Patients with stroke often develop mood and emotional disturbances, including post-stroke depression (PSD), post-stroke emotional incontinence (PSEI), and post-stroke anger (PSA) [1-3]. A recent meta-analysis showed that about one-third of patients have PSD at any time within 5 years after stroke [4]. The prevalence of PSEI and PSA in the early phase has been reported to be 6% to 34% and 15% to 35%, respectively [3].

The pathogenesis of these mood and emotional disturbances is similar, but not identical. For PSEI and PSA, serotonergic dysfunction caused by stroke lesions appears to play an important role [3]. Although this could also account for the development of PSD, the pathophysiology of PSD appears more complex because of the strong involvement of familial/psychosocial factors [5]. Thus, the pathophysiology of PSD may differ according to the time point after the stroke [5,6].

The prevalence of and associated factors for these disturbances in the long-term period (LTP) have been rarely studied [3,4]. The aim of this study was to investigate the prevalence of and associated factors for PSD, PSEI, and PSA at LTP, using follow-up data from the Efficacy of Early Administration of Escitalopram on Depressive and Emotional Symptoms and Neurologic Dysfunction After Stroke (EMOTION) trial, a double- blind, placebo-controlled, randomized study.

Seven of 17 hospitals that had previously participated in the EMOTION trial were included in this sub-study. The other centers could not perform this study because of the investigators being transferred to other hospitals, loss of interest, or practical difficulties in performing the study (e.g., IRBs allowing direct but not telephone interviews). The seven hospitals had previously enrolled 356 patients in the EMOTION trial, 222 of whom consented to participate. The major reasons for non-participation were the inability to establish contact (46 patients, 34.3%), patients’ reluctance to participate (31 patients, 23.1%), missing data (26 patients, 19.4%), and patients passing away (22 patients, 16.4%) (Figure 1).

Fifty-one of 222 patients were interviewed face-to-face at the outpatient clinic, and the others were interviewed via telephone by a trained investigator. The investigators obtained written consent (or verbal consent on the telephone) and interviewed them using a pre-established questionnaire. If possible, relatives or caregivers were also interviewed to confirm the patient’s responses. All interviews were conducted between May 2017 and May 2018.

The duration from the index stroke to LTP ranged from 35 to 83 months (60 [range, 50 to 70]). All participants lived in their homes, except for four who were in the sanitarium. Nine of 222 patients (4.1%) experienced recurrent stroke. Twenty-nine of 222 (13.0%) were taking antidepressants, and 28 (12.6%) were regularly visiting psychiatric clinics. Of 99 patients with PSD at LTP, only 20 (20.2%) were being treated. A total of 78 patients (35.1%) had low social support. There were no significant differences in baseline characteristics between participants and non-participants, except that non-participants were older and had higher MADRS and NIHSS scores (Table 1).

Compared to the mRS 6 months post-stroke (3.0 [2.0 to 4.0]), the median mRS at LTP (1.50 [0.75 to 3.00], P<0.01) decreased (Figure 2A), whereas the prevalence of PSD (35.6% to 44.6%, P=0.03) and mean MADRS (6.2 to 8.7, P<0.01) increased (Figure 2B). The prevalence of PSEI (11.7% to 6.3%, P=0.05) and mean anger score at LTP (21.6 to 16.2, P<0.01) decreased (Figure 2C and D).

We studied the relationships between demographic and clinical factors, and the use of escitalopram, and PSD, PSEI, and anger score at 6 months and at LTP in a simple regression analysis (Table 2). Variables showing significant effects (P<0.05) were baseline PSEI and NIHSS for PSEI at 6 months; left side lesion, baseline PSEI, and NIHSS for PSEI at LTP; and female sex, baseline anger score for PSA at both 6 months and LTP. Female sex, baseline MADRS, PSD, and NIHSS were significant factors for PSD at 6 months, while age, hyperlipidemia, baseline MADRS, PSD, and NIHSS were significantly associated with PSD at LTP.

Multiple logistic regression analysis after adjustment for important factors (P<0.1) showed that the following factors were significantly (P<0.05) associated: baseline PSD and NIHSS for PSD at 6 months, hyperlipidemia and baseline PSD for PSD at LTP, baseline PSEI and NIHSS for PSEI at 6 months, left-side lesion and baseline PSEI at LTP, age and baseline anger score for PSA at 6 months, and baseline anger score and NIHSS for PSA at LTP (Table 3). There was no autocorrelation or multicollinearity in the multiple regression test.

Figure 3 shows the relationships of ESSI with MADRS and with anger score at LTP. The correlation test showed a significant negative relationship between MADRS and ESSI (ρ=- 0.39, P<0.01) and between anger score and ESSI (ρ=-0.28, P<0.01).

Table 4 shows the effect of low social support and mRS at LTP on PSD, PSEI, and anger scores at LTP. In the simple regression test, the effect of low social support and mRS on PSD was statistically significant. The significance remained in the multiple regression test after adjustment for age, sex, baseline PSD, baseline NIHSS, regular psychiatric clinic visits at LTP, antidepressant use at LTP, and recurrent stroke at LTP. However, the effect of low social support on PSD at LTP was greater than that of mRS. There was no interaction between low social support and mRS at LTP (P=0.53). Neither mRS nor low social support was significantly associated with PSEI at LTP. For PSA, the effect of low social support on the anger score at LTP was significant in the adjusted test. We additionally tested the relationship of ESSI or mRS with “PSD treatment” among patients with PSD at LTP. ESSI was positively associated with psychiatric clinic visits in the logistic regression test (odds ratio, 1.13; 95% confidence interval, 1.02 to 1.25; P=0.03), while mRS was not (P=0.11).

