A longitudinal study of cognitive function in multiple sclerosis: is decline inevitable?

Journal of Neurology https://doi.org/10.1007/s00415-020-09720-8 ORIGINAL COMMUNICATION A longitudinal study of cognitive function in multiple sclerosis: is decline inevitable? Marina Katsari1 · Dimitrios S. Kasselimis1,2 · Erasmia Giogkaraki1 · Marianthi Breza1 · Maria‑Eleftheria Evangelopoulos1 · Maria Anagnostouli1 · Elisabeth Andreadou1 · Costas Kilidireas1 · Alia Hotary3 · Ioannis Zalonis1 · Georgios Koutsis1 · Constantin Potagas1 Received: 13 October 2019 / Revised: 15 January 2020 / Accepted: 20 January 2020 © Springer-Verlag GmbH Germany, part of Springer Nature 2020 Abstract Background Numerous cross-sectional studies report cognitive impairment in multiple sclerosis (MS), but longitudinal studies with sufficiently long-term follow-up are scarce. Objective We aimed to investigate the cognitive 10-year course of a cohort of MS patients. Methods 59 patients with clinically isolated syndrome (CIS) or relapsing–remitting (RR) MS were evaluated with Rao’s Brief Repeatable Battery of Neuropsychological Tests at baseline and follow-up (at least 10 years later). They constituted 47.2% of 124 consecutive CIS and RRMS patients originally evaluated at baseline. Patients assessed at follow-up were well matched for baseline clinical characteristics with dropouts. Results The proportion of MS patients with overall cognitive impairment was increased by 10% within the 10-year period. When grouped on the basis of impairment in specific cognitive domains at baseline, patients originally impaired showed improvement at follow-up, while the opposite trend was observed for patients non-impaired at first assessment. A detailed case-by-case investigation revealed mixed evolution patterns, several patients fail in fewer domains at follow-up compared to baseline or failing at different domains at follow-up compared to baseline. Conclusions This study suggests a more fluid picture for the evolution of cognitive function in a subgroup of MS patients and contradicts the concept of an inevitable, progressively evolving “dementia”. Keywords Multiple sclerosis · Cognitive impairment · Longitudinal study · CIS · RRMS Introduction Cognitive impairment is reported in up to 65% of patients with MS, with prominent deficits in information processing speed [1-3] and long-term memory [4-6], and has been Marina Katsari, Dimitrios S. Kasselimis, Georgios Koutsis, and Constantin Potagas have equally contributed to this work. * Dimitrios S. Kasselimis dkasselimis@gmail.com 1 1st Department of Neurology, Eginition Hospital, National and Kapodistrian University of Athens, 74 Vas. Sofias Av., 11528 Athens, Greece 2 Department of Psychiatry, School of Medicine, University of Crete, Heraklion, Greece 3 School of Medicine, National and Kapodistrian University of Athens, Athens, Greece variably referred to as “subcortical dementia”, [7] or “white matter dementia” [8]. Although such terms suggest inevitable and extensive cognitive decline in MS, other reports limit dementia prevalence to 22% of the general MS population [9]. Within this context, the concept of inevitable and extensive decline of cognitive function should be examined cautiously, through the interpretation of well-designed longitudinal studies of cognitive function in MS. There are only three longitudinal studies with adequately long intervals between initial assessment and follow-up (10–20 years) to provide important insights into the pattern of cognitive evolution in MS. Two of these studies observed deterioration in simple and complex auditory attention span and episodic verbal learning and memory, with one showing additional worsening in visuospatial memory and cognitive flexibility, while the other shows additional deterioration in information processing speed and visual construction [10, 11]. The third study found significant deterioration only in 13 Vol.:(0123456789) Journal of Neurology information processing speed and complex attention [12]. Two of these studies also noted that patients with better baseline cognitive performance seemed to experience the greatest decline over time [11, 12]. All three aforementioned studies have limitations. The study of Amato and colleagues [10] although well designed and masterfully completed, achieving a dropout rate of only 10%, is restricted to early-onset MS. The other two studies recruited patients within the framework of phase III immunotherapy trials, which again might not be representative of relapsing–remitting (RR) MS patients [11, 12]. Furthermore, a high proportion of dropouts, exceeding 70%, characterize one of these studies, further limiting its scope [11]. Given the above, we felt that there was a clear need for further longitudinal studies with adequate sample size and long inter-testing intervals to properly assess the evolution of cognitive dysfunction in patients with RRMS. The present study addresses the evolution