Background

Demographic, ecological, and politico-economic factors together with comorbidities have influenced the severity of Coronavirus disease 2019 (COVID-19) worldwide [1]. Among others, this pandemic has also raised health concerns related to people with multiple sclerosis (PwMS) in the acute phase of the outbreak. These people have been considered at higher risk to contract SARS-CoV-2 infection and have a more severe COVID-19 course due to their clinical disability, comorbidities, disease-associated immune alterations and the use of disease modifying therapies (DMTs) [2, 3].

In the general population more than 6,500,000 deaths have been reported in the world [4]. Data collected during the pandemic suggest that COVID-19 may determine acute detrimental immunologic, metabolic, and thromboembolic abnormalities. Furthermore, long-term persistence of neurologic and neuropsychiatric manifestations was also described in COVID-19 survivors [5] and was associated with pro-inflammatory unbalance [6], suggesting that such immunologic changes may persist long after the resolution of the infection.

Iaffaldano et al. [7] have recently observed that in PwMS presence of comorbidities, female sex and younger age may increase the risk to develop a COVID infection by approximately 70%, 25% and 10% respectively. Furthermore, PwMS who developed COVID-19 were more frequently treated during the course of the disease with high efficacy DMT or moderate efficacy DMTs followed by high efficacy DMT strategies (18.23% vs 15.60%, 30.17% vs 23.11%).

Among PwMS, older age, male sex, concomitant comorbidities, higher disability, ongoing anti CD-20 therapy and recent methylprednisolone administration were associated with a worse prognosis [3, 8,9,10,11]. These findings are in line with those obtained from large-scale cohorts in Western Countries which suggest that in the general population risk factors for severe COVID-19 disease course are older age, male sex, and concomitant comorbidities [12, 13]. The definition of such demographic and clinical features are helpful for the identification of those patients at high-risk of severe COVID-19, but still their utility is limited in terms of preventive interventions as they are non-modifiable traits. In contrast, studies on the general population have highlighted that also modifiable factors such as higher body mass index (BMI), higher glycated hemoglobin and smoking [14], are associated with a worse outcome, opening the way to tailored preventive strategies.

So far, no study has tried to identify lifestyle and other modifiable variables associated with an increased risk of contracting SARS-CoV-2 infection in PwMS. Therefore, in this case–control study we focused on the identification of anthropometric features, life-style habits, working activities and living environment conditions associated with COVID-19 in PwMS. Clinical data including pre-morbid disability, treatment, and comorbidities were also evaluated.

Methods

Study population, study setting, and data collection

We enrolled patients afferent to the Multiple Sclerosis Center of the IRCCS San Raffaele Hospital, who had a diagnosis of MS according to 2017 McDonald Criteria [15]. Data were retrospectively collected in PwMS who had a confirmed COVID-19 infection between March 2020 and March 2021 (MS-COVID). Inclusion criteria were: (i) typical clinical manifestations of COVID-19, such as fever, cough and dyspnea [16], plus (ii) having a positive Nucleic Acid Amplification Test (NAAT) for SARS-CoV-2 at the time of the symptoms or a positive IgG or IgM SARS-CoV-2 serology within the three months following typical COVID-19 symptoms. Data collection was obtained by direct patient interview within the first year after infection.

At the same time, we randomly selected a group of PwMS without a history of symptoms suggestive of COVID-19 infection and without COVID-19 diagnosis between March 2020 and March 2021 (MS-NCOVID). MS-COVID and MS-NCOVID patients were 1:2 matched by age (± 4 years), sex, expanded disability status scale (EDSS, ± 0.5), disease duration (± 3 years), and line of treatment (first or second line). First line treatments were glatiramer acetate, interferon, teriflunomide, dimethyl fumarate, and azathioprine, while second line treatment were fingolimod, natalizumab, cladribile, rituximab, ocrelizumab or alemtuzumab. As per historical moment, all participants were not vaccinated against SARS-CoV-2.

