Posttraumatic Stress Disorder, Antidepressant Use, and Hemorrhagic Stroke in Young Men and Women: A 13-year Cohort Study

Abstract

Background and Purpose:

Antidepressants are commonly prescribed for posttraumatic stress disorder (PTSD) and may increase the risk of bleeding, including hemorrhagic stroke.

Methods:

We prospectively examined independent effects of PTSD, selective serotonin and norepinephrine reuptake inhibitors [SSRI and SNRI] on the risk of incident hemorrhagic stroke in a nationwide sample of 1.1 million young and middle-aged Veterans. Time-varying multivariate Cox models were used to examine hemorrhagic stroke risk by PTSD status and use of SSRI or SNRI while adjusting for demographics, lifestyle factors, stroke and psychiatric comorbidities. Sensitivity analyses controlled for healthcare utilization.

Results:

During 13 years of follow-up (2.14 years on average), 507 patients (12% women) suffered a hemorrhagic stroke. The overall incidence rate was 1.70 events per 10,000-person years. In unadjusted models, PTSD was associated with an 82% greater risk of new-onset hemorrhagic stroke (HR, 1.82; 95% CI, 1.48–2.24), SSRI use was associated with a greater than 2-fold risk (HR, 2.02; 95% CI, 1.66–2.57), and SNRI use was associated with a 52% greater risk (HR, 1.52; 95% CI, 1.08–2.16). In fully-adjusted models, effects of PTSD and SNRI were attenuated (aHR, 1.03; 95% CI, 0.81–1.34; aHR, 1.19; 95% CI, 0.83–1.71), but SSRI use remained associated with a 45% greater risk of hemorrhagic stroke (aHR, 1.45; 95% CI, 1.13–1.85). Hypertension, drug abuse, and alcohol abuse were also associated with increased stroke risk. Non-obesity and being non-Hispanic were protective factors. In sensitivity analyses, healthcare utilization was a small but significant predictor of stroke.

Conclusions:

In the largest known investigation of PTSD and antidepressant-associated risk for hemorrhagic stroke in young adults, use of SSRIs, but neither PTSD nor SNRIs were independently associated with incident stroke. SNRIs may be preferable for treating PTSD and comorbid conditions, although pursuing other modifiable risk factors and non-pharmacological treatments for PTSD also remains essential.

Introduction

In adults less than 45 years old, hemorrhagic stroke, including subarachnoid and intracerebral hemorrhage, represent up to 50% of incident stroke cases,^1,2 and is more common earlier in adulthood.³ As the economic burden and physical ramifications of hemorrhagic stroke may exceed that of ischemic stroke,¹ hemorrhagic stroke may herald particularly grave and long-lasting consequences for patients who are younger. Correspondingly, understanding the previously undescribed or underappreciated risk factors for incident hemorrhagic stroke among younger adults is of paramount importance.

Posttraumatic stress disorder (PTSD) is a prevalent and treatable mental health condition among younger adults,⁴ particularly Veterans,⁵ that develops from exposure to severe psychological stress or trauma and certain contexts.^6,7 PTSD has been associated with increased ischemic stroke risk among younger adults and may also confer vulnerability for early-onset hemorrhagic stroke by modulating inflammatory and atherosclerotic pathways.⁸ Furthermore, patients with PTSD often have more stroke risk factors, including hypertension and polysubstance use, than those with other anxiety disorders.^9,10 Although psychological stress has been linked to increased hemorrhagic stroke among adults aged 54 to 64,^11,12 it is unclear if PTSD increases the risk of hemorrhagic stroke overall, and particularly among those who are younger.

Selective serotonin reuptake inhibitors (SSRIs) and serotonin-norepinephrine reuptake inhibitors (SNRIs) are widely used as first-line pharmacotherapies for PTSD (e.g., fluoxetine, sertraline, venlafaxine) because of their safety and tolerability,¹³ and are among the most widely prescribed drugs in the United States (U.S.).^14,15 Due to their antiplatelet effects, SSRI/SNRI use increases the risk of abnormal bleeding, including intracranial hemorrhage.^16,17 While plausible, whether PTSD alone, SSRI or SNRI use alone, or PTSD compared to SSRI/SNRI use increases incident hemorrhagic stroke risk among young adults remains unknown.

In this study, we assessed the independent and combined risk for incident hemorrhagic stroke that is associated with PTSD and the use of SSRI/SNRI medication in a nationwide sample of 1.1 million young and middle-aged U.S. Veterans who served in support of Iraq and Afghanistan conflicts (Operations Enduring Freedom, Iraqi Freedom, or New Dawn [OEF/OIF/OND]). We hypothesized that PTSD and SSRI/SNRI use would each be associated with a significant risk of hemorrhagic stroke, but that use of antidepressant medication would be the primary predictor of stroke, more so than the effect of PTSD.