of cognitive functions in a population of consecutive patients with clinically isolated syndrome (CIS) or RRMS. It aims to investigate cognitive changes over a long time-frame (at least 10 years), and identify possible inter-individual differences in cognitive course, on the basis of initial cognitive status. Methods Patients The present neuropsychological study is based on the longitudinal evaluation of a cohort of 124 consecutive patients with CIS or RRMS originally presenting at an MS specialist center in Athens, Greece, more than 10 years ago [13]. The follow-up assessment was not preplanned and from the initially recruited cohort, 59 (47.2%) were available for cognitive reevaluation (14 with CIS and 45 with RRMS, at baseline), as shown in Table 1. The remaining 65 could not be cognitively reassessed for variable reasons (Fig. 1). The initial assessment took place during the period 2004–2008 and the second from 2015 to 2017. Patients had neuropsychological —testing and clinical appraisal (including calculation of the Expanded Disability Status Scale-EDSS) at both time periods [14]. Out of 59 patients with longitudinal follow-up, 15 had no relapses, while 44 had at least one relapse in the intervening 10-year period. More specifically, there were 13 patients with 1 relapse, 14 with 2, 7 with 3, 4 with 4, 4 with 5, 1 with 6, and 1 patient with 7 relapses, in total. At the first assessment, 13 patients had active disease on the basis of MRI findings, without however demonstrating any symptoms. At the second assessment, there was no evidence of disease activity on MRI, nor any clinical relapses. Predetermined exclusion criteria for cognitive follow-up were co-morbidity with other major psychiatric, 13 neurological, or metabolic disorders that could independently affect cognitive function, and steroid treatment up to 2 months before the planned neuropsychological assessment. Patients with cognitive follow-up were compared for baseline demographic and clinical characteristics to patients without cognitive follow-up (Table 1). An effort was made to also compare clinical data at follow-up between these two groups. However, although disease course and EDSS were available for all patients with cognitive follow-up, they could only be determined for 25 patients without cognitive followup. Out of these 25 cases, the EDSS was actually calculated by phone interview on 17 cases (Table 1) [15]. All patients gave written informed consent. The study was approved by the Eginition hospital ethics committee. Neuropsychological assessment The Brief Repeatable Neuropsychological Battery (BRNB) [7] is a battery of neuropsychological tests, which includes the Symbol Digits Modalities Span (SDMT), to test the speed of information processing, the Buschke Selective Reminding Test (SRT) for verbal learning and memory, consisting of the Long-Term Storage (LTS) as an indicator for long-term storage/process of learning, the Consistent Long-Term Storage (CLTS) as an indicator of consolidation, and the Delayed Recall (SRTD), as an indicator for delayed retrieval, the Paced Auditory Serial Addition Test (PASAT), for working memory, in two forms, PASAT3, at a rate of 3 s per digit, and PASAT2, at a rate of 2 s per digit), the 10/36 Spatial Recall Test (SPART) for visual learning and memory, with Spatial Recall Test-immediate (SPARTi) for immediate retrieval, and Spatial Recall Test-delayed (SPARTd) for delayed retrieval, and, finally, the Word List Generation (WLG), to test semantic verbal fluency and cognitive flexibility. Statistical analyses We first calculated z-scores for all measures, based on normative data published for each neuropsychological test in the Greek population [16]. Given that demographic factors, such as age and years of formal schooling, have been shown to affect performance on most cognitive tasks, we used these standardized scores in all subsequent analyses. We then compared performance on all tests between the two time points using paired sample t test. Basic assumptions of normality and equality of variances were assessed with Kolmogorov–Smirnov and Levene’s test, respectively. In case of severe violations, the non-parametric Wilcoxon was used. To decrease the likelihood of Error Type I, the level of statistical significance was adjusted using a Bonferroni correction, and set at α = 0.01. Journal of Neurology Table 1 Basic demographic and clinical characteristics of 59 Greek multiple sclerosis patients that were cognitively reassessed 10 years following original assessment, at baseline and follow-up MS patients with cogni- MS patients without tive follow-up cognitive follow-up N (%) Gender Male Female Education (years) Age at onset (years) Age at baseline (years) Baseline disease duration (years) Age at follow-up (years) Follow-up disease duration (years)* Disease course at baseline CIS RRMS Disease course at follow-upa CIS RRMS SPMS EDSS at baseline EDSS at follow-upa,b p value 59 (47.2) 