Objective To investigate the rate of serious bacterial, viral or opportunistic infection in patients with rheumatoid arthritis (RA) starting tocilizumab (TCZ) versus tumour necrosis factor inhibitors (TNFi) or abatacept.
Methods Using claims data from US Medicare from 2010 to 2015, and IMS and MarketScan from 2011 to 2015, we identified adults with RA who initiated TCZ or TNFi (primary comparator)/abatacept (secondary comparator) with prior use of ≥1 different biologic drug or tofacitinib. The primary outcome was hospitalised serious infection (SI), including bacterial, viral or opportunistic infection. To control for >70 confounders, TCZ initiators were propensity score (PS)-matched to TNFi or abatacept initiators. Database-specific HRs were combined by a meta-analysis.
Results The primary cohort included 16 074 TCZ PS-matched to 33 109 TNFi initiators. The risk of composite SI was not different between TCZ and TNFi initiators (combined HR 1.05, 95% CI 0.95 to 1.16). However, TCZ was associated with an increased risk of serious bacterial infection (HR 1.19, 95% CI 1.07 to 1.33), skin and soft tissue infections (HR 2.38, 95% CI 1.47 to 3.86), and diverticulitis (HR 2.34, 95% CI 1.64 to 3.34) versus TNFi. An increased risk of composite SI, serious bacterial infection, diverticulitis, pneumonia/upper respiratory tract infection and septicaemia/bacteraemia was observed in TCZ versus abatacept users.
Conclusions This large multidatabase cohort study found no difference in composite SI risk in patients with RA initiating TCZ versus TNFi after failing ≥1 biologic drug or tofacitinib. However, the risk of serious bacterial infection, skin and soft tissue infections, and diverticulitis was higher in TCZ versus TNFi initiators. The risk of composite SI was higher in TCZ initiators versus abatacept.
- rheumatoid arthritis
- bacterial infections
- serious infections
- tumor necrosis factor inhibitor
- biologic therapy
Statistics from Altmetric.com
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.
- rheumatoid arthritis
- bacterial infections
- serious infections
- tumor necrosis factor inhibitor
- biologic therapy
What is already known about this subject?
The evidence is limited on serious infection risk with non-tumour necrosis factor inhibitor (TNFi) biologics, such as tocilizumab (TCZ), in a head-to-head comparison with different biologic disease-modifying antirheumatic drugs (DMARDs).
In randomised controlled trials, the serious infection rates were higher in the TCZ monotherapy group than methotrexate, and in the 4 mg/kg and 8 mg/kg TCZ plus DMARD group compared with placebo plus DMARD group.
What does this study add?
This large multidatabase observational study found no difference in the risk for composite serious infection requiring hospitalisation in patients with rheumatoid arthritis (RA) who initiated TCZ versus TNFi after failing ≥1 other biologic drug or tofacitinib.
The risk of serious bacterial infection, skin and soft tissue infections, and diverticulitis was, however, higher in TCZ versus TNFi initiators.
The risk of composite serious infections was higher in TCZ initiators when compared with abatacept initiators.
How might this impact on clinical practice or future developments?
This head-to-head comparative safety information between TCZ and TNFi or abatacept may help make a better-informed treatment decision for RA in clinical practice.
