Sex Differences in the Risk of Stroke Associated With Traditional and Non-Traditional Factors in a US Cohort of People With HIV Infection

Abstract

Background and Objectives

Although stroke risk associated with HIV may be greater for women than men, little is known about whether the impact of different factors on cerebrovascular risk varies by sex in people with HIV (PWH) and contributes to stroke risk disparities in this population. The primary objective of this study was to examine whether sex modifies the effect of demographics, cardiometabolic factors, health-related behaviors, and HIV-specific variables on stroke risk in PWH from the Centers for AIDS Research Network of Integrated Clinical Systems (CNICS) cohort.

Methods

In this observational cohort study, we analyzed data from clinical encounters for PWH followed at 5 CNICS sites from approximately 2005 to 2020. All potential stroke events were adjudicated by neurologists. Patient-reported outcomes collected at clinic visits, including substance use and depression, were also available. We used Cox proportional hazards models to determine whether sex modified the association of predictors of interest with incident stroke.

Results

Among 13,573 PWH (19% female sex at birth, mean age 44 years, mean follow-up 5.6 years), female sex was associated with a higher risk of stroke only among individuals aged 50 years or younger (hazard ratio [HR] 2.01 at age 40 [1.25–3.21] vs HR 0.60 at age 60 [0.34–1.06]; p = 0.001 for the interaction). Younger female participants who developed a stroke were more likely to have treated hypertension, a higher cardiovascular risk score, and detectable HIV than younger male participants whereas these factors were comparable by sex among older participants who developed a stroke. Sex modified the effect of detectable HIV (HR 4.66 for female participants [2.48–8.74] vs HR 1.30 for male participants [0.83–2.03]; p = 0.001 for the interaction), methamphetamine use (HR 4.78 for female participants [1.47–15.56] vs HR 1.19 for male participants [0.62–2.29]; p = 0.04 for the interaction), and treated hypertension (HR 3.44 for female participants [1.74–6.81] vs HR 1.66 for male participants [1.14–2.41]; p = 0.06 for the interaction) on stroke risk.

Discussion

Younger female participants with HIV were at elevated cerebrovascular risk compared with younger male participants. Several risk factors had a greater adverse effect on stroke risk in female participants than in male participants, including HIV viremia, methamphetamine use, and treated hypertension. These findings underscore the importance of a personalized approach to predict and prevent cerebrovascular risk among PWH.

Introduction

The increased risk of stroke associated with HIV infection may be more marked for women than for men, especially at younger ages.^1-4 Various mechanisms underlying this differential stroke risk by sex have been proposed. Excess cerebrovascular risk may be related to greater immune activation and inflammation and a stronger effect of inflammation on the development of cardiovascular disease (CVD) and stroke, in women compared with men with HIV.^5-7 Reduced ovarian reserve and estrogen depletion during the menopausal transition may lead to increased gut permeability, heightened immune activation, and higher incidence of inflammation-related comorbidities in women with HIV.^8,9 Although the impact of cardiometabolic factors on stroke risk varies by sex in the general population^10,11 and potentially also in people with HIV (PWH),¹ little is known about how these differences contribute to stroke risk disparities in PWH. Characterizing drivers of sex differences in stroke risk will provide insight into whether personalized cerebrovascular risk prediction tools and sex-specific guidelines are needed for stroke prevention for PWH.

We examined whether sex modifies the effect of cardiometabolic risk factors, health-related behaviors, and HIV-specific variables on stroke risk in PWH from the Centers for AIDS Research Network of Integrated Clinical Systems (CNICS) cohort, a longitudinal cohort of PWH in clinical care at multiple sites in the United States, composed of approximately 20% of female participants. All results are presented by sex, using language such as “female participant,” “male participant,” “individual assigned female sex at birth,” or “individual assigned male sex at birth.” However, in statements based on data from other studies, we generally use the terms “women” and “men” in line with the terms used in those references, most of which do not specify whether data captured were on sex at birth, gender, or both.

Methods

Study Design and Population

In this observational study, we analyzed data from CNICS, which includes comprehensive clinical information from all outpatient and inpatient encounters for PWH receiving care at 8 US sites. To be eligible for inclusion in the CNICS cohort, PWH must have had at least 2 visits within 12 months. For this study, individuals cared for at 5 sites participating in central stroke adjudication (Johns Hopkins University, University of Alabama at Birmingham, University of California San Diego, University of North Carolina at Chapel Hill, and University of Washington) were eligible. All sites routinely capture patient-reported outcomes through self-administered assessments at clinic visits.