65 (52.8) – 21 38 13.86 ± 2.70 30.09 ± 10.36 33.90 ± 9.21 3.88 ± 5.20 45.44 ± 9.54 15.38 ± 5.85 22 43 13.91 ± 2.88 29.00 ± 8.80 35.10 ± 8.07 6.21 ± 6.08 47.57 ± 7.78 17.50 ± 5.41 14 45 14 51 0.771d 5 49 5 1.26 ± 1.22 1.98 ± 1.61 0 15 10 1.81 ± 1.45 3.71 ± 2.14 0.001d 0.838c 0.932c 0.531c 0.440c 0.033c 0.371c 0.130c 0.033b,c 0.0001c Additionally, baseline and available follow-up clinical data of patients that dropped out from the original cohort of 124 consecutive patients are presented and compared to the group with cognitive follow-up Data are mean ± SD. The two groups with and without cognitive follow-up were generally well matched for baseline demographic and clinical characteristics, with two exceptions: disease duration and EDSS score at baseline were somewhat lower for the group with cognitive follow-up. Furthermore, baseline cognitive characteristics were similar between the 59 patients with cognitive follow-up and the 65 patients without, with the exception of performance on SDMT, which was significantly better for the group with cognitive follow-up (not shown in table). Regarding matching for clinical data at the time of follow-up, based on the limited data available from 25 dropouts, EDSS scores and proportion of patients with secondary progressive MS were significantly higher in the dropout group a Disease course and EDSS available for all patients with cognitive follow-up, but only for 25 patients without cognitive follow-up b Out of 25 cases without cognitive follow-up that had EDSS data, 17 were phone-EDSS; EDSS: expanded disability status scale; CIS: clinically isolated syndrome; RRMS: relapsing–remitting multiple sclerosis; SPMS: secondary progressive multiple sclerosis c Chi-square test d One-way ANOVA Based on the z-scores, and following the rationale of a recent study [11] that analyzed groups on the basis of baseline performance, we calculated the proportion of impaired and non-impaired patients, at baseline and follow-up assessment (common definition of impairment: performance less than –1.5 standard deviations, in at least three tests of the BRNB). We calculated composite sores to reduce the number of variables involved, by conducting a factor analysis that indicated which scores load on the same factor. For neuropsychological tests that had high enough loadings on the same factor, the composite score was the mean z-score. Standardized composite scores were used to identify patients who were impaired or non-impaired on different cognitive domains, corresponding to the factors derived by the aforementioned analysis (a patient defined as impaired in one cognitive domain, if the composite score was less than −1.5 standard deviations). This was done for both times of testing, to identify possible transitions of individual patients from the impaired to the non-impaired group and vice versa, through a detailed case-by-case investigation. Furthermore, we conducted pairwise comparisons between the two testing moments, separate for the impaired and non-impaired at baseline patients, using Repeated Measures Analysis of Variance (ANOVA). Time of testing was the within-subject factor and group (impaired vs. non-impaired at first measurement) was the between-subject factor, for all cognitive domains revealed by the aforementioned factor analysis. 13 Journal of Neurology Fig. 1 Patient flow diagram of original cohort of 124 consecutive patients with clinically isolated syndrome or relapsing– remitting multiple sclerosis cognitively assessed in 2004–2008, illustrating main reasons for dropout from follow-up cognitive assessment Correlation analyses were performed to assess possible associations between EDSS and cognitive scores. We did run t tests for all neuropsychological scores, to investigate possible differences in cognitive performance between subgroups with or without subclinical disease activity at first assessment (see description of patients above). The two groups did not differ with regard to age and years of formal schooling. Bonferroni correction was implemented to adjust the significance level for multiple comparisons. To assess possible effects of intervening relapses on follow-up neuropsychological performance, we performed comparisons (independent-samples t test) between subgroups with or without relapses in the intervening 10-year period with regard to all cognitive scores at second assessment (see description of patients above), and cognitive change between assessments. The two groups did not differ with regard to age and years of formal schooling. We also conducted correlation analyses to assess possible relationships between number of relapses and follow-up performance or cognitive change between assessments with regard to all neuropsychological measures, for the whole sample and separately for the subgroup of 44 patients who had intervening relapses. Bonferroni correction was implemented to adjust the significance level for multiple comparisons. 