Risk of infection in ocrelizumab-treated adults with multiple sclerosis aged 55 years and older: a retrospective cohort study

Summary

Background

Pivotal trials for ocrelizumab in multiple sclerosis (MS) excluded adults >55 years old. This study aimed to compare rates of infection and infection-related hospitalization between ocrelizumab-treated and unexposed older adults with MS.

Methods

A retrospective cohort study was performed, analyzing the electronic medical records, June 2017–January 2026, at the Massachusetts General Hospital, comparing 800 older adults with MS who received at least 600 mg of ocrelizumab IV to 800 older adults who never received B-cell depleting therapies. Infection and related hospitalization rates were modeled using inverse probability of treatment weighting and negative binomial regression with person-time offsets.

Findings

Among 1600 adults (n = 1111, 69.4% female; n = 1496, 93.5% White; median ages 64 (range 55–88) for ocrelizumab and 67 years (range 55–98) for unexposed), median follow-up was 3.5 vs. 8.2 years. There were 429 infection-related hospitalizations (164 ocrelizumab, 265 unexposed), 10 infection-associated deaths (2 ocrelizumab, 8 unexposed), and 4538 infections (1597 ocrelizumab, 2941 unexposed). After adjustment, ocrelizumab was associated with a higher infection-associated hospitalization rate (IRR 1.67; 95% CI 1.15–2.43; p = 0.008) and infections (IRR 1.42; 95% CI 1.20–1.68; p < 0.001). Higher disability and comorbidity independently increased the risk of both outcomes. Female sex was associated with a higher infection risk (IRR 1.36, 95% CI 1.12–1.64, p = 0.002), driven mostly by urinary tract infections, which accounted for 91.46% of the observed sex difference.

Interpretation

In adults aged ≥55 years with MS, ocrelizumab was associated with higher rates of infection-related hospitalization and infections. Findings support individualized risk-benefit discussion, especially regarding UTIs in older patients of female sex, dedicated UTI monitoring and prevention protocols for older ocrelizumab-treated patients, and inclusion of this age group in future clinical trials.

Funding

Investigator-initiated grant from Genentech.

Research in context.

Evidence before this study

The authors performd a search of PubMed.gov for the keyword “ocrelizumab” as well as more refined searches for “ocrelizumab” AND [“infection” OR “elderly” OR “age” OR “outcomes”] for published studies of people treated with ocrelizumab. The initial search was performed for publications from March 1, 2008 [first year of indexed publication using “ocrelizumab” in the title] to March 1, 2025 [the date of the search]. Older age has been shown to increase the risk of infection in small, convenience cohorts of multiple sclerosis (MS) patients treated with ocrelizumab. Larger studies of rituximab-treated patients with MS, of all ages, have shown that infection risk increases with treatment duration, disability level, obesity, and age. Hypogammaglobulinema from B-cell depleting treatment explains less than a quarter of all infections in MS patients. Recently published clinical trial data find that carefully selected MS patients >55 years old continue to benefit from ocrelizumab through preservation of hand function.

Added value of this study

We quantitatively report the risk of infection in a high number of well-characterized MS patients treated with ocrelizumab, using a comparison group of patients never exposed to B-cell depleting treatment in the same older age group. All MS patients were evaluated and treated by neuroimmunologists after careful clinical selection to take ocrelizumab. We separate infections, hospitalizations for infections, deaths related to infection, and the type of infection. We report the relevance of urinary tract infections in older women with MS who take ocrelizumab.

Implications of all the available evidence

Even among carefully selected adults >55 years old with MS, ocrelizumab increases the risk of infection and hospitalizations from infections. Disability level is a leading determinant in their risk of infection, suggesting early line ocrelizumab treatment, preventing later life disability—has both efficacy and safety implications for older adults with MS. Many ocrelizumab-associated infections could be prevented through focusing on urological health in MS, particularly in women, which requires greater clinical emphasis, guidelines development, and health advocacy.

Introduction

Ocrelizumab, a humanized anti-CD20 monoclonal antibody, is U.S. FDA-approved and widely prescribed for the treatment of relapsing and primary progressive multiple sclerosis (MS) in adults and has demonstrated efficacy and safety in several randomized trials.1, 2, 3, 4, 5, 6, 7, 8 Since its approval in 2017, more than 400,000 people have been treated with ocrelizumab globally.⁷ However, the initial clinical trials for ocrelizumab do not reflect the older age ranges of patients now commonly encountered in the MS clinic. In the OPERA I and II trials, the mean age of treated participants was 37 years.³ In the ORATORIO trial, the average age of participants was 44 years, with older subgroups stratified at 45 years and above.² Since people with MS > 55 years old were excluded,^2,3,8,9 treatment decisions on ocrelizumab for people with MS in this age group rely on extrapolation from substantially younger populations, despite the physiological, social, and comorbidity-related differences that accompany aging.^10,11 Real world studies of ocrelizumab-treated people with MS have reported that more than half experience at least one infection during treatment, with respiratory and urinary tract infections being the most common.^12,13 Urinary tract infections (UTIs), in particular, are among the most frequently reported infections in people with MS treated with disease-modifying therapies.^14,15 This increased susceptibility to UTIs is likely due to both disease- and treatment-related factors. Neurogenic bladder dysfunction is common in people with MS due to spinal cord and brainstem lesions which predispose them to recurrent UTIs independent of treatment. Anti-CD20 B-cell depleting therapies further compound this risk by impairing immune responses needed to effectively clear pathogens. This is supported by other studies that found that higher efficacy DMTs, including anti-CD20 therapies carry greater UTI risk than moderate efficacy DMTs (adjusted HR 1.21, 95% CI 1.14–1.30),¹⁶ and that rituximab and natalizumab specifically showed increased UTI risk relative to other DMT classes.¹⁷ Infection-related hospitalizations have also been noted to occur among older MS patients with higher disability.¹²

This evidence gap is becoming more critical as the demographic profile of people with MS shifts. The U.S. and global populations are aging, and improved life expectancy, especially among women, means that more people with MS are living longer.¹⁸ Between 2011 and 2021, the prevalence of MS among older adults in the USA increased by 15%; >50% were taking a disease modifying therapy (DMT) for MS.¹⁹

Concerns regarding infection risk play a central role in treatment decision-making for older people with MS, especially in those with stable disease.²⁰ A large, population-based Swedish registry identified people with MS > 60 years have a threefold risk of serious infection compared to controls.²¹ Age-related immunosenescence and increased susceptibility to infections may alter both treatment risks and therapeutic response.22, 23, 24 Observational studies find that age >50 years is a risk factor for infections, hypogammaglobulinemia, or both, during immunosuppressive therapy.^25,26 Several real-world studies have raised questions about hypogammaglobulinemia with long-term anti-CD20 exposure in relation to infectious risk, particularly in older patients.^25,27,28

This single-institution, propensity score weighted, retrospective cohort study investigates the infection-related safety profile of ocrelizumab in people with MS aged ≥55 years, since the introduction of ocrelizumab. We compare the incidence of infections, infection-related hospitalizations, and deaths between people with MS receiving ocrelizumab IV vs. unexposed.

