Post-stroke depression in patients with large spontaneous intracerebral hemorrhage

Abstract

Objective:

To determine factors associated with post-stroke depression (PSD) and relationship between PSD and functional outcomes in spontaneous intracerebral hemorrhage (ICH) using prospective data from a large clinical trial.

Methods:

MISTIE III, a randomized, multicenter, placebo-controlled trial, was conducted to determine if minimally invasive surgery with thrombolysis improves outcome compared to standard medical care. Our primary outcome was post-stroke depression at 180 days. Secondary outcomes were change in blinded assessment of modified Rankin Scale (mRS) from 30 to 180 days, and from 180 to 365 days. Logistic regression models were used to assess the relationship between PSD and outcomes.

Results:

Among 379 survivors at day 180, 308 completed Center for Epidemiologic Studies Depression Scale, of which 111 (36%) were depressed. In the multivariable analysis, female sex (Adjusted Odds Ratio [AOR], 95% Confidence Interval [CI]: 1.93 [1.07–3.48]), Hispanic ethnicity (3.05 [1.19–7.85]), intraventricular hemorrhage (1.88 [1.02–3.45]), right-sided lesions (3.00 [1.43–6.29]), impaired mini mental state examination at day 30 (2.50 [1.13–5.54]), and not being at home at day 30 (3.17 [1.05–9.57]) were significantly associated with higher odds of PSD. Patients with PSD were significantly more likely to have unchanged or worsening mRS from day 30 to 180 (42.3% vs. 25.9%; p=0.004), but not from day 180 to 365.

Conclusion:

We report high burden of PSD in patients with large volume ICH. Impaired cognition and not living at home may be more important than physical limitations in predicting PSD. Increased screening of high-risk post-stroke patients for depression, especially females and Hispanics may be warranted.

INTRODUCTION

Post-stroke depression (PSD) is relatively under-studied in spontaneous intracerebral hemorrhage (ICH), the most disabling form of stroke.¹ The highest incidence of PSD occurs within the first year after ICH, estimated at 15–23%, and may decline over time.^2–4

Depression after stroke in general is more prevalent among females, with history of depression, higher stroke severity, and lower level of independence and social support, dependent on study population, and methodology.^5–7 PSD following acute ischemic stroke is independently associated with mortality and lower quality of life (QoL).^8–10 However, PSD is often undiagnosed and left untreated.¹¹ Thus, patients with PSD may not reach their full potential for functional recovery and may experience lower perceived QoL and higher levels of cognitive impairment.^{6, 10–11}

Refining our ability to anticipate PSD has major implications for post-stroke rehabilitation. Investigating associations between early disability and development of PSD, and the impact on a later recovery trajectory is critical to provide a framework for future studies aimed at limiting PSD. No study has examined both early and late changes in functional outcomes based on PSD diagnosis after ICH. The objective of this study was to investigate demographic and clinical features of ICH which correlate best with PSD, and the trajectory of functional outcome improvement or worsening before and after PSD diagnosis. We hypothesized that higher ICH severity, early cognitive impairment, and female sex would be independently associated with PSD and that PSD would impact both early and later outcome trajectory.

METHODS

Parent study design

We performed a secondary analysis of patients enrolled in the MISTIE III trial (Minimally Invasive Surgery Plus Alteplase for Intracerebral Hemorrhage Evacuation), a randomized, controlled, open-label, phase 3 trial that evaluated minimally invasive surgery (MIS) with thrombolysis (n=250) compared to standard medical care (n=249).¹² Patients were aged 18 years or older with spontaneous, non-traumatic, supratentorial ICH of >30 mL, premorbid modified Rankin Score (mRS) 0 or 1, and presentation Glasgow Coma Scale (GCS) ≤14 or National Institutes of Health Stroke Scale (NIHSS) ≥6.¹² Patients randomized to MIS received up to 9 doses of alteplase every 8 hours via intrahematomal catheter until hematoma volume was reduced to ≤15 mL.

Standard protocol approvals, registrations, and patient consents

The MISTIE III trial was performed at 78 hospitals in the US, Canada, Europe, Australia, and Asia following local institutional review board and country ethics approval.¹² The study is registered with ClinicalTrials.gov, NCT01827046. Written informed consent for research was obtained from all participants (or legal representatives or surrogates when applicable). The Johns Hopkins Hospital institutional review board approved the de-identified analysis of previously collected patient data.