Biologic disease-modifying antirheumatic drugs (bDMARDs) are effective in controlling rheumatoid arthritis (RA) activity and are usually considered when remission or low disease activity is not achieved with conventional synthetic DMARDs (sDMARDs). While high disease activity in RA is a risk factor for infections regardless of treatment,1 an important concern with bDMARD treatment in RA is the potential risk of serious infections (SIs) given their potent immunosuppressive effects.2–5 Since the biologic drugs act on diverse cellular and cytokine targets, the SI risk could be different among these biologics. Systematic reviews suggested the higher risk of bacterial infection, SIs and tuberculosis in patients treated with tumour necrosis factor inhibitor (TNFi) compared with conventional sDMARDs3 and non-TNF biologics, including abatacept and rituximab.5 However, the evidence is limited on SI risk with non-TNFi biologics, such as tocilizumab (TCZ), in a head-to-head comparison with different bDMARDs. In 24-week randomised controlled trials (RCTs), the SI rates were higher in the TCZ monotherapy group than methotrexate (3.6 vs 1.5/100 patient-years), and also in the 4 mg/kg and 8 mg/kg TCZ plus DMARD group compared with placebo plus DMARD group (4.4 and 5.3 vs 3.9/100 patient-years).6–9
Previous safety studies on TCZ did not include a direct comparison with a different biologic drug, and if they did, were small and did not evaluate the risk of specific SIs. Among the studies comparing TCZ with another biologic, a cohort study in Medicare population found no difference (HR 1.10, 95% CI 0.89 to 1.34) in the hospitalised infection for TCZ versus abatacept,10 and a prospective cohort study using Japanese RA registry observed no difference (HR 2.23, 95% CI 0.93 to 5.37) in the SI rate with TCZ versus anti-TNF use.11 However, a recent study using the British Society for Rheumatology Biologics Register for Rheumatoid Arthritis (BSRBR-RA) found an increased SI risk (HR 1.22, 95% CI 1.02 to 1.47) in TCZ compared with etanercept.12 Contrary findings from these studies highlight the need for further high-quality, large population-representative safety evidence with head-to-head comparison of TCZ with other biologics.
The objective of this study was to investigate the rate of serious bacterial, viral or opportunistic infection in patients with RA initiating TCZ compared with other biologics—TNFi or abatacept—in a multidatabase, population-based cohort. We also compared the risk of specific SIs in TCZ initiators with other biologics.
We used three large US healthcare claims databases—Medicare 2010–2015, IMS ‘PharMetrics’ Plus 2011–2015 and Truven ‘MarketScan’ 2011–2015—for this cohort study. Both PharMetrics and MarketScan are widely representative of a national commercially insured population in the USA13–15 and contain longitudinal medical and pharmacy claims from several different managed care plans. Medicare, a federally funded programme, provides healthcare coverage (part A for inpatient care, part B for outpatient medical services and part D for prescription drugs) for nearly all legal residents of the USA aged ≥65 and some disabled patients aged <65.16
Study cohort and design
We identified adult (≥18 years) patients with RA with one inpatient visit or ≥2 outpatient visits (7–365 days apart) using the International Classification of Diseases 9th or 10th Revision codes for RA. These patients were required to have continuous enrolment during the baseline period of 180 days before initiation of TCZ or a comparator drug (index date), which could be either TCZ or TNFi (primary cohort), or TCZ or abatacept (secondary cohort). All patients were required to have used ≥1 different biologic agent or tofacitinib any time prior, to minimise confounding by disease duration or severity between the TCZ and TNFi/abatacept groups. During the baseline 180-day period, TCZ initiators in both cohorts were naïve to TCZ, TNFi initiators in the primary cohort were naïve to a given specific TNFi and TCZ, and abatacept initiators in the secondary cohort were naïve to abatacept. A claims-based algorithm that combined diagnosis codes for RA with ≥1 DMARD dispensing record had a positive predictive value of ≥86% for identifying patients with RA.17
We excluded nursing home residents and those with pre-existing malignancy prior to and at the index date from both cohorts. We also excluded rituximab users as rituximab is often used in patients with history or diagnosis of malignancy and has a longer duration of action. Patients could enter the cohort only once when they first met the inclusion/exclusion criteria.
The primary outcome was a composite of any SI, including bacterial, viral or opportunistic infection, and the secondary outcomes were prespecified types of SIs, including serious bacterial infection, herpes zoster, opportunistic infection, pyelonephritis/urinary tract infection, skin and soft tissue infection, tuberculosis, viral hepatitis, diverticulitis, pneumonia/upper respiratory tract infection, and septicaemia/bacteraemia. All the outcomes were defined with validated claims-based algorithms18–20 using the principal discharge diagnosis (online supplementary table 2). Secondary analyses were conducted using any position of discharge diagnosis.
