Do immune checkpoint inhibitors affect the course of multiple sclerosis? A systematic review and meta‐analysis

Abstract

Background and purpose

Elderly people with multiple sclerosis (pwMS) present higher probability of malignancies. Immune checkpoint inhibitors (ICIs) improve cancer prognosis but pose risk of disease flares in people with pre‐existing autoimmune conditions, including MS. Data addressing the impact of ICIs on MS are scarce. This systematic review and meta‐analysis evaluates the effects of ICIs on MS disease activity.

Methods

A systematic literature search in Google Scholar and PubMed, following PRISMA 2020 guidelines, identified five observational studies. Data on clinical and neuroradiological outcomes were analyzed using random‐effects models.

Results

The clinical activity meta‐analysis included 90 pwMS undergoing ICI therapy (median follow‐up = 0.62–1.85 years, 103.74 patient‐years). The pooled relapse rate was 5.45 per 100 patient‐years (95% confidence interval [CI] = 1.86–14.92). Median time to relapse was 1 month after the ICI start (range = 0.4–6 months). No relapse occurred after 58 years. The neuroradiological activity meta‐analysis was conducted on 36 pwMS (median magnetic resonance imaging [MRI] follow‐up = 0.75–1.85 years, 41.94 patient‐years). The pooled new MRI lesion rate was 24.9 per 100 patient‐years (95% CI = 10.9–47.3), with median time to new MRI lesions of 3 months (range = 1–6 months). In 80% of cases, disease‐modifying treatment (DMT) was suspended at ICI initiation.

Conclusions

We found a low relapse rate in pwMS following ICI treatment, with no events in older pwMS. The risk of neuroradiological activity appears higher, but mainly occurs in pwMS who discontinued DMT. All events occurred within the first 6 months of ICI therapy. These conclusions are based on small observational studies, highlighting the urgent need for further research on this topic.

INTRODUCTION

Multiple sclerosis (MS) is a chronic inflammatory, demyelinating disease of the central nervous system. In recent years, a demographic shift has occurred in people with MS (pwMS), with an increasing prevalence of elderly individuals affected by the disease [1]. Epidemiological studies indicate that the prevalence of MS peaks between the ages of 45–64 years in the USA [2] and Europe [3]. Concurrently, there is a rising incidence of late onset MS, defined by the onset of MS after the age of 50 years [4, 5].

The management of elderly pwMS introduces new challenges, including the possible occurrence of age‐related comorbidities [1]. One significant comorbidity is cancer, as pwMS show an increased risk of developing neoplasms after the age of 50 years [6, 7], as occurs in the general population. In addition, a recent meta‐analysis found that greater age is associated with increased malignancy incidence in pwMS treated with depletive disease‐modifying treatment (DMT), especially after the age of 45 years [7]. The demographic shift, therefore, makes it increasingly common to manage cancer in pwMS.

Recently, immune checkpoint inhibitors (ICIs) have demonstrated significant prognostic impact in oncology, becoming a widely used therapeutic option for a broad range of tumors [8]. Consequently, many individuals with MS may require such treatments.

ICIs, which target pathways like CTLA‐4, PD‐1, and PD‐L1, enhance antitumor immunity by blocking inhibitory signals that restrict T‐cell activation [8]. On the other hand, the modulation of immune checkpoints leads to a significant risk of immune‐related adverse events (irAEs) [9], with 1%–2% represented by neurological events [10]. Similarly, ICI treatment in people with a pre‐existing autoimmune disease may increase the risk of disease flares [11].

Regarding MS, the available data on the potential impact of ICIs on the course of the disease are limited. Few reports have documented disease reactivation—some severe—in pwMS treated with ICIs [8, 12, 13, 14, 15, 16, 17]. Additionally, there have been cases of newly diagnosed MS following ICI treatment [18] or conversion from radiologically isolated syndrome to MS [19]. Moreover, ICI administration has been associated with disease worsening in an animal model of MS [8]. However, pwMS were excluded from ICI clinical trials, and observational studies on ICI treatment in patients with autoimmune diseases have included only a few MS cases. Consequently, the more robust data come from recent small‐scale observational studies.

