HIV Infection and Incidence of Ischemic Stroke
HIV Infection and Incidence of Ischemic Stroke
We conducted a cohort study of HIV-positive and HIV-negative members of Kaiser Permanente Northern and Southern California (KPNC and KPSC, respectively), large integrated healthcare systems providing comprehensive medical services to over 30% of insured Californians. The study population was selected from a previously described cohort. HIV-positive individuals were identified using HIV registries that include all known HIV/AIDS cases since the early 1980s for KPNC and 2000 for KPSC, with HIV-positive status confirmed by review of medical charts or medical center case lists. Individuals not included in the registries were considered HIV-negative. Eligible individuals included adult (≥18 years of age) Kaiser Permanente members during 1996–2011 for KPNC and 2000–2011 for KPSC. HIV-negative members were frequency-matched 10 : 1 to HIV-positive individuals by age (5-year age groups), sex, medical center, and initial calendar year of follow-up. The start of follow-up for each individual (baseline) was the earliest date on or after January 1, 1996 (January 1, 2000, for KPSC), when eligibility criteria were met. Individuals were followed until the earliest of an ischemic stroke, death, termination of membership, or 31 December 2011. The institutional review boards at KPNC and KPSC approved this study with a waiver of written informed consent.
HIV registries were used to obtain HIV status, race/ethnicity, and HIV-transmission risk factor. For KPNC only, additional data were extracted by chart review on dates of first ART use and beginning of known HIV infection.
Data extracted from Kaiser Permanente's electronic medical record (EMR) included laboratory test results (CD4 cell counts and HIV RNA levels); pharmacy prescription fills (ART, lipid-lowering therapy, and antihypertension medications); age; health plan enrollment periods; dates of death from hospital records, California death certificates, and Social Security Administration datasets; and inpatient or outpatient clinical diagnoses, including overweight/obesity [International Classification of Disease codes, version 9 (ICD-9): 278, 259.9, V85, internal weight/height codes], smoking/tobacco use (ICD-9: 305.1, V15, V65, 649, internal social history codes), drug/alcohol abuse (ICD-9: 291, 292, 303–305.0, 305.2–305.5), diabetes (ICD-9: 250.xx, 357.2, 362.0, 366.41, 648.0, laboratory measurements, or medication use for KPSC and an internal Diabetes Registry for KPNC), and hypertension (ICD-9: 401–405 or antihypertension medication use).
The outcome of interest was incident ischemic stroke, defined as a primary diagnosis in the EMR of an acute ischemic cerebrovascular event (ICD-9: 433.x1; 434, excluding 434.x0; or 436). This selection of codes has been found to have high specificity and positive predictive value for ischemic stroke. Cases were limited to ischemic stroke because of the known atherosclerotic risks of ART use that can lead to ischemic events, and because, in contrast to hemorrhagic stroke, its incidence among HIV-infected individuals has reportedly increased over time. Individuals were excluded if they had an ischemic stroke diagnosis within 2 years prior to baseline.
Variables for analyses included age (<40, 40–49, 50–64, ≥65 years), sex, race/ethnicity (white, black/African American, Hispanic/Latino, Asian/Pacific Islander, other; with imputed data where missing), calendar era (1996–1999, 2000–2003, 2004–2007, 2008–2009, 2010–2011), and socioeconomic status (SES, calculated as neighborhood SES index based on census measures of income, occupation, and education). Other covariates were smoking, alcohol/drug abuse, overweight/obesity, diabetes, hypertension, and lipid-lowering therapy use, all defined as ever/never from 2 years prior to baseline through the end of follow-up. Among HIV-positive individuals, we evaluated any prior ART use, recent (i.e., prior 6 months) and duration of protease inhibitor use, recent and duration of nonnucleoside reverse transcriptase inhibitor (NNRTI) use, duration of known HIV infection (≥10, 5–9.9, <5 years), HIV-transmission risk factor (MSM, heterosexual sex, injection drug use, unknown), recent (i.e. prior 6 months) and nadir (i.e., lowest recorded in Kaiser Permanente) CD4 cell counts (<200, 200–499, ≥500 cells/μl), and recent HIV RNA levels (≥10 000, 500–9999, <500 copies/ml). Time-dependent variables were age, calendar era, diabetes, hypertension, lipid-lowering therapy use, ART use (prior and duration), duration of known HIV infection, and recent CD4 cell counts and HIV RNA levels; these were updated continuously except for CD4 and HIV RNA variables, which were updated at 6-month intervals.
We compared individual characteristics by HIV status at baseline using the Pearson χ test for categorical variables and the Kruskal–Wallis test for continuous variables. We computed ischemic stroke incidence rates per 100 000 person-years by HIV status. We used Poisson regression models to estimate the percentage change in rates over calendar time by HIV status and stratified by age group. Using HIV-negative individuals as the reference group, rate ratios for stroke were obtained from Poisson models for HIV infection overall and stratified by sex and immunodeficiency characteristics. We used likelihood-ratio tests to assess effect modification across sex strata, and χ tests for trend across CD4 and HIV RNA strata. Adjusted models included terms for HIV infection, age, sex, calendar era, race/ethnicity, SES, smoking, alcohol/drug abuse, overweight/obesity, diabetes, hypertension, and lipid-lowering therapy.
