Objectives To evaluate the risk of gastrointestinal perforation (GIP) in subjects with rheumatoid arthritis (RA) treated with antitumour necrosis factor (anti-TNF) therapy compared with non-biological disease-modifying antirheumatic drugs (nbDMARDs).
Methods Using data from the British Society for Rheumatology Biologics Register, we compared the incidence of GIPs between 11 881 anti-TNF-treated and 3393 nbDMARD-treated RA patients using Cox regression modelling. Hazard ratios (HRs) with confidence intervals (CI) were calculated. Adjustment was made for potential confounders including current steroid use. The study covered the time period between 2001 and 2011.
Results There were 42 (upper 20, lower 22) GI perforations: five in the nbDMARD cohort and 37 in the anti-TNF cohort. After adjustment, treatment with TNF antagonists was associated with an HR of 1.6 (95% CI 0.4 to 6.0) for all GIPs, 2.7 (95% CI 0.4 to 18.1) for lower GIPs and 0.9 (95% CI 0.1 to 5.8) for upper GIPs. Current use of steroids was the single most important predictor of GI perforation with an adjusted HR of 2.9 (95% CI 1.5 to 5.4), but this risk was confined to lower GIPs (HR 8.0, 95% CI 2.6 to 24.1).
Conclusions We have not found a statistically significant association between anti-TNF treatment and the risk of GIP.
- Rheumatoid Arthritis
- DMARDs (synthetic)
Statistics from Altmetric.com
Gastrointestinal perforation (GIP) is a serious and potentially life-threatening condition occasionally observed in patients with rheumatoid arthritis (RA). These patients may be at an increased risk of GIP because of prescribed medication, and/or because of the consequences of the disease process itself (eg, vasculitis).1 GIPs have been reported mainly in users of non-steroidal anti-inflammatory drugs (NSAIDs),2 high-dose glucocorticoids3–6 and IL-6 antagonist tocilizumab.7–9 Anecdotal cases of GIP have been also reported in RA patients treated by antitumour necrosis factor (anti-TNF) agents,10–16 but the possible risk of GIPs conferred by anti-TNF treatment has not been, to our knowledge, rigorously quantified. Our primary aim was to test the hypothesis that in RA patients, anti-TNF therapy increases the risk of GIP compared with non-biological disease-modifying antirheumatic drugs (nbDMARDs).
The British Society for Rheumatology Biologics Register-Rheumatoid Arthritis (BSRBR-RA) is a national prospective cohort study that was established in 2001. Patients with active RA who were starting treatment with anti-TNF therapy were enrolled for observational follow-up. A comparison cohort of RA patients with active disease currently receiving nbDMARDs was recruited and followed up in parallel. Baseline information was collected regarding demographics, disease severity, concomitant medication and comorbidity. There were three sources of data collection during follow-up: consultant questionnaires, patient questionnaires and diaries, and the UK national cancer and death register (National Health Service Information Centre). Adverse events from these sources were coded using Medical Dictionary for Regulatory Activities (MedDRA) terms.
Case definition and verification
GIP was defined as a complete non-traumatic penetration of the wall of the oesophagus, stomach, or small or large intestine. We used all relevant MedDRA codes (see online supplementary table S1) to identify GIPs. All events were then verified by a physician (JZ) from the source documentation. According to their anatomical location, the events were divided into upper and lower GIPs. Events were attributed to the anti-TNF agent if they occurred while the patient was receiving anti-TNF therapy or within 90 days of the stop date (defined as 7 days after the last dose of etanercept (ETN), 14 days after adalimumab (ADA) and 60 days after the last infusion of infliximab (INF)). Events were attributed to the most recent drug in patients who switched anti-TNF therapy. If the patient died within 30 days after the event, any death was considered to be associated with GIP.
