点击显示 收起
Infectious Diseases Service, University Hospital, Institute of Microbiology, University of Lausanne, Swiss HIV Cohort Study Data Center
Lipidology, University Medical Policlinic, Lausanne
Infectious Diseases, University Hospital, Bern, Infectious Diseases, University Hospital, Zurich, Infectious Diseases, Kantonsspital, St. Gallen, Infectious Diseases, University Hospital, Geneva, Infectious Diseases, University Hospital, Basel
Ospedale Civico, Lugano, Switzerland
Background.
Single-nucleotide polymorphisms in genes involved in lipoprotein and adipocyte metabolism may explain why dyslipidemia and lipoatrophy occur in some but not all antiretroviral therapy (ART)treated individuals.
Methods.
We evaluated the contribution of APOC3 -482CT, -455TC, and 3238CG; 2 and 4 alleles of APOE; and TNF -238GA to dyslipidemia and lipoatrophy by longitudinally modeling >2600 lipid determinations and 2328 lipoatrophy assessments in 329 ART-treated patients during a median follow-up period of 3.4 years.
Results.
In human immunodeficiency virus (HIV)infected individuals, the effects of variant alleles of APOE on plasma cholesterol and triglyceride levels and of APOC3 on plasma triglyceride levels were comparable to those reported in the general population. However, when treated with ritonavir, individuals with unfavorable genotypes of APOC3 or APOE were at risk of extreme hypertriglyceridemia. They had median plasma triglyceride levels of 7.33 mmol/L, compared with 3.08 mmol/L in the absence of ART. The net effect of the APOE*APOC3*ritonavir interaction was an increase in plasma triglyceride levels of 2.23 mmol/L. No association between TNF -238GA and lipoatrophy was observed.
Conclusions.
Variant alleles of APOE and APOC3 contribute to an unfavorable lipid profile in patients with HIV. Interactions between genotypes and ART can lead to severe hyperlipidemia. Genetic analysis may identify patients at high risk for severe ritonavir-associated hypertriglyceridemia.
Adverse metabolic effects of potent antiretroviral therapy (ART) have become a major concern. Hyperlipidemia increases the risk of cardiovascular disease [1]. Lipoatrophy may stigmatize the patient and is associated with insulin resistance and hyperlipidemia [2, 3]. Both hyperlipidemia and lipoatrophy have been linked to advancing age, male sex, a CD4 T cell count <200 cells/L, and the type of ART used [2,4]. The use of ritonavir (RTV) and RTV-boosted protease inhibitors (PIs) are the most significant predictors of hypercholesterolemia and hypertriglyceridemia [4], whereas use of stavudine (d4T), didanosine (ddI), and, in some studies, PIs are associated with lipoatrophy [2]. However, these complications do not occur in all patients, despite similar exposure to ART and comparable demographic, immunologic, and virologic characteristics. The reasons for this discrepancy may be related to host genetic factors.
Single-nucleotide polymorphisms (SNPs) in APOE (specifically the APOE 4 and 2 alleles) are linked to hyperlipidemia and cardiovascular events [5] in the general population. SNPs of APOC3 at positions -455 and -482 in the promoter region and at position 3238 in the 3 untranslated region (the SstI allele) are associated with hypertriglyceridemia [6, 7]. A SNP at position -238 in the promoter region of TNF (previous designation, TNF-) has been linked to the more-rapid onset of lipoatrophy in ART-treated patients [8], and the rare allele was found to be present more frequently in patients with HIV who have lipoatrophy than in those who do not have the condition [9]. The APOE genotype may be functionally associated with hyperlipidemia, as the role of APOE in the transport and clearance of lipoprotein remnants from the bloodstream would suggest [5]. Through the inhibition of the activity of lipoprotein lipase, APOC3 modulates lipolysis and hepatic clearance of plasma triglycerides (TGs) [6, 7]. TNF expression in adipose tissue has been implicated in lipoatrophy because of its modulation of lipolysis and adipocyte differentiation [10]. Most reports on the metabolic complications associated with APOE and APOC3 polymorphisms have been based on studies of the general population. In HIV-infected individuals, the interpretation of the data has been limited by the small number of patients studied [11, 12], contradictory findings regarding a link between APOE genotype and hyperlipidemia [1214], and a focus on PI-treated men [1315]. The present study used a longitudinal approach to model the influence of APOE and APOC3 polymorphisms on ART-associated hyperlipidemia, as well as the influence of the TNF -238 polymorphism on lipoatrophy, in the context of frequent changes of therapy and other contributing factors.
