The Preferred Reporting Items for Systematic Reviews and Meta-analyses (PRISMA) (19) statement, the Joanna Briggs recommendations for systematic reviews of observational epidemiological studies reporting prevalence and cumulative incidence data (20), and the Meta-analysis of Observational Studies in Epidemiology (MOOSE) indications (21) were followed to conduct this systematic review with meta-analysis. The protocol was registered in the International Prospective Register of Systematic Reviews (PROSPERO) with the registration number CRD42022380463. As this research does not involve the direct recruitment of subjects, the local ethics committee’s approval and the written consent form were not required.

A systematic literature search using four electronic academic.

databases—PubMed/MEDLINE, Scopus, SciVerse ScienceDirect, Web of Science — was performed. The following search terms were used: (Employ* OR unemploy* OR occupation* OR “work” OR vocation* OR “workplace” OR “workforce” OR “labour force” OR “labor force” OR Career* OR Job* OR “worker” OR “fitness for work”) AND (“Multiple sclerosis” OR “Demyelinating Autoimmune Diseases” OR “Demyelinating Autoimmune Disorders” OR “Clinically Isolated Syndrome” OR “Demyelinating”) AND (“SARS-CoV-2” OR “COVID-19” OR “Coronavirus disease 2019”). The detailed search strategy is presented in the Supplementary Table S1. The search results were exported and managed in Mendeley 1.19.8 (Elsevier, New York, United States).

Two independent and previously trained investigators (BV and AR) carried out the selection of the studies, one being blind to the decision of the other. In case of conflicting views, a senior investigator (GD) was consulted to promote a discussion and reach a consensus. After removing duplicate entries, we performed a first screening of titles and abstracts to assess their potential relevance and remove those off-topic. Then, the full manuscripts were carefully read to determine their final eligibility. The inclusion criteria were framed according to the PICOS acronym. We included any peer-reviewed original article that enrolled adult patients with the diagnosis of MS and assessed any occupational variable during the COVID-19 pandemic. The studies could have chosen any comparator or intervention. There were no time limits and no language restrictions. We accepted cross-sectional studies, longitudinal studies, and experimental studies, and articles designed as reviews, conference abstracts, letters to the editor, expert opinions, commentaries, case reports, case series, editorials were excluded. We also excluded different published articles that reported the same result from the same study population.

The primary outcomes were the prevalence of unemployment, retirement, and employment status change among people with MS during the COVID-19 pandemic. Other outcomes included the modality and characteristics of work: type of work, full-time work, part-time work, and remote work. We also searched for data from studies that addressed any change in the work status due to the COVID-19 outbreak. The prevalence of all these outcomes was calculated considering the number of events as the numerator and the number of study participants as the denominator. The outcome measures were regarded as categorical variables and reported as percentages.

The data extracted included information about the first author, country of the study, year of publication, sample size, mean age of participants, higher educational attainment (defined as more than 12 years of schooling), gender distribution, study design, mean duration of the disease, MS phenotype (progressive or relapsing–remitting), Expanded Disability Status Scale (EDSS) scores, history of past COVID-19 infection, use of disease-modifying drugs (DMDs) and the prevalence of anxiety or depression. When available, we recorded a description of which DMD they were using as well as the prevalence of PwMS under high-efficacy therapies (22). We considered as “high efficacy therapy” as the immunotherapies recommended by the current guidelines of treatment of MS that have a greater efficacy: natalizumab, ocrelizumab, rituximab, alemtuzumab, siponimod, cladribine, fingolimod. We considered as “other immunosuppressants” the drugs that are not in the first line of treatment of MS but have a “non-specific” immunosuppressive action: cyclophosphamide, methotrexate, azathioprine, mycophenolate mofetil. This decision is in line with the classification used in the studies included in the review. In the case of articles missing essential data, we contacted the corresponding author to obtain more information by e-mail. The study was excluded whenever our attempt to contact failed. When a multicenter study reported the results according to each country, the information was treated as if it came from two different studies. All extracted data were double-checked 1 month after the initial extraction to optimize intra-rater reliability and minimize the risk of bias. The quality assessment was performed with the Critical Appraisal Checklist for Studies Reporting Prevalence Data which was developed and validated by the Joanna Briggs Institute (23). It comprises 9 questions for which researchers can answer “yes,” “no,” “unclear” or “not applicable (NA)” in response to each item. The greater the number of “no” or “uncertain” selected, the greater the risk of bias in each category and in each study. The critical appraisal was carried out considering the variables of interest in our review. This step was also carried out by two independent and previously trained investigators (BV and AM), being a third researcher (GD) always consulted in case of discrepancy.