Follow-up occurred from approximately 2005 to 2020, with end dates varying by site. The start of the observation period was defined as the latest of 3 possible dates: (1) initiation of stroke surveillance at each site, (2) first clinical assessment of patient-reported outcomes at each site, or (3) initial CNICS visit plus 6 months. The end of the observation period was defined as the earliest of 3 possible dates: (1) stroke event, (2) last CNICS visit plus 9 months, (3) death, or (4) administrative censoring. We excluded individuals with a history of stroke before the start of the observation period or those who did not complete at least one patient-reported outcome assessment before the end of stroke surveillance.

Stroke Outcomes, Predictors, and Covariates

Potential stroke events in CNICS are identified centrally from the data repository and adjudicated by neurologists, as previously described.¹² Events that are adjudicated as strokes are then classified by stroke type (i.e., ischemic stroke, intraparenchymal hemorrhage, and subarachnoid hemorrhage) and ischemic stroke subtype based on Trial of Org 10172 in Acute Stroke Treatment (TOAST) criteria.¹³ The outcome of interest in this study was a composite of all stroke types.

We obtained data on demographics, cardiometabolic comorbidities, health-related behaviors, and HIV-specific parameters from the start of the observation period. Most variables were gathered from the central data repository, including age, sex assigned at birth captured by self-report, race, ethnicity, laboratory values, medication prescriptions, medical diagnoses, and systolic and diastolic blood pressure. Treated hypertension was defined as hypertension requiring pharmacologic treatment. Diabetes mellitus was defined as (1) hemoglobin A1c >6.5% or (2) use of a diabetes-specific medication such as insulin or (3) use of a diabetes-related medication frequently but not exclusively used to treat diabetes (e.g., biguanides) and having a diagnosis of diabetes mellitus. We used the Chronic Kidney Disease Epidemiology Collaboration equation without race to calculate the estimated glomerular filtration rate (eGFR). We calculated the atherosclerotic cardiovascular disease (ASCVD) score, which predicts 10-year risk of atherosclerotic CVD using sex, age, race, systolic blood pressure, hypertension treatment, diabetes, tobacco use, and total and high-density lipoprotein (HDL) cholesterol. HIV-specific risk factors included CD4 cell count, HIV viral load, and antiretroviral therapy (ART) use.

Additional information was obtained from clinical assessments of patient-reported outcomes implemented in CNICS as part of clinical care visits.¹⁴ These assessments include questions on physical activity¹⁵; substance use from the Alcohol, Smoking and Substance Involvement Screening Test¹⁶; alcohol use from the Alcohol Use Disorders Identification Test¹⁷ (AUDIT-C; scores range from 0 to 12 with higher scores indicating a greater likelihood that alcohol use may be affecting a patient's health and safety); and depression with the Patient Health Questionnaire-9¹⁸ (PHQ-9; scores range from 0 to 27, with higher scores corresponding to more severe depression).

Statistical Analyses

Missing data were handled by multiple imputation with chained equations (m = 10). Predictors of interest were demographic and clinical characteristics that (1) were significantly associated with incident stroke in a multivariable Cox proportional hazards model or (2) differed between female and male participants who had an incident stroke. We created Cox proportional hazards models to determine whether sex at birth modified the association of predictors of interest with incident stroke after adjusting for age and a stroke risk score. The stroke risk score was generated by fitting a logistic regression model for the association of race and ethnicity, cohort entry year, site, mode of HIV transmission, systolic and diastolic blood pressure, cholesterol, triglycerides, eGFR, physical activity, cocaine (never/former/current), cigarette (never/former/current), and alcohol use with stroke and calculating the predicted probability of having a stroke given these participant characteristics. In sensitivity analyses, we used inverse probability of censoring weighting created with race, ethnicity, cohort entry year, site, HIV transmission mode, cigarette use, alcohol use, and physical activity. p < 0.05 was considered statistically significant.

Standard Protocol Approvals, Registrations, and Patient Consents

All patients provided written informed consent to participate in the CNICS study. All participating sites (Johns Hopkins University, University of Alabama at Birmingham, University of California San Diego, University of North Carolina at Chapel Hill, and University of Washington) received human subject approval for CNICS.

Data Availability

Data that support the findings of this study are available from the corresponding author on reasonable request and with the permission of the CNICS Research Coordinating Committee.

Results

Of 13,573 PWH included in the analysis, 19% were assigned female sex at birth, and the mean age was 44 years. The racial makeup of the cohort was similar for individuals who identify as White and Black, at 42% and 40%, respectively. The mean follow-up time was 5.6 years [standard deviation (SD) 3.4 years] overall, 5.8 years (SD 3.3 years) for female participants and 5.5 years (SD 3.5 years) for male participants. Treated hypertension was common, present in one-quarter of the cohort, as was current cigarette smoking in 38%. The mean CD4 count was 533 cells/mm³, and 19% had an HIV viral load >400 copies/mL. Additional demographic and clinical characteristics are listed in Table 1.