Methods

Patient Population and Data Sources

Although data sharing agreements prohibit us from making the dataset publicly available, access may be granted from the corresponding author upon receiving a reasonable request. The sample was comprised of Veterans from the roster maintained by the Department of Defense’s Manpower Data Center, consisting of those who served in support of OEF/OIF/OND and enrolled in Veterans Affairs (VA) care (N=1,063,973). Further details concerning the dataset are found elsewhere.^8,10 All patients had at least 1 inpatient or 2 outpatient visits at a VA facility from October 1, 2001 through November 1, 2014, and no history of ischemic or hemorrhagic stroke, or transient ischemic attack, at their first VA encounter. The final sample was 988,090. There were no demographic differences between included and excluded patients. The VA Connecticut Healthcare System approved procedures and waiver of informed consent.

Clinical diagnoses, administrative, and pharmacy data were acquired through the VA Corporate Data Warehouse, including outpatient encounters and inpatient hospitalizations (dates and International Classification of Diseases, Ninth Revision, Clinical Modification [ICD-9-CM] codes). Baseline was specified as the date of a patient’s first VA clinical encounter. The earliest stroke was classified as the index event. Diagnoses were included if the ICD-9-CM code was recorded during 1 hospitalization or 2 outpatient encounters prior to the index stroke to ensure accuracy.⁸ Diagnoses and prescriptions after the index stroke were excluded. Follow-up time extended from the date of the first clinical encounter until a stroke event or the last date of follow-up.

Hemorrhagic Stroke

The primary outcome was the first occurrence of hemorrhagic stroke, defined as new ICD-9-CM codes (430: subarachnoid hemorrhage, 431: intracerebral hemorrhage, 432: other and unspecified intracranial hemorrhage) during the observation period. Codes are concordant with formal adjudication.¹⁸

PTSD

A PTSD diagnosis was required on at least 1 inpatient or 2 outpatient visits (ICD-9-CM code: 309.81). There criteria were also used to identify mental health diagnoses in VA administrative data.

SSRI/SNRI Medication

Patients were categorized as taking an SSRI if their medication record listed citalopram, escitalopram, fluoxetine, fluvoxamine, paroxetine, and sertraline or an SNRI if the record included desvenlafaxine, duloxetine, or venlafaxine from baseline through follow-up.

Other Risk Factors for Stroke

A priori covariates included demographic data (sex, race/ethnicity [Black, Hispanic, White, Other], and marital status), body mass index (BMI; dichotomized as ≥30 kg/m²:obese vs. non-obese¹⁴) and smoking status [current, former, never]) collected closest to baseline. Validated ICD-9-CM codes were used to identify diagnoses of hypertension, traumatic brain injury (TBI), atrial fibrillation, lipid disorders, vascular disease, chronic kidney disease (CKD), diabetes mellitus, chronic obstructive pulmonary disease (COPD), major depressive disorder (MDD), generalized anxiety disorder (GAD), PTSD, alcohol abuse, and drug abuse (aggregated from abuse of cocaine, opioids, sedative/hypnotics, other drugs, and polysubstance abuse). Baseline systolic and diastolic blood pressure were also included. VA pharmacy records were used to obtain prescription data for antihypertensive medications (e.g., angiotensin-converting enzyme inhibitors, angiotensin receptor blockers, beta-blockers, diuretics), SSRI, and SNRI. Healthcare utilization was represented by the number of primary care visits in the 2 years after military discharge.

Statistical Analysis

Analyses were conducted using SAS V9.4. A two-sided p-value of <0.05 indicated statistical significance. There was less than 5% missing data for all variables except smoking status (26%), BMI and TBI (22%), and healthcare utilization (12%). Missing data were rectified using regression-based multiple imputation with 5 imputed datasets. There were no significant differences between the original dataset and the product of multiple imputation. Next, we calculated the incidence rates of stroke per 10,000 person‐years for hemorrhagic stroke overall, and then by PTSD, SSRI, and SNRI status. We then used Kaplan-Meier survival analyses to construct plots displaying participants’ probability of stroke over time, stratified by the three groups. Log‐rank tests were used to determine if time to event differed by exposure. For each patient, follow-up time was assessed separately and concluded on whichever occurred first: the first stroke event or the date of data censoring (last VA encounter). As associations between PTSD, use of SSRI/SNRI medication, hemorrhagic stroke, and other clinical variables were time-dependent, variables were treated as time-varying in statistical models, with each risk factor coded as 0 prior to, and 1 following, the diagnosis date. Clinical diagnoses and prescriptions that occurred after the index stroke were excluded.

