Reply to “Comparative Safety Profiles of Ocrelizumab and Rituximab in Multiple Sclerosis Treatment Using Real‐World Evidence”

We thank Dr. Copeland for the thoughtful comments regarding our study “Comparative Safety Profiles of Ocrelizumab and Rituximab in Multiple Sclerosis Treatment Using Real-World Evidence.” We appreciate the close reading of the work and the opportunity to address the points raised. The letter argues that the magnitude of the difference in infection-related hospitalizations reported in our analysis is biologically implausible, as both agents target CD20. Although ocrelizumab and rituximab share a therapeutic target, monoclonal antibodies directed against the same antigen can differ substantially in structure and effector function. Rituximab is a chimeric antibody, whereas ocrelizumab is humanized and engineered to enhance Fc-receptor–mediated activity. These differences influence downstream mechanisms of B-cell depletion, including complement-dependent cytotoxicity and antibody-dependent cellular cytotoxicity, as well as depletion kinetics in peripheral blood and extravascular sites and patterns of immune reconstitution. Such distinctions among anti-CD20 antibodies have been described in both experimental and clinical studies.1-3 For this reason, the observation of different safety profiles among agents within the same therapeutic class should not be considered inherently implausible. The correspondence also highlights the absence of mediation through hypogammaglobulinemia in our exploratory analysis. This mediation analysis was included to examine one potential pathway and was not intended to establish a single mechanism linking treatment exposure to infection risk. Susceptibility to infection in patients receiving B-cell–depleting therapies likely reflects multiple processes, including alterations in immune cell function, B-cell repopulation dynamics, and host factors beyond circulating immunoglobulin levels. The absence of detectable immunoglobulin G-driven mediation, therefore, does not invalidate the association observed between treatment exposure and hospitalization. Additional concerns relate to model specification, temporal confounding, and covariate extraction from clinical documentation. As described in the article and in our previous reply, the analysis incorporated adjustments for more than 20 demographic, clinical, and socioeconomic variables using propensity score matching and an outcome model within a typical doubly robust framework. Calendar time of treatment initiation was included as a covariate, and we performed a restricted analysis limited to patients who received ocrelizumab after approval to align treatment eras. Importantly, the association between rituximab exposure and hospitalization was observed in both the University of California San Francisco cohort and an independent University of California-wide dataset analyzed separately. Observing a similar signal across datasets analyzed with different specifications supports the robustness of the finding. We agree that observational analyses cannot eliminate all potential sources of residual confounding. Nonetheless, real-world datasets remain essential for evaluating the safety of widely used therapies, particularly for outcomes that are uncommon or difficult to capture in randomized trials. Our findings should, therefore, be interpreted as intended, an observational safety signal that merits further investigation, rather than as a definitive causal conclusion. All authors contributed to this reply. G.C. has nothing to report. S.L.H. currently serves on the scientific advisory board of Accure, Alector, Annexon, and Hinge Therapeutics and has received travel reimbursement and writing support from F. Hoffmann-La Roche and Novartis AG for anti-CD20-therapy-related meetings and presentations. These companies had no input in the work presented here. B.A.C.C. has received research support from Genentech and Kyverna. These companies had no input in the work presented here. S.E.B. has nothing to report.