Demographic, anthropometric, clinical, laboratory, and life-style features

In both groups, we retrospectively collected demographic and anthropometric features, clinical data, laboratory tests, working activity, living condition, and life-style habits.

Demographic variables were age and sex, while anthropometric variables were height, weight, and BMI. Clinical variables included: Expanded Disability Status Scale (EDSS) [17], MS clinical phenotype (relapsing–remitting [RR], primary progressive [PP], secondary progressive [SP]), disease duration, time from MS diagnosis, number of comorbidities and ongoing treatment with DMTs and, if so, administration of first- or second-line therapies. Laboratory tests were performed in the six months before COVID-19, according to the routine monitoring of these patients for DMT, or before data collection for MS-NCOVID and included vitamin D levels (ng/ml). Classes of working activities were: (1) no external contacts (unemployed, telework or working alone on workplace), (2) work requiring direct external contact (with colleagues or public), (3) health-care work. Living conditions were assessed by the evaluation of number of cohabitants, presence and number of school-aged children. Life-style behaviours such as smoke (never smoker, previous smoker, current smoker) were also collected.

Statistical analysis

For all statistical analyses, the significant level was 0.05. All statistical analyses were performed using R 3.5.0 (http://www.R-project.org/).

Between-group comparison of demographic, anthropometric, clinical, laboratory, and life-style features

Continuous variables were reported as median and interquartile range, while categorical variables as absolute and relative frequencies. Disease duration was computed at COVID-19 onset for the MS-COVID group and at the time of the interview for the MS-NCOVID group. Single DMTs in the two groups were reported only descriptively.

Logistic regression analysis for predicting MS-COVID vs MS-NCOVID status

Univariate and multiple logistic regression analyses were employed to evaluate the association between patients’ features and SARS-CoV-2 infection. The features resulting matched between the MS-COVID and MS-NCOVID groups were not included in the analysis. In the univariate analysis, p-values were adjusted with Bonferroni’s correction to account for multiple testing. In the multiple regression analysis, the final model was obtained with a backward procedure of variable selection.

Post-hoc Bayesian network and association analysis

A Bayesian Network model was estimated considering the same features evaluated in the logistic regression analysis, to explore how variables are associated with COVID-19 infection accounting for their complex relationships. Since this analysis required the usage of only categorical variables, numerical variables were categorized either with clinical cut-off or quartiles (number of comorbidities: 0, 1, ≥ 2; Number of cohabitants: 0, 1, 2, ≥ 3; BMI: underweight, normal, overweight; Vitamin D levels: < 23, ≥ 23 and < 31, ≥ 31 and < 39, ≥ 39). The BN was estimated using the Tabu greedy search (TABU) algorithm with Akaike information criterion (AIC) score function. For variables resulting associated with COVID-19 infection, we explored the association with other patients’ features with Fisher’s exact test or Mann–Whitney U test.

Results

Between-group comparison of matched features in MS-COVID and MS-NCOVID patients

As expected, MS-COVID and MS-NCOVID patients were similar in terms of a priori matched variables, specifically: age, sex distribution, disease duration, EDSS, disease phenotype and line of DMTs (Table 1). MS-COVID and MS-NCOVID patients were mainly female (66% and 65% respectively), and the median age was of 41 and 45 years. The median disease duration was relatively long in both groups (10 and 12 years) and disability was usually mild (median EDSS 1.5 and 2.0). The majority of patients had a relapsing–remitting disease (95% in both groups) and were treated with first-line DMTs (56–52%). Detailed specifics of ongoing DMT are summarized in Table 2.

