Objective To estimate the effects of biological drugs on the risk of demyelinating events in rheumatoid arthritis (RA).
Methods Case–control analyses nested in an administrative database cohort.
Results Initially the risk of demyelinating events appeared to be increased after exposure to anakinra and decreased after exposure to antitumour necrosis factor (anti-TNF) agents. However, this apparent differential risk was due to more anakinra use (and avoidance of anti-TNF agents) in persons at high risk for demyelinating events. In individuals not at high risk, the adjusted rate ratio was 1.31 (95% CI 0.68 to 2.50) after exposure to anti-TNF agents and 0.80 (95% CI 0.29 to 2.24) after exposure to anakinra.
Conclusions When accounting for differential prescription patterns, there was a trend towards more events after exposure to anti-TNF agents. When studying rare but important potential drug associations, pharmacoepidemiological studies are valuable but must be carefully performed.
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.
Historically, multiple sclerosis (MS) and other inflammatory demyelinating syndromes were reported only rarely in rheumatoid arthritis (RA).1 2 Since the advent of therapies targeting tumour necrosis factor (TNF), however, numerous case reports have surfaced.3,–,6 Given the paucity of data, we evaluated the incidence of demyelinating events in RA and the influence of drug exposures. This afforded an opportunity to describe ‘channeling bias’, a phenomenon in observational studies resulting from systematic differences in drug prescriptions. This bias may occur once a specific drug becomes associated with an adverse event. In this setting, patients with a possible history of (or risk factors for) that event may be preferentially prescribed a newer agent on the presumption that the new drug is safer. Although observational studies are a useful way to study drug effects, careful analysis is required.
Patients and methods
We assembled a cohort of patients with RA using the PharMetrics Patient-Centric Database (analysed earlier for other outcomes).7 Inclusion criteria were >1 physician billing diagnosis for RA plus >1 disease-modifying agent (disease-modifying antirheumatic drug (DMARD), traditional/biological) dispensed after 1 January 1995. Our exclusion criteria included a history of a confirmed demyelinating event prior to cohort entry defined by ≥2 billing codes (≥8 weeks apart but within 2 years) or ≥1 hospital admission for MS, transverse myelitis or optic neuritis.
Entry into the cohort was the date of the first prescription for either a traditional DMARD or a biological agent (anti-TNF agents etanercept and infliximab or the interleukin 1 receptor antagonist anakinra). Subjects were followed until event, death or end of study (31 December 2005). A newly confirmed demyelinating event was defined by ≥1 hospitalisation diagnosis or ≥2 billing codes.
Case–control analyses were performed using time-dependent drug exposures, selecting 10 random controls for each case, matching on age, sex, calendar time of cohort entry and ensuring that each control remained at risk for a first-time event at the index date (date the case occurred). Drug exposures during the year before the index date were identified from prescription data. No direct measure of RA disease severity was available, but we did control for physician visits as a correlate of disease severity, as well as for non-steroidal anti-inflammatory drugs and glucocorticosteroids.8 9
Conditional logistic regression was used to estimate crude and adjusted rate ratios (RR) for each traditional DMARD or biological agent of interest. Current medication exposure was defined as a prescription dispensed during the 60 days before the index date. DMARD exposure groups in the models included methotrexate, antimalarial agents (hydroxychloroquine/chloroquine), leflunomide, all other traditional DMARDs (sulfasalazine, cyclophosphamide, azathioprine), anti-TNF agents and anakinra (FDA approved in 2001). No other biological agent was on the healthcare plan formulary in 2005.
The RA cohort comprised 104 958 subjects, 87% of whom were female, average follow-up 1.9 person-years. Mean (SD) age at cohort entry was 52.1 (13.7) years. There were 253 newly confirmed demyelinating events, 209 coded as MS and 44 as optic neuritis (with no cases of transverse myelitis). In the sample overall the risk of demyelinating events initially appeared to be increased with anakinra and decreased with anti-TNF agents (see adjusted RR estimates, table 1). However, after reviewing the history of the 253 ‘newly confirmed’ demyelinating events (based on one hospital admission or ≥2 physician visits within a specific period), 171 individuals actually had evidence of a possible demyelinating event prior to cohort entry based on a single physician visit for MS or optic neuritis.
