There is limited information about risk factors or protective factors associated with CI in MS. Among demographic and clinical correlates, aging is clearly associated to a decline of neuropsychological performance [22],while conflicting evidence suggests that male sex is a risk factor for progression of CI [23]. Fatigue, depression and disease duration are weakly correlated with cognitive capacity [24]. CI is more frequent in patients who are in the progressive phase of the disease and the profile and severity of cognitive deficits seem to be mostly driven by age and disability accrual [22]. Several neuroimaging studies have extensively explored MRI correlates of CI in MS, highlighting the relevance of white and grey matter changes and, in particular, brain volume loss [25].

Our results are in line with previous cross-sectional and longitudinal observations pointing to the association of CI with older age at onset, aging and disability [22,26]. Early and appropriate treatment of the disease with disease modifying drugs can therefore represent a key strategy to improve both the physical and cognitive outcome of the subject [27].

As for comorbidities, they have been associated with a worse disease outcome [28]. In particular, cardiovascular risk factors have been associated with brain lesion burden and brain atrophy [29]. In our sample, using the failure of two tests as the cut-off point to define CI, a history of thyroid disease was associated with poorer cognitive performance. While further studies should better analyze this association, thyroid dysfunction represents a relevant, modifiable risk factor that is common in young adults with MS.

Among lifestyle factors, smoking is a well-recognized risk factor for Alzheimer disease and is related to preclinical changes in the brain, higher risk of cognitive decline, and increased risk of dementia [30]. In MS smoking is both a susceptibility risk factor for MS and a prognostic factor, associated with disease progression. Moreover, it has been associated with increased lesion volumes and brain atrophy in multiple sclerosis [31] and CI in one study [32]. Prolonged use of inhaled or ingested street cannabis in patients with MS has been associated with poor performance on cognitive domains commonly affected in this population [33]. In a neuropsychological and f-MRI study [34] cannabis use was associated with compromised cerebral compensatory mechanisms, already faulty in MS. Lastly, regarding alcohol consumption, chronic heavy intake is a well-established cause of brain atrophy and dementia [35] although this association has not been specifically explored in MS.

In our study smoking, cannabis and alcohol intake were not associated with cognitive functioning. In our sample, however, heavy drinkers were not represented. Moreover, we can hypothesize that, on the one hand, our negative results may be due to relatively small sample size and, to the other hand, that diagnosis of a chronic disease itself may have led the patient to modify his/her lifestyle orienting the subject towards healthier inhabits. Finally, we cannot rule out the possibility of under-report in our patients, particularly for alcohol or cannabis use.

As for potential protective factors, in our study higher cognitive reserve stands out as the most consistent, potentially modifiable protective factor. In fact, in all the analyses, better cognitive performance was associated with higher cognitive reserve—expressed as premorbid IQ—which confirms the findings of other studies [36,37]. In a previous work [37] we found that the protective role of cognitive reserve mainly applied to the early stages of the disease and within a hypothetical “threshold” of brain atrophy. However, in our sample of subjects with a mean age of 45 years and a mean disease duration of 11 years this protective role was still evident, highlighting the potential of preventive strategies focusing on intellectual enrichment in this population of patients.

It is noteworthy that in our sample hormonal therapy in female patients resulted to be protective against CI. This was represented by estroprogestinic oral contraceptives whose exact dosage and duration of intake were, however, not recorded. A few randomized clinical trials have reported positive effects of estrogen therapy on cognitive performance [38–42] and the topic has been extensively reviewed elsewhere [43,44]. In animal studies the estrogens has been showed to have task-specific effects on cognitive performance and these effects might be influenced by age and time after loss of ovarian function [45]. One of the proposed mechanisms is that estradiol may enhance performance by increasing cholinergic activity in the hippocampus and cerebral cortex. Furthermore, estradiol effects on hippocampal neurons might be modulated by cholinergic activity, enabling this hormone to produce lasting changes in cortical connectivity and function [45]. These effects might be moderated by aging, that decreases cholinergic activity [45].

Hormonal therapy was not retained in our multivariable analysis, and that—in light of the younger age of the women taking estroprogestinics in our sample—might be due to the more important influence of aging, rather than the hormonal treatment, on cognitive functions.

In our sample, it is intriguing that physical activity in childhood and adolescence emerged as a protective factor in the univariate and remained as a trend in the multivariable analysis. However, in our study we could not document the potential impact of different levels of physical activity, possibly due to the relatively small sample size and the low proportion of patients engaged in moderate or high intensity activity.

There is growing evidence deriving from animal studies [46] and human studies in children and elderly people [47] that physical exercise is connected with greater hippocampal volume, higher white matter integrity and more efficient white matter activity [47]. Thus, physical activity may improve cognitive functioning and memory, possibly enhancing hippocampal function [48]. In MS, physical exercise may be protective against the development of cognitive dysfunction and exert a synergistic effect together with cognitive rehabilitation in patients with established CI [49]. However, there is limited information on the potential role of physical activity in childhood and adolescence. In the general population this has been associated with reduced morbidity [50], including a decreased risk of MS in a recent study [51]. In the context of cognitive functioning, physical exercise early in life might contribute to the building of the subject cognitive reserve. On the other hand, physical activity could be related to higher socio-economic status, an acknowledged proxy of cognitive reserve. Based on our findings, further studies focusing on the pediatric age, using a precise classification of physical exercise in terms of intensity and duration appear to be highly advisable.

Furthermore, we could not demonstrate any protective role of caffeine intake, that in healthy adults has been associated with improvement of alertness, vigilance, attention and reaction time, and, less consistently, memory and higher-order executive functions [52].

Finally, we did not explore the potential impact on cognition of different levels of vitamin D, whereas one MS study has suggested an association between vitamin D supplementation and better cognitive performance [53].

Our study has a few limitations, mainly represented by the cross-sectional design and retrospective assessment of exposures and potential recall bias. Therefore, the results need to be confirmed in a larger sample with a more accurate assessment of exposure to hypothesized factors. Finally, the statistical method used—the stepwise multivariable logistic analysis–although widely used and appealing because of its simplicity may have some limitations. In fact, it may be less efficient especially for small studies, where problems such as over-fitting due to data sparsity and collinearity may arise. Another caveat is the potential bias due to confounding [54] Causal diagrams utilize assumptions regarding the underlying causal relationships between relevant variables to perform confounder selection rather than relying on observed statistical associations. In this respect, the use of multiple models based on multiple potential direct acyclic graphs (DAGs) could be preferred [55]

Since MS is a multifactorial disease deriving from a complex interplay between genetic and environmental factors, further investigation is also required on potential genetic susceptibility to CI [56].

In conclusion, this exploratory study underscores the complexity of physiopathological mechanisms underlying CI in MS and provides a few clues to further research in this area.