Demographic and clinical characteristics are presented overall, and according to PTSD, and SSRI/SNRI status for patients with an incident hemorrhagic stroke. Groups were compared using t-tests and the chi-square statistic. Independent effects of PTSD, and SSRI/SNRI use on risk for hemorrhagic stroke were initially tested separately using two multivariate-adjusted time-varying Cox proportional-hazards models with 95% confidence intervals (CIs). Unadjusted effects of each predictor were examined first. Next a series of four models were created to sequentially adjust for potential confounders: 1) demographics (age, sex [Female as the reference group], and race/ethnicity [White as the reference]); 2) lifestyle factors (obesity, smoking status [“Never smoked” as the reference], alcohol and drug abuse); 3) prevalent stroke comorbidities (hypertension, lipid disorders, and diabetes), and; 4) psychiatric comorbidities (MDD and GAD). Tests of the psychiatric comorbidities and PTSD indicated that multicollinearity was not a concern (Tolerance=0.849–0.891, variance inflation factor=1.122–1.178). Unadjusted and adjusted models were also constructed with PTSD and SSRI/SNRI to compare the independent effects on risk of incident hemorrhagic stroke.

As patients with a PTSD diagnosis or those taking SSRI/SNRI medication may be more likely to seek medical care than comparison groups, we performed sensitivity analyses to assess for surveillance bias. Therefore, healthcare utilization was included in the 2 fully adjusted models for PTSD and SSRI/SNRI status, respectively, and in the fully adjusted model including all predictors.

Results

Table 1 provides characteristics from the complete cohort and for patients with a diagnosis of hemorrhagic stroke by PTSD diagnosis and SSRI/SNRI use. Overall, the cohort was young, with a mean±SD age of 30.18±9.19 years. Most patients were male (88%), White (64%), non-Hispanic (89%), and unmarried (54%). A quarter of the cohort was obese (26%) and over half were current or former smokers (57%). The most common diagnoses were PTSD (28%), lipid disorders (18%), and hypertension (13%). Approximately 37% and 7% of the cohort was prescribed an SSRI or SNRI, respectively. Characteristics of those with a stroke diagnosis were also compared based on receipt of SSRI or SNRI medication (see Data Supplement Table I).

Table 1.

Cohort Characteristics by Index Stroke, PTSD Diagnosis and Use of SSRI/SNRI Medication

