Introduction

Multiple sclerosis (MS) and neuromyelitis optica spectrum disorder (NMOSD) are chronic inflammatory demyelinating diseases of the central nervous system (CNS). Most patients with MS or NMOSD experience weakness, sensory change or visual impairment depending on lesion location.

Previous epidemiologic studies have found that MS is linked to an increased risk of stroke, with an estimated hazard ratio (HR) from 1.3 to 4.0 (online supplemental etable 1)1–7 and cardiovascular diseases, including stroke, seem to contribute more to mortality than expected.8 9 These associations have important clinical implications, as stroke is a leading cause of disability and death globally,10 and stroke in patients with MS will further exacerbate the disease burden. One important limitation of several previous studies is that they failed to control for potential comorbidity confounders, such as hypertension, diabetes mellitus (DM) and dyslipidemia,3 4 6 which were more prevalent in MS than in the general population.11

On the other hand, the risk of stroke has not been determined in population-based studies of NMOSD. Compared with MS, NMOSD has different geographical and racial distributions emphasising the distinctive immunopathogenesis of both conditions, whereas East Asians have a higher prevalence of NMOSD compared with Caucasians, MS is more prevalent in European and North American populations than in Asian populations.12 Although a positive association between MS and cerebral small vessel disease burden was found in a postmortem pathologic study,13 little is known about vascular changes in the brain during NMOSD.

The present study investigated the stroke (ischaemic or haemorrhagic) risk of patients with MS or NMOSD compared with control populations using the Korean National Health Insurance Service (KNHIS) database. We also compared the risk of stroke between MS and NMOSD.

MethodsData source and study design

We used the Korean National Health Insurance (KNHI) database. In Korea, the government operates a health insurance programme for approximately 97% of the entire population.14 Of the KNHI database, claims data that were gathered for information on records of prescription and diagnosis of the International Classification of Diseases 10th revision (ICD-10) are included. The KNHI database also contains demographic factors (eg, age, sex, place of residence and income level) that link to the death registry to qualify the enrolees. Detailed information about the KNHI database has been published elsewhere.15

The requirement for participant’s consent was waived since we used retrospective deidentified data collected in the KNHI database.

MS/NMOSD definition and control matching

We selected patients with MS/NMOSD and their matched controls, who had the same criteria as that reported for the prevalence and incidence of MS/NMOSD in the KNHI database.16 Patients with MS/NMOSD met the following criteria: (1) ≥2 outpatient or hospitalisation claims with the ICD-10 code (G35.0 for MS and G36.0 for NMOSD) and (2) registration in the Rare Intractable Diseases (RID) system with the code for MS/NMOSD. The index date of MS/NMOSD cases was designated as the first date of claims for diagnosis. A special copayment reduction programme in Korea, the RID registration programme, has been in operation to support work-up and treatment for diseases since 2009. Once physicians submit the diagnosis for applicable diseases, the KNHI committee certifies the registration of patients into the programme. A standard for RID registration of MS/NMOSD in Korea is based on the 2005 McDonald criteria for MS17 or 2006 Wingerchuk criteria (2010 - 2015)/revised 2015 NMOSD diagnostic criteria (after 2016).18 19

A total of 2368/2333 MS/NMOSD incident patients aged 19 years or older from 1 January 2010 to 31 December 2017 were identified. Among these patients, those who (1) had a previous history of ischaemic stroke (I63 or I64) or haemorrhagic stroke (I60, I61 or I62) before an MS/NMOSD diagnosis (N=736 for MS, N=587 for NMOSD); (2) had incident ischaemic or haemorrhagic stroke within 1 year from the index date (N=76 for MS, N=54 for NMOSD); (3) had unavailable data for matching variable (N=9 for MS, N=3 for NMOSD) and (4) were deceased at index date (N=1 for MS) were excluded. Finally, 1541/1687 MS/NMOSD patients were included (figure 1).

Figure 1

Flow diagram of study participants. MS, multiple sclerosis; NMOSD, neuromyelitis optica spectrum disorder.

Five controls were selected for each MS/NMOSD case. We matched the year of the index date of NMO/MS cases to the controls who were free of a stroke diagnosis and were alive in that year based on age, sex and the presence of DM, hypertension and dyslipidaemia. Comorbidities were based on claims data before the index date as follows: hypertension (≥1 claim per year for the ICD-10 codes I10‒I11 or≥1 claim per year for prescription of antihypertensive medication), DM (≥1 claim per year under ICD–10 codes E10–14 and ≥1 claim per year for a prescription of antidiabetic medication) and dyslipidaemia (≥1 claim per year under ICD-10 code E78 and ≥1 claim per year for the prescription of a lipid-lowering agent) (figure 1). The index date of controls corresponded to that of matched MS/NMOSD cases. After applying the same exclusion criteria as for the patient population, a total of 16 140 (7705 MS/8435 NMOSD patients, respectively) matched controls were added to the analysis.