Analysis was restricted to patients with a physician diagnosis of RA who had at least one returned consultant follow-up questionnaire prior to 31 March 2011 (the end of follow-up for this analysis). All patients were followed from start of an anti-TNF therapy (or registration for the nbDMARD cohort) until death, first episode of GIP or last follow-up, whichever came first. Patients within the anti-TNF cohort who stopped therapy for a reason other than GIP contributed follow-up time until 90 days after their stop date. If a patient then switched to another anti-TNF agent, they contributed subsequent follow-up time to the second anti-TNF agent. Crude incidence rates were calculated as the number of episodes of GIP per 100 000 exposed patient years (pyrs) of follow-up with 95% CIs. Survival analyses, performed using a Cox proportional hazards model, were used to compare the rates of GIPs between cohorts. Stratification by deciles of propensity score for treatment by TNF antagonists was used to adjust for confounding between the groups. Current steroid use was entered into the Cox regression model as a time-varying covariate. The baseline covariates balanced by the propensity score model were identified from an a priori list including age, gender, smoking, history of peptic ulcer disease, RA disease severity, use of steroids, non-selective NSAIDs, COXII inhibitors, aspirin and gastroprotective drugs. Because of significant associations with the outcome of interest found in univariate analyses, we have also included weight, history of hypertension and the number of previous nbDMARDs in the propensity score model. For all analyses, missing baseline data were replaced using multiple imputations. All analyses were performed using Stata V.10 software (StataCorp, College Station, Texas, USA). For further details of the methodology see online supplementary text.
This analysis included 15 274 patients (3393 nbDMARD, 11 881anti-TNF cohort).
The anti-TNF cohort comprised 4129 patients starting ETN, 3483 patients starting INF and 4269 patients starting ADA. The anti-TNF cohort was younger, had more severe disease of a longer duration and had greater exposure to steroids at baseline (table 1). There were 42 (upper 20, lower 22) GIPs: five in the nbDMARD cohort and 37 in the anti-TNF cohort (table 2). In all, 12 patients (29% of GIPs) died within 30 days following GIP. Overall, GIPs contributed to 2.0% of all deaths recorded in the cohort during the follow-up time.
The crude incidence rate was 38 per 100 000 pyrs (95% CI 12 to 89) in the nbDMARD cohort and 72 per 100 000 pyrs (95% CI 50 to 99) in the anti-TNF cohort. The unadjusted HR for anti-TNF therapy was 2.0 (95% CI 0.8 to 5.0) and fully adjusted HR was 1.6 (95% CI 0.4 to 6.0) (table 3). We conducted sensitivity analyses to calculate HRs in those on TNF drugs excluding the 90-day lag window, which did not alter the results (see online supplementary table S2). Incident rates of upper and lower GIPs and their associated unadjusted and fully adjusted HR are shown in table 3. In univariate analyses, higher age, use of steroids at baseline, higher Health Assessment Questionnaire scores, presence of renal disease, history of hypertension and smoking were significantly associated with the risk of lower GIP, while use of non-selective NSAIDs at baseline and low weight were significantly associated with the risk of upper GIP (see online supplementary table S3). Current use of steroids was the single most important predictor of GIP (when analysed as a time-varying variable together with strata of propensity score) with an overall HR of 2.9 (95% CI 1.5 to 5.4), but this increase of risk was confined to lower GIPs (HR 7.9, 95% CI 2.6 to 24.1), rather than upper GIPs (HR 1.2, 95% CI 0.5 to 3.0).
The data from this BSRBR cohort show that treatment with ETN, INF or ADA was not associated with a statistically significant change in the risk of all GIPs. However, because of the low number of events, our study had 85% power to detect a 3.5-fold increase of risk of all GIPs in the anti-TNF cohort. Therefore, differences of a smaller magnitude cannot be excluded. However, hazard estimates for anti-TNF in the range observed in this study would translate into only a small increase of absolute risk, since GIP is a rare event, with a crude incidence rate of less than 1 per 1000 pyrs observed in the nbDMARD cohort. The outcome of patients who had GIP was serious: 29% of these patients died within 30 days after the event and GIPs contributed to 2% of all deaths observed in the cohort during the follow-up time. The hazard associated with anti-TNF treatment seemed to be increased for lower but not upper GI perforations (HR 2.7 vs 0.9, respectively).