PATIENTS AND METHODS
Patients.
Study participants were followed in the Swiss HIV Cohort Study (SHCS) (http://www.shcs.ch) and were treated for HIV infection at 1 of 7 Swiss medical centers. The genetics project of the SHCS was approved by the ethics committees of all participating centers. Patients were included if they had given written, informed consent for genetic testing.
Lipid determinations included the routine assessments of plasma total cholesterol, high-density lipoprotein (HDL) cholesterol, and TG levels. The level of non-HDL cholesterol (NHC) was calculated by subtracting the level of HDL cholesterol from the level of total cholesterol. Patients who had a Centers for Disease Control and Prevention stage C AIDS-defining event in the preceding 3 months or a CD4 T cell count <100 cells/L were excluded, to limit confounding by HIV-associated wasting and opportunistic illness. Lipoatrophy was defined by the patient-reported presence of fat loss at 1 body site (face, neck, arms, legs, or buttocks) that was confirmed by the physician on physical examination. No lipoatrophy assessments were made by use of computed tomography scanning or dual-energy x-ray absorptiometry. The fasting state (with fasting defined as 8 h without caloric intake) was recorded each time that a blood sample was taken.
Allelic discrimination.
Genomic DNA was extracted from frozen leukocyte pellets. APOC3 promoter -482CT [7], APOC3 3UTR 3238CG [7], APOE coding region 2060TC [5], and TNF promoter -238GA [8, 9] were investigated by TaqMan allelic discrimination (Applied Biosystems) (table 1). The polymorphisms of APOC3 -455TC, APOE 2060TC, and APOE 2198CT were assessed by PCR amplification and restriction fragmentlength polymorphism (RFLP) analysis using either FokI [13, 16] or HhaI [17] digestion. All equivocal results were repeated and confirmed by a second observer. Results were entered into the central SHCS database in a blinded fashion.
For statistical analysis, all variant alleles were assessed individually, and, for the purpose of clarity, were reported using previously reported composite genotypes [13, 18]. For analysis of plasma TG levels [18], we assessed (1) the presence of APOE 3/3 versus 2-containing APOE genotypes (i.e., 2/2, 2/3, and 2/4) versus APOE 3/4 and APOE 4/4, as well as (2) APOE 3/3 versus all alleles other than 3/3. For the analysis of plasma cholesterol levels [18], we assessed APOE 3/3 versus 2-containing APOE genotypes (i.e., 2/2, 2/3, and 2/4) versus APOE 3/4 and APOE 4/4. For APOC3, the cumulative number of variant alleles was used, as reported elsewhere [13], and thus we assessed common alleles, 1 or 2 variant alleles, and all 3 variant alleles.
Statistical analysis.
The data were analyzed longitudinally by modeling the individual effects of the different covariables on plasma lipid levels and on body fat loss. The study period was defined as the follow-up period beginning after the first study visit that included measurement of plasma lipid levels (available since 1999) until 22 March 2004, when information from the SHCS database was downloaded. The analysis used a well-established statistical method developed for longitudinal data analysis, population-averaged marginal modeling [19, 20]. A multivariate distribution was fitted to the data by use of score-like methods (generalized estimating equations) [19, 20]. The advantages of this approach include the ability to accommodate repeated measurements at irregular intervals, the straightforward interpretation of regression coefficients, and the computational efficiency for relatively large data sets. Covariables included in the regression model were ART regimen, age, sex, body mass index (BMI), waist circumference, fasting state, presumed mode of HIV transmission, ethnicity, diabetes mellitus status, treatment with lipid-lowering agents, smoking status, CD4 T cell count, HIV load, and the genetic factors tested. The influence of ART and of lipid-lowering agents on plasma lipid levels was assumed to be rapid and reversible [21]. For evaluation of the effects of the genetic variants on lipoatrophy, the cumulative exposure to ART was used, as was reported elsewhere [2, 22].
