The results in this present study highlight that a majority of patients on BCDT show disease control at 18 months independent of their MS phenotype when considering NEDA-3 and MEDA status with brain and spinal cord analysis. We further suggest that sustained B-cell depletion observed in most of patients at 18 months is insufficient in predicting MS disease control. We show that long-term anti-CD20 therapies comes at the cost of significant treatment-related hypoIgG after 30 months of BCDT and infectious SAEs that are associated with longer treatment duration. Furthermore, prior treatment to BCDT should be taken into consideration given that prior LIT is associated with lower absolute IgG levels and hypoIgG in patients on long-term BCDT.

Our study supports previous results regarding the efficacy of BCDT with a reduction in ARR at 12 and 18 months [22, 31,32,33]. We observed a considerable proportion of patients who were both NEDA-3 and MEDA at 12 and 18 months. No studies have analyzed NEDA-3 outcomes in BCDT including brain and spinal cord MRI studies, and MEDA outcomes have yet to be reported for patients on BCDT. Post-hoc analysis of OPERA I and II showed 72.2% of patients were NEDA-3 after brain MRI re-baseline between 24–96 weeks post-BCDT [34]. Our results taking into account brain and spinal cord MRI are similar considering that 82.8% of RMS patients in our study attained NEDA-3 at 18 months. The fact that we compared MRI imaging at this time-point to the re-baselined MRI provides a strength to our study, since an early re-baselined MRI may reflect persistent inflammatory activity prior to BCDT efficacy and thereby negatively impact NEDA-3 status achievement [34]. Indeed, we observed that failure of NEDA-3 and MEDA at 12 months was primarily due to MRI activity. At 12 months, it is possible that MRI activity may reflect the presence of new lesions prior to treatment initiation, or that new lesions may appear prior to treatment efficacy considering that anti-CD20 therapies have yet to reach its full efficacy during the first six months [34, 35]. MEDA has been shown to be associated with minimal risk for increased long-term disability in RMS patients on interferon beta or glatiramer acetate [30]; nevertheless, future studies will need to address if achieving MEDA status is sufficient to prevent long-term disability in patients on BCDT.

Age has been associated with greater inflammatory disease control [36]. A study by Cellerino et al. observed improved disease control in younger RMS patients with regards to NEDA-3 status at 24 months post-BCDT [36]. However, our study did not find an association with age, which may be due to the inclusion of PMS patients in our cohort who tend to show less clinical and radiological inflammatory activity at an older age. Additionally, lower NEDA at 12 months could be due to the comparison of the re-baselined MRI to the MRI prior to BCDT.

Our results support a growing body of literature that following absolute CD19+ B-cell counts post-BCDT is not sufficient, per se, to predict disease control [10, 14, 15, 27, 37]. While our study did not perform B-cell subset analyses, it has been shown that B-cell subsets post-BCDT tend to be more naive and transitional, and less towards a memory-B phenotype, which is thought to be implicated in MS pathogenesis [27, 37, 38]. Repopulation of CD27+ memory B-cells after rituximab remains low even at Week 52 post-infusion [38]. The slower repopulation kinetics of memory-B cells may also explain as to why early CD19+ B-cell repopulation in patient cohorts with either extended BCDT dosing intervals above the standard 6-month reinfusion or interruption do not show significant clinical worsening or new MRI lesions [10, 14, 15, 39, 40]. BCDT reinfusion based on CD27+ memory B-cells has been adopted by certain groups treating patients with neuromyelitis optica spectrum disorder, myasthenia gravis, as well as patients with MS [11, 16,17,18,19]. Nevertheless, randomized control trials and defined memory B-cell reconstitution cutoffs need to be properly established in MS in order to guide dosing intervals.

Given that NEDA-3 and MEDA criteria were not achieved in patients that showed sustained B-cell depletion, we also looked at other biological markers that could be predictive of MS disease control. Although it has been shown that CD20+ T lymphocytes enriched in the CD8+ T-cell compartment are proinflammatory and are present in periphery of MS patients [38, 41, 42], no T-cell lymphocyte subset was associated with MS disease control. Furthermore, it has been previously observed that disease control in patients treated by rituximab for rheumatoid arthritis (RA) was better in patients with IgM hypogammaglobulinemia [9]; however, we did not observe any Ig isotype as a predictor for disease control.

Over 50% of patients were hypoIgM at 30 months, while just 15% of patients were hypoIgG at this time point, similarly to a previous study investigating long-term rituximab treatment in MS and its effects on gammaglobulin levels [43]. Observational studies suggest that treatment-induced hypogammaglobulinemia in MS is associated with an increased risk of infection, although increased risk of SAE is debated [23, 43, 44]. In our study, SAEs while on long-term BCDT were infrequent, and our results are consistent with previous studies in autoimmune neurological and rheumatological diseases [20, 44,45,46]. Similar to previous studies, patients with SAE had a longer treatment duration than those without SAE, thus suggesting that cumulative long-term exposure to IV BCDT confers a risk of SAE in MS patients [23, 44]. Predictors of SAE, such as hypogammaglobulinemia, were not possible considering the low number of events over the study period.

