|Home | About | Journals | Submit | Contact Us | Français|
Objectives. To evaluate the risk of serious infections (SIs) in patients with RA treated with anti-TNF therapy with emphasis on the risk across different ages.
Methods. Using data from the British Society for Rheumatology Biologics Register, a prospective observational study, we compared the risk of SI between 11798 anti-TNF-treated patients and 3598 non-biologic DMARD (nbDMARD)-treated patients.
Results. A total of 1808 patients had at least one SI (anti-TNF: 1512; nbDMARD: 296). Incidence rates were: anti-TNF 42/1000 patient-years of follow-up (95% CI 40, 44) and nbDMARD 32/1000 patient-years of follow-up (95% CI 28, 36). The adjusted hazard ratio (adjHR) for SI in the anti-TNF cohort was 1.2 (95% CI 1.1, 1.5). The risk did not differ significantly between the three agents adalimumab, etanercept and infliximab. The risk was highest during the first 6 months of therapy [adjHR 1.8 (95% CI 1.3, 2.6)]. Although increasing age was an independent risk factor for SI in both cohorts, there was no difference in relative risk of infection in patients on anti-TNF therapy in the older population. There was no difference in hospital stay for SI between cohorts. Mortality within 30 days of SI was 50% lower in the anti-TNF cohort [odds ratio 0.5 (95% CI 0.3, 0.8)].
Conclusions. These data add to currently available evidence suggesting that anti-TNF therapy is associated with a small but significant overall risk of SI. This must be balanced against the risks associated with poor disease control or alternative treatments.
RA has detrimental effects on a wide range of health outcomes, reaching far beyond the damage to the musculoskeletal system. RA is associated with increased mortality and comorbidity from a number of causes compared with the general population, including infection [1, 2]. Biologic therapies now bring the opportunity to target precise pathways within the immune system and inflammatory pathway leading to improved outcome for RA patients through better disease control and thus a reduction in the associated comorbidity and mortality. The anti-TNF drugs were the first biologic agents to become established in the management of RA patients who fail to respond to traditional non-biologic DMARDs (nbDMARDs). However, TNF plays an important role in host defence [3, 4]. Therefore, the introduction of anti-TNF agents was accompanied by a need to study and understand the effect on infection risk of modifying this key pathway.
This article summarizes the latest results from the British Society for Rheumatology Biologics Register (BSRBR) on rates of all serious infections (SIs), comparing nbDMARD with anti-TNF, as well as comparing between the three anti-TNF agents that were licensed in the UK between 2001 and 2009 [infliximab (INF), etanercept (ETN) and adalimumab (ADA)]. As follow-up has accrued, the BSRBR has acquired sufficient statistical power to be able to examine time-varying risk, age-specific risks and differences in outcome of infection.
The BSRBR is a large national prospective study, established to investigate the long-term safety of exposure to biologic agents in patients with RA. Full details of the BSRBR methodology have been published previously . In brief, the study commenced in 2001 alongside national recommendations that all RA patients prescribed anti-TNF therapy within the UK should be registered with the BSRBR. Patients were recruited to the ETN and INF cohorts from 2001 onwards, while recruitment to the ADA cohort started in 2003. Recruitment targets of 4000 patients per treated cohort were met for the ETN cohort in 2005, for INF in 2007 and for ADA in 2008. Before recruitment targets were met, we estimated that >80% of anti-TNF-treated patients with RA in the UK were registered on the BSRBR. A comparison cohort of biologic-naïve patients with active RA [defined as a 28-joint DAS (DAS-28) >4.2] was recruited in parallel. These patients had active disease despite current treatment with an nbDMARD. Patients who were prescribed biologics were recruited from across the whole of the UK (over 250 hospitals), whereas controls were recruited from just 29 centres across the UK. The 29 control centres are distributed across the UK and include a mixture of secondary and tertiary care rheumatology units in rural and urban settings in deprived and affluent areas. Ethics approval for this study was obtained in December 2000 from the Multicentre Research Ethics Committee (MREC) for the Northwest of England.
Baseline information included demographics, disease duration, HAQ score, DAS-28 score, steroid use, smoking history and comorbidity. Steroid use was defined as actively receiving oral steroids at the time of recruitment.
Information on adverse events was collected in three ways: 6 monthly questionnaires were sent to the treating rheumatology team for 3 years and annually thereafter; questionnaires were sent to the patients every 6 months (for 3 years); flagging with the UK National Health Service Information Centre (NHS-IC), which informed the register of any deaths and the cause of death.