Exploratory analyses confirmed the published effects of individual ART agents on plasma HDL cholesterol, NHC, and TG levels and on lipoatrophy [3, 4, 23]. These data supported the rationale for dividing the ART regimens into the following groups: (1) no ART at the time that a blood sample was taken, (2) PI-sparing ART, (3) single-PI ART, and (4) RTV-containing ART. For lipoatrophy analysis, the ART regimens were divided into the following groups [2, 22]: (1) ART that did not include ddI, d4T, or a PI; (2) ART that included ddI and/or d4T but no PI; (3) ART that included a PI but not ddI or d4T; and (4) ART that included ddI and/or d4T and a PI.
When the date of the assessment of plasma lipid levels did not coincide with that of the assessment of waist circumference and weight, the values of these latter covariables were imputed by linear interpolation at the date of the assessment of plasma lipid levels. For appropriate confounding adjustment, cross-sectional and longitudinal relationships between the continuous explanatory variables and the response were distinguished [20, 24, 25]. Adequate functional form was assessed using fractional polynomials [26, 27]. The correlation structure was assumed to be autoregressive [19], because the clustering was at the individual level (repeated measurements). However, when other (e.g., exchangeable) correlation structures were used, the results were similar. Goodness of fit was evaluated visually by comparing mean observed and predicted measurements on time plots of values in strata defined by polymorphisms and ART group [20].
For analysis of lipoatrophy, the dependent variable was binary, and the logit link was used. The correlation structure was assumed to be exchangeable, because exploratory analyses indicated that lipoatrophy was rarely reversible over the follow-up period. A robust sandwich estimate of variances-covariances was used, which produces valid SEs even if the correlation structure was misspecified [19, 20].
Two-way as well as 3-way interactions between genotypes and ART groups were first tested globally. If the global test was significant, subcategories contributing to the P value were screened for, and significant and consistent strata were retained. The regrouping of subcategories into strata was performed using the global 2 test, to address the issue of multiple comparisons.
Patients who had elevated plasma HDL cholesterol, NHC, and TG levels were matched by age (±5 years), BMI (±3 units), sex, and ART group to patients who had normal values for these lipid parameters, and a single-time-point, cross-sectional analysis was performed. All statistical analyses were conducted using Stata (version 8.2; StataCorp).
RESULTS
Patients.
The clinical characteristics of the study patients (n = 329) are summarized in table 2. We compared their characteristics with those of the entire population of patients enrolled in the SHCS who had 1 follow-up visit during 19992003 (n = 6724). Compared with this population of the SHCS, the study patients were more likely to have suppressed HIV viremia, the duration of their follow-up period was longer, and there were fewer deaths and dropouts. These characteristics define the selection bias of patients in the SHCS Genetic Cohort, who were chosen for the quality of their clinical documentation and their ability to sign the consent form for genetic analysis.
The allelic frequencies were as follows: APOC3 -455TC, 0.410; APOC3 -482CT, 0.303; APOC3 3238CG, 0.092; APOE 2060TC, 0.093; APOE 2198CT, 0.055; and TNF 238GA, 0.035. These were comparable to the published allelic frequencies in ethnically similar populations [28]. The distribution of the genotypes was as follows: APOE 3/3, 238 patients (72.3%); APOE genotypes other than 3/3, 91 patients (27.7%; 2/2, 2/3, or 2/4, 35 patients [10.6%]; 3/4 or 4/4, 56 patients [17.0%]); common alleles of APOC3, 120 patients (36.5%); 1 or 2 variant alleles of APOC3, 150 patients (45.6%); and all 3 variant alleles of APOC3, 59 patients (17.9%).
During the follow-up period, a total of 2608 determinations of plasma cholesterol (both HDL and NHC) levels (mean, 7.4 determinations of plasma cholesterol levels/patient) and 2741 determinations of plasma TG levels (mean, 7.8 determinations of plasma TG levels/patient) were recorded (range, 318 total lipid determinations/patient). The mean number of determinations of plasma TG levels during the follow-up period was comparable in subjects with low and high median plasma TG levels (patients with low median plasma TG levels, 7.97 determinations/patient; patients with high median plasma TG levels, 8.25 determinations/patient); the median number of determinations was identical (7 determinations/patient). Thus, potential bias due to overrepresentation of subjects with high plasma TG levels is unlikely. The median plasma lipid levels during the follow-up period are shown in table 2. Results were unchanged when plasma lipid levels in patients who had an AIDS-defining event in the preceding 3 months or a CD4 T cell count <100 cells/L were included.
Effects of ART regimen on plasma lipid levels.