Study outcome and follow-up

We identified newly diagnosed stroke as the primary study outcome. Both ischaemic and haemorrhagic stroke cases were defined separately. Newly diagnosed ischaemic stroke was defined as ≥1 hospitalisation record with ICD-10 codes of I63 or I64 and claims for brain imaging including CT or MRI.20 For haemorrhagic stroke, ≥1 hospitalisation record with ICD-10 codes of I60, I61 or I62 and claims for brain imaging including CT or MRI were required.20 To assess the outcome, patients and controls were followed from 1 year after the index date to the date of incident ischaemic stroke, haemorrhagic stroke or death, or until the end of the study period (31 December 2019), whichever came first.

Statistical analysis

Descriptive statistics were presented as numbers (percentages) for categorical variables and means±SD for continuous variables. We compared two groups using the χ2 or Fisher’s exact test for categorical variables, as appropriate, and Student’s t test for continuous variables. The incidence probability of stroke was plotted using the Kaplan-Meier model and the difference between patients with MS or NMOSD and matched controls was assessed by the log-rank test. The association between MS/NMOSD prevalence and the incidence of ischaemic or haemorrhagic stroke was estimated by Cox proportional hazards regression using crude and multivariable models adjusted for sex, age, income and comorbidities (DM, hypertension and dyslipidaemia). The hazard of stroke was also compared between patients with MS and NMOSD using the Cox proportional hazards model. HRs and 95% CI were calculated.

Subgroup analyses were performed by sex, age and the presence of hypertension, DM and dyslipidaemia, and the interactions between the variables were also tested. Forest plots for the HRs and 95% CI by subgroups were established. All statistical analyses were performed using SAS statistical package V.9.4 (SAS Institute, Cary, North Carolina), and p<0.05 was considered statistically significant.

ResultsBaseline characteristics

Matching variables including age, sex and the presence of DM, hypertension and dyslipidaemia were well distributed between MS/NMOSD patients and their matched controls (table 1). In patients with MS and their matched controls, 40% of patients were men and 11.9% of them were aged over 65 years. Among these patients, 26% were comorbid with hypertension, followed by dyslipidaemia (17.9%) and DM (9.7%). In patients with NMOSD and their matched controls, 34.3% of patients were men and 8.7% of them were aged over 65 years. Among these patients, 19.6% were comorbid with dyslipidaemia, followed by hypertension (17.8%) and DM (12.7%). MS/NMOSD patients were more likely to be within the low-income group (p<0.001 for MS and p=0.035 for NMOSD). Mean durations of follow-up after the index date were 4.6/4.4 years for patients with MS/NMOSD and 4.8/4.5 for their matched controls (figure 1).

Table 1

Baseline characteristics of study participants

Stroke risk in patients with MS/NMOSD compared with matched controls

The cumulative incidence curves for MS/NMOSD patients compared with their matched controls are shown in figures 2 and 3. A higher risk of stroke was found in MS (adjusted HR (aHR) = 2.78, 95% CI 1.91 to 4.05) and in NMOSD (aHR 1.69, 95% CI 1.10 to 2.61) compared with their controls (figures 4 and 5). During follow-up, the incidence of stroke (both ischaemic and haemorrhagic) was 5.91 per 1000 person-years for patients with MS and 2.25 for their matched controls (online supplemental etable 2). Among patients with NMOSD, the incidence rate of stroke was 3.67 per 1000 person-years compared with 2.20 for their matched controls (online supplemental etable 2).

Figure 2

Kaplan-Meier curves for the risk of overall (A), ischaemic (B) or haemorrhagic stroke (C) in patients with MS and their matched controls. MS, multiple sclerosis

Figure 3

Kaplan-Meier curves for the risk of overall (A), ischaemic (B) or haemorrhagic (C) stroke in patients with NMOSD and their matched controls. NMOSD, neuromyelitis optica spectrum disorder.

Figure 4

Forest plots portraying the hazard ratios and 95% CIs for the associations between demographics and vascular risk factors, and stroke–overall (A), ischaemic (B) and haemorrhagic stroke (C) incidence in MS. Strata were age (20–39, 40–64 and≥65 years), sex (male or female), diabetes mellitus (no or yes), hypertension (no or yes) and dyslipidaemia (no or yes). All models were adjusted for age, sex, income, and the presence of diabetes mellitus, hypertension or hyperlipidaemia. p was determined using the F-statistic for interaction. MS, multiple sclerosis

Figure 5

Forest plots portraying the hazard ratios and 95% CIs for the associations between demographics and vascular risk factors, and stroke–overall (A), ischaemic (B) and haemorrhagic stroke (C) incidence in NMOSD. Strata were age (20–39, 40–64 and≥65 years), sex (male or female), diabetes mellitus (no or yes), hypertension (no or yes) and dyslipidaemia (no or yes). All models were adjusted for age, sex, income, and the presence of diabetes mellitus, hypertension or hyperlipidaemia. p was determined using the F-statistic for interaction. NMOSD, neuromyelitis optica spectrum disorder.

