Skip navigation
Link to Back issues listing | Back Issue Listing with content Index | Subject Index

Adverse events with TNF antagonists

Randomised trials
Malignancy
Serious infections
Observational studies
Malignancies
Serious infections
Comment

There is no doubt that TNF-antagonists – antibodies that mop up various forms of TNF in the body – have contributed to substantial benefits for patients with rheumatological and other disorders where the immune system is abnormal. Much greater improvements are now possible, and the benchmarks for treatment have been raised markedly.

That said, fiddling with the immune system brings dangers, including cancer and serious infections. Trying to measure the rate at which these rare events occur, and whether the rates are different with TNF-antagonist treatment is hard because of limited numbers. New studies provide at least some insight, and highlight the problems. It is probably best to look at the evidence from randomised trials first, then compare it with evidence from observational studies.


Randomised trials

A systematic review and meta-analysis [1] used trials of infliximab and adalimumab in rheumatoid arthritis (etanercept was excluded because of a somewhat different mechanism of action). It examined published material for serious adverse events, used FDA information, and discussed serious adverse events with trialists and manufacturers for clarification. The denominator was the number of patients with at least one dose of drug, and the numerator patients with at least one serious malignancy or infection. Nine trials lasting at least 12 weeks (eight were six months to one year) randomised 5,014 patients.

A separate analysis of etanercept [2] included 4,322 patients in 22 trials, with almost 7,000 patient years of follow up, using an integrated safety database.


Malignancy

There were 29 malignancies in 3,192 patients (0.9%) on infliximab or adalimumab, and 3 in 1,428 (0.2%) on placebo (Figure 1); six other lymphomas were detected during follow up, but not included in the analysis. The odds ratio was 3.3 (1.2 to 9.1), and the number of patients who needed to be treated for six to 12 months with infliximab or adalimumab for one to be harmed was 154 (91 to 500).



Figure 1: Cancer rates in randomised trials of infliximab and adalimumab in RA





With etanercept the number of malignancies was not given, but it was claimed that the incidence of overall malignancies remained stable with time and was comparable to that expected in the rheumatoid arthritis population.


Serious infections

There were 126 serious infections in 3,493 patients (3.6%) on infliximab or adalimumab, and 26 in 1,512 (1.7%) on placebo (Figure 2). The odds ratio was 2.0 (1.3 to 3.1), and the number of patients who needed to be treated for six to 12 months with infliximab or adalimumab for one to be harmed was 59 (39 to 125).



Figure 2: Serious infection rates in randomised trials of infliximab and adalimumab in RA





For etanercept, medically important infections (intravenous antibiotic or hospital admission) were claimed not to be higher than in control groups.


Observational studies

A series of observational studies come from Sweden [3-5], where the mix of anti-TNF treatments was predominantly etanercept and infliximab. These used several sources of information, a prevalent cohort of 53,000 rheumatoid arthritis patients, an incidence cohort of 3,700 rheumatoid arthritis patients, and one rheumatoid arthritis cohort of 4,160 patients treated with TNF antagonists. These had approximately 300,000, 13,000, and 10,000 person years of follow up respectively, with average follow ups of 6, 3.5 and 2.5 years per patient.

Linkage between different registries and cancer and other registries meant that outcomes could be collected. The entire Swedish population served as a control group for calculation of standardised rates.


Malignancies

Haematopoietic malignancies [3] occurred in 507 patients, at rates higher than the general population (Table 1). When compared with the prevalent and incident rheumatoid arthritis cohorts not receiving TNF, patients receiving TNF antagonists had a relative risk of 1.1 (95% CI 0.6 to 2.1).

Solid cancers [4] occurred in 3,584 patients, at rates little greater than in the general population (Table 1). When compared with the prevalent and incident rheumatoid arthritis cohorts not receiving TNF, patients receiving TNF antagonists were little different (Table 1) overall, or when analysed by duration of observation.



Table 1: Cancer rates in observational study of infliximab and etanercept in Sweden



RA cohort
Haematopoietic malignancies
Patient years of observation
Standardised incidence rate
(95% CI)
Haematopoietic malignancy
Prevalent
481
297102
1.7 (1.5 to 1.8)
Incident
15
13292
1.6 (0.9 to 2.6)
anti-TNF
11
9715
2.1 (1.1 to 3.8)
Solid cancer
Prevalent
3379
297102
1.05 (1.01 to 1.08)
Incident
138
13292
1.1 (0.9 to 1.3)
anti-TNF
67
9715
0.9 (0.7 to 1.2)



Serious infections

The evidence here comes from another analysis of cohort studies in Sweden [5]. It was limited by the small number of tuberculosis cases in patients treated with infliximab or etanercept (15, 10 of them pulmonary). Only four of these were included in a statistical analysis. The best guess is that tuberculosis rates are increased in rheumatoid arthritis patients treated with TNF antagonists.


Comment

It is always tempting to dismiss observational studies when they disagree with randomised trials (or meta-analyses of them). This may be hasty, especially when both types of study are well done. It is always worth asking a few questions, especially when we are dealing with serious, but rare, adverse events.

The first question is about size, about the number of events. Because they are rare events, malignancies and serious infections are unlikely to be numerous. If a statistical analysis is based on a handful of events, potential interference from the random play of chance is possible. One meta-analysis, for instance, had three malignancies with placebo, and one observational study used four and two cases of tuberculosis to calculate statistics.

While on size, observational studies can involve a lot of patients, and here observational studies had up to four times as many patient years of observation than did a meta-analysis of randomised trials.

Which brings us to time, especially important for adverse events. Some may occur early with treatment, others late, or they can be constant over time. Comparing short with long duration studies can be problematical.

And finally, what treatments are being used? None of the observational studies seemed to include adalimumab, or any other newer agents. So while meta-analyses concentrated on some agents but not others, observational studies reported on a different mix of agents being used in clinical practice. The question, then, is whether we are comparing like with like.

If there is an lesson here, it is that for serious but rare adverse events, rushing to judgement may not be prudent. If there is an answer, it is that tinkering with the immune system can produce great benefit, but with a risk of something bad happening. Until we can say with certain who will benefit and who will be at risk, we have to live with that uncertainty.

Reference:

  1. T Bongartz et al. Anti-TNF antibody therapy in rheumatoid arthritis and the risk of serious infections and malignancies. JAMA 2006 295: 2275-2285.
  2. R Fleischmann et al. Long term safety of etanercept in elderly subjects with rheumatic diseases. Annals of the Rheumatic Diseases 2006 65: 379-384.
  3. J Askling et al. Haematopoietic malignancies in rheumatoid arthritis: lymphoma risk and characteristics after exposure to tumour necrosis factor antagonists. Annals of the Rheumatic Diseases 2005 64: 1414-1420.
  4. J Askling et al. Risks of solid cancers in patients with rheumatoid arthritis and after treatment with tumour necrosis factor antagonists. Annals of the Rheumatic Diseases 2005 64: 1421-1426.
  5. J Askling et al. Risk and case characteristics of tuberculosis in rheumatoid arthritis associated with tumour necrosis factor antagonists in Sweden. Arthritis & Rheumatism 2005 52: 1986-1992.

previous or next story