Information processing speed modulation by electrical brain stimulation in multiple sclerosis: towards individually tailored protocols

Abstract

Information processing speed is a core cognitive process, highly relevant in everyday life and the most frequent and disabling cognitive symptom in patients with relapsing multiple sclerosis. Correlational evidence from brain imaging suggests involvement of the superior parietal lobe in the speed component of information processing, thereby providing a neurobiological foundation for neuromodulatory interventions. By using regionally specific, focalized transcranial direct current stimulation (tDCS) in healthy individuals and patients with relapsing multiple sclerosis, we provide causal evidence for superior parietal lobe involvement in information processing speed and identified a clinically relevant predictor of tDCS response in patients with relapsing multiple sclerosis. The study employed a registered, randomized, sham tDCS-controlled, three-way-blinded, cross-over trial and a mixed-factors design with eight arms [between-subjects: group (patients/healthy controls; N = 32/group); tDCS polarity (excitatory/inhibitory); within-subjects: stimulation (active/sham tDCS)]. Concurrently with tDCS (1.5 mA; active: 20 min; sham: 40 s), participants completed a computerized version of the Symbol Digit Modalities Test, the current gold standard for quantifying information processing speed impairment in patients with relapsing multiple sclerosis. Data were analysed in a Bayesian framework with generalized linear mixed models. Bayesian modelling provided strong causal evidence of bilateral superior parietal lobe involvement in information processing speed and a double dissociation of stimulation response in patients and controls (i.e. a significant three-way interaction of group × stimulation × polarity). Healthy individuals showed the expected canonical pattern of significantly reduced and increased response latency during anodal or cathodal tDCS, respectively. Across the patient groups, a reversed pattern was found and tDCS response was predicted by baseline Symbol Digit Modalities Test performance. More impaired patients benefited from cathodal tDCS, while less impaired patients benefited from anodal tDCS. For standardized Symbol Digit Modalities Test scores, the transition from beneficial to non-beneficial effects (anodal: < -0.58; cathodal: > -0.70) was consistent across the patient groups. tDCS was well tolerated, with no evidence for differences in mild adverse effects across groups and tDCS conditions. Blinding integrity was confirmed and behavioural outcomes were not explained by factors unrelated to tDCS. Our results provide direct causal evidence for superior parietal lobe involvement in information processing speed in health and disease and suggest that the degree of information processing speed impairment in the patients reflects compensatory or dysfunctional neuroplastic processes that can be counteracted by tDCS in a polarity-specific way. Identified standardized transition scores for the effectiveness of excitatory or inhibitory tDCS will inform future individually tailored stimulation protocols in patients with relapsing multiple sclerosis (trial registration: NCT04667221).

Keywords: Bayesian modelling; focalized transcranial direct current stimulation; information processing speed; relapsing multiple sclerosis; superior parietal cortex.

Graphical Abstract

Figure 1

Study overview. (A) The study comprised three sessions: a neuropsychological baseline assessment, followed by two experimental sessions (either active or sham tDCS); stimulation order was counterbalanced in each participant group (i.e. 50% of participants received active tDCS first, followed by sham tDCS; 50% received sham first, followed by active stimulation). Stimulation sessions were separated by at least 1 week. (B) Procedure of the experimental sessions: nine blocks of the modified SDMT were completed with either active (anodal, cathodal) or sham stimulation. Breaks in between blocks were self-paced but could not exceed 60 s.

Figure 2

Experimental task: examples of (A) coherent and (B) incoherent trials of the modified SDMT. At the top of the image the legend of the current block is displayed, which indicates correct symbol–digit combinations. At the bottom, a ‘probe’ combination is shown and participants had to indicate coherent or incoherent combinations by button press.

Figure 3

Current flow simulations bilateral SPL-tDCS. (A) shows peak current intensity in the target region (bilateral superior parietal cortex) and (B) medial view illustrates current intensities to deeper brain regions. Note: Standard Simulation of Non-invasive Brain Stimulation (SimNIBS) parameters and a Montreal Neurological Institute 152 standard brain were used for simulations. Current flow patterns are shown for the anodal tDCS condition; cathodal stimulation has a reversed polarity but identical current distribution.

Figure 4

Response latency effects. Conditional effects plot based on a GLMM with lognormal link function. Response latency effects of anodal and cathodal stimulation in both subject groups, controlled for session effects, motor slowing (via TMT-A), and the number of correct answers in the paper–pencil SDMT. Compared to sham stimulation, healthy controls exhibited shorter response latencies during anodal stimulation and increased response latencies during cathodal stimulation. Across the groups, this effect was reversed in patients with RMS, i.e. response latencies were shorter during cathodal stimulation and longer during anodal stimulation (relative to sham stimulation; β = −0.10, 95% CI = −0.11 to −0.08, evidence ratio = ∞, N_subjects = 62, N_observations = 52 764).

Figure 5

Conditional effects plot. Based on GLMMs with lognormal link function. The three-way interactions for the association between response latency during either placebo (sham) and anodal or cathodal tDCS are illustrated. (A) Study-specific distribution of raw scores of the baseline SDMT (β = 0.06, 95% CI = 0.04–0.08, evidence ratio = ∞, N_subjects = 62, N_observations = 52 764). (B) Age-corrected SDMT norms (β = 0.06, 95% CI = 0.04–0.08, evidence ratio = ∞, N_subjects = 62, N_observations = 52 764). Controls did not reach z-scores less than −1 indicating that the patient group drives the reversal of the AeCi response. The x-axis shows the raw or z-scores of the baseline. Note the raw SDMT scores were standardized during model computation, but rescaled for plotting (M = 48.15, SD = 11.93). The y-axis shows mean response latency during sham or active (anodal, cathodal) tDCS conditions. Ribbons represent a 95% credible interval. Dots represent mean response latency of individual participants across experimental blocks and baseline SDMT scores of the participants.