Table 1.

Baseline Demographic and Clinical Characteristics of People With HIV Overall and With and Without Incident Stroke

Characteristic, mean (SD) or N (%)	Overall (N = 13,573)	Incident stroke (N = 162)	No incident stroke (N = 13,411)
Stroke type
Ischemic	—	135 (83%)	—
Hemorrhagic	—	17 (10%)	—
Other	—	10 (6%)	—
Demographics
Age, y	44 (11)	51 (10)	44 (11)
Female sex at birth	2,541 (19)	41 (25)	2,500 (19)
Race and ethnicity
White	5,702 (42)	63 (39)	5,639 (42)
Black	5,416 (40)	82 (51)	5,334 (40)
Hispanic	1847 (14)	13 (8)	1834 (14)
Other	608 (4)	4 (2)	604 (5)
Cardiometabolic risk factors
Treated hypertension	3,448 (25)	82 (51)	3,366 (25)
Systolic blood pressure, mm Hg	125 (15)	131 (20)	125 (15)
Diastolic blood pressure, mm Hg	77 (11)	80 (12)	77 (11)
Total cholesterol, mg/dL	173 (42)	176 (43)	172 (42)
HDL cholesterol, mg/dL	42 (16)	46 (20)	42 (15)
LDL cholesterol, mg/dL	101 (35)	101 (31)	101 (35)
Statin use	2,290 (17)	56 (35)	2,234 (17)
Diabetes mellitus	1,157 (9)	32 (20)	1,125 (8)
Body mass index, kg/m2	26 (6)	27 (7)	26 (6)
eGFR, mL/min/1.73 m2	96 (21)	90 (24)	96 (21)
ASCVD risk score, %	5.8 (7.4)	12.2 (10.7)	5.7 (7.3)
Mental health and health-related behaviors
PHQ-9 score	5.8 (6.3)	6.2 (6.2)	5.8 (6.3)
Current methamphetamine use	1,102 (8)	13 (8)	1,089 (8)
Current cocaine use	949 (7)	12 (7)	937 (7)
Current injection drug use	228 (2)	2 (1)	226 (2)
Cigarette use
Never	5,189 (38)	56 (35)	5,133 (38)
Former	3,219 (24)	33 (20)	3,186 (24)
Current	5,165 (38)	73 (45)	5,092 (38)
AUDIT-C score	2.2 (2.5)	1.8 (2.3)	2.2 (2.5)
Physical activity in last month
None	1,912 (14)	23 (14)	1,889 (14)
Moderate	8,874 (65)	116 (72)	8,758 (65)
Vigorous	2,787 (21)	23 (14)	2,764 (21)
HIV-related and other variables
On antiretroviral therapy	12,678 (93)	156 (96)	12,522 (93)
Initiation of antiretroviral therapy within the past 6 mo	1,110 (8)	5 (3)	1,105 (8)
Detectable viral load (>400 copies/mL)	2,587 (19)	43 (27)	2,544 (19)
CD4 count (cells/mm3)	533 (306)	448 (331)	534 (306)
Hepatitis C infection	1,823 (13)	44 (27)	1,779 (13)

Abbreviations: ASCVD = Atherosclerotic Cardiovascular Disease; AUDIT-C = Alcohol Use Disorders Identification Test; eGFR = estimated glomerular filtration rate; HDL = high-density lipoprotein; LDL = low-density lipoprotein; PHQ-9 = Patient Health Questionnaire-9.

A total of 162 incident stroke events occurred during the observation period, of which most were ischemic stroke events (83%). Of the 135 ischemic stroke events, small vessel disease was the most common subtype (25%), followed by cardioembolic stroke (21%), stroke of undetermined etiology (21%), large vessel atherosclerosis (19%), and stroke of other determined etiology (14%).

The stroke incidence rate per 1,000 person-years by sex at birth and age was 1.2 (95% CI 0.9–1.6) for male participants aged 50 years or younger, 2.5 (95% CI 1.7–3.7) for female participants aged 50 years or younger, 4.1 (95% CI 3.3–5.3) for male participants older than 50 years, and 3.5 (95% CI 2.2–5.8) for female participants older than 50 years. In the overall cohort, older age (hazard ratio [HR] 1.63 per 10 years, 95% CI 1.35–1.97, p < 0.001) but not female sex (HR 1.15, 95% CI 0.76–1.74, p = 0.50) was associated with higher stroke risk (Table 2, Figure 1). Stratified by age group, younger female participants had higher rates of stroke than younger male participants but fewer stroke events compared with older female and male participants (Figure 2). Other variables associated with higher stroke risk included treated hypertension, statin use, diabetes, and detectable viral load.

