ABSTRACT
Background
Fatigue in people with multiple sclerosis (PwMS) is common and disabling, yet poorly understood. This study aimed to clarify the distinction between long‐term fatigue and task‐induced fatigability and their respective pathophysiological processes, and to examine the contribution of decreased cognitive processing speed (CPS) to the relationship between structural brain damage, fatigue, and fatigability.
Methods
Fatigue was induced in 34 PwMS (18–49 years) and 32 age−/sex‐matched healthy controls (HCs, 24–44 years) using an individually calibrated task based on each participant's CPS. Perceived task‐induced fatigue was assessed from pre‐ to post‐task change in self‐reported fatigue, and objective fatigability from within‐task changes in performance, vigilance, and pupil diameter. Brain volumes and lesions were quantified on structural 3 T MRI.
Results
PwMS showed higher long‐term fatigue and slower CPS (p < 0.001). Once cognitive load was calibrated to individual CPS, objective markers of fatigability (performance, vigilance, pupil changes) no longer differed from HCs (all p > 0.05), whereas perceived task‐induced fatigue remained elevated. Both CPS and long‐term fatigue—but not fatigability—were associated with putamen atrophy and lesion burden. Reduced CPS mediated the associations between putamen atrophy and long‐term fatigue (indirect effect, 95% CI [−54.48, −0.97]).
Conclusions
In PwMS, the association between structural brain damage and long‐term fatigue is mediated by reduced CPS, whereas objective fatigability normalizes once cognitive load is calibrated to individual CPS. These findings support adapting cognitive load to each patient's processing capacity rather than reducing task demands uniformly to help mitigate the daily‐life impact of fatigue.
Keywords: cognitive fatigue, cognitive processing speed, fatigability, multiple sclerosis, putamen
In people with multiple sclerosis, task‐induced fatigability normalized once cognitive load was calibrated to each patient's processing speed, whereas perceived fatigue remained elevated. Long‐term fatigue—but not fatigability—was associated with putamen and callosal atrophy, an association mediated by slowed processing speed, which emerges as a key link between structural brain damage and fatigue.
1. Introduction
Fatigue is one of the most frequent and disabling symptoms in people with multiple sclerosis (PwMS) [1]. Defined as a perceived exhaustion of internal resources [2], it affects 36.5% to 90% of patients [3, 4], often appears early in the disease [5], markedly reduces quality of life [6] and increases unemployment risk [7]. Yet its pathophysiology remains poorly understood. Beyond the inflammation‐driven “sickness behavior” [8], slowed information processing caused by demyelination and disconnection has drawn particular attention [9], as reduced cognitive processing speed (CPS) is one of the most consistently impaired functions in MS [10] and a plausible driver of the effort and exhaustion patients report. How exactly slowed CPS relates to fatigue, however, remains unresolved.
Addressing this requires distinguishing two concepts often conflated under “fatigue”. Recent work recommends reserving fatigue for the subjective, self‐reported perception of exhaustion, and fatigability for the measurable change in performance or physiological output over a defined task [11, 12, 13, 14]—a distinction supported by the observation that these subjective and objective components are only moderately coupled [15, 16]. Accordingly, a task‐induced rise in the subjective dimension is best described as “increased fatigue” rather than “fatigability.” Fatigue can be further divided into state and long‐term (trait) fatigue (Figure 1).
FIGURE 1.
Conceptual framework of fatigue and fatigability. Fatigue comprises two distinct dimensions that are only weakly coupled. The subjective dimension is self‐reported fatigue, assessed with rating scales: Repeated state‐fatigue ratings (here with VAS) yield long‐term (trait) fatigue when averaged over time, whereas their pre‐ to post‐task difference (ΔVAS) reflects the increase in task‐induced fatigue. The objective dimension is fatigability, defined as the task‐induced change in an objective criterion, and comprises two components: Performance/behavioral fatigability (here decline in task accuracy and in vigilance, ΔPVT 1/RT) and physiological fatigability (change in pupil diameter, Δpupil). FSS, Fatigue Severity Scale; PVT, psychomotor vigilance test; RT, reaction time; VAS, visual analogue scale of fatigue.
CPS may be tied to both. Because information is processed more slowly, PwMS must sustain a higher relative cognitive load to meet the same demands, so that a workload that is moderate for controls becomes effortful and accelerates the performance decline that defines fatigability. On this view, the rising fatigue reported during sustained tasks may not reflect a genuine vulnerability to fatigability, but rather a mismatch between task demands and available processing capacity—a distinction that can only be resolved by adapting cognitive load to each individual's CPS.
At the brain level, fatigue has been linked to fronto‐striatal‐limbic, frontal‐cingulate and parietal networks [17], though findings in MS remain inconsistent [18, 19]. Notably, the fronto‐thalamo‐striatal injury most consistently associated with fatigue in PwMS—particularly thalamic damage and fronto‐striatal disconnection [20]—coincides with the circuit whose disruption most reliably tracks slowed CPS: fronto‐striatal‐thalamic atrophy and white‐matter damage repeatedly predict slower processing speed [21, 22, 23], and putamen atrophy in particular has been linked to slowed CPS independently of other regions, including the thalamus [24]. This convergence raises the possibility that slowed CPS is a key link between structural damage and fatigue. The mechanistic chain, however, remains unclear, as some studies find the fatigue–CPS association mediated by depression or abolished after adjustment [25, 26]. Whether CPS mechanistically links brain damage to long‐term fatigue therefore remains an open question.
This study aimed to clarify the role of CPS in fatigue and fatigability in PwMS versus healthy controls (HC), combining structural MRI with an individually calibrated fatigue‐induction task, the Timeload–Dual Back (TloadDback) [2, 27]. Grounded in the Time‐Based Resource‐Sharing Model [28]—in which limited resources are shared across functions, so that shortening stimulus duration raises cognitive load—the task sets a personalized high‐load condition from each participant's fastest successful CPS, allowing fatigability to be assessed independently of baseline differences in processing capacity.
We tested two hypotheses: (i) that task‐induced fatigability can be normalized when cognitive load is adapted to individual CPS, such that objective markers (performance and physiological markers) no longer differ from HC; and (ii) that reduced CPS mediates the relationship between structural brain damage—particularly putamen atrophy—and fatigue.