In the present paper, we systematically review the current literature concerning the possible impact of ICIs on MS course. In addition, we conducted a meta‐analysis to synthesize the currently available evidence on the effect of ICI treatment on MS course.

METHODS

Search strategy

We carried out a systematic review of the literature following PRISMA 2020 guidelines [20]. We searched papers concerning the possible impact of ICIs on MS disease course on Google Scholar and PubMed. We searched for peer‐reviewed papers published in the English language until early May 2024. For the Google Scholar search, we used the following strategy: “intitle”: “multiple sclerosis” AND (“immune checkpoint inhibitors” OR “checkpoint inhibitors”). The search identified 123 papers. For PubMed search, we used the following strategy: “multiple sclerosis” AND (“immune checkpoint inhibitors” OR “checkpoint inhibitors”). The search identified 39 papers. Paper titles and abstracts were screened by authors S.G. and L.S., including all the studies potentially relevant to the search topic. Among the 123 studies screened on Google Scholar, 10 were identified as relevant and underwent a full‐text assessment. Among them, we retained six observational studies [8, 10, 15, 16, 17, 21] and three case reports [12, 13, 22]. One paper was excluded for not being an original work; another one was excluded for being the abstract of a paper subsequently published and included among the four observational studies. We excluded a case report because it had not undergone peer review. Among the 39 papers screened on PubMed, we identified four papers as possibly relevant to the topic. Among them, one was excluded for not being an original work, whereas the other three were observational studies already included following the search on Google Scholar.

We also analyzed the bibliographies of the included studies to check for possible papers that our search may have missed. In this way, we included another case report relevant to the search topic [14].

Data extraction

We conducted a meta‐analysis of the current evidence, focusing on the observational studies reporting clinical and neuroradiological outcomes of pwMS treated with ICIs. Among the observational studies selected by our search strategy, we excluded the paper by Garcia et al. [15] from our meta‐analysis because its methodology differed significantly from that of the other observational studies. Whereas the five included studies focus on cohorts of pwMS treated with ICIs, the study by Garcia et al. is based on the totality of adverse events reported to the US Food and Drug Administration (FDA). Without knowing the number of MS patients treated, this study cannot provide incidence rates of neuroradiological and clinical activity. We extracted the following data from the included papers: sample size, study follow‐up, number of pwMS experiencing new magnetic resonance imaging (MRI) activity and relapses following treatment with ICIs, number of pwMS on DMT at ICI initiation, proportion of relapsing–remitting MS (RRMS), sex, age, disease duration, disease course, and data concerning previous disease activity.

Statistics

We calculated the event rates per patient‐year for relapses and new MRI lesions; patient‐years were derived by multiplying the sample size by the study follow‐up duration expressed in years. Effect size (ES) was calculated by logit‐transforming the event rates per patient‐year. In cases of zero events, we applied a proportional correction by adding 1 divided by the sample size of the study (1/N) to prevent undefined logit values [23]. The pooled ES, along with its 95% confidence interval (CI), was estimated through random‐effects models using the DerSimonian and Laird method. Random‐effects models were employed to prevent the overrepresentation of large samples and to capture the uncertainty arising from heterogeneity among studies. Between‐study heterogeneity and inconsistency were assessed using Cochran Q statistic and the I ² index, respectively. As previously published, we classified I ² ≤ 25% as marginal, 25%–75% as moderate, and ≥75% as substantial inconsistency [24]. We assessed the risk of publication bias using Kendall τ and Egger Z tests for asymmetry. Meta‐regression equations were used to identify variables possibly influencing the pooled ES. The normality assumption was verified for all variables included in the models; if the normality assumption was violated, the variable was log‐transformed. Two‐tailed p‐values < 0.05 were considered significant. We used the JASP software version 0.18.3.0 for statistical analysis.