Table 2.

Risk Factors for Incident Stroke in the Overall Cohort of People With HIV in Adjusted Analyses^a

	Hazard ratio	95% CI	p Value
Demographics
Age (per 10 y)	1.63	1.35–1.97	<0.001
Female sex at birth	1.15	0.76–1.74	0.50
Race and ethnicity
White	—	—	—
Black	1.12	0.74–1.70	0.58
Hispanic	0.67	0.36–1.25	0.21
Other	0.67	0.24–1.87	0.45
Cardiometabolic risk factors
Treated hypertension	1.53	1.06–2.22	0.02
Systolic blood pressure (per 10 mm Hg)	1.05	0.91–1.21	0.49
Diastolic blood pressure (per 10 mm Hg)	1.08	0.87–1.35	0.49
Total cholesterol (per 50 mg/dL)	0.91	0.52–1.60	0.75
HDL cholesterol (per 10 mg/dL)	1.14	0.96–1.36	0.12
LDL cholesterol (per 30 mg/dL)	0.96	0.68–1.35	0.82
Statin use	1.65	1.12–2.44	0.01
Diabetes mellitus	1.61	1.05–2.48	0.03
Body mass index (per 5 kg/m2)	1.12	0.96–1.32	0.15
eGFR (per 30 mL/min/1.73 m2)	0.95	0.73–1.23	0.69
Mental health and health-related behaviors
PHQ-9 score (per 5 points)	1.09	0.95–1.24	0.21
Current methamphetamine use	1.17	0.63–2.16	0.62
Current cocaine use	0.91	0.49–1.72	0.78
Cigarette use
Never	—	—	—
Former	0.88	0.57–1.37	0.58
Current	1.39	0.96–2.02	0.08
AUDIT-C score (per 5 points)	0.91	0.64–1.30	0.62
Physical activity in last month
None	—	—	—
Moderate	1.20	0.76–1.90	0.43
Vigorous	0.96	0.53–1.75	0.90
HIV-related and other variables
Detectable viral load (>400 copies/mL)	1.53	1.04–2.25	0.03
CD4 count (per 100 cells/mm3)	0.89	0.84–0.95	<0.001

Abbreviations: AUDIT-C = Alcohol Use Disorders Identification Test; eGFR = estimated glomerular filtration rate; HDL = high-density lipoprotein; LDL = low-density lipoprotein; PHQ-9 = Patient Health Questionnaire-9.

^aModel adjusted for all the variables included in the table and Centers for AIDS Research Network of Integrated Clinical Systems (CNICS) site.

Figure 1. Factors Associated With Stroke in People With HIV From CNICS.

In the overall cohort, cardiovascular risk factors, including treated hypertension and diabetes mellitus, and a detectable viral load were associated with an increased risk of stroke while a higher CD4 count was protective against stroke. In addition to the factors shown in the figure, the multivariable model was also adjusted for CNICS site, systolic blood pressure, diastolic blood pressure, total cholesterol, HDL cholesterol, LDL cholesterol, and physical activity. CNICS = Centers for AIDS Research Network of Integrated Clinical Systems.

Figure 2. Kaplan-Meier Estimate for Survival Free of Stroke by Sex at Birth and Age at Baseline.

Female participants younger than 50 years had higher rates of stroke compared with male participants in the same age range. A similar pattern was not observed among female and male participants older than 50 years, who had more comparable rates of stroke.

Among individuals who had an incident stroke, the race/ethnicity distribution differed by sex (Table 3), reflecting the composition of the cohort. However, we did not observe a statistically significant interaction between sex and race/ethnicity on stroke risk. A diagnosis of treated hypertension was more prevalent among female than male participants who developed a stroke. BMI was higher among female participants compared with male participants who had stroke events, as were total cholesterol and HDL cholesterol. Current cocaine use was more common among female participants than male participants who had strokes, whereas male participants were more likely to have previously used cocaine and methamphetamine. A higher mean score on the PHQ-9 was observed among female participants compared with male participants who had strokes, although the difference did not reach statistical significance. Furthermore, female participants with an incident stroke were more likely to have a detectable viral load than male participants (Table 3).

Table 3.