To identify ischemic stroke risk factors among HIV-positive individuals, we obtained adjusted rate ratios for recent and nadir CD4 cell count, recent HIV RNA level, prior ART use, HIV-transmission risk factor, and duration of known HIV infection, in addition to age, sex, calendar era, race/ethnicity, SES, smoking, alcohol/drug abuse, overweight/obesity, diabetes, hypertension, and lipid-lowering therapy. In the subset with a complete history of protease inhibitor and NNRTI use, we examined the association of cumulative protease inhibitor and NNRTI use in models adjusting for baseline CD4, baseline HIV RNA, and the covariates described above, with additional models that also included terms for recent protease inhibitor and NNRTI use.
Opportunistic infections with neurologic involvement can mimic the symptoms of ischemic stroke; because these infections mainly occur at low CD4 cell counts, it has been suggested that their misclassification may induce a noncausal association between immunodeficiency and ischemic stroke. To assess potential misclassification, an HIV physician (coauthor D.B.K.) conducted a chart review of HIV-positive cases diagnosed within KPNC to identify events caused by opportunistic infections or other central nervous system processes that may mimic strokes. A random sample of 15 cases was selected for review from each of three recent CD4 cell count strata (<200, 200–499, and ≥500 cells/μl) to determine if any observed misclassification was more frequent at lower CD4 cell counts. Overall, 44/45 (98%) were identified as true ischemic strokes, with the one noncase from a patient with CD4 cell count less than 200 cells/μl. Based on this review, we concluded there was minimal outcome misclassification, and that differential misclassification by CD4 cell count was also minimal. To further exclude the possibility of misclassification, we conducted a sensitivity analysis in which we fit adjusted models comparing stroke incidence by HIV status, with HIV-positive individuals stratified by recent CD4 cell count and excluding cases occurring within three months of an opportunistic infection with neurologic involvement (i.e. toxoplasmosis, cytomegalovirus, cryptococcosis, central nervous system lymphoma, or progressive multifocal leukoencephalopathy).
Analyses were conducted in SAS 9.1 (Cary, North Carolina, USA). Statistical tests were two-sided, and statistical significance was defined as P < 0.05.
Methods
Study Design, Setting, and Population
We conducted a cohort study of HIV-positive and HIV-negative members of Kaiser Permanente Northern and Southern California (KPNC and KPSC, respectively), large integrated healthcare systems providing comprehensive medical services to over 30% of insured Californians. The study population was selected from a previously described cohort. HIV-positive individuals were identified using HIV registries that include all known HIV/AIDS cases since the early 1980s for KPNC and 2000 for KPSC, with HIV-positive status confirmed by review of medical charts or medical center case lists. Individuals not included in the registries were considered HIV-negative. Eligible individuals included adult (≥18 years of age) Kaiser Permanente members during 1996–2011 for KPNC and 2000–2011 for KPSC. HIV-negative members were frequency-matched 10 : 1 to HIV-positive individuals by age (5-year age groups), sex, medical center, and initial calendar year of follow-up. The start of follow-up for each individual (baseline) was the earliest date on or after January 1, 1996 (January 1, 2000, for KPSC), when eligibility criteria were met. Individuals were followed until the earliest of an ischemic stroke, death, termination of membership, or 31 December 2011. The institutional review boards at KPNC and KPSC approved this study with a waiver of written informed consent.
Study Measurements
HIV registries were used to obtain HIV status, race/ethnicity, and HIV-transmission risk factor. For KPNC only, additional data were extracted by chart review on dates of first ART use and beginning of known HIV infection.
Data extracted from Kaiser Permanente's electronic medical record (EMR) included laboratory test results (CD4 cell counts and HIV RNA levels); pharmacy prescription fills (ART, lipid-lowering therapy, and antihypertension medications); age; health plan enrollment periods; dates of death from hospital records, California death certificates, and Social Security Administration datasets; and inpatient or outpatient clinical diagnoses, including overweight/obesity [International Classification of Disease codes, version 9 (ICD-9): 278, 259.9, V85, internal weight/height codes], smoking/tobacco use (ICD-9: 305.1, V15, V65, 649, internal social history codes), drug/alcohol abuse (ICD-9: 291, 292, 303–305.0, 305.2–305.5), diabetes (ICD-9: 250.xx, 357.2, 362.0, 366.41, 648.0, laboratory measurements, or medication use for KPSC and an internal Diabetes Registry for KPNC), and hypertension (ICD-9: 401–405 or antihypertension medication use).