Methods

Standard protocol approvals, registrations, and patient consents

This protocol was approved by both the Mass General Brigham (MGB) Institutional Review Board (protocol no. 2023P002647) and the Northwestern University Institutional Review Board (protocol no. STU00226450). The need for individual participant consent was waived by the MGB IRB for this retrospective, non-interaction, medical records-based study.

Data source

All participants were evaluated at the Massachusetts General Hospital (MGH). The electronic health records system was queried to identify participants aged ≥55 years for the diagnosis of MS. The study observation period was June 17, 2017, the first day ocrelizumab therapy was given clinically at MGH, until January 5, 2026, the day of termination of data collection for this study.

Patient selection

The inclusion criteria were: (1) adult aged 55 years or older at the index date, (2) confirmed diagnosis of MS (including relapsing and progressive phenotypes), meeting McDonald (2017) criteria,²⁹ (3) followed by a neurologist at MGH, and (4) sufficient records on MS and medications (i.e., documentation of MS treatment at MGH), defined as a medical record containing at least three retrievable neurology notes and/or documented MS history details, extending beyond a diagnosis code alone (e.g., “MS, dx 2007”). Participants whose records could not be accessed were excluded.

Participants were exposed if they received ≥1 complete dose (i.e. 600 mg) of ocrelizumab intravenously during the study period. Participants who received two induction doses (300 mg each) were included and considered as having received one dose. The date of initiation of ocrelizumab for each participant marks the index date for the ocrelizumab group. The exposed group was followed from their index date to the earliest of (1) six months after their last dose of ocrelizumab, (2) date of termination of data collection for this study, or (3) date of death. Treatment duration was calculated as the time between the date of the first complete infusion (600 mg total, whether 300 mg twice or 600 mg once) and six months after the last recorded dose.

Participants were unexposed if they never received B-cell depleting therapy in their lifetimes. Those in the unexposed group were either (1) not treated with a DMT or (2) treated with a non-B-cell depleting DMT. The index date for the unexposed participants was set as June 17, 2017, the first day that ocrelizumab was clinically administered at MGH, and they were followed until the study censoring date or death. The uniform index date for controls ensures that unexposed patients were not selected for downstream survival after ocrelizumab became a counterfactual clinical choice.

Exclusion criteria were (1) a lack of medical history on file, (2) history of chemotherapy, noting cyclophosphamide, alemtuzumab, mitoxantrone, methotrexate, azathioprine, or mycophenolate mofetil were acceptable if they were given for MS or comorbidity management, (3) concurrent participation in a clinical trial involving an experimental or blinded study drug, or (4) history of hematopoietic stem cell transplant.

A pre-specified target of 800 ocrelizumab-exposed people with MS and 800 unexposed (1:1 allocation) was determined based upon study resources and an estimation of the total number of ocrelizumab-treated participants seen at MGH to date. A proprietary search engine of the medical records (Research Patient Data Registry) and the institutional clinical experience led to the anticipated target number of ocrelizumab-exposed people as 800. As the sample size was determined by patient availability rather than a prospective power calculation, we conducted a post-hoc simulation-based analysis to estimate the minimum detectable incidence rate ratio (IRR) at 80% power with a two-sided alpha of 0.05. The minimum detectable IRR was 1.59 for infection-related hospitalizations and 1.31 for infections.

Each patient record was reviewed by at least one study investigator for eligibility. There were 1346 charts reviewed to identify people with MS for the ocrelizumab-exposed group, and 3576 charts reviewed to identify unexposed group (59.4% vs. 22.4% screen-positive for exposed and unexposed, respectively) (Fig. 1).

Fig. 1.

Flowchart of participants. We queried the electronic health record system of Massachusetts General Hospital to identify potentially eligible participants with multiple sclerosis. The study observation period went from June 17, 2017 to January 5, 2026. The electronic records were then manually reviewed for eligibility and included in the study.

Data extraction

Data for all included people with MS were individually extracted from the electronic health records system into a datasheet. Key variables included age, sex, race, ethnicity, comorbidities by International Classification of Diseases codes, MS disease phenotype, age of MS diagnosis, number and names of prior DMTs/immunosuppressive therapies, name of current DMT/therapy, dates of current therapy, number of infections on current therapy, types of infections (and pathogens when available), number of hospitalizations for infections, baseline and most recent Expanded Disability Status Scale (EDSS) scores,³⁰ serum IgG levels, John Cunningham virus (JCV) antibody serology, and death. Baseline EDSS was defined as the most recent measurement recorded within 12 months prior to the index date (date of first ocrelizumab infusion for the ocrelizumab group and June 17, 2017, for the unexposed group). Infections were identified using 1) laboratory-confirmed infections (e.g., positive SARS-CoV-2 test), 2) clinician-documented infections in outpatient, inpatient, or emergency department, or 3) patient-reported infections in portal messages. For the ocrelizumab group, infection events were only included if they occurred between the index date and six months after the person's last dose of ocrelizumab. For the unexposed group, all infection events occurring between the study index date (June 17, 2017) and the censoring date were included. Follow-up criteria differed between groups to reflect the nature of each exposure. For ocrelizumab-treated people with MS, follow-up was six months to capture infections attributable to B-cell depletion of ocrelizumab. For unexposed people with MS, since no equivalent treatment endpoint exists, follow-up continued until the censoring date.