Comparison of Demographic and Clinical Characteristics by Sex at Birth Among People With HIV With Incident Stroke

Characteristic, mean (SD) or N (%)	Female sex at birth (N = 41)	Male sex at birth (N = 121)	p Value
Demographics
Age, y	47 (12)	52 (9)	0.03
Race and ethnicity			0.03
White	10 (24)	53 (44)
Black	29 (71)	53 (44)
Hispanic	2 (5)	11 (9)
Other	0 (0)	4 (3)
Cardiometabolic risk factors
Treated hypertension	27 (66)	55 (45)	0.02
Systolic blood pressure, mm Hg	133 (23)	130 (19)	0.5
Diastolic blood pressure, mm Hg	81 (13)	79 (12)	0.5
Total cholesterol, mg/dL	189 (47)	171 (41)	0.049
HDL cholesterol, mg/dL	53 (17)	43 (21)	0.02
LDL cholesterol, mg/dL	108 (34)	98 (30)	0.2
Statin use	16 (39)	40 (33)	0.5
Diabetes mellitus	11 (27)	21 (17)	0.2
Body mass index, kg/m2	30 (8)	26 (5)	0.004
eGFR, mL/min/1.73 m2	86 (35)	91 (19)	0.4
eGFR <30 mL/min/1.73 m2	4 (10)	0 (0)	<0.001
ASCVD risk score, %	11.7 (12.1)	12.4 (10.3)	0.8
Mental health and health-related behaviors
PHQ-9 score	7.8 (6.4)	5.6 (6.1)	0.06
Methamphetamine use			0.02
Never	34 (83)	73 (60)
Former	4 (10)	38 (31)
Current	3 (7)	10 (8)
Cocaine use			0.007
Never	23 (56)	48 (40)
Former	12 (29)	67 (55)
Current	6 (15)	6 (5)
Current injection drug use	2 (5)	0 (0)	0.02
Cigarette use			0.3
Never	17 (41)	39 (32)
Former	5 (12)	28 (23)
Current	19 (46)	54 (45)
AUDIT-C score	1.5	1.9	0.3
Physical activity in last month			0.4
None	8 (20)	15 (12)
Moderate	29 (71)	87 (72)
Vigorous	4 (10)	19 (16)
HIV-related and other variables
On antiretroviral therapy	40 (98)	115 (96)	0.6
Detectable viral load (>400 copies/mL)	18 (44)	25 (21)	0.004
Current CD4, cells/mm3	504 (430)	429 (290)	0.3
Hepatitis C infection	7 (17)	37 (31)	0.09

Abbreviations: ASCVD = Atherosclerotic Cardiovascular Disease; AUDIT-C = the Alcohol Use Disorders Identification Test; eGFR = estimated glomerular filtration rate; HDL = high-density lipoprotein; LDL = low-density lipoprotein; PHQ-9 = Patient Health Questionnaire-9.

In multivariable models, age modified the effect of sex at birth on stroke risk (p = 0.001 for the interaction). At age 40, the risk of stroke was 2-fold higher for female participants than for male participants in the cohort (HR 2.01, 95% CI 1.25–3.21, p = 0.004) whereas by age 50, female sex was no longer a significant risk factor of stroke. By age 60, female sex approached being a protective factor against stroke (HR 0.60, 95% CI 0.34–1.06, p = 0.08) (Figure 3). In an exploratory analysis, we compared demographic and clinical characteristics between female and male participants who had an incident stroke stratified by age to identify factors potentially related to stroke risk in younger female participants (eTable 1). Treated hypertension and diabetes were more common in female participants compared with male participants aged 50 years or younger who had an incident stroke, whereas the prevalence of these risk factors was comparable between female and male participants older than 50 years who had a stroke. BMI and ASCVD risk scores were also higher in younger female participants who had an incident stroke compared with younger male participants but similar between older female and male participants. In addition, compared with younger male participants who developed a stroke, younger female participants were less likely to be virologically suppressed while this measure was more comparable between older female and male participants who had a stroke.

Figure 3. Differential Impact of Female Sex at Birth on Stroke Risk by Age.

At younger ages (40 years), female sex compared with male sex at birth was associated with an increased risk of stroke. At older ages (60 years), female sex at birth was associated with a lower risk of stroke.

Sex at birth modified the effect of a detectable viral load and methamphetamine use on stroke risk (Table 4). Having a detectable viral load had a greater adverse effect on stroke risk in female participants (HR 4.66, 95% CI 2.48–8.74, compared with HR 1.30 for male participants, 95% CI 0.83–2.03; p = 0.001 for the interaction), independent of age, stroke risk score, and methamphetamine use. Similarly, methamphetamine use had a greater negative impact on stroke risk in female participants (HR 4.78, 95% CI 1.47–15.56, compared with HR 1.19 for male participants, 95% CI 0.62–2.29; p = 0.001 for the interaction), as did treated hypertension (HR 3.44, 95% CI 1.74–6.81, compared with HR 1.66 for male participants, 95% CI 1.14–2.41; p = 0.06 for the interaction). No significant interaction was found between sex at birth and race/ethnicity, systolic or diastolic blood pressure, total or HDL cholesterol, statin use, diabetes, BMI, renal function, PHQ-9 score, cocaine use, cigarette use, or CD4 count. Results were comparable in sensitivity analyses using inverse probability of censoring weighting.