13 Results Demographic and clinical characteristics at baseline and follow‑up Demographic and clinical characteristics of the MS patient cohort at baseline and follow-up are presented in Table 1. To assess possible ascertainment bias, data are presented both for the 59 MS patients with cognitive follow-up and for the 65 dropouts. All analyses that follow are restricted to the group of 59 patients that were cognitively reassessed after a 10-year interval. Overall cognitive impairment at baseline and follow‑up The proportion of MS patients with overall cognitive impairment (defined as failing at least three out of nine tests of BRNB) increased from 42% at baseline to 52.5% at 10-year follow-up. Independent-samples t tests did not show significant differences between subgroups with or Journal of Neurology without subclinical disease activity in any of the neuropsychological scores at first assessment. Similarly, no significant differences were found between subgroups with or without relapses in the intervening 10-year period, with regard to follow-up neuropsychological performance and cognitive change between assessments. There was also no significant correlation between number of intervening relapses and any of the above cognitive scores. Factor analyses of neuropsychological performance Two different factor analyses were conducted, one for baseline measures (KMO = 0.630) and one for follow-up measures (KMO = 0.617). Factor analysis for baseline measures revealed three factors: PASAT 2 and 3, SDMT, and WLG loaded on factor 1 (WLG loads marginally; 0.538); SRT-LTS, SRT-CLTS, and SRTD loaded on factor 2; SPARTi and SPARTd loaded on factor 3. Factor analysis for follow-up measures revealed four factors: SDMT and WLG loaded on factor 1; PASAT2 and 3 loaded on factor 2; SRT-LTS, SRT-CLTS, and SRTD loaded on factor 3; SPARTi and SPARTd loaded on factor 4. Based on these two factor analyses, we decided to form three composite scores and two simple-measures scores (reflecting five cognitive domains) for the whole data, separately calculated from the two assessments: verbal memory (VM) composite score (verbal memory domain), calculated by SRT-LTS, SRT-CLTS, and SRTD z-scores; visuospatial memory (VisSpM) composite score (visuospatial memory domain), calculated by SPARTi, and SPARTd z-scores; working memory (WM) composite score (working memory domain), calculated by PASAT 2, and PASAT3 z-scores; processing speed domain (z-score of SDMT); and cognitive flexibility/verbal fluency domain (z-score of WLG). Group differences in neuropsychological performance between baseline and follow‑up Two sets of pairwise comparisons were conducted for the whole sample. First, using z-scores of individual neuropsychological tests, and second using cognitive domain scores. Results revealed significant differences between the two testing times for SRTD [Z = − 3.377, p = 0.001] and the VM composite score [t(58) = 2.892, p = 0.005]. Results of pairwise comparisons for the five different cognitive domain scores are shown in more detail in Fig. 2. We then used Repeated-Measures ANOVA in each of the five cognitive domain scores for subgroups of patients impaired and unimpaired at baseline. An overview of these results for each of the five cognitive domains is illustrated in Fig. 3. More specifically, in the WM domain, there was a significant effect of time of testing (F = 4.329, p = 0.042), but also a significant interaction between time of testing and the between-subject factor (impaired vs. non-impaired in WM at baseline) (F = 18.345, p = 0.00008). Follow-up simple main effects (effect of time of testing separately for each group) were found significant for the non-impaired group [t(36) = 2.096, p = 0.043] and the impaired group [t(17) = − 3.362, p = 0.004]. In the VM domain, there was no significant effect of time of testing (F = 0.967, p = 0.33), but a significant interaction between time of testing and the between-subject factor (impaired vs. non-impaired in VM at baseline) (F = 9.994, p = 0.003). Follow-up simple main effects (effect of time of testing separately for each group) were found significant for the non-impaired group [t(44) = 3.981, p = 0.0003], but not for the impaired group [t(13) = − 1.730, p = 0.107]. In the VisSpM domain, there was no significant effect of time of testing (F = 0.112, p = 0.739), but a significant interaction between time of testing and the between-subject factor (impaired vs. non-impaired in VisSpM at baseline) (F = 25.219, p = 0.000005). Follow-up simple main effects (effect of time of testing separately for each group) were found significant for the non-impaired group [t(33) = 4.707, p = 0.00004], and the impaired group [t(24) = − 2.683, p = 0.013]. In the processing speed domain, there was no significant effect of time of testing (F = 0.246, p = 0.622), nor a significant interaction between time of testing and the betweensubject factor (impaired vs. non-impaired on SDMT at baseline) (F = 3.020, p = 0.088). In the cognitive flexibility/ verbal fluency domain, there was a significant effect of time of testing (F = 5.070, p = 0.029), but also a significant interaction between time of testing and the between-subject factor (impaired vs. non-impaired on WLG at baseline) (F = 7.600, p = 0.008). Follow-up simple main effects (effect of time of testing separately for each group) were found non-significant for the non-impaired group [t(45) = 0.610, p = 0.545]. A marginal difference was found for the impaired group [t(10) = − 2.176, p = 0.055]. Cognitive evolution of individual patients grouped according to baseline performance Table 2 illustrates the cognitive evolution of individual MS patients on the five different cognitive domains assessed. Although the total number of unimpaired patients has halved from 18 at baseline to 9 at 10-year follow-up and there is a general trend of failing in more domains over time, it becomes clear from Table 2 that this does not represent a straightforward linear evolution of cognitive dysfunction. 13 Journal of Neurology Fig. 2 Diagrams presenting mean cognitive performance (z-scores) at baseline and follow-up for the five different cognitive domains (a: processing speed; b: cognitive flexibility/verbal fluency; c: verbal memory; d: visuospatial memory; E: working memory). Significant differences between times of testing were found only for the verbal memory domain (c) Cognitive scores and physical disability There were no significant associations between EDSS score at baseline and cognitive scores at follow-up, between cognitive scores at baseline and EDSS score at follow-up, or between changes in cognitive scores over time and EDSS score at follow-up. Also, there were 13 no significant associations between EDSS score and concurrent cognitive scores, for baseline or follow-up assessment. Journal of Neurology Fig. 3 Diagrams presenting mean cognitive performance (z-scores) at baseline and follow-up for the two groups formed on the basis of patient performance at baseline (impaired vs. non-impaired), separately for the five different cognitive domains (a: processing speed; b: cognitive flexibility/verbal fluency; c: verbal memory; d: visuospatial memory; e: working memory). The diagrams show significant differences between times of testing and interactions (for details, see text) Discussion Our results suggest that over a period of 10 years, overall cognitive impairment in patients with an original diagnosis of CIS and RRMS increases by about 10%, from 42% at baseline to 52.5% at follow-up. This seems to be driven primarily by evolving dysfunction in verbal memory, as no significant change was seen in other cognitive domains. These general results, however, mask an interesting phenomenon, which is revealed when patients are divided into two groups: those impaired in a particular domain at baseline and those unimpaired at baseline. Patients originally impaired in specific cognitive domains show improvements as a group at follow-up, whereas patients with intact function at baseline show deterioration. Furthermore, looking at individual patient performance at baseline and follow-up in all five cognitive domains investigated, one can discern several cases who have failed in fewer domains at follow-up compared to baseline or who have failed at different domains at follow-up compared to baseline. This suggests a much 13 Journal of Neurology Table 2 Individual performance on different cognitive domains at baseline and follow-up of 59 patients with MS Patient 2 3 4 5 6 7 8 9 10 11 CogFlex VM 2nd Assessment VisSpM WM ProSp CogFlex VM VisSpM WM ProSp Patients unimpaired in all domains at baseline 1 1st Assessment 12 13 14 15 16 17 18 20 21 22 23 24 25 26 27 28 29 30 31 13 Patients impaired in one domain at baseline 19 Journal of Neurology Table 2 (continued) 32 33 34 36 37 38 39 40 41 42 43 44 45 46 47 Patients impaired in more than one domain at baseline 35 48 49 50 51 52 53 54 55 56 57 58 59 The Table illustrates the cognitive evolution of individual MS patients on the 5 different cognitive domains assessed, by pictorially demonstrating whether patients were impaired (dark grey rectangles) or non-impaired (light grey rectangles) in each domain at baseline and at 10-year follow-up. Patients are divided based on whether they had impairment in none, one or more than one cognitive domains at baseline. Areas left white indicate missing data. CogFlex: Cognitive Flexibility; VM: Verbal Memory; VisSpM: Visuospatial Memory; WM: Working Memory; ProSp: Processing Speed. 