Data collection and outcomes

Demographic and clinical data were collected prior to randomization, including age, sex, ethnicity (Hispanic or not), race (white, black, other), and geographic location (North America, Europe, Asia/Australia). The trial neuroimaging center evaluated non-contrast CT and MRI brain scans blinded to outcomes and study arm assignment. Hematoma volumes and intraventricular hemorrhage (IVH) volumes were assessed using semi-automated planimetry. We defined lobar ICH location if selectively involving cerebral cortex, underlying white matter, or both.¹³ Deep ICH was defined as selective involvement of thalami, basal ganglia, or both. End-of-treatment (EOT) volume was defined as the parenchymal hematoma volume at 24 hours post last dose of alteplase (MIS group) and at the median surgical EOT time (3 days) for patients in the standard of care (SOC) group. Pre-randomization CT measurements were as follows: ICH volume and location (affected hemisphere, deep or lobar), and presence of IVH. Baseline co-morbidities included: diabetes, hyperlipidemia, hypertension, cardiovascular disease (coronary artery disease or congestive heart failure), anticoagulation, and prior history of hemorrhage or stroke. Patient location was defined by physical residence at days 30, 180, and 365: home, rehabilitation center, long-term care facility, or acute care. Total number of days in hospital was also collected from admission. Fazekas scale score was used to assess white matter hyperintensities on MRI during the acute phase.¹⁴ Total Fazekas scale score ≥3 or deep Fazekas scale score ≥2 were considered as severe white matter hyperintensities.^15,16

Functional outcome was determined using mRS, dichotomized as good (0–3) vs. poor (4–5), at 30, 180, and 365 days; the latter was the pre-specified primary outcome of the main trial.¹² We also collected Barthel Index (BI), National Institutes of Health Stroke Scale (NIHSS), EuroQol Visual Analogue Scale (EQ-VAS),¹⁷ and Mini-Mental State Examination (MMSE) at 30, 180, and 365 days. The EQ-VAS measures self-reported overall health on a vertical visual analogue scale ranging from 0 (worst possible health state) to 100 (best possible health state). MMSE was dichotomized as normal cognition (MMSE ≥24) and cognitively impaired (MMSE <24).¹⁸

The 20-item Center for Epidemiologic Studies Depression Scale (CES-D) was used to screen for depressive symptoms.¹⁹ The assessment is a self-report depression scale designed for a general population, with scores ranging from 0 (no symptoms) to 60 (extreme symptoms). A score of ≥16 is indicative of depression.²⁰ The assessment was patient reported, and was collected by trained study staff at 180 days. Our study population consisted of patients who had completed CES-D at 180 days. We analyzed the primary binary outcome of CES-D as ≥16 versus CES-D <16 at 180 days.

Statistical Analysis

We summarized normally distributed continuous variables as means with standard deviations, while non-normally distributed variables were reported as medians with interquartile range (IQR). For univariate analyses, we used Wilcoxon rank sum test or Student’s t-test for continuous variables depending on the normality of distribution, and χ2 test or Fisher’s exact test for categorical variables.

For each analysis, patients were grouped into two categories: CES-D and no CES-D at day 180, and with and without improvement (by one or more mRS levels) between days 30 and 180 and days 180 and 365. Clinical characteristics were compared between CES-D and no CES-D and between improved mRS from day 180 to 365 versus no improvement or worsening using univariate analysis with appropriate tests. Variables with a predetermined significance of p-value <0.1 for the outcome of interest as well as variables considered to be clinically associated with depression were entered into logistic regression models and compared using Akaike Information Criterion (AIC) to select the optimal set of covariates for each model.

For the analysis of variables associated with CES-D, the population was divided into two groups: (1) all sites and (2) by four geographic regions in the United States (US) (Midwest, Northeast, South, and West). These models included demographics (age, female sex, and Hispanic ethnicity), ICH severity characteristics (stability ICH volume, presence of IVH, ICH location, and ICH-affected hemisphere), assessments at 30 days (EQ-VAS, NIHSS, impaired cognition, and patient not at home), geographic location (North America vs. other and within four US regions), GCS score at randomization, and two or more co-morbidities. The multivariable model to assess variables associated with improved mRS from day 180 to 365 included CES-D, demographics (age, female sex, and Hispanic ethnicity), ICH severity characteristics (stability ICH volume, presence of IVH, ICH location, and ICH affected hemisphere), assessments at 30 days (EQ-VAS, NIHSS, impaired cognition, and patient not at home), geographic location (North America vs. other and within four US regions), GCS score at randomization, and presence of two or more co-morbidities.