	TOTAL (N=988,090)		HEMORRHAGIC STROKE (N=507)
		P*	PTSD (n=240)	No PTSD (n=267)	P	SSRI/SNRI (n=349)	No SSRI/SNRI (n=158)	P
Demographics
Age (years)†	30.18±9.19	<0.001	30.66±9.83	32.96±10.87	0.01	31.39±10.36	32.94±10.58	0.12
Male	867,970(87.84)	0.07	214(89.17)	245(91.76)	0.32	312(89.40)	147(93.04)	0.19
Race/Ethnicity
White	637,113(64.48)	<0.001	158(65.83)	149(55.81)	0.17	219(62.75)	88(55.70)	0.62
Black	155,871(15.77)		36(15.00)	48(17.98)		55(15.76)	29(18.35)
Hispanic	109,179(11.05)		35(14.58)	48(14.58)		55(15.76)	28(17.72)
Married	451,301(45.67)	0.24	116(48.33)	126(47.19)	0.55	171(49.00)	71(44.94)	0.54
Obesity	256,603(25.96)	<0.001	93(38.75)	99(37.08)	0.70	136(38.97)	56(35.44)	0.45
Smoking Status
Current	309,703(42.31)	<0.001	139(57.92)	86(32.21)	<0.001	175(50.14)	50(31.65)	<0.001
Former	104,705(14.31)		35(14.58)	47(17.60)		48(13.75)	34(21.52)
Never	317,493(43.38)		66(27.50)	134(50.19)		126(36.10)	74(46.84)
Blood Pressure
Systolic blood pressure	127.03±14.50	0.02	129.4±18.42	127.7±16.88	0.28	128.63±18.24	128.30±16.26	0.85
Diastolic blood pressure	77.28±11.22	0.04	78.68±13.42	78.09±13.25	0.62	78.37±13.58	78.37±12.77	0.99
Stroke Subtype					0.33			0.20
Subarachnoid	104(0.01)	<0.001	52(21.67)	52(19.48)		75(21.49)	29(18.35)
Intracerebral	173(0.02)	<0.001	74(30.83)	99(37.08)		125(35.82)	48(30.38)
Other/unspecified	230(0.02)	<0.001	114(47.50)	116(43.45)		149(42.69)	81(51.27)
Stroke Comorbidities
Hypertension	129,613(13.12)	<0.001	99(41.25)	110(41.20)	0.99	151(43.27)	58(36.71)	0.16
Lipid disorders	178,593(18.07)	<0.001	89(37.08)	85(31.84)	0.21	128(36.68)	46(29.11)	0.10
Traumatic brain injury	51,813(5.24)	<0.001	130(54.17)	115(43.07)	0.01	168(48.14)	77(48.73)	0.90
Atrial fibrillation	2,491(0.25)	<0.001	3(1.25)	5(1.87)	0.57	5(1.43)	3(1.90)	0.70
Chronic kidney disease	4,306(0.42)	<0.001	7(2.92)	9(3.37)	0.77	12(3.44)	4(2.53)	0.59
Diabetes	29,559(2.99)	<0.001	20(8.33)	28(10.49)	0.41	30(8.60)	18(11.39)	0.32
COPD	12,869(1.30)	<0.001	27(11.25)	19(7.12)	0.11	38(10.89)	8(5.06)	0.03
Psychiatric History
Major depressive disorder	95,132(9.63)	<0.001	90(37.50)	27(10.11)	<0.001	112(32.09)	5(3.16)	<0.001
Generalized anxiety disorder	155,687(15.76)	<0.001	111(46.25)	59(22.10)	<0.001	149(42.69)	21(13.29)	<0.001
PTSD	276,141(27.95)	<0.001	240(100)	0(0)	-	223(53.90)	17(10.76)	<0.001
Substance Use
Alcohol abuse	107,517(10.88)	<0.001	103(42.92)	45(16.85)	<0.001	121(34.67)	27(17.09)	<0.001
Drug abuse	53,541(5.42)	<0.001	78(32.50)	17(6.37)	<0.001	86(24.64)	9(5.70)	<0.001
Medications
Antihypertensive therapy	122,029(12.35)	<0.001	112(46.67)	104(38.95)	0.08	162(46.42)	54(34.18)	0.01
SSRI	376,005(37.03)	<0.001	218(90.83)	123(46.07)	<0.001	341(97.71)	8(2.29)	<0.001
SNRI	67,630(6.66)	<0.001	64(26.67)	21(7.87)	<0.001	85(24.36)	264(75.64)	<0.001
Healthcare Utilization†	4.09±4.05	<0.001	8.56±12.05	9.42±23.87	0.61	9.17±18.78	8.67±20.14	0.79

Abbreviations: COPD, chronic obstructive pulmonary disease; PTSD, posttraumatic stress disorder; SSRI and SNRI, selective serotonin and norepinephrine reuptake inhibitors;

^*Comparing those with and without a stroke.

^†Mean±SD and N(%) otherwise.

During an average follow-up period of 2.14±2.67 years, 507 patients (48 women) suffered a hemorrhagic stroke. Overall stroke incidence was 1.70 events/10,000-person years. The absolute incidence rates for PTSD, SSRI, and SNRI use were 1.83, 2.97, and 2.18, respectively. A PTSD diagnosis and SSRI use were each associated with a greater risk of stroke across follow-up (ps<0.001), but this association was not found for SNRIs (p=0.10; Figures 1a–1c). Those with a diagnosis of PTSD before stroke were more likely to be current smokers, to have a history of TBI, a diagnosis of MDD, GAD, alcohol abuse, and drug abuse, and to be prescribed SSRI/SNRI. Patients who were prescribed an SSRI/SNRI prior to stroke were more likely to smoke, to be diagnosed with COPD and all psychiatric and substance use conditions, and to be prescribed antihypertensive therapy.

Figure.

Survival curves depicting risk for hemorrhagic stroke by posttraumatic stress disorder (PTSD; A), selective serotonin reuptake inhibitors (SSRI; B), and selective norepinephrine reuptake inhibitors (SNRI) status (C).

In unadjusted model, PTSD was associated with a significantly greater risk of hemorrhagic stroke (hazard ratio [HR], 1.82; 95% CI, 1.48–2.24; p=0.03). After adjusting for demographics, lifestyle factors, and stroke comorbidities, the association between PTSD and hemorrhagic stroke remained (Table 2, Model 3); however, when psychiatric comorbidities were added, the PTSD and hemorrhagic stroke risk relationship became non-significant (adjusted HR [aHR], 1.22, 95% CI, 0.97–1.53; Model 4). In the fully-adjusted model, patients who were Hispanic showed a greater risk of stroke (aHR, 1.49, 95% CI, 1.16–1.93). Additionally, obesity, drug and alcohol abuse, and diagnoses of MDD and GAD were each associated with increased risk of hemorrhagic stroke, with hypertension as the leading risk factor (aHR, 3.31, 95% CI, 2.58–4.25).