Table 1 Demographic and clinical features (non-modifiable factors) of MS patients with and without history of COVID-19
Full size table
Table 2 Ongoing treatment in the study population
Full size table

Associations between risk factors and SARS-CoV-2 infection

Table 3 reports the descriptive statistics of anthropometric features, laboratory variables, working activities, living conditions and life-style habits in the two groups. At univariate logistic regression analysis (Table 3), higher vitamin D levels resulted protective for COVID-19 disease (odds ratio [OR]; 95% confidence interval, [CI]: 0.94 [0.91;0.95], adjusted p value < 0.0001). In contrast, the number of cohabitants (OR [95% CI]: 1.36 [1.53;1.61], adjusted p value = 0.004) and the working activity during the pandemic (jobs with direct external contact: OR [95% CI]: 2.34 [1.53;3.61], adjusted p-value = 0.0009; healthcare jobs: OR [95% CI]: 3.55 [1.75;7.27], adjusted p value = 0.0040) resulted risk factors for SARS-CoV-2 infection.

Table 3 Univariate associations between features not matched a priori and SARS-CoV-2 infection
Full size table

Smoking status emerged as protective against SARS-CoV-2 infection (OR [95% CI] = 0.33 [0.19;0.56], adjusted p value = 0.0004).

The final multiple logistic regression model confirmed the results of the univariate logistic regression analysis, as reported in Table 4.

Table 4 Multivariate associations between features not matched a priori and SARS-CoV-2 infection
Full size table

Bayesian network and association analysis

Given the protective association between active smoking and COVID-19 disclosed by the logistic regression analysis, we estimated a Bayesian Network (BN) model to assess possible complex relationships among the variables analysed, considering the same features evaluated in the logistic regression analysis. Specifically, we found in the BN that a specific working activity was also associated with smoking status, and by a further association analysis this resulted significant (adjusted p = 0.010), as healthcare professionals were usually non-smokers. The BN is shown in Fig. 1, while results of between-group comparisons of non-smokers during the pandemic (i.e., both never smokers and previous smokers) and active smokers are reported in Table 5.

Fig. 1
figure 1

Bayesian Network analysis for exploring the dependence structure among the variables and SARS-CoV-2 infection. Numerical variables were categorized as described in the Statistical methods

Full size image
Table 5 Between-group comparisons of features not matched a priori in patients smokers and non-smokers during the pandemic
Full size table

Discussion

This is a single-center case–control study aimed to evaluate modifiable risk factors for SARS-CoV-2 infection in PwMS. Comorbidities were also included in the analysis to avoid their nuisance effects in the interpretation of data. In our analysis, at multiple logistic regression, higher levels of vitamin D (OR 0.93, p < 0.0001) emerged as protective factors against SARS-CoV-2 infection. In contrast, higher number of cohabitants (OR 1.26, p = 0.02) and works requiring direct external contact (OR 2.61, p = 0.0002) or in the healthcare sector (OR 3.73, p = 0.0019) resulted risk factors for COVID-19 disease. In addition to higher vitamin D levels, also active smoking status (OR 0.27, p < 0.0001) emerged as protective factors against SARS-CoV-2 infection.

In the general population, other studies found that people with 25OHD levels ≥ 30 ng/ml had lower risk of SARS-CoV2 infection, lower risk of severe COVID-19 and lower COVID-19 mortality than 25OHD-deficient patients [18,19,20]. Vitamin D exerts immunoregulatory properties on both innate and adaptive immune systems. Its deficit is associated with increased autoimmunity as well as an increased susceptibility to infection [21], possibly explaining why lower vitamin D levels confer a higher risk of SARS-CoV-2 infection even in the general population. Moreover, there is established evidence that vitamin D deficiency is relevant for MS pathogenesis, since it was consistently found as a significant risk factor for the development of the disease [22]. More recently, low levels of vitamin D were also reported as a negative prognostic factor in patients with MS, given their association with more severe MRI activity, cognitive impairment and evolution to secondary progressive disease course [22,23,24]. These observations strengthen the indication of maintaining 25(OH)D levels among PwMS of at least 30 ng/ml (75 nmol/L), and preferably 40–60 ng/mL (100–150 nmol/L) [25], especially during the ongoing pandemic.