2. Methods
2.1. Participants
Thirty‐six patients with relapsing–remitting MS (McDonald 2017 criteria [29]) were recruited at the Neurology Department of the Hôpital Universitaire de Bruxelles (HUB, 2021–2024), together with 33 age‐ and sex‐matched healthy controls (HCs). All participants gave written informed consent (ethics HUB P2020/708). Inclusion criteria included normal or corrected vision/hearing, fluent French, and no neurological (other than MS in patients), psychiatric or substance‐use disorders. Exclusion criteria were major depression (Beck Depression Inventory, BDI‐II [30] ≥ 28), pregnancy, or upper‐limb impairment. Patients also required Expanded Disability Status Scale (EDSS) < 6 and ≥ 3 months relapse‐free. Two patients lacked MRI because of claustrophobia and one HC was excluded for an extreme TloadDback calibration score (> 3SD). Final samples comprised 34 PwMS (26 women, 18–49 years) and 32 HCs (21 women, 24–44 years). Pupil data were available on 33 PwMS and 31 HC due to technical issues.
2.2. Procedure
The experiment took place over two consecutive days (Figure 2). On Day 1, participants completed questionnaires on long‐term fatigue (Fatigue Severity Scale, FSS [31]), depression (BDI‐II), and anxiety (State and Trait Anxiety Inventory, STAI‐Y [32]), then underwent a 1.5 h TloadDback calibration in the early afternoon (1 pm). On Day 2, morning MRI acquisitions were followed by the TloadDback fatigue‐induction task at 1 pm with simultaneous eye‐tracking (~1 h). Visual Analog Scale of Fatigue (VASf [33]) and a 5‐min Psycho‐motor vigilance test (PVT [34]) were administered immediately before and after the task.
FIGURE 2.
TloadDback Cognitive Fatigue Induction Protocol. Day 1 is comprised of the long‐term fatigue questionnaire (Fatigue Severity Scale, FSS) together with the Beck Depression Inventory (BDI‐II) and the State–Trait Anxiety Inventory (STAI‐Y), followed by the calibration task. The calibration phase of the TloadDback task is used to determine the individual's maximum cognitive load, represented by the fastest CPS (shortest Stimulus Time Duration, STD) for successful completion of the task (> 85% accuracy) carried out in a stepwise block format consisting of 30 letters and 30 numbers in alternation for each block (1‐back and parity test). Day 2 consists of pre‐task measures of state fatigue and vigilance (VASf and PVT); the TloadDback fatigue induction session tailored to the individual's CPS, with an eye tracker to measure pupil dilation over the task; and post‐task assessments of vigilance and state fatigue (PVT and VASf, reversed). Task‐induced fatigue was assessed by changes between post and pre‐task state fatigue (ΔVASf). Fatigability was assessed by computing over‐task changes in performance (accuracy across quartiles), vigilance (ΔPVT), and pupil diameter (Δpupil).
2.3. CPS and Fatigue‐Induction Task (TloadDback)
The TloadDback task followed the protocol previously detailed in Borragán et al. [2].
2.3.1. Calibration Phase (Day 1)
This phase determines each individual's shortest CPS. After a familiarization session (two 30‐trial blocks of a number‐parity task and a 1‐back letter task separately, then one 60‐trial block alternating both), participants completed blocks of 60 alternating trials with a stepwise decreasing stimulus time duration (STD), starting at 1500 ms. The participants were asked to respond as quickly and accurately as possible. Block performance was computed with a weighted formula (accuracy for letters = 65% and for digits = 35% of the total score). If accuracy exceeded 85%, STD was reduced by 100 ms; if it fell below 85%, the same STD was repeated. The task ended when accuracy was below 85% for three consecutive blocks at the same STD. The shortest STD maintained above threshold indexed the individual CPS. This calibration phase has shown robust test–retest reliability [35].
2.3.2. Fatigue‐Induction Phase (Day 2)
Participants performed a single 16‐min block at their individual shortest STD (fastest CPS), ensuring each participant experienced a sustained high cognitive load. Performance was analyzed across four 4‐min quartiles to track changes over time [2].
2.4. Fatigue and Fatigability Measures
Long‐term fatigue was measured with the FSS. State fatigue was assessed immediately before and after the TloadDback fatigue‐induction task using a visual analogue scale (VASf); the pre‐ to post‐task increase in VASf (ΔVASf) indexed the subjective, task‐induced rise in fatigue.
Fatigability, the objective decline in task performance, was quantified as the change in accuracy across task quartiles [2, 27]. Additional objective markers of task‐induced change were also acquired. Vigilance was indexed by reciprocal reaction times (1/RT) on the PVT, compared pre‐ and post‐task. Pupil diameter—reflecting cognitive effort and fatigue state [36]—was recorded continuously during the task with a Tobii Pro Fusion eye tracker (120 Hz) [37] and averaged per 4‐min quartile, with the difference between the first and last quartile computed as a within‐task change.
2.5. MRI Acquisitions
2.5.1. Volumes
MRI data acquisitions were carried out with a 3 T SIGNA PET‐MR scanner (GE HealthCare, Milwaukee, WI, USA) using a 32‐channels head receiver coil. 3D Brain Volume (BRAVO) T1‐weighted (TR/TE/Flip Angle: 8.3 ms/3.1 ms/12°; Field of view covering the whole head; resolution 1 mm × 1 mm × 1 mm) and 3D T2‐FLAIR images (TR/TE/TI: 7200 ms/120 ms/2041 ms; Field of view covering the whole head; in‐place resolution 1 mm × 1 mm, slice thickness 1.4 mm) were acquired for determining structural volumes and lesions. Structural parcellation was obtained using the Desikan‐Killiany Atlas [38]. Volumes were calculated using FreeSurfer 6 image software [39]. Bilateral structure volumes were computed as the mean of left and right volumes. Lesion filling was not performed as previous literature has shown MS lesion filling did not change FreeSurfer output [40]. All structural volumes were normalized using the estimated total intracranial volume in milliliters and then represented as percentages.
