Carter, Rosalind J. PhD*; Wiener, Jeffrey MS†; Abrams, Elaine J. MD‡; Farley, John MD§; Nesheim, Steven MD∥; Palumbo, Paul MD¶; Bulterys, Marc MD, PhD†; for the Perinatal AIDS Collaborative Transmission Study-HIV Follow-up after Perinatal Exposure (PACTS-HOPE) Group
Highly active antiretroviral therapy (HAART) consisting of protease inhibitors (PIs) or nonnucleoside reverse transcription inhibitors (NNRTIs) in combination with nucleoside reverse transcription inhibitors (NRTIs) has dramatically reduced morbidity and mortality among HIV-infected children.1 However, with more widespread use, there have been an increasing number of reports describing a variety of metabolic abnormalities, including dyslipidemia, insulin resistance, and changes in body fat distribution.2-10 Several studies in adults not affected by HIV11,12 have demonstrated a correlation between high cholesterol levels during adolescence and the development of atherosclerotic lesions as young adults, suggesting that dyslipidemia among HIV-infected youth may ultimately increase their long-term risk for cardiovascular disease. Among adults, HAART therapy has also been associated with an elevated risk of myocardial infarction13,14 that seems to increase with prolonged HAART exposure.15,16
A number of small clinical studies2-4 have documented an increased risk of atherogenic dyslipidemia among children on PI-inclusive antiretroviral (ARV) regimens. Recent cross-sectional analyses of large cohorts in Europe5 and the United States10 have confirmed these findings. Evidence of a direct effect of PI-inclusive regimen on lipid values comes from studies of HIV-infected adults and children17 that have demonstrated that, compared with pretreatment measures, lipid values substantially increase after PI therapy is initiated. In addition, seronegative adult volunteers18 were shown to develop dyslipidemia after a short course of treatment with a PI. Furthermore, McComsey et al19 found that children enrolled in an efavirenz substitution study had a significant reduction in cholesterol and triglyceride levels after eliminating PIs from their ARV regimens.
Whereas there is increasing evidence of a direct relationship between PI treatment and dyslipidemia, not all children on these regimens develop hypercholesterolemia (HC). In general, the cross-sectional design and small number of lipid measurements in most pediatric studies limit more complex analyses identifying risk factors for dyslipidemia.
In this study, we examined the effects of PIs on lipid measures in a large cohort of perinatally HIV-infected children followed from birth. We modeled dyslipidemia longitudinally as a function of the child s ARV regimen at the time of their lipid measurement while controlling for important covariates including viral load, CD4 T-lymphocyte count, and body mass index (BMI) at the same time point. This allowed us to include a large number of lipid measurements and examine how changes in children s ARV regimens affected lipid levels over time.
Between April 1986 and September 1998, women at risk for HIV-1 infection were enrolled in the Perinatal AIDS Collaborative Transmission Study (PACTS), a longitudinal prospective study of perinatal HIV-1 transmission and the natural history of pediatric HIV-1 disease. The study was conducted at health care centers in Newark, New Jersey; Baltimore, Maryland; Atlanta, Georgia; and New York City and funded by the Centers for Disease Control and Prevention (CDC).20,21 Evaluations of women and infants were carried out at regular intervals and included questionnaires, medical record abstraction, disease staging, and phlebotomy. In 2001, all surviving HIV-infected children not lost to follow-up throughout the duration of PACTS and a frequency-matched random sample of the HIV-exposed, uninfected children originally enrolled in PACTS and their families were asked to'participate in a new protocol, the HIV Follow-up of Perinatally Exposed Children Study (PACTS-HOPE), which focused on pediatric HIV disease progression and psychosocial development.22 As part of the PACTS-HOPE study, information, including nonfasting total serum cholesterol and triglycerides, CD4 T-lymphocyte count and percent, and HIV-1 RNA viral load (copies/mL), and ARV regimen changes were abstracted from medical records for all clinic visits from the child s last PACTS study visit. For a few children, this included visits as early as 1997. Most chart abstractions covered 1999-2004. Data files from PACTS and PACTS-HOPE were linked, providing continuous information on ARV use from birth through April 2004.