All data about any occupational outcome were synthesized narratively according to each study. Quantitative data were pooled in a meta-analysis. Any study that determined as an inclusion criterion the need for some specific occupational characteristic (e.g., subjects necessarily employed) was excluded from the meta-analysis due to the clear addition of a selection bias in the calculation of pooled prevalence. We used the random-effects model based on the binomial distribution to calculate the pooled estimates of the prevalence of unemployment, retirement, and remote working among PwMS with their respective confidence intervals (CIs). Potential influences on prevalence estimates were investigated using subgroup analyses and meta-regression. Therefore, we identified a priori potential variables that could be associated with the estimates: age, sex, educational level, disease duration, progressive MS phenotype, EDSS, past COVID-19 infection, DMDs, anxiety, and depression. Potentially statistically significant differences in the effect size of each country were assessed with ANOVA. We assessed the heterogeneity between estimates using the I² statistic and a visual examination of the forest plot. Substantial heterogeneity was considered when I² exceeded 75% (24). To investigate the presence of publication bias, we employed Egger’s linear regression test (25) and examined funnel plots visually. Additionally, we conducted sensitivity analysis, excluding any potential outliers. A p < 0.05 was considered statistically significant. All statistical analyses were performed using STATA/BE 17.0.

Some studies report some associations between employment status and clinical and demographic characteristics. In an American study, PwMS who self-denominated Hispanic, Latinx or Spanish origin had a statistically significantly higher rate of job loss. Moreover, the educational level was also associated with the prevalence of unemployment among PwMS. The proportion of workers who had a university degree and lost their jobs due to COVID-19 was four times lower than that of other workers (14.3% vs. 55.6%, p = 0.012) (65). The employment status of PwMS during the COVID-19 pandemic was also associated with the presence of anxious and depressive symptoms (38). In line with this finding, Altunan et al. showed that the use of dysfunctional strategies of coping was statistically higher among the unemployed group of PwMS (56). Regarding the relationship between the occupational outcomes of PwMS and the vaccination against COVID-19, only two studies provided information. Ciotti et al. found that vaccination status was not statistically associated with employment status (66). In contrast, Abbasi et al. reported that being unemployed is associated with COVID-19 vaccine hesitancy/rejection (36).

The results of the present review indicate that a large proportion of PwMS has experienced job changes due to the COVID-19 pandemic. Pokryszko-Dragan et al. reported that 33% of PwMS faced any work-related problem (50). Among these, 41.7% had their job suspended, 14.6% were fired, 14.6% had some difficulties in turning to remote work, 8.3% had an increased workload and 20.8% feared the infection at the workplace. In line with these findings, Bishop et al. demonstrated that, among participants previously engaged in full-time employment before the onset of the COVID-19 pandemic, 17.4% reported job loss due to the pandemic, while 39.1% reported other alterations in their employment status (65). The most commonly reported changes encompassed a reduction in working hours or shifts, transitioning to remote work, and switching departments or job roles within their respective organizations. Capuano et al. found that only 10.5% of the research subjects continued to work in the same modality as before the lockdown, while 23.9% started working from home (45). Motolese et al. also described that 26.3% had a working hour reduction and only 43.2% kept the usual workplace (46). In a different scenario, an American cross-sectional study revealed that more than half (55.8%) of the PwMS did not have their jobs affected by the pandemic (67). The other participants who had employment somehow changed included 22.7% who reported increased remote work, 11.2% who reported working more than usual, and 11.2% who had to work with children in the house. Zanghì et al. showed that their study participants reported a reduction in the working hours with the lockdown, however, the average pre-pandemic working hours went from 41.3 (SD 4.2) hours per week to an average of 36.2 (SD 3.2) hours per week, indicating that the reduction may not be so pronounced (the authors did not address this data with a statistical test) (42).

There were studies in which no subjects had COVID-19 and others in which the entire sample was diagnosed with the infection. Few studies sought to classify the entire sample according to the type of work performed. In a Polish study, among those who were working at the time of the study, 82.2% were white-collar workers and 17.8% were blue-collar workers (51). In an Italian study, 77.5% of the sample were employees or self-employed, while 22.5% were executives with management roles (42). Arrambide et al. found that healthcare workers with MS had a significantly higher incidence of death from COVID-19 than other PwMS (55).