2.5.2. Lesion Segmentation
The segmentation of T2 FLAIR hyperintense lesions was obtained using the Lesion Segmentation Toolbox‐Lesion Growth Algorithm (LST‐LGA) [41] with the following parameters: an initial threshold of 0.2, the Markov Random Field parameter of 1, and the maximum number of iterations for the LGA of 50. Lesion maps were verified by an MS‐certified neurologist (SE) to confirm there were no major errors requiring manual correction.
2.6. Statistics
Repeated measures ANOVAs (with Greenhouse–Geisser correction) were used to test differences in fatigue states (between‐group), fatigability (within‐subjects, pre‐ vs. post‐task), and the difference in fatigability between PwMS and HC (group × session interaction). Between‐group comparisons were assessed with Mann–Whitney U tests. Kendall's tau‐B correlations examined associations between structural volumes, lesion measures, and clinical variables both within and across groups. Mediation analyzes were performed in JASP [42] using the lavaan syntax of the SEM package, with standardized estimates derived from 5000 bootstrapped samples and a maximum likelihood (ML) estimator [43]. Significance was set at p < 0.05 and adjusted for multiple comparisons using false discovery rate (FDR).
3. Results
3.1. Group Characteristics
Demographic and psychosocial analyzes revealed no significant differences between HC and PwMS in terms of age, sex, relationship status, laterality, and sleepiness (Table 1). Level of education and occupational status were significantly lower in PwMS, while levels of depression, anxiety, and insomnia were higher. As expected, long‐term fatigue (assessed using the FSS) was significantly higher in PwMS and was positively associated with disease severity (EDSS; B = 0.495, p FDR = 0.001; Table S1). Structural MRI analyzes showed significant decreased volumes in PwMS compared to HC of total white matter, corpus callosum, thalamus, putamen, nucleus accumbens, precuneus, prefrontal cortex, and parietal lobules (Table 2 and Figure 3).
TABLE 1.
Demographic and clinical characteristics in patients with multiple sclerosis and healthy controls.
| HC (n = 32) | PwMS (n = 34) | χ 2 | p | |||
|---|---|---|---|---|---|---|
| n | % | n | % | |||
| Sex (M/F) | 11/21 | F = 66% | 8/26 | F = 76% | 0.946 | 0.331 |
| Education a | 3 | 100% | 3 | 85.3% | 6.012 | 0.049* |
| Relationship Status | 12 | 37.5% | 18 | 52.9% | 1.585 | 0.208 |
| Occupational Status | 29 | 90.6% | 23 | 67.7% | 5.208 | 0.022* |
| Median | [Min–Max] | Median | [Min–Max] | U | p | |
| Age | 31 | [24–44] | 34 | [18–49] | 476.00 | 0.389 |
| Laterality | 79.17 | [−100 to 100] | 66.76 | [−58.33 to 100] | 569.00 | 0.751 |
| Depressive Symptoms (BDI‐II) | 3 | [0–24] | 14 | [1–36] | 197.50 | < 0.001*** |
| State Anxiety (STAI‐Y) | 26 | [20–46] | 35.5 | [20–59] | 278.50 | < 0.001*** |
| Trait Anxiety (STAI‐Y) | 37 | [20–58] | 46.5 | [25–65] | 365.00 | 0.022* |
| Insomnia Index (ISI) | 6 | [0–13] | 11 | [1–24] | 259.00 | < 0.001*** |
| Epworth Sleepiness Scale | 6.5 | [0–14] | 8 | [2–18] | 400.00 | 0.065 |
| Long‐term Fatigue (FSS) | 25 | [11–55] | 39.5 | [15–62] | 190.00 | < 0.001*** |
| Disease Descriptives | Median | Range | n/a | n/a | ||
| Duration (years) | 7.5 | [0.5–22] | ||||
| Disease Severity (EDSS) | 1.5 | [0–4] | ||||
| Lesion Count | 18 | [0–39] | ||||
| Lesion Volume (%) | 0.00024 | [0–0.005] | ||||
Education level is based on 3 levels: primary, secondary and university. BDI‐II: Beck Depression Inventory, version 2; STAI‐Y: State–Trait Anxiety Inventory; ISI: Insomnia Severity Index; FSS: Fatigue Severity Scale; EDSS: Expanded Disability Status Scale. Between‐group data were analyzed using the Mann–Whitney U test. Sex was analyzed using frequency distribution analysis.
p < 0.05.
p < 0.001.
TABLE 2.
Structural brain volumes in patients with multiple sclerosis and healthy controls.
| Volumes (%) | HC (n = 32) | PwMS (n = 34) | U | p FDR | ||
|---|---|---|---|---|---|---|
| Median | SD | Median | SD | |||
| Total Gray Matter | 43.753 | (±1.479) | 45.492 | (±5.894) | 504.000 | 0.615 |
| Total White Matter | 29.811 | (±1.570) | 27.877 | (±3.540) | 819.000 | < 0.001*** |
| Thalamus | 0.906 | (±0.054) | 0.840 | (±0.142) | 772.000 | 0.006** |
| Putamen | 0.656 | (±0.063) | 0.564 | (±0.123) | 823.000 | < 0.001*** |
| Pallidum | 0.218 | (±0.031) | 0.207 | (±0.026) | 683.000 | 0.131 |
| Caudate | 0.463 | (±0.059) | 0.450 | (±0.057) | 675.000 | 0.146 |
| Nucleus accumbens | 0.091 | (±0.016) | 0.074 | (±0.023) | 841.000 | < 0.001*** |
| Cingulate Cortex | 1.391 | (±0.092) | 1.402 | (±0.155) | 523.000 | 0.887 |
| Precuneus | 1.403 | (±0.089) | 1.187 | (±0.118) | 1038.000 | < 0.001*** |
| Insular Cortex | 1.000 | (±0.085) | 1.031 | (±0.097) | 450.000 | 0.295 |
| Prefrontal Cortex | 8.201 | (±0.459) | 7.389 | (±0.543) | 946.000 | < 0.001*** |
| Parietal Lobules | 3.386 | (±0.217) | 3.044 | (±0.284) | 915.000 | < 0.001*** |
| Temporal Lobe | 6.797 | (±0.391) | 6.801 | (±0.754) | 562.000 | 0.887 |
| Occipital Lobe | 3.118 | (±0.310) | 3.292 | (±0.462) | 440.000 | 0.259 |
| Cerebellum | 7.166 | (±0.465) | 7.200 | (±0.920) | 542.000 | 0.985 |
| Corpus Callosum | 0.243 | (±0.033) | 0.193 | (±0.037) | 888.000 | < 0.001*** |
Note: Mann–Whitney U tests.
p < 0.01.
p < 0.001.