Infection events were categorized by site as: urinary tract (UTI), non-SARS-CoV-2-upper respiratory tract, SARS-CoV-2, lower respiratory/pneumonia, skin/soft tissue, sepsis, and other infections. Pneumonia was further subclassified as SARS-CoV-2, non-SARS-CoV-2-viral, bacterial, fungal, aspiration not otherwise specified, or unknown.

Statistical analysis

Baseline demographic and clinical characteristics were summarized by group. Follow-up time was calculated for each participant in person-months. Data were reviewed to ensure consistent coding of variables and clarify missing data. Distributions of variables were assessed using histograms. To address missing data, we used multiple imputation and conducted our analyses on the imputed data. Variables were imputed using predictive mean matching.^29,30 Stability of imputations were assessed with trace plots, and by comparing means and standard deviations. Rubin's rule was used to pool estimates across imputed data sets.³¹ All statistical analyses were done with R v4.6.0.

Inverse probability of treatment weighting

To account for baseline differences between the Ocrelizumab-treated group and the unexposed group, we used inverse probability of treatment weighting (IPTW) to balance the pre-treatment characteristics of the two cohort groups and control for confounding. Covariate balancing propensity scores (CBPS) were calculated, which maximize for the likelihood of treatment assignment as well as for covariate balance using generalized method of moments estimation.³² The model included treatment group (ocrelizumab exposed/unexposed) as a dependent variable, and known risk factors or confounders, such as age at enrollment, sex, duration of MS, EDSS score, smoking status, the number of prior MS therapies at the time of enrollment, the MS phenotype, and the number of comorbidities as independent variables. These variables were recorded before the time of a person's first ocrelizumab infusion or at the unexposed group's start date. Since the objective of this study was to estimate the treatment effect specifically within the ocrelizumab-treated population, weights were estimated as the average treatment effect on the treated (ATT).³³ The balance of covariates between the two treatment groups was assessed using standardized mean differences (SMDs) of the independent variables, and an SMD <0.1 was assumed to indicate adequate balance. After weighting, balance was achieved for all baseline variables. The SMDs and a love plot of covariate balance are shown in the Supplementary Materials (Supplemental Material 2 and 3). Weights were calculated for each imputed dataset and then pooled across imputed datasets using Rubin's rule.³¹

Outcome model

The pre-defined primary outcome was the number of hospitalizations for infection, expressed as a rate per person-months. The pre-defined secondary outcome was the number of reported infection events, also expressed as a rate per person-months. Both models used the same structure and differed in the dependent variable. A generalized linear model with a negative binomial distribution and a log-link function was used to model the overdispersed count-outcomes over the course of the observational period. The models included the same set of predictors as the CBPS model, with the addition of the treatment group as a binary variable. As follow-up times varied between people with MS, the models were offset by the natural logarithm of the observed person-time. This approach allows modeling the rates of infection over an individual observation period. To check for potential misspecifications, model diagnostics included assessments of the dispersion parameter, zero inflation, and variance inflation factor. To calculate conservative standard errors for the model of this weighted pseudo-population, sandwich standard errors were calculated. Confidence intervals were calculated at the 95% level. Statistical significance was set at p < 0.05.

To quantify the contribution of UTIs to sex differences in overall infection risk, we used g-computation based on the fitted negative binomial models. Infection rates were predicted under counterfactual scenarios in which all participants were set to male or female sex, using models for all infections and for non-UTI infections. The proportion of sex difference attributable to UTIs was calculated as 1 minus the ratio of the non-UTI to total predicted differences. These confidence intervals were calculated using the Wilson score method. Due to inconsistent follow-up schedules and missing data for longitudinal measurements, we did not do a sequential target trial emulation.

Sensitivity analyses

We calculated e-values to assess the sensitivity of the observed associations to unmeasured confounding.³⁴ For missing data, we ran three sensitivity analyses. First, we used complete case analyses instead of multiple imputation analyses. Second, we assessed the robustness of our findings to potential underreporting of infections in the unexposed group. We simulated higher numbers of infections and hospitalizations in the unexposed group and ran the outcome models for underreporting rates of up to 25%. Third, we used pattern mixture models to assess the sensitivity of the observed associations to missingness-not-at-random (MNAR). For that, variables with missing data were imputed with systematically different values than the observed cases. This allows estimation on how severe deviations would have to be in order to impact our primary conclusions.

We conducted five sensitivity analyses to assess the impact of subgroups or risk factors on our outcomes. First, we modelled the association of ocrelizumab treatment with non-UTI infections. The purpose of this analysis was to evaluate, whether excluding the large number of UTI reports would change our conclusions. Second, to assess whether follow-up during COVID might have introduced a systematic bias between groups, we fitted the models including the fraction of total observation time that a participant was followed during the COVID-19 pandemic as an additional variable. Third, we ran the models on the subset of persons that have received B-cell depleting therapies in the past (alemtuzumab, rituximab, ofatumumab, cladribine) in order to assess whether their medication history affects our results. Fourth, we excluded persons that have received ever received B-cell depleting therapies or other strong immunosuppressants (methotrexate, mitoxantrone, cyclophosphamide). Fifth, we included only unexposed participants on a non-B-cell depleting disease modifying therapy as a comparator group.

To assess the effect of time under ocrelizumab treatment on infection outcomes, we added a subgroup analysis with the ocrelizumab group. We fit a negative binomial generalized additive model on the outcome of infection counts with a cubic spline function for time under ocrelizumab treatment, and age, sex, EDSS at baseline, smoking status as predictors. We used the model to predict infection outcomes on a grid of values, and to inspect the shape of the effect of time on ocrelizumab on infection counts. The partial effects plot can be found in the supplementary materials.

Role of the funding source

The funder had no role in the study design, data collection, analysis, interpretation, or in the writing of this manuscript.

Results

A total of 1600 adults (ocrelizumab exposed n = 800, unexposed n = 800) age ≥55 years old with a diagnosis of MS were included (age range 55–88 years (ocrelizumab) and 55–98 years (unexposed)). The total follow-up time was 112,023.71 person-months: 34,977.2 (ocrelizumab) + 77,046.5 (unexposed). Median person-months in follow-up were 41.8 (3.5 years) (ocrelizumab) and 98.2 (8.2 years) (unexposed).

1496 (93.5%) of the participants were recorded as White, and 1111 (69.4%) were female. The median age was 64 years (ocrelizumab) and 67 years (unexposed) (Table 1). In both groups, the largest proportion of participants were in the five-year epoch of 60–64 years old.