Laboratory studies of total serum cholesterol and triglycerides, CD4 T-lymphocyte count, and HIV-1 RNA were performed by commercial laboratories as part of routine clinical care. Most lipid measures were based on nonfasting blood. Most triglyceride measurements were obtained in conjunction with cholesterol measurements. HC was defined23 as total cholesterol ≥200 mg/dL. Hypertriglyceridemia (HT) was defined5 as triglycerides ≥150 mg/dL. Because most samples were tested using the NASBA HIV-1 RNA quantification kit, whose lower limit of detection is 400 copies/mL, viral load <400 copies/mL was defined as undetectable. Immunologic categories, based on age and CD4 T-lymphocyte count, were defined according to the CDC pediatric classification system.24 BMI was calculated as weight over height squared (kg/m2). Age- and sex-adjusted BMI percentiles were used to compare children ≥90th percentile with those <90th percentile. Pubertal stage of patients was defined according to Tanner criteria by the child s physician during clinical visits. BMI and Tanner staging were recorded every 6 months based on clinical exam. Lipodystrophy diagnosis was based on clinical observation and not on systematic evaluation of all children. Each study site noted any diagnosis of lipoatrophy (extremity wasting) or lipohypertrophy (dorsocervical fat accumulation, or truncal adiposity) during reviews of the child s medical record every 6 months.
Children were classified by ARV regimen based on the medications they were taking on the day of the lipid measurement. Most children had been on a regimen for 30days before lipid measurement. Our definition of "ritonavir-containing regimens" included both regimens where ritonavir was the primary PI, and where ritonavir was used in small doses to boost the effect of another PI. In calculating the number of PIs included in a regimen, medications that combined a PI with a boosting dose of ritonavir (eg, Kaletra) were coded as a single PI. Children who changed to PI-inclusive regimens during the period of observation contributed data to both PI-inclusive and non-PI-inclusive categories of the ARV treatment variable, according to when lipids were measured relative to the date of medication regimen change. Only viral load and CD4+ T lymphocyte count data, used to determine CDC immunologic category, taken within 30 days of the lipid measurement were used; however, most laboratory data were obtained on the same day as the lipid measurement.
Informed consent from parents or legal guardians was obtained, as well as the child's assent when age appropriate. The study was approved by the Institutional Review Boards of participating hospitals as well as the CDC.
The time of the child's first abstracted cholesterol measurement varied between 1997 and 2004 because lipid values were obtained for clinical purposes and enrollment was staggered over a 2-year period. Due to changes in standard of care across the study time frame and variability in clinical care practices between hospitals, some children had more lipid measurements available for analysis than others. Potential univariate associations between mean lipid levels and demographic and clinical characteristics were identified by computing mean values for all available lipid measurements.
HC and HT were longitudinally modeled as binary responses using generalized estimating equations (GEE),25 including factors identified in the univariate analysis as possible covariates. A covariate was included in the GEE model if a Wald χ2 test was statistically significant at α = 0.05. GEE models were constructed for the entire set of lipid measurements, as well as for the subset of measurements taken while on a PI-inclusive regimen. An autoregressive correlation structure was used for all models to adjust for measurements taken over time. This method assumes that repeated observations for the same person will be more correlated the closer they are taken to one another and can effectively include persons with a varying number of unequally spaced observations. Odds ratios (ORs) and 95% confidence intervals (CIs) were calculated for each covariate in the models. All data analyses were performed using SAS version 9.1 (SAS Institute Inc., Cary, NC).