CLINICAL DISEASE ACTIVITY FOLLOWING ICI TREATMENT

For the meta‐analysis of disease activity, clinical data were available from five studies [8, 10, 16, 17, 21] (Table 1). The pooled cohort of pwMS consisted of 90 pwMS. The median follow‐up ranged from 0.62 to 1.85 years, with a total of 103.74 patient‐years. The mean age ranged from 48 to 67.4 years, and the disease duration ranged from 10 to 26.8 years. Data for malignancy type were available in four studies (n = 51) [8, 10, 16, 21]. The most frequent malignancies were melanoma (45%) and lung cancer (33.3%), with other malignancies accounting for 21.7% (n = 3 breast cancer, n = 2 urothelial cancer, n = 1 cutaneous squamous cell carcinoma, n = 1 tongue squamous cell cancer, n = 1 colorectal cancer, n = 1 Merkel carcinoma, n = 1 sarcoma, n = 1 Hodgkin lymphoma). In one case, the treatment with ICIs (pembrolizumab) was used to treat progressive multifocal leukoencephalopathy [10]. Concerning the ICIs administered, data were available in only three studies [8, 10, 21]. Pembrolizumab was the most used ICI (56.1%), followed by nivolumab (31.7%), and ipilimumab (19.51%); atezolizumab and durvalumab were used in only two cases [8]. The percentage of pwMS on DMT before ICI treatment ranged from 24.2% to 77.8%, and 18.7% to 35.7% of pwMS continued DMT during ICI administration. Additionally, the percentage of DMT discontinuation at the start of ICI treatment ranged from 38.5% to 66.7%.

TABLE 1.

Summary of the studies included in the meta‐analysis.

First author (year)	Sample size, N	Patients with MRI data, n	Age, years	Sex ratio, male‐to‐female	Disease duration, years	Patients with progressive disease, %	Patients on DMT prior ICI treatment, %	Patients on DMT during ICI treatment, %	Percentage of DMT discontinuation, %	Clinical follow‐up, years	MRI follow‐up, years	Patients with relapse, n	Patients with new MRI lesions, n
Chavaz (2023) [16]	11	nr	51	0.375	10	18	45.5	nr	nr	0.62	na	1	na
Conway (2023) [21]	16	12	67.4	0.3	26.8	25	56.3	18.7	66.7	1.85	nr	0	0
Androdias (2024) [8]	18	18	48	0.5	16	39	77.8	35.7	64.3	0.96	0.75	1	3
Nylander (2024) [10]	7	6	55.4	0.43	18.2	42.9	57.1	28.6	50	1.27	1.04	0	2
Quinn (2024) [17]	38	nr	66	0.29	20	10.5	34.2	21.1	38.5	1.08	na	1	na

Note: Data are reported as median or mean, consistently with what is reported in the study.

Abbreviations: DMT, disease‐modifying treatment; ICI, immune checkpoint inhibitor; MRI, magnetic resonance imaging; na, not applicable; nr, not reported.

Among pwMS included, three experienced a relapse, leading to a pooled ES (rate of relapse event) of 5.45 per 100 patient‐years (95% CI = 1.86–14.92, k = 5; Figure 1a). We found marginal inconsistency among studies (Q = 2.88, p = 0.57; I ² = 0%), with no significant publication bias (Kendall τ = 0.3, p = 1; Egger Z = −5.71, p = 0.71).

FIGURE 1.

Forest plots of the relapse rate (a) and the new magnetic resonance imaging lesion rate (b). The relapse rate is represented as the event rate per patient‐year. The 95% confidence interval (CI) is provided in parentheses for each study and for the pooled effect size. In the graphs, the squares represent the event rate of individual studies, with sizes proportional to the sample sizes and horizontal bars representing the 95% CI. The pooled effect size is represented by the red diamond, with the width representing the 95% CI. The dashed vertical line indicates the pooled effect size on the x‐axis. RE, random effect.

Meta‐regression analyses were conducted to assess the possible impact of age, disease duration, proportion of RRMS, proportion of pwMS on DMT during ICI treatment, and proportion of pwMS discontinuing DMT at ICI initiation. Except for a trend toward a negative correlation between every 10 years of disease duration and the rate of relapse events (β = −2.187, p = 0.099), we found no correlation between the tested variables and the rate of relapses (not shown). This may at least partially be explained by the limited number of available studies. Because single‐patient data concerning previous disease activity and ongoing DMT treatment were available in only one [8] and two studies [8, 10], respectively, we were not able to check any possible correlation between these variables and relapse rate.

