Pharmacoepidemiological studies of rheumatoid arthritis : methodological considerations and applications

Abstract

Study I: Biological disease-modifying anti-rheumatic drugs (DMARDs) and the risk of gastrointestinal perforations: Study I was motivated by previous signals of an increased risk of lower gastro-intestinal perforations among rheumatoid arthritis (RA) patients treated with tocilizumab. The primary aim of study I was to compare the incidence of lower gastro-intestinal perforations between RA patients initiating the biological DMARDs tocilizumab, abatacept, rituximab and tumor necrosis factor inhibitors (TNFi). Secondary comparisons with bionaïve RA patients and general population controls were made. We designed a cohort study which included RA patients identified in the Swedish National Patients Register (NPR) and we identified biological DMARD treatments in the Swedish Rheumatology Quality Register (SRQ). General population controls, matched by sex, age and location to biological DMARD treated RA patients, were available. The main outcome was hospitalization or death due to lower gastro-intestinal perforations identified using a prespecified list of International Classification of Disease, 10th revision (ICD-10) codes in the NPR and the Swedish Causes of Death Register. In line with previous studies, we observed an increased risk of lower gastro-intestinal perforations among patients treated with tocilizumab compared to patients treated with TNFi (hazard ratio of 2.2, 95% confidence interval 1.3 to 3.8), and there was no evidence of an increased risk among patients treated with abatacept or rituximab. Also, compared to general population controls, only RA patients treated with tocilizumab had an increased risk of lower gastro-intestinal perforations after adjustment for sex, age and baseline glucocorticoids use. The absolute risk of lower gastro-intestinal perforations was low even under treatment with tocilizumab (~4 events /1000 person-years), but considering the potential for serious complications, the presence of additional risk factors, such as older age and use of glucocorticoids, should be evaluated before deciding to initiate tocilizumab, and patients should be monitored for diverticulitis and lower gastro-intestinal perforations during treatment.

Study II: Comparative effectiveness of baricitinib, tofacitinib and biological DMARDs in RA: Study II was motivated by a lack of evidence for the relative effectiveness of the Janus Kinase inhibitor (JAKi) baricitinib compared to non-TNFi biological DMARDs. The aim of study II was to compare the effectiveness of the JAKis baricitinib and tofacitinib with that of biological DMARDs. Study II was a cohort study which included RA patients who initiated baricitinib, tofacitinib, abatacept, interleukin-6 inhibitors (IL-6i), rituximab, and TNFi, as identified in SRQ. In the primary analysis, these patients were followed for one year from treatment initiation, at the end of which the proportions of treatment responders were evaluated using to the following measures: EULAR disease activity score assessed on 28 joints (DAS28) good response, health assessment questionnaire disability index (HAQ-DI) improvement > 0.2 units compared to baseline, and clinical disease activity index (CDAI) remission. Patients who discontinued treatment before one year were classified as “non-responders”. Improvements at three-months compared to baseline in DAS28, HAQ-DI, and CDAI, as well as drug retention over followup were also assessed. After confounding adjustment, one-year treatment response proportions were consistently higher on baricitinib compared to TNFi, even though the differences were small. Comparisons with non-TNFi biological DMARDs also favored baricitinib, but not consistently. There was no evidence that response proportions on tofacitinib were different from those on baricitinib or biological DMARDs, but the sample of tofacitinib treated patients was small, limiting precision. Drug retention was significantly higher on baricitinib compared to alternatives, and the magnitude of three-months improvements followed a similar pattern to one-year treatment responses. In conclusion, our results show that baricitinib and tofacitinib are at least as effective as biological DMARDs for treating RA.

Study III: Emulation of the SWEFOT trial in observational data: Study III was motivated by a lack of confidence in the comparative effectiveness evidence generated by observational studies, which could prove a valuable complement to randomized controlled trial (RCT) comparative efficacy evidence. It has been suggested that designing observational studies to mimic RCTs may reduce bias. A sensible approach for testing the validity of trial emulations in observational data is to emulate an existing trial protocol, and then compare the results. Therefore, the aim of study III was to emulate the protocol of the Swedish Farmacotherapy (SWEFOT) pragmatic trial in an observational study including a non-overlapping sample of participants coming from the same source population (Swedish RA patients) and to compare the results. In SWEFOT, methotrexate (MTX) insufficient responders were randomized to receive additional infliximab or sulfasalazine (SSZ) + hydroxychloroquine (HCQ). In the observational study, we included RA patients initiating infliximab (N = 313) or SSZ + HCQ (N = 196) after MTX, identified using data from SRQ and the Prescribed Drugs Register, and mimicking the SWEFOT eligibility criteria. The primary outcome was the proportion of EULAR DAS28 good responders at 9 months, classifying patients who discontinued treatment as non-responders. The proportions of responders in the observational emulation were comparable to those in SWEFOT: 39% (vs. 39% in SWEFOT) for infliximab and 28% (vs. 25%) for SSZ + HCQ. The crude observed response ratio was 1.39 (95% confidence interval: 1.04 to 1.86), increasing to 1.48 (95% confidence interval 0.98 to 2.24) after confounding adjustment, compared to 1.59 (95% confidence interval 1.10 to 2.30) in SWEFOT. Thus, by designing our observational study to emulate SWEFOT, we could closely replicate the trial results, favoring infliximab over SSZ + HCQ combination therapy at 9 months.

Study IV: Glucocorticoid exposure and the risk of serious infections in RA: Study IV was motivated by the apparent disagreement between RCT and observational study results regarding the risk of serious infections associated with the use of glucocorticoids in RA, revealed in a meta-analysis. We hypothesized that the conflicting results could be explained by improper representations of a time-varying exposure and improper confounding adjustment, in previous observational studies. Thus, in study IV we aimed to contrast the incidence of hospitalization for infections (serious infections) between different oral glucocorticoid time-varying dose histories over three years, adjusting for time-varying confounding and selection bias using inverse probability weighting (IPW). We identified 9639 patients newly diagnosed with RA in SRQ and followed them for three years after their first rheumatology visit. To allow the exposure to vary over time, and to adjust for time-varying confounding, each participant’s follow-up was divided into 90-day periods. The average daily-dose of dispensed oral prednisone was calculated in each period, categorizing it into “no use”, “low” (≤ 10mg/day) and “high” (> 10mg/day) doses. Timevarying confounders were measured before each exposure period. The incidence of serious infections over follow-up was modelled by pooled logistic regression, allowing separate effects for different periods of the exposure history. An increased incidence rate of serious infections was associated with higher glucocorticoid doses and the association was stronger for more recent compared to past exposure. Compared to no glucocorticoids, exposure to low doses during the first year added 1.8 serious infection cases per 100 patients (95% confidence interval 0.8 to 2.8) at three years, while exposure to high doses added 4.1 (95% confidence interval 2.5 to 5.8) cases. Hence, our results broadly agree with previous observational studies showing a dose dependent increased risk of infections associated with (recent) use of oral glucocorticoids.