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


In this large prospective VLBM study, we show that lesions in the left frontal cortex increase the risk of depressive symptoms 6 months poststroke while lesions within the right hemisphere are unrelated to depressive symptoms. In the following, we argue for the importance of damage to the left frontal cortex as a clinically relevant biological factor in a biopsychosocial model of PSD.

One fifth of our patients presented with depressive symptoms 6 months after stroke. One explanation for this rate, which is lower than the 30% reported in the literature,[1,11] is the exclusion of patients with prior strokes or history of depression from our study. In these patient groups, a higher prevalence of PSD has been observed.[25] In accordance with previous studies, we found that severer deficits in the first weeks after stroke correlate with a higher risk for depression after 6 months.[1,3] Lesion volume, however, was not associated with depression after 6 months which was unexpected,[11] but has been observed before.[26] The main reason might be the large proportion of patients with smaller lesions in our cohort. Based on the longitudinal assessment, we were able to determine that higher HADS-D and HADS-A scores within the first weeks increase the risk of depressive symptoms after 6 months. Indeed, as demonstrated by multiple logistic regression, only higher HADS-D scores in the acute phase were associated with depression after 6 months. In line with this, the application of HADS or the similar PHQ-9 (Patient Health Questionnaire-9) has already been proposed as an early screening tool to detect patients at risk.[27]

The primary focus of our study, however, was to identify neural substrates of depressive symptoms poststroke. We used VLBM to test whether specific lesion locations are associated with depressive symptoms and identified a cluster in the left frontal cortex (red voxels in Figure 3). Despite better overall lesion coverage, no right hemisphere voxels were significantly associated with depressive symptoms. Quite the contrary, for the vast majority of right hemisphere voxels analyzed, we observed a reversed direction with positive z-values (cyan voxels in Figure 3). Because of the mass univariate approach with correction for multiple comparisons, there is an inherent trade-off between false positive and false negative results in VLBM analyses and VLBM is especially vulnerable to false negative results (type-II-errors).[28] With permutation-based testing and FWE-correction at voxel-level, we used sufficiently strict and state-of-the-art methods to control for false positive results and are thus confident that the identified association of lesions in the left frontal cortex and depressive symptoms is a true positive finding.[12] In addition, the opposite direction of statistical measures in the right hemisphere point to a true negative result rather than a consequence of low power or a too conservative statistical approach. To this end, our results clarify on the question of whether lesion location contributes to depressive symptoms. The literature on lesion location and PSD was not only divided on the relevance of left frontal lesions but also on the role of the right hemisphere with recent meta-analyses both, describing[29] and dismissing[30] associations of right hemisphere lesions and PSD. Conflicting results of prior studies have likely arisen from the use of varying techniques with rather low anatomic precision for inference of structure-function relationships. We are therefore confident that our VLBM analysis in a large patient cohort indeed resolves some of the controversies on lesion location and PSD—namely that left frontal lesions but not right hemispheric lesions are associated with greater depressive symptoms. Of interest, our results echo the first studies on lesion location and PSD, which found that greater depressive symptoms were associated with left anterior lesions quantified as distance of the anterior border of the ischemic lesion from the left frontal pole.[9] The dominance of left-sided lesions in PSD also corresponds to most theories on the asymmetrical foundation of emotion.[10,31]

