Association of Lesion Location and Depressive Symptoms Poststroke

Julian Klingbeil, MD; Max-Lennart Brandt; Max Wawrzyniak, MD; Anika Stockert, MD; Hans R. Schneider; Petra Baum, MD; Karl-Titus Hoffmann, MD; Dorothee Saur, MD


Stroke. 2021;52(3):830-837. 

In This Article


Demographics and Clinical Characteristics

Descriptive data for demographic variables and behavioral measures are summarized in Table 1. Overall, patients in our cohort were mildly affected with a mean Barthel-Index of 80.6 and a mean NIHSS of 2.5. Six months poststroke, 53 patients (19.6%) had a HADS-D score >7 indicative for PSD. Antidepressant medication was recorded in 12 patients (3.7%) at discharge from hospital (t1) and in 18 patients (6.5%) at t2. Of these, 50.5% and 55.5% were still rated as depressed at t1 and t2, respectively. Patients lost to follow-up did not differ significantly from the remaining cohort in lesion volume (Mann-Whitney U test: P=0.50) or HADS-D at t1 (P=0.73). Results of the logistic regressions are summarized in Table 2. In the separate logistic regressions, symptoms of depression (HADS-D) and anxiety (HADS-A) and severity of stroke symptoms (NIHSS) at t1 increased the risk for of PSD+ at t2. Lower scores on the Barthel-Index as a measure of worse functional status were not significant. Age, lesion volume, and sex were not associated with PSD. In the multiple logistic regression, HADS-D at t1 was the only independent predictor of PSD at t2, but the explained variance of the model was relatively small (Nagelkerke R 2=0.22, Cohen's f2=0.05).

Voxel-based Lesion Behavior Mapping

Maximum lesion overlap (22/270) was located in the right insula (Figure 2A). Only voxels affected in ≥5 patients (346 030 voxels, corresponding to 7.38% of all voxels) could be submitted to VLBM. Thus, large sections of the left and right hemisphere, the brain stem, and cerebellum were not included in the analysis (Figure 2B). For PSD+ compared with PSD−, no significant lesion behavior association was identified (p(FWE) <0.05 corresponding to z < −4.31, peak z = −4.29; Figure II in the Data Supplement). When HADS-D was introduced as a continuous rather than binary variable for depressive symptoms, VLBM identified a significant cluster spanning 28 voxels (peak MNI coordinates: −35/17/−19 mm, z = −4.21) in the left orbitofrontal cortex (Harvard Oxford atlas) where damage was associated with greater depressive symptoms (p(FWE)<0.05, corresponding to z < −4.17; Figure 3). The cluster largely locates to the pars orbitalis of the left inferior frontal gyrus (Anatomy toolbox). It is surrounded by further voxels with high but not significant z-values located along the anterior border of the lesion overlap within the left frontal lobe spanning the orbitofrontal cortex, inferior frontal gyrus, the frontal operculum cortex, and the anterior insula. For the significant cluster, Cohen's ds was above 2.2 (Figure IIIB in the Data Supplement). With respect to the accepted threshold of 0.8, this indicates a very large effect size.

Figure 3.

Voxel-based lesion behavior mapping (VLBM) results for depressive symptoms.
Results of VLBM analysis for depressive symptoms as a continuous measure in Hospital Anxiety and Depression Scale-depression subscale at t2. Brightness indicates absolute z value. Values <0 are given in yellow and >0 in cyan. Significance threshold with p(family wise error) <0.05 was z < −4.17 (corresponding to 28 voxels highlighted in red around MNI [Montreal Neurological Institute] coordinate −35/17/−19 mm). This figure appears in color online.