Cerebral Small Vessel Disease and Depression among Intracerebral Hemorrhage Survivors

Abstract

Background and Purpose:

Intracerebral Hemorrhage (ICH) is an acute manifestation of Cerebral Small Vessel Disease (CSVD), usually Cerebral Amyloid Angiopathy (CAA) or Hypertensive Arteriopathy (HTNA). CSVD-related imaging findings are associated with increased depression incidence in the general population. Neuroimaging may therefore provide insight on depression risk among ICH survivors. We sought to determine whether CSVD CT and MRI markers are associated with depression risk (before and after ICH), depression remission, and effectiveness of antidepressant treatment.

Methods:

We analyzed data from the single-center longitudinal ICH study conducted at Massachusetts General Hospital (MGH-ICH). Participants underwent CT and MRI imaging, and were followed longitudinally. We extracted information for neuroimaging markers of CSVD subtype and severity. Outcomes of interest included pre-ICH depression, new-onset depression after ICH, resolution of depressive symptoms, and response to antidepressant treatment.

Results:

We followed 612 ICH survivors for a median of 47.2 months. Multiple CSVD-related markers were associated with depression risk. Survivors of CAA-related lobar ICH were more likely to be diagnosed with depression before ICH (Odds Ratio 1.68, 95% Confidence Interval [CI] 1.14–2.48) and after ICH (Sub-Hazard Ratio 1.52, 95% CI: 1.12–2.07), less likely to achieve remission of depressive symptoms (SHR 0.69, 95% CI: 0.51–0.94), and to benefit from antidepressant therapy (p = 0.041). CAA disease burden on MRI was associated with depression incidence and treatment resistance (interaction p = 0.037), whereas HTNA disease burden was only associated with depression incidence after ICH.

Conclusions:

CSVD severity is associated with depression diagnosis, both before and after ICH. CAA-related ICH survivors are more likely to experience depression (both before and after ICH) than patients diagnosed with HTNA-related ICH, and more likely to report persistent depressive symptoms and display resistance to antidepressant treatment.

INTRODUCTION

Intracerebral hemorrhage (ICH) is one of the most severe forms of stroke, accounting for almost half of all stroke related-morbidity and mortality.^{1, 2} Depression is a common occurrence after stroke (including ICH), and is associated with decreased quality of life, poor long-term functional outcomes and increased mortality.^{3, 4} Timely identification and effective treatment of depression are therefore a crucial aspect of post-stroke care, especially for ICH survivors.

Cerebral Small Vessel Disease (CSVD), a chronic disorder of the microscopic vessels of the brain, is the most common underlying cause of spontaneous ICH.^{5, 6} Hypertensive Arteriopathy (HTNA, also known as arteriolosclerosis or deep perforators arteriopathy) and Cerebral Amyloid Angiopathy (CAA) are responsible for the overwhelming majority of ICH cases.^6–9 CSVD is typically assessed via neuroimaging markers of disease etiology and severity, including hematoma location (lobar vs. non-lobar), cerebral microbleeds (CMB), white matter hyperintensities (WMH), lacunes, cortical superficial siderosis (cSS), and enlarged perivascular spaces (EPVS).¹⁰ Prior studies found CSVD markers to associate with late-life depression and apathy in the general population. ^{3, 11}

We therefore sought to determine whether information on CSVD subtype and severity derived from neuroimaging are associated with prevalence of depression before ICH, new-onset depression incidence after ICH, resolution of depressive symptoms, and resistance to antidepressant treatment. We specifically planned to test the informative potential on depression outcomes of interest of: 1) CT-based hematoma location; 2) MRI-based ICH etiological classification; 3) individual MRI-based CSVD markers; 4) validated scores summarizing global CSVD, CAA-specific and HTNA-specific burden on MRI.^12–14 To do so, we leveraged longitudinal data from a single center prospective study of ICH with baseline MRI scans (at time of ICH) and repeated measures of depression during follow-up.