In our cohort, the strongest associations of lesion location and depressive symptoms were found in the pars orbitalis of the inferior frontal gyrus (Brodmann area 47/12). This region may be assigned to both, the orbitofrontal cortex, and the ventrolateral prefrontal cortex (VLPFC).[32] While the bulk of the orbitofrontal cortex was not affected by stroke in our cohort (see Figure 2), lesions in other parts of the VLPFC and the anterior insula surrounding the significant cluster also showed a greater depressive symptom load on average although there was no significant association (brighter yellow voxels in Figure 3). Functionally, the VLPFC and the anterior insula (also jointly referred to as the fronto-insular cortex) are an integral part of the salience network (SN).[33,34] The SN detects and filters salient stimuli and supports a variety of complex brain functions including emotions, communication, social behavior, and self-awareness.[35] All major nodes of the SN are affected in patients with major depressive disorder.[36] Decreased intrinsic connectivity in the SN has been related to major depressive disorder in general.[37] Moreover, lesions in the SN impair cognitive control[38] and, in turn, impaired cognitive control is associated with a more chronic course of PSD.[11] Linked to the potential effects of impaired cognitive control on depression, the only other positive VLBM study for PSD, which was restricted to left hemisphere lesions in 39 patients, identified a region of the central executive network itself, the DLPFC, where lesions caused greater depressive symptoms after stroke.[13] Dysfunction of the DLPFC is a plausible mechanism for causation of PSD because excitation of this region with noninvasive brain stimulation is an established treatment of major depressive disorder and a potential treatment for PSD.[39] Recently, another study based on lesion-network symptom-mapping also identified the left DLPFC to be associated with depression.[15] For this study, various cohorts of patients with brain lesions due to intracerebral hemorrhage, ischemic stroke, or penetrating traumatic injury were pooled. Functional connectivity between lesion locations and a region within the left DLPFC was higher in patients with depression than in those without. Based on the concept that dysfunction in this region (which was spared by lesions in most patients) causes depressive symptoms, a depression circuit was defined based on whole-brain functional connectivity of the DPLFC. By calculating network damage scores, possible prognostic utility in identifying patients at risk for depression was determined for this brain circuit, which largely corresponds to the central executive network. Indeed, patients with lesions in regions defined by positive connectivity to the DPLFC were up to 4× more likely to suffer from depression. The VLPFC maps onto this depression circuit and the fronto-insular cortex identified in our study shares functional connectivity with the DLPFC.[33] To conclude, we interpret these 3 lesion studies as converging evidence for a causal role of the left lateral prefrontal cortex in PSD.

There are several limitations that need to be addressed. First, our primary VLBM analysis that tested for an association of PSD (defined as HADS-D >7) with lesion locations was negative. In light of the significant results when using the raw scores from the HADS-D, the negative result is likely due to the lower statistical power of the Liebermeister test together with the insufficient discrimination about PSD based on a binary cutoff value. Second, while HADS is an established tool to screen for depression and anxiety in the hospital setting, it mainly focuses on anhedonia.[16] It might prove fruitful in the future to differentiate lesion effects on subsymptoms of PSD (eg, rumination, dysphoria, and cognitive control of negative emotions) and to assess the influence of other stroke-related cognitive deficits based on more detailed psychiatric and behavioral testing. Third, the area deemed to have a sufficient overlap for analysis was small (Figure 2B). It did not include several potentially important regions such as the left DLPFC. The exclusion of patients with severe aphasia in combination with the overall greater reluctance of patients with severe deficits to participate certainly aggravated this problem. Therefore, our cohort is dominated by patients with rather small lesions and mild deficits. Poor power in brain regions less often affected by stroke is a general weakness of VLBM.[12] A power calculation based on our results (Figure IIIA in the Data Supplement) identified the left fronto-insular cortex as the only region where our observations are likely to be reproduced, but only when cohorts of the same or even larger size are recruited. This may explain why other VLBM studies of comparable size[14,15] failed to detect a lesion association. Finally, lesions of patients with PSD show weak overlap in previous studies[15] and in our cohort and the majority of lesions lie outside the fronto-insular cortex (see Figure I in the Data Supplement). In these patients, depressive symptoms may rather be caused by psychosocial factors or an affection of the depression circuit. But if the left fronto-insular cortex is damaged, the lesion site itself does seem to be a relevant factor as indicated by the large effect sizes in and around this region (Figure IIIB in the Data Supplement). To which degree these effects are indirect, that is, stroke causes cognitive deficits (eg, memory, language, attention), which trigger reactive depressive symptoms, or direct, that is, stroke causes changes in mood leading to depressive symptoms, remains an open question.