Mechanisms of ischemic stroke in HIV-infected patients

The presence of stroke, either cerebral infarction or intracranial hemorrhage, has been reported in several autopsy series of HIV-infected brains, with prevalence rates ranging between 6% and 34%.^1–5 However, most of these cerebrovascular pathologic findings were clinically asymptomatic and stroke did not appear to be diagnosed prior to time of death.^3–5 The prevalence of clinically diagnosed stroke syndrome has been reported to be between 0.5% and 3.75% in more than 2,000 patients with AIDS or AIDS related complex (ARC) from six clinical series.¹ Meanwhile, the adjusted relative risks of stroke for patients with AIDS in a recent population-based study were 9.1 for ischemic infarction and 12.7 for intracranial hemorrhage.⁶

Causes uncommon in the general population of patients with stroke have been noted to play a more frequent role in immunodepressed patients with HIV, such as vasculitis⁷ and perhaps hypercoagulability.^8–10 Studies investigating stroke pathophysiology in HIV-infected patients have been small and predate the development of more effective treatment regimens against HIV.

With the introduction of combination antiretroviral therapy (CART) a steep reduction in mortality and morbidity has been observed in HIV-infected subjects in the United States and other Western countries. The development and inclusion of protease inhibitors in this antiretroviral regimen has been highly responsible for its success. However, treatment with CART containing protease inhibitors has been associated with severe premature atherosclerotic vascular disease, including premature myocardial infarctions and stroke.^11,12

The goal of our study was to ascertain the distribution of stroke mechanisms in a relatively large population of HIV-infected patients and to assess which variables contribute to the determination of the stroke mechanism, including exposure to CART.

Methods.

We reviewed a comprehensive database of admission and discharge codes of patients admitted to the Special Immunology, Internal Medicine, and Neurology Services of Jackson Memorial Hospital (a large community-based teaching hospital which serves the 2,376,000 residents of Miami-Dade County, FL [US Census Bureaus, quickfacts.census.gov; accessed on September 7, 2006]), with diagnosis of HIV infection and stroke from 1996 to 2005. Reporting the diagnosis of HIV is required in Miami-Dade County since 1997. Between 1997 and 2002, 8,616 newly diagnosed HIV infections were reported. In 2003 there were just over 12,000 people living with AIDS in Miami, which has the fifth largest AIDS population of any city in the United States (HIV/AIDS Epidemiologic Profile Miami-Dade County 2004, www.aidnset.org; accessed on September 7, 2006). In 2002, the highest concentration of AIDS cases (>600/100,000) were reported in the address codes served by Jackson Memorial Hospital (Miami-Dade County Health Department, Vital and Morbidity Statistics 2002).

We retrieved the information on patients whose admission or discharge diagnosis included the International Classification of Diseases (ICD)-9 codes of cerebrovascular disease (430–438) along with HIV infection (042–044).

Pertinent clinical and radiologic data were then reviewed to confirm the diagnosis of stroke. Acute stroke was defined as new focal neurologic deficits with documentation of corresponding acute ischemic infarction or hemorrhage on brain imaging. Confirmed diagnosis of HIV infection, either prior to the admission or made during hospitalization, was required. Patients with exacerbation of previous neurologic deficits, postictal deficits, and nonvascular organic brain lesions, such as tumors, parasitic cysts, or progressive multifocal leukoencephalopathy (PML), were excluded.

We collected demographic information, date of stroke, time of diagnosis of HIV infection, CD4 count, viral load, and antiretroviral treatment (current, previous, or naïve). We also looked for history of hypertension, diabetes mellitus, cigarette smoking, alcohol and drug abuse, renal disease, coronary artery disease, atrial fibrillation, or cardiomyopathy. We collected the results of diagnostic studies, including carotid ultrasound, transthoracic echocardiogram, transesophageal echocardiogram, EKG, Holter monitor, transcranial Doppler ultrasound (TCD), CT of the head, MRI and magnetic resonance angiography (MRA) of the brain, and conventional cerebral angiography. We searched for documentation of CNS infection or tumor, pathology reports, lumbar puncture, or toxicology screen and tabulated the measurements of serum lipids, homocysteine, glycohemoglobin A1C, and hypercoagulability markers.

Mechanisms of stroke were classified according to TOAST criteria¹³ into the following groups: large artery atherosclerosis, cardioembolism, small vessel occlusion, stroke of other determined etiology, and stroke of undetermined etiology, the latter including two or more causes identified, negative evaluation, or incomplete evaluation (in patients with stroke of undetermined cause, we required CSF analysis, testing for hypercoagulability, and an angiographic study for the evaluation to be deemed complete). When the mechanism was not readily apparent, we defined it by consensus among two or more of the investigators. Cases of large artery atherosclerosis and small vessel occlusion were considered atherothrombotic strokes; and all other strokes were considered non-atherothrombotic.