Table 1.

Demographic and clinical variables of the patient population in this study.

	Ocrelizumab-Exposed	Ocrelizumab-Unexposed	Overall
	(N = 800)	(N = 800)	(N = 1600)
Sex
Female	509 (63.6%)	602 (75.3%)	1111 (69.4%)
Male	291 (36.4%)	198 (24.8%)	489 (30.6%)
Race
White	738 (92.3%)	758 (94.8%)	1496 (93.5%)
Black	36 (4.5%)	22 (2.8%)	58 (3.6%)
Unknown/other	26 (3.3%)	20 (2.5%)	46 (2.9%)
Age (years)
55–59	206 (25.8%)	135 (16.9%)	341 (21.3%)
60–64	229 (28.6%)	178 (22.3%)	407 (25.4%)
65–69	179 (22.4%)	154 (19.3%)	333 (20.8%)
70–74	118 (14.8%)	139 (17.4%)	257 (16.1%)
75–79	43 (5.4%)	111 (13.9%)	154 (9.6%)
80+	25 (3.1%)	83 (10.4%)	108 (6.8%)
Current MS phenotype
PPMS	158 (19.8%)	107 (13.4%)	265 (16.6%)
RRMS	425 (53.1%)	405 (50.6%)	830 (51.9%)
SPMS	180 (22.5%)	124 (15.5%)	304 (19.0%)
Unknown/other	37 (4.6%)	164 (20.5%)	201 (12.6%)
Years since MS diagnosis
<10	180 (22.5%)	85 (10.6%)	265 (16.6%)
10–19	215 (26.9%)	199 (24.9%)	414 (25.9%)
20–29	208 (26.0%)	231 (28.9%)	439 (27.4%)
30–39	112 (14.0%)	140 (17.5%)	252 (15.8%)
40+	41 (5.1%)	59 (7.4%)	100 (6.3%)
Unknown	44 (5.5%)	86 (10.8%)	130 (8.1%)
No. of prior therapies
0	195 (24.4%)	370 (46.3%)	565 (35.3%)
1	189 (23.6%)	192 (24.0%)	381 (23.8%)
2	171 (21.4%)	94 (11.8%)	265 (16.6%)
3	126 (15.8%)	37 (4.6%)	163 (10.2%)
4	62 (7.8%)	16 (2.0%)	78 (4.9%)
5+	49 (6.1%)	8 (1.0%)	57 (3.6%)
Unknown	8 (1.0%)	83 (10.4%)	91 (5.7%)
EDSS baseline
Median (IQR)	3.5 (3.5)	2.5 (5.0)	3.0 (4.0)
Missing/Unknown	346 (43.3%)	499 (62.4%)	845 (52.8%)
Most recent EDSS
Median (IQR)	5.0 (4.0)	3.0 (4.5)	3.5 (4.5)
Missing/Unknown	315 (39.4%)	486 (60.8%)	801 (50.1%)
Smoking
No	734 (91.8%)	735 (91.9%)	1469 (91.8%)
Yes	56 (7.0%)	59 (7.4%)	115 (7.2%)
Unknown	10 (1.3%)	6 (0.8%)	16 (1.0%)
Comorbidities
Anxiety	125 (15.6%)	136 (17.0%)	261 (16.3%)
Asthma	54 (6.8%)	71 (8.9%)	125 (7.8%)
COPD	69 (8.6%)	92 (11.5%)	161 (10.1%)
Cancer (any)	84 (10.5%)	117 (14.6%)	201 (12.6%)
Adrenal cancer	1 (0.1%)	0 (0%)	1 (0.1%)
Bladder cancer	2 (0.3%)	5 (0.6%)	7 (0.4%)
Breast cancer	22 (2.8%)	30 (3.8%)	52 (3.3%)
CNS cancer	2 (0.3%)	1 (0.1%)	3 (0.2%)
Colorectal cancer	9 (1.1%)	16 (2.0%)	25 (1.6%)
Endometrial cancer	3 (0.4%)	2 (0.3%)	5 (0.3%)
Head and neck cancer	1 (0.1%)	1 (0.1%)	2 (0.1%)
Hematological cancer	7 (0.9%)	9 (1.1%)	16 (1.0%)
Lung cancer	3 (0.4%)	5 (0.6%)	8 (0.5%)
Ovarian cancer	2 (0.3%)	5 (0.6%)	7 (0.4%)
Peritoneal carcinoma	1 (0.1%)	0 (0%)	1 (0.1%)
Prostate cancer	8 (1.0%)	7 (0.9%)	15 (0.9%)
Skin cancer	14 (1.8%)	25 (3.1%)	39 (2.4%)
Thyroid cancer	6 (0.8%)	4 (0.5%)	10 (0.6%)
Unknown cancer type	2 (0.3%)	1 (0.1%)	3 (0.2%)
Uterus cancer	1 (0.1%)	0 (0%)	1 (0.1%)
Kidney cancer	0 (0%)	1 (0.1%)	1 (0.1%)
Pancreas cancer	0 (0%)	4 (0.5%)	4 (0.3%)
Sarcoma	0 (0%)	1 (0.1%)	1 (0.1%)
Chronic kidney disease	14 (1.8%)	46 (5.8%)	60 (3.8%)
Chronic liver disease (incl. Hep. C)	28 (3.5%)	58 (7.3%)	86 (5.4%)
Depression	179 (22.4%)	186 (23.3%)	365 (22.8%)
Diabetes	78 (9.8%)	126 (15.8%)	204 (12.8%)
Heart disease	93 (11.6%)	169 (21.1%)	262 (16.4%)
Hypertension	173 (21.6%)	258 (32.3%)	431 (26.9%)

Current DMT for unexposed group at baseline is reported in Appendix 29.

Relapsing remitting MS was more frequently listed in both groups (n = 425 (53.1%) in ocrelizumab and 405 (50.6%) in unexposed group). Disease duration since MS diagnosis was somewhat longer in the unexposed group (median 22 years) although overall similar to the ocrelizumab group (median 19 years). The unexposed group had a larger proportion of participants in the categories ≥20 years since MS diagnosis.