Stratification showed that anakinra was associated with an increased risk for a ‘newly confirmed’ demyelinating event only in subjects with a history of a possible event before enrolment in the cohort (table 1). In this subgroup, anti-TNF agents appeared as strongly protective against a confirmed demyelinating event. Reviewing prescription records established that anti-TNF agents were usually prescribed as the initial biological agent unless subjects had a history (prior to cohort entry) of a possible demyelinating event, in which case anakinra was preferentially prescribed. Among subjects with no physician visits for suspected MS or optic neuritis prior to cohort entry (table 1), there was a trend towards more demyelinating events after exposure to anti-TNF drugs.
In quantifying the risk of drug-related demyelinating events in RA, we have demonstrated the potential for channeling bias in observational studies. In this study we assigned cohort entry as the date of the first RA-related drug prescription and excluded subjects with a history of a confirmed demyelinating event (based on >2 physician billing diagnoses or ≥1 hospitalisation) prior to cohort entry. This meant that some patients with a single physician visit for suspected MS or optic myelitis prior to RA cohort entry were included in our initial analyses. A significant number (N=171) of these went on to become ‘confirmed’ demyelinating' events. The apparent increased risk of demyelinating events with anakinra exposure, suggested by our initial analyses, is undoubtedly due to ‘channeling’—that is, some patients who had a single physician diagnosis for a demyelinating event at some time prior to cohort entry were preferentially prescribed anakinra rather than anti-TNF agents due to physician concern that anti-TNF therapy can precipitate demyelination.10 These subjects were at a high risk for later being confirmed with a demyelinating event, but this risk was due to their past history, not to the anakinra. Of note, for the category ‘other DMARDs’, initial analyses showed an adjusted RR of 1.48 (95% CI 1.01 to 2.19), also suggesting some channeling—that is, preferential use of older second-line agents instead of anti-TNF agents in persons at higher risk for demyelinating events.
TNFα blockade had been investigated in MS before its use in RA with poor results. Case reports of demyelination after exposure to anti-TNF agents began to appear a decade ago10 and coincided with increasing availability of anakinra; this set the stage for channeling. Persons at high risk for MS (eg, a first-degree relative with a demyelinating disease) might also be preferentially prescribed agents other than anti-TNF agents, contributing to an observed artifactual increased risk for anakinra.
In an administrative database where each subject may contribute only a few years of follow-up, incomplete medical history is a potential limitation. National administrative databases can have the advantage of a longer observation interval (since residents remain in the database until death or migration). In an RA cohort assembled from Canadian administrative data (a total of 158 011 patient-years of observation, average ≥6 years/subject), we ascertained only 38 cases of new onset MS in the RA sample, giving an incidence rate of 24 cases/100 000 patient-years. This incidence is very similar to recent figures for the North American general population.11 In contrast, the ‘incidence’ of MS in the PharMetrics Patient-Centric Database was 127/100 000 patient-years, suggesting that prevalent MS cases were undetected at baseline and only ascertained (hence appearing as ‘incident’) during follow-up.
As well as possibly exacerbating existing MS, our calculations suggest that anti-TNF agents may promote the risk of truly incident demyelinating events by about 30%. Although MS is not a frequent occurrence in patients with RA, it may be a devastating blow to those who develop it. Patients should therefore be informed of the possibility of an increased risk with anti-TNF therapy.
Our work draws attention to the ever-present possibility of channeling in observational studies of drug effects. This potential bias should be considered whenever a study suggests a specific drug as protective against an outcome (such as heart failure, infection, malignancy) if this drug might preferentially be avoided with a prior history, clinical suspicion or risk factors related to the outcome of interest. When studying rare outcomes in a rare disease, carefully performed pharmacoepidemiological studies remain of considerable value.
Funding SB is the recipient of career awards from the Canadian Institutes for Health Research (CIHR), Fonds de recherche en santé du Québec (FRSQ), Canadian Arthritis Network and of support from the McGill University Research Institute and Faculty of Medicine. SS is the recipient of the James McGill professorship and the CIHR distinguished investigator award. The McGill Pharmacoepidemiology Research Unit is funded by the FRSQ. CR is the recipient of a postdoctoral fellowship from the Multiple Sclerosis Society of Canada.
Ethics approval This study was conducted with the approval of the Commission d'acces a l'information du Quebec.
Competing interests None.
Provenance and peer review Not commissioned; externally peer reviewed.