FIGURE 3.
Illustration of deep brain regions with decreased volumes in PwMS compared to HC: Putamen (blue), thalamus (red), nucleus accumbens (green), along with the corpus callosum (light blue); total white matter, precuneus, prefrontal cortex and parietal lobules are not shown for better visibility.
3.2. CPS, Task‐Induced Fatigue and Fatigability
During the calibration phase of the TloadDback (Figure 2), PwMS needed longer STD, that is, longer processing times, than HC to perform equally (Figure 4a, PwMS: mean = 991.18 ± 223.43 ms, median = 900.0 ms vs. HC: mean = 806.25 ± 152.27 ms, median = 800.0 ms, U = 286.0, p < 0.001). Once adapted to individual CPS, performance across the fatigue‐induction task remained stable and comparable between groups (Figure 4b, main effect of performance along quartiles: F(2.23, 142.59) = 2.15, p = 0.115, main effect of group: F(1,64) = 0.82, p = 0.369, interaction performance × group: F(2.23, 142.59) = 1.03, p = 0.366). Vigilance and pupil diameter (Figure 4d,e) both decreased after the task, but their evolution was similar between groups (ANOVA 1/RT: main effect of group: F(1,64) = 2.54, p > 0.05, main effect of sessionpre/post: F(1,64) = 41.01, p < 0.001, interaction: F(1,64) = 0.84, p > 0.05; ANOVA pupil: main effect of group: F(1,62) = 5.42, p = 0.023, main effect of sessionpre/post: F(1,62) = 42.68, p < 0.001, interaction: F(1,62) = 1.38, p > 0.05). Thus, once cognitive load was individually calibrated, objective markers of fatigability (performance, vigilance, pupil) showed comparable task‐induced change in both groups.
FIGURE 4.
CPS and fatigue induction in HC (red) and PwMS (blue). (a) During the calibration phase, the HC group showed significantly faster CPS (shorter STD) than the PwMS group (p < 0.001). (b) During the fatigue‐induction task (adapted for individual CPS), quartile performance levels and their evolution were similar between groups. (c) Perceived State Fatigue was higher in PwMS both pre‐ and post‐task and increased more from pre‐ to post‐task (PwMS: VASfpre = 3.544 ± 1.867, VASfpost = 5.533 ± 2.310, HC: VASfpre = 2.245 ± 1.613, VASfpost = 3.510 ± 2.116). In contrast, (d) vigilance (PwMS: PVT 1/RTpre = 2.713 ± 0.289, 1/RTpost = 2.569 ± 0.353; HC: 1/RTpre = 2.802 ± 0.198, 1/RTpost = 2.694 ± 0.270) and (e) pupil diameter (PwMS: Pupilpre = 4.115 ± 0.775, pupilpost = 3.888 ± 0.680; HC: Pupilpre = 3.815 ± 0.613, pupilpost = 3.458 ± 0.616) declined similarly in both groups. Thus, objective markers of fatigability changed comparably across groups, whereas subjective task‐induced fatigue (c) was greater in PwMS. Error bars are standard errors of the means.
In contrast, perceived task‐induced fatigue diverged. Perceived state fatigue was higher after the task than before in both groups (Figure 4c) but was higher in PwMS than in HCs both before and after the task and increased further post‐task, indicating greater baseline and task‐induced fatigue in PwMS (ANOVA main effect of group: F(1,64) = 12.91, p < 0.001, main effect of sessionpre/post: F(1,64) = 93.85, p < 0.001, interaction: F(1,64) = 4.64, p = 0.035).
Longer processing time (STD) was associated, in PwMS, with higher long‐term fatigue (FSS, B = 0.397, p FDR = 0.002) and post‐task state fatigue (VASf, B = 0.347, p FDR = 0.014; pre‐task p > 0.05), and with lower post‐task vigilance (PVT‐1/RT, B = −0.264, p FDR = 0.038; pre‐task p > 0.05; pupil pre and post p > 0.05). Conversely, processing time was not associated with the task‐induced changes, neither in fatigue (ΔVASf), nor in vigilance or pupil diameter (ΔPVT‐1/RT, Δpupil diameter, all ps > 0.05). Like long‐term fatigue, longer processing time was associated with EDSS (B = 0.417, p FDR = 0.008, Table S1).
3.3. Psychosocial Factors
Depression symptoms (BDI‐II) were associated with long‐term fatigue (FSS, B = 0.300, p FDR = 0.033) and with post‐task state fatigue (VASf, B = 0.351, p FDR = 0.028) and post‐task vigilance (B = −0.346, p FDR = 0.014). However, they were not associated with task‐induced fatigue (ΔVASf, B = 0.105, p FDR = 0.55), nor with task performance or CPS (Tables S2 and S3). State and trait anxiety (STAI), insomnia severity, educational levels, and occupational status were not associated with any measure of fatigue, nor with behavioral and physiological measures (Tables S2 and S3). Neither depression nor anxiety was associated with brain volumes or lesion burden (Table S4).
3.4. CPS, Fatigue, Fatigability and Brain Volumes
Given that both fatigue and slowed CPS in MS have been linked to fronto‐thalamo‐striatal and long‐range connectivity disruption, including fronto‐parietal and white‐matter damage [17, 21, 22, 23, 24], we focused our analyzes on these structures. In PwMS, long‐term fatigue was significantly associated with reduced putamen volume and with lesion burden. Longer processing time was associated with reduced volumes of the putamen, thalamus, caudate, and corpus callosum, and with lesion burden (Table 3).
TABLE 3.