A total of 2694 cholesterol and 2541 triglyceride measurements were recorded for 178 (98%) of 182 study participants. Eighty-three children (47%) had HC at least once during the study period, with a total of 539 measurements ≥200 mg/dL. On average, children had 15 cholesterol measurements (range, 1-63) abstracted from clinical records between 1997 and 2004. Of the 178 children, 69 (39%) were male; 143 (80%) were black, and 17 (10%) were white (Table 1). Overall, 34 (19%) were of Hispanic ethnicity. The mean age of children at the time of their first abstracted cholesterol measurement was 6.3 years (range, 0.5-15). At the time of the first abstracted cholesterol measurement, 33 (19%) had an undetectable viral load and 92 (52%) were CDC immunologic category 1. Nineteen (26%) had a BMI ≥90th percentile and approximately 75% were Tanner I pubertal stage. One hundred (56%) children were receiving PI-inclusive therapy for a mean duration of 15.6 months (range, 1-58) before the first abstracted cholesterol measurement.
From 1997 to 2004, 134 (75%) of 178 children had been on a PI-inclusive HAART regimen. The median duration of the child's first PI-inclusive regimen was 4.8 years (range, 0.2-7.3 years). The most common PIs used were nelfinavir (38%), ritonavir (23%), and lopinavir/ritonavir (20%). NNRTIs were used by 75 (42%) of 178 children during the study time frame; 22 children (12%) were on HAART regimens consisting of NNRTIs and NRTIs only. Among the children on NNRTI-containing regimens, 53% used nevirapine and 42% efavirenz.
As a preliminary step, we examined the association between mean cholesterol and clinical and demographic indicators across all cholesterol measurements (Table 2). These relationships are likely confounded by the number of lipid results available for a child but serve to identify potential covariates for longitudinal models. Measurements from children on PI-inclusive regimens had higher mean cholesterol (175 mg/dL) than from children on no ARV therapy (137 mg/dL), mono (137 mg/dL), dual (147 mg/dL), or HAART with NNRTI but no PIs (149 mg/dL). Higher mean cholesterol values were also found among children younger than 11 years, males, those with undetectable viral load, children classified as immunologic category 1, and children on regimens that included stavudine. Differences in mean cholesterol were minimal for BMI, race, ethnicity, and Tanner staging of breast, pubic hair, and genital development.
HT was more common than HC. One hundred twenty children (67%) met criteria for HT at least once during the study period, where 671 triglyceride measurements out of 2541 total were ≥150 mg/dL. Similar to cholesterol, median triglyceride values were higher among children with ARV regimens that included either PIs or stavudine, age younger than 6 years, those with undetectable viral load, and children classified as immunologic category 1 (Table 2). In addition, measurements from males and children with BMI ≥90th percentile were elevated compared with those from females and children whose BMI was below the 90th percentile.
The final GEE model for HC included time, ARV regimen, CDC immunologic category based on age and CD4+ T lymphocyte count, and viral load as covariates PI-inclusive ARV regimen, undetectable viral load, and category 3 immunosuppression were strong, independent predictors of HC (Table 3). Children on PI-inclusive regimens were over 4 times more likely to have HC than those on ARV regimens that did not include PIs (OR = 4.3; 95% CI: 2.3-8.2; P < 0.001). Undetectable viral load was also associated with an increased risk of HC (OR = 2.5; 95% CI: 1.5-4.1; P < 0.001). Finally, HC was significantly less likely to occur among severely immunosuppressed (immunologic category 3) children compared with immunocompetent children (OR = 0.42; 95% CI: 0.19-0.95; P = 0.04). There were no significant interactions between the covariates in any of the multivariate models. Although current stavudine use was associated with higher mean cholesterol in preliminary analyses, there was no increased risk independent of PI use.
When modeling HT, ARV regimen and BMI were the only significant variables. HT was strongly associated with PI-inclusive regimens (Table 3). In addition, BMI ≥90th percentile more than doubled the risk for HT. Age, viral load, and immunologic category were not associated with HT.
