METABOLIC SYNDROME AND RELATED DISORDERS Volume 11, Number 6, 2013  Mary Ann Liebert, Inc. Pp. 417–426 DOI: 10.1089/met.2013.0017

Association of Clinical and Therapeutic Factors with Incident Dyslipidemia in a Cohort of Human Immunodeficiency Virus–Infected and Non-Infected Adults: 1994–2011 Avnish Tripathi, MD, PhD,1 Jeanette M. Jerrell, PhD,2 Angela D. Liese, PhD,3,4 Jiajia Zhang, PhD,3 Ali A. Rizvi, MD,5 Helmut Albrecht, MD,6 and Wayne A. Duffus, MD, PhD6

Abstract Objective: The aim of this study was to determine the incidence rate of dyslipidemia in a retrospective cohort of human immunodeficiency virus (HIV)-infected and non-HIV-infected adults and to evaluate the association of incident dyslipidemia with exposure to combination antiretroviral therapy (cART). Methods: The study cohort included HIV-infected individuals and a matched group of non-HIV-infected individuals served through the South Carolina Medicaid database in 1994–2011. Linkage with the HIV/AIDS surveillance database provided time-varying viro-immunological status. Time-dependent proportional hazards analysis and marginal structural models were used to assess the demographic, therapeutic, and clinical factors associated with incident dyslipidemia. Results: Among 13,632 adults with a median age of 39 years, the overall incidence rate per 1000 person years of dyslipidemia was higher in cART-treated compared to cART-naı¨ve and matched non-HIV groups (24.55 vs. 14.32 vs. 23.23, respectively). Multivariable results suggested a significantly higher risk of dyslipidemia in the cART-treated HIV-infected group [adjusted hazard ratio (aHR) = 1.18; 95% confidence interval (CI) = 1.07–1.30] and a significantly lower risk in the cART naı¨ve HIV-infected group (aHR = 0.66; CI = 0.53–0.82) compared to the control non-HIV-infected group. Marginal structural modeling suggested a significant association between incident dyslipidemia and exposure to both protease inhibitor- [adjusted rate ratio (aRR) = 1.27; CI = 1.08–1.49] and non-nucleoside reverse transcriptase inhibitor- (aRR = 1.78; CI = 1.19–2.66) based cART regimens. Pre-existing hypertension, obesity, and diabetes increased the risk of dyslipidemia, whereas hepatitis C virus, lower CD4 + T cell count, and higher HIV viral load had a protective effect. Conclusions: Incident dyslipidemia is lower in the early stages of HIV infection, but may significantly increase with cumulative exposure to cART. Viro-immunological status and underlying comorbidities have a strong association with the onset of dyslipidemia.

Introduction

A

dvances in combination antiretroviral therapy (cART) have dramatically improved survival in people living with human immunodeficiency virus/acquired immunodeficiency syndrome (HIV/AIDS) (PLWHA).1,2 By 2015, more

than half of the HIV-infected population will be older than 50 years.3,4 As the HIV population ages, there is growing concern about their increased mortality and morbidity due to the cardiometabolic disorders associated with long-term cART use.4–8 Previous studies have suggested an increased risk of dyslipidemia in HIV infection,4,9–13 which increases the risk

1

Division of Internal Medicine, Department of Medicine, University of Mississippi School of Medicine, Jackson, Mississippi. Department of Neuropsychiatry and Behavioral Science, University of South Carolina School of Medicine, Columbia, South Carolina. 3 Department of Epidemiology and Biostatistics, Arnold School of Public Health, University of South Carolina, Columbia, South Carolina. 4 Center for Research in Nutrition and Health Disparities, Arnold School of Public Health, University of South Carolina, Columbia, South Carolina. 5 Division of Endocrinology, Department of Medicine, University of South Carolina School of Medicine, Columbia, South Carolina. 6 Division of Infectious Diseases, Department of Medicine, University of South Carolina School of Medicine, Columbia, South Carolina. 2

417

418 of atherosclerosis and, consequently, cardiovascular disorders.14 The pathogenesis of dyslipidemia in HIV infection is not completely understood, but it is largely believed to be a complex interplay of the proinflammatory effect of HIV viremia, individual sociodemographic and lifestyle factors, underlying co-morbid conditions, and the adverse effects of long-term highly active antiretroviral therapy (HAART) use.15–17 Of the three main classes of antiretroviral medications used in HIV infection, protease inhibitors (PIs), non-nucleoside reverse transcriptase inhibitors (NNRTIs), and nucleoside/ nucleotide reverse transcriptase inhibitors (NRTIs), PIs have been prominently associated with an increased risk of atherogenic lipid abnormalities.4,10,14,18 Considerable within-class heterogeneity appears to exist for specific NRTIs and NNRTIs in their associated risk for incident dyslipidemia, thus warranting further investigation.4,14,19,20 Viro-immunological status and disease stage (HIV vs. AIDS) has also been significantly associated with the risk of dyslipidemia in PLWHA.14 Too frequently, previous research examining the prevalence of dyslipidemia in HIV has employed cross-sectional study designs and lacked a non-HIV-infected comparison group. To better understand the association of incident dyslipidemia and HIV infection, it is essential to account for changes in clinical and therapeutic factors over time.8 Furthermore, literature on the incidence of dyslipidemia in the HIV population in the Deep South, which is one of the most impacted regions in the United States, is scarce. In this investigation, we examine the incidence rate of dyslipidemia over time in a large retrospective cohort of HIV-infected individuals compared to a matched group of non-HIV-infected individuals, as well as the relative risk of dyslipidemia associated with exposure to different classes of cART after accounting for time-varying changes in viro-immunological status and other pertinent confounders.