Correlations between brain volumes, fatigue and processing times.
| Volumes (%) | Long‐term fatigue (FSS) | Processing times (STD) | ||
|---|---|---|---|---|
| B | p FDR | B | p FDR | |
| Putamen | −0.347 | 0.034* | −0.403 | 0.013* |
| Thalamus | −0.236 | 0.138 | −0.322 | 0.031* |
| Pallidum | 0.056 | 0.861 | −0.060 | 0.730 |
| Caudate | −0.189 | 0.238 | −0.299 | 0.038* |
| Prefrontal Cortex | 0.034 | 0.889 | −0.079 | 0.713 |
| Parietal Lobules | 0.005 | 0.965 | 0.002 | 0.988 |
| Corpus Callosum | −0.268 | 0.108 | −0.372 | 0.014* |
| Total White Matter | −0.164 | 0.283 | −0.152 | 0.371 |
| Lesion Count | 0.471 | < 0.001*** | 0.388 | 0.004** |
| Lesion Volume | 0.330 | 0.007** | 0.313 | 0.010** |
Note: Kendall's Tau Correlations.
p < 0.05.
p < 0.01.
p < 0.001.
In contrast, neither state (pre and post VASf) nor task‐induced fatigue (ΔVASf) was associated with any structural volumes or with lesion burden (Tables S2 and S3). We ran additional analyzes to test if task‐induced fatigue could be linked to structures involved in interoception and self‐monitoring, such as the insula and the anterior cingulate cortices, but no association was found (ΔVASf with Insula, B = 0.11, p = 0.35; ACC, B = 0.09, p = 0.45). For vigilance, both state (pre‐ and post‐task) and task‐induced measures were associated with lesion count and volume, whereas pupil diameter was not (lesion count: PVT‐1/RTpre: B = −0.344, p FDR = 0.01; PVT‐1/RTpost: B = −0.470, p FDR < 0.01; ΔPVT‐1/RT: B = 0.261, p FDR = 0.031, see also Tables S5 and S6).
3.5. CPS as Mediator of Brain–Long‐Term Fatigue Effects
We then conducted mediation models to test the hypothesis that slower CPS capabilities may mediate the link between altered brain structures and long‐term fatigue in PwMS. We conducted mediation models testing processing time (STD) as a mediator between, on one hand, volumes of the corpus callosum, the putamen, and lesion count, and, on the other hand, fatigue severity (FSS), with depression included as a covariate. The total effect of corpus callosum atrophy on long‐term fatigue was statistically significant (b = −141.7, 95% bootstrap CI [−248.72, −23.84]), along with the indirect effect via the processing time (b = −58.57, 95% CI [−146.14, −1.65]), while the direct effect was not (b = −83.1, 95% CI [−196.45, 53.62]), suggesting a full mediating role of CPS (see also Table S7). The same mediation effect was observed with the putamen (total effect b = −48.39, CI [−80.48, −15.12]; direct effect b = −29.06, CI [−66.29, 6.22]; indirect effect b = −19.33, CI [−54.48, −0.97]; Table S8). Lesion count was significantly associated with fatigue severity, both in the total (b = 0.67, p < 0.001) and the direct effect (b = 0.54, p = 0.002), but with no evidence that this relationship was mediated by processing time (indirect effect: b = 0.13, p = 0.168; Table S9).
4. Discussion
This study aimed to clarify the role of CPS in cognitive fatigue and fatigability in pwMS. By adapting cognitive load to individual processing speed, we were able to isolate markers of fatigability independently of baseline limited cognitive resources. We showed that fatigability did not differ from HCs when accounting for individual CPS. In addition, we explored how processing speed may contribute to the relationship between structural brain alterations and long‐term fatigue, and showed that CPS mediates the link between brain atrophy—especially the putamen and corpus callosum—and fatigue.
The results confirmed the elevated long‐term fatigue in PwMS compared to HC, which was associated with disease severity (EDSS) and depressive symptoms. PwMS also required more time than HC to accomplish the same task with the same level of accuracy, highlighting slowed processing speed. However, when cognitive load was adjusted to individual CPS, all objective markers of fatigability—including task performance, physiological responses, and vigilance—were normalized and followed a similar trajectory to those observed in HC. These findings confirm and extend the preliminary results of two previous behavioral studies conducted in small samples of PwMS using the TloadDback task, which reported a similar pattern of performance evolution across groups over the course of the task [27, 36]. This suggests that objective fatigability in PwMS is not altered per se but rather results from reduced cognitive resources (in relation to reduced CPS) in daily life. Also, when cognitive load was individualized, none of the fatigability markers were associated with brain volume or lesion burden. This is consistent with the finding that the process of fatigability throughout the task was similar between HC and PwMS and suggests it may result from normal functional mechanisms rather than the consequences of structural damage.
Notably, perceived task‐induced fatigue remained elevated in PwMS even though objective markers did not. This dissociation is consistent with evidence that perceived and objective task‐induced changes are only moderately correlated and are better assessed separately [15, 16]. Subjective experience may be additionally shaped by interoceptive, affective, and metacognitive processes [12, 44]. However, in our study, perceived task‐induced fatigue was not related to affective symptoms (depression and anxiety), to interoceptive and self‐monitoring structures such as the anterior cingulate or insula, and to lesion burden. Its origin therefore remains to be established and may lie in functional alterations not captured here.
We further investigated how increased information processing times may contribute to the association between structural damage and long‐term fatigue. Both processing times and long‐term fatigue were associated with lesion burden (count and volume) and structural damage in white matter (corpus callosum) and the thalamo‐striatal network. Mediation analyzes showed that reduced processing speed mediated the relationship between corpus callosum and putamen atrophy and fatigue severity, even after accounting for depressive symptoms. Demyelination of long‐distance fibers and axonal degeneration in the thalamo‐striato‐cortical network are likely major contributors to impaired neuronal conduction in PwMS, leading to decreased CPS and increased cognitive load and fatigue in daily life. Lesion count was significantly associated with fatigue severity, but with no evidence that this relationship was mediated by processing speed. These results suggest that lesion burden contributes directly to fatigue, rather than through its impact on cognitive slowing. This may indicate that lesion count itself is not responsible for reduced CPS, but that the location of lesions and their disruption of long‐range connectivity and thalamo‐cortical loops may be more critical. Beyond this processing‐speed pathway, it is also possible that the putamen may contribute to fatigue independently of CPS, through its role in attention and motivation [45], and in action planning and execution [46].