Despite shorter disease duration, the disability level, measured by EDSS, was higher in ocrelizumab-treated people with MS. The median baseline EDSS (defined as the score prior to first ocrelizumab exposure) was 3.5, increasing to 5.0 at the most recent assessment. By comparison, the baseline and most recent EDSS scores for the unexposed were 2.5 and 3.0, respectively. Baseline EDSS was missing for 346 (43.3%) ocrelizumab-treated and 499 (62.4%) unexposed participants; most recent EDSS was missing for 315 (39.4%) and 486 (60.8%), respectively (Table 1). There were 176 ocrelizumab and 85 unexposed people with an EDSS of ≥5.

Most ocrelizumab-treated people (n = 408, 51.1%) took ocrelizumab as their third or higher line DMT (Tables 1 and 2). By contrast, nearly half of the unexposed group (n = 370, 46.3%) had never taken a DMT. Current DMT use in the unexposed group at baseline is reported in Supplemental Material 29. Hypertension, heart disease, depression, and anxiety were the most common comorbidities in both groups (Table 1).

Table 2.

Disease modifying therapies prior to beginning of the study.

	Ocrelizumab-Exposed	Ocrelizumab-Unexposed
Interferon Beta	246	142
Glatiramer acetate	224	123
Dimethyl fumarate	154	57
Natalizumab	103	13
Fingolimod	88	10
Rituximab	66	0
Teriflunomide	57	33
Cyclophosphamide	35	2
Mycophenolate Mofetil	33	4
Methotrexate	11	6
Daclizumab	6	1
Azathioprine	5	1
Alemtuzumab	4	1
Siponimod	4	3
Diroximel Fumarate	3	10
Ofatumumab	3	0
Basiliximab	2	0
Mitoxantrone	2	0
Cladribine	1	0
Sulfasalazine	1	0
Ozanimod	0	1

A patient may have had multiple different therapies before the beginning of this study.

Serum IgG was measured in 87.3% (n = 698) of the ocrelizumab group and 23.8% (n = 190) of the unexposed group. Hypogammaglobulinemia, defined as a serum IgG level below 600 mg/dL at any point, was observed in 191 participants: 167 in the ocrelizumab and 24 in the unexposed groups. JCV antibody testing was performed in 598 (74.8%) participants in the ocrelizumab group and 212 (26.5%) participants in the unexposed group. Among them, 406 (67.9%) and 117 (55.2%), respectively, tested seropositive (titer >0.4).

A total of 4538 infection events were recorded (1597 from the ocrelizumab group and 2941 from the unexposed group) (Table 3). A higher proportion of the ocrelizumab group had no infections compared with the unexposed group (n = 210, 38.8% vs. n = 157, 19.6%). In contrast, recurrent infections (≥5 during follow-up) were more common in the unexposed: 14.1% (n = 113, ocrelizumab) vs. 27.8% (n = 222, unexposed). Among the ocrelizumab-exposed group who experienced at least one infection, the median time to first infection was 414 days (1.14 years, 95% CI 368–510 days) after ocrelizumab initiation. In comparison, the time to first infection was longer for the unexposed group with a median of 876 days (95% CI 737–987 days, p < 0.001).

Table 3.

Infections and hospitalizations events.

	Ocrelizumab-Exposed	Ocrelizumab-Unexposed	Overall
	(N = 800)	(N = 800)	(N = 1600)
Total number of hospitalizations	164	265	429
Number of infection-related hospitalizations
0	679 (84.9%)	669 (83.6%)	1348 (84.3%)
1	91 (11.4%)	79 (9.9%)	170 (10.6%)
2	21 (2.6%)	24 (3.0%)	45 (2.8%)
3	7 (0.9%)	14 (1.8%)	21 (1.3%)
4+	2 (0.3%)	14 (1.8%)	16 (1.0%)
Total no. of recorded infections	1597	2941	4538
Infection events per patient
0	310 (38.8%)	157 (19.6%)	467 (29.2%)
1	164 (20.5%)	155 (19.4%)	319 (19.9%)
2	91 (11.3%)	107 (13.4%)	198 (12.3%)
3	80 (10.1%)	83 (10.4%)	163 (10.3%)
4	42 (5.3%)	76 (9.5%)	118 (7.4%)
5	31 (3.9%)	52 (6.5%)	83 (5.2%)
6+	82 (10.3%)	170 (21.3%)	252 (15.8%)
Number of infections events by subtype
UTI	700	1139	1839
SARS-CoV-2	289	356	645
Pneumonia	73	149	222
Bacterial	8	63	71
SARS-CoV-2 pneumonia	33	8	41
Aspiration	5	28	33
Non-SARS-CoV-2 viral	2	1	3
Fungal	0	1	1
Unknown	25	48	73
Upper-respiratory tract	216	368	584
Lower-respiratory tract	86	230	316
Soft tissue	84	238	322
Sepsis	25	55	80
Other	124	406	530

Urinary tract infections (UTIs) were the most common infection subtype, followed by COVID-19 and other non-COVID upper respiratory infections in both groups (Fig. 2). Lower respiratory, soft-tissue, and other infections were more frequently recorded in the unexposed group than in the ocrelizumab group. Sepsis was uncommon in both groups, with 25 events (ocrelizumab) and 55 events (unexposed). There were 222 pneumonia cases: 73 (ocrelizumab) and 149 (not). SARS-CoV-2 pneumonia was most common in the ocrelizumab group (n = 33, 45.2%), while bacterial pneumonia was most common in the unexposed group (n = 63, 42.3%). The non-COVID infection incidence rates were 37.4 vs. 33.6 per 1000 person-months (ocrelizumab vs. unexposed). Two people with MS (never exposed to ocrelizumab) had active tuberculosis.

Fig. 2.

Total infection-related adverse events by treatment group (ocrelizumab-exposed group in peach and the unexposed group in turquoise). A, Total number of infection-related hospitalizations and infection events by treatment group. B, Distribution of reported infection subtypes by treatment group.