The pooled overall effect size for the prevalence of unemployment was 0.39 (95% CI = 0.32–0.45, I² = 98.63%; Figure 2). There was no statistically significant difference between the pooled estimates of unemployment among the countries (p = 0.14). The country with the highest prevalence was Australia (ES = 0.58, 95% CI = 0.53–0.63), and the country with the lowest effect size was Chile (ES = 0.10, 95% CI = 0.07–0.14). Regarding the prevalence of retirement, the pooled overall effect size was 0.20 (95% CI = 0.15–0.24, I² = 95.66%; Figure 3). The effect sizes also did not vary significantly by country (p = 0.62), with Italy and Chile having the lowest values (ES = 0.10, 95% CI = 0.04–0.17 and ES = 0.10, 95% CI = 0.07–0.13, respectively) and Portugal the highest (ES = 0.29, 95% CI = 0.23–0.34). Considering the proportion of subjects with MS either unemployed or retired, the pooled overall effect size was 0.47 (95% CI = 0.42–0.53, I² = 97.82%; Figure 4).

On meta-regression analysis, unemployment was negatively associated with the use of other immunosuppressants (p = 0.032; Supplementary Figure S1). The pooled prevalence of PwMS who were unemployed or retired was positively associated with the progressive phenotype of the disease (p = 0.017; Figure 5A) and the use of glatiramer acetate (p = 0.004; Supplementary Figure S2), but negatively associated with hospitalization due to COVID-19 (p = 0.008; Figure 5B) and the use of other immunosuppressants (p = 0.032; Supplementary Figure S3), siponimod (p < 0.001; Figure 5C), and cladribine (p = 0.021; Figure 5D). In parallel, the estimate of the prevalence of retirement was negatively associated with the use of high-efficacy therapies (p = 0.004; Figure 5E), fingolimod (p = 0.014; Figure 5F), cladribine (p < 0.001; Supplementary Figure S4) and the presence of depressive symptoms (p < 0.001; Supplementary Figure S5). The inspection of the funnel plot and the results of the Egger’s test showed a publication bias in the pooled estimates of unemployment (p = 0.02; Supplementary Figure S6) and retirement (p = 0.014; Supplementary Figure S7) but not in the estimate of PwMS who were either unemployed or retired (p = 0.509; Supplementary Figure S8). The sensitivity analysis did not show any other statistically significant results.

The pooled proportion of PwMS that reported any change of the employment status during the COVID-19 pandemic was 0.43 (95% CI = 0.36–0.50, I² = 86.62%; Figure 6) while the pooled prevalence of PwMS who worked remotely during this period was 0.37 (95% CI = 0.15–0.58, I2 = 98.91%; Figure 7). There was no statistically significant difference considering the study origin for the combined estimates of change in employment status (p = 0.30) and remote working (p = 0.37). The change in employment status was negatively associated with the duration of MS (p = 0.03; Figure 8A) but positively associated with the progressive phenotype of the disease (p < 0.001; Figure 8B). There were no statistically significant results in the meta-regression analyses for the proportion of PwMS working remotely.

Regarding the workload of the subjects included in the studies, the estimates of the effect sizes of the people working part-time and full-time were 0.10 (95% CI = 0.08–0.12, I² = 64.15%; Supplementary Figure S9) and 0.34 (95% CI = 0.25–0.43, I² = 96.1%; Supplementary Figure S10). The pooled prevalence of PwMS who were working part-time was positively influenced by the older average age of the individuals (p = 0.034; Figure 8C), the diagnosis of COVID-19 (p = 0.005; Supplementary Figure S11), the use of fingolimod (p = 0.032; Supplementary Figure S12), alemtuzumab (p = 0.04; Supplementary Figure S13) and cladribine (p = 0.02; Supplementary Figure S14), and the presence of anxious (p = 0.001; Figure 8D) and depressive symptoms (p = 0.003; Figure 8E). In contrast, the prevalence of PwMS who had a full-time job was associated with younger mean age (p < 0.001; Supplementary Figure S15), a shorter disease duration (p < 0.001; Supplementary Figure S16), the progressive form of MS (p < 0.001; Supplementary Figure S17), the use of any DMD (p = 0.038; Figure 8F), and the use of glatiramer acetate (p = 0.001; Supplementary Figure S18), alemtuzumab (p = 0.035; Supplementary Figure S19) and natalizumab (p = 0.011; Supplementary Figure S20). There were publication biases in the results of the meta-analysis of the estimates of PwMS working part-time (p = 0.005; Supplementary Figure S21), full-time (p = 0.005; Supplementary Figure S22) and that had a change in the employment status (p = 0.0059; Supplementary Figure S23).