Methods Participants There were 8939 HIV-infected adults ( ‡ 18 years) served through the South Carolina (SC) Medicaid system from January 1, 1994, through December 31, 2011, who were enrolled for at least 9 months of every year of the study. Exclusion criteria included death within 6 months of entry in cohort, those who had less than 30 days between the first and last visit in the cohort, or diagnostic codes of cocaine abuse or dependence due to the known toxic effect of cocaine on cardiometabolic function. A local optimal propensity score matching methodology (i.e., the GREEDY algorithm) was used to randomly select a 1:1 non-HIV-infected control group also served through the SC Medicaid program during the same study period, matched on age, race/ethnicity, gender, total months of enrollment, and year of entry into the cohort.21 The final HIV-infected case cohort consisted of 6816 patients, which when combined with the matched non-HIV-infected patients, yielded a total study cohort of 13,632. The comparison between HIV-infected and matched non-HIV-infected group on demographics and baseline medical conditions at the time of enrollment are presented in Appendix 1. Their medical and pharmacy Medicaid claims, demographics, the International Classification of Disease-Ninth Revision (ICD-9) and Current Procedural Terminology (CPT) codes associated with each medical visit,

TRIPATHI ET AL. and all medications prescribed to these individuals were linked with the enhanced HIV/AIDS Reporting System (eHARS) surveillance database for viral load (VL) and CD4 + T cell counts for all HIV-infected patients. This study was approved by the SC Department of Health and Human Services Research Committee and the SC Department of Health and Environmental Control Institutional Review Board.

Variable definition HIV-infected individuals were categorized as cART-treated if they were prescribed medications for ‡ 30 days, cumulatively; those not prescribed any antiretroviral medications were categorized as cART naı¨ve. Incident dyslipidemia, not occurring within 6 months of the first date of Medicaid enrollment (washout period), was defined as: (1) ‡ 2 medical visit claims, at least 30 days apart, with an ICD-9 code 272.0x (pure hypercholesterolemia), 272.1x (pure hyperglyceridemia), 272.2x (mixed hyperlipidemia), 272.3x (hyperchylomicronemia), or 272.4x (other, unspecified hyperlipidemia); or (2) a prescription of any lipid-lowering drug for ‡ 30 days. Use of a 6-month washout period mitigated misclassification of a prevalent condition as incident. Similar definitions were used for the covariates: diabetes (250.xx and/or prescription of antidiabetic medications, including insulin, for ‡ 30 days), essential hypertension (401.xx and/or prescription of antihypertensive medications for ‡ 30 days), obesity/overweight (278.xx), hepatitis B virus (070.3x), hepatitis C virus (HCV) (070.5x), and tobacco use (469.0x and 350.1x and/or any prescription for smoking cessation).

Statistical analyses Exploratory and descriptive analyses were performed to determine data distribution and assess univariate association of covariates with the main outcome variable, incident dyslipidemia. The incidence density rate of dyslipidemia per 1000 person-years (PY) was calculated separately for the three main exposure groups: HIV-infected HAART naı¨ve, HIV-infected HAART-treated, and non-HIV-infected. The stratified incidence rates were computed by gender, race/ ethnicity, and age categories. Time-dependent proportional hazards analysis was used to estimate the relative risk [hazard ratio (HR)] of new-onset dyslipidemia among the three exposure groups after accounting for time-varying and fixed confounding factors. Comorbid metabolic disorders (i.e., essential hypertension, diabetes, obesity), and HCV and HBV co-infection were used as time-varying covariates. To explore the association of antiretroviral treatment and viro-immunological control per month with development of dyslipidemia, a subgroup analysis was conducted only among HIV-infected individuals. In this analysis, time-dependent exposure to PI- and NNRTI-based regimens and viro-immunological status per person-month were included in the model as predictor variables. Because NRTI medications are common components of PI- and NNRTI-based regimens, exposure to these medications was not included as a covariate to avoid collinearity. Modeling exposure to PIs and NNRTIs as timedependent variables also allowed accounting for individuals who were treated with both types of regimens over the study period.

Table 1.

Characteristics of the Total Cohort and New Onset Dyslipidemia Group (1994–2011)

Variable HIV exposure category HAART naı¨ve HAART treated Non HIV-infected Gender Female Male Race Black Others White Age categories 18–29 years 30–44 years 45–64 years ‡ 65 years Median age (years) Median months enrolled First enrollment year 1994–1996 1997–1999 2000–2002 2003–2005 2006–2008 2009–2011 Co-morbid hypertension Yes No Co-morbid diabetes Yes No Documented obesity Yes No Co-morbid hepatitis B Yes No Co-morbid hepatitis C Yes No AIDS Yes No PI-based treatment Yes No NNRTI-based treatment Yes No NRTI-based treatment Yes No Any antiretroviral Yes No Median Log10 VLc Median CD4 cellsd Documented tobacco use Yes No

All subjects (n = 13,632)

New-onset dyslipidemia ‘‘No’’ New-onset dyslipidemia ‘‘Yes’’ (n = 11,619; 85%) (n = 2013; 15%) Odds ratio P valuea

1338 (9.82) 5478 (40.18) 6816 (50.0)

1242 (92.83) 4568 (83.39) 5809 (83.23)

96 (7.17) 910 (16.61) 1007 (14.77)

0.45 1.15 Ref.