Table 2.

Multivariate Models of PTSD and Risk for Incident Hemorrhagic Stroke^*

	Model 1		Model 2		Model 3		Model 4
PTSD	1.84(1.49–2.26)	<0.001	1.41(1.12–1.76)	0.003	1.32(1.06–1.66)	0.01	1.22(0.97–1.53)	0.10
Demographics
Age	1.02(1.01–1.02)	0.001	1.02(1.01–1.03)	<0.001	1.01(0.99–1.02)	0.09	1.01(0.99–1.02)	0.08
Female	0.75(0.55–1.02)	0.08	0.80(0.58–1.09)	0.15	0.85(0.62–1.16)	0.30	0.81(0.60–1.11)	0.20
Black	0.99(0.77–1.27)	0.008	0.90(1.02–1.31)	0.77	0.90(0.70–1.17)	0.44	0.94(0.73–1.22)	0.65
Hispanic	1.40(1.09–1.80)	0.009	1.45(1.13–1.87)	0.004	1.44(1.12–1.86)	0.005	1.49(1.16–1.93)	0.002
Lifestyle Factors
Obesity			1.45(1.21–1.75)		1.59(1.32–1.92)	<0.001	1.59(1.32–1.92)	<0.001
Current smoker			1.22(0.93–1.60)	0.15	1.22(0.93–1.59)	0.15	1.22(0.93–1.59)	0.15
Former smoker			1.07(0.87–1.31)	0.55	1.06(0.86–1.30)	0.58	1.06(0.86–1.30)	0.59
Drug abuse†			2.01(1.41–2.88)	<0.001	1.99(1.39–2.85)	<0.001	1.84(1.28–2.64)	<0.001
Alcohol abuse†			2.58(1.95–3.43)	<0.001	2.40(1.81–3.19)	<0.001	2.28(1.72–3.02)	<0.001
Stroke Comorbidities†
Hypertension					3.39(2.64–4.35)	<0.001	3.31(2.58–4.25)	<0.001
Lipid disorders					0.81(0.61–1.08)	0.15	0.80(0.60–1.07)	0.13
Diabetes					1.22(0.76–1.97)	0.41	1.22(0.76–1.97)	0.41
Psychiatric Comorbidities†
MDD							1.37(1.01–1.85)	0.04
GAD							1.38(1.07–1.79)	0.01

Abbreviations: GAD, generalized anxiety disorder; MDD, major depressive disorder; PTSD, posttraumatic stress disorder;

^*Hazard ratios (HRs), 95% CIs, and P values.

^†Diagnoses are prior to the index stroke event and are time-dependent.

In the unadjusted model, use of SSRI (HR, 2.02; 95% CI, 1.66–2.47, p<0.001) and SNRI (HR, 1.52; 95% CI, 1.08–2.16, p=0.02) were each associated with a greater risk of hemorrhagic stroke. In fully-adjusted analyses (Table 3, Model 4), only SSRIs remained a significant predictor (aHR, 1.47, 95% CI, 1.17–1.84; p=0.001). Hispanic ethnicity, obesity, drug and alcohol abuse were each associated with a significantly greater risk of incident hemorrhagic stroke. Hypertension was again the primary independent risk factor for stroke (aHR, 3.25, 95% CI, 2.53–4.17; p<0.001). Notably, neither a diagnosis of MDD or GAD was significant.

Table 3.

Multivariate Models of SSRI or SNRI Medication Use and Risk for Incident Hemorrhagic Stroke^*