There were a total of 9169 determinations of the type of ART regimen in use during the study period. No ART was in use at 1078 (11.8%) time points. PI-sparing ART, single-PI ART, and RTV-containing ART were in use at 2965 (32.3%), 3242 (35.4%), and 1884 (20.5%) time points, respectively. The influence of ART regimen on plasma lipid levels is summarized in table 3. The population-averaged plasma HDL cholesterol level was higher in patients treated with PI-sparing ART (P = .006). The plasma NHC level was lower in patients treated with PI-sparing ART (P = .019) and higher in patients treated with single-PI ART (P = .01) or RTV-containing ART (P < .001). Plasma TG levels were significantly higher only in patients treated with RTV-containing ART (P < .001). Because RTV can be used at different dosages, we compared plasma TG levels in patients treated with RTV alone (600 mg of RTV twice daily) or in combination with lopinavir (100 mg of RTV twice daily) or saquinavir (400 mg of RTV twice daily). The median plasma TG levels during the follow-up period in these 3 groups of patients were comparable (3.19 vs. 3.00 vs. 2.89 mmol/L, respectively; P = .63). Thus, patients treated with RTV were analyzed as a single group. As was reported elsewhere [29], there was a trend toward higher mean plasma TG levels in patients treated with PI-sparing ART that contained efavirenz (0.47 mmol/L [95% confidence interval {CI}, -0.08 to 1.02 mmol/L] increase in plasma TG levels; P = .09).
Effects of variant alleles on plasma TG levels.
APOE genotypes other than the common 3/3 genotype were associated with higher mean plasma TG levels (P < .001) (table 3). The increase in plasma TG levels was 0.9 mmol/L (95% CI, 0.431.36 mmol/L) (P < .001) for APOE genotypes 2/2, 3/2, and 4/2 and was 0.61 mmol/L (95% CI, 0.231.00 mmol/L) (P = .002) for APOE genotypes 3/4 and 4/4. Thus, as regards plasma TG levels, APOE genotypes other than 3/3 were analyzed as a single group. Plasma TG levels were significantly elevated in patients who had all 3 variant alleles of APOC3 (P = .006) but not in patients who had only 1 or 2 variant alleles of APOC3 (P = .83).
The most unfavorable composite genotype consisted of APOE genotypes other than 3/3 plus all 3 variant alleles of APOC3 (table 4). The predicted net effect of the APOE*APOC3 interaction was a mean 1.24 mmol/L (95% CI, 0.452.03 mmol/L) increase in plasma TG levels (P = .02). This unfavorable composite genotype was associated with elevations of plasma TG levels in patients in all ART groups; however, the most striking genotype-ART interaction was observed when patients who had the unfavorable composite genotype were treated with RTV-containing ART (figure 1). Here, the predicted net effect of the APOE*APOC3*RTV interaction was a mean 2.23 mmol/L (95% CI, 1.333.14 mmol/L) increase in plasma TG levels (P < .001). These estimates represent the contribution of the interactions in addition to the specific effects of the covariables listed in table 3. Thus, the mean plasma TG level in patients who had the unfavorable composite genotype was 3.08 mmol/L if no ART was used and increased to 7.33 mmol/L when RTV was used in the regimen (P < .001) (figure 1).
When patients were exposed to RTV, extreme hypertriglyceridemia (defined as at least 2 determinations at which plasma TG levels were above the 95th percentile of the study population [7.0 mmol/L]) was recorded in 4 (15.4%) of 26 patients who had common alleles of APOE and APOC3, in 3 (20.0%) of 15 patients who had all 3 variant alleles of APOC3 and common alleles of APOE, in 1 (11.1%) of 9 patients who had APOE non-3/3 genotypes and common alleles of APOC3, but in 6 (60.0%) of 10 patients who had the unfavorable composite genotype (P = .026). The TNF -238GA polymorphism had no effect on plasma TG levels (P = .973).
An additional, single-time-point, cross-sectional analysis of 75 patients who had plasma TG levels 2.3 mmol/L and 75 control subjects who had normal plasma TG levels and were matched to patients by age, BMI, sex, and ART group was underpowered for detection of the deleterious effects of variant alleles of APOC3 or APOE. For example, individuals who had APOE non-3/3 genotypes had an odds ratio of 1.41 (95% CI, 0.653.06) for the development of hypertriglyceridemia (P = .379).
