Response to: Correspondence on “Associations of baseline use of biologic or targeted synthetic DMARDs with COVID-19 severity in rheumatoid arthritis” by Sparks et al

Abstract

We appreciate the interest in our paper and the opportunity to provide additional clarification and context. It is now generally accepted that there are different phases of COVID-19 illness. Our study investigated baseline use, presumably at the earliest phase of infection where viral replication is most pronounced. Clinical trials have typically tested the efficacy of medication use among hospitalised patients where pathologic hyperinflammation may mediate poor clinical outcomes. Therefore, the timing of immunomodulator use related to the COVID-19 disease course may have different effects on clinical outcomes, and trial results may not be able to be extrapolated to observational studies of baseline medication use. Observational results may lend insight into key immune mechanisms that could alter the trajectory of COVID-19, but the results could be confounded. To limit possible confounding by indication, our study used an active comparator study design that included patients with a single disease, rheumatoid arthritis (RA), who were taking biologic or targeted synthetic diseasemodifying antirheumatic drugs (b/tsDMARD). In doing so, we included patients with similar disease severity who had the opportunity to receive advanced therapy. We compared each class of b/ tsDMARD to tumour necrosis factor inhibitors (TNFi) since this was the most commonly used b/tsDMARD. Of note, medications were collected by mechanism of action, rather than individual drugs. While this observational study design minimised possible confounding, all patients were on some immunomodulator regimen, so it is possible that some of these medications might have had favourable effects on the COVID-19 disease trajectory. We agree that our finding associating rituximab for use in RA with worsened COVID-19 severity compared with TNFi use has biologic plausibility due to the importance of the adaptive immune response to infections and to SARSCoV-2 in particular. A previous GRA study also associated rituximab use with increased COVID-19related mortality compared with methotrexate. Additionally, another study found that patients with multiple sclerosis (MS) who used rituximab were more likely to be hospitalised than patients without MS. Along with other studies on this topic, 6 these collectively indicate that patients treated with rituximab are at risk for poor COVID-19 outcomes. This is particularly concerning given that the SARSCoV-2 antibody response after COVID-19 vaccination is blunted for rituximab users. However, additional studies are needed to understand whether this blunted antibody response is associated with worse clinical outcomes. We also agree that our findings suggest that interleukin-6 inhibitor (IL6i) or TNFi users may have had a relatively less severe COVID-19 course compared with users of the other b/tsDMARD classes. The RECOVERY trial, which enrolled the largest sample size to date, and REMAPCAP trial, which enrolled critically ill patients, each demonstrated benefit of tocilizumab with regards to mortality in hospitalised patients with severe COVID-19 pneumonia. However, these trial results cannot necessarily be extrapolated to patients with chronic rheumatic diseases using IL6i as outpatients at the time of SARSCoV-2 infection. Early infection, characterised by viral replication, may respond differently in the context of IL6i than later phases of disease, which are characterised by hyperinflammation. In addition to findings from our study, other observational studies of patients with immunemediated diseases have also found that TNFi users may have more favourable outcomes when compared with users of alternative medications. While trial data regarding the efficacy of TNFi for acute COVID-19 are pending, the available evidence to date provides reassurance with regards to COVID-19 outcomes to users of TNFi or IL6i for immunemediated disease. Abatacept use in RA was not associated with COVID-19 severity compared with TNFi use in our primary analysis. In the secondary propensity scorematched analysis, there was a statistical association between abatacept and disease severity, as measured by an ordinal COVID-19 severity scale (OR 1.60, 1.02 to 2.51). However, there were no associations of abatacept with any of the binary outcomes examined (hospitalisation, oxygenation, ventilation, death). While the Spanish study mentioned could suggest favourable COVID-19 outcome in abatacepttreated patients, sample size was low for firm conclusions. Thus, the relationship of abatacept with COVID-19 outcomes requires further study. We found that baseline use of JAK inhibitors (JAKi) for RA at COVID-19 onset was associated with poor outcomes compared with TNFi use. While further replication of our study is required, a previous nationwide Swedish study also suggested that JAKi use may be associated with hospitalisation and death due to COVID-19 compared with conventional synthetic DMARD use, similar to our results. Concomitant glucocorticoid use is unlikely to explain our results. While residual confounding is always possible in an observational study, we adjusted for both use and dose of glucocorticoids. Furthermore, similar proportions of JAKi and TNFi users were on prednisone doses of 10 mg/day or more at baseline (4.6% vs 4.4%). There are now several trials 13 indicating the beneficial effects of baricitinib and tofacitinib for COVID-19, particularly among those in the hyperinflammatory phase (as opposed to the early phase characterised by viral replication). Additionally, these benefits have been observed when JAKi are used in combination with glucocorticoids for COVID-19 treatment. While these results may initially seem to conflict with our findings, we agree that this may be explained by a matter of timing of the use of JAK inhibitors during acute COVID-19. Participants in several of these trials (patients with RA were excluded) were often treated with JAKi in hospitalised patients who may correspond to the hyperinflammatory phase of COVID-19. In contrast, patients in our study had been exposed to JAKi prior to SARSCoV-2 infection, which may result in prolonged SARSCoV-2 viral replication and, therefore, lead to a more severe course. This would also align with findings regarding glucocorticoids, where baseline use or use in patients not requiring oxygen has been associated with a more severe COVID-19 disease course, while dexamethasone use for hospitalised patients requiring oxygen clearly demonstrates clinical benefit. While we caution against overinterpretation of our observational findings, the American College of Rheumatology currently recommends discontinuing JAK inhibitors at onset SARSCoV-2 infection. In summary, each class of b/tsDMARD that we studied could alter the trajectory of the COVID-19 disease course. Our observational study of baseline use of b/tsDMARD needs to be interpreted in the appropriate context of clinical trials of these medications where the intervention occurred at specific moments of clinical deterioration.