For the primary as-treated analysis, patients were followed from the day after the index date until the earliest of: treatment discontinuation +60 days, switching between exposure groups, outcome occurrence, disenrollment, death or study end period. For the secondary intention-to-treat up to 180 days (ITT 180-day) analysis, follow-up time ended on the earliest occurrence of the 180th day of follow-up, outcome occurrence, disenrollment, death or the end of study period.
During the 6-month baseline period, we measured ≥70 predefined variables potentially related to RA severity or duration, or development of infection in each database. We assessed the index year, demographics, comorbid conditions, combined comorbidity index,21 claims-based index of RA severity index,22 use of DMARDs (during all available data) and other prescription drugs including use of systemic steroids, non-steroidal anti-inflammatory drugs and analgesics, use of prophylactic antibiotics/antivirals, vaccination, and history of any invasive procedures or surgery. We also measured markers of healthcare utilisation intensity, including receipt of cancer screening tests and physician orders of outpatient laboratory tests for acute phase reactants.23
We compared baseline patient characteristics of TCZ and TNFi, as well as TCZ and abatacept initiators, before and after propensity score (PS)-matching to adjust for ≥70 baseline covariates within each database.24 Multivariable logistic regression models estimated patients’ PS values, that is, the predicted probability of TCZ versus a comparator drug conditioning on all the confounders. In the PS models, the index bDMARD was specified as the dependent variable and all confounders were entered as independent variables without further variable selection. For the primary cohort, we used PS-matching with a variable ratio up to 1:3 comparing TCZ and TNFi to increase the cohort size/study power. The matching ‘caliper’ was 0.02 on the PS scale.25 26
The number of events, person-time, incidence rate (IR) and IR differences for each outcome were reported, and HR with 95% CIs comparing treatment groups were estimated using Cox proportional hazard models conditioning on the PS-matching set to account for the variable ratio.27 28 The database-specific HRs were combined by an inverse variance-weighted, fixed-effects model.
We conducted prespecified subgroup analysis by gender, age, and baseline use of methotrexate, oral steroids, antibiotics and recent use of oral steroids. We also evaluated the association between TCZ use and SI versus a TNFi in the first 30, 90, 180 and 365 days of index therapy use to determine whether the rate of SI was constant over the treatment duration.
For the TCZ versus abatacept comparison, we used a 1:1 fixed-ratio PS-matching with calliper of 0.02. All the analytic steps were repeated for the secondary comparison.
We identified 141 869 patients with RA who started either TCZ or TNFi across the three databases. After applying the exclusion criteria (figure 1), our study cohort included a total of 62 563 patients with RA starting either TCZ (n=16 719) or TNFi (n=45 844). The variable ratio 1:3 PS-matching maintained 63.7% (n=39 829) of patients with RA in the final cohort: 13 102 TCZ and 26 727 TNFi initiators. The secondary comparison cohort comprised 10 414 PS-matched pairs of TCZ and abatacept initiators (online supplementary figure 1).
Even before PS-matching, patient characteristics were generally similar in TCZ and TNFi groups (online supplementary table 1). After variable ratio 1:3 PS-matching (table 1), all baseline characteristics were well balanced between the TCZ and TNFi groups with a standardised difference <10%.26 31 The mean age was 72 years in Medicare, 51 in IMS and 53 in MarketScan. At baseline, 69%–73% of patients used methotrexate and 70%–79% used corticosteroids. In the 1:1 PS-matched secondary comparison cohort, the baseline characteristics of TCZ and abatacept initiators were also well balanced (online supplementary table 2).