Few studies have investigated the cumulative effect of previous DMTs prior to BDCT induction and their influence on biological parameters. DMTs such as fingolimod and dimethyl fumarate are known to induce lymphocytopenia, and have been shown to impact T-lymphocyte subsets and increase the risk of T-lymphocytopenia at baseline and up to 12 months after BCDT induction [25]. Our study confirms and extends the results of this former study, as we observed a significant decrease in the absolute CD45+, CD3+ and CD4+ counts and CD4+ lymphocytopenia until 24 months in patients having received prior LIT including fingolimod and dimethyl fumarate, but also cytotoxic immunosuppressive agents. Although we grouped all prior LIT together for analysis in this study, it should be highlighted that S1P receptor modulators (siponimod and fingolimod) included in the analysis differ mechanistically from the other LITs. Given that these treatments prevent the egress of lymphocytes from peripheral lymphoid tissue, this may therefore impact the findings in patients switching directly from an S1P receptor modulator.

In addition to differences in T-lymphocyte subsets, we also observed a significant decrease in absolute IgG levels at baseline and after 24 months in BCDT patients having received prior LIT. A higher proportion of patients with hypoIgG was also observed at 30 months in this population. It has been shown that cyclophosphamide treatment prior to rituximab in anti-neutrophil cytoplasm antibody associated vasculitis was associated with decreased serum IgG concentrations [47], yet concomitant use of methotrexate and rituximab in rheumatoid arthritis patients was observed to be a protective factor in the risk of developing hypogammaglobulinemia [48]. With regard to MS, prior fingolimod treatment has been shown to influence hypogammaglobulinemia [49], while Ig levels remain relatively stable in patients over at least 96 weeks of dimethyl fumarate [50]. MS patients treated with natalizumab have reduced Ig levels over time [49, 51], which may be attributed to impairment of B-cell maturation in the periphery [52]. In our study, it could be possible that prior LIT in BCDT treated patients exerts a synergistic effect on decreasing IgG levels at later BCDT cycles by accelerating the depletion of IgG producing mature B-cells in the peripheral blood. Nonetheless, the ultimate consequences of prior LIT exposure in relation to BCDT is not clear, considering we did not observe an increase in serious opportunistic infections or malignancies in our cohort. Further studies are needed in order to uncouple the role of prior LIT with respect to BCDT.

This study has several limitations, of which the retrospective and observational design, thereby limiting the possibility for collecting all adverse events throughout the study period. However, we focused on grade-3 or above SAEs, which are often documented, and ultimately reduced the likelihood of missing safety data. Furthermore, the monocentric design of our study most likely led to a limited sample size. The retrospective design of the study, inclusion of certain patients in randomized control trials, and loss to follow up, all contributed to missing biological data. Nevertheless, our clinic is an MS expert center with exhaustive clinical, biological and brain and spinal MRI activity in routine, and only 15% of patients lacked initial clinical or biological data. Of note, NEDA takes into account confirmed clinical progression, and therefore NEDA attainment could not be calculated for some patients given the lack of EDSS score at 12 and 18 months. Although we observed limited variations in perfusion intervals, most notably during the COVID-19 pandemic, we observed a significantly longer perfusion interval in comparing before and after March 2020. These results are unsurprising, given that we share space with elective orthopedic surgery, which was less active during the COVID-19 pandemic, and therefore our perfusion clinic was able to maintain clinical activity during this time since. Nevertheless, this difference in delay was clinically irrelevant with regards to the results.

Given the high proportion of patients with MS disease control after the first year of treatment, especially when considering less strict disease activity such as MEDA, it is tempting to suggest a possible opportunity to readapt BCDT at the 18-month time period. Furthermore, given the risk of developing significant treatment-induced hypoIgG at 30 months and the non-negligible risk of SAE occurring after a mean of 4 cycles of BCDT (i.e. 18-months), this time point may be useful in future studies that look to tailor anti-CD20 therapies. Rheumatological diseases have readapted their treatment strategy by a “treat-to-target approach” based on clinical activity [9], yet with respect to certain autoimmune neurological diseases the treatment strategy is based on biological parameters rather than new clinical activity [17, 18]. Our study suggests that sustained CD19+ B-cell depletion is insufficient to predict clinical or radiological disease control in MS patients, and therefore may not be a useful guide in order to aid neurologists in tailoring BCDT reinfusion in this patient population. These data highlight other variables that may need to be considered in the balance for BCDT tailoring, such as prior LIT. A randomized controlled trial would be useful in order to investigate a potential induction strategy followed by maintenance therapy, similar to other autoimmune diseases treated with BCDT.