Diagnosis of vasculitis required pathologic or angiographic evidence and microbiologic confirmation of the infectious agent for cases of infectious vasculitis. Ischemic infarctions were ascribed to hypercoagulability only when there was proof of a hypercoagulable condition in the absence of other potential causes for the infarction despite a comprehensive evaluation.

For the statistical analysis, we used the χ² and Fisher exact test for categorical variables and the Student t-test to compare continuous variables. Analysis of variance was used to compare means between multiple groups and logistic regression analysis was used to analyze covariates predictors of outcome.

Results.

We identified 82 HIV-infected patients with acute stroke, 77 with ischemic infarctions (94%), and 5 with intraparenchymal hemorrhages (5%). Mean age was 42 years (range 3 to 72), 49 patients (59%) were male, and 73 (89%) were African American. Seventy-five patients had a previous diagnosis of HIV infection (92%) and 7 patients were diagnosed during the acute stroke evaluation (8%). Mean CD4 count at the time of stroke diagnosis was 113 cells/mm³ (range 2 to 659); 66 patients (84%) had CD4 count <200 cells/mm³ and 43 (54%) had CD4 count <100 cells/mm³. Antiretroviral therapy had been used by 48 patients (58%), including 30 patients (37%) who were on CART at the time of the stroke.

Tobacco use was the most prevalent traditional vascular risk factor (51%), followed by hypertension (42%). History of hyperlipidemia was uncommon and results of lipid panels at the time of the stroke (tested in 49 patients) were most often normal: mean total cholesterol 169 mg/dL (range 66 to 305), high-density lipoprotein 37 mg/dL (range 12 to 73), low-density lipoprotein 98 mg/dL (range 32 to 222), and triglycerides 147 mg/dL (range 58 to 575). Homocysteine levels were tested in 17 patients, with a mean of 8.6 mg/dL (range 5.35 to 14). History of diabetes mellitus was present in only 6 patients (7%). Only one patient had diagnosis of atrial fibrillation. Brain MRI was performed in 57% of patients, carotid ultrasound in 57%, echocardiogram in 84%, and conventional cerebral angiography in 12%. Thirty-two patients (39%) underwent lumbar puncture and CSF analysis was diagnostic in 15% of those cases.

Among 77 patients with ischemic stroke, 10 had large artery atherosclerosis (13%), including 4 with extracranial internal carotid artery stenosis (5%), 5 with middle cerebral artery stenosis (6.5%), and 1 had vertebrobasilar stenosis (1.3%). In 15 patients, the mechanism was defined as cardioembolism (19%), due to the presence of cardiomyopathy with reduced left ventricular ejection fraction (<30%) or evidence of intracardiac thrombus, and no coexistent atherosclerotic disease. Fifteen patients had a small vessel occlusion (19%) confirmed with imaging studies. Eighteen patients were classified as other determined etiology (23%), including 10 cases of vasculitis (13%), 1 vertebral artery dissection (1.3%), and 7 cases of hypercoagulability (9%). There were 19 cases of stroke of undetermined etiology, including 3 with two or more causes identified, 3 with negative evaluation, and 13 with incomplete evaluation (17%).

Anticardiolipin antibodies were investigated in 31 patients (38%) and found in 29% of these cases. Protein S level was evaluated in 22 patients (26%) and deficiency was documented in 45% of these individuals. Among the 7 patients with ischemic stroke ascribed to hypercoagulability, 6 had anticardiolipin antibodies in the serum and 4 had protein S deficiency. The laboratory findings and associated clinical conditions of these patients are shown in table 1.

Diagnosis of vasculitis was interpreted as confirmed in six cases, probable in three, and possible in one (table 2). When we compared patients with vasculitis with the rest of the population, there were no differences in age (39 ± 13. 9 vs 42.1 ± 12.7; p = 0.2) or CD4 count (113 ± 81.7 vs 104.6 ± 133.8; p = 0.6), but vasculitis was predictive of poor outcome (OR 3.07, 95% CI 0.78 to 12.06).

Patients with atherothrombotic strokes were older than patients with non-atherothrombotic strokes (45.3 ± 3 vs 40.1 ± 1.6; p = 0.04). Seventeen patients with atherothrombotic strokes had been exposed to antiretroviral therapy (68%), including 11 (44%) who were receiving CART at the time of the stroke. Meanwhile, 28 patients (54%) with non-atherothrombotic strokes had been exposed to antiretroviral agents, and 18 (35%) were currently using CART when the stroke occurred. Neither antiretroviral treatment at any time nor current use of CART correlated with the type of ischemic stroke. There were no also differences in CD4 counts at the time of the stroke between the atherothrombotic and non-atherothrombotic groups (mean 81.7 ± 105.3 vs 117.8 ± 137.8; p = 0.06).