To assess the effect of outlying observations, we conducted sensitivity analyses by excluding the 21 children who had a single elevated cholesterol measurement ≥200 mg/dL and 44 children who had a single elevated triglyceride value ≥150 mg/dL and re-running the respective GEE models. Excluding children with a single elevated lipid value had little effect on the model results, and all reported associations remained significant.
Children on PI-Inclusive Regimens and Risk Factors for HC and HT
For the subset of children on PI-inclusive regimens, the final GEE model for HC included time since initiation of first PI-inclusive regimen, number of PI drugs in the regimen, age,'viral load, and CDC immunologic categories. HC was significantly and independently associated with viral load and'a regimen containing multiple PIs (Table 4). Children with undetectable viral loads were 2.5 times more likely to have HC than children with detectable viral loads (OR = 2.5; 95% CI: 1.6-4.1; P < 0.001). Similar to models using all available measurements, children 11 years and older and those most immunosuppressed (immunologic category 3) were less likely to have HC than immunocompetent children and those in the youngest age category (0-5 years).
BMI >90th percentile and receiving a ritonavir-inclusive regimen were independently associated with an increased risk for HT (Table 4). In contrast to HC, neither viral load nor immunologic status was associated with HT.
Duration of PI Use and Lipid Values
Median cholesterol values plotted by duration of first PI-inclusive regimen in 6 month intervals are shown in Figure 1. This trend is easier to assess visually using median rather than mean values, given that they reduce the effects of outliers. Although 40% of children had changes in their PI-inclusive regimen during the study time frame, we focused on the first PI regimen for simplification. Median cholesterol values increased significantly from pretreatment values (0-6 months before PI use) to median values 6 months after initiating PI-inclusive regimens. Median cholesterol values continued to increase, particularly after the first 6 months of the initial PI-inclusive regimen through 24 months duration. Median cholesterol changed minimally after 24 months on the first PI-inclusive regimen.
There was no relationship between the duration of first PI-inclusive regimen and median triglyceride values.
Hyperlipidemia seems to be consistent over time. Among the 83 children who had HC at least once, 62 (75%) had HC on a subsequent measurement. Similarly, of the 120 children who had HT at least once, 76 (63%) had HT on at least one other occasion.
The prevalence of clinical lipodystrophy in this cohort was 5.6% (10/178). All 10 children were currently receiving or had received in the recent past regimens that included PIs and stavudine. The most common characteristics were as follows: truncal adiposity (n = 3), buffalo hump (n = 2), extremity wasting (n = 1), or "other evidence of lipodystrophy" (n = 4). Of these 10 children, at diagnosis, 4 had cholesterol ≥200 mg/dL, 6 had triglycerides ≥150 mg/dL, and 3 children had both HC and HT.
In our longitudinal study, children on PI-inclusive regimens had a substantially higher risk of both HC and HT than children not receiving PIs. Independent of ARV regimen, children with undetectable viral loads had an increased risk of HC; severely immunosuppressed children were significantly less likely to have HC. Among children on PI-containing regimens, having an undetectable viral load and taking multiple PIs were associated with a higher risk of HC. Severe immunosuppression and age 11 years and older was associated with a lower risk of HC.
Previous pediatric studies have also noted an association between current PI use and higher mean total cholesterol.2-5,8,10,17,29 Recent cross-sectional analyses of large pediatric cohorts in the United States10 and Spain26 have also reported an association between undetectable viral load and HC. A number of pediatric studies2,5,9,26 have shown a strong association between stavudine-inclusive regimens and redistribution of body fat. However, none,2,5,10 including this study, has demonstrated an increased risk of dyslipidemia that is independent of PI use. This may be explained by the fact that most children receiving stavudine-inclusive regimens are also receiving PIs.
The increased risk associated with undetectable viral load as well as a reduced risk for older and severely immunosuppressed children may be best explained by adherence. Whether measured by caregiver s self-report in pediatric studies10 or by pharmacy refill records for HIV-infected adults,27 better adherence is independently associated with HC. In addition, among children, better adherence is directly related to undetectable viral load,28 high CD4+ T lymphocyte count, and younger age.30 These findings suggest that the relationship between undetect viral load and HC may be drug mediated rather than immunologically mediated.