Statistical analyses were performed using Stata (version 14.0, College Station, TX) and R (version 4.0.2, R Project for Statistical Computing). Area under the receiver-operating characteristic curve (AUC) was obtained to estimate the rate of successful classification. Hosmer–Lemeshow test was used to assess goodness of fit for logistic regression models. All analyses were two-tailed, and significance level was determined by p<0.05.

Data Availability Statement

The MISTIE III trial data, including de-identified participant data, is available at the National Institute of Neurological Disorders and Stroke data archive (https://www.ninds.nih.gov/Current-Research/Research-Funded-NINDS/Clinical-Research/Archived-Clinical-Research-Datasets). Those seeking access must complete a NINDS data request form and receive approval.

RESULTS

Of 499 randomized patients, 102 (20.4%) were excluded from the analysis due to death, 96 (19.2%) or loss to follow-up, 6 (1.2%), leaving 397 (79.6%) survivors who completed the 180-day visit. Of these, 308 (77.6%) completed the CES-D; 49 (12.3%) had no CES-D performed due to coma, obtunded, and non-communicative states, 16 (4.0%) had visit assessments completed by proxy, one (0.3%) had an incomplete assessment, and 23 (5.8%) were missing for unknown reasons (Figure 1). Compared to those who completed the CES-D, survivors without CES-D were more likely to be white, non-Hispanic with more severe ICH (larger ICH volume, left-sided lesions and deep location, lower GCS, higher NIHSS at baseline) and higher residual ICH volume post-operatively (Table e-1).

Figure 1: Analysis Population.

Abbreviations: CES-D = Center for Epidemiological Studies Depression scale; IRB = Institutional Review Board; US = United States.

Of 308 who completed CES-D, 246 (79.9%) participated in North America, and 62 (20.1%) participated in European, Australian, or Asian sites. The prevalence of depression (CES-D ≥16) was 111 (36.0%) at 180 days. While Hispanic and black patients were more likely to have depression compared to other racial/ethnic groups, other patient demographics and comorbidities were similar amongst patients with and without depression (Table 1). Patients with depression had higher clinical severity at baseline including higher rates of IVH, right hemisphere ICH, severe white matter hyperintensities, and higher admission NIHSS. Comparison of day 30 and day 180 clinical assessments showed that depressed patients at day 180 were less likely to live at home at day 30 or day 180; had worse performance on mRS, MMSE (day 30 only), NIHSS, and BI; and reported lower quality of life on the EQ-VAS compared to non-depressed patients at both time points. EOT ICH volume was lower in patients without PSD compared to those with (median 22.7 vs. 29 mL; P=0.04).

Table 1:

Univariate associations of depressed versus not depressed mood at 180 days among survivors (N=308)

Characteristics	Depressed (N=111, 36.0%)	Not Depressed (N=197, 64.0%)	p-value
Surgical treatment arm	54 (48.7)	114 (57.9)	0.119
Demographics
Age at consent (years), median (IQR)	62 (50–69)	60 (50–68)	0.667
Female sex, n (%)	44 (39.6)	69 (35.0)	0.420
Ethnicity: Hispanic or Latino, n (%)	19 (17.1)	14 (7.1)	0.006
Race, n (%)
White	84 (75.7)	147 (74.6)	0.045
Black	24 (21.6)	30 (15.2)
Other	3 (2.7)	19 (9.6)
Unknown	0 (0.0)	1 (0.5)
Co-morbidities, n (%)
On anticoagulants	5 (4.5)	12 (6.1)	0.558
Diabetes	33 (29.7)	42 (21.3)	0.099
Hyperlipidemia	38 (34.2)	70 (35.5)	0.819
Cardiovascular disease	16 (14.4)	21 (10.7)	0.331
Hypertension	108 (97.3)	189 (95.9)	0.551
Prior hemorrhage, stroke	4 (3.6)	10 (5.1)	0.560
Pre-randomization CT findings
Stability ICH volume (mL), median (IQR)	45.2 (37.5–55.7)	40.5 (32.1–51.4)	0.013
Stability IVH present, n (%)	72 (64.9)	100 (50.8)	0.017
ICH right hemisphere, n (%)	74 (66.7)	98 (49.8)	0.004
ICH deep location, n (%)	72 (64.9)	108 (54.8)	0.086
Other findings
EOT ICH volume (mL), median (IQR)	29.0 (12.2–42.6)	22.7 (9.6–34.9)	0.038
Severe white matter hyperintensities, n (%)	57 (58.8), n=97	81 (45.5), n=81	0.036
GCS at randomization, n (%)
Mild (13–15)	28 (25.2)	31 (15.7)	0.126
Moderate (9–12)	44 (39.6)	89 (45.2)
Severe (3–8)	28 (25.2)	31 (15.7)
NIHSS total at randomization, median (IQR)	19 (15–22)	17 (13–21)	0.015
Antidepressant medication (prior to randomization), n (%)	0 (0.0)	2 (1.0)	0.537
Antidepressant medication (at 180 days post-randomization), n (%)	4 (3.6)	6 (3.1)	0.751
Assessments at 30 days
Patient not living at home, n (%)	106 (95.5)	156 (79.2)	<0.0001
mRS poor 4–5, n (%)	103 (92.8)	152 (77.2)	0.001
EQ-VAS score, n (%)	50 (25–60), n=93	50 (30–70), n=182	0.031
Impaired MMSE, n (%)	75 (72.8), n=103	101 (55.2), n=183	0.004
BI score, median (IQR)	15 (0–40)	40 (10–70)	<0.0001
NIHSS score, median (IQR)	14 (9–19), n=110	10 (5–16), n=191	<0.0001
Geographic region
North America, n (%)	90 (81.1)	156 (79.2)	0.691
US Region, n (%)
Midwest	26 (29.6)	35 (23.0)	0.023
Northeast	13 (14.8)	40 (26.3)
South	38 (43.2)	45 (29.6)
West	11 (12.5)	32 (21.1)