To our knowledge, our study is the first to investigate the longterm (average 5 years) changes in PSD, PSEI, and PSA. We found that the prevalence of PSD and mean MADRS score gradually increased over time. We had previously found that the MADRS scores at the early stage of stroke decreased along with improvement in neurological deficits, probably due to psychological responses associated with improving neurological deficits [7]. Our results are not consistent with a previous longitudinal study from the South London Stroke Register (SLSR). In this study, the prevalence of PSD defined by depression subscale score >7 in the Hospital Anxiety and Depression Scale was stationary over time: 33% (30% to 36%), 28% (25% to 30%), and 31% (27% to 34%) at 3 months, 1 year, and 5 years after stroke, respectively [17]. Direct comparisons should be made cautiously because our study and the SLSR study are methodologically different. We conducted direct or phone interviews, whereas a postal interview was used in the SLSR study. While 222 of 356 patients (62.4%) were followed up at LTP in our study, only 585 out of 3,689 (15.9%) were followed up at 5 years in the SLSR study.

Nevertheless, the gradual increase in the degree and prevalence of PSD over time in our study needs to be discussed. Although mRS scores were still independent factors associated with PSD at LTP (Table 4), they gradually decreased over time in our study (Figure 2). Thus, the increasing prevalence of PSD at LTP is unlikely to be attributed to worsening functional disability [18]. We found that low social support was significantly associated with PSD at LTP. Although we did not investigate ESSI at earlier time points, the strong relationship between ESSI and MADRS at LTP (Figure 3) suggests that the lack of social support may be one of the reasons for the persistent or increasing prevalence of PSD. We also found that only approximately 20% of patients with PSD were under treatment. Although the treatment status was not described in the SLSR study, this factor may have contributed to the difference. In Korea, some depressive patients do not visit psychiatric clinics because of the social stigma attached to psychiatric diseases [19]. More elderly Koreans regard depression as a “personal weakness” or “normal aging” instead of a “disease” than Americans [20]. It was reported that the use of antidepressants in Korea was one of the lowest among OECD countries [21]. All these factors may have contributed to the prevalent PSD at LTP in our cohort.

We noted that aside from baseline PSD and NIHSS scores, hyperlipidemia was a factor associated with PSD at LTP (Table 3). Although a study from Taiwan showed that patients with hyperlipidemia had a high risk of depression [22], a meta-analysis of the relationship between vascular risk factors and late-life depression did not show such an association [23]. Further studies are required to confirm the relationship between PSD and hyperlipidemia.

We found that the prevalence and severity of PSEI decreased from 6 months post-stroke to LTP (Figure 2). This result was consistent with a report that revealed that the prevalence of PSEI decreased from 21% at 6 months post-stroke to 11% at 12 months post-stroke [24]. Numerous studies have shown that PSEI is pathophysiologically closely associated with serotonergic system dysfunction in the brain in stroke patients [1,25-28]. Thus, the decreasing prevalence of PSEI at LTP may be attributed to the recovery of the damaged serotonergic system in the brain over time. Unlike PSD, PSEI was not associated with ESSI in our study. The anger score also decreased from 6 months post-stroke to LTP (Figure 2). This may also be explained by the improved brain serotonergic system over time. However, unlike PSEI, the anger score was associated with low social support (Figure 3), although the relationship was not as strong as that with PSD. It has been shown that although PSA is closely associated with brain serotonergic system dysfunction, it is also related to depression and frustration [3].

This study has a few limitations. First, since only Korean patients were enrolled, it may be difficult to apply the results to other ethnicities. Second, as we examined social support only once at LTP, its impact on patients’ emotions at 6 months remains unknown. Third, we could not categorize lesion location precisely because the number of patients with PSD, PSEI, and high anger score were small. Fourth, since LTP was not pre-defined in this sub-study, the range of LTP was rather broad (35 to 85 months). Finally, many patients did not participate in the study. Compared with participants, non-participants were older and had higher initial MADRS and NIHSS scores (Table 1). In our study, multiple regression analysis showed that baseline PSD, but not MADRS score, was associated with PSD at LTP, and the initial NIHSS score was associated with PSA at LTP. Although it is difficult to precisely assess the possible impact of these differences on our results, the frequencies of post-stroke mood/emotional disturbances at LTP may have been underestimated in our study. This is an inherent problem in the research on depression; individuals with depression are more cautious and hesitant about clinical trial participation than those without [29]. However, we do not think that this limitation greatly influenced our main findings, as our primary aim was to observe the long-term change in post-stroke emotional disorders rather than assess the exact prevalence of these symptoms in a certain period.

Despite these limitations, our data showed that the prevalence and degree of PSEI and anger score decreased, whereas those of PSD increased at LTP in our cohort. Our results have several practical implications. Physicians may consider cessation of medications for PSEI at LTP but need to be aware of the increased prevalence of PSD. As PSD and PSA are associated with a lack of social support, strategies to improve social support may have to be developed to prevent and manage PSD and PSA at LTP.