We have confirmed the findings of previous studies, showing that current use of glucocorticoids confers significant risk for lower GIPs.3–6 ,17 ,18 Curtis et al recently studied GIPs in RA patients within two large administrative US databases.17 ,18 They found antecedent diverticulitis to be a risk factor for GIP. BSRBR does not collect baseline information on previous diverticulitis, and so we could not explore this factor. Our analysis excluded events which occurred in patients after stopping anti-TNF therapy or after starting a subsequent non-anti-TNF biologic. These patients are likely to represent the most severe patients and possibly those most at risk of GIP, which may partially explain lower observed incident rates for all GIPs (and lower GIPs), than reported by Curtis et al. Administrative databases have the advantage of covering large numbers of patients, but are designed for other purposes and capture of events of interest and important confounders may be incomplete or often impossible. The main strengths of the BSRBR data have been detailed previously19 and relate to the prospective nature of the study, and the detailed data collection during follow-up.
Our study also has several limitations to be noted. First, the BSRBR does not collect longitudinal data on NSAID use, and thus we could adjust only for NSAID medication at baseline. Second, despite the fact that BSRBR is one of the largest prospective longitudinal observational drug studies in rheumatology, the low incidence of GIPs especially in the control group precluded any definitive answers concerning the risk associated with the intervention. As GIPs are rare events of heterogeneous nature with many possible confounding factors, the question concerning the risk associated with anti-TNF drugs probably cannot be solved within any single prospective longitudinal cohort study, but may call for pooled analysis of existing national biologics registries.
In conclusion, anti-TNF therapy was not associated with a statistically significant risk of all GIPs in this large cohort of RA patients. Current use of steroids was the single most important predictor of GIP, but this increase of risk was confined to lower GIPs.
MedDRA is a registered trademark of the International Federation of Pharmaceutical Manufacturers and Associations (IFPMA). The authors acknowledge the enthusiastic collaboration of all consultant rheumatologists and their specialist nurses in the UK in providing the data. In addition, the authors acknowledge the support from the BSR Executive, the members of the BSR Registers and Research Committee and the BSRBR Project Team in London for their active role in enabling the register to undertake its tasks. The authors also acknowledge the seminal role of the BSR Clinical Affairs Committee for establishing national biological guidelines and recommendations for such a register.
Handling editor Tore K Kvien
Contributors JZ performed the analysis and wrote the main draft of the article. DS and KH supervised the conduct of the analysis. RD, JG, AL and LM assisted with the analytic design. ML supervised the use of statistical methods. KW and AL assisted with identification of cases. All authors contributed significantly to the analysis and/or to the final draft.
Funding The BSR commissioned the BSRBR as a UK-wide national project to investigate the safety of biologic agents in routine medical practice. BSR receives restricted income from UK pharmaceutical companies, presently Abbott Laboratories, Amgen, Schering Plough (now MSD) and Wyeth Pharmaceuticals (now Pfizer). This income finances a wholly separate contract between the BSR and the University of Manchester who provide and run the BSRBR data collection, management and analysis services. The principal investigators and their team have full academic freedom and are able to work independently of pharmaceutical industry influence. All decisions concerning analyses, interpretation and publication are made autonomously of any industrial contribution.
Competing interests DPMS and KLH are principal investigators on the BSRBR. BSR receives restricted income from UK pharmaceutical companies, presently Abbott Laboratories, Biovitrum, Merck Sharp & Dohme, Pfizer and Roche. This income finances a wholly separate contract between the BSR and the University of Manchester. The principal investigators and their team have full academic freedom and are able to work independently of pharmaceutical industry influence. All decisions concerning analyses, interpretation and publication are made autonomously of any industrial contribution. Members of the Manchester team, BSR trustees, committee members and staff complete an annual declaration in relation to conflicts of interest. The authors declare no other conflicts of interest.
Ethics approval This study was obtained in December 2000 from the Multicentre Research Ethics Committee (MREC) for the Northwest of England.
Provenance and peer review Not commissioned; externally peer reviewed.
If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.