Abbreviations: BI = Barthel Index, CT = computer tomography, EOT = end of treatment, GCS = Glasgow Coma Scale, ICH = intracerebral hemorrhage, IVH = intraventricular hemorrhage, IQR = interquartile range, mL = milliliters, MMSE = Mini Mental State Examination, mRS = modified Rankin Scale, NIHSS = National Institute of Health Stroke Scale, US = United States

Values are n (%) for binary or categorical measures, median (interquartile range) for continuous measures.

In the multivariable logistic regression model for all sites (Figure 2), factors significantly associated with a higher odds of depression included female sex, Hispanic ethnicity, presence of IVH at onset, right hemisphere lesions, impaired cognition at 30 days, and not being at home at 30 days. The AUC for the model was 0.745. Considering US sites only, female sex, Hispanic/Latino ethnicity, right hemisphere lesions, and not living at home by day 30 remained significant, whereas presence of IVH and impaired MMSE had weaker associations with CES-D. Geographic location, however, was an important variable, with patients living in the South and Midwest having higher odds of post-stroke depression compared to the West, respectively. The AUC for this model was 0.791.

Figure 2: Forest plot of factors associated with PSD at 180 days using baseline variables and clinical assessments at 30 days.

Abbreviations: EQ-VAS = EuroQoL Visual Analogue Scale; GCS = Glasgow Coma Scale; ICH = Intracerebral Hemorrhage; IVH = Intraventricular Hemorrhage; MMSE = Mini Mental State Examination; mRS = modified Rankin Scale; US = United States.

Univariate analysis of association of variables collected at 180 days with PSD are reported in Table e-2. Multivariable analysis of factors associated with PSD at 180 days using baseline and clinical assessments at 180 days are reported in Table e-3. Results were similar to models using day 30 data except that poor mRS 4–5 and EQ-VAS at 180 days were significantly associated with depression, whereas patient location and impaired MMSE were not. The AUC for this model was 0.771. Considering US sites only, female sex, Hispanic/Latino ethnicity, right-sided lesions, and EQ-VAS at 180 days remained significant. Region location was an important variable, with patients living in the South and Midwest having higher odds of PSD compared to the West. The AUC for the model was 0.790.

Figure 3 shows the ordinal mRS at day 30 and the distribution of depression at day 180. Almost all depressed patients at day 180 had poor mRS at day 30, but 152 patients (60%) with poor day 30 mRS were not depressed at day 180. Patients with PSD at day 180 were significantly more likely to have no change or worsening in mRS from day 30 to 180 compared to patients without PSD (42.3% vs. 25.9%; p=0.004) (Figure 4). There was no association between late changes (day 180 to 365) in mRS with pre-existing PSD compared to not being depressed at day 180; mRS improved in 26% and 23.6% in non-PSD and PSD patients respectively (p=0.645). From day 30 to 365, only 10 patients were reported to have received pharmacotherapy for depression.