Risk of hospitalised SI
For the primary as-treated analysis, a total of 618 SIs occurred in the TCZ group (13 198 person-years) and 1155 in the TNFi group (28 950 person-years) across three databases during a mean (SD) follow-up of 0.9 (0.7) years and a maximum observational period of 5.9 years (table 2). Nearly one-third had ≥1 year follow-up and over 9% of all patients had ≥2 years’ follow-up. The overall IR of composite SI events was 4.68/100 person-years in TCZ and 3.99/100 person-years in TNFi users. The database-specific IRs ranged from 3.07/100 person-years (MarketScan) to 7.05 (Medicare) in TCZ initiators and from 2.53 (MarketScan) to 7.20 (Medicare) in TNFi initiators. For the primary comparison of TCZ versus TNFi, the composite SI risk was not different with a pooled HR of 1.05 (95% CI 0.95 to 1.16).
In the secondary cohort, 388 SI events (4.51/100 person-years) occurred in TCZ initiators over 8599 person-years and 295 in abatacept initiators (3.24/100 person-years) over 9094 person-years (table 3). Contrary to the results of primary analyses, the composite SI risk was higher with TCZ than abatacept use (pooled HR 1.40, 95% CI 1.20 to 1.63) (table 3).
The ITT 180-day analysis (tables 2 and 3) results were consistent with the primary findings for both primary (pooled HR 1.09, 95% CI 0.96 to 1.23) and secondary (pooled HR 1.34, 95% CI 1.11 to 1.63) comparison.
Risk of secondary outcomes
As summarised in table 4 and online supplementary table 3, among the secondary infection endpoints, 522 serious bacterial infection events were observed in TCZ and 865 in the TNFi group with IR/100 person-years of 3.95 and 2.97, respectively. The risk of serious bacterial infection was elevated (pooled HR 1.19, 95% CI 1.07 to 1.33) with TCZ compared with the TNFi. The risk for skin and soft tissue infections (pooled HR 2.38, 95% CI 1.47 to 3.86, IR/100 person-years 0.28 vs 0.12) and diverticulitis (pooled HR 2.34, 95% CI 1.64 to 3.34, IR/100 person-years 0.52 vs 0.21) was also higher in TCZ than TNFi users. The risk for herpes zoster, opportunistic infection, pneumonia/upper respiratory tract infection, pyelonephritis/urinary tract infection, tuberculosis, septicaemia/bacteraemia and viral hepatitis was not different in the TCZ and TNFi groups.
The risk of serious bacterial infection was more pronounced in TCZ than abatacept (pooled HR 1.50, 95% CI 1.27 to 1.78). Similarly, the greater risk was seen for diverticulitis (pooled HR 1.79, 95% CI 1.12 to 2.84), pneumonia/upper respiratory tract infection (pooled HR 1.37, 95% CI 1.04 to 1.80) and septicaemia/bacteraemia (pooled HR 1.39, 95% CI 1.08 to 1.78) with TCZ than with abatacept use.
Subgroup and sensitivity analyses
We found no difference in the risk of composite SI events between TCZ initiators and TNFi initiators in the subgroup analysis by age, sex, baseline methotrexate use and baseline antibiotic use (table 5). Similarly, the risk associated with TCZ versus TNFi was not different in patients with or without baseline and recent (60 days prior to index date) oral steroid use (table 5). The sensitivity analyses with maximum follow-up of 90, 180 and 365 days showed findings consistent with the primary analyses; however, higher risk of composite SIs was observed with a maximum follow-up of 30 days in TCZ versus TNFi, suggesting the short-term SI risk associated with TCZ use (online supplementary table 4). The results of the secondary analyses defining the outcomes based on any position of discharge diagnosis were consistent with the primary analyses (online supplementary table 5).