Other studies showed that housing conditions, such as overcrowded housing, are an important risk factor for SARS-CoV-2 infection [1, 26]. Household exposure can be a relevant risk factor for SARS-CoV-2 infection in patients who rarely leave their home. In line with this, a study demonstrated that the odd for COVID-19 seropositivity among cancer patients was related to intrafamiliar exposure [27]. To the best of our knowledge, no other study demonstrated that specifically the number of cohabitants also increases the risk in PwMS. This result is particularly challenging if we consider that reduced human interaction related to COVID-19 has a profound impact on organic and mental health [1, 26, 27]. Although it is associated with an increased risk of COVID-19 infection, we still encourage social interactions between PwMS and their relatives, but we recommend strict adherence to COVID-19 preventive measures, such as FFP2 mask wearing and frequent hand sanitizing.

A working activity requiring a direct contact with people outside the family (i.e., colleagues or public, healthcare jobs) resulted a risk factor for COVID-19 disease. In line with this hypothesis, distant working was largely adopted, especially during the first phase of the pandemic. However, many employers are now reducing distant work opportunities. Our results suggest that this policies could be harmful for PwMS, where we should encourage teleworking whenever possible. This is of increasing importance because of the ongoing energetic crisis, as patients may favour public to private transportation, with the former associated with increased COVID-19 [28].

Unexpectedly, we found that active smoking status was associated with a reduced risk of developing COVID-19. Smoking is associated with increased severity of COVID-19 infection in the general population [29,30,31,32,33], but initial studies found it to be negatively associated with the risk of developing COVID-19 in the general population [34,35,36]. Subsequent studies demonstrated that smoking is positively associated with the risk of developing COVID-19 in the general population [29, 31, 32]. Although a consistent answer to this “smoking paradox” is hard to extract from the literature, the protective effect of smoking was hardly criticised [37]. To assess whether other features associated with smoking can explain such a risk reduction in our sample, we applied a Bayesian Network analysis which found an association between non-smoking status and working in the healthcare sector, which was, as expected [38], strongly associated in the multiple logistic regression with the risk of SARS-CoV-2 infection (OR [95% CI]: 3.73 [1.64;8.70]). Another possible explanation is that smokers may have chronic mild respiratory symptoms, leading to an underestimation of possible COVID-19 related symptoms and therefore a lower rate of COVID-19 diagnosis. Further studies are clearly needed to clarify the association between smoking and COVID-19. Nevertheless, we strongly discourage smoking among PwMS considering its detrimental effect on disease progression, disease activity, atrophy and disability [39].

This study is not without limitations. First, data collection was retrospective. However, data for this study were collected through a direct patients-neurologist’s interview and not only though medical records, which should have increased the accuracy of the information collected. Another limitation is the inclusion of patients infected during the first pandemic wave, but this had the advantage of avoiding the confounding effect of vaccination, which we know may have a different efficacy according to patients’ treatment [40]. Furthermore our study population had relatively mild clinical disability (reflecting our centre’s population). Further studies are needed to establish whether these findings are applicable to patients with more severe levels of clinical disability.

Although MS-NCOVID patients had no history of symptoms suggestive of COVID-19 infection and no COVID-19 diagnosis between March 2020 and March 2021, they were not necessarily tested negative for COVID-19 serological status. Therefore it is possible that some patients may have developed asymptomatic COVID-19 and have been misallocated in the MS-NCOVID group.

Even though patients were matched for line of treatment, the frequency of ongoing anti CD-20 therapy was higher in MS-COVID group (Table 2). Anti CD-20 therapies are associated with more severe COVID-19 [2, 3], but were not identified as specific non-modifiable risk factor for COVID-19 [7].

In conclusion, our study demonstrated that several modifiable factors can favour or contrast SARS-CoV-2 infection in PwMS, and may help clinicians to protect those patients with a priori risk factors of severe COVID-19 severity. Pursuing an educational policy towards employers aimed at emphasizing the community and social importance of telework as well as the maintenance of protective personal measures and behaviours on public transportation and outdoor places may prevent unnecessarily risk of infection in these patients. Moreover, this study reinforces the importance of careful monitoring and supplementation of vitamin D levels in PwMS.