Incident cases of SIs were identified from all three sources of follow-up. SIs were defined as those requiring i.v. antibiotics or hospitalization, or those resulting in death. This analysis was confined to cases that were reported or verified by the patient’s rheumatologist. Thus, patient-reported SIs were only included in the analysis if later verified by a consultant. Events were ascribed to the anti-TNF agent if they occurred while the patient was receiving anti-TNF therapy or within 90 days of the first missed dose. Events were attributed to the most recent drug in patients who switched anti-TNF therapy. Patients were censored from this analysis after their first episode of SI.
Analysis was restricted to patients with a physician diagnosis of RA. All patients had to have at least one returned consultant follow-up questionnaire before 31 December 2008 (the end of follow-up for this analysis). All patients were followed from the start of anti-TNF therapy (or registration for the DMARD cohort) until death, first SI or last follow-up. Patients within the anti-TNF cohort who stopped therapy for a reason other than SI contributed to follow-up time until 90 days after their first missed dose. If a patient then switched anti-TNF agent, they contributed subsequent follow-up to the new anti-TNF cohort.
Crude incidence rates were calculated as the number of episodes of SI per 1000 patient-years of follow-up with 95% CI. To estimate risk differences between the groups, a survival analysis was performed using a Cox proportional hazards model. Adjustment was made for age, gender and calendar year of recruitment. Multi-variable regression was performed with additional confounders identified from an a priori list including smoking, diabetes, chronic obstructive pulmonary disease (COPD), steroid use and disease severity (HAQ, DAS-28 and disease duration as continuous variables). MTX exposure was adjusted for as a time-varying covariate. Results are presented for both the whole follow-up period and limited to pre-specified time windows: 0–6, 6–12, 12–24 and 24–36 months of treatment. The cohort was then divided into four groups according to age at registration: <55, 55–64, 65–74 and ≥75 years. Stratified risk of infection within each age group was examined as above and the Wald test was used to look for evidence of a trend.
Finally, outcome following infection was assessed in two ways: (i) the length of hospital stay was compared between the two groups using the Mann–Whitney U-test and (ii) mortality within 30 days following diagnosis of SI was compared between the two groups using logistic regression (adjusted for age, gender, comorbidity, smoking, disease duration and severity, entry year and baseline steroid use). Missing baseline data were replaced using multiple imputations. All analyses were done using Stata 10.1 (StataCorp., College Station, TX, USA).
In total, 15396 patients were eligible for inclusion in this analysis: 11798 in the anti-TNF cohort and 3598 in the nbDMARD cohort. The baseline characteristics of the patients are shown in Table 1. In total, 3366 (22%) patients switched biologic during the follow-up period. Baseline characteristics relate to the first anti-TNF agent prescribed. The nbDMARD cohort was older and included a higher proportion of men. Thirty-six per cent of the nbDMARD cohort and 23% of the anti-TNF cohort were aged >65 years. Although disease activity was higher in the anti-TNF cohort, both cohorts had high mean levels of disease activity. Characteristics were similar across the three anti-TNF cohorts at baseline. The median duration of follow-up was 3.9 [interquartile range (IQR) 2.4, 4.9] years in the anti-TNF cohort and 2.6 (IQR 1.4, 3.8) years in the nbDMARD cohort.
Within both cohorts, comorbidity (diabetes and COPD) increases with age up to 75 years (supplementary table 1, available as supplementary data at Rheumatology Online). As one might expect, disease duration at study entry increases with age in both the nbDMARD and anti-TNF cohorts with the difference between the nbDMARD and anti-TNF groups being similar in each age band. Mean disease activity is similar across the age bands for both the nbDMARD and anti-TNF cohorts. Baseline mean HAQ score increases with age in the nbDMARD cohort but not in the anti-TNF cohort. Thus, there is a greater difference in the baseline mean HAQ score in the nbDMARD and anti-TNF cohorts in the youngest age band than in the oldest age band. Baseline steroid exposure rose dramatically with increasing age in both cohorts, with 18% of the nbDMARD cohort and 40% of the anti-TNF cohort aged <55 years receiving steroid at baseline, compared with 36% of the nbDMARD cohort and 55% of the anti-TNF cohort aged >75 years.
In total, 1808 patients experienced at least one SI (nbDMARD: 296; anti-TNF: 1512; Table 2). Approximately one-third of patients who suffered one SI were reported as having a second SI during the subsequent follow-up. The proportion did not differ significantly between the two cohorts (32% DMARD, 28% TNF, P=0.123). All further analysis considers only the first-reported SI per subject.