Our study also found that PI-containing regimens and high BMI were independently associated with an increased risk of HT. Among children on PI-inclusive regimens, BMI ≥90th percentile and those whose regimens included ritonavir were at an increased risk for HT. The association between HT and ritonavir alone, or a PI combination containing ritonavir, has been noted both among children5,17 as well as adults.31 The positive correlation between BMI and elevated triglycerides has been reported by studies of cardiovascular disease risk in general pediatric populations.34,35 However, this is the first study to describe an independent association between HT and BMI among HIV-infected children.
Evaluating the cumulative risk of exposure to HAART therapy and dyslipidemia is a pressing concern for clinicians. We found that an increasing linear relationship between median cholesterol and duration of first PI-inclusive regimen did not persist after 24 months on a regimen. Our data show that after 24 months, median cholesterol remains fairly constant. Earlier studies found that cholesterol measurements taken after initiating a PI-inclusive regimen are substantially higher compared with pretreatment levels.17 However, Cheseaux et al17 found no significant increase in either mean cholesterol or triglyceride levels comparing lipid levels after 12 months of PI-inclusive regimen with levels after 24 and 36 months of treatment. The stabilizing relationship with extended durations of PI usage may explain why most cross-sectional studies have failed to detect an association of PI duration and HC or HT.2,10
To our knowledge, this is the first analysis to examine the effect of ARV regimen on lipid values over time. Our approach accounts for changes in medical regimens and includes all lipid measurements taken on each regimen. In addition, the use of longitudinal models allowed us to adjust for confounding due to some children contributing more lipid measurements to the analysis than others (eg, more frequent lipid measurements among children receiving PIs or with a history of HC). The large number of lipid measurements included in this analysis provided sufficient statistical power to identify risk factors for HC and HT among children receiving PI-containing regimens. Previous clinical studies were limited by small sample sizes in evaluating the role of immunologic parameters, specific PI medications, and age as independent risk factors for dyslipidemia.
Nonetheless, the present study has several limitations. First, total cholesterol and triglyceride levels were mostly based on nonfasting tests conducted as part of the child s clinical care. Although nonfasting tests may result in temporary elevated triglyceride levels due to recent consumption of high-fat foods, we do not think this resulted in overestimating the prevalence of hyperlipidemia. Among healthy adults, a comparison of fasting and nonfasting total cholesterol values showed a small, statistically significant, although clinically unimportant, increase in nonfasting values.32 These differences had little effect when classifying patients into "high cholesterol" groups; agreement between fasting and nonfasting blood testing was 90%. Similarly, data from a study3 of HIV-infected children showed similar group mean lipid levels from nonfasting and fasting tests. Here, the use of fasting or nonfasting cholesterol values did not effect the significant relationship between PI use and lipid levels. A second limitation is that not all children in our analysis had pretreatment lipid measurements, thus limiting our ability to conclude that the initiation of the PI-inclusive regimen preceded HC. Missing data may be because clinicians did not start routine monitoring of lipids until 1997, after the introduction of HAART and when an association with elevated lipids was suspected. Another limitation of this study is the lack of an active assessment of adherence. Finally, the substantially lower prevalence of clinical lipodystrophy in our cohort than has been described in other studies5,7,9 may be due to the lack of standardized criteria for evaluating lipodystrophy, which may have resulted in an underestimate of children who were mildly dysmorphic.
