The study sample was composed of prospectively enrolled EwMS and HCs. The inclusion criteria for EwMS were (1) age ≥50 years old, (2) neuropsychological and clinical examination, and (3) MS diagnosis per 2010-revised McDonald criteria (Polman et al., 2011). The exclusion criteria for the EwMS consisted of (1) history of psychiatric and major depressive disorder prior to the onset of MS, (2) active relapse within 30 days of the examination, and (3) use of corticosteroids within 30 days of the study. Inclusion and exclusion criteria for HCs were (1) ≥50 years old, (2) neuropsychological examination, (3) self-reported intact cognition, (4) no history of neurological or medical illness that might impact cognitive function, and (5) no diagnosis of major depressive disorder. Furthermore, HCs with mini-mental state examination (MMSE) performance lower than 28 were excluded. Both the EwMS and HCs were additionally screened with the beck depression inventory-II fastscreen (BDI-FS).

Clinical examination was performed by a neurologist and physical disability was assessed using the expanded disability status scale (EDSS) (Kurtzke, 1983). The Timed 25-Foot Walk (T25FW) and 9-Hole Peg Test (9PHT) were additional measures of lower and upper extremity function, respectively. The EwMS were classified by disease course as relapsing-remitting MS (RRMS) and progressive MS (PMS). The study was approved by the Institutional Review Board (IRB) and all participants signed an informed consent form.

The neuropsychological examination was performed by trained examiners under supervision of a board-certified neuropsychologist (RHBB) and included tests traditionally regarded as MS and AD-specific (Strauss et al., 2006). The cognitive domains, their participating tests, and corresponding references are shown in Table 1.

The exam included the Minimal Assessment of Cognitive Function in Multiple Sclerosis (MACFIMS) battery (Benedict et al., 2002) which evaluates several domains, including: processing speed/working memory [Paced Auditory Serial Addition Test (PASAT) (Gronwall, 1977) and Symbol Digit Modalities Test (SDMT) (Smith, 1982)], learning and memory [learning trials for California Verbal Learning Test–second edition (CVLT-II) (Delis et al., 2000) and Brief Visuospatial Memory Test-Revised (BVMT-R) (Benedict, 1997)], language [Controlled Oral Word Association Test (COWAT) (Ruff et al., 1996)] and executive function [number of correct sorts and description scores from the Sorting subtest of the Delis-Kaplan Executive Function System (D-KEFS) (Delis et al., 2001)].

Participants also completed neuropsychological measures that are traditionally utilized for detecting AD-specific cognitive impairments (Strauss et al., 2006). These tests included those measuring the domains of memory [Logical Memory from the Wechsler Memory Scale-Revised (WMS-R) (Wechsler, 1945)], language [Boston Naming Test (BNT) (Goodglass et al., 1983)], two categories of semantic fluency (total number of categorical items generated in one minute; animals and supermarket items), and visuospatial skill [Beery Visual-Motor Integration (VMI) (Beery, 2010) and clock drawing test (Agrell and Dehlin, 1998)]. The MMSE was also performed as an additional assessment of descriptive global cognitive functioning. For all performed psychometric tests, higher scores indicate better cognitive performance.

All statistical tests were performed using SPSS version 25.0 (IBM, Armonk, NY, United States). The distribution of the variables was examined using both Kolmogorov-Smirnov and Shapiro-Wilk tests of normality. The differences in demographics and cognitive raw scores between EwMS and HCs were determined by χ²–test for categorical variables, ANOVA or t-test for continuous, normally distributed variables, and Mann Whitney U-test for continuous, not normally distributed variables. To control for the multiple comparisons, Benjamini-Hochberg procedure was employed and adjusted p < 0.05 were considered statistically significant.

In order to delineate the degree of cognitive impairment across various cognitive domains while controlling for age, z-scores derived from the age-matched HCs were calculated. EwMS with z-score performance lower than -1.5 on at least two different cognitive domains were classified as cognitively impaired. The number of tests within each corresponding domain are shown in Table 1.

The impact of CPS on differences in cognitive performance between EwMS and HCs was additionally determined by analysis of covariance (ANCOVA) adjusted for SDMT scores.

Lastly, a multivariate logistic regression model determining the specific cognitive differences between the EwMS and the HCs was constructed. The model was composed of an initial force-entered block which corrected for differences in sex, age, and years of education. Furthermore, a second step-wise inclusion block determined the strength and order of cognitive variables which provided the greatest explanatory value in differentiating EwMS and HCs. For each derived step, the respective R², Wald coefficient, and p-values were reported.

The EwMS (n = 104) and HCs (n = 56) were similar in age (62.1 vs. 62.3 years, p = 0.919), years of education (15.2 vs. 15.6 years, p = 0.213) and sex ratio (75.0% vs. 62.5% females, p = 0.09). The MS group had a mean disease duration of 20.5 years, median EDSS scores of 3.5 (IQR 3.0–6.0), and ratio of RRMS/PMS of 73/31. Regarding disease modifying therapy (DMT), most EwMS were prescribed interferon-β (32, 30.8%), followed by glatiramer acetate (28, 17.3%), oral DMT medications (15, 14.4%), natalizumab (6, 5.8%), off-label medication (1, 0.9%), and the rest of the EwMS were not on any DMT (32, 30.8%).

