Changes in Lipid Profile Over 24 Months Among Adults on First-Line Highly Active Antiretroviral Therapy in the Home-Based AIDS Care Program in Rural Uganda

Background: Use of highly active antiretroviral therapy (HAART) has been linked to dyslipidemia and increased risk of cardiovascular disease (CVD) in HIV-infected patients in industrialized countries. The effects of HAART on lipid metabolism among sub-Saharan Africans, for whom access to antiretroviral therapy is expanding, remain largely unknown.

Methods: From July 2003 to May 2004, 987 antiretroviral-naive patients with symptomatic HIV disease or a CD4 count <250 cells/mm³ were started on HAART in the Home-Based AIDS Care (HBAC) Program in Tororo, Uganda. The HBAC Program provided weekly drug delivery and field-based clinical monitoring. Nonfasting repository sera from a subset of 374 patients were analyzed for levels of total cholesterol (TC), direct low-density lipoprotein cholesterol (LDL-c), direct high-density lipoprotein cholesterol (HDL-c), and triglycerides (TG) at baseline (before HAART) and after 12 and 24 months of HAART using Randox enzymatic kits (Crumlin, United Kingdom).

Results: The 374 patients evaluated (49% women, mean age = 39 years, CD4 count = 124 cells/mm³, body mass index = 19.7 kg/m²) received initial HAART composed of stavudine, lamivudine, and either nevirapine (365 patients [98%]) or efavirenz (9 patients [2%]). During 24 months, 99 (26%) patients had single drug substitutions from stavudine to zidovudine and 27 (7%) had single drug substitutions from nevirapine to efavirenz. At baseline, the mean serum lipid concentrations were 120 mg/dL for TC, 53 mg/dL for LDL-c, 29 mg/dL for HDL-c, and 123 mg/dL for TG; values were generally comparable for men and women. During 24 months of treatment, TC increased by a mean of 31 mg/dL, LDL-c by a mean of 26 mg/dL, and HDL-c by a mean of 19 mg/dL, whereas the TC/HDL-c ratio decreased from a mean of 4.6 to 3.4 (all changes, P < 0.001). TG levels initially decreased and then returned to baseline levels by 24 months. At baseline and 24 months, respectively, TC was ≥200 mg/dL for 2% and 10% of patients, LDL-c was ≥130 mg/dL for 1% and 6%, HDL-c was <40 mg/dL for 88% and 41%, and TG were ≥150 mg/dL for 23% and 20%.

Conclusions: Rural Ugandans with advanced HIV disease initiating nevirapine- or efavirenz-based HAART experienced infrequent elevations in TC, LDL-c, and TG at baseline and after 24 months of therapy. Increases in HDL-c levels were substantial and proportionally greater than increases in TC or LDL-c levels. The risk of CVD and how it is affected by lipid changes in this rural African population are unknown. However, the changes we observed after 24 months of HAART seem unlikely to increase the risk of CVD.

Rapid expansion of highly active antiretroviral therapy (HAART) programs in resource-constrained settings, such as sub-Saharan Africa, has generated the need for data on HAART-associated adverse event rates in the local patient populations. Dyslipidemia is a well-recognized complication of HAART that has been linked to increased risk for cardiovascular morbidity in HIV-infected patients.^1,2 Studies in industrialized countries have shown that certain antiretrovirals, particularly stavudine and protease inhibitors (PIs), can increase levels of total cholesterol (TC), low-density lipoprotein cholesterol (LDL-c), and triglycerides (TG), with less or no effect on levels of high-density lipoprotein cholesterol (HDL-c),^3-6 changes that increase the risk for cardiovascular disease (CVD).^1,2,7-9 By contrast, HAART regimens containing nevirapine or efavirenz have been associated with less atherogenic lipid profiles in the industrialized countries.^5,10 First-line regimens recommended by the World Health Organization (WHO) include stavudine or zidovudine combined with lamivudine and either nevirapine or efavirenz,¹¹ but data on the effects of these regimens on lipid levels from populations in more resource-constrained settings are scarce.^12-14 HIV-infected patients who initiate HAART in resource-constrained settings may experience different rates of lipid abnormalities than patients in industrialized countries, because they are more likely to have advanced HIV disease and poor nutritional status and to begin treatment with PI-sparing regimens. A small study from India and a substudy of data from the larger multicontinental 2NN study found that treatment with stavudine, lamivudine, and nevirapine was associated with substantial increases in TC and TG in patients treated on average for 18 to 20 months, with close to 40% of Indian patients achieving a TC level ≥200 mg/dL.^12-14 We studied lipid levels before initiation of antiretroviral therapy (ART) and after 12 and 24 months on WHO-recommended first-line HAART regimens in a subset of participants from the Home-Based AIDS Care (HBAC) Program in rural Uganda.

METHODS

HBAC Program Protocol

The HBAC Program is based in the Tororo and Busia districts of eastern Uganda, a rural area populated predominantly by subsistence farmers clustered in small villages without access to a municipal water supply.¹⁵ The local diet consists primarily of plantains, maize, beans, peas, and cassava, supplemented occasionally with meat and fish.¹⁶ Persons eligible to be screened for participation in the program were clients of The AIDS Support Organization (TASO), an indigenous nongovernmental organization that has provided HIV/AIDS care and support services in Uganda since 1987. Entry criteria for HIV-infected persons to the ART program were a CD4 count <250 cells/mm³; severe HIV disease, defined as WHO stage 3 or 4 disease; or a history of recurrent herpes zoster. The HBAC Program participants included in the present analysis all started antiretrovirals between July 2003 and May 2004 as part of a randomized ART efficacy