Figure 3: Frequency PSD at 180 days by ordinal modified Rankin Scale (mRS) score at 30 days.

mRS scores range from 0 (no disability) to 5 (severe disability).

Figure 4: Association between change in mRS before and after evaluation of depression.

Early improvement in mRS was associated with less PSD at day 180 (p=0.004), while there was no association between changes in mRS with pre-existing PSD (p=0.645). ^a represents 1 (0.9%) subject who had worsening of mRS in the early depressed group.

In the multivariable logistic regression model for variables associated with improvement in mRS from day 180 to 365 (Table 2), factors that were associated with lower odds of improvement in mRS were higher ICH volume (0.97 [0.95–1.00], p=0.031), higher EQ-VAS score at 30 days (0.99 [0.97–1.00], p=0.049), and ICH deep location (0.49 [0.25–0.93], p=0.029), whereas right hemisphere lesions (2.39 [1.05–5.43], p=0.037) and moderate GCS (2.26 [1.10–4.66], p=0.027) compared to mild GCS were associated with higher odds of improvement. The AUC for the model was 0.700. Considering US sites only, factors that were associated with lower odds of improvement in mRS were higher ICH volume (0.96 [0.93–0.99], p=0.015) and impaired MMSE at 30 days (0.34 [0.10–0.80], p=0.022; AUC=0.755). The AUC for the model was 0.755. PSD at day 180 was not significantly associated with mRS change at one year in the multivariable model.

Table 2:

Factors associated with any increase/improvement in mRS from 180 to 365 days

	Model 1 (N=308) All sites		Model 2 (N=240) US regions
Variables	Odds Ratio (95% CI)	P-value	Odds Ratio (95% CI)	P-value
N	261		208
Depressed at 180 days	0.94 (0.44–2.00)	0.867	1.33 (0.53–3.29)	0.543
Female sex	0.71 (0.37–1.35)	0.297	0.48 (0.21–1.09)	0.080
Hispanic or Latino ethnicity	0.44 (0.11–1.82)	0.258	0.50 (0.10–2.51)	0.399
ICH deep location	0.49 (0.25–0.93)	0.029	0.53 (0.25–1.12)	0.095
Mild GCS (13–15)	Reference group		Reference group
Moderate GCS (9–12)	2.26 (1.10–4.66)	0.027	2.25 (0.91–5.54)	0.079
Severe GCS (3–8)	1.26 (0.49–3.29)	0.630	1.25 (0.43–3.63)	0.678
Stability ICH volume, mL	0.97 (0.95–1.00)	0.031	0.96 (0.93–0.99)	0.015
Stability IVH present	0.79 (0.41–1.50)	0.466	1.00 (0.44–2.27)	0.994
ICH right hemisphere	2.39 (1.05–5.43)	0.037	1.58 (0.58–4.35)	0.374
EQ-VAS at 30 days	0.99 (0.97–1.00)	0.049	0.99 (0.97–1.00)	0.063
Impaired MMSE at 30 days	0.62 (0.28–1.36)	0.237	0.34 (0.14–0.86)	0.022
Patient not at home at 30 days	1.19 (0.50–2.85)	0.692	0.95 (0.34–2.64)	0.922
South, US region			1.23 (0.35–4.29)	0.745
Midwest, US region			1.05 (0.30–3.65)	0.943
Northeast, US region			3.10 (0.89–10.9)	0.076
West, US region			Reference group
AUC	0.700		0.755
Hosmer–Lemeshow test, p-value	0.168		0.075

Abbreviations: AUC = Area Under Curve; ICH = Intracerebral Hemorrhage; EQ-VAS = EuroQol Visual Analogue Scale; GCS = Glasgow Coma Scale; MMSE = Mini Mental State Examination; US = United States

DISCUSSION

In a large prospective observational cohort of patients with moderate to large ICH, we found that female sex, Hispanic ethnicity, right-sided ICH, presence of IVH, and both cognitive impairment and non-home residence at day 30 were independently associated with PSD at day 180. Nearly 4 out of 10 ICH survivors at 180 days displayed significant depression, which is substantially higher than observed in prior ICH populations and may be due to higher clinical severity in this cohort. Clinical improvement over the first 6 months, which occurred in 74% of non-depressed subjects, was significantly less likely (58%) in patients diagnosed with PSD at this time point, though PSD was not associated with improvement in mRS over the subsequent 6 months. These findings suggest that the trajectory of improvement over the first 6 months is a critical factor in the development of PSD.