	Model 1		Model 2		Model 3		Model 4
SSRI	2.05(1.68–2.50)	<0.001	1.69(1.37–2.09)	<0.001	1.58(1.27–1.96)	<0.001	1.47(1.17–1.84)	0.001
SNRI	1.55(1.10–2.19)	0.01	1.35(0.95–1.92)	0.09	1.29(0.91–1.83)	0.16	1.20(0.84–1.71)	0.32
Demographics
Age	1.02(1.01–1.02)	0.002	1.02(1.01–1.03)	<0.001	1.01(0.99–1.02)	0.08	1.01(1.00–1.02)	0.08
Female	0.70(0.52–0.96)	0.03	0.76(0.56–1.04)	0.08	0.76(0.56–1.04)	0.09	0.80(0.58–1.09)	0.15
Black	1.02(0.79–1.31)	0.90	1.04(0.81–1.33)	0.78	0.92(0.71–1.19)	0.53	0.93(0.72–1.20)	0.59
Hispanic	1.42(1.10–1.83)	0.007	1.46(1.13–1.88)	0.004	1.45(1.12–1.87)	0.004	1.45(1.12–1.87)	0.004
Lifestyle Factors
Obesity			1.47(1.22–1.77)	<0.001	1.60(1.32–1.93)	<0.001	1.60(1.32–1.93)	<0.001
Current smoker			1.22(0.93–1.59)	0.14	1.21(0.93–1.59)	0.16	1.03(0.93–1.59)	0.16
Former smoker			1.03(0.83–1.26)	0.81	1.03(0.83–1.26)	0.81	1.03(0.84–1.27)	0.77
Drug abuse†			1.90(1.33–2.72)	<0.001	1.89(1.32–2.70)	<0.001	1.81(1.26–2.59)	0.001
Alcohol abuse†			2.40(1.82–3.17)	<0.001	2.26(1.71–3.00)	<0.001	2.19(1.66–2.90)	<0.001
Stroke Comorbidities†
Hypertension					3.28(2.56–4.21)	<0.001	3.25(2.53–4.17)	<0.001
Lipid disorders					0.80(0.60–1.06)	0.11	0.80(0.59–1.05)	0.10
Diabetes					1.22(0.75–1.96)	0.42	1.22(0.75–1.96)	0.42
Psychiatric Comorbidities†
MDD							1.25(0.92–1.69)	0.16
GAD							1.26(0.97–1.64)	0.09

Abbreviations: GAD, generalized anxiety disorder; MDD, major depressive disorder; PTSD, posttraumatic stress disorder

^*Hazard ratios (HRs), 95% CIs, and P values.

^†Diagnoses are prior to the index stroke event and are time-dependent.

In the unadjusted model comparing PTSD, SSRI, and SNRI use, SSRI/SNRI use remained significant predictors of incident hemorrhagic stroke (SSRI:HR, 1.84; 95% CI, 1.46–2.31, p<0.001; SNRI:HR, 1.46; 95% CI, 1.03–2.07, p=0.03) while PTSD was not (HR, 1.24; 95% CI, 0.98–1.58, p=0.08). In the fully adjusted model, SSRI use remained significantly associated with stroke (HR, 1.45; 95% CI, 1.13–1.85, p<0.001; Table 4, Model 4), while PTSD and SNRI use were not. Hispanic ethnicity, obesity, alcohol and drug abuse, and hypertension were again predictive of stroke. Model fitness improved from the model with demographics only (Model 1: c-statistic=0.56) to the fully-adjusted model (Model 4: c-statistic=0.71, p for improvement <0.001). In exploratory analyses, the stroke variable was separated into subtypes: subarachnoid, intracerebral, and other/unspecified (Supplemental Tables II–X). In the fully-adjusted model of all predictors, SSRI use was associated with a greater incidence of subarachnoid stroke only (HR, 1.89; 95% CI, 1.14–3.13, p=0.01).

Table 4.

Multivariate Models of PTSD, SSRI/SNRI Medication Use, and Risk for Incident Hemorrhagic Stroke^*

	Model 1		Model 2		Model 3		Model 4
PTSD	1.24(0.97–1.58)	0.08	1.07(0.83–1.37)	0.60	1.05(0.82–1.34)	0.71	1.03(0.81–1.34)	0.80
SSRI	1.86(1.48–2.34)	<0.001	1.64(1.30–2.09)	<0.001	1.55(1.22–1.97)	<0.001	1.45(1.13–1.85)	0.003
SNRI	1.49(1.05–2.11)	0.03	1.34(0.94–1.90)	0.11	1.28(0.90–1.81)	0.17	1.19(0.83–1.71)	0.34
Demographics
Age	1.02(1.01–1.03)	<0.001	1.02(1.01–1.03)	<0.001	1.01(0.99–1.02)	0.08	1.01(0.99–1.02)	0.07
Female	0.72(0.53–0.98)	0.04	0.76(0.56–1.04)	0.09	0.82(0.60–1.12)	0.20	0.80(0.58–1.09)	0.16
Black	1.01(0.79–1.30)	0.92	1.04(0.80–1.33)	0.80	0.92(0.71–1.19)	0.53	0.93(0.72–1.20)	0.59
Hispanic	1.42(1.10–1.82)	0.007	1.46(1.13–1.88)	0.004	1.45(1.12–1.87)	0.004	1.45(1.12–1.87)	0.004
Lifestyle Factors
Obesity			1.47(1.22–1.77)	<0.001	1.60(1.33–1.93)	<0.001	1.60(1.32–1.93)	<0.001
Current smoker			1.21(0.93–1.59)	0.16	1.21(0.93–1.59)	0.16	1.21(0.93–1.59)	0.16
Former smoker			1.03(0.83–1.26)	0.81	1.03(0.83–1.26)	0.80	1.03(0.84–1.27)	0.77
Drug abuse†			1.89(1.32–2.71)	0.002	1.88(1.31–2.69)	<0.001	1.80(1.26–2.59)	<0.001
Alcohol abuse†			2.37(1.79–3.14)	<0.001	2.24(1.69–2.96)	<0.001	2.18(1.65–2.89)	<0.001
Stroke Comorbidities†
Hypertension					3.28(2.56–4.41)	<0.001	3.24(2.53–4.16)	<0.001
Lipid disorders					0.79(0.60–1.06)	0.12	0.80(0.59–1.05)	0.10
Diabetes					1.22(0.75–1.96)	0.42	1.22(0.75–1.96)	0.42
Psychiatric Comorbidities†
MDD							1.24(0.92–1.69)	0.17
GAD							1.26(0.97–1.64)	0.09