METHODS

Data Availability

The authors certify they have documented all data, methods, and materials used to conduct the research presented. Anonymized data pertaining to the research presented will be made available upon reasonable request from external investigators.

Study Eligibility and Patients’ Recruitment

We performed longitudinal analyses of prospectively collected data from the ongoing ICH study conducted at Massachusetts General Hospital (Boston, MA, USA).^15–17 Study subjects were consecutive patients admitted with a diagnosis of spontaneous ICH to Massachusetts General Hospital (MGH) in the period between January 1, 1998 and December 31, 2017. All participants enrolled were 18 years or older at time of acute primary ICH. Patients with intracranial hemorrhage secondary to trauma, rupture of an aneurysm or vascular malformation, brain tumor or conversion of an ischemic infarct were not considered eligible, and were consequently excluded from enrollment. For the purposes of the present study, patients who did not undergo an MRI within 90 days of initial ICH were not included. Study protocols and procedures were approved by the institutional review board at Massachusetts General Hospital. Written informed consent was obtained from all study participants or surrogates.

Participants’ Enrollment and Baseline Data Collection

At time of enrollment, study staff gathered information on demographics, medical and family history, and medication use via in-person research interview.¹⁶ This in-person interview was complemented and expanded upon via semi-automated review of Electronic Medical Records (EMR). We diagnosed participants with pre-ICH history of depression in the presence of both: 1) self-reported or informant-reported prior diagnosis of depressive disorders, and 2) documented history of depression (by DSM-5 criteria), based on review of medical records.

Neuroimaging

CT and MRI scans analyzed were deidentified, digitalized, and analyzed as previously described (Supplemental Methods).¹⁸ Admission CT scans were analyzed to determine ICH location, hematoma volume, and existence of intraventricular blood according to previously validated methodology¹⁹. MRI-based markers of CSVD were rated according to the Standards for Reporting Vascular Changes on Neuroimaging (STRIVE) consensus criteria, and included WMH, lacunes, CMB, EPVS and cSS.¹⁰ All participants were classified based on location of ICH, CMBs, and presence of cSS as HTNA-related ICH, CAA-related ICH and mixed-ICH, as previously described (Supplemental Methods).²⁰ Based on recently described and validated CSVD scores, we rated global, HTNA-related and CAA-related CSVD burden using three separate scales (Table I).^12,13

Longitudinal follow-up

ICH survivors and their caregivers were followed via phone interviews by trained study staff blinded to baseline and neuroimaging information at 3 and 6 months after ICH, and every 6 months thereafter.¹⁶ Telephone-based collection of follow-up data was supplemented with semiautomated review of longitudinal EMR.¹⁵ We defined depression incidence post-ICH as participants meeting both of the following criteria: 1) documented evidence in medical records (by DSM-5 criteria) or billing information of a new depression diagnosis; and 2) 4-item version of the Geriatric Depression Scale (GDS-4) score of 2 or above.²¹ We also administered the Katz and Lawton questionnaires for Activities of Daily Living (ADLs) and Instrumental Activities of Daily Living (IADLs).^{22, 23}

Variables’ Definition and Exposures of Interest

Age at index ICH was analyzed as a continuous variable. Race / ethnicity was analyzed as a categorical variable, with white patients as the reference group owing to their numerical preponderance. Educational level was dichotomized using a cutoff of 12 or more years of education. CSVD MRI markers were analyzed and scores created as described above. We defined depression for outcome analyses as described above. Study participants without history of pre-ICH depression were diagnosed with incident depression when first meeting these criteria. Patients diagnosed with incident depression after ICH were considered to have undergone remission of depressive symptoms when meeting both of the following criteria: 1) documented evidence in medical records or billing information of resolution of depressive symptoms; and 2) GDS-4 score of less than 2. We also extracted data on use and dosing of antidepressant medications based on previously described methods.¹⁷