Our findings carry practical implications. Because a rise in perceived fatigue during sustained cognitive activity may reflect either genuine fatigability or a mismatch between task demands and processing capacity, assessing and managing cognitive fatigue may require accounting for individual CPS rather than applying a uniform load. Concretely, calibrating task difficulty to each patient's processing speed could equate relative load across patients in rehabilitation or fatigue‐management settings, while adapting the pace and temporal structure of daily and occupational tasks (time per task, pacing, fractionation) may matter more than reducing their quantity alone. Because our study was mechanistic and did not test load‐adaptive interventions, these implications remain to be validated in specific interventional studies.
Several limitations should be noted. Our cohort comprised patients with mild disability (EDSS < 6, median 1.5), partly because the sustained, motorically and cognitively demanding paradigm is difficult to administer in more advanced disability. Disease duration, however, spanned a wide range (up to 22 years), so our sample was not restricted to early disease. Our findings therefore apply to mildly disabled MS and cannot be extrapolated to more disabled patients, in whom the CPS–fatigue relationship may differ. Similarly, patients with major depression (BDI‐II ≥ 28) were excluded: while this limits generalizability to patients with significant comorbid depression, it was a deliberate choice to minimize confounding by mood, given the strong association between depression and fatigue in MS. Within the included range, depression was related to overall fatigue but not to task‐induced fatigue or brain structure, indicating that our structural findings are unlikely to be mood‐driven; whether this extends to patients with major depression requires dedicated study. In addition, the FSS captures both cognitive and motor dimensions of fatigue; the putamen–fatigue association may therefore partly reflect motor fatigue and difficulties in movement execution or automation [47, 48], although patients with significant motor disability were excluded.
In conclusion, reduced CPS mediates the relationship between structural brain damage and long‐term fatigue in PwMS, and objective task‐induced fatigability normalizes once cognitive load is calibrated to individual CPS—whereas perceived fatigue remains elevated. These findings clarify the pathophysiology of cognitive fatigue in MS and identify slowed processing speed as a key link between structural damage and fatigue. They suggest that managing this fatigue may benefit from adapting cognitive load to each patient's processing capacity rather than reducing task demands uniformly—a principle that warrants validation in future studies.
Author Contributions
Jeromy Hrabovecky: conceptualization, investigation, writing – original draft, methodology, visualization, formal analysis, data curation. Xavier De Tiège: writing – review and editing, funding acquisition, conceptualization, resources, supervision. Mélanie Strauss: conceptualization, funding acquisition, writing – review and editing, project administration, resources, supervision. Philippe Peigneux: conceptualization, writing – review and editing, supervision. Sophie Elands: investigation, writing – review and editing. Antonin Rovai: investigation, writing – review and editing, methodology, formal analysis. Guillermo Borragán: conceptualization, writing – review and editing, methodology. Nicolas Gaspard: resources, writing – review and editing. Hichem Slama: supervision, methodology, conceptualization, writing – review and editing, formal analysis.
Funding
J.H. was supported by the Fonds Erasme pour la recherche médicale (Brussels, Belgium; Convention de recherche “Les Voies du Savoir II”) and GE HealthCare (Wauwatosa, WI, USA); S.E. was supported by the Fonds Erasme; N. G. was supported by research grants from the FNRS, Fonds Erasme, Innoviris and the Fondation Jaumotte‐Demoulin (Brussels, Belgium); X.D.T is a clinical researcher at the Fonds de la Recherche Scientifique (FRS‐FNRS, Brussels, Belgique); M.S. is Postdoctorate Clinical Master Specialist at the FRS‐FNRS. This project was supported by the Fonds Erasme (Clinical Research Project attributed to M.S.) and GE HealthCare (Research Project attributed to X.D.T). The PET‐MR project at the Université Libre de Bruxelles and Hopital Universitaire de Bruxelles is financially supported by the Association Vinçotte Nuclear (AVN, Brussels, Belgium).
Conflicts of Interest
S.E. served on advisory boards for Alexion, Novartis Sanofi, Biogen and BMS and has spoken for Roche and Novartis in 2024 and 2025. N.G. consulted and served on the speakers' bureau of Angelini Pharma and UCB Pharma, and on the advisory board of Bioserenity.
Supporting information
Table S1: Disease parameters and fatigue, vigilance and pupil diameter in PwMS.
Table S2a: Affective factors and fatigue, vigilance and pupil diameter in PwMS.
Table S2b: Demographic and sleep factors and fatigue, vigilance and pupil diameter in PwMS.
Table S3: Cognitive Processing Times and Task Performance with Psycho‐social Factors.
Table S4: Psycho‐affective factors and brain volumes and lesion burden in PwMS.
Table S5: Pre‐ and post‐task measures, brain volumes and lesion burden in PwMS.
Table S6: Correlation analyzes of task‐induced fatigue and fatigability with brain regions in PwMS.
Table S7: Mediation analysis ‐ Corpus callosum.
Table S8: Mediation analysis ‐ Putamen.
Table S9: Mediation analysis ‐ Lesion count.
Acknowledgements
We acknowledge all the participants, patients, and healthy volunteers for their contribution to this work.
Data Availability Statement
The data that supports the findings of this study are available on request from the corresponding author. The data are not publicly available due to privacy or ethical restrictions.