The unadjusted rates of SI were higher in the anti-TNF cohort (42 vs 32 events per 1000 patient-years of follow-up). Univariate analysis of the a priori list of potential confounders identified age, male gender, DAS-28, disease duration, diabetes, COPD, baseline steroids and smoking all as significant predictors of infection (supplementary table 2, available as supplementary data at Rheumatology Online). The adjusted rate of SI was 20% higher in the anti-TNF cohort than in the nbDMARD cohort [adjusted hazard ratio (adjHR) 1.2 (95% CI 1.1, 1.5)]. The highest crude SI rate was seen with INF [46/1000 (95% CI 42, 50)], followed by ADA [43/1000 (95% CI 39, 47)] and ETN [38/1000 (95% CI 35, 42)]. However, in the adjusted analysis, there was no significant difference in SI rates between the three anti-TNF agents. The analysis was also performed excluding patients who switched biologic (censoring them at the time of first switch): adjHR 1.2 (95% CI 1.0, 1.4).
The adjHR of SI in the anti-TNF vs the DMARD cohort was the highest in the first 6 months of therapy [adjHR in all anti-TNF cohort 1.8 (95% CI 1.3, 2.6)]. The risk then decreased over time with the lowest risk observed between 24 and 36 months [adjHR in all anti-TNF cohort 0.9 (95% CI 0.6, 1.3)]. A similar pattern was observed when analysis was stratified by anti-TNF therapy.
The crude rate of infection increased markedly with increasing age in both cohorts (Table 3). However, the adjHR was similar across the age bands (Table 3), with no significant trend (P=0.210). An alternative analysis dividing the cohort into those aged under or over 65 years found similar adjHR in both age groups over the entire period of follow-up [adjHR 1.3 (95% CI 1.1, 1.6) for those <65 years and adjHR 1.1 (95% CI 0.8, 1.4) for those >65 years] as well as in an analysis limited to the first 6 months of therapy [adjHR 1.9 (95% CI 1.1, 3.0) for those <65 years and adjHR 1.6 (95% CI 0.9, 2.8) for those >65 years].
The duration of hospital stay did not differ between anti-TNF [median stay 6 days (IQR 3–12)] and nbDMARD [median stay 7 days (IQR 3–14)] treated patients (Table 4). However, there was a much lower 30-day mortality rate among patients in the anti-TNF cohort [anti-TNF 7%; DMARD 16%; P<0.001; adjusted odds ratio 0.5 (95% CI 0.3, 0.8)].
RA patients have been recognized to be at increased risk of infection for several decades . Some of the increased risks is attributable to the disease process itself and some to the immunosuppressive properties of its treatment, in particular CSs. TNF plays an important role in the control of infection. In particular, TNF release from macrophages appears crucial in the maintenance and formation of granulomata, as well as playing a critical role in the defence against invasion by intracellular organisms. TNF also has roles in leucocyte trafficking and IC clearance . TNF inhibition is a risk factor for a variety of infections in animal models [8, 9]. However, anti-TNF therapies may also have potential beneficial effects on the immune system in disease states such as RA by reducing the immune abnormalities intrinsic to the disease.
Several approaches have been used to try and quantify the risks of infection associated with anti-TNF therapy in RA. Meta-analyses of randomized controlled trials (RCTs) have produced conflicting results [10, 11]. However, although the RCT is the gold standard for evaluating drug efficacy, RCTs often lack power (in terms of both numbers and duration of follow-up) to evaluate specific adverse events. In addition, patients enrolled in RCTs are a very select group of patients who do not wholly reflect the RA cohorts seen in routine clinical practice . These shortcomings can be addressed by large-scale prospective observational studies, although these lack the advantages of randomization and so are subject to confounding by indication.
The observational design of the BSRBR means patients are not randomized to their respective treatment, and therefore there may be differences in clinical characteristics. This needs to be considered when comparing rates of SI between the two cohorts as any observed difference may reflect the characteristics of the patients selected for treatment rather than an effect of the treatment itself. The detailed collection of baseline data within the BSRBR allows for adjustment of these patient characteristics that may potentially confound any association between anti-TNF and SI. However, despite this there remains the possibility of residual unmeasured confounding.
CS exposure deserves particular attention in this analysis. Although adjustment has been made for baseline CS exposure, it has not been possible to adjust for CSs as a time-varying covariate. This is because the BSRBR does not have information on whether patients were receiving steroids immediately before any infection. Patients who commence and respond to anti-TNF are likely to require less steroids during their follow-up period (and the converse may be true with the nbDMARD cohort). If this is the case, our adjusted model may be underestimating the effect of anti-TNF.