The EwMS were more disabled when compared to the HCs in both lower and upper extremity function (T25FW, 6.7s vs. 4.7s, p < 0.001; and 9PHT, 24.7s vs. 21.5s, p < 0.001). The MS group also had higher depression scores (BDI-FS of 1.0 vs. 0.0, p = 0.005). All demographic and clinical information are shown in Table 2.

The raw scores from the neuropsychological examination of the EwMS and HCs on both the MACFIMS and AD-specific batteries are shown in Table 3. EwMS had lower raw scores when compared to the age-matched controls on both total learned and short delay recall in CVLT-II (49.7 vs. 54.9, d = 0.45, p = 0.018 and 9.8 vs. 11.2, d = 0.38, p = 0.045, respectively), SDMT (46.3 vs. 55.4, d = 0.91, p < 0.001), letter verbal fluency (35.8 vs. 43.0, d = 0.67, p = 0.001) and category verbal fluency (17.3 vs. 20.3, d = 0.67, p = 0.001; 20.9 vs. 25.7, d = 0.84, p < 0.001; and 38.2 vs. 45.9, d = 0.87, p < 0.001, for animal category, supermarket category and total score, respectively). Furthermore, the EwMS had lower immediate recall and delayed recall scores on WMS-R logical memory performance (23.7 vs. 26.4, d = 0.39, p = 0.045 and 18.8 vs. 21.7, d = 0.38, p = 0.048). No other group differences were observed.

To investigate the influence of processing speed on our findings, we performed an analysis of covariance in which the COWAT differences (letter-based and category-based fluency) were adjusted by CPS performance (SDMT score as covariate). The performance on SDMT had a significant influence on letter-based verbal fluency performance (F_1,156 = 39.5 r = 0.49, p < 0.001). In the final model, EwMS and HCs were not significantly different in letter-based fluency scores when CPS was controlled (F_1,156 = 3.9, p = 0.057). Processing speed also influenced the performance on total categorical verbal fluency (F_1,156 = 34.2, r = 0.425, p < 0.001) and on both animal-specific (F_1,156 = 12.2, r = 0.242, p = 0.001), and supermarket items-specific category (F_1,156 = 39.7, r = 0.476, p < 0.001). Despite the large SDMT effect, categorical verbal fluency remained significantly different between the EwMS and HCs (adjusted p = 0.036, adjusted p < 0.001, and adjusted p = 0.001, for animals, supermarket items, and total categorical verbal fluency, respectively).

Based on the HC-derived z-scores, the differences in processing speed and verbal fluency represented the most commonly affected domains in the EwMS. More than a quarter (27.9%) of EwMS had categorical verbal fluency impairment relative to HCs (22.1 and 24.0% for individual animal and supermarket tests, respectively). 25.0% were impaired on the letter-based verbal fluency, and 25.9% as impaired on the SDMT. Overall, 47.1% of the EwMS were classified as cognitively impaired on at least two different domain-specific tests. Table 4 portrays the impairment rate observed for each cognitive test.

Both the multivariate logistic regression model blocks and derived step-wise predictors are shown in Table 5. The diagnosis of MS was best predicted by differences in category-based verbal fluency (Wald = 9.935, p = 0.002) and SDMT (Wald = 13.937, p < 0.001). Together with the force-entered block containing age, sex and years of education, the final step-wise model was able to explain 27% of the diagnosis variance (Nagelkerke R² = 0.27). In terms of odds ratio (OR), for every point decrease in categorical verbal fluency, the chance of MS diagnosis increased for 8.1% [Exp(B) = 0.919]. Similarly, for every point decrease in SDMT, the change of MS diagnosis is increased for 9.3% [Exp(B) = 0.907].

Lastly, we investigated the associations between verbal fluency and cognitive processing speed with the overall disability scores as depicted by EDSS. Within the total sample of EwMS, EDSS scores were associated with both total verbal fluency (Spearman’s ranked correlation, r_s = -0.292, p = 0.003) and SDMT performance (r_s = -0.398, p < 0.001). Due to the nature of EDSS scoring system, we further delineated the EwMS into groups within the lower range of the score (EDSS scores between 0 and 4.0 which depend and reflect the neurological examination) and within the higher range of the score (EDSS scores ≥4.0 which solemnly depends on the walking ability) (van Munster and Uitdehaag, 2017; Jakimovski et al., 2018a). As expected, the associations between cognitive performance and EDSS scores were mainly present within the subset of EwMS with lower EDSS scores (total verbal fluency r_s = -0.341, p = 0.001 and SDMT r_s = -0.398, p < 0.001).