Abbreviations: GAD, generalized anxiety disorder; MDD, major depressive disorder; PTSD, posttraumatic stress disorder

^*Hazard ratios (HRs), 95% CIs, and P values.

^†Diagnoses are prior to the index stroke event and are time-dependent.

In a sensitivity analysis, healthcare utilization (i.e., number of primary care visits) was added to see if the variable would reduce the effects observed in primary analyses. In fully-adjusted separate models of PTSD and SSRI/SNRI, more frequent utilization was associated with a slightly greater risk of stroke in the PTSD model (utilization aHR, 1.02; 95% CI, 1.01–1.02, p<0.001), the SSRI/SNRI model (utilization aHR, 1.02; 95% CI, 1.01–1.02, p<0.001), and the fully-adjusted model with all predictors (utilization aHR, 1.02, 95% CI, 1.01–1.02; p<0.001).

Discussion

In this large nationwide sample of 1.1 million young and middle-aged Veterans, the usual risk factors for hemorrhagic stroke in later life –hypertension, obesity, and substance abuse^19,20 – were associated with hemorrhage, while being non-Hispanic was protective. Patients with PTSD showed a significantly greater risk of hemorrhagic stroke in unadjusted analyses, supporting our hypothesis that young adults with PTSD are particularly vulnerable, and corresponding with earlier work among Veterans and the general population.^8,11,12 However, the effect of PTSD became non-significant in adjusted models. In other unadjusted analyses, use of SSRI and SNRI medication were each independent predictors of hemorrhagic stroke. Yet, in fully-adjusted and combined models with PTSD, SSRI medication emerged as the leading predictor of stroke whereas the underlying conditions indicating use of those medications (i.e., PTSD, GAD, MDD) were not. While SSRI/SNRI use confers a greater risk of brain hemorrhage among older adults,^17,21 SSRI, but not SNRI use appears to play a similar role in the etiology of hemorrhagic stroke among younger patients.^20,22

Modifiable Risk Factors

SSRIs are the first-line treatment recommendation for PTSD and related conditions, suggesting that patients taking SSRIs require more vigilance in risk factor control, which could be the most effective way to mitigate the risk of hemorrhage. Obesity, hypertension, and drug and alcohol abuse were each independently associated with an increased risk of stroke, mirroring previous observations,^22–23 and surprisingly, smoking was not associated with stroke. Although unexplored herein, there may be dose-response relationships between hypertension, tobacco use, and risk of hemorrhage. While rates of traditional vascular risk factors were lower than in older samples,²⁴ cerebrovascular risk may still be more pronounced among younger adults,^3,23 suggesting that treating traditional risk factors is essential for stroke prevention.^25,26 Sensitivity analyses revealed that greater healthcare utilization was also associated with an increased risk of hemorrhagic stroke, indicating more opportunities to discuss, educate, and treat modifiable risk factors among high-risk patients, with several avenues for patient-provider collaboration: weight, blood pressure, and substance abuse.¹⁴ In another national cohort of adults aged 18–45, including records from 325 million visits, modifiable risk factors for stroke were identified in >20% of patients during primary care visits.²⁷ In this study, hypertension was associated with a 3-fold greater risk of stroke. Prioritizing hypertension management as part of a multipronged behavioral approach is likely a particularly impactful stroke prevention strategy for younger patients.