< 0.001 < 0.001

5893 (43.23) 7739 (56.77)

5007 (84.97) 6612 (85.44)

806 (15.03) 1127 (14.56)

1.04 Ref.

0.441

9642 (70.73) 1147 (8.41) 2843 (20.86)

8211 (85.16) 967 (84.31) 2441 (85.86)

1431 (14.84) 180 (15.69) 402 (14.14)

1.06 1.13 Ref.

0.935 0.278

373 9343 3762 154 38 51

(2.74) (68.54) (27.60) (1.13) (31–46) (23–105)

352 8120 3019 128 37 43

(94.37) (86.91) (80.25) (83.12) (30–45) (19–88)

21 1223 743 26 42 120

(5.63) (13.09) (19.75) (16.88) (34–49) (76–175)

Ref. 2.52 4.12 3.40 1.28b 1.06b

5083 2187 2049 1680 1443 1190

(37.29) (16.04) (15.03) (12.32) (10.59) (8.73)

4250 1803 1725 1417 1287 1137

(83.61) (82.44) (84.19) (84.35) (89.19) (95.55)

833 384 324 263 156 53

(16.39) (17.56) (15.81) (15.65) (10.81) (4.45)

1.04 1.13 Ref. 0.99 0.65 0.25

< 0.001 < 0.001

0.679 < 0.001 0.052 < 0.001 < 0.001

< 0.001 0.039 < 0.001

5095 (37.38) 8537 (62.62)

3509 (68.87) 8110 (95.00)

1586 (31.13) 427 (5.00)

8.58 Ref.

< 0.001

1972 (14.47) 11,660 (85.53)

1150 (58.32) 10,469 (89.79)

822 (41.68) 1191 (10.21)

6.28 Ref.

< 0.001

1340 (10.83) 12,292 (90.17)

838 (62.54) 10,781 (87.71)

502 (37.46) 1511 (12.29)

4.27 Ref.

< 0.001

477 (3.50) 13,155 (96.50)

396 (83.02) 11,223 (85.31)

81 (16.98) 1932 (14.69)

1.19 Ref.

0.178

1114 (8.17) 12,518 (91.83)

944 (84.74) 10,675 (85.28)

170 (15.26) 1843 (14.72)

1.04 Ref.

0.624

5560 (40.79) 8072 (59.21)

4790 (86.15) 6829 (84.60)

770 (13.85) 1243 (15.54)

0.88 Ref.

< 0.001

3642 (26.72) 9990 (73.28)

2999 (82.34) 8620 (86.29)

643 (17.66) 1370 (13.71)

1.35 Ref.

< 0.001

3150 (23.11) 10,482 (76.89)

2464 (78.22) 9155 (87.34)

686 (21.78) 1327 (12.66)

1.92 Ref.

< 0.001

5440 (39.91) 8192 (60.09)

4531 (83.29) 7088 (86.52)

909 (16.71) 1104 (13.48)

1.28 Ref.

< 0.001

5478 8154 3.38 202

4568 7051 3.53 179

910 1103 2.74 377

1.27 Ref. 0.37b 1.07b

< 0.001

(40.18) (59.82) (2.23–4.41) (79–426)

4094 (30.03) 9538 (69.97)

(83.39) (86.47) (2.47–4.57) (66–391)

3232 (78.94) 8387 (87.93)

(16.61) (13.53) (2.08–3.64) (235–561)

862 (21.06) 1151 (12.07)

1.94 Ref.

< 0.001 < 0.001 < 0.001

Data are presented as n (%). a P values are obtained from the univariate logistic regression. b Odds ratio for continuous variable are given for increments of 5 years, 6 months, 5 log10, and 50 cell count for age, months of enrollment, log 10 VL, and CD4 + T cell count, respectively. c,d The median values presented in the table for the log10 VL and CD4 T cell count represent the median of the mean value for a person over the entire period of observation. This enables to assess the viro-immunological control of a person over the entire follow-up period. HAART, highly active antiretroviral therapy; AIDS, acquired immunodeficiency syndrome; PI, protease inhibitors; NNRTI, non-nucleoside reverse transcriptase inhibitors; NRTI, nucleoside/nucleotide reverse transcriptase inhibitors; VL, viral load.