Table 4.

Interaction Between Sex at Birth and Predictors of Interest on Stroke Risk^a,b

Hazard ratio (95% CI)	Female sex at birth	Male sex at birth	p Value for interaction
Increase in age by 10 y	1.12 (0.82–1.53)	2.05 (1.73–2.44)	0.001
Detectable viral load	4.66 (2.48–8.74)	1.30 (0.83–2.03)	0.001
Methamphetamine use	4.78 (1.47–15.56)	1.19 (0.62–2.29)	0.04
Treated hypertension	3.44 (1.74–6.81)	1.66 (1.14–2.41)	0.06

^aAdjusted for age (centered at 40) and stroke risk score created from race/ethnicity, HIV risk factor, AUDIT-C score, cigarette smoking status, cohort entry year, site, physical activity level, eGFR, cocaine use status, systolic and diastolic blood pressure, total cholesterol, HDL, LDL, and triglycerides.

^bNo significant interaction effect observed between sex at birth and race, blood pressure measurements, statin use, diabetes mellitus, body mass index, total cholesterol, HDL cholesterol, eGFR, current cocaine use, cigarette use, PHQ-9 score, and CD4 count.

Discussion

The impact of several variables, including age, detectable HIV viral load, methamphetamine use, and treated hypertension, on stroke risk differed by sex in the CNICS cohort. Most of these factors had a greater adverse effect on stroke risk in female participants than in male participants in our study, which could partially account for higher rates of stroke observed in women compared with men with HIV.^1,3

We found that the risk of stroke associated with female sex at birth varied by age. Younger female participants with HIV were at higher risk of stroke compared with younger male participants whereas this increased risk of stroke in female participants was not observed above age 50. Viewed another way, the effect of advancing age on stroke risk in PWH diverged by sex. For male participants, stroke risk doubled for every 10 years of age. Conversely, for female participants who contended with the highest risk of stroke at younger ages, older age did not correspond to greater stroke risk. These findings match observations from other HIV cohorts, including the AIDS Clinical Trials Group Longitudinal Linked Randomized Trials cohort and Partners cohort, in which higher stroke incidence among women compared with men was most pronounced at younger ages.^1-3

Because stroke is uncommon at younger ages, even a modest absolute difference in stroke incidence could appear as a relatively large discrepancy between women and men with HIV, especially in the absence of an HIV-uninfected control group with which to compare stroke rates. This may be relevant because elevated stroke risk in young women compared with young men may not be unique to PWH. Data from large population-based studies have indicated that, in recent decades, young women may be experiencing higher rates of stroke than young men.^19,20 However, in the Partners cohort, which included individuals with and without HIV, not only did young women with HIV experience higher absolute rates of stroke than young men with HIV, but the relative risk was higher in young women with HIV compared with young women without HIV and exceeded the same risk comparison by HIV status for men.^1,2 This suggests that excess risk of stroke for younger female participants with HIV observed in our study is unlikely to be fully explained by temporal trends in the general population.

Similar to other US cohorts of PWH, the racial composition of CNICS differs by sex, with a larger proportion of female participants who are Black vs male participants who are White, reflecting the disproportionate impact of the HIV epidemic on certain US populations. This is a relevant consideration because stroke incidence rates and stroke-related mortality in the general population differ by race/ethnicity, with Black individuals at higher risk compared with White individuals, especially at younger ages.²¹ The observed sex-by-age interaction on stroke risk in our study remained significant after adjusting for race and ethnicity, implying that, at younger ages, female participants were at higher risk of stroke independent of race. Furthermore, we did not find a significant interaction between sex and race on stroke risk. Given the relatively modest number of incident strokes in the cohort, however, we were not able to evaluate for a 3-way interaction between age, sex, and race or to stratify the analyses by race (i.e., comparing stroke rates in Black female participants with Black male participants and in White female participants with White male participants).

