Distinct influence of different vascular risk factors on white matter brain lesions in multiple sclerosis

Objective To determine if vascular risk factor (VRF), that is, smoking, arterial hypertension (HT), dyslipidaemia and diabetes, have an effect on multiple sclerosis (MS) pathology as measured by MS typical brain lesions, we have compared brain MRIs from patients with MS with and without VRF age-matched and sex-matched. Methods Brain MRIs from five centres were scored for the presence of Dawson’s fingers (DF) and juxtacortical lesions (JCL). A regression model was built to predict the effect of each individual VRF on DF and JCL, considering age and disease duration. Results 92 MS cases without VRF and 106 MS with one or more VRF (80 ever-smokers, 43 hypertensives, 25 dyslipidaemics and 10 diabetics) were included. Ever-smoking associated with a higher burden of DF (Exp(B)=1.29, 95% CI 1.10 to 1.51, p<0.01) and JCL (Exp(B)=1.38, 95% CI 1.21 to 1.57, p<0.01). No other VRF had an impact on DF. Dyslipidaemia associated with increased JCL (Exp(B)=1.30, 95% CI 1.10 to 1.56, p<0.01) but HT did not associate with any of the outcomes. Conclusions Individual VRF appear to affect MS-specific lesions differently. An increase in MS lesions was mainly seen in smokers; however, this VRF is most likely to be present from onset of MS, and other VRF effects may be partly mitigated by treatment. Our findings support that treating VRF and cessation of smoking may be important in the management of MS.


InTrOduCTIOn
Vascular risk factors (VRF) and comorbidities are associated with worse clinical outcomes in multiple sclerosis (MS). 1 Higher T2 white matter brain lesions (WML) volumes have been seen in patients with MS with concomitant arterial hypertension (HT) and smoking. 2 3 It is not clear whether or not this increase of lesion load is due to a higher burden of MS or vascular WML nor whether the effects of individual VRF on MS WML differ. Several lesion characteristics segregate more clearly with MS, namely juxtacortical U-shaped or S-shaped, ovoid or elongated well-demarcated lesions perpendicular to the wall of lateral ventricle (Dawson's fingers). 4 In the present study, we aimed to determine if there is an excess of these 'MS-specific' lesions on conventional MRI in MS with VRF compared with MS without VRF and to explore the effect of each individual VRF on 'MS-specific' lesions.

MATerIAl And MeThOds study design and cohorts
A multicentre retrospective case-control study was set up to compare MS cases with and without VRF. Anonymised clinical and imaging data were obtained from five MRI in MS (MAGNIMS-www. magnims. eu) network centres: Graz, London, Rome, Barcelona and Amsterdam. Each centre contributed pairs of age-matched and sex-matched MS cases with and without VRF. The imaging and clinical data were collated centrally in Oxford, according to local ethics regulations, quality controlled and blinded for the visual scoring. MS cases were diagnosed according to the 2010 McDonald criteria and exclusion criteria were presence of previous stroke or high risk for cardioembolic stroke (eg, atrial fibrillation, valvular heart disease), other neurological diseases or known non-MS brain lesions.
Demographic and clinical data were provided including: sex, age, disease duration and disease modifying treatment (DMT) at the time of the MRI (yes, no and if latter if they have ever had treatment). The presence or absence of the following VRF was recorded according to previously used definitions: 5 6 (1) HT (ever); (2) hypercholesterolemia (ever); (3) diabetes mellitus (DM); (4) self-reported smoking status-positive if patients ever smoked >10 cigarettes a day for at least 6 months (table 1). T1 and T2-Fluid Attenuated Inversion Recovery (FLAIR)weighted brain imaging sequences obtained from 3 T brain MRIs were provided, with variable sequence protocol depending on the source of the data but of established high quality.

Visual scoring
A single investigator (RG), blinded to the VRF status, performed visual scoring of the brain MRIs. Inter-rater and intrarater variability was tested on a subset (n=100) by repeat blind scoring by RG and two other independent blinded evaluators (MJ, GP). MS-like WML recorded were: (1) the number of Dawson's fingers (defined as well demarcated periventricular lesions, with elongated, ovoid or flame-like shape, perpendicular to the wall of the lateral ventricles), touching the ventricular lining, on axial T2-FLAIR; (2) number of juxtacortical lesions (well demarcated lesions that touched the cortex on both axial T2-FLAIR and T1); (3) U-shaped or S-shaped juxtacortical lesions-a subset of juxtacortical lesions with S or U shapes.

statistical analysis
Inter-rater agreement was measured by free marginal multirater kappa coefficient. 7  Results are presented as the exponentials of the regression coefficients (Exp(B)) and respective CI. A conservative p<0.01 was used to assess significance. All analyses were performed with SPSS V.25.