MRI DISEASE ACTIVITY FOLLOWING ICI TREATMENT

Among the five studies included in the meta‐analysis, only three provided neurological data and were analyzed for MRI disease activity. These studies accounted for 36 pwMS. The median MRI follow‐up ranged from 0.75 to 1.85 years, with a total of 41.94 patient‐years. Compared to the total population, the MRI cohort showed similar age, disease duration, malignancy types, and ICIs administered. Both the percentage of pwMS on DMT before ICI treatment and the percentage of DMT discontinuation at the start of ICI treatment were higher compared to the total population (57.1%–77.8% and 50%–66.7%, respectively). Differences between total population and MRI cohort are provided in Table S1.

Among the pwMS included, five developed new MRI lesions, leading to a pooled ES (rate of new MRI lesions) of 24.9 per 100 patient‐years (95% CI = 10.9–47.3, k = 3; Figure 1b). We found marginal inconsistency among studies (Q = 2.02, p = 0.365; I ² = 0.86%) with no significant publication bias (Kendall τ = −0.33, p = 1.0; Egger Z = 0.19, p = 0.7). Meta‐regression analyses were conducted to assess the possible impact of age, disease duration, proportion of RRMS, proportion of pwMS on DMT during ICI treatment, and proportion of pwMS discontinuing DMT at ICI initiation. We found no correlation between the tested variables and the rate of new MRI lesions (not shown).

Single‐patient data were available from two studies, accounting for the total of five pwMS experiencing MRI disease activity following ICI initiation [8, 10]. Among them, four were female, with a median age of 53 years (range = 26–66 years). The median time to new MRI lesions was 3 months, with all events occurring within the first 6 months of ICI treatment, mirroring the timing observed for clinical relapses (range = 1–6 months). In 80% of pwMS with new MRI lesions, DMT was suspended at ICI initiation. Notably, in two cases (40%), the suspended DMT was natalizumab, known for its risk of rebound following discontinuation. The most frequent ICI used in pwMS with new MRI lesions was pembrolizumab (80%), whereas another patient received a combination of nivolumab and ipilimumab.

DISCUSSION

The need to manage malignancies in pwMS is increasing, alongside the general aging of this population [1]. Because the introduction of ICIs has dramatically improved the prognosis of many cancers, the use of these drugs in pwMS is being increasingly considered. However, given that ICIs are associated with a significant risk of irAEs, concerns exist about their use in individuals with pre‐existing autoimmune diseases, including MS. Prior studies focusing on ICI treatment in people with pre‐existing autoimmunity found that 27%–75% of these patients experienced irAEs or disease flares [10]. However, these studies included only a few pwMS, thus not providing conclusive data on the risk of disease reactivation in MS. On the other hand, MS relapses following ICI treatment have been described in different case reports [12, 13, 14] and in a study that analyzed the totality of the adverse events of ICIs reported to the FDA [15]. In this case, the authors identified a total of 14 MS relapses, with only eight cases occurring in individuals with a documented history of MS. Notably, in this paper, three relapses were severe, and two of them led to death. Due to the nature of the study, however, this paper did not provide the total number of pwMS treated with ICIs, preventing an estimation of the relapse rate. Furthermore, there is a lack of clinical and MRI data, with a possible overassessment or misdiagnosis of the clinical events. Overall, although such studies draw attention to the possibility of disease relapses during ICI treatment, they are intrinsically affected by reporting bias, risking the perception of such occurrences as common.

In recent years, a few observational studies on pwMS treated with ICIs were published [8, 10, 16, 17, 21]. These studies, despite their limitations, provide valuable insights by assessing the incidence of adverse events within a cohort of pwMS treated with ICIs. However, all the available studies so far are limited by their retrospective nature and the small sample sizes, reducing the possibility of generalizing results. Against this background, we conducted a systematic review and meta‐analysis of the literature, to synthesize the current evidence regarding the potential impact of ICIs on disease activity in MS.

