BOLD signal variability as potential new biomarker of functional neurological disorders

Highlights • Higher BOLD variability in somatomotor, salience and limbic networks in FND.• Better clinical outcome at follow-up associated with higher variability in SMA.• Higher insular variability at baseline predicted a worse clinical outcome.• BOLD signal variability might present a prognostic and state biomarker for FND.


Fig. S1.
Flowchart illustrating the dropouts and number of subjects excluded based on different criteria during initial assessment and follow up after 8 months.

Control analyses
A control analysis was performed to assess the robustness of our results, whereby we performed the same analysis for the differences in BOLD signal variability between FND and HC with the correction of the effect of depression, anxiety, and psychotropic medication (dichotomized into yes/no including one or more of the following: benzodiazepines, antidepressants neuroleptics, antiepileptics, opioids).
None of the HC used psychotropic medication at the time of the study.32 patients were under current psychotropic medication.To do so, we used BDI, STAI-S and psychotropic medication as additional covariates of no-interest in the voxel-wise t-test.We also quantified the results on a network-level by overlaying significant clusters with the YEO network atlas.

Whole brain longitudinal analysis
To firstly look of the evolution of SDBOLD from T1 to T2 in FND patients, a voxel-wise t-test was performed using age and gender as variables of no interest comparing the scans from T1 to scans at T2.
To correct for multiple comparisons, a family-wise error correction (FEW, P < 0.01) was applied at the cluster level

Control analyses -Corrected for age, gender, depression, anxiety and psychotropic medication
The analysis of the differences in SDBOLD between FND and HC revealed 3 significant clusters (pFWE < 0.05, minimum cluster size = 50 Voxels) and one cluster showing a trend (pFWE = 0.061, cluster size = 42 voxels), where FND patients showed higher SDBOLD compared to HC across brain regions including the insula, the supplementary motor cortex (SMA) and basal ganglia (Fig. S2).Characteristics of these clusters are shown in Table S4.61% percent of the voxels in the control analysis were overlapping with the original analysis.The mapping to the YEO network atlas showed, that the voxels within these clusters were mostly overlapping with the limbic network (55%), somatomotor network (16%) and default mode network (22%).

Whole brain longitudinal analysis
The analysis of the differences in SDBOLD between start of the study and follow-up revealed 3 significant clusters (PFWE < 0.05, minimum cluster size = 50 voxels), where FND patients at follow-up showed higher SDBOLD compared to inclusion.The correlation of the ∆symptom severity and ∆SDBOLD showed a significant negative correlation between ∆CGI-1 and the SMA.A positive ∆CGI-1 means a worse general impression was reported at T2 compared to T1, a positive ∆SDBOLD means a higher SDBOLD at T2 compared to T1. Together this indicates that an improvement of the symptom severity represented by a negative ∆CGI-1 correlates with an increased SDBOLD in the SMA at T2 compared to T1 represented by a positive ∆SDBOLD (Fig. 2).There were no significant correlations with ∆S-FMDRS.
In the predictive GLM (corrected for anxiety and depression) the SDBOLD in the left insula could predict ∆CGI-1 (ß= 0.11, p-value = 0.025), as well as in the right insula (ß= 0.47, p-value = 0.045).Thus, a higher SDBOLD at T1 was linked to an improvement of the CGI-1.There were no significant predictions for the S-FMDRS.

Fig. S4 :
Fig. S4: Correlation of clinical scores and ∆SDBOLD in different Brain regions.(A, B) Correlation of ∆SDBOLD in the right SMA/right Hippocampus and ∆CGI.(C,D) Correlation of of ∆SDBOLD in the left SMA/left Insula and ∆SF-36 general health.

Table S1
Demographic and clinical features at baseline of individuals with functional neurological disorder (FND) who engaged in follow-up compared to those who did not participate.Abbreviations: FUP = Patients who participated in the follow-up, Non-FUP = Patients who did not participate in the follow-up.

Table S2 :
Significant Clusters of voxel-wise contrast analysis FND > HC

Table S3 :
Percentage of voxels in the significant cluster of the selected AAL2 regions.

Table S5 :
Significant clusters of T2 > T1 in control analysis corrected for age and gender