419

420 Marginal structural analysis has been used to explore causal/temporal associations in observational data sets. We used marginal structural models with the inverse probability of treatment-weighted (IPTW) estimators to further assess the association of exposure to PI- and NNRTI-based regimens on the development of dyslipidemia. The IPTW estimators took into account the probability of receiving the particular class of cART (e.g., PI or NNRTI), depending on previous exposure history and the probability of remaining in the dataset in any person-month. The predictor variables used for computing IPTW weights included individual characteristics, baseline and time-dependent co-morbid conditions, and medications for each person-month from the initiation of treatment of interest to being censored. While calculating the IPTW for PI exposure, exposure to NNRTIs over time is included as a covariate and vice versa. Overall, marginal structural models produce an artificial randomization in observational data similar to randomized controlled trials and, therefore, further reduce the residual confounding which is inherent in conventional methodology such as the Cox proportional hazards analysis. To obtain a final parsimonious model, each multivariable model was reduced by backward elimination using a cutoff P value of 0.08 and ensuring that removal of a variable did not result in more than 10% change in the dependent variable. The adjusted rate ratio (aRR) along with 95% confidence interval (CI) and associated P value are reported. P values of less than 0.05 (two-sided tests) were considered statistically significant and the Tukeyadjusted P value was used for pairwise multiple comparisons, where they were performed. All statistical analyses were performed in SAS software, version 9.2 (SAS Institute, Cary, North Carolina).

TRIPATHI ET AL.

Results The median age of the study cohort patients was 39 years [interquartile range (IQR), 31–46 years), and the majority was male (57%) and African American (71%) (Table 1). Patients were followed for a median of 70 months (IQR, 23–105) or a total of 88,359 PY. In HIV-infected individuals, the median log10 VL was 3.37 (IQR, 2.36–4.42) and median CD4 + T cell count was 208 cells/mm3 (IQR, 78–426 cells/mm3). In the HIV-infected group, 5478 (80%) were prescribed ART. Of these, 3642 (66%) and 3150 (46%) were treated with PI- and NNRTI-based regimens, respectively, at least once over the study period (not presented in Table 1). In addition, 1943 (35%) were treated with both PI- and NNRTI-based regimens over the study period. Overall, 15% of the cohort developed incident dyslipidemia, 37% had essential hypertension, 15% had diabetes mellitus, and 10% had documented obesity; 8% had a HCV co-infection and 30% had documented tobacco use (Table 1). Stratified incidence rates per 1000 PY are presented in Fig. 1. The overall incidence rate of dyslipidemia was found to be higher in cart-treated HIV-infected individuals compared to cART-naı¨ve HIV-infected and the matched group of non-HIV-infected individuals (24.55 vs. 14.32 vs. 23.23, respectively). Appendix 2 presents stratified incidence rates along with associated 95% CIs. The multivariable time-dependent Cox proportional hazards analysis suggested that the risk of developing incident dyslipidemia was significantly higher in the cART-treated HIV-infected group [adjusted hazard ratio (aHR) = 1.18; CI = 1.07–1.30] and significantly lower in the cART-naı¨ve HIV-infected group (aHR = 0.66; CI = 0.53–0.82) compared to the non-HIV-infected group (Table 2). Adjusted failure curves from this analysis are presented in Fig. 2, which are

FIG. 1. Comparison of incidence rate of dyslipidemia per 1000 person-years in human immunodeficiency virus (HIV)infected and non-HIV-infected control population in South Carolina. cART, combination antiretroviral therapy.

INCIDENCE OF DYSLIPIDEMIA IN HIV INFECTION

421

Table 2. Time-Dependent Proportional Hazards Analysis: Factors Associated with Increased Risk of Dyslipidemia in HIV-Infected Individuals

Unadjusted models Variables Subject category cART naı¨ve HIV-infected cART treated HIV-infected Non HIV-infected (controls) Gender Female Male Race/ethnicity Black Others1 White Age at enrollment (years) Pre-existing hypertension Yes No Pre-existing diabetes Yes No Pre-existing obesity Yes No Pre-existing hepatitis B Yes No Pre-existing hepatitis C Yes No Tobacco use (ever) Yes No Concomitant PI exposure Yes No Concomitant NNRTI exposure Yes No Concomitant NRTI exposure Yes No CD4 + T cell count cells/mm3 < 200 200–499 > 499 HIV load log(10) copies/mL

Model 1: HIV-infected versus non-HIV-infected groups

Model 2: HIV-infected group: Exposed to PIs and/or NNRTIs versus not exposed

HR

95% CI

aHR

95% CI

aHR

0.65 1.05 1.00

0.52–0.80 0.96–1.15 Ref.

0.66 1.18 1.00

0.53–0.82 1.07–1.30 Ref.

N/A

0.74 1.00

0.67–0.80 Ref.

0.85 1.29 1.00 1.01

0.76–0.95 1.08–1.54 Ref. 1.00–1.02

0.76 0.97 1.00 1.03

0.68–0.86 0.81–1.17 Ref. 1.03–1.04

0.68 0.90 1.00 1.03

0.56–0.82 0.67–1.20 Ref. 1.02–1.04

3.71 1.00

3.37–4.09 Ref.

2.60 1.00

2.34–2.89 Ref.

2.44 1.00

2.05–2.91 Ref.

4.01 1.00

3.54–4.55 Ref.

2.46 1.00

2.14–2.82 Ref.

2.16 1.00

1.71–2.72 Ref.

2.18 1.00

1.91–2.49 Ref.