The BSRBR has previously reported a 20% non-significant increase in the rate of SI between TNF-treated individuals and controls . In addition, we have reported that intracellular bacterial species (e.g. Listeria, Salmonella) occurred more frequently in the anti-TNF cohort and specifically Mycobacterium tuberculosis infections were substantially increased [13, 14]. Despite longer follow-up, this updated analysis also shows a 20% increase in risk, which is now statistically significant. A small but significant increase in the risk of SI has also been reported by the German and Swedish Biologics Registries [15, 16].
However, presenting a single estimate for the risk of SI is misleading as the risk is not constant over time. In the first 6 months of therapy, the risk of SI was 80% higher in the anti-TNF-treated cohort than in the non-biologic-treated controls. Askling and Dixon  reviewed all published papers on infection risk associated with anti-TNF therapy in RA and found that the risk was highest in the first few months and then declined. This variation in risk is probably explained by a combination of factors. First, there will be a number of patients who are at higher baseline risk. When these individuals develop an SI, they will stop their drug and no longer contribute to the anti-TNF cohort. This results in a depletion of susceptible individuals (a healthy user effect), and so reduces the apparent risk of the drug. In addition to this, there may be a true time-dependent change in the drug safety profile. A persistent blockade of one cytokine pathway may lead to up-regulation of other immune-signalling pathways that can compensate for the lack of TNF. Also as patients become established on anti-TNF therapy, their RA becomes better controlled, their dose of steroids can be reduced, they become more mobile and disease-driven alterations in natural immunity reduce. This may in part explain why one meta-analysis of RCT data showed an increased risk , given that RCTs predominantly focus on the early period of exposure.
We also considered the question of differential risk between the three anti-TNF agents. The German registry reported a lower risk of Herpes varicella zoster skin infections in patients treated with ETN than in those treated with the mAb therapies (INF and ADA). We reported a similar pattern with M. tuberculosis infections [18, 19]. However, we found no significant difference in risk between the three agents for overall SIs, either in the first 6 months or overall.
We next examined the risk across different age groups. Managing RA in older patients is challenging, as it is complicated by both comorbidity and polypharmacy. Traditionally, clinical practice has been to assume a higher risk with nbDMARDs in the elderly and to prefer to use low doses of oral steroids . This pattern was seen in the BSRBR data set, with higher baseline steroid exposure among those patients aged ≥75 years. Steroids, even in low doses, are an important risk factor for SI [21–25]. Despite the high background risk of infection in the elderly patient with RA, our data do not support the notion that anti-TNF therapy increases this risk any further in the elderly patient than in the younger patient. This finding is consistent with post hoc analyses from clinical trials of ETN vs placebo, which have compared rates of infection in those under or over 65 years of age [26, 27]. Several observational studies have also addressed this issue. Schneeweiss et al.  did not find an increased risk of infection in anti-TNF users compared with those receiving MTX among US Medicare beneficiaries. Genevay et al.  also reported no differences in rates of discontinuation or SI in a Swiss cohort of 1571 RA patients divided into younger (<65 years) or older (>65 years) age groups. Our results add to this literature by presenting data stratified into even older age groups. However, it is important to note that although the relative risk of infection was similar in all age groups, given the higher background risk in those patients aged >65 years, there is higher absolute risk with anti-TNF therapy in this age group. To put this into context, during the first 6 months of treatment, in those aged <65 years, 25 (95% CI 20, 31) RA patients would need to be treated to observe one additional SI, while in those aged >65 years the equivalent number is 19 (95% CI 16, 23).
It is possible that physicians may have a lower threshold for hospital admission for an RA patient with infection if the patient is receiving anti-TNF than if they are not. Thus, although the patients in the anti-TNF cohort had more infections that satisfied the definition of SI, these infections may not, in fact, have been more severe. To explore this further, we compared the length of hospital stay and 30-day mortality risk between our two cohorts. The similar length of hospital stay suggests that the infections in anti-TNF-treated patients were not more severe than those in nbDMARD-treated patients. The 50% lower 30-day mortality rate in the anti-TNF cohort is intriguing. In animal models of sepsis, treatment with anti-TNF therapy before the onset of sepsis was found to be beneficial, presumably by suppressing the inflammatory response . Human trials did not show any significant advantage of adding TNF inhibitors to a sepsis regimen [31, 32], although the human trials could not initiate anti-TNF therapy until after sepsis was established. In the context of the BSRBR, patients had effectively been pre-treated with anti-TNF therapy before developing any infection. It is important to stress that the design of this study precludes any definite proof of causality in this association. However, the size and strength of this association would certainly support the need for additional research in this area.