429 infection-related hospitalizations were recorded: 164 in the ocrelizumab exposed and 265 in the unexposed groups. 84.9% (n = 679) of exposed and 83.6% (n = 669) of unexposed experienced no infection-related hospitalizations, while recurrent hospitalizations (≥3 events) occurred in 9 (1.1%) and 28 (3.5%) participants. Among the ocrelizumab group who experienced ≥1 hospitalization, the median time to first hospitalization was 829 days post-ocrelizumab initiation. Infections by type in the ocrelizumab-exposed group were as follows: UTI (n = 64), COVID-19 (n = 44), pneumonia, including influenza and respiratory syncytial virus (RSV) (n = 16), cellulitis (n = 5), bacteremia (n = 4), wound infection (n = 4), appendicitis (n = 2), gallbladder infection/cholecystitis (n = 2), unspecified or other infections (n = 15), and multiple concurrent infections (n = 8). The incidence rate of infection-related hospitalizations was 4.69/1000 person-months (ocrelizumab) vs. 3.44/1000 person-months (unexposed). There were 10 deaths: 2 cases, 8 unexposed.

After adjusting for sex, age, comorbidity burden, baseline EDSS, and MS disease duration, ocrelizumab was associated with higher rates of infection-related hospitalizations compared with unexposed people with MS (incidence rate ratio (IRR) 1.67, [1.15–2.43], p = 0.008) (Table 4). Higher comorbidity burden (IRR per additional comorbidity, 1.34 [1.19–1.52], p < 0.001) and higher baseline EDSS (IRR per 1-point increase, 1.56 [1.37–1.76], p < 0.001), were also independently associated with greater hospitalization risk. Age, sex, MS disease duration, and MS phenotype were not statistically significantly associated with infection-related hospitalization risk.

Table 4.

Model estimates for infection-related hospitalizations and reported infections.

	Hospitalization IRR (95% CI)	p-value	Infection IRR (95% CI)	p-value
Ocrelizumab treatment	1.667 (1.146–2.426)	0.0075	1.418 (1.200–1.675)	<0.0001
Female sex	0.673 (0.437–1.035)	0.071	1.360 (1.124–1.644)	0.0015
Age	1.014 (0.986–1.042)	0.34	1.004 (0.992–1.017)	0.51
Comorbidity sum	1.341 (1.186–1.516)	<0.0001	1.213 (1.145–1.284)	<0.0001
EDSS at baseline	1.559 (1.377–1.766)	<0.0001	1.161 (1.096–1.231)	<0.0001
MS duration	1.004 (0.983–1.027)	0.70	1.002 (0.992–1.012)	0.70
MS Phenotype: PPMS	0.696 (0.371–1.306)	0.26	0.894 (0.670–1.191)	0.44
MS Phenotype: SPMS	0.970 (0.553–1.699)	0.91	1.075 (0.815–1.418)	0.61
No. of prior therapies	0.961 (0.810–1.141)	0.65	0.965 (0.885–1.052)	0.42
Active smoking	0.436 (0.228–0.834)	0.012	0.657 (0.485–0.890)	0.0067

Ocrelizumab exposure was associated with a higher incidence rate of infections and infection-related hospitalizations.

In the corresponding model for infection incidence (Table 4), ocrelizumab was associated with higher infection rates (IRR 1.42 [1.20–1.68], p ≤ 0.001). Similar to hospitalization risk, comorbidity burden (IRR per additional comorbidity, 1.21 [1.14–1.28], p < 0.001) and baseline EDSS (IRR per 1-point increase, 1.16 [1.10–1.23], p < 0.001) were associated with significantly higher infection incidence. Female sex was associated with higher infection rates (IRR 1.36, [1.12–1.64], p = 0.002).

The observed IRRs for both primary outcomes exceed the minimum detectable effect sizes identified in the post-hoc power simulation, as described in Methods.

Using g-computation, we estimated that UTIs accounted for 91.46% of the observed difference in overall infection risk between male and female participants (95% CI 87.35–94.32%) There was no statistically significant independent association with age or MS disease duration.

To further characterize UTI risk, we fit a separate negative binomial model with only UTI events, using the same covariate set as the primary models. In this model, ocrelizumab exposure (IRR 1.94, 95% CI 1.42–2.65, p < 0.001), female sex (IRR 1.97, 95% CI 1.30–2.99, p = 0.001), comorbidity burden (IRR 1.20, 95% CI 1.07–1.35, p = 0.001), and higher baseline EDSS (IRR 1.36, 1.25–1.48, p < 0.001) were each independently associated with higher UTI risk (Supplemental Material 18).

To evaluate what strength an unobserved confounder would need to have in order to explain away the observed association, we calculated e-values for the association of ocrelizumab with more infections and hospitalizations. For the association of ocrelizumab with hospitalizations, the e-value was 2.72 (lower bound 1.56), and for the association with infection reports, it was 2.19 (lower bound 1.69).

The effective sample size (ESS) after multiple imputation and weighting was 800 for the ocrelizumab group, and 237.91 for the unexposed group; on complete-cases data, the ESS was 409 and 104.37, respectively. Using complete-case analysis, instead of multiple imputation, in our model did not change the conclusions (IRR for ocrelizumab and hospitalization 2.09, 95% CI 1.27–3.43, p = 0.004; IRR for ocrelizumab and infection 1.39, 95% CI 1.14–1.71, p = 0.001). Pattern mixture modelling, where we simulated up to ±2 points of bias in baseline EDSS measurements, revealed no changes to the significance of the associations between ocrelizumab and infections or ocrelizumab and hospitalizations.

Discussion

Clinical outcomes data from large groups of older adults with MS allow prescribing decisions based on real-world evidence, with measured assessments of risks vs. benefits. To date, most data on older people with MS have been inferred from smaller cohorts of rituximab-treated people with MS or other autoimmune diseases, or extrapolated from initial trials.^11,12,28,35 Our 1:1 allocated study of 1600 older adults with MS, each diagnosed, evaluated, and characterized by a neurologist since the introduction of ocrelizumab onto the U.S. market, allows longer-term follow-up of ocrelizumab-treated people with MS, including the risks of infections, hospitalization-associated infections, and deaths.

We identified a high number of ocrelizumab-treated older adults. This involves a high number of people with MS who aged into the defined range after they initiated ocrelizumab. Our study includes all adults with ocrelizumab exposure at this age range, regardless of baseline risk factors, exposure duration, and disease duration and severity. Each person was treated by a neurologist, usually an MS subspecialized neurologist, at our academic institution. Each chart was reviewed individually for the purposes of this analysis. We found that ocrelizumab-treated people with MS had a higher rate of hospitalizations for infections than the unexposed, the latter encompassing mostly people with MS on either no DMT or lower-efficacy agents. This association should be interpreted cautiously given the potential for residual confounding by disability, as discussed further below.