History of cocaine use was present in 53 patients; 10 of these patients (19%) were in the atherothrombosis group and 20 patients (37%) in the non-atherothrombosis group (p = 0.66). However, when toxicology testing was performed (in 31 patients with ischemic stroke), cocaine was not detected in any patient with atherothrombotic stroke (0/6 patients), but was detected in 13/25 tested patients (52%) with non-atherothrombotic stroke (p = 0.02). Table 3 summarizes the comparison of different variables between patients with atherothrombotic and non-atherothrombotic strokes.

Regarding the five cases of hemorrhagic stroke, four hematomas were located in basal ganglia and one was cerebellar. Two cases were secondary to hypertension, two cases related to cocaine abuse, and in one case the etiology remained undetermined. These patients were not taken into account for the statistical analysis of mechanisms of stroke, but were included in the outcome analysis.

Ten patients died (12.5%) and strokes were severely disabling in 18 (22.5%) cases. Patients with unfavorable outcome (death or severe disability) had a lower CD4 count (56.8 ± 59.7 vs 139.6 ± 153.5; p = 0.004). All patients who died had a CD4 count <100 cells/mm³ (mean 22.6 ± 20.3).

Discussion.

In this series of consecutive HIV-infected patients with acute stroke we found that the mechanisms of ischemia were variable. Most of our patients were young and severely immunocompromised. Traditional risk factors for vascular disease had relatively low prevalence. Conversely, mechanisms which are unusual in the general stroke population, such as vasculitis and hypercoagulability, played a greater role in our cohort of HIV-infected patients.

Cardiac disease is common in HIV-infected patients, especially in later stages of the infection,¹⁴ and dilated cardiomyopathy was deemed the responsible mechanism for the stroke in almost 20% of our patients. Hemostatic abnormalities have been previously documented in HIV-infected patients with ischemic stroke, in particular protein S deficiency^8,10 and antiphospholipid antibodies.^8,15 However, the pathogenic role of protein S deficiency in HIV-infected patients with ischemic stroke has been questioned.¹⁶ Using very restrictive criteria to attribute stroke to hypercoagulability, we found that protein S deficiency was frequently encountered (45%) but rarely the only or most likely cause of cerebral infarction. Our experience confirms that infectious vasculitis is relatively common in HIV-infected patients with stroke and advanced immunosuppression and argues for performing lumbar puncture in these cases.

The relative high frequency of cryptogenic strokes observed in our series has also been noted by others.^10,17,18 Following our strict definition to consider a diagnostic evaluation complete in patients with stroke of undetermined cause, we found an elevated rate of cases with incomplete evaluation. In light of the increased frequency of atypical stroke mechanisms in HIV-infected patients, comprehensive diagnostic evaluations (possibly including CSF analysis, testing for hypercoagulable conditions, and angiographic studies) are justified.

We found an association between current cocaine use and non-atherothrombotic mechanisms. The clinical implications of this association are difficult to establish since the group of non-atherothrombotic strokes included various mechanisms and also cryptogenic cases. Vasoconstriction, increased platelet aggregation, and vasculitis have been postulated as possible mechanisms of brain ischemia in cocaine users.¹⁹

Use of CART, particularly containing protease inhibitors, has been associated with the development of metabolic syndrome, accelerated atherosclerosis, and increased risk of vascular events.^11,12,20,21 We did not find any correlation between antiretroviral therapy and stroke mechanism. However, this study does not exclude the possibility that patients with better responses to treatment and lesser degrees of immunosuppression might have increased risk of atherosclerosis-related strokes after long exposure to CART.

Our study has several limitations. We used diagnostic codes to identify our study subjects and cannot exclude the possibility of having missed some HIV-infected patients with stroke in our screening process; this could potentially represent a bias against milder strokes. The lack of a comparison group of HIV-infected patients without strokes limits our ability to test if risk factors (such as hypercoagulable conditions) were more prevalent in the patients with stroke. Differences in the comprehensiveness of the diagnostic evaluations need to be considered when interpreting our results.

The high representation of severely immunodepressed patients and lack of information on the duration of exposure to CART and compliance with this therapy limited our ability to test the hypothesis that CART could increase the risk of atherothrombotic strokes. Our strict criteria to ascribe infarctions to vasculitis or hypercoagulability may have underestimated the impact of these mechanisms. In addition, we did not investigate the interesting hypothesis that vascular changes provoked by the HIV infection itself might predispose to cerebral infarctions. This possibility is supported by epidemiologic data⁶ and clinical series,¹⁷ changes in vasomotor reactivity observed in patients with HIV infection without cerebrovascular symptoms,²² the independent association of HIV infection with intima-media thickness of the carotid artery,²³ and the reports of cerebral vasculopathy with aneurysm formation in HIV-infected children²⁴ and young adults.²⁵ A form of HIV vasculopathy may have accounted for some of the strokes that remained unexplained despite extensive evaluation in our series.