Statistical Methods

Overall Analysis Plan

Our initial analysis plan focused on exploring associations between CT-based ICH location and MRI markers of CSVD with the following depression-related outcomes of interest: 1) pre-ICH depression; 2) new-onset depression after ICH; 3) resolution of depressive symptoms; and 4) resistance to antidepressant therapy (using dedicated interaction analyses, see below). We also performed sensitivity analyses for early (onset within 6 months of ICH) vs. delayed (onset after 6 months from ICH) depression after hemorrhagic stroke (Table III). For each outcome, we first explored associations with ICH location and then with individual CSVD markers in univariable and multivariable models. Finally, we investigated associations between CSVD MRI scores and outcomes of interest. We present our results in agreement with the TRIPOD guidelines.

Univariable and Multivariable Association Analyses

We investigated the association between CSVD markers and pre-ICH depression using logistic regression models. To determine whether CSVD markers are associated with new-onset depression after ICH incidence and depression resolution, we employed a time-to-event analysis framework. All participants diagnosed with depression were separately included in longitudinal analyses of depression resolution (as defined above), with inception either at time of acute ICH (for individuals with history of pre-ICH depression) or depression diagnosis (for those who developed depression after ICH). We conducted additional analyses including interaction terms between CSVD markers of interest and antidepressant use to identify associations with treatment-resistance. Antidepressant use was entered in the model as a dichotomous (i.e. yes/no) time-varying exposure. Information on individual antidepressant agents and dosing were entered as covariates (thus modeling dose changes and switches between different agents), and also handled as time-varying exposures. All time-to-event univariable analyses relied on the Log-rank test, while multivariable analyses were carried using Fine and Gray competing risk regression (with death specified as a competing outcome in all models).

For all analyses we pre-specified adjustment for established predictors of post-stroke depression risk, including age, sex, race/ethnicity, education, history of stroke or TIA before index ICH, NIH Stroke Scale at index ICH presentation, hematoma volume, recurrent stroke (ischemic or hemorrhagic), functional performance of ADLs and IADLs, and antidepressant use.³ We also selected all variables with p<0.20 for association with the dependent outcome variable for inclusion in subsequent multivariable analyses. We then generated a minimal model via backward elimination of non-significant variables (at p>0.05), excluding the CSVD marker of interest.

Multiple Testing Adjustment

We corrected for multiple testing using the Benjamini-Hochberg False Discovery Rate (FDR) method for adjustment.²⁴ We report p-values after adjustment of original results using the False Discovery Rate (FDR) method with the notation “pFDR”. All significance tests were 2-tailed, and significance was set at pFDR < 0.05. All analyses performed using R software (The R Foundation for Statistical Computing), version 3.6.2.

RESULTS

Study Participants

We initially screened for inclusion in subsequent analyses 1549 consecutive primary ICH cases enrolled in our longitudinal study (Figure 1). After application of pre-specified inclusion and exclusion criteria, we analyzed data for 612 participants. The majority of ineligible participants (485 of 937, 52%) were excluded because of mortality within 90 days of ICH. Missing or inadequate MRI data represented the second most common exclusion criterion, resulting in an additional 279 ICH survivors being excluded (30% of all excluded individuals). We found no significant differences when comparing ICH survivors with vs. without available MRI data (Table II). In addition, availability of MRI data was not associated with depression risk before ICH (Odds Ratio [OR] 1.28, 95% Confidence Interval [CI] 0.83 – 1.97) or during follow-up (Hazard Ratio [HR] 0.79, 95 CI 0.47 – 1.33). Among study participants median time from hospital arrival to MRI scan was 4.2 days (Inter-Quartile Range [IQR]: 3.3 – 5.8). We present key characteristics for participating ICH survivors in Table 1.

Figure 1. Study Enrollment Flowchart.