Our findings can be benchmarked against a large Swedish register-based study, which reported a threefold higher risk of serious infection among people with MS compared to the general population.²¹ This study, however, examined MS status rather than treatment exposure within an MS population. They also used a narrower outcome definition of “serious infections,” defined as an infection that led to hospitalization and/or death, compared to our infection count. Using our most comparable outcome, infection-related hospitalization rate, the Swedish cohort's rate for progressive MS was 32.6 per 1000 person-years (2.72 per 1000 person-months), which is lower than the rate we observed both in our ocrelizumab (4.69 per 1000 person-months) and unexposed groups (3.44 per 1000 person-months).

This difference may reflect differences in hospital admission thresholds and study populations between the US and Sweden. Notably, the Swedish cohort also identified UTIs as the infection subtype with the biggest excess infection risk associated with MS disease severity, which is consistent with our findings. In their model restricting the outcome to only UTIs, baseline EDSS was independently associated with higher UTI risk (IRR 1.36, 95% CI 1.25–1.48, p < 0.001) alongside ocrelizumab exposure (IRR 1.94, 95% CI 1.42–2.65, p < 0.001) and female sex (IRR 1.94, 95% CI 1.30–2.99, p = 0.001).

The risk of infection should be interpreted in the context of participant selection. Overall, the ocrelizumab-treated participants had higher baseline disability, which was a significant difference compared to the unexposed. Disability is a major reason to prompt ongoing treatment in MS, even as it portends a higher risk of infections, for example, due to immobility or spinal cord dysfunction that may also lead to bladder dysfunction and UTIs.

However, it is worth noting that there was substantial missingness in baseline EDSS in both groups (n = 346, 43.3% ocrelizumab; n = 499, 62.4% unexposed). Thus, if unmeasured disability is higher among ocrelizumab-treated participants with missing EDSS scores than our model captures, the true association between ocrelizumab treatment and infection risk would be smaller than observed in this study. Additionally, we did not account for MS relapses that might have occurred in proximity to the baseline, which could transiently elevate EDSS scores, or most recent EDSS measurements. As we do not have data on relapses in both groups, the direction of any resulting bias on our estimates is unclear. Recent results from the phase IIIb ORATORIO-HAND trial demonstrated that ocrelizumab retains clinical efficacy in people with MS who are older and have greater disability (median EDSS 6.0), including slowing disability progression and preservation of upper limb function, as measured by the 9-hole peg test.³⁶ In our cohort, disability at the time of last follow-up for the ocrelizumab group was similarly high (median EDSS score 5.0), reflecting significant impairments in daily activities.³⁰ The ocrelizumab-exposed group with MS overall had tried at least two prior DMTs and had a higher need for treatment compared to the unexposed group. Ocrelizumab-unexposed people with MS had lower average EDSS scores, longer MS disease duration, and half of whom had never taken a DMT.

While hospitalizations for infections were higher among ocrelizumab-treated participants, the number of fatal infections in both groups was very low (2 in the ocrelizumab group, 8 unexposed). Since our study spanned the COVID-19 pandemic, the number of observed infections may be higher here than in a more typical timeframe.

There were notable differences in follow-up duration between the two groups (3.5 years in ocrelizumab vs. 8.2 years in the unexposed group). Such differential surveillance and reporting practices may also contribute to the higher infection rate in the ocrelizumab group; ocrelizumab-treated participants with more frequent physician contact may have more comprehensive capture of individual infection episodes compared to those with less structured follow-up. We also identified several other important and previously recognized risk factors for infections. Increasing comorbidity burden was associated with a higher risk of infection and infection-related hospitalizations across groups. The high proportion of comorbidities in this study's people with MS reflects the U.S. burden in general, and in fact may be lower than the nationally recognized prevalences of hypertension, diabetes, anxiety, and depression.37, 38, 39, 40 We also note a very low prevalence of smoking habits in our older adults with MS, making the relationship between smoking and ocrelizumab on infections difficult to assess here.

Although male sex was associated with lower overall infection risk, participants of male sex had a higher risk for infection-related hospitalizations, suggesting differences in infection severity. The excess infection risk in older participants of female sex were driven by their higher incidence of UTIs and is worthy of future work, both for treatment and prevention guidance for older people of female sex with MS as well as more research on preventative approaches. While these variables were included in both the weighting and outcome model to balance the characteristics of treatment groups, residual confounding by unmeasured variables remains possible.

We analyzed available serum immunoglobulin G levels by group. 698 (87.3%) of ocrelizumab-treated people with MS had at least one IgG measurement, and of them, 167 (23.9%) had at least one IgG level consistent with hypogammaglobulinemia. Given the lack of consistent testing in the unexposed group (only n = 190, 23.8% had ≥1 measurement), we cannot comment on the excess prevalence of hypogammaglobulinemia in ocrelizumab-treated participants but note that approximately 1 in 4 older adults on ocrelizumab experienced at least one laboratory value consistent with hypogammaglobulinemia.

The relationship between infections and ocrelizumab has been studied by others.^12,13,41,42 Several factors may contribute to infections in ocrelizumab-treated people with MS. Epidemiologically, there is confounding by indication; high-risk people with MS, including those with higher EDSS scores and those who have already tried other high-efficacy therapies, are more likely to be selected for ocrelizumab at older ages.⁴³ Our ocrelizumab group includes participants who have previously taken natalizumab, rituximab, and other high efficacy DMTs. The benefit–risk profile of high-efficacy drugs in older adults with MS is not straightforward. While ocrelizumab has demonstrated efficacy in slowing disability progression, the relative benefit may diminish with increasing age as inflammatory disease activity naturally declines.^24,36

The reasons ocrelizumab may increase the risk of infections, independent of other risk factors, almost certainly relate to its primary mechanism of action: selective, sustained depletion of CD20-positive B cells, leading to decreased antibody-mediated immune responses to new pathogens.³ The MUSETTE study found that higher doses of initial ocrelizumab did not improve MS disease outcomes but increased the risk of infections.⁴⁴ The relevance of hypogammaglobulinemia for ocrelizumab-associated infections is not settled, as different reports conflict.^13,45 Others have found neutropenia to be a significant finding in chronic ocrelizumab treatment, but not necessarily an infectious risk.^46,47 Age itself is an important and leading risk factor for most infections, due to immunosenescence.²¹ This can impact a wide variety of factors, including vaccine response, pathogenicity of a virus, colonization, and the severity and recovery from an illness.