By the end of our study in 2004, 75% of the children in our cohort were on regimens that included PIs. We can expect that most these children will remain on such regimens for the foreseeable future. Although it seems that median total cholesterol levels do not linearly increase with longer duration of PI exposure after 24 months on the same regimen, our study does show that clinically significant hyperlipidemia tends to be persistent over time. Therefore, lipid levels of children receiving ARV therapy should be evaluated regularly. For children with atherogenic dyslipidemia, improvements in diet and exercise should be the first intervention as recommended by the National Cholesterol Education Program.23 Lipid-lowering drug therapy should be considered, although the efficacy of pharmacologic treatment of PI-associated lipidemia in a pediatric population is unknown. Furthermore, there is a great need to explore alternative ARV treatment strategies including the use of new PIs such as atazanavir, which seems to have a minimal effect on lipid profiles in children,33 and CD4-guided interruptions to minimize chronic exposure to therapy.
Our results show that children on PI-inclusive HAART regimens, particularly those who respond well to this regimen, as demonstrated by viral suppression and the lowest category of immunosuppression, are at highest risk for hyperlipidemia. Whereas there is legitimate concern about the risk for accelerated atherosclerotic disease among these children, this future risk must be considered in the context of the current beneficial effect of viral suppression as a result of effective PI-inclusive regimens.
The risk for atherosclerotic disease among HIV-infected children, and especially children with PI-associated dyslipidemia, is uncertain. A number of studies among adults have found that thickening of the intima-media layer of the arterial wall, a risk factor for coronary heart disease, is more pronounced in patients receiving PI-containing regimens than in ARV-naive patients and individuals without HIV infection.36,37 We are currently evaluating the effect of PI-inclusive regimens on carotid intima-media thickness in a subset of our cohort to heighten our understanding of the relationship between dyslipidemia in HIV-infected children and subsequent risk for CHD. In addition, more longitudinal analyses of pediatric cohorts should further clarify the cumulative effects of PI over time, particularly for children started on HAART at a young age.
The authors would like to thank the Bronx Lebanon Hospital: Saroj Bakshi, Caroline Nubel, Elizabeth Adams, and Aida Rivas; the Centers for Disease Control and Prevention: Linda Koenig, Darcy Freedman, April Bell, Bob Yang, Shawn Wei, Siva Rangarajan, and Mary Glenn Fowler; the Columbia University Mailman School of Public Health: Louise Kuhn; the Emory University School of Medicine: Vickie Grimes, Francis Lee, Andre Nahmias, Mary Sawyer, Michael Lindsey, and Kevin Sullivan; the Harlem Hospital Center: Susan Champion, Julia Floyd, Cynthia Freeland, Margaret Heagarty, and Pamela Prince; the Jacobi Hospital Center: Jacobo Abadi, Joanna Dobroszycki, Adell Harris, Genevieve Lambert, Michael Rosenberg, and Andrew Wiznia; the Metropolitan Hospital Center: Mahrukh Bamji, Grace Canillas, Lynn Jackson, and Nancy Cruz; the Medical and Health Research Association of New York City, Inc.: Tina Alford, Mary Ann Chiasson, Elisa Rivera, Eileen Rillamas-Sun, Donald Thea, and Jeremy Weedon; the Montefiore Medical Center: Julia Arnsten, Anna Winston, Valerie Nedwin, and Ellie Schoenbaum; the University of Medicine and Dentistry of New Jersey: Linda Bettica, Thomas Denny, Lucia Ejiofor, Susan Abudato, Jennis Hannah, Mary Jo Hoyt, Carmen Laracuente, Wei Lu, Deborah Storm, and Jeffrey Swerdlow; and the University of Maryland School of Medicine: Vickie Tepper, Katie Peery, Susan Hines, Sue Lovelace, Prasanna Nair, and Peter Vink.
1. Gortmaker SL, Hughes M, Cervia J, et al. The impact of protease inhibitor combination therapy on mortality among children and youth infected with HIV-1. N Engl J Med. 2002;345:1522-1528.
2. Bitnun A, Sochett E, Babyn P, et al. Serum lipids, glucose homeostasis and abdominal adipose tissue distribution in protease-inhibitor-treated and naive HIV-infected children. AIDS. 2003;17:1319-1327.