References
- 1. Penner I. K. and Paul F., “Fatigue as a Symptom or Comorbidity of Neurological Diseases,” Nature Reviews. Neurology 13, no. 11 (2017): 662–675. [DOI] [PubMed] [Google Scholar]
- 2. Borragán G., Slama H., Bartolomei M., and Peigneux P., “Cognitive Fatigue: A Time‐Based Resource‐Sharing Account,” Cortex 89 (2017): 71–84. [DOI] [PubMed] [Google Scholar]
- 3. Oliva Ramirez A., Keenan A., Kalau O., Worthington E., Cohen L., and Singh S., “Prevalence and Burden of Multiple Sclerosis‐Related Fatigue: A Systematic Literature Review,” BMC Neurology 21, no. 1 (2021): 468. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 4. Barbi C., Pizzini F. B., Tamburin S., et al., “Brain Structural and Functional Alterations in Multiple Sclerosis‐Related Fatigue: A Systematic Review,” Neurology International 14, no. 2 (2022): 506–535. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 5. Bertoli M. and Tecchio F., “Fatigue in Multiple Sclerosis: Does the Functional or Structural Damage Prevail?,” Multiple Sclerosis Journal 26, no. 14 (2020): 1809–1815. [DOI] [PubMed] [Google Scholar]
- 6. Young C. A., Mills R., Rog D., et al., “Quality of Life in Multiple Sclerosis Is Dominated by Fatigue, Disability and Self‐Efficacy,” Journal of the Neurological Sciences 426 (2021): 117437. [DOI] [PubMed] [Google Scholar]
- 7. Hechenberger S., Helmlinger B., Penner I. K., et al., “Psychological Factors and Brain Magnetic Resonance Imaging Metrics Associated With Fatigue in Persons With Multiple Sclerosis,” Journal of the Neurological Sciences 454 (2023): 120833. [DOI] [PubMed] [Google Scholar]
- 8. Dantzer R., Heijnen C. J., Kavelaars A., Layé S., and Capuron L., “The Neuroimmune Basis of Fatigue,” Trends in Neurosciences 37, no. 1 (2014): 39–46. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 9. Andreasen A. K., Spliid P. E., Andersen H., and Jakobsen J., “Fatigue and Processing Speed Are Related in Multiple Sclerosis,” European Journal of Neurology 17, no. 2 (2010): 212–218. [DOI] [PubMed] [Google Scholar]
- 10. Benedict R. H. B., Amato M. P., DeLuca J., and Geurts J. J. G., “Cognitive Impairment in Multiple Sclerosis: Clinical Management, MRI, and Therapeutic Avenues,” Lancet Neurology 19, no. 10 (2020): 860–871. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 11. Kluger B. M., Krupp L. B., and Enoka R. M., “Fatigue and Fatigability in Neurologic Illnesses: Proposal for a Unified Taxonomy,” Neurology 80, no. 4 (2013): 409–416. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 12. Enoka R. M., Almuklass A. M., Alenazy M., Alvarez E., and Duchateau J., “Distinguishing Between Fatigue and Fatigability in Multiple Sclerosis,” Neurorehabilitation and Neural Repair 35, no. 11 (2021): 960–973. [DOI] [PubMed] [Google Scholar]
- 13. Behrens M., Gube M., Chaabene H., et al., “Fatigue and Human Performance: An Updated Framework,” Sports Medicine 53, no. 1 (2023): 7–31. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 14. Penner I. K., Grothe M., and Chan A., “Fatigue: A Common but Poorly Understood Symptom in Neurological and Non‐Neurological Diseases,” Nature Reviews. Neurology 21, no. 12 (2025): 706–720. [DOI] [PubMed] [Google Scholar]
- 15. Loy B. D., Taylor R. L., Fling B. W., and Horak F. B., “Relationship Between Perceived Fatigue and Performance Fatigability in People With Multiple Sclerosis: A Systematic Review and Meta‐Analysis,” Journal of Psychosomatic Research 100 (2017): 1–7. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 16. Hu M., Muhlert N., Robertson N., and Winter M., “Perceived Fatigue and Cognitive Performance Change in Multiple Sclerosis: Uncovering Predictors Beyond Baseline Fatigue,” Multiple Sclerosis and Related Disorders 32 (2019): 46–53. [DOI] [PubMed] [Google Scholar]
- 17. Schumann A., Di Giuliano M., Schulz S., et al., “Mapping the Brain's Fatigue Network: A Transdiagnostic Systematic Review and Meta‐Analysis on Functional Correlates of Mental Fatigue,” Brain Communications 7, no. 5 (2025): fcaf315. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 18. Meijboom R., Foley P., MacDougall N. J. J., et al., “Fatigue in Early Multiple Sclerosis: MRI Metrics of Neuroinflammation, Relapse and Neurodegeneration,” Brain Communications 6, no. 5 (2024): fcae278. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 19. Franceschini A., Preziosa P., Valsasina P., et al., “Lesion Location and Functional Connections Reveal Cognitive Impairment Networks in Multiple Sclerosis,” Annals of Clinical Translational Neurology 13, no. 2 (2026): 242–255. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 20. Capone F., Collorone S., Cortese R., Di Lazzaro V., and Moccia M., “Fatigue in Multiple Sclerosis: The Role of Thalamus,” Multiple Sclerosis Journal 26, no. 1 (2020): 6–16. [DOI] [PubMed] [Google Scholar]
- 21. Johnen A., Bürkner P. C., Landmeyer N. C., et al., “Resolving the Cognitive Clinico‐Radiological Paradox: Microstructural Degeneration of Fronto‐Striatal‐Thalamic Loops in Early Active Multiple Sclerosis,” Cortex 121 (2019): 228–239. [DOI] [PubMed] [Google Scholar]
- 22. Amin M. and Ontaneda D., “Thalamic Injury and Cognition in Multiple Sclerosis,” Frontiers in Neurology 11 (2021): 623914. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 23. Manca R., Sharrack B., Paling D., Wilkinson I. D., and Venneri A., “Brain Connectivity and Cognitive Processing Speed in Multiple Sclerosis: A Systematic Review,” Journal of the Neurological Sciences 388 (2018): 115–127. [DOI] [PubMed] [Google Scholar]