Concerning the possibility of experiencing a clinical relapse following ICI treatment in pwMS, we observed 5.45 relapse events per 100 patient‐years. Compared to other autoimmune diseases, the risk of relapse in MS appears lower. Furthermore, analyzing single‐patient data, we found that 60% of pwMS experiencing a relapse were not on DMT at ICI initiation. Despite insufficient data on DMT treatment in pwMS not experiencing a relapse, this finding suggests a potential protective role of DMT against ICI‐induced relapses. Our meta‐regression found no correlation between patient characteristics and the relapse rate, likely due to the small number of studies included and their small sample sizes. When analyzing the totality of pwMS experiencing a relapse after ICI initiation, we found no patients aged >58 years. This may suggest a possible protective role of age against ICI‐induced disease activity, as observed in the general MS population [1].

When analyzing MRI disease activity, we found 24.9 new MRI lesion events per 100 patient‐years, indicating a much higher rate of neuroradiological activity compared to clinical disease activity. However, this finding should be interpreted cautiously. Only three studies reported MRI outcomes, significantly impacting the quality of our analysis. Additionally, 80% of pwMS with new MRI lesions had suspended DMT at ICI initiation, and notably, 40% had suspended natalizumab, posing a concrete risk of disease rebound. This likely led to an overestimation of the new MRI lesion event rate in our analysis. Conversely, only 20% of pwMS with new MRI lesions were on DMT, suggesting also in this case a protective effect of DMT against ICI‐induced MS disease activity.

As observed in other autoimmune diseases [25, 26, 27], all disease activity events—both clinical and neuroradiological—occurred in the first 6 months of ICI treatment. This highlights the importance of clinical and MRI monitoring of pwMS, especially within the initial months of ICI administration. The effect of ICI rechallenging following MS reactivation remains an open question due to the lack of data. This strategy was reported in two cases [8, 14], with only one of them resulting in a second relapse in the following months [14].

The present study should be read in the light of its limitations. First, the small number of studies available and their small size strongly impact the estimation of the relapse and new MRI lesion rates of our meta‐analysis and prevent the identification of possible variables that may influence the likelihood of disease activity. A very important limitation arises also from the inclusion of a study that has not yet undergone peer review, which accounts for a large portion of the patients in our meta‐analysis (38 patients) [17]. Although this may result in significant bias, we decided to include it due to the limited number of studies on the topic. Moreover, all included studies are retrospective. As observed by previous authors, these studies may suffer from treatment selection bias, as a higher number of pwMS with inactive disease might have been selected for treatment with ICIs [21]. This may be evidenced by the low percentage of pwMS on DMT before ICI treatment in most of the studies. Furthermore, the use of a correction for handling zero‐event studies may lead to an overestimation of event rates, particularly in smaller studies, thereby potentially inflating the pooled effect size. Finally, a very important limitation is the complete absence of data on the potential impact of ICIs on disease progression.

Despite these limitations, the present meta‐analysis suggests that the administration of ICIs in pwMS with cancer results in a low and manageable risk of relapse. Therefore, given the significant prognostic impact on malignancies, their use should be considered in pwMS. Strategies to minimize the risk of disease reactivation could include the continuation of DMT during ICI administration and intensive monitoring of disease activity during the first 6 months of treatment.

However, larger, multicentric, long‐term, prospective studies are needed to more accurately estimate the risk of disease activity and, most importantly, to provide data on the potential impact on future disease progression.

AUTHOR CONTRIBUTIONS

Stefano Gelibter: Conceptualization; methodology; data curation; formal analysis; writing – original draft; writing – review and editing. Lorenzo Saraceno: Conceptualization; writing – original draft; writing – review and editing. Emanuela Susani: Conceptualization; writing – original draft; writing – review and editing. Fiammetta Pirro: Writing – review and editing. Maria Sessa: Writing – review and editing; supervision; conceptualization. Alessandra Protti: Methodology; writing – review and editing; conceptualization; supervision.

CONFLICT OF INTEREST STATEMENT

The authors declare no conflicting interests concerning the present study.

Supporting information

Table S1.

Gelibter S, Saraceno L, Susani E, Pirro F, Sessa M, Protti A. Do immune checkpoint inhibitors affect the course of multiple sclerosis? A systematic review and meta‐analysis. Eur J Neurol. 2025;32:e16547. doi: 10.1111/ene.16547

DATA AVAILABILITY STATEMENT

The data that support the findings of this study are available from the corresponding author upon reasonable request.