Effects of variant alleles on plasma NHC and HDL cholesterol levels.
The 3/4 and 4/4 APOE genotypes were associated with elevated plasma NHC levels (P = .038) (table 3). As expected, the APOE genotype had no influence on plasma HDL cholesterol levels. There was a trend toward lower plasma NHC levels in patients who had 1 or 2 variant alleles of APOC3 (P = .088) and in patients who had all 3 variant alleles (P = .052). No significant interaction between various genotypes and ART group were identified (table 4). No effect of the TNF -238GA polymorphism on either plasma HDL cholesterol levels or plasma NHC levels was seen (P = .592 and P = .881, respectively).
Lipid-lowering agents were given more frequently to patients who had APOE genotypes other than 3/3 (12.2% of determinations) and to patients who had variant alleles of APOC3 (10.8% of determinations) than to patients who had the APOE 3/3 genotype (8.7% of determinations) (P = .007) or to patients who had the common alleles of APOC3 (7.6%) (P = .023); the models were adjusted for treatment with lipid-lowering agents.
Lipoatrophy.
A total of 2328 assessments of lipoatrophy (mean, 7.1 assessments/patient; range, 39 assessments/patient) were made during the follow-up period in 325 patients. The prevalence of lipoatrophy was 25.2%. Baseline age (P = .004), BMI (P = .016), and waist circumference (P = .004) were associated with lipoatrophy (table 5). Lipoatrophy was associated with cumulative exposure to ddI/d4T, irrespective of the associated use of a PI (P < .001). Exposure to a PI in the absence of ddI and/or d4T was not associated with lipoatrophy (P = .153). The odds ratio for the development of lipoatrophy was higher in patients who had 3 variant alleles of APOC3 (P = .046) but was not higher in patients who had the TNF -238GA polymorphism (P = .625).
DISCUSSION
We report on the identification of a significant genotype-ART interaction that may lead to severe hyperlipidemia in HIV-infected patients treated with ART. Known genetic variants in APOC3 and APOE contributed significantly to elevations in plasma TG levels to a degree comparable to factors such as age or the use of RTV-containing ART. However, the effect of unfavorable APOC3 alleles was observed only in patients who received ART. Variant alleles of APOC3 and APOE had no significant effect on plasma HDL cholesterol levels, whereas the effect of variant alleles of APOE on plasma NHC levels was additive to that of single-PI ART. Our results extend the findings reported by Fauvel et al. [13], who documented an association between variant alleles of APOC3 and hypertriglyceridemia in 60 men treated with PIs.
The most relevant finding of our study is the identification of a complex genotype-ART interaction that may lead to extreme hypertriglyceridemia when individuals with an unfavorable composite APOC3-APOE genotype are exposed to RTV. Such a phenotypic expression of a latent genetic predisposition to hyperlipidemia has been reported by Rodondi et al. [18]. In this pharmacogenetic study, these authors identified an elevated risk of future hyperlipidemia in APOE 2 and 4 carriers who developed hypertriglyceridemia during retinoic acid treatment for acne. Conversely, patients who did not have elevations in their plasma TG levels during retinoic acid therapy seemed to be protected against future hyperlipidemia.
We did not identify a genetic predisposition to lipoatrophy in patients in our data set who had the TNF -238GA polymorphism. This is in contrast to the findings of 2 other studies [8, 9]. Nolan et al. [8] used a time-to-event model to evaluate the occurrence of lipoatrophy in 191 white men. The frequency of the rare allele (f = 0.065) was higher than that in the present study (f = 0.035). After 2 years of follow-up, the effect of the TNF polymorphism was modest in the study by Nolan et al., but the effect became more pronounced over time. It is possible that we might identify a significant effect of TNF -238GA with an extended follow-up period. The greater diagnostic accuracy in the study by Nolan et al. [8] (by the use of radiological assessments of body composition) may further explain the differences between the results of the studies. Although there was an association between lipoatrophy and variant alleles of APOC3, the biological mechanism is unclear, and this association needs to be confirmed in other patient populations before it can be considered a true finding.