When we further considered either ischaemic or haemorrhagic stroke separately as a study outcome, a higher risk of ischaemic stroke was found in MS (aHR=2.63, 95% CI 1.70 to 4.07) and marginally in NMOSD (aHR=1.60, 95% CI=0.99 to 2.57) compared with their controls (figures 4 and 5; and online supplemental etable 3). A higher risk of haemorrhagic stroke was also associated with MS (aHR=2.93, 95% CI=1.48–5.80) but not with NMOSD (aHR=1.71, 95% CI=0.62 to 4.72) compared with their controls (figures 4 and 5; and online supplemental etable 4).

The risk for overall, ischaemic and haemorrhagic stroke was not different between patients with MS and NMOSD after controlling for age, sex, income and comorbidities (hypertension, DM and dyslipidaemia) (p=0.216, 0.560 and 0.097, respectively) (table 2).

Table 2

Comparison of stroke risks between patients with MS and NMOSD

Subgroup analysesAge

When stratified by age, the risk of stroke was not different across all age groups in patients with MS (p for interaction=0.861) (figure 4). However, the risk for stroke tended to be more prominent in younger NMOSD patient groups. Patients with NMOSD aged less than 40 (aHR=3.50, 95% CI 1.33 to 9.21) had higher risk of stroke than people aged 40–64 (aHR=1.95, 95% CI 1.12 to 3.39) and older than 65 (aHR=0.57, 95% CI 0.17 to 1.89) in an order with borderline significance (p for interaction=0.069) (figure 5). When we only include ischaemic stroke as a primary end point, a similar non-significant trend was observed in patients with MS/NMOSD (figures 4 and 5).

Sex

Sex-stratified differences were not found in the risk of stroke for patients with MS or NMOSD (p for interaction=0.527 or 0.165, respectively) (figures 4 and 5). However, for NMOSD, an increased risk of stroke was observed in women (aHR=2.15, 95% CI 1.26 to 3.67), which was not found in men (aHR=1.12, 95% CI 0.52 to 2.40) (figure 5). When only ischaemic stroke was considered, the results appeared to be consistent (figures 4 and 5). The difference by sex was not found in the risk of haemorrhagic stroke in patients with MS/NMOSD compared with their controls (p for interaction=0.774 and 0.434, respectively) (figures 4 and 5).

Comorbidities

There was no significant effect on modification by the presence of comorbidities (DM, hypertension or dyslipidaemia) on the risk of stroke in both patient groups (figures 4 and 5). Furthermore, when ischaemic or haemorrhagic stroke was separately considered as the primary outcome, the results according to the presence of comorbidities were also consistently presented.

Discussion

In this study, we explored the risk of stroke in NMOSD compared with the general population and compared the stroke risk between patients with MS and NMOSD within the same population. This population-based cohort study indicated that the patients with MS had 2.78-fold higher risk of overall stroke (HR 2.63 for ischaemic stroke and 2.93 for haemorrhagic stroke) compared with controls matched by age, sex, DM, hypertension and dyslipidaemia. Those with NMOSD were associated with 1.69-fold higher risk of overall stroke compared with matched controls. Increased risk of ischaemic stroke alone was also found in NMOSD (HR 1.60) with marginal statistical significance. Although the risk estimates seemed lower in NMOSD than in MS, the risk did not differ significantly between them.

An increased risk of stroke was found in MS, which was consistent with the results of previous studies.1–7 In addition, the presence of hypertension, DM or dyslipidaemia did not alter the stroke risk, while several studies did not consider comorbidities that might increase the vascular risk. This suggests that the high prevalence of comorbidities in patients with MS6 21 does not fully account for the increased stroke risk. Some potential common risk factors for MS and stroke have been suggested, such as obesity in childhood and early adolescence, the presence of T-cells specific to Epstein-Bar virus in atherosclerotic or demyelinating plaques, low levels of vitamin D, vascular endothelial cell activation, and the excess burden of small vessel disease.13 22 In addition, a recent report suggested that there were several notable shared genes between ischaemic stroke and MS.23 Systemic inflammation during disease course could promote atherosclerosis mostly through endothelial dysfunction.24 The integrity of blood-brain barrier in MS is also disrupted during inflammation, increasing the permeability of the vessels.25 CNS inflammation may provide a nidus for focal haemorrhage. However, there were only few studies that focused on the risk for haemorrhagic stroke in MS.26 In this study, the risks of ischaemic and haemorrhagic stroke were similar in MS. Further studies are needed to confirm change in risk of haemorrhagic stroke and to clarify the exact mechanism of cardiovascular risk and the effect of its management on the long-term health of patients with MS.