PTSD, Other Psychiatric Conditions, and Hemorrhagic Stroke

A PTSD diagnosis is increasingly associated with a greater risk of ischemic stroke among younger patients,⁸ and with incident hemorrhagic stroke among middle-aged and older patients.^11,12 Thus, it was surprising that there was not a similar, independent effect of PTSD in fully-adjusted models of hemorrhagic stroke. The most likely explanation relates to other psychiatric diagnoses that are highly comorbid with PTSD – MDD and GAD.²⁸ PTSD was significantly associated with incident hemorrhagic stroke until psychiatric comorbidities were introduced. No condition that is commonly treated with SSRI/SNRI remained a significant predictor in the final model, but SSRI use was associated with a greater than 2-fold risk of hemorrhagic stroke. Thus, the psychotherapeutics, rather than the underlying mental health conditions per se, may elevate vascular risk in younger adults with PTSD. Cognitive behavioral therapy (i.e., CBT) offers an excellent but comparatively underused tool for the integrated treatment of PTSD and other behavioral risk factors for stroke.²⁹ CBT could dually reduce patients’ use of SSRIs and risk of brain hemorrhage.

Antidepressants and Mechanisms

Our findings add to a growing, cautionary literature indicating that SSRI use contributes to intracranial bleeding and the potential for hemorrhage.^16,17,21,30 SSRIs likely increase cerebrovascular risk via serotonin inhibition in the brain and peripheral platelets.^17,30,31 Medications that inhibit serotonin reuptake are recommended by the VA and other organizations to treat PTSD, and other psychiatric and non-psychiatric conditions.^13,31–33 As the level of inhibition differs by medication,^17,34 some therapeutics may involve a greater inherent risk of bleeding and stroke.³⁵ It is unclear if the effects of SSRIs are dose-dependent, but SSRIs may increase the risk of subarachnoid stroke in particular. Furthermore, concurrently using SSRIs and other agents known to increase risk of hemorrhagic stroke (e.g., anticoagulants), may lead to worse cerebrovascular consequences.^30,31 To appropriately prescribe SSRIs, clinicians should weigh a patient’s medical history and the potential for adverse vascular effects with their needs and potential advantages. For example, SSRIs are often given post-stroke to improve cognitive recovery and reduce the risk of depression.¹⁷ Non-pharmacological treatments should be considered for younger patients who are more vulnerable to hemorrhagic stroke.

Limitations

As this cohort consisted of primarily younger, male, OEF/OIF/OND Veterans, analyses should be replicated with a larger proportion of women, and results may not be generalizable to older Veterans or to adults without a history of military service. However, other studies of non-Veterans also indicate that SSRI/SNRI medications increase the risk of stroke.^16,17,21,30 Although we examined three different types of hemorrhage, some stroke cases may also have been misclassified, potentially altering the results. SSRI use may also be associated with other types of stroke in this population, which should be explored subsequently. As a well-known stroke risk factor, TBI was included in sample characteristics but information was unavailable related to skull fractures, concussion, or other severe head injuries and potential stroke predictors. TBI was excluded from the model, allowing us to investigate other, more nuanced effects. Thus, findings may not generalize to those with PTSD and TBI. As this sample was young, only 8 patients with incident stroke had a diagnosis of atrial fibrillation, and few were likely taking an anticoagulant, so those variables were not controlled in analyses. Although not assessed, other medications may also exacerbate the hazard associated with SSRIs. Finally, we could not distinguish between persistent vs. remitted PTSD and lacked information about cause of death for any patients who may have died during follow-up. A competing risk model should be used to examine deaths, and suicide more specifically, among those with and without PTSD and using antidepressants.

Summary

In a cohort of 1.1 million young Veterans, using SSRI medication conferred a greater independent risk for early hemorrhagic stroke compared to a diagnosis of PTSD or use of SNRI medication. SSRIs are commonly prescribed to Veterans and to manage the mental health sequelae of trauma in non-VA settings. Further research is needed to understand the cerebrovascular consequences of SSRI use in younger adults treated for PTSD and other psychiatric conditions. Additional granular studies are required to examine effects of medication subtypes and hemorrhagic stroke risk among younger adults to inform clinical recommendations. Providers should review the risk of hemorrhagic stroke with patients when prescribing SSRIs. Pursuing non-pharmacological treatments also remains essential to mitigate the risk of stroke among young adults.

Non-standard Abbreviations and Acronyms

aHR

Adjusted hazard ratio

GAD

Generalized anxiety disorder

HR

Hazard ratio

ICD-9-CM

International Classification of Diseases, Ninth Revision, Clinical Modification

MDD

Major depressive disorder

PTSD

Posttraumatic stress disorder

SNRI

Selective norepinephrine reuptake inhibitors

SSRI

Selective serotonin reuptake inhibitors

TBI

traumatic brain injury

VA

Veterans Affairs medical center