In this large US population-based multidatabase study, we found no difference in the composite SI risk between TCZ and TNFi initiators who previously used≥1 other bDMARDs or tofacitinib. However, the risk of serious bacterial infections, skin and soft tissue infection, and diverticulitis was higher in TCZ initiators compared with TNFi initiators, and was more pronounced when compared with abatacept initiators. Although the risk estimate for skin and soft tissue infections, and diverticulitis was higher with TCZ use, the rate difference per 100 patients were 0.16 and 0.31, respectively, and the absolute risks in both groups were small; thus, estimates can be imprecise in such analyses. Despite the elevated risk observed for the secondary infections, the null finding for the composite SI was possibly due to the additional SIs included in the primary endpoint definition that were not evaluated separately, and because more prevalent types of SIs were not different between TCZ and TNFi. The IRs/100 person-years observed in our study for composite SIs (4.68 in TCZ and 3.99 in TNFi) and bacterial infection (3.95 in TCZ and 2.97 in TNFi) were similar to previous studies.12 32–34 The serious bacterial infection IR/100 person-years of 5.63 in TCZ and 5.26 in the TNFi group in our Medicare population were similar to the IR of 4.89 from the previous Medicare study.34
The results from our study are in line with the previously identified safety profile of TCZ from RCTs and other real-world data. A prospective cohort study using Japanese RA registry showed no significant difference in the SI rate with TCZ versus anti-TNF use (HR 2.23, 95% CI 0.93 to 5.37).11 A recent observational cohort study using BSRBR-RA showed an increased risk (HR 1.22, 95% CI 1.02 to 1.47) in TCZ compared with etanercept for SI, defined as an infection resulting in death, hospitalisation or requiring intravenous antimicrobial therapy.12 Our primary pooled HR is within the 95% CIs from this BSRBR-RA study, but slightly different from their point estimate, which could be due to the differences in population, comparison group and outcome definition. Our results are comparable with a long-term postmarketing observational surveillance study in Japan,32 which found no increases in the proportions of patients with SI over the 3 years of TCZ treatment, and the findings from the German biologics register showing higher relative risk (RR) of infection among TCZ users (RR 1.15) and lower risk among abatacept users (RR 0.82) compared with anti-TNF therapy.35 Contrary to our secondary comparison group findings, a cohort study using 2006–2011 Medicare data has previously suggested no difference in the risk for hospitalised infection (adjusted HR 1.10, 95% CI 0.89 to 1.34) in TCZ versus abatacept users,10 which could be due to differences in the population and outcome definition (principal vs any diagnosis). The current study adds to the literature by providing risk estimates for serious bacterial infection, skin and soft tissue infections, and diverticulitis with TCZ compared with TNFi in three large cohorts of patients with RA from routine care. These findings are clinically important considering the higher risk of infections previously seen in a recent BSRBR report describing an elevated risk of SI associated with the use of adalimumab, etanercept and infliximab (adjusted HR 1.2, 95% CI 1.1 to 1.5).36
The main strength of this study is the generalisability and large size as we used three large US claims databases, which was necessary to achieve an adequate study size since TCZ is a relatively new therapy. Unlike enrollees in clinical trials, our study cohort is representative of older adults (Medicare) and the working population and their dependents (PharMetrics and MarketScan). In addition to TCZ versus TNFi comparison, we also compared TCZ with abatacept for better confounding control among these exposures and better understanding of the risk attributable to each agent. Our study provides the HRs and the absolute risk of SIs and several specific SIs. This study used rigorous pharmacoepidemiological approaches to provide findings with high validity, including the new-user design, specific outcome definition, use of active comparator and variable ratio PS-matching to perform appropriate statistical control for a large number of confounders and to minimise confounding by indication and immortal time bias.37 38 Because TCZ is mostly used as a second-line treatment for RA in routine practice, the requirement of prior biologic or targeted sDMARD use in both TCZ and TNFi/abatacept starters likely further reduced the degree of confounding by RA duration or severity between the two groups. Furthermore, given the known differences in patient demographics and other characteristics across the three databases, to optimise confounding control in each data source, we did all the analyses separately in each of the three databases and combined the results using a meta-analysis technique. The large size of the study allowed us to evaluate the risk of several specific SIs. Lastly, our sensitivity and subgroup analyses showed consistent results.