1.61 1.00

1.39–1.86 Ref.

1.29 1.00

1.01–1.66 Ref.

1.08 1.00

0.76–1.52 Ref.

1.05 1.00

0.83–1.33 Ref.

0.61 1.00

0.47–0.79 Ref.

0.45 1.00

0.32–0.64 Ref.

1.37 1.00

1.25–1.50 Ref.

1.19 1.00

1.09–1.31 Ref.

2.19 1.00

1.92–2.49 Ref.

N/A

2.09 1.00

1.78–2.45 Ref.

3.06 1.00

2.67–3.50

N/A

2.27 1.00

1.93–2.68 Ref.

2.93 1.00

2.56–3.36 Ref.

N/A

N/A

0.48 0.76 1.00 0.99

0.41–0.56 0.65–0.89 Ref. 0.98–0.99

N/A

0.66 0.77 1.00 0.86

N/A

95% CI

0.53–0.81 0.63–0.92 Ref. 0.80–0.92

1 The ‘‘other’’ category is comprised of Hispanic, Asian, and Pacific Islander groups. Data presented are hazards ratio (HR) or adjusted hazards ratio (aHR) along with their associated 95% confidence interval (95% CI). Model 1 included both HIV infected and non-HIV-infected control group, whereas Mode 2 included on HIV-infected groups. HIV, human immunodeficiency virus; PI, protease inhbitors; NNRTIs, non-nucleotide reverse transcriptase inhibitors; cART, combination antiretroviral therapy; N/A, not applicable.

adjusted for gender, race, age, year of enrollment, and tobacco use, but not for time-dependent co-morbidities. Black race/ethnicity compared to white was associated with a significantly lower risk (aHR = 0.76; CI = 0.68–0.86), whereas older age (aHR = 1.03; CI = 1.03–1.04) was associated with a higher risk of incidental dyslipidemia. Furthermore, pre-

existing hypertension (aHR = 2.60; CI = 2.34–2.89), diabetes (aHR = 2.46; CI = 2.14–2.82), obesity (aHR = 1.61; CI = 1.39– 1.86), and tobacco use (aHR = 1.19; CI = 1.09–1.31) were associated with a higher risk, whereas, pre-existing hepatitis C was associated with a significantly lower risk of incident dyslipidemia (aHR = 0.61; CI = 0.47–0.79) (Table 2).

422

FIG. 2. Failure curves from the Cox proportional hazards analysis for the probability of incident dyslipidemia among human immunodeficiency virus (HIV)-infected combination antiretroviral therapy (cART)-treated, HIV-infected cART-naı¨ve, and non-HIV-infected comparison groups.

INCIDENCE OF DYSLIPIDEMIA IN HIV INFECTION Table 3. Results of Marginal Structural Models with Stabilized Inverse Propensity Treatment Weights to Examine the Causal Association Between Combination Antiretroviral Therapy Classes and Development of Dyslipidemia Antiretroviral exposure aRRa PIs NNRTIs

1.27 1.78

95% CI 1.08–1.49 1.19–2.66

Mean P value IPTW 0.004 0.005

1.13 1.78

Range IPTW 0.003–1219 0.0003–15a

Adjusted relative risk (aRR) represents measure of association obtained from the parsimonious multivariable marginal structural models obtained through backward elimination. The covariates remaining in these models were similar to that given for the Cox proportional hazards analysis in Table 2. The associated 95% confidence intervals (95% CI) and P values have also been presented. The mean IPTWs should ideally be as close to one as possible and the range should not be too high. a The upper limit of IPTW was truncated to the maximum of 15 to stabilize the MSM model. CI, confidence interval; IPTW, inverse propensity treatment weights; PIs, protease inhibitors; NNRTIs, non-nucleotide reverse transcriptase inhibitors.

The impact of cART medication classes on development of incident dyslipidemia was explored next in the HIV-infected cohort. A significantly higher risk of incident dyslipidemia was independently associated with cumulative exposure to both PI-based regimens (aHR = 2.09; CI = 1.78–2.45) and NNRTI-based regimens (aHR = 2.27; CI = 1.93–2.68). In addition, a poorer viro-immunological status was associated with a lower risk of dyslipidemia after controlling for other confounding factors (aHR = 0.66; aHR = 0.77; aHR = 0.86, respectively, for lower CD4 + counts, and HIV load) (Table 2). Results of the marginal structural models created to explore exposure to PIs and NNRTIs also suggested a significant association between the development of incident dyslipidemia and cumulative exposure to PIs (aRR = 1.27; CI = 1.08–1.49) and NNRTIs (aRR = 1.78; CI = 1.19–2.66). However, the measure of association was slightly lower than in the conventional Cox proportional hazards analysis (Table 3).