Unlike the decline in stroke incidence across older age groups in the general population,^19,22 rates of stroke in the young have been on the rise²³ and potentially more so for young women than men.²⁴ An uptick in the prevalence of cardiometabolic risk factors in the young may partially account for the growing burden of stroke in this population.^25,26 Our data suggest that the same may be true for younger female participants with HIV who had a stroke as they tended to have a higher BMI and worse CVD risk profile than younger male participants who had a stroke. Hypertension, among other cardiometabolic risk factors, was more prevalent among younger female participants who developed strokes than younger male participants whereas the frequency of CVD risk factors was comparable between older female and male participants who developed strokes. Beginning as early as the third decade, women have a steeper rise in blood pressure than men that persists throughout the life course,²⁷ which could lead to variation in the timing and presentation of cerebrovascular disease in women vs men. Among individuals who had a stroke in our study, ASCVD risk was also higher for younger female participants compared with younger male participants but similar when comparing older participants by sex. The higher ASCVD risk score among younger female participants is especially remarkable as the formula, by design, assigns higher risk to the male sex.

The risk of stroke associated with hypertension for female participants in the cohort was more than twice that for male participants. This is in line with data from the general population and PWH that point to a larger magnitude of the effect of several cardiometabolic risk factors, including hypertension, on stroke risk in women compared with men.^1,10,11,28 By contrast, similar to other studies,^10,29 at any given systolic or diastolic blood pressure, no difference in stroke risk was noted by sex. These divergent findings suggest that disparities in the management and control of hypertension, which may be less optimal for non-White women,³⁰ may be contributing to the greater negative effect of hypertension on stroke risk in women, including women with HIV.

While rates of cigarette, cocaine, and methamphetamine use did not differ when comparing female participants and male participants who had strokes within age strata, methamphetamine use had a greater adverse effect on stroke risk in female participants compared with male participants in our study. In general, more methamphetamine-associated stroke events, which are typically small vessel disease stroke events,³¹ occur in men,³² consistent with higher rates of methamphetamine use among men. As women are increasingly recognized to be at higher risk of microvascular dysfunction,³³ women may be uniquely vulnerable to the impact of methamphetamine use on small vessel pathophysiology, leading to greater cerebrovascular injury and stroke risk.

Higher rates of comorbid mental health conditions in younger women with HIV, including depression and stress, which are strongly associated with CVD and stroke³⁴ and have been linked to carotid arterial inflammation in PWH,³⁵ could underlie excess stroke risk observed in this population. Elevated stroke risk among younger female participants in our study was independent of depression and antidepressant use. Younger female participants who developed a stroke were more likely to have a detectable viral load and lower CD4 cell count compared with younger male participants, whereas a similar difference by sex was not observed among older participants who developed a stroke. In light of data pointing to immune reconstitution-related inflammation as a potential contributor to stroke risk in PWH,³⁶ we investigated recent (within 6 months) ART initiation at baseline, which was not associated with stroke risk. However, we did not analyze recent ART initiation as a time-dependent variable, which may have been more informative to understand the relationship between immune reconstitution and stroke risk.

Among female participants but not male participants, having a detectable viral load was associated with more than 4-fold higher risk of stroke. Viremia has been consistently identified as a predominant risk factor of stroke in PWH, with a magnitude of effect equivalent to major risk factors such as hypertension and advancing age.^1,3 Early in the course of HIV infection, women have lower viral loads but similar disease progression as men.³⁷ This may be attributed, in part, to a more robust immunologic response to HIV observed in women at a given level of viremia.³⁸ While, during early infection, this vigorous immune response may lead to better virologic control in women, the long-term consequences of enhanced immune activation may offset the benefit, resulting in elevated risk of inflammation-related comorbid conditions, including stroke.^7,39

Our findings must be interpreted in the context of various limitations. First, we did not have data on atrial fibrillation, a major risk factor of stroke that may affect women more than men.⁴⁰ Second, we did not have information on social determinants of health, central to which is poverty, which may outstrip even the strongest cardiometabolic risk factors in terms of stroke risk.⁴¹ Third, we were unable to test for a 3-way age-by-sex-by-race interaction or to stratify analyses by race. Finally, we did not have data on sex-specific and gender-specific factors (e.g., menopause, oral contraception, and gender-affirming medications) that could influence differences in stroke risk between women and men. Strengths include use of a large, comprehensive, and diverse study population; stroke adjudication by neurologists who review the primary clinical data, thereby avoiding pitfalls of relying on administrative codes to define stroke; and access to patient-reported outcomes on substance use, depression, and other covariates relevant to stroke.

We found that, among younger PWH, individuals assigned female sex at birth were at elevated risk of stroke compared with individuals assigned male sex at birth. The profile of a female participant in CNICS who developed a stroke was an individual who, despite younger age, had a remarkably high CVD risk profile that exceeded that of younger male participants and equaled or surpassed CVD risk observed in older PWH. However, CVD risk may not tell the whole story as other non-traditional risk factors (e.g., HIV viremia and methamphetamine use) had a greater negative impact on stroke risk in female participants than male participants. In line with a recent executive order to prioritize women's health research across the lifespan,⁴² investigation into mechanisms underlying sex differences in stroke risk will be critical to understanding how these findings might translate into sex-specific stroke prevention strategies for PWH and the general population.