3. Lainka E, Oezbek S, Falck M, et al. Marked dyslipidemia in human immunodeficiency virus-infected children on protease inhibitor-containing antiretroviral therapy. Pediatrics. 2002;110:e56.
4. Melvin AJ, Lennon S, Mohan KM, et al. Metabolic abnormalities in HIV type1-infected children treated and not treated with protease inhibitors. AIDS Res Hum Retrovir. 2001;17:1117-1123.
5. European Paediatric Lipodystrophy Group. Antiretroviral therapy, fat redistribution and hyperlipidemia in HIV-infected children in Europe. AIDS. 2004;18:1443-1451.
6. Vigano A, Mora S, Testolin C, et al. Increased lipodystrophy is associated with increased exposure to highly active antiretroviral therapy in HIV-infected children. J Acquir Immune Defic Syndr. 2003;32:482-489.
7. Jaquet D, Levine M, Ortega-Roriguez E, et al. Clinical and metabolic presentation of the lipodystrophic syndrome in HIV-infected children. AIDS. 2000;14:2123-2128.
8. Amaya R, Kozinetz C, McMeans A, et al. Lipodystrophy syndrome in human immunodeficiency virus-infected children. Pediatr Infect Dis J. 2002;21:405-410.
9. Arpadi SM, Cuff PA, Horlick MN, et al. Lipodystrophy in HIV-infected children is associated with high viral load and low CD4 lymphocyte count and CD4 lymphocyte percentage at baseline and use of protease inhibitors and stavudine. J Acquir Immune Defic Syndr. 2001;27:30-34.
10. Farley J, Gona P, Crain M, et al. Prevalence of elevated cholesterol and associated risk factors among perinatally HIV-infected children (4-19years) in PACTG 219C. J Acquir Immune Defic Syndr. 2005;38: 480-487.
11. Li S, Chen W, Srinivasan SR, et al. Childhood cardiovascular risk factors and carotid vascular changes in adulthood. JAMA. 2003;290:2271-2276.
12. Raitakari OT, Juonala M, Kahonen M, et al. Cardiovascular risk factors in childhood and carotid artery intima-media thickness in adulthood. JAMA. 2003;290:2277-2283.
13. Holmberg S, Moorman A, Williamson J, et al. Protease inhibitors and cardiovascular outcomes in patients with HIV-1. Lancet. 2002;360:1747-1748.
14. Friis-Moller N, Sabin CA, Webber R, et al. Combination antiretroviral therapy and the risk of myocardial infarction. N Engl J Med. 2003;349:1993-2003.
15. El-Sadr W, Reiss P, De Wit S, et al. Relationship between prolonged exposure to combination ART and myocardial infarction: effect of sex, age and lipid changes. Presented at: 12th Conference on Retroviruses and Opportunistic Infections; 2005; Boston.
16. Mary-Krause M, Cotte L, Simon A, et al. Increased risk of myocardial infarction with duration of protease inhibitor therapy in HIV-infected men. AIDS. 2003;17:2479-2486.
17. Cheseaux JJ, Jotterand V, Aebi C, et al. Hyperlipidemia in HIV-infected children treated with protease inhibitors: relevance for cardiovascular diseases. J Acquir Immune Defic Syndr. 2002;30:288-293.
18. Purnell JQ, Zambon A, Knopp RH, et al. Effect of ritonavir on lipids and post-heparin lipase activities in normal subjects. AIDS. 2000;14:51-57.
19. McComsey G, Bhumbra N, Jen-Fu M, et al. Impact of protease inhibitor substitution with efavirenz in HIV-infected children: results of the first pediatric switch study. Pediatrics. 2003;111:e275.
20. Bulterys M, Nesheim S, Abrams E, et al. Lack of evidence of mitochondrial dysfunction in the offspring of HIV infected women: retrospective review of perinatal exposure to antiretroviral drugs in the Perinatal AIDS Collaborative Transmission Study. Ann N Y Acad Sci. 2000;918:212-221.