Black box presents initial inclusion criteria. White boxes represent individuals retained at each step of the evaluation process. Boxes with dashed lines represent exclusion criteria and excluded individuals. Grey boxes represent patients experiencing depression-related outcomes of interest. Abbreviations: FU = Follow-up; ICH = Intracerebral Hemorrhage.

Table 1.

Participants’ Characteristics

Variable	Pre-ICH Depression	No Pre-ICH Depression
		New-Onset Depression after ICH		No Depression after ICH
	No. (%)	No. (%)		No. (%)
No. of Individuals	104 (100)	255 (100)		253 (100)
Demographics
Age at Enrollment (Mean, SD)	69.6 (12.5)	71.2 (12.8)		70.3 (13.1)
Sex (Male)	53 (51)	138 (54)		133 (53)
Race / Ethnicity*
White	83 (80)	201 (79)		232 (92)
Black	7 (7)	15 (6)		11 (4)
Hispanic	8 (8)	15 (6)		15 (6)
Other	6 (6)	18 (7)		1 (1)
Education (≥12 years)*	53 (51)	133 (52)		175 (69)
Medical History
Tobacco Use (Current)	10 (10)	28 (11)		28 (11)
Alcohol Consumption (≥ 3 oz/day)	19 (18)	51 (20)		37 (15)
Hypertension	78 (75)	189 (74)		191 (75)
Diabetes	19 (18)	54 (21)		48 (19)
Prior ICH	4 (4)	10 (4)		6 (2)
Prior Ischemic Stroke / TIA*	10 (10)	28 (11)		17 (7)
CT Imaging Data
ICH Location*
Lobar	58 (56)	143 (56)		129 (51)
Non-lobar	43 (41)	110 (43)		120 (47)
Mixed locations	3 (3)	2 (1)		4 (2)
ICH Volume (cc, median (IQR)*	17.2 (5.3 – 29.2)	20.5 (6.6 – 30.1)		16.8 (5.5 – 27.4)
Intraventricular Extension*	26 (25)	77 (30)		61 (24)
No. of Individuals	104 (100)	255 (100)	253 (100)
ICH Etiological Classification *
CAA-related	57 (55)	143 (56)	109 (43)
HTNA-related	21 (20)	54 (21)	73 (29)
Mixed	26 (25)	59 (23)	70 (28)
Medication Use during Follow-Up
Antiplatelet Agents	16 (15)	41 (16)	41 (16)
Oral Anticoagulants	11 (11)	26 (10)	25 (10)
Statins	33 (32)	89 (35)	87 (34)
CSVD MRI Markers
Lacunes (presence)	36 (35)	87 (34)	80 (32)
Basal Ganglia EPVS (count median, IQR)*	1 (1–2)	2 (1–2)	1 (1–2)
Centrum Semiovale EPVS (count median, IQR)*	2 (2–3)	2 (2–3)	2 (1–2)
Lobar CMB (presence)*	58 (56)	145 (57)	121 (48)
Non-lobar CMB (presence)*	34 (33)	89 (35)	59 (23)
White Matter Hyperintensities
Periventricular (score median, IQR)*	2 (1–2)	2 (1–2)	2 (1–2)
Deep (score median, IQR)*	2 (1–2)	2(1–2)	1 (1–2)
Cortical Superficial Siderosis
Focal (n)	10 (10)	26 (10)	19 (8)
Disseminated (n)	6 (6)	20 (8)	3 (1)
CSVD MRI Scores
Global CSVD Score (median, IQR)*	3 (2–3)	3 (2–3)	2 (1–2)
HTNA-CSVD Score (median, IQR)*	2 (1–3)	2 (2–3)	2 (1–2)
CAA-CSVD Score (median, IQR)*	2 (1–2)	2 (1–3)	1 (0–1)

All values reported as number and percentages, unless otherwise specified

^*pFDR <0.05 for comparison across categories

Abbreviations: CAA = Cerebral Amyloid Angiopathy; CMB = Cerebral Microbleed, CSVD = Cerebral Small Vessel Disease; EPVS = Enlarged Perivascular Spaces; HTNA = Hypertensive Arteriopathy; ICH = Intracerebral Hemorrhage; IQR = Inter-Quartile Range; pFDR = p-value after multiple testing adjustment; SD = Standard Deviation; WMH = White Matter Hyperintensity.