13 Journal of Neurology more fluid picture for the evolution of cognitive function in patients with MS, at least in the relatively milder half of MS, and contradicts the concept of an inevitable, progressively evolving ‘dementia’. The ~ 10% increase in overall cognitive impairment seen over 10 years in the present study cannot be directly compared to results from the three previous long-term studies [10-12], partly because of differences in the neuropsychological battery used and partly because of differences in the definitions for cognitive impairment. We used a definition of at least three impaired tests, to be consistent with our earlier publication [13], whereas the other reports have used a less strict definition of at least two impaired tests. Nevertheless, some useful observations can be made. In the 10-year follow-up study of Amato and colleagues [10], the percentage of cognitively impaired patients increased from 26% at baseline to 56% at follow-up. In contrast, Schwid and colleagues [11], and Strober and colleagues [12] (10 years and 18 years follow-up, respectively) observed smaller deteriorations over time of 5% and 18%, respectively. Factors that may contribute to this discrepancy are the highly variable rate of dropouts across these studies (ranging from 5 to 71%), as well as baseline differences in mean disease duration, ranging from 1.6 [10] to 7 years [11, 12], while our patients had mean disease duration of 4.6 years at baseline. Based on the literature, the presence of cognitive impairment seems to particularly increase beyond the fifth year of illness [17]. Several longitudinal studies have not shown significant cognitive deterioration, after short time intervals (follow-up of 2–4 years) in patients with CIS or early MS [18-21]. In this context, the large proportional increase in cognitively impaired patients over 10 years found in the study of Amato and colleagues [10] could be attributed to the above described delayed onset of cognitive deterioration. Regarding the specific deterioration in verbal learning and recall, significant for the whole cohort at 10-year followup, this was not seen in an analogous fashion in two previous long-term studies of cognitive evolution in MS. Schwid and colleagues [12] in fact, had virtually no change in the proportion of patients failing SRT subscores at baseline and 10-year follow-up, whereas Strober and colleagues [11], observed deterioration in verbal learning within a context of more widespread changes [11, 12]. Results from shorter longitudinal studies are contradicting [22, 23]. The cause for these discrepancies is not self-evident, and factors such as proportion of dropouts, disease duration at baseline, sample size and characteristics, and specific definition of cognitive impairment may partly underlie these differences. When patients are divided at baseline into intact and impaired in specific cognitive domains, it becomes clear that these subgroups behave differently in terms of evolution of cognitive function. Patients with intact memory at baseline (verbal, visuospatial, and working memory domains), 13 deteriorate significantly in these domains over a 10-year interval. In the case of processing speed, patients intact at baseline did not deteriorate significantly, in contrast to a previous long-term study that applied analogous methodology [11]. This discrepancy may be related to the proportion of patients converting to SPMS, which was much lower in our study [11, 13]. It is noteworthy that Schwid and colleagues [11], have also argued that declines in cognitive scores were consistently greater in patients with better baseline performance [12]. Our results also showed that patients impaired at baseline demonstrated significantly improved performance in working memory and verbal fluency. This unexpected observation of cognitive improvement in this specific subgroup of patients has not been previously acknowledged. Strober and colleagues [11], which used the same methodology, did not observe an analogous improvement in their cohort [11]. However, the longer follow-up (18 years) and high proportion of patients converting to progressive disease in their cohort precludes close comparison. The present observation, which needs replication in further studies, calls for a word of caution when interpreting uncontrolled observational studies evaluating interventions aimed at improving cognitive function in patients with MS. This ‘bouncing-back’ of cognitive function in impairedat-baseline patients is an intriguing phenomenon that may reflect the compensation capabilities of the brain in ΜS patients. Functional MRI studies report that patients with MS, without impairment in particular cognitive tasks, consistently show increased cerebral activation and more widely distributed cortical recruitment than healthy controls, and modifications of functional connectivity within cognitionrelated regions [24-26]. Yet, the whole picture seems more complex, given that resting-state functional connectivity studies suggest that increased activation can be either adaptive or maladaptive in nature, depending on the progression of the disease [27, 28]. Beyond mechanisms of brain plasticity that may compensate for more permanent focal neuronal damage, there are two factors that could have reversibly affected cognition at baseline. First, the