Glossary

ART

antiretroviral therapy

ASCVD

Atherosclerotic Cardiovascular Disease

CNICS

Centers for AIDS Research Network of Integrated Clinical Systems

CVD

cardiovascular disease

eGFR

estimated glomerular filtration rate

HDL

high-density lipoprotein

HR

hazard ratio

PWH

people with HIV

Appendix. Authors

Name	Location	Contribution
Felicia C. Chow, MD, MAS	Departments of Neurology and Medicine (Infectious Diseases), University of California, San Francisco	Drafting/revision of the manuscript for content, including medical writing for content; major role in the acquisition of data; study concept or design; analysis or interpretation of data
Robin M. Nance, PhD	Department of Medicine, University of Washington, Seattle	Drafting/revision of the manuscript for content, including medical writing for content; analysis or interpretation of data
Kyra Becker, MD	Department of Neurology, University of Washington, Seattle	Drafting/revision of the manuscript for content, including medical writing for content; major role in the acquisition of data
Emily L. Ho, MD, PhD	Department of Neurology, University of Washington, Seattle; Swedish Neuroscience Institute, Swedish Medical Center, Seattle, Washington	Drafting/revision of the manuscript for content, including medical writing for content; major role in the acquisition of data
Andrew Huffer, MD	Department of Neurology, University of Washington, Seattle	Drafting/revision of the manuscript for content, including medical writing for content; major role in the acquisition of data
Rizwan Kalani, MD	Department of Neurology, University of Washington, Seattle	Drafting/revision of the manuscript for content, including medical writing for content; major role in the acquisition of data
Christina M. Marra, MD	Department of Neurology, University of Washington, Seattle	Drafting/revision of the manuscript for content, including medical writing for content; major role in the acquisition of data
Joseph R. Zunt, MD, MPH	Departments of Medicine, Neurology, Epidemiology, and Global Health, University of Washington, Seattle	Drafting/revision of the manuscript for content, including medical writing for content; major role in the acquisition of data
Laura Bamford, MD	Department of Medicine, University of California, San Diego	Drafting/revision of the manuscript for content, including medical writing for content
Greer A. Burkholder, MD	Department of Medicine, University of Alabama, Birmingham	Drafting/revision of the manuscript for content, including medical writing for content
Edward Cachay, MD	Department of Medicine, University of California, San Diego	Drafting/revision of the manuscript for content, including medical writing for content
Joseph J. Eron, MD	Department of Medicine, University of North Carolina, Chapel Hill	Drafting/revision of the manuscript for content, including medical writing for content
Jeanne Keruly, MS	Department of Medicine, Johns Hopkins University	Drafting/revision of the manuscript for content, including medical writing for content
Mari M. Kitahata, MD, MPH	Department of Medicine, University of Washington, Seattle	Drafting/revision of the manuscript for content, including medical writing for content
Sonia Napravnik, PhD	Department of Medicine, University of North Carolina, Chapel Hill	Drafting/revision of the manuscript for content, including medical writing for content; analysis or interpretation of data
Michael S. Saag, MD	Department of Medicine, University of Alabama, Birmingham	Drafting/revision of the manuscript for content, including medical writing for content
Amanda L. Willig, PhD, RD	Department of Medicine, University of Alabama, Birmingham	Drafting/revision of the manuscript for content, including medical writing for content
Richard D. Moore, MD	Department of Medicine, Johns Hopkins University	Drafting/revision of the manuscript for content, including medical writing for content
David L. Tirschwell, MD, MSc	Department of Neurology, University of Washington, Seattle	Drafting/revision of the manuscript for content, including medical writing for content; major role in the acquisition of data; study concept or design; analysis or interpretation of data
Joseph A. Delaney, PhD	Department of Epidemiology, University of Washington, Seattle; University of Manitoba, Winnipeg, Canada	Drafting/revision of the manuscript for content, including medical writing for content; study concept or design; analysis or interpretation of data
Heidi M. Crane, MD, MPH	Department of Medicine, University of Washington, Seattle	Drafting/revision of the manuscript for content, including medical writing for content; major role in the acquisition of data; study concept or design; analysis or interpretation of data

Study Funding

This work was supported by the NIH's National Heart, Lung, and Blood Institute (R01 HL126538) and National Institute of Neurologic Disorders and Stroke (R01 NS126086), as well as the American Heart Association (13GRNT14560022).

Disclosure

The authors report no relevant disclosures. Go to Neurology.org/N for full disclosures.