Follow-up Information

Participants were followed for a median time of 47.2 months (Inter-Quartile Range [IQR]: 35.4 – 57.8), with yearly loss to follow-up of 1.2%. Among study participants, 104 (17%) carried a diagnosis of pre-ICH depression. Of these, 93 (89%) demonstrated active symptoms of depression following ICH at some point during follow-up. An additional 255 participants (42%) were diagnosed with new-onset depression after ICH. In total, 348/612 study participants (57%) qualified for a diagnosis of depression at any time point before or after ICH. After onset of depression, we followed ICH survivors for resolution of symptoms. During a median follow-up time of 25.8 months (IQR 18.1 – 36.2) we observed resolution of depression in 181 study participants (52%) in this subgroup.

Cerebral Small Vessel Disease and Pre-ICH Depression

In univariable and multivariable models, lobar ICH survivors were more likely to be diagnosed with depression before ICH (OR 1.68, 95% CI 1.14 – 2.48, pFDR = 0.014), while survivors of deep, cerebellar and mixed location ICH were not. Upon analyzing MRI-defined hemorrhage etiology, we likewise found that participants presenting with CAA-related ICH were more likely to be diagnosed with depression before ICH (OR 1.96, 95% CI 1.19 – 3.23, pFDR = 0.01). We also found that lobar lacunes, enlarged Centrum Semiovale (CSO) perivascular spaces, lobar cerebral microbleeds, deep white matter hyperintensities and disseminated cortical superficial siderosis were all associated with pre-ICH depression (Table 2). In terms of composite CSVD scores, higher values of the global CSVD score and the CAA score (but not the HTNA CSVD score) were associated with greater pre-ICH depression prevalence.

Table 2:

Depression Outcomes among ICH Survivors and CSVD Markers

Neuroimaging Variables	Pre-ICH Depression			New-Onset Depression after ICH			Depression Resolution
	OR	95% CI	pFDR	SHR	95% CI	pFDR	SHR	95% CI	pFDR
Lacunes
Lobar Lacunes (≥ 1)	1.41	1.05 – 1.89	0.031	1.62	1.11 – 2.36	0.019	0.86	0.74 – 0.99	0.048
Non-lobar Lacunes (≥ 1)	1.25	0.99 – 1.58	0.094	1.50	1.04 – 2.16	0.033	0.91	0.82 – 1.01	0.12
Enlarged Perivascular Spaces
Basal Ganglia EPVS (≥ 10)	1.47	0.98 −2.21	0.077	1.33	1.05 – 1.68	0.021	0.88	0.76 – 1.03	0.096
Centrum Semiovale EPVS (≥ 10)	1.26	1.02 – 1.56	0.040	1.39	1.04 – 1.86	0.030	0.73	0.52 – 1.02	0.089
Cerebral Microbleeds
Lobar CMBs (≥ 2)	1.29	1.02 – 1.64	0.042	1.40	1.03 – 1.84	0.036	0.74	0.55 – 1.00	0.066
Deep CMBs (≥ 1)	1.22	0.97 – 1.54	0.13	1.18	1.01 – 1.38	0.045	0.85	0.70 – 1.03	0.13
White Matter Hyperintensities
Deep WMHs (grade 2 or 3)	1.36	1.04 −1.78	0.029	1.55	1.10 – 2.18	0.017	0.91	0.83 – 0.99	0.057
Periventricular WMHs (grade 3)	1.14	0.91 – 1.43	0.34	1.44	0.72 −2.89	0.39	0.87	0.63 – 1.20	0.44
Cortical Superficial Siderosis
Focal	1.48	0.76 – 2.89	0.31	0.85	0.60 – 1.21	0.42	1.27	0.82 – 1.97	0.33
Disseminated	1.39	1.05 – 1.85	0.028	1.26	1.04 – 1.53	0.024	0.81	0.67 – 0.98	0.039
MRI-based CSVD Scores *
Global SVD Score	1.28	1.04 – 1.58	0.036	1.29	1.05 – 1.58	0.022	0.67	0.35 – 1.27	0.28
CAA-specific SVD Score	1.37	1.08 – 1.74	0.015	1.49	1.13 – 1.97	0.009	0.78	0.64 – 0.96	0.026
HTNA-specific SVD Score	1.12	0.96 – 1.31	0.21	1.32	1.03 – 1.70	0.041	0.88	0.75 – 1.04	0.16

