A clinically significant mental health disorder occurs in greater than one-third of poststroke patients and may be associated with increased morbidity and mortality [
Although there has been a previous systematic review on associations of PSA [
Our review updates and complements an earlier systematic review on the associations of PSA by Menlove et al. which reported searches from inception to May 2014. Their search strategy can be found in their primary publication [
We applied our search strategy from Jan 2014 to 19th July 2015 using the databases MEDLINE, EMBASE, CINAHL PLUS, AMED, PsychINFO, and ProQuest dissertation. There was an overlap in the search period (four months) with Menlove’s review to ensure studies that were in print were not missed. We used search terms synonymous with stroke and anxiety obtained from a Cochrane review of interventions for PSA [
One author (FW) screened all titles and excluded the obviously irrelevant. The remaining citations were exported to EndNote and two authors (FW and HYC) independently screened the abstracts to decide which to obtain as a full text. The two authors independently read all full texts and decided which studies fulfilled the eligibility criteria. Any discrepancies were dealt with through discussion.
Our inclusion criteria were as follows: (1) studies on stroke patients including ischaemic, haemorrhagic, or unspecified stroke subtype; (2) incidence studies, cohort studies, cross-sectional studies, case control studies, or case series that made use of consecutive patient recruitment within clearly defined geographical and time-limited boundaries; (3) assessed anxiety using a validated anxiety scale or clinical diagnostic criteria in a psychiatric interview; (4) used regression analyses; and (5) reported in English.
Studies were excluded if they (1) had mixed populations (unless separate results for stroke patients were reported); (2) were limited to select patient characteristics (such as age, gender, and lesion side); (3) used retrospective recruitment or reporting of mood; (4) did not measure anxiety specifically; (5) contained insufficient data for reporting of associations; (6) included participants <18 years old; (7) were case reports or included <10 participants; (8) were intervention studies; or (9) included >25% of participants with TIA rather than stroke.
Two authors (FW and HYC) independently extracted data onto electronic tables. Variables included demographics, time points, and measures for assessing anxiety, proportion who had PSA, statistical methods for analysing associations, and factors including PSD, prestroke anxiety, prestroke depression, locus of control, coping, confidence, fatigue, and sleep disturbance.
The quality of each study was assessed independently by two authors (FW and HYC) using The Strengthening the Reporting of Observational studies in Epidemiology (STROBE) checklist [
We included studies [
The authors conducted this systematic review within the Centre for Cognitive Ageing and Cognitive Epidemiology (CCACE) which is funded by the Medical Research Council (MRC) and the Biotechnology and Biological Sciences Research Council (BBSRC).
In order to carry out a meta-analysis of the association between PSA and the factors of interest, we calculated odds ratios (ORs) and 95% confidence intervals (95% CI) from raw data where possible. For meta-analysis, if raw data were not available, ORs and 95% CIs reported in the studies were used. If only correlation coefficients were available, these were converted into ORs using conventional methods [
Random-effects modelling was used to determine the summary estimate of OR [
Our searches identified 2061 studies. We obtained 100 full texts and six [
Flow diagram showing study selection process.
The 24 studies recruited a total of 15448 participants, with a mean age ranging from 51.7 to 75.2 years (though four studies did not report mean age [
To assess the presence of anxiety, 13 studies used the Hospital Anxiety and Depression Scale (HADS-A) [
A range of cut-offs were used in the thirteen studies that measured anxiety using HADS-A. Seven used a cut-off of ≥8 [
Sixteen studies excluded patients with moderate or severe cognitive or communication impairment, including aphasia [
The quality of studies is summarised in Supplemental Table I. According to the STROBE checklist the background and aims of the majority were clear, and the discussion was satisfactory, with weaknesses mainly in the methodology: two out of 24 studies provided sample size calculations; three explained how missing data were addressed; six indicated the number of participants with missing data for each variable of interest; two described some efforts to address potential sources of bias; and three out of 11 cohort studies described how loss to follow-up was addressed. Ten studies gave sources of funding.
We were only able to carry out meta-analysis on the association of PSA with PSD as the other factors did not have sufficient data.
Thirteen studies (
Of the 13 studies which had reported the association between PSD and PSA and the additional study from which we were able to calculate an OR, 13 [
Random-effects meta-analysis for the association between poststroke anxiety and depressive symptoms. Horizontal axis represents the odds ratio (OR) comparing the occurrence of depressive symptoms in patients with and without poststroke anxiety. Horizontal error bars represent the 95% confidence interval (95% CI) of the OR from individual studies. The vertical line represents the summary OR. Symbol size represents the natural log of the number of participants in that study.
Three studies reported adjusted ORs [
The asymmetrical funnel plot (Figure
Funnel plot for publication bias. The vertical axis represents the inverse standard error. The horizontal axis represents the natural log odds ratio of the associations between poststroke anxiety and depressive symptoms. The vertical bar represents the summary estimate of odds ratios.
One study [
One study [
One study [
Two studies (
Four studies reported the association with cognitive impairment. Two studies [
One study [
Meta-analysis of 13 studies of 2408 patients demonstrated a statistically significant association between PSA and PSD (
To our knowledge this is the only meta-analysis on the associations of PSA and psychological factors; although a systematic review was published previously [
Our review has some limitations. Firstly, we could not include five studies; four of which were published only in abstract form and the other as a dissertation. We contacted the authors but received no response. Secondly, in our meta-analysis of the association with poststroke depressive symptoms, only three studies reported ORs that were adjusted for potential confounding factors; therefore the results from the 10 other studies included may overestimate the associations found. When we analysed the studies with adjusted and unadjusted ORs separately we found the unadjusted group had a higher summary effect than the adjusted group although there was no significant heterogeneity between the groups. There were also limitations of the individual studies: most did not report sample size calculations or describe efforts to address potential sources of bias; very few studies provided reasons for nonparticipation at each stage or explained how missing data was addressed; most studies were hospital based and 16 studies excluded patients with significant cognitive or communication impairment, limiting generalisability and making it difficult to draw conclusions on the association of PSA with cognitive impairment.
Our results and another review [
The association we found at a single time point does not allow us to determine the direction of causality. It is plausible that another factor may cause both depression and anxiety. Therefore more longitudinal studies exploring possible direction of causality are required.
Locus of control, coping strategies, and confidence are potentially modifiable factors that could be targeted for interventions to treat anxiety after stroke. High levels of external locus of control have been found to correlate with higher levels of anxiety in patients with a generalised anxiety disorder and in patients with multiple sclerosis (MS) and ovarian cancer [
Significant associations between anxiety and depression after stroke suggest that PSD should be screened for in patients with PSA and vice versa. The presence of anxiety or depression before stroke may be useful in identifying those most at risk of developing PSA. Other potential modifiable factors that could be targeted in interventions include fatigue and sleep disturbance. Further research into the association of PSA with potential treatment targets is required.
The authors declare that they have no competing interests.
Thanks are due to Sheila Fisken (Academic Support Librarian for the University of Edinburgh) for assistance with the search strategy and Maureen Harding for obtaining some of the full texts. This paper was funded by Medical Research Council (MRC) and the Biotechnology and Biological Sciences Research Council (BBSRC), Grant no. MR/K026992/1. Simiao Wu received funding from Chinese Government and University of Edinburgh and Ho-Yan Yvonne Chun from Stroke Association Princess Margaret Fellowship and Chief Scientist Office Clinical Academic Fellow.