Longitudinal modeling represents a powerful approach to quantitating the individual contribution of genetic factors and effects of ART in the context of complex traits. We exploited prospectively collected data consisting of >2600 lipid determinations in 329 individuals, during a study period of >3 years. This allowed us to evaluate the influence of the polymorphisms in APOC3 and APOE as well as multiple intervening factors of relevance to plasma lipid levels (age, sex, ethnicity, smoking status, mode of HIV transmission, ART group, CD4 T cell count, HIV viremia, BMI, fasting state, treatment with lipid-lowering agents, and diabetes mellitus status). The study patients served as their own controls, through multiple changes in ART and periods when ART was not used. This approach was selected in response to the well-recognized dilemmas in genetic association studies, which regard, particularly, the issues of statistical power, phenotype definition, and the use of appropriate statistical tools. We also attempted a conventional matched control analysis using a single time point, which proved underpowered for identification of the genetic effects. Potential bias arising from overrepresentation of subjects with high plasma lipid levels seems unlikely, because we found that the frequency of determinations was comparable in subjects with low and high plasma lipid levels.
The relatively high frequency of variant alleles of APOE and APOC3 might suggest a potential role for genetic testing before initiation of ART. This merits formal evaluation in a randomized trial. For example, 27.7% of patients in this study were carriers of variant alleles of APOE, 17.9% were carriers of all 3 variant alleles of APOC3, and 5.8% were carriers of both. The genotyping of treatment-naive patients might be an efficient method to determine the advisability of the administration of RTV-containing ART. Given the risk of hypertriglyceridemia, it might be prudent to select an alternative ART regimen for patients who have unfavorable genotypes.
SWISS HIV COHORT STUDY
The members of the Swiss HIV Cohort Study are M. Battegay, E. Bernasconi, J. Bni, H. Bucher, P. Bürgisser, S. Cattacin, M. Cavassini, R. Dubs, M. Egger, L. Elzi, P. Erb, K. Fantelli, M. Fischer, M. Flepp, A. Fontana, P. Francioli (president), H. Furrer (chairman of the Clinical and Laboratory Committee), M. Gorgievski, H. Günthard, B. Hirschel, L. Kaiser, C. Kind, T. Klimkait, U. Lauper, B. Ledergerber, M. Opravil, F. Paccaud, G. Pantaleo, L. Perrin, J.-C. Piffaretti, M. Rickenbach (head of the Data Center), C. Rudin (chairman of the Mother and Child Substudy), P. Schmid, J. Schüpbach, R. Speck, A. Telenti, A. Trkola, P. Vernazza (chairman of the Scientific Board), R. Weber, and S. Yerly.
References
1. Friis-Moller N, Sabin CA, Weber R, et al. Combination antiretroviral therapy and the risk of myocardial infarction. N Engl J Med 2003; 349:19932003. First citation in article
2. Bernasconi E, Boubaker K, Junghans C, et al. Abnormalities of body fat distribution in HIV-infected persons treated with antiretroviral drugs: the Swiss HIV Cohort Study. J Acquir Immune Defic Syndr 2002; 31:505. First citation in article
3. Carr A, Samaras K, Thorisdottir A, Kaufmann GR, Chisholm DJ, Cooper DA. Diagnosis, prediction, and natural course of HIV-1 protease-inhibitorassociated lipodystrophy, hyperlipidaemia, and diabetes mellitus: a cohort study. Lancet 1999; 353:20939. First citation in article
4. 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; 189:105674. First citation in article
5. Mahley RW, Rall SC Jr. Apolipoprotein E: far more than a lipid transport protein. Annu Rev Genomics Hum Genet 2000; 1:50737. First citation in article
6. Li WW, Dammerman MM, Smith JD, Metzger S, Breslow JL, Leff T. Common genetic variation in the promoter of the human apo CIII gene abolishes regulation by insulin and may contribute to hypertriglyceridemia. J Clin Invest 1995; 96:26015. First citation in article
7. Dammerman M, Sandkuijl LA, Halaas JL, Chung W, Breslow JL. An apolipoprotein CIII haplotype protective against hypertriglyceridemia is specified by promoter and 3 untranslated region polymorphisms. Proc Natl Acad Sci USA 1993; 90:45626. First citation in article