Our reported rate of PSD (36%) is higher than that reported in other studies of ICH patients (Table 3), which range from 15% to 23%.^2–4 The highest rate of PSD was reported by Koivunen et al.,² who studied relatively younger ICH patients (mean age 50 years). ICH severity was not significantly associated with PSD except for baseline hydrocephalus and pain assessments. Two previous clinical trials, the Factor Seven for Acute Hemorrhagic Stroke (FAST) Trial and the Diagnostic Accuracy of MRI in Spontaneous Intracerebral Hemorrhage (DASH) study, evaluated PSD in ICH patients at 90 days and one year post-stroke.^3,4 PSD was common in both trials (20% and 15%, respectively), but while clinical severity indicators, comorbidities, and female sex predicted PSD in the FAST trial, the DASH trial found that PSD was not associated with initial hemorrhage severity but was associated with worse 1-year outcomes and with late worsening of disability (between 3 and 12 months).³ Both FAST and DASH studies focused on middle-age ICH patients with smaller ICH size and less long-term impairment. Both studies had similar ICH characteristics, though the DASH study had more patients with lobar ICH location. In adjusted analyses, age, anticoagulation at admission, and one year mRS were significantly associated with PSD in the DASH study, whereas female sex, previous comorbidity, and BI and NIHSS assessments at day 15 were associated with PSD at 90 days in the FAST trial. Although PSD in the MISTIE III trial was higher than in other studies, of ICH severity indicators, only IVH was significantly associated with PSD on adjusted analysis. The DASH and FAST trials did not include ethnicity, patient living location, cognition, and laterality of ICH, which were all important in our study (Table 3). Particularly, impaired cognition and not living at home appear to be key factors associated with depression and may be more important than physical limitations, provided that home residence is possible. Considering US sites only, the presence of IVH and impaired MMSE had weaker associations with CES-D and varied by US region (Table e-4). IVH and impaired MMSE had higher frequencies in depressed patients in the Midwest and South compared to Northeast and West. Therefore, one explanation for patients living in the South and Midwest having higher odds of post-stroke depression compared to other US regions is that they had higher rates of IVH and impaired cognition which were independent predictors of PSD in the full cohort. Small numbers of patients with PSD in the Northeast and West may also have limited our ability to observe similar associations in these regions. We did not find literature describing the prevalence of depression in the general population by US regions.

Table 3:

Studies Measuring Post-Stroke Depression in ICH Patients.

Studies	Koivunen2	DASH3	FAST4	MISTIE III
Location	Finland	USA (State: California)	122 sites in 22 countries	78 sites in 9 countries
Study type	Prospective	Prospective	Prospective	Prospective
Overall study duration	10 years	365 days	90 days	365 days
Total sample analyzed	130	89	596	308
Time of PSD measurement	Every year	365 days	90 days	180 days
Prevalence of PSD	30 (23.1%)	13 (15%)	120 (20%)	111 (36%)
PSD measures	HADSa BDI-IIb	HDRSc	HDRSc	CES-Dg
Univariate analysis: statistically significant predictors of PSD	Hydrocephalus BPId and PASSe Employment status	Anticoagulation (admit) mRS 0–1 vs 2–5 at 1 year Change in mRS (day 90 to 1 year)	History of comorbidity Female gender Baseline ICH volume Midline shift presence Day 15 assessments: GCS, BI, NIHSS, and mRS 0–1, 2–3, vs. 4–5 Use of antidepressants EQ-VAS	Hispanic/Latino ethnicity Presence of IVH ICH right hemisphere NIHSS at randomization Day 30 and 180 assessments: patient location, mRS 0–3 vs 4–5, EQ-VAS, impaired cognition, NIHSS, and BI score
Multivariable analysis: statistically significant predictors of PSD	Hydrocephalus PASS score	Age Anticoagulation 1-year mRS	Female History of comorbidity Day 15 assessments: BI and NIHSS	Female Hispanic/Latino ethnicity Presence of IVH ICH right hemisphere Day 30 assessments: patient location and impaired cognition

Abbreviations: BDI-II^b: Beck Depression Inventory-II; BI = Barthel Index; BPI^d= Brief Pain Inventory; CES-D^g = Center for Epidemiologic Studies Depression Scale; EQ-VAS= EuroQol Visual Analogue Scale; GCS = Glasgow Coma Scale; HADS^a = Hospital Anxiety and Depression Scale; HDRS^c= Hamilton Depression Rating Scale; ICH = Intracerebral Hemorrhage; IVH = Intraventricular Hemorrhage; mRS = modified Rankin Scale; NIHSS = National Institute of Health Stroke Scale; PASS^f= Pain Anxiety Symptom Scale; PSD = Post-stroke Depression