The stroke risk in NMOSD was also higher than matched controls. Although the HR for stroke appeared to be higher in MS than in NMOSD, there was no significant difference. The stroke risk of NMOSD has not been evaluated, although there were only a few cases of acute ischaemic or haemorrhagic stroke during intravenous high-dose steroid treatment.27 28 A study using a database of US claims found that comorbidities such as hypertension, DM and cerebrovascular disease were more prevalent in patients with NMOSD compared with non-NMOSD controls.29 Earlier pathologic studies have also demonstrated thickened and hyalinised small blood vessels and perivascular inflammation with immunoglobulin and complement deposition within demyelinating lesions in NMO.30 Small vessel pathologies can lead to both ischaemic and haemorrhagic consequences. In addition, inflammatory vascular markers contributing to endothelial dysfunction increased in NMOSD as in MS during acute relapses.31 32 In this study, the HRs for ischaemic and haemorrhagic stroke were similar in NMOSD. While the risk of ischaemic stroke was marginally significant, the risk of haemorrhagic stroke was not statistically significant, possibly due to the small number of stroke cases. Our findings should be interpreted with caution and further population-based studies are warranted to confirm whether additional stroke risk exists in NMOSD.

There seems to be a sex difference in the association between MS/NMOSD and ischaemic stroke risk, although this was not statistically confirmed possibly due to the small number of events. Previously, sex-related differences in clinical features have been reported in MS (eg, faster disease progression in males)33 and NMOSD (eg, a longer interval from onset to diagnosis).34 This potential sex difference in MS and NMOSD features could contribute to the risk of ischaemic stroke and should be further evaluated.

Stroke risk was marginally different across age groups in MS and NMOSD. However, younger patients under 65 years of age tended to be at an especially higher risk than controls of the same age. A possible explanation could be that the relative contribution of MS or NMOSD as a risk factor for stroke is stronger in younger patients. Older people may be exposed to vascular risk factors such as hypertension, DM, smoking and alcohol for a longer period than younger people, so the possibility of complications related to those factors is higher in older people than in younger people. On the other hand, in young adults who are generally healthy, the disease itself can be a major cause of vascular complications, such as stroke. In addition, it has been known that relapse rates are age and time related; they are relatively high during the early disease course and decline with increasing age in both MS and NMOSD, which suggests that the endothelial dysfunction secondary to high inflammatory activity may increase stroke risk in younger patients.35–37

Our study has the following clinical implications. As the stroke risk increases in MS and NMOSD, more attention is needed for managing stroke risk factors along with disease control in patients with these conditions. Acute stroke in MS/NMOSD could be misdiagnosed or underdiagnosed due to the overlap of clinical symptoms and MRI features between relapse and stroke.38 However, early diagnosis is very important considering that hyperacute stroke management is crucial for a better prognosis, and reduced neural plasticity in the MS brain could lead to slow recovery following stroke.39 In the near future, studies should be pursued to determine whether the same primary and secondary preventive stroke treatment strategies can also be applied to patients with MS/NMOSD.

This study has several limitations. First, the antiaquaporin-4 antibody test for clinical use was approved by the Korean Ministry of Health and Welfare in 2015, but before that this test had only been done for research purposes. Therefore, patients who met the 2015 NMOSD diagnostic criteria may have been diagnosed somewhat later after the onset of the disease.16 In addition, the anti-MOG antibody test was not used clinically until October 2019 in Korea. A subset of recurrent MOGAD could be classified as seronegative NMOSD or MS. However, the seroprevalence of MOG antibodies was much lower than that of AQP4 antibodies in Korea,40 and the proportion of MOGAD cases based on MS or NMO code might be small. In addition, we speculate that their inclusion in the MS or NMOSD group would not increase stroke risk because MOGAD has a more benign clinical course than MS/NMOSD. Second, since our data were claim based, there was no information on disease activity, disease duration, severity of risk factors (hypertension, DM and dyslipidaemia), smoking, alcohol drinking, body mass index and physical activity, which might also affect the stroke risk. For example, smoking is associated with enhanced risk of both MS41 and stroke and can act as a potential confounder. Previous studies from a UK primary care database have considered smoking,2 3 but we could not due to lack of data. Third, drug use information was not fully considered in the data analyses. Systemic steroid use may increase the cardiovascular risk.42

We found that the stroke risk was 1.7-fold higher in patients with NMOSD than in a matched control population, which seems to be slightly lower than, but not statistically different from that in patients with MS (2.8-fold). Further investigations in MS and NMOSD could determine the pathophysiology of increased stroke risk and its potential sex differences.