This systematic review and meta-analysis summarizes the most up-to-date information on a range of imaging markers associated with PSD and PSA during the acute, post-acute, and chronic stroke phase. Meta-analyses indicated that PSD in the post-acute phase was significantly more frequent in patients with frontal or basal ganglia lesions. No significant association was found between PSD and lesion laterality in the post-acute and chronic stroke phase. Nevertheless, it is of interest to mention that left-sided stroke occurred more often in the PSD group in the acute phase. This result became insignificant after the inclusion of four recent large studies (Chen et al. 2016; Metoki et al. 2016; Wei et al. 2016; Zhang et al. 2016), which differed from the other studies in that they reported a relatively low PSD prevalence (median 18.6%, IQR 17.4–30.2). Frequency of PSD was equal for ischemic and hemorrhagic stroke in all stroke phases, but PSA was more frequent after hemorrhagic stroke in the acute phase, whereas it was more frequent after ischemic stroke in the post-acute phase. Since only four PSA studies were available, this finding should be interpreted with caution. Also, PSA did not depend on lesion laterality or location, but again the amount of available PSA studies was small in general.

Our meta-analysis updates and extends previous studies. The meta-analysis by Wei et al. (2015) on lesion laterality and PSD found a significant association between right hemispheric lesions and risk of PSD in the post-acute stroke phase (1–6 months). In contrast to Wei et al. (2015), we defined the post-acute period as 15 days to 6 months, which could explain the difference in results. In agreement with Caeiro et al. (2013a) the prevalence of PSA was not associated with lesion laterality. Both meta-analyses did not study associations with markers other than lesion laterality and lesion type, while the review of van Dalen et al. (2013) evaluated associations between PSA and lesion location only qualitatively and concluded that no clear association could be found.

2011c 2011 2001 2015b 2015a 4 Studies Phase Circuits - network PSD Terroni et al. (2011) Acute Disruption of limbic-cortical-striatal-pallidal-thalamic circuit, Medial PFC dysfunction Yang et al. (2015b) Acute Frontal lobe, insula, limbic system, parietal lobe, basal ganglia, temporal lobe Vataja et al. (2001) Post-acute Higher number and lesion volume in (left) prefronto-subcortical circuit Tang et al. (2011c) Post-acute Lesions in frontal subcortical circuits PSA Yang et al. (2015a) Acute Limbic system, basal ganglia, insula, frontal, temporal, parietal, occipital lobe The present findings suggest that lesion location is an important risk factor for PSD in the post-acute stroke phase. However, in the past few years the hypothesis of PSD and PSA being associated with damage to specific lesion locations has been shifted to the idea that damage to a neuronal network involved in affect is underlying the development of PSD and PSA (Tang et al.; Terroni et al.; Vataja et al.), with different sub-circuits involved in PSD (Yang et al.) and PSA (Yang et al.), see Table. DTI is a promising tool to identify more accurately how these brain networks are affected after stroke. The qualitative overview of imaging markers associated with PSD and PSA showed that not only direct stroke-related features such as lesion location, lesion volume, and number of lesions, but also other neurovascular, non-directly stroke-related but often co-occurring features, such as degree of WMH, cerebral microbleeds, and atrophy, were frequently associated with PSD. With respect to PSA, associations with degree of WMH, lesion volume, and number of lesions were found in some extent. Co-occurring vascular lesions may make a stroke patient more vulnerable for developing PSD and PSA. Therefore, future studies should focus on a broader range of imaging markers, including lesion volume, atrophy, WMH, and cerebral microbleeds, and also how lesion-related markers may interact with co-occuring indirect vascular markers. Besides, advanced imaging techniques (e.g. DTI, fMRI) are needed to evaluate how microstructural abnormalities and changes in functional connectivity contribute to the development of PSD and PSA.

Our study has the following strengths. A large amount of publications on PSD were identified, resulting in a rich pooled cohort of studies that were not included in earlier meta-analyses (Chen et al. 2013, 2016; Gozzi et al. 2014; Jiang et al. 2014; Metoki et al. 2016; Saxena and Suman 2015; Terroni et al. 2015; Wei et al. 2016; Wichowicz et al. 2015; Zhang et al. 2016). Furthermore, beside information on lesion laterality, also data on other imaging markers was retrieved for quantitative and qualitative analysis. Therefore, the present review provides an up-to-date and extended overview of findings on the association between imaging markers and risk of PSD and PSA.

One limitation of the present study was the small amount of studies on PSA, which made it difficult to perform sub-analyses. Therefore, future studies are needed on imaging markers of PSA, covering a broad range of imaging markers. Nevertheless, as heterogeneity was small between PSA studies, we believe that the results are still of importance, but should be interpreted with caution as the generalizability and validity is compromised in comparison with meta-analyses including a larger amount of studies. In addition, moderate to high (unexplained) heterogeneity between studies in some meta-analyses indicated large differences in methodology between studies. Particularly the use of different scales and cut-offs to define the presence of depression and apathy and different imaging methods (CT vs. MRI) are of influence on the comparability of findings. Also differences in eligibility criteria (e.g. exclusion of patients with aphasia, differences in age range) can create heterogeneity among studies. Meta-regression analyses were performed to identify potential sources of heterogeneity, and only study phase for PSD studies and imaging method for PSA studies could be identified. However, in addition to the included variables, also other potential variables (e.g. years of education, cognitive status), that could not be included in the analyses due to the large variability in the methods and availability of data between studies, might explain some of the between-study difference in effect estimates. Therefore, we performed random-effects meta-analyses, which take the heterogeneity between studies into account.