To our knowledge, MISTIE III is the only randomized clinical trial to use the CES-D to evaluate PSD in spontaneous ICH. Other commonly used depression scales in studies of ischemic and hemorrhagic stroke include the Hamilton Depression Rating Scale (HDRS), Beck Depression Inventory-II (BDI-II), and Hospital Anxiety and Depression Scale (HADS). The CES-D has been extensively studied in several large community samples and is an easily administered, feasible instrument for screening for post-stroke depression.²⁰ The well-established BDI-II was compared with CES-D (correlation coefficient was between 0.72 and 0.87) and with HADS (correlation coefficient was between 0.62 and 0.71),²⁰ and HDRS was compared with CES-D (correlation coefficient was between 0.60 and 0.80).²¹ Thus, the CES-D assessment aids in identifying subjects who are experiencing clinical depression with good sensitivity and specificity.²²

Our study showed that patients with PSD were more likely to have no change or worsening in mRS from 30 to 180 days compared to patients without PSD, who were more likely to show improvement in mRS. This is consistent with results of the DASH study, which found a signification association between depression at 365 days and worsening mRS from 90 to 365 days compared to non-depressed patients. Moreover, the depression score was higher in subjects with worsened mRS. Our finding of no association between PSD at day 180 and change in mRS from 180 to 365 days suggests that PSD may not lead to worsened disability although disability scores may have stabilized during this time period.

In studies of ischemic stroke, rates of PSD range from 5% to 37%.^23–26 Factors associated with PSD include: age, sex, employment status, marital status, mRS pre- and post-stroke, and social support. McCarthy et al.²³ reported 37% of patients were considered at risk for depression at 3 months post-stroke with significantly higher rates in patients under 65 years of age. Higher post-stroke depression scores were noted in patients that had difficulty paying bills on time and reported family or health-related stress at baseline. Volz et al.²⁴ reported 29.5% of patients were considered at risk for depression six months post-stroke. Baseline depression status and social support were major predictors of depression. Williams et al.²⁵ reported only 5% were indicative of depression 30 days post-stroke, and about 5% received a different mental health diagnosis within 3 years of stroke. These patients were also younger, more likely to be white, and less likely to be alive by at 3-year follow-up. Desmond et al.²⁶ reported 11.2% of patients were depressed in ischemic stroke compared to 5.2% in stroke-free control patients. Depression was associated with severe stroke, especially in certain vascular territories (affecting limbic structures), with dementia, and female sex. Thus, factors associated with PSD following ischemic stroke are variable, with cognition and social factors possibly having a larger impact than stroke severity, similar to ICH-associated PSD.

Our study demonstrated an association between right hemisphere ICH and PSD, which has been reported in other studies including hemorrhagic and ischemic stroke populations.²⁷ Of the 191 participants excluded from our study sample, 127 (66.5%) had left hemisphere ICH. We hypothesize that language impairment contributed to inability to assess PSD using the CES-D for participants with left hemisphere ICH, resulting in a higher sampling of PSD among participants with right hemisphere ICH. This rationale has also been used to explain association of higher PSD in right-sided ischemic stroke.

Similar to PSD in ischemic stroke populations, we also found an association between PSD and female sex in ICH (26). This may reflect the prevalence of major depression in females which is higher compared to males in the general population and correlates with hormonal changes especially during puberty, before menstruation, and at the perimenopause phase²⁹ Dong et al. reported that sex differences in prevalence of PSD at 90 days after first-ever stroke was not significant overall, but varied by pre-stroke depression status with women having higher frequency of pre-stroke depression requiring medication compared to men.³⁰ Factors associated with secondary depression such as PSD appear to be different in males and females. Paradiso studied PSD in 301 patients with ischemic stroke where women were twice as frequently diagnosed with major depression as men.³¹ Women with PSD had a greater frequency of left hemisphere lesions than men, and risk was higher with a prior diagnosis of depression and with cognitive impairment. Among men, impairment in ADLs and social function were associated with a greater risk of PSD. Dong et al reported that Hispanics had higher prevalence of PSD at 90 days than non-Hispanic whites. The main contributor to this finding was lower educational attainment.³⁰