Discussion These results add to the knowledge base in several ways. We addressed the risk of incident dyslipidemia in the HIVinfected population in the US Deep South, a region disproportionately affected by one of the highest national rates of mortality and morbidity among PLWHA.22 Methodologically, we included time-varying exposure history to cART medications, pre-existing cardiometabolic conditions, and viro-immunological status for each person-month in the analysis. Our results suggest a substantially higher risk of incident dyslipidemia among cART-treated, HIV-infected adults and an independent temporal association with cumulative exposure to both PI- and NNRTI-based regimens. These results also suggest that HIV infection alone, unconfounded by cART exposure, exerted a protective effect against incident dyslipidemia. Some previous studies have produced similar results,23,24 noting a significant decline in total serum cholesterol immediately after HIV seroconversion and a subsequent return to preconversion levels after initiation of cART. Alternatively, several cross-sectional

423 studies have reported increased levels of total cholesterol, including high levels of atherogenic low-density lipoprotein particles in HAART-naı¨ve HIV-infected individuals.12,19 Furthermore, among HIV-infected individuals, a declining immune status, i.e., higher VL and lower CD4 + T cell count, was associated with a significantly lower risk of dyslipidemia. Enkhmaa et al. recently suggested in a biethnic population study that lower CD4 + T cell count and higher HIV viral load were significantly associated with lower levels of allele-specific apolipoprotein (a) [Apo(a)] and concluded that HIV subjects with improving disease status may be at a higher cardiovascular risk.25 Taken together, these results may indicate that newly diagnosed HIV-infected patients have lower serum cholesterol levels on initial examination but, as the disease progresses, healthcare providers should regularly evaluate and manage dyslipidemia as recommended by professional organizations.26 Many previous studies have reported a significant class effect of PIs on the development of dyslipidemia.4,12,14,16,26 However, the literature suggests mixed results for the class effect of NNRTIs and NRTIs,4,14,19,20 which may be attributed to greater within-class heterogeneity in the lipid impact profiles of some medications.11 Nevertheless, the strong association of NNRTIs and NRTIs with development of incident dyslipidemia in this investigation may suggest inadequate clinical use of regimens with better lipid profiles and, therefore, not only warrants further investigation, but also improved clinical attention.27,28 Previous studies have also shown mixed results for racial differences in the prevalence and incidence of dyslipidemia,29–31 whereas our results suggest a significantly lower incidence and adjusted risk of dyslipidemia among blacks as compared to whites. Moreover, in the very few studies examining the association of HCV infection and the development of dyslipidemia, a majority demonstrate a protective effect in both HIV-infected and non-HIV-infected populations.32 In our study, a significant negative association between HCV infection and the development of incident dyslipidemia was also found. Finally, consonant with the extant literature, pre-existing hypertension, diabetes, and obesity were the most significant predictors for incident dyslipidemia in this investigation.14,27,33 These results underscore the importance of early diagnosis and management of cardiometabolic conditions, especially in cART-treated HIV-infected individuals. These results should be interpreted with following limitations in mind. The data were collected for administrative and billing purposes, and neither the recruitment nor the data collection was under control of the investigator. Nevertheless, we used conservative criteria for defining both the outcome and predictor variables to mitigate the risk of variable misclassification. Information on potentially important covariates such as diet, physical activity, and family history of metabolic conditions was lacking in this data set. The Medicaid-based findings may not be generalizable to other states or insured populations. In conclusion, there is growing consensus that dyslipidemia is an important therapeutic target to mitigate the increasing risk of cardiovascular disorders, including atherosclerosis, among PLWHA. Healthcare providers should make every effort to evaluate and manage dyslipidemia in the early stages of HIV by exploring the options of starting or switching to cART regimens with better lipid profiles and

424 directly managing apparent lipid abnormalities with recommended lifestyle modifications and medications.

Acknowledgments This work was partially supported by funding from the Mid-Atlantic American Heart Association Pre-Doctoral Fellowship. The views expressed do not necessarily represent those of the funding source or official findings of the South Carolina Department of Health and Human Services (Medicaid) or the South Carolina Department of Health and Environmental Control.

Author Disclosure Statement No competing financial interests exist.

References 1. Walensky RP, Paltiel AD, Losina E, et al. The survival benefits of AIDS treatment in the United States. J Infect Dis 2006;194:11–19. 2. Palella FJ, Jr., Delaney KM, Moorman AC, et al. Declining morbidity and mortality among patients with advanced human immunodeficiency virus infection. HIV Outpatient Study Investigators. N Engl J Med 1998;338:853–860. 3. High KP, Effros RB, Fletcher CV, et al. Workshop on HIV Infection and Aging: What is known and future research directions. Clin Infect Dis 2008;47:542–553. 4. Giannarelli C, Klein RS, Badimon JJ. Cardiovascular implications of HIV-induced dyslipidemia. Atherosclerosis 2011; 219:384–389. 5. Palella FJ, Jr., Baker RK, Moorman AC, et al. Mortality in the highly active antiretroviral therapy era: Changing causes of death and disease in the HIV outpatient study. J Acquir Immune Defic Syndr 2006;43:27–34. 6. Aberg JA. Cardiovascular complications in HIV management: Past, present, and future. J Acquir Immune Defic Syndr 2009;50:54–64. 7. Ho JE, Hsue PY. Cardiovascular manifestations of HIV infection. Heart 2009;95:1193–1202. 8. Friis-Moller N, Reiss P, Sabin CA, et al. Class of antiretroviral drugs and the risk of myocardial infarction. N Engl J Med 2007;356:1723–1735. 9. Pefura Yone EW, Betyoumin AF, Kengne AP, et al. First-line antiretroviral therapy and dyslipidemia in people living with HIV-1 in Cameroon: A cross-sectional study. AIDS Res Ther 2011;8:33. 10. Dube MP, Cadden JJ. Lipid metabolism in treated HIV Infection. Best Pract Res Clin Endocrinol Metab 2011;25: 429–442. 11. Crane HM, Grunfeld C, Willig JH, et al. Impact of NRTIs on lipid levels among a large HIV-infected cohort initiating antiretroviral therapy in clinical care. AIDS 2011;25: 185–195. 12. Armstrong C, Liu E, Okuma J, et al. Dyslipidemia in an HIVpositive antiretroviral treatment-naive population in Dar es Salaam, Tanzania. JAIDS J Acq Immune Defic Syndr 2011; 57(2):141–145. 13. Tebas P. HIV and cardiometabolic abnormalities: New perspectives and treatment update. J Acquir Immune Defic Syndr 2008;49(Suppl 2):S77–S78. 14. Green ML. Evaluation and management of dyslipidemia in patients with HIV infection. J Gen Intern Med 2002;17: 797–810.