possibility of an “isolated cognitive relapse”, which is difficult to control for, given that such phenomena are neither associated with subjective changes in mood or fatigue levels, nor with alteration in cognitive abilities insight [29]. Second, anxiety, which has been shown to be a significant predictor of cognitive performance independently of other clinical confounding variables [30]. Regarding the association of cognition with physical disability and the ability of one to act as a predictor of the other, we did not find significant correlations of EDSS scores with cognitive domains at any time-point. Several studies have found correlations between EDSS and cognitive deficits [13, 17, 31], but others have demonstrated that only high EDSS Journal of Neurology scores show this association to cognition [1, 32, 33]. Ιn our study, mean EDSS score was low in both assessments and less than 10% of patients reached secondary progression, and this might explain the aforementioned lack of correlation with cognition. To summarize the main limitations of the present study, one should first mention the 53% dropout rate, with all the negative consequences accompanying it. Second, the exclusion of primary progressive MS further limits the generalizability of our results. Third, the picture of cognitive evolution would have been more complete had 2- and 5-year assessments been included to give more information about the sequence of changes in different cognitive domains over time. Finally, a more extensive neuropsychological battery would have offered more detailed and varied observations for different cognitive domains. Particularly, tests of special executive functions (organization and planning) should have also been included. In conclusion, this study is in contrast to the notion that MS will inevitably result in progressive cognitive impairment [7, 8], and suggests a more fluid picture for the evolution of cognitive function in patients with MS, at least in the relatively milder half of the disease, and contradicts the concept of an inevitable, progressively evolving “dementia”. Cognitive impairment at baseline does not by default lead to progressive decline. Although the overall proportion of cases with cognitive impairment increases over time, there are several patients who improve over the years in specific cognitive domains, while they may or may not decline in other domains. Data availability statement Anonymized data will be shared on request by any qualified investigator. Funding This research is co-financed by Greece and the European Union (European Social Fund—ESF) through the Operational Programme «Human Resources Development, Education and Lifelong Learning» in the context of the project “Strengthening Human Resources Research Potential via Doctorate Research” (MIS-5000432), implemented by the State Scholarships Foundation (ΙΚΥ). Compliance with ethical standards Conflicts of interest The authors report no conflicts of interest. Disclosures Dr Koutsis has received research grants from Teva and Genesis Pharma and provided consultation services for and received honoraria from Novartis, Genzyme, Genesis Pharma, Specifar, Pfizer, and Teva. Dr Breza reports no disclosures. Dr Zalonis reports no disclosures. Dr Potagas reports no disclosures. Dr Evangelopoulos has provided consultation services for and received honoraria from Novartis, Biogen, and Teva. Dr Anagnostouli has received research grants from Biogen, Merck- Serono, Novartis, Teva, Bayer, and Gen- zyme, as well as lecture-fees from Novartis, Teva, Biogen and Genzyme. Dr Andreadou has received research grants from Biogen, MerckSerono, Novartis, and Sanofi-Aventis, as well as lecture-fees from Teva. Dr Kilidireas has received research grants from Biogen, Novartis, Teva, and Merck-Serono. Dr Kasselimis is supported by IKY Scholarships Programme co-financed by the European Union (European Social Fund [ESF]) and Greek national funds through the action entitled “Reinforcement of Postdoctoral Researchers,” in the framework of the Operational Programme “Human Resources Development Program, Education and Lifelong Learning” of the National Strategic Reference Framework (NSRF) 2014–2020. Ethical approval The study was approved by the Eginition hospital ethics committee. References 1. Bergendal G, Fredrikson S, Almkvist O (2007) Selective decline in information processing in subgroups of multiple sclerosis: an 8 year old longitudinal study. Eur Neurol 57:193–202 2. DeLuca J, Chelune GJ, Tulsky DS et al (2004) Is speed of processing or working memory the primary information processing deficit in multiple sclerosis? J Clin Exp Neuropsychol 26:550–562 3. Janculjak D, Mubrin A, Brinar V et al (2002) Changes of attention and memory in a group of patients with multiple sclerosis. Clin Neurol Neurosurg 104:221–227 4. Brassington JC, Marsh NV (1998) Neuropsychological aspects of multiple sclerosis. Neuropsychol Rev 8:43–77 5. DeLuca