This study has several limitations. First, CES-D was collected only at day 180, precluding longitudinal assessment. Second, patients were not screened for depression prior to randomization, so baseline history of depression was not validated. Of the 308 patients, two (1.0%) were on antidepressants prior to study randomization and 10 patients (3.2%) were on antidepressants at any time from day 30 to 365, suggesting a low rate of pre-stroke depression. Of these 10 patients, only one was on antidepressants after 180 days indicating possible under-treatment of this condition. Third, socioeconomic data of potential importance to a diagnosis of depression such as education, income, marital status, physical activity, and occupation was not collected. Fourth, the MISTIE III trial excluded patients with lower severity ICH (ICH volume <30 mL), which limits generalizability to small ICH size, but provides a good balance of baseline clinical factors, and 2/3 of patients had deep ICH, which is less frequently represented in large case series. Higher clinical severity is also a strength compared to other studies that have evaluated predominantly mild to moderate severity ICH. Fifth, we identify that the CES-D is limited in several respects. (i) It may be difficult to evaluate using CES-D in patients with cognitive dysfunction; (ii) CES-D is limited in measurement of some intrinsic features of depressive mood such as strong sadness, feelings of worthlessness, suicidal ideation, and guilt compared to other depression instruments although the scale has acceptable reliability and validity when used for assessing depression in suicide attempters;²⁸ (iii) It may also be difficult to distinguish between stroke symptoms and PSD symptoms. Slow movement, decreased speech volume, and decreased appetite are common symptoms after a stroke as are emotional instability and feeling of fatigue which are all also symptoms of depression. Finally, one-fifth of survivors were unable or did not complete CES-D assessments, presenting a bias in demographics and clinical severity compared to those who completed the CES-D.

Our data suggest a high frequency of depression (36%) in patients with moderate to large ICH, especially among survivors with cognitive impairment and those not living at home. Cultural or socioeconomic factors associated with geographic location may also be important. PSD has major implications on post-stroke rehabilitation, including functional recovery and quality of life. Therefore, effective strategies to screen and manage PSD in ICH patients are needed, as well as longitudinal PSD assessment. The CES-D provides a reliable and feasible approach to screening for PSD in ICH patients. Healthcare professionals and clinicians may have overlooked or deferred treatment of PSD in ICH. Early recognition of PSD has the potential to address emotional and perhaps physical needs of patients with large ICH. Prospective studies are needed to provide evidence-based approaches to address PSD and its treatment among ICH survivors. Interventions early in the recovery course may be useful.

Appendix 1: Authors

Name	Location	Contribution
Radhika Avadhani, MS	Johns Hopkins University, Baltimore, MD	Study concept and design, acquisition, analysis, and interpretation of data, statistical analysis, drafting of the manuscript; No disclosure to report
Richard E. Thompson, PhD	Johns Hopkins University, Baltimore, MD	Acquisition, analysis, and interpretation of data, statistical analysis, critical revision of the manuscript for important intellectual content; No disclosure to report
Lourdes Carhuapoma, MS	Johns Hopkins Hospital, Baltimore, Maryland	Critical revision of the manuscript for important intellectual content; No disclosure to report
Gayane Yenokyan, PhD	Johns Hopkins University, Baltimore, MD	Statistical analysis, critical revision of the manuscript for important intellectual content; No disclosure to report
Nichol McBee, MPH	Johns Hopkins University, Baltimore, MD	Critical revision of the manuscript for important intellectual content, administrative, technical, and material support; No disclosure to report
Karen Lane, CCRP	Johns Hopkins University, Baltimore, MD	Critical revision of the manuscript for important intellectual content; No disclosure to report
Noeleen Ostapkovich, MS	Johns Hopkins University, Baltimore, MD	Critical revision of the manuscript for important intellectual content; No disclosure to report
Agnieszka Stadnik, MS	University of Chicago, Chicago, IL	Critical revision of the manuscript for important intellectual content; No disclosure to report
Issam A. Awad	University of Chicago, Chicago, IL	Critical revision of the manuscript for important intellectual content; No disclosure to report
Daniel F. Hanley	Johns Hopkins University, Baltimore, MD	Study concept and design, analysis, and interpretation of data, critical revision of the manuscript for important intellectual content, managed study funding and personnel; No disclosure to report
Wendy C. Ziai	Johns Hopkins Hospital, Baltimore, Maryland	Study concept and design, acquisition, analysis, and interpretation of data, statistical analysis, drafting of the manuscript, critical revision of the manuscript for important intellectual content, study supervision; No disclosure to report