5 (95% CI 1.3 15.8)) or if non-melanoma skin cancers were omitted (OR 3.7 (95% CI 1.0 13.2)).
When stratified by dose of TNF inhibitor, high-dose therapy was significantly associated with an increased risk of malignancy (OR 4.3 (95% CI 1.6 11.8). However, low-dose therapy was not significantly associated with an increased risk (OR 1.4 (95% CI 0.3 5.7).
126 serious infections were reported in the combined TNF inhibitor treated group compared to 26 in the placebo treated group, corresponding to a two-fold increased risk (95% CI 1.3-3.1). When stratified by dose of TNF inhibitor, both high- and low dose therapy were significantly associated with an increased risk of infection (OR 2.3 (95% CI 1.5-3.6) and OR 1.8 (95% CI 1.1- 3.1), respectively).
Conclusions: Meta-analysis of pooled data from randomized efficacy trials of adalimumab and infliximab indicate and increased risk of malignancy among RA patients treated with high-dose TNF inhibitor therapy, and an increased risk of serious infection at both high and low dose compared to patients randomized to placebo.
Editorial Comment: While the analysis appears to be sound and the outcomes non-random, there are several issues that might suggest the conclusions of this carefully constructed meta-analysis may not be robust for both malignancies and infection.
For malignancies:
Lead time. It is generally felt that a lead time of many months to several years precedes clinically diagnosable cancer. Thus, it is problematic to attribute solid tumors, such as lung, breast, and colon cancer, to a therapy initiated only a few weeks to months prior to diagnosis. Even the sub-analysis excluding malignancies occurring in the first 6 weeks of the trial is problematic from this standpoint, and it would be interesting to know if the results remain significant if another lead-time cutpoint (such as a still conservative 3 months) were used instead. Non-melanoma skin cancers are thought to have an even longer lead time and are thus even more problematic when trying to ascribe temporal causality with treatment. Furthermore, due to their generally benign nature, non-melanoma skin cancers are not typically omitted from this type of malignancy assessment.
For infection:
Classification of serious infections. A common definition of serious infection will include hospitalization, the need for parenteral antibiotics, typical signs and symptoms (i.e. fever, chills), preferably with an identified organism. Here, all reported infections that received any antimicrobial (and thus, conceivably, patients with rhinorrhea who received an antibiotic would be classified as having a serious infection). The authors do not provide additional information regarding the nature of the individual infections, although those deemed serious in the original publications (such as pneumonias and other infections requiring hospitalization and parenteral antibiotics) are accounted for in the individual manuscripts. A more complete picture would involve a separate analysis of these infections only.
Dose. The original publication from the infliximab ATTRACT trial (included in the meta-analysis presented here) identified an increased risk for serious infection in the patients initiated on a dose of 10 mg/kg. This was confirmed in START (also included here). For this reason, initiating infliximab at this dose has not been recommended and is not novel information.
Another issue is the classification of high- vs. low-dose therapy, in which there is overlap of dosing between the two classification criteria. In addition, the standard dose of adalimumab (40mg every other week) is classified here as high-dose.
The results presented here, while provocative, require additional confirmation. The concept of using meta-analysis for safety data is novel and avoids the problem of treatment allocation encountered in population based epidemiologic studies and safety registries. However, the advantage gained by having non-biased treatment allocation may be lost and surpassed by the issues raised above.