There are limitations to this study. While we adjusted for >70 baseline variables potentially related to SI risks using PS-matching, residual confounding is still possible. Information on confounders, such as RA duration, RA severity, DMARD treatment history, body mass index, smoking, alcohol intake or other behavioural risk factors, is also not captured in the claims data. It is possible that TCZ patients received higher number of biologics or other DMARDs than the TNFi group. Outcome misclassification is a possibility as we relied on the diagnosis codes for identifying SIs. However, our endpoints were identified by a principal inpatient code to minimise the possibility of outcome misclassification. Even with the large study cohort size, the number of secondary outcomes was relatively small. A possibility of surveillance bias may exist as rheumatologists may have decided to hospitalise patients with infections preferentially if they were treated with TCZ versus other biologics. The mean follow-up was short (<1 year), although most of these SIs are thought to occur in a few months after starting a treatment. Lastly, we did not examine the risk of recurrent infections.
In conclusion, this large multidatabase observational study found no difference in the risk for composite SI requiring hospitalisation in patients with RA who initiated TCZ versus TNFi after failing ≥1 other biologic drug or tofacitinib. However, we found an increased risk of serious bacterial infection, skin and soft tissue infections, and diverticulitis in TCZ versus TNFi initiators. In the secondary cohort comparing TCZ with abatacept, higher risks for composite SI, serious bacterial infection, diverticulitis, pneumonia/upper respiratory tract infection and septicaemia/bacteraemia were observed in the TCZ group. This head-to-head comparative safety information between TCZ and TNFi or abatacept may help better manage patients with RA in clinical practice.
Disclaimer : This is a summary of a scientific article written by a medical professional (“the Original Article”). The Summary is written to assist non medically trained readers to understand general points of the Original Article. It is supplied “as is” without any warranty. You should note that the Original Article (and Summary) may not be fully relevant nor accurate as medical science is constantly changing and errors can occur. It is therefore very important that readers not rely on the content in the Summary and consult their medical professionals for all aspects of their health care and only rely on the Summary if directed to do so by their medical professional. Please view our full Website Terms and Conditions.
Copyright © 2019 BMJ Publishing Group Ltd & European League Against Rheumatism. Medical professionals may print copies for their and their patients and students non commercial use. Other individuals may print a single copy for their personal, non commercial use. For other uses please contact our Rights and Licensing Team.
Handling editor Josef S Smolen
Contributors AP had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis. Study concept and design: SCK, RJD, AP. Analysis of data: AP. Interpretation of data and drafting of the manuscript: all authors.
Funding This study was funded by Roche but solely conducted at the Brigham and Women’s Hospital. The funder was given the opportunity to make non-binding comments on a draft of the manuscript, but the authors retained the right of publication and to determine the final wording.
Competing interests SCK has received research grants to the Brigham and Women’s Hospital from Roche, Pfizer and Bristol-Myers Squibb. DHS has received research grants to the Brigham and Women’s Hospital from Pfizer, Genentech, Amgen, AbbVie, Bristol-Myers Squibb and Corrona. AP has nothing to disclose. AJSO has nothing to disclose. RJD reports serving as the principal investigator on research grants from Merck and Vertex to the Brigham and Women’s Hospital for unrelated projects. SG, MB and KS are employed by Genentech. SS is consultant to WHISCON and to Aetion, a software manufacturer of which he also owns equity. SS is the principal investigator of research grants to the Brigham and Women’s Hospital from Boehringer Ingelheim.
Patient consent for publication Not required.
Ethics approval The Institutional Review Board of the Brigham and Women’s Hospital approved the study protocol and patient privacy precautions.
Provenance and peer review Not commissioned; externally peer reviewed.