- 24. Akaishi T., Fujimori J., and Nakashima I., “Basal Ganglia Atrophy and Impaired Cognitive Processing Speed in Multiple Sclerosis,” Cureus 16, no. 1 (2024): e52603. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 25. Biasi M. M., Manni A., Pepe I., et al., “Impact of Depression on the Perception of Fatigue and Information Processing Speed in a Cohort of Multiple Sclerosis Patients,” BMC Psychol 11, no. 1 (2023): 208. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 26. Golan D., Doniger G. M., Wissemann K., et al., “The Impact of Subjective Cognitive Fatigue and Depression on Cognitive Function in Patients With Multiple Sclerosis,” Multiple Sclerosis Journal 24, no. 2 (2018): 196–204. [DOI] [PubMed] [Google Scholar]
- 27. Borragán G., Gilson M., Atas A., et al., “Cognitive Fatigue, Sleep and Cortical Activity in Multiple Sclerosis Disease,” Frontiers in Human Neuroscience 12 (2018): 378. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 28. Barrouillet P., Bernardin S., and Camos V., “Time Constraints and Resource Sharing in Adults' Working Memory Spans,” Journal of Experimental Psychology. General 133, no. 1 (2004): 83–100. [DOI] [PubMed] [Google Scholar]
- 29. Thompson A. J., Banwell B. L., Barkhof F., et al., “Diagnosis of Multiple Sclerosis: 2017 Revisions of the McDonald Criteria,” Lancet Neurology 17, no. 2 (2018): 162–173. [DOI] [PubMed] [Google Scholar]
- 30. Beck A. T., Steer R. A., and Brown G., Beck Depression Inventory‐II (Psychological Corporation, 1996). [Google Scholar]
- 31. Krupp L. B., LaRocca N. G., Muir‐Nash J., and Steinberg A. D., “The Fatigue Severity Scale: Application to Patients With Multiple Sclerosis and Systemic Lupus Erythematosus,” Archives of Neurology 46, no. 10 (1989): 1121–1123. [DOI] [PubMed] [Google Scholar]
- 32. Spielberger C. D., State‐Trait Anxiety Inventory for Adults (Mind Garden, 1983). [Google Scholar]
- 33. Lee K. A., Hicks G., and Niño‐Murcia G., “Validity and Reliability of a Scale to Assess Fatigue,” Psychiatry Research 36, no. 3 (1991): 291–298. [DOI] [PubMed] [Google Scholar]
- 34. Wilson A., Dollman J., Lushington K., and Olds T., “Reliability of the 5‐Min Psychomotor Vigilance Task in a Primary School Classroom Setting,” Behavior Research Methods 42, no. 3 (2010): 754–758. [DOI] [PubMed] [Google Scholar]
- 35. Hrabovecky J. M., De Tiège X., Samyn C., et al., “Is a Single Calibration for the TloadDback Cognitive Fatigue Induction Task Reliable?,” Frontiers in Psychology 16 (2025): 1561819. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 36. Guillemin C., Read J., Requier F., et al., “Pupil Response Speed as a Marker of Cognitive Fatigue in Early Multiple Sclerosis,” Multiple Sclerosis and Related Disorders 65 (2022): 104001. [DOI] [PubMed] [Google Scholar]
- 37. Tobii Pro Fusion Eye Tracker, Version 1 (Tobii Pro, 2020), https://www.tobii.com/products/eye‐trackers/screen‐based/tobii‐pro‐fusion. [Google Scholar]
- 38. Desikan R. S., Ségonne F., Fischl B., et al., “An Automated Labeling System for Subdividing the Human Cerebral Cortex on MRI Scans Into Gyral Based Regions of Interest,” NeuroImage 31, no. 3 (2006): 968–980. [DOI] [PubMed] [Google Scholar]
- 39. Fischl B., “FreeSurfer,” NeuroImage 62, no. 2 (2012): 774–781. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 40. Guo C., Ferreira D., Fink K., Westman E., and Granberg T., “Repeatability and Reproducibility of FreeSurfer, FSL‐SIENAX and SPM Brain Volumetric Measurements and the Effect of Lesion Filling in Multiple Sclerosis,” European Radiology 29, no. 3 (2019): 1355–1364. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 41. Schmidt P. and Wink L., “LST: A Lesion Segmentation Tool for SPM. Manual/Documentation for Version 2.0.15,” (2017).
- 42. JASP Team , “JASP, Version 0.18.03,” (2024), https://jasp‐stats.org/.
- 43. Hayes A. F., Introduction to Mediation, Moderation, and Conditional Process Analysis: A Regression‐Based Approach (Guilford Publications, 2017). [Google Scholar]
- 44. Manjaly Z. M., Harrison N. A., Critchley H. D., et al., “Pathophysiological and Cognitive Mechanisms of Fatigue in Multiple Sclerosis,” Journal of Neurology, Neurosurgery, and Psychiatry 90, no. 6 (2019): 642–651. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 45. Fleming S. M. and Dolan R. J., “The Neural Basis of Metacognitive Ability,” Philosophical Transactions of the Royal Society of London. Series B, Biological Sciences 367, no. 1594 (2012): 1338–1349. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 46. O'Doherty J. P., Dayan P., Friston K., Critchley H., and Dolan R. J., “Temporal Difference Models and Reward‐Related Learning in the Human Brain,” Neuron 38, no. 2 (2003): 329–337. [DOI] [PubMed] [Google Scholar]
- 47. Middleton F. A. and Strick P. L., “Basal Ganglia and Cerebellar Loops: Motor and Cognitive Circuits,” Brain Research. Brain Research Reviews 31, no. 2–3 (2000): 236–250. [DOI] [PubMed] [Google Scholar]
- 48. Bostan A. C., Dum R. P., and Strick P. L., “Functional Anatomy of Basal Ganglia Circuits With the Cerebral Cortex and the Cerebellum,” Progress in Neurological Surgery 33 (2018): 50–61. [DOI] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.
Supplementary Materials
Table S1: Disease parameters and fatigue, vigilance and pupil diameter in PwMS.
Table S2a: Affective factors and fatigue, vigilance and pupil diameter in PwMS.
Table S2b: Demographic and sleep factors and fatigue, vigilance and pupil diameter in PwMS.
Table S3: Cognitive Processing Times and Task Performance with Psycho‐social Factors.
Table S4: Psycho‐affective factors and brain volumes and lesion burden in PwMS.
Table S5: Pre‐ and post‐task measures, brain volumes and lesion burden in PwMS.
Table S6: Correlation analyzes of task‐induced fatigue and fatigability with brain regions in PwMS.
Table S7: Mediation analysis ‐ Corpus callosum.
Table S8: Mediation analysis ‐ Putamen.
Table S9: Mediation analysis ‐ Lesion count.
Data Availability Statement
The data that supports the findings of this study are available on request from the corresponding author. The data are not publicly available due to privacy or ethical restrictions.