TRIPATHI ET AL. 15. Anastos K, Lu D, Shi Q, et al. Association of serum lipid levels with HIV serostatus, specific antiretroviral agents, and treatment regimens. J Acquir Immune Defic Syndr 2007;45:34–42. 16. Goff DC, Bertoni AG, Kramer H, et al. Dyslipidemia prevalence, treatment, and control in the Multi-Ethnic Study of Atherosclerosis (MESA). Circulation 2006;113:647–656. 17. Palios J, Kadoglou NP, Lampropoulos S. The pathophysiology of HIV-/HAART-related metabolic syndrome leading to cardiovascular disorders: the emerging role of adipokines. Exp Diabetes Res 2012;2012:103063. 18. Samaras K, Wand H, Law M, et al. Prevalence of metabolic syndrome in HIV-infected patients receiving highly active antiretroviral therapy using International Diabetes Foundation and Adult Treatment Panel III criteria: associations with insulin resistance, disturbed body fat compartmentalization, elevated C-reactive protein, and [corrected] hypoadiponectinemia. Diabetes Care 2007;30:113–119. 19. Clotet B, van der Valk M, Negredo E, et al. Impact of nevirapine on lipid metabolism. J Acquir Immune Defic Syndr 2003;34(Suppl 1):S79–S84. 20. Buscher AL, Giordano TP. Gaps in knowledge in caring for HIV survivors long-term. JAMA 21;304:340–341. 21. Coca-Perraillon M. Local and global optimal propensity score matching. Statistics and Data Analysis 2007. Available at http://www2.sas.com/proceedings/forum2007/185-2007 .pdf Accessed December 3, 2011. 22. Reif S, Geonnotti K, Whetten K. HIV infection and AIDS in the Deep South. Am J Public Health. 2006;96:970. 23. Riddler SA, Smit E, Cole SR, et al. Impact of HIV infection and HAART on serum lipids in men. JAMA 2003;289:2978– 2982. 24. Shor-Posner G, Basit A, Lu Y, et al. Hypocholesterolemia is associated with immune dysfunction in early human immunodeficiency virus-1 infection. Am J Med 1993;94:515–519. 25. Enkhmaa B, Anuurad E, Zhang W, et al. HIV disease activity as a modulator of lipoprotein(a) and allele-specific apolipoprotein(a) levels. Arterioscler Thromb Vasc Biol 2013;33: 387–392. 26. Dube´ M, Stein J, Aberg J, et al. Guidelines for the evaluation and management of dyslipidemia in human immunodeficiency virus (HIV)-infected adults receiving antiretroviral therapy: recommendations of the HIV Medicine Association of the Infectious Disease Society of America and the Adult AIDS Clinical Trials Group. Clin Infect Dis 2003;37:613–627. 27. Dube MP. Guidelines for the evaluation and management of dyslipidemia in human immunodeficiency virus (HIV)infected adults receiving antiretroviral therapy: recommendations of the HIV Medical Association of the Infectious Disease Society of America and the Adult AIDS Clinical Trials Group. Clin Infect Dis 2003;37:613–627. 28. Expert Panel on Detection E, Treatment of High Blood Cholesterol in A. Executive Summary of the Third Report of the National Cholesterol Education Program (NCEP) Expert Panel on Detection, Evaluation, and Treatment of High Blood Cholesterol in Adults (Adult Treatment Panel III). JAMA 2001;285:2486–2497. 29. Pullinger CR, Aouizerat BE, Gay C, et al. Metabolic abnormalities and coronary heart disease risk in human immunodeficiency virus-infected adults. Metab Syndr Relat Disord 2010;8:279–286. 30. O’Meara JG, Kardia SL, Armon JJ, Brown CA, Boerwinkle E, Turner ST. Ethnic and sex differences in the prevalence, treatment, and control of dyslipidemia among hypertensive adults in the GENOA study. Arch Intern Med 2004;164:1313– 1318.