Data availability
Raw anonymised data are available for appropriate requests. Substantial inter-rater agreement was found for the number of: Dawson's fingers (0.75), juxtacortical lesions (0.72) and U-S shape juxtacortical (0.65) lesions.

The effects of the presence of VrF on lesions
The number of juxtacortical lesions was 27.0% higher in MS with VRF (Exp(B)=1.27, 95% CI 1.11 to 1.44, p<0.001) compared with MS without VRF in a Poisson model adjusting for disease duration (Model 1). No significant differences between these two groups were found regarding the number of U-shaped or S-shaped juxtacortical lesions or of Dawson's fingers (table 1, Model 1).

The effects of the presence of individual VrF on lesions
In a model to predict the number of Dawson's fingers, considering each of the VRF and adjusting for disease duration and age (Model 2), ever smoking associated with a higher burden of Dawson's fingers, while no significant association was found with the other VRF (table 1, Model 2). In a similar model to predict the number of juxtacortical lesions, smoking and dyslipidaemia were associated with a higher burden of juxtacortical lesions and diabetes with a lower number of these lesions. No association was found between HT and juxtacortical lesion number.
None of the individual VRF was associated with the number of U-shaped or S-shaped juxtacortical lesions (table 1, Model 2).
All the studied 'MS-specific' lesions significantly increased with disease duration but interestingly a reverse association was seen with age.
To visually illustrate the effects of smoking alone considering the previous results, we removed patients with smoking plus other VRFs to produce a pure 'MS smokers only' group and a 'MS with non-smoking VRF only' group and compared these groups with 'MS without VRF'.

discussion
This study is novel in studying the effect of different VRF on lesions with features more usually associated with MS, rather than total lesion numbers which does not distinguish between vascular and MS pathology. A key finding of our study is that VRF should not be considered as a composite, as each has a different association with 'MS-specific' lesions. Smoking increased both Dawson's fingers and juxtacortical lesion number, but dyslipidaemia only increased the latter. HT showed no effect on MS-like lesions, while DM was associated with a lower number of juxtacortical lesions.
The documented effect of smoking and hyperlipidaemia on 'MS-specific' lesions supports a direct influence on MS pathology, possibly through promotion of acute inflammation. 2 8-11 In line with our results, HT 12 and DM 13 and total VRF 5 have failed to show an impact on T2 lesion volume in MS, although MS-specific lesion locations were not studied, and individual VRFs, including smoking, were not adjusted for. As expected, disease duration had a significant effect on lesions but less expected age had an inverse effect when disease duration was included in the model. This interesting observation is in line with our postmortem studies showing that the number of active plaques decreases with age at death. 14 DM and age-related cerebrovascular changes may have an unforeseen negative impact on classic MS WML development.
Our results need to be interpreted with caution since the number of patients included in each individual VRF group was low (from 10 in diabetes group to 80 in the smoking group) and no information regarding VRF treatment that may have an impact on MS WML 2 15 was available. Additionally, VRFs may be present for different lengths of time and be managed more or less effectively, for example smoking usually starts in adolescence before the onset of MS and it is not treatable by medication, whereas HT, DM and dyslipidaemia tend to develop later and are treated; thus, smoking may be a higher VRF in our cohort of patients. Finally, our study may have been underpowered to detect an association between the studied VRF and the less commonly found, but possibly more specific, U-shaped or S-shaped juxtacortical lesions.
Despite these limitations, our study highlights that individual VRF should not be grouped together, that they appear to have an effect on MS-like lesions (not just brain lesions per se) and reinforces the importance of supporting patients in giving up smoking. To better understand the effect of each VRF on specific types of WML, future prospective studies matching age across all groups, and risk grading the individual VRFs (eg, pack years for smoking, time and control of individual VRFs) are warranted.