Important practice patterns must be taken into account, particularly when prescribing ocrelizumab. We have relatively standard prescribing of 600 mg IV ocrelizumab every 26 weeks, with overall very limited variations on this practice. While others anecdotally may prescribe over longer epochs, follow CD20-positive B-cell repopulation on serial laboratory draws, or follow other factors, the almost universal practice at our center includes dosing as per the U.S. FDA-approved label. The use of intravenous immunoglobulin for hypogammaglobulinemia, particularly at mild levels or without persistence, is not routine.

Although we did not specifically collect individual vaccination histories, Massachusetts has a high overall vaccination rate for influenza (53.7% in 2024 vs. 41.3% national average) and has consistently reported adult vaccination coverage above the national average.^48,49 Thus, we would anticipate that the rates of vaccine-preventable infections in our cohort are comparable to, if not lower than, other regions, although this was not directly assessed in our study.

Several limitations are notable. First, the retrospective nature of our design means that we do not have consistent follow-up intervals, assessments, or other aspects of the participant evaluation. Observational studies are subject to specific sources of bias and do not replace randomized controlled trials; the lack of standard follow-up in the unexposed group, in contrast to the ocrelizumab group's more structured clinical contact schedule, may inflate ocrelizumab's association with infection risk. Participants were prescribed ocrelizumab by their treating neurologist. Patients who were considered too high risk or with too mild a disease course would not be treated with ocrelizumab, particularly in the earlier years of its availability. If high-risk patients were less likely to receive ocrelizumab, the study could underestimate ocrelizumab's infection risk. Furthermore, immortal time bias, i.e. that the ocrelizumab-group participants had to live long enough to be considered for ocrelizumab, cannot be ruled out and would also lead to an underestimation of ocrelizumab-associated infection risk. Important variables demonstrated significant missingness, notably EDSS, which may further introduce confounding; if those with greater disability were less likely to have recorded EDSS values, the infection risk may be underestimated in the unexposed group. While the model conclusions remain unchanged in sensitivity analyses to the missingness-at-random (MAR) assumption, missingness-not-at-random (MNAR) remains a potential source of bias in this study. Additionally, while baseline EDSS was required to be recorded within 12 months of the index date for both groups, the time elapsed between baseline EDSS assessment, and the index date was not calculated for each group. Due to missing data, we did not conduct a sequential target trial emulation, which could have strengthened causal inference and avoided a uniform index date for the unexposed group.

Furthermore, our cohort was drawn from a single tertiary academic center in Massachusetts with a predominantly White population (93.5%), limiting our study's generalizability. Detailed ethnicity data were not collected, and thus, we were unable to assess whether infection risk differs by ethnicity. If other subpopulations experience higher infection incidences, this can potentially lead to larger effect sizes of ocrelizumab treatment on infection outcomes.

Finally, this analysis was not population-based. It is possible, although not likely, that infections were not reported or that people with MS out-migrated without documentation in the medical record. This may lead to underestimation of infection risk, particularly for the control group, which did not have clinical follow-ups at fixed intervals. We also did not collect certain variables of potential interest, such as body mass index and neutrophil counts. Omission of these potential confounding variables could bias our estimates away from the null. People with MS with prior DMT treatments to ocrelizumab would be at a potentially higher long-term risk of infection. Since ocrelizumab-treated participants had substantially greater prior exposure to other high-efficacy DMTs than unexposed participants, not fully accounting for this cumulative treatment burden likely biases our results away from the null as well. Our observation time on ocrelizumab-treated participants is still less than five years on average, and our assessment combines a wide range of older ages, from 55 to >80 years old. Thus, longer-term infection risk associated with sustained B-cell depleting therapy may not be fully captured, and our estimates may underestimate the true risk associated with prolonged ocrelizumab exposure. Further study of the changing role of the immune system in MS across different ages is warranted. As residual confounding cannot be ruled out in our study, future research should, as a first step, aim to replicate this study with an independent cohort.

Our study also had several important strengths. We carefully identified the largest group of well-characterized people with MS who are ≥55 years old to date with ocrelizumab treatment. This includes a detailed assessment of their medical history, MS disease history, DMT treatment course, and both infectious and overall outcomes. There were important absences in the incidence of new infections. We found a single case of primary CNS lymphoma in a participant on ocrelizumab and 8 cases of shingles in the ocrelizumab group. The low prevalence of hospitalization-associated infectious fatalities is notable, particularly given the long observation period, including the COVID-19 pandemic. We identify the high relevance of comorbidity management in the outcome of people with MS in both ocrelizumab-treated and untreated groups, including vascular disease risk factors and diseases of lifestyle and aging. We further identify the major relevance of UTIs and a sex-based impact on MS outcomes. Finally, we demonstrate that baseline EDSS score is particularly relevant to treatment outcomes for infections.

As the treatment of MS changes to early, aggressive, high-efficacy agents and the diagnosis of MS includes earlier disease phenotypes, use of biomarkers, and careful early monitoring, the future of older adults’ treatment in MS will also change. We anticipate that lower disability, as measured by EDSS score, at the initiation of ocrelizumab will be highly favorable for future patients. Meanwhile, there is an unmet need for a detailed prospective study of older adults, including ocrelizumab dosing strategies to minimize infectious risks, structured monitoring protocols for UTIs and other infections in this demographic, and expanded inclusion of older adults in future DMT clinical trials to inform such protocols.

Contributors

SL and FJM contributed to the conception and design of the study; SL, RS, NE, JH, JS, MM, and SS contributed to the acquisition and analysis of data; SL, RS, and FJM verified the underlying data reported in the manuscript; SL, RS, and FJM contributed to drafting the text or preparing the figures. All authors read and approved the final version of the manuscript.

Data sharing statement

The data supporting the findings of this study contain identifiable individual-level information and cannot be shared publicly due to ethical and legal restrictions. The statistical analysis code is available at https://codeberg.org/zhili97/mgh_ms.git.

Declaration of interests

FJM has received research funding support from Amgen, Genentech (the manufacturer of ocrelizumab), Novartis, and TG Therapeutics, and consulting fees from Alexion, Amgen, Genentech/Roche, and Novartis. All other authors have nothing to report.