INCIDENCE OF DYSLIPIDEMIA IN HIV INFECTION 31. Sharma MD, Pavlik VN. Dyslipidaemia in African Americans, Hispanics and Whites with type 2 diabetes mellitus and hypertension. Diabetes Obes Metab. 2001;3:41–45. 32. Stapleton JT, Bennett K, Bosch RJ, et al. Effect of antiretroviral therapy and hepatitis C co-infection on changes in lipid levels in HIV-infected patients 48 weeks after initiation of therapy. HIV Clin Trials 2007;8:429–436. 33. Grinspoon S, Carr A. Cardiovascular risk and body-fat abnormalities in HIV-infected adults. N Engl J Med 2005; 352:48–62.

425 Address correspondence to: Avnish Tripathi, MD, PhD Division of Internal Medicine Department of Medicine University of Mississippi School of Medicine 2500 North State Street Jackson, MI 39216 E-mail: [email protected]

(Appendix follows/)

426

TRIPATHI ET AL. Appendix 1. Comparison of Demographic Characteristics and Baseline Medical Conditions Between HIV-Infected and the Matched Non-HIV-Infected Study Cohorts Served Through the South Carolina Medicaid (1994–2011) All subjects (n = 13,632)

Variablea Gender Female Male Race Black Others White Age categories 18–29 years 30–44 years 45–64 years ‡ 65 years Median age (years) Median months enrolled Baseline hypertension Yes No Baseline diabetes Yes No Baseline obesity Yes No Co-morbid hepatitis B Yes No Co-morbid hepatitis C Yes No Documented tobacco use Yes No

HIV-infected (case) (n = 6816; 50%)

Non-HIV infected (control) (n = 6816; 50%)

P valuea

5893 (43.23) 7739 (56.77)

2953 (21.66) 3863 (28.34)

2940 (21.57) 3876 (28.43)

0.822

9642 (70.73) 1147 (8.41) 2843 (20.86)

4856 (35.62) 560 (4.11) 1400 (10.27)

4786 (35.11) 587 (4.31) 1443 (10.59)

0.408

373 9343 3762 154 38 51

186 4705 1857 68 38 53

187 4638 1905 86 38 50

0.369

(2.74) (68.54) (27.60) (1.13) (31–46) (23–105)

(1.36) (34.51) (13.62) (0.50) (31–46) (22–106)

(1.37) (34.02) (13.97) (0.63) (30–46) (23–103)

0.408 0.059

1894 (13.89) 11,738 (86.11)

921 (6.76) 5895 (43.24)

973 (7.14) 5843 (42.86)

0.198

764 (5.60) 12,868 (94.40)

387 (2.84) 6429 (47.16)

377 (2.77) 6439 (47.23)

0.710

200 (1.47) 13,432 (98.53)

102 (0.75) 6714 (49.25)

98 (0.72) 6718 (49.28)

0.776

477 (3.50) 13,155 (96.50)

431 (3.16) 6385 (46.84)

46 (0.34) 6770 (49.66)

< 0.001

1114 (8.17) 12,518 (91.83)

927 (6.80) 5889 (43.20)

187 (1.37) 6629 (48.63)

< 0.001

4094 (30.03) 9538 (69.97)

1994 (14.63) 4822 (35.37)

2100 (15.40) 4716 (34.60)

0.048

Data presented are n (%), except for continuous variables where data are presented as median (interquartile range). a Baseline medical conditions are defined as relevant co-morbidities including hypertension, diabetes, and obesity diagnosed within 3 months of enrollment in the study. However hepatitis B and C are presented as cross-section data because of relative small numbers overall.

Appendix 2. Incidence Rate of Dyslipidemia per 1000 Person-Years of Follow-Up of Persons Served Through the South Carolina Medicaid Program from January 1, 1994, Through December 31, 2011 Characteristics Overall Gender Male Female Race White Black Othersa Age categories 18–29 years 30–44 years 45–64 years ‡ 65 years

cART naı¨ve HIV-infected

Non-HIV-infected (controls)

24.55 (22.98–26.21)

14.32 (11.60–17.48)

23.23 (21.81–24.71)

29.00 (36.49–31.68) 20.81 (18.86–22.91)

11.71 (8.24–16.14) 16.64 (12.67–21.47)

26.04 (24.00–28.21) 20.06 (18.16–22.10)

30.57 (26.58–35.00) 22.33 (20.61–24.16) 33.22 (26.3–41.4)

12.98 (7.10–21.78) 13.89 (10.85–17.52) 21.37 (10.67–38.23)

22.01 (18.90–25.49) 22.79 (21.18–24.49) 31.65 (25.45–38.9)

11.67 25.37 44.67 37.25

7.24 14.65 21.76 21.72

8.63 23.5 38.73 30.98

cART-treated HIV-infected

(9.82–13.77) (23.10–27.81) (39.76–14 9.92) (14.98–76.75)

(3.96–12.15) (10.69–19.61) (15.07–30.41) (4.48–63.48)

(7.03–10.48) (21.46–25.68) (35.04–42.70) (17.71–50.32)

Data are incidence rates were 1000 person-years of follow-up with their associated 95% confidence interval around the incidence rate given in parenthesis. a ‘‘Other’’ race/ethnicity includes Hispanics, Asians, Pacific Islanders, and other non-black non-white race ethnicity. cART, combination antiretroviral therapy; HIV, human immunodeficiency virus.