The outcome of interest was incident ischemic stroke, defined as a primary diagnosis in the EMR of an acute ischemic cerebrovascular event (ICD-9: 433.x1; 434, excluding 434.x0; or 436). This selection of codes has been found to have high specificity and positive predictive value for ischemic stroke. Cases were limited to ischemic stroke because of the known atherosclerotic risks of ART use that can lead to ischemic events, and because, in contrast to hemorrhagic stroke, its incidence among HIV-infected individuals has reportedly increased over time. Individuals were excluded if they had an ischemic stroke diagnosis within 2 years prior to baseline.
Statistical Analysis
Variables for analyses included age (<40, 40–49, 50–64, ≥65 years), sex, race/ethnicity (white, black/African American, Hispanic/Latino, Asian/Pacific Islander, other; with imputed data where missing), calendar era (1996–1999, 2000–2003, 2004–2007, 2008–2009, 2010–2011), and socioeconomic status (SES, calculated as neighborhood SES index based on census measures of income, occupation, and education). Other covariates were smoking, alcohol/drug abuse, overweight/obesity, diabetes, hypertension, and lipid-lowering therapy use, all defined as ever/never from 2 years prior to baseline through the end of follow-up. Among HIV-positive individuals, we evaluated any prior ART use, recent (i.e., prior 6 months) and duration of protease inhibitor use, recent and duration of nonnucleoside reverse transcriptase inhibitor (NNRTI) use, duration of known HIV infection (≥10, 5–9.9, <5 years), HIV-transmission risk factor (MSM, heterosexual sex, injection drug use, unknown), recent (i.e. prior 6 months) and nadir (i.e., lowest recorded in Kaiser Permanente) CD4 cell counts (<200, 200–499, ≥500 cells/μl), and recent HIV RNA levels (≥10 000, 500–9999, <500 copies/ml). Time-dependent variables were age, calendar era, diabetes, hypertension, lipid-lowering therapy use, ART use (prior and duration), duration of known HIV infection, and recent CD4 cell counts and HIV RNA levels; these were updated continuously except for CD4 and HIV RNA variables, which were updated at 6-month intervals.
We compared individual characteristics by HIV status at baseline using the Pearson χ test for categorical variables and the Kruskal–Wallis test for continuous variables. We computed ischemic stroke incidence rates per 100 000 person-years by HIV status. We used Poisson regression models to estimate the percentage change in rates over calendar time by HIV status and stratified by age group. Using HIV-negative individuals as the reference group, rate ratios for stroke were obtained from Poisson models for HIV infection overall and stratified by sex and immunodeficiency characteristics. We used likelihood-ratio tests to assess effect modification across sex strata, and χ tests for trend across CD4 and HIV RNA strata. Adjusted models included terms for HIV infection, age, sex, calendar era, race/ethnicity, SES, smoking, alcohol/drug abuse, overweight/obesity, diabetes, hypertension, and lipid-lowering therapy.
To identify ischemic stroke risk factors among HIV-positive individuals, we obtained adjusted rate ratios for recent and nadir CD4 cell count, recent HIV RNA level, prior ART use, HIV-transmission risk factor, and duration of known HIV infection, in addition to age, sex, calendar era, race/ethnicity, SES, smoking, alcohol/drug abuse, overweight/obesity, diabetes, hypertension, and lipid-lowering therapy. In the subset with a complete history of protease inhibitor and NNRTI use, we examined the association of cumulative protease inhibitor and NNRTI use in models adjusting for baseline CD4, baseline HIV RNA, and the covariates described above, with additional models that also included terms for recent protease inhibitor and NNRTI use.
Opportunistic infections with neurologic involvement can mimic the symptoms of ischemic stroke; because these infections mainly occur at low CD4 cell counts, it has been suggested that their misclassification may induce a noncausal association between immunodeficiency and ischemic stroke. To assess potential misclassification, an HIV physician (coauthor D.B.K.) conducted a chart review of HIV-positive cases diagnosed within KPNC to identify events caused by opportunistic infections or other central nervous system processes that may mimic strokes. A random sample of 15 cases was selected for review from each of three recent CD4 cell count strata (<200, 200–499, and ≥500 cells/μl) to determine if any observed misclassification was more frequent at lower CD4 cell counts. Overall, 44/45 (98%) were identified as true ischemic strokes, with the one noncase from a patient with CD4 cell count less than 200 cells/μl. Based on this review, we concluded there was minimal outcome misclassification, and that differential misclassification by CD4 cell count was also minimal. To further exclude the possibility of misclassification, we conducted a sensitivity analysis in which we fit adjusted models comparing stroke incidence by HIV status, with HIV-positive individuals stratified by recent CD4 cell count and excluding cases occurring within three months of an opportunistic infection with neurologic involvement (i.e. toxoplasmosis, cytomegalovirus, cryptococcosis, central nervous system lymphoma, or progressive multifocal leukoencephalopathy).
Analyses were conducted in SAS 9.1 (Cary, North Carolina, USA). Statistical tests were two-sided, and statistical significance was defined as P < 0.05.