^*effect sizes calculated for 1-point increase

Each MRI marker or score was entered in a separate logistic regression or competing risk model including adjustment for age, sex, race/ethnicity, education, history of stroke or TIA before index ICH, NIH Stroke Scale at index ICH presentation, hematoma volume, recurrent stroke (ischemic or hemorrhagic), functional performance of ADLs and IADLs, and antidepressant use.

Associations achieving statistical significance (p < 0.05) were bolded.

Abbreviations: 95% CI = 95% Confidence Interval; CAA = Cerebral Amyloid Angiopathy; CSVD = Cerebral Small Vessel Disease; EPVS = Enlarged Perivascular Spaces; HTNA = Hypertensive Arteriopathy; OR = Odds Ratio; pFDR = p-value after multiple testing adjustment; SE = Standard Error; SHR = Sub-Hazard Ratio; WMHs = White Matter Hyperintensity.

Cerebral Small Vessel Disease and New-Onset Depression after ICH

In both univariable and multivariable models, lobar ICH survivors were more likely to be diagnosed with new-onset depression after ICH (Sub-Hazard Ratio [SHR] 1.52, 95% CI 1.12 – 2.07, pFDR = 0.010). Similarly, survivors of CAA-related ICH (defined based on MRI data) were at highest risk for new-onset depression after ICH (SHR 1.84, 95% CI: 1.20 – 2.82 pFDR = 0.008). Among specific MRI markers, lacunes (both lobar and non-lobar), enlarged perivascular spaces (CSO and basal ganglia), cerebral microbleeds, deep white matter hyperintensities and disseminated cortical superficial siderosis were associated with new-onset depression after ICH (Table 2). The global CSVD, CAA and HTNA scores were all associated with diagnosis of new-onset depression after ICH. We illustrate the association between hemorrhage etiology and new-onset depression after ICH in Figure 2.

Figure 2. Hemorrhage Etiology and Risk of Depression after ICH.

Abbreviations: CAA = Cerebral Amyloid Angiopathy, CSVD = Cerebral Small Vessel Disease, HTNA = Hypertensive Arteriopathy; ICH = Intracerebral Hemorrhage.

Cerebral Small Vessel Disease Severity and Depression Resolution after ICH

We found in both univariable and multivariable analyses that survivors of lobar ICH diagnosed with depression were less likely to experience resolution of depressive symptoms (SHR 0.69, 95% CI 0.51 – 0.94, pFDR = 0.028). Survivors of CAA-ICH (defined based on MRI data) were also less likely to achieve resolution of depressive symptoms (SHR 0.72, 95% CI: 0.56 – 0.93, pFDR = 0.019). Upon analyzing individual CSVD markers, we found that only lobar lacunes and disseminated cortical superficial siderosis were associated with lower likelihood of depression resolution (Table 2). Similarly, when examining all three CSVD MRI scores, only CAA severity was associated with decreases likelihood of resolution of depressive symptoms among ICH survivors.