J, Barbieri-Berger S, Johnson SK (1994) The nature of memory impairments in multiple sclerosis: acquisition versus retrieval. J Clin Exp Neuropsychol 16:183–189 6. Rao SM, Grafman J, DiGuilio D et al (1993) Memory dysfunction in multiple sclerosis: its relation to working memory, semantic encoding and implicit learning. Neuropsychology 7:364–374 7. Rao SM, Leo GJ, Bernardin L et al (1991) Cognitive dysfunction in multiple sclerosis. I. Frequency, patterns, and prediction. Neurology 41(5):685–691 8. Filley CM (2012) White matter dementia. Ther Adv Neurol Disord 5(5):267–277 9. Benedict RH, Bobholz JH (2007) Multiple sclerosis. Semin Neurol 27(1):78–85 10. Amato MP, Ponziani G, Siracusa G et al (2001) Cognitive dysfunction in early-onset multiple sclerosis: a reappraisal after 10 years. Arch Neurol 58(10):1602–1606 11. Strober LB, Rao SM, Lee JC, Fischer E et al (2014) Cognitive impairment in multiple sclerosis: an 18year follow-up study. Mult Scler Relat Disord 3:473–481 12. Schwid SR, Goodman AD, Weinstein A et al (2007) Cognitive function in relapsing multiple sclerosis: minimal changes in a 10-year clinical trial. J Neurol Sci 255(1):57–63 13. Potagas C, Giogkaraki E, Koutsis G et al (2007) Cognitive impairment in different MS subtypes and clinically isolated syndromes. J Neurol Sci 267(1):100–106 14. Kurtzke JF (1998) Rating neurologic impairment in multiple sclerosis: an expanded disability status scale (EDSS). Neurology 33(11):1444–1452 15. Lechner-Scott J, Kappos L, Hofman M et al (2003) Can the Expanded Disability Status Scale be assessed by telephone? Mult Scler 9(2):154–159 16. Mantzavinou P (2016) Rao’s brief repeatable battery in multiple sclerosis. Normative data in the Greek population (oral presentation). In: Cognitive impairment in MS: from pathology to phenomenology, Poros, Greece, 17–20 June, 2016 (in Greek) 13 Journal of Neurology 17. Achiron A, Chapman J, Magalashvili D et al (2013) Modeling of cognitive impairment by disease duration in multiple sclerosis: a cross-sectional study. PLoS ONE 8(8):e71058 18. Jønsson A, Andresen J, Storr L et al (2006) Cognitive impairment in newly diagnosed multiple sclerosis patients: a 4-year follow-up study. J Neurol Sci 245(1–2):77–85 19. Mariani C, Farina E, Cappa SF et al (1991) Neuropsychological assessment in multiple sclerosis: a follow-up study with magnetic resonance imaging. J Neurol 238(7):395–400 20. Patti F, Failla G, Ciancio MR et al (1998) Neuropsychological, neuroradiological and clinical findings in multiple sclerosis. A 3 year follow-up study. Eur J Neurol 5(3):283–286 21. Uher T, Blahova-Dusankova J, Horakova D et al (2014) Longitudinal MRI and neuropsychological assessment of patients with clinically isolated syndrome. J Neurol 261(9):1735–1744 22. Piras MR, Magnano I, Canu ED et al (2003) Longitudinal study of cognitive dysfunction in multiple sclerosis: neuropsychological, neuroradiological, and neurophysiological findings. J Neurol Neurosurg Psychiatry 74(7):878–885 23. Glanz BI, Healy BC, Hviid LE et al (2012) Cognitive deterioration in patients with early multiple sclerosis: a 5-year study. J Neurol Neurosurg Psychiatry 83(1):38–43 24. Audoin B, Ibarrola D, Ranjeva JP et al (2003) Compensatory cortical activation observed by fMRI during a cognitive task at the earliest stage of multiple sclerosis. Hum Brain Mapp 20(2):51–58 25. Mainero C, Caramia F, Pozzilli C et al (2004) fMRI evidence of brain reorganization during attention and memory tasks in multiple sclerosis. Neuroimage 21(3):858–867 13 26. Rocca MA, Valsasina P, Ceccarelli A et al (2009) Structural and functional MRI correlates of Stroop control in benign MS. Hum Brain Mapp 30(1):276–290 27. Filippi M, Rocca MA (2013) Present and future of fMRI in multiple sclerosis. Expert Rev Neurother 13(sup2):27–31 28. Leavitt VM, Wylie G, Genova HM et al (2012) Altered effective connectivity during performance of an information processing speed task in multiple sclerosis. Mult Scler 18(4):409–417 29. Pardini M, Uccelli A, Grafman J et al (2014) Isolated cognitive relapses in multiple sclerosis. J Neurol Neurosurg Psychiatry 85(9):1035–1037 30. Ribbons K, Lea R, Schofield PW et al (2016) Anxiety levels are independently associated with cognitive performance in an Australian multiple sclerosis patient cohort. J Neuropsychiatry Clin Neurosci 29(2):128–134 31. Amato MP, Ponziani G, Pracucci G et al (1995) Cognitive impairment in early-onset multiple sclerosis: pattern, predictors, and impact on everyday life in a 4-year follow-up. Arch Neurol 52(2):168–172 32. Bsteh G, Ehling R, Lutterotti A et al (2016) Long term clinical prognostic factors in relapsing-remitting multiple sclerosis: insights from a 10-year observational study. PLoS ONE 11(7):e0158978 33. Drew M, Tippett LJ, Starkey NJ et al (2008) Executive dysfunction and cognitive impairment in a large community-based sample with multiple sclerosis from New Zealand: a descriptive study. Arch Clin Neuropsychol 23(1):1–9