21. Abrams EJ, Wiener J, Carter R, et al. Maternal heath factors and early pediatric antiretroviral therapy influence the rate of perinatal HIV-1 disease progression in children. AIDS. 2003;17:867-877.
22. Freedman D, Koenig L, Wiener J, et al. Challenges in enrollment of perinatally HIV-infected and HIV-exposed but uninfected children into a perinatal HIV cohort study: Strategies for locating and recruiting hard-to-reach families. Paediatr Perinat Epidemiol. In press.
23. Lauer R, Barness LA, Clark R, et al. Report of the expert panel on blood cholesterol levels in children and adolescents. Pediatrics. 1992;89(S):525-584.
24. Centers for Disease Control and Prevention. 1994 revised classification system for human immunodeficiency virus infection in children less than 13 years of age. MMWR Morb Mortal Wkly Rep. 1994;43(RR-12):1-10.
25. Zeger SL, Liang KY. Longitudinal data analysis for discrete and continuous outcomes. Biometrics. 1986;42:121-130.
26. Ramos JT, Guillen S, Resino R, et al. High prevalence of fat redistribution and hyperlipidemia in a large cohort of HIV-infected children. Presented at: 12th Conference on Retroviruses and Opportunistic Infections; 2005; Boston.
27. Vergis EN, Paterson DL, Wagener MM, et al. Dyslipidaemia in HIV-infected patients: association with adherence to potent antiretroviral therapy. Int J STD AIDS. 2001;12:463-468.
28. Farley J, Hines S, Musk A, et al. Assessment of adherence to antiviral therapy in HIV-infected children using the Medication Event Monitoring System, pharmacy refill, provider assessment, caregiver self-report, and appointment keeping. J Acquir Immune Defic Syndr. 2003;33:211-218.
29. Beregszaszi M, Dollfus C, Levine M, et al. Longitudinal evaluation and risk factors of lipodystrophy and associated metabolic changes in HIV-infected children. J Acquir Immune Defic Syndr. 2005;40:161-168.
30. Mellins CA, Brackis-Cott E, Dlezal C, et al. The role of psychosocial and family factors in adherence to antiretroviral treatment in human immunodeficiency virus-infected children. Pediatr Infect Dis J. 2004;23:1035-1041.
31. Fontas E, van Leth F, Sabin CA, et al. Lipid profiles in HIV-infected patients receiving combination antiretroviral therapy: are different antiretroviral drugs associated with different lipid profiles? J Infect Dis. 2004;189:1056-1074.
32. Craig SR, Amin RV, Russel DW, et al. Blood cholesterol screening: influence of fasting state on cholesterol results and management decisions. J Gen Intern Med. 2000;15:395-399.
33. Rutstein R, Samson P, Aldrovandi G, et al. Effect of atazanavir on serum cholesterol and triglyceride levels in HIV-infected infants, children and adolescents: pediatric AIDS Clinical Trials Group (PACTG) 1020A. Presented at: 12th Conference on Retroviruses and Opportunistic Infections; 2005; Boston.
34. Geiss HC, Parhofer KG, Schwandt P, et al. Parameters of childhood obesity and their relationship to cardiovascular risk factors in healthy prepubescent children. Int J Obes Relat Metab Disord. 2001;25:830-837.
35. Flodmark CE, Sveger T, Nilsson-Ehle P, et al. Waist measurement correlates to a potentially atherogenic lipoprotein profile in obese 12-14-year-old children. Acta Paediatr. 1994;83:941-945.
36. Mercie P, Thiebaut R, Lavignolle V, et al. Evaluation of cardiovascular risk factors in HIV-1 infected patients using carotid intima-media thickness measurement. Ann Int Med. 2002;34:55-63.
37. Seminari E, Pan A, Voltini G, et al. Assessment of atherosclerosis using carotid ultrasonography in a cohort of HIV-positive patients treated with protease inhibitors. Atherosclerosis. 2002;162:433-438.
© 2006 Lippincott Williams & Wilkins, Inc.