Cerebral Small Vessel Disease and Depression Treatment Resistance after ICH

We conducted pre-specified interaction analyses to determine whether CSVD subtype and/or severity were associated with treatment resistance of depressive symptoms following ICH, after accounting for medication type, dosing and duration of therapy. Among ICH etiological subtypes, CAA-related ICH was associated with decreased likelihood of response to antidepressant treatment (SHR 0.72, 95% CI = 0.53 – 0.97, interaction pFDR = 0.041). Mixed etiology and HTNA-related ICH diagnoses were not associated with antidepressant treatment response (both interaction pFDR > 0.20). Among validated CSVD scores, only increase in the CAA score was associated with decreased response to antidepressant treatment (Table IV). We illustrate the association between antidepressant treatment and mood symptoms as a function of CSVD severity in Figure 3.

Figure 3. CSVD Severity and Response to Antidepressant Treatment after ICH.

Plot presenting the association between antidepressant use and resolution of depressive symptoms within study subgroups defined by Global (top), CAA (middle), and HTNA (bottom) validated MRI-based burden scores. Abbreviations: CAA = Cerebral Amyloid Angiopathy, CSVD = Cerebral Small Vessel Disease, HTNA = Hypertensive Arteriopathy.

DISCUSSION

We provide evidence that neuroimaging markers of CSVD evaluated at time of an acute ICH are associated with depression diagnosis among survivors, both prior to and after the acute hemorrhagic stroke event. Survivors of CAA-related lobar ICH are significantly more likely to be diagnosed with depression, less likely to experience resolution of depressive symptoms, and showed decreased benefit from antidepressant treatment. Taken together, these findings highlight the relevance of MRI-based CSVD assessment in the evaluation and treatment of mood disorders after ICH.

Our results expand upon established associations between white matter disease, lacunes and depressive symptoms in the general elderly population. ^{11, 25–27} We also provide novel evidence of associations between EPVS, cSS, and CMB with post-stroke depression risk. Taken together, available evidence indicates that vascular lesions are central to the etiopathogenesis of post-stroke depression (or at least post-ICH depression), echoing the “vascular depression” hypothesis.^{3, 28–30} Additional studies, ideally including novel neuroimaging or biochemical markers of CSVD, will ultimately be required to definitively clarify the mechanisms linking CSVD with neuropsychiatric sequelae of primary ICH. Of note, in this and prior studies, we found higher prevalence of depression among ICH survivors than generally reported among ischemic stroke patients.^{3, 17} This is likely to reflect systematic underlying differences, chiefly more severe disability and presence of severe and progressive CSVD among ICH survivors.

In contrast to a previous study ³¹, we found stronger association with significantly larger effect size for lobar CMB in the development of pre- and post-ICH depression compared to deep CMB. In our study, other CAA-related MRI markers (disseminated cSS and lobar lacunes) were also strongly associated with likelihood of depression diagnosis before and after ICH. It is certainly possible that systematic differences in study population (e.g. study setting, demographics, risk factor profiles) resulted in higher incidence of CAA in our study. However, cortical brain networks are considered central to mood regulation and emotion perception,^{32, 33} and damage to these networks may ultimately lead to vascular-related mood disorders. ^{28, 34, 35} Of note, while our findings suggest CAA is involved in the pathogenesis of vascular depression, we were unable to fully account for the effects of parenchymal amyloid burden (related to Alzheimer’s disease) on post-ICH depression. Additional studies are therefore warranted to explore the relative contributions of CAA and Alzheimer’s disease pathologies on depression risk after hemorrhagic stroke.