8. Nolan D, Moore C, Castley A, et al. Tumour necrosis factor gene -238G/A promoter polymorphism associated with a more rapid onset of lipodystrophy. AIDS 2003; 17:1213. First citation in article
9. Maher B, Alfirevic A, Vilar FJ, Wilkins EG, Park BK, Pirmohamed M. TNF- promoter region gene polymorphisms in HIV-positive patients with lipodystrophy. AIDS 2002; 16:20138. First citation in article
10. Bastard JP, Caron M, Vidal H, et al. Association between altered expression of adipogenic factor SREBP1 in lipoatrophic adipose tissue from HIV-1infected patients and abnormal adipocyte differentiation and insulin resistance. Lancet 2002; 359:102631. First citation in article
11. Lister RK, Youle M, Nair DR, Winder AF, Rustin MH. Latent dysbetalipoproteinaemia precipitated by HIV-protease inhibitors. Lancet 1999; 353:1678. First citation in article
12. Grunfeld C, Doerrler W, Pang M, Jensen P, Weisgraber KH, Feingold KR. Abnormalities of apolipoprotein E in the acquired immunodeficiency syndrome. J Clin Endocrinol Metab 1997; 82:373440. First citation in article
13. Fauvel J, Bonnet E, Ruidavets JB, et al. An interaction between apo C-III variants and protease inhibitors contributes to high triglyceride/low HDL levels in treated HIV patients. AIDS 2001; 15:2397406. First citation in article
14. Behrens G, Schmidt HH, Stoll M, Schmidt RE. ApoE genotype and protease-inhibitorassociated hyperlipidaemia. Lancet 1999; 354:76. First citation in article
15. Bonnet E, Ruidavets JB, Tuech J, et al. Apoprotein c-III and E-containing lipoparticles are markedly increased in HIV-infected patients treated with protease inhibitors: association with the development of lipodystrophy. J Clin Endocrinol Metab 2001; 86:296302. First citation in article
16. Groenendijk M, Cantor RM, Blom NH, Rotter JI, de Bruin TW, Dallinga-Thie GM. Association of plasma lipids and apolipoproteins with the insulin response element in the apoC-III promoter region in familial combined hyperlipidemia. J Lipid Res 1999; 40:103644. First citation in article
17. Hixson JE, Vernier DT. Restriction isotyping of human apolipoprotein E by gene amplification and cleavage with HhaI. J Lipid Res 1990; 31:5458. First citation in article
18. Rodondi N, Darioli R, Ramelet AA, et al. High risk for hyperlipidemia and the metabolic syndrome after an episode of hypertriglyceridemia during 13-cis retinoic acid therapy for acne: a pharmacogenetic study. Ann Intern Med 2002; 136:5829. First citation in article
19. Zeger SL, Liang KY. Longitudinal data analysis for discrete and continuous outcomes. Biometrics 1986; 42:12130. First citation in article
20. Diggle PJ. Analysis of longitudinal data. Oxford: Oxford University Press, 2002. First citation in article
21. Purnell JQ, Zambon A, Knopp RH, et al. Effect of ritonavir on lipids and post-heparin lipase activities in normal subjects. AIDS 2000; 14:517. First citation in article
22. Lichtenstein KA, Ward DJ, Moorman AC, et al. Clinical assessment of HIV-associated lipodystrophy in an ambulatory population. AIDS 2001; 15:138998. First citation in article
23. Periard D, Telenti A, Sudre P, et al. Atherogenic dyslipidemia in HIV-infected individuals treated with protease inhibitors. The Swiss HIV Cohort Study. Circulation 1999; 100:7005. First citation in article
24. Ware JH, Dockery DW, Louis TA, Xu XP, Ferris BG Jr, Speizer FE. Longitudinal and cross-sectional estimates of pulmonary function decline in never-smoking adults. Am J Epidemiol 1990; 132:685700. First citation in article
25. Ware JH. Linear models for the analysis of longitudinal studies. Am Stat 1985; 39:95101. First citation in article
26. Royston P, Ambler G, Sauerbrei W. The use of fractional polynomials to model continuous risk variables in epidemiology. Int J Epidemiol 1999; 28:96474. First citation in article
27. Royston P, Altman DG. Regression using fractional polynomials of continuous covariates. Appl Stat 1994; 43:42967. First citation in article
28. Battino M, Giunta S, Galeazzi L, et al. Coenzyme Q10, antioxidant status and ApoE isoforms. Biofactors 2003; 18:299305. First citation in article
29. Doser N, Sudre P, Telenti A, et al. Persistent dyslipidemia in HIV-infected individuals switched from a protease inhibitorcontaining to an efavirenz-containing regimen. J Acquir Immune Defic Syndr 2001; 26:38990. First citation in article