These data add to the currently available evidence suggesting that anti-TNF therapy is associated with a small but significant overall risk of SI. The risk is highest in the first 6 months of therapy and then falls. This should be explained to the individual patient. Increasing age is an important risk factor for infection in patients with RA. Some of this increased risk may be related to steroid usage. To date, there is no convincing evidence that the relative risk of infection with anti-TNF therapy also increases with age. Further research is needed to help clinicians balance the risk of treatment options in elderly subjects. Finally, there is a significantly lower 30-day mortality rate following SI in RA patients treated with anti-TNF agents.
Supplementary data are available at Rheumatology Online.
The authors acknowledge the enthusiastic collaboration of all consultant rheumatologists and their specialist nurses in the UK in providing the data. In addition, we acknowledge the support from Dr Ian Griffiths (Past) and Prof David Isenberg (Current), Chairs of the BSRBR Management Committee, Prof Gabriel Panayi, Prof David G. I. Scott, Dr Andrew Bamji and Dr Deborah Bax, presidents of the BSR during the period of data collection, for their active role in enabling the Register to undertake its tasks and to Samantha Peters (CEO of the BSR), Mervyn Hogg and members of the BSRBR Scientific Steering Committee. We also acknowledge the seminal role of the BSR Clinical Affairs Committee for establishing national biologic guidelines and recommendations for such a register. Finally, we would like to acknowledge the substantial contribution of Andy Tracey, Katie McGrother, Xuejuan Fan and Dr Mark Lunt in database design and manipulation and Prof Alan Silman in his prior role as a principal investigator of the BSRBR. The BSR receives restricted financial support from Abbott Laboratories, Amgen, Biovitrum, Schering Plough, Wyeth Pharmaceuticals and Roche. This income finances a separate contract between the BSR and the University of Manchester who provide and run the BSRBR. All decisions concerning analysis, interpretation and publications are made autonomously of any industrial contribution. The British Society for Rheumatology (BSR) commissioned the Biologics Register (BSRBR) as a UK-wide national project to investigate the safety of biologic agents in routine medical practice. D.P.M.S. and K.L.H. are principal investigators on the BSRBR. The BSR receives restricted income from UK pharmaceutical companies, presently Abbott Laboratories, Amgen, Schering Plough and Wyeth Pharmaceuticals. 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. Members of the Manchester team, BSR trustees, committee members and staff complete an annual declaration in relation to conflicts of interest.
Funding: Funding to pay the Open Access publication charges for this article was provided by the BSRBR.
Disclosure statement: The authors have declared no conflicts of interest.
The BSRBR Control Centre Consortium consists of the following institutions (all in the UK): Antrim Area Hospital, Antrim (Dr Nicola Maiden), Cannock Chase Hospital, Cannock Chase (Dr Tom Price), Christchurch Hospital, Christchurch (Dr Neil Hopkinson), Derbyshire Royal Infirmary, Derby (Dr Sheila O’Reilly), Dewsbury and District Hospital, Dewsbury (Dr Lesley Hordon), Freeman Hospital, Newcastle-upon-Tyne (Dr Ian Griffiths), Gartnavel General Hospital, Glasgow (Dr Duncan Porter), Glasgow Royal Infirmary, Glasgow (Prof Hilary Capell), Haywood Hospital, Stoke-on-Trent (Dr Andy Hassell), Hope Hospital, Salford (Dr Romela Benitha), King’s College Hospital, London (Dr Ernest Choy), Kings Mill Centre, Sutton-in-Ashfield (Dr David Walsh), Leeds General Infirmary, Leeds (Prof Paul Emery), Macclesfield District General Hospital, Macclesfield (Dr Susan Knight), Manchester Royal Infirmary, Manchester (Dr Ian Bruce), Musgrave Park Hospital, Belfast (Dr Allister Taggart), Norfolk and Norwich University Hospital, Norwich (Prof David Scott), Poole General Hospital, Poole (Dr Paul Thompson), Queen Alexandra Hospital, Portsmouth (Dr Fiona McCrae), Royal Glamorgan Hospital, Glamorgan (Dr Rhian Goodfellow), Russells Hall Hospital, Dudley (Prof George Kitas), Selly Oak Hospital, Selly Oak (Dr Ronald Jubb), St Helens Hospital, St Helens (Dr Rikki Abernethy), Weston General Hospital, Weston-super-Mare (Dr Shane Clarke/Dr Sandra Green), Withington Hospital, Manchester (Dr Paul Sanders), Withybush General Hospital, Haverfordwest (Dr Amanda Coulson), North Manchester General Hospital (Dr Bev Harrison), Royal Lancaster Infirmary (Dr Marwan Bukhari) and the Royal Oldham Hospital (Dr Peter Klimiuk).