We also provide novel evidence linking increasing CSVD severity with treatment resistance for post-stroke depression. Previous studies have shown that late-life “vascular” depression is less likely to respond to treatment with antidepressant medications. ³ Our findings directly implicate CSVD biomarkers as associated with decreased response to treatment of post-stroke depression..²⁸ Additionally, we previously showed that CSVD severity modifies the association between use of SSRIs (considered first-line treatment for post-stroke depression) and recurrent ICH risk.¹⁷ If confirmed in randomized controlled trials, our findings could pave the way for CSVD markers being leveraged to inform post-stroke depression treatment, specifically by quantifying individuals’ risk-benefit balance in achieving remission of depressive symptoms while limiting risk for recurrent bleeding.³⁶

Our study has some limitations. First, all study participants were enrolled at a tertiary referral center with dedicated expertise in ICH care, leading to potential severity and imaging-availability bias. Second, due to study design we are limited to describing associations between CSVD markers and depression-related outcomes, without any ability to establish causal links. This is particularly relevant for treatment resistance, as antidepressant medication use was not pre-determined in a randomized fashion. Third, we did not include in our imaging protocol advanced MRI-based markers of CSVD severity, such as connectivity or microstructural integrity information derived from fMRI or DTI sequences respectively. However, our imaging protocol is more closely aligned to clinical MRI scans already being performed on ICH survivors in current clinical practice – thus increasing the relevance of our findings. Fourth, diagnosis of depression was formulated through medical records review and brief patient interviews, thus potentially limiting diagnostic accuracy. Neurocognitive symptoms more frequently resulting from lobar ICH (e.g. aphasia, neglect) in particular may have been misdiagnosed as depression, thus biasing our results. However, we previously demonstrated our approach has excellent (>90%) sensitivity and specificity for depression diagnosis compared to in-person clinician evaluation.¹⁷ We also correctly modeled the effects of antidepressant treatment, suggesting only limited confounding due to neurocognitive symptoms –since they are not expected to respond to pharmacological therapy. Nonetheless, availability of medical records may have resulted in delayed diagnosis of depression after ICH. We also did not screen for presence of depression during the acute ICH hospitalization; our findings, therefore, have limited applicability to depressive symptoms in the immediate phase after ICH and their potential association with ICH recovery. Finally, focusing on hemorrhagic stroke survivors likely introduced bias in our analyses of pre-ICH depression. However, these analyses are intended to complement our findings related to depression incidence after ICH, rather than explore mood disorders among CSVD patients at risk for ICH. Our study displays various strengths. First, we were able to leverage a relatively large sample of ICH survivors, using standardized neuroimaging and longitudinal follow-up procedures leveraging concordant information derived from multiple source (patients, informants, electronic medical records, medical billing, medication refill data).^{15, 17, 19} Finally, to our knowledge, ours is the first study providing specific evidence for the association between CAA and post-ICH depression. We also provide novel evidence that CSVD burden affects likelihood of depressive symptom resolution, and is associated with resistance to antidepressant treatment.

In summary, we provide evidence that MRI-based markers of CSVD evaluated at time of an acute ICH event are associated with likelihood of being diagnosed with depression, both before and after the hemorrhagic stroke. CAA severity (as defined on MRI) was associated with resistance to medication treatment among ICH survivors. These findings highlight the central role played by CSVD in mood-related outcomes among survivors of primary ICH. Clinicians caring for ICH survivors could leverage MRI imaging to identify individuals at high risk for depression, such as those with evidence of underlying CAA. Additional investigative efforts (especially dedicated clinical trials) are warranted to determine whether MRI neuroimaging could further inform clinical management of depressive symptoms after ICH.

Non-standard Abbreviations and Acronyms

BG

Basal Ganglia

CI

Confidence Interval

CAA

Cerebral Amyloid Angiopathy

CMB

Cerebral Microbleed

CSO

Centrum Semiovale

cSS

Cortical Superficial Siderosis

CSVD

Cerebral Small Vessel Disease

EPVS

Enlarged Perivascular Spaces

HTNA

Hypertensive Arteriopathy

ICH

Intracerebral Hemorrhage

IQR

Interquartile Range

MGH

Massachusetts General Hospital

MGH-ICH

Intracerebral Hemorrhage Study conducted at Massachusetts General Hospital

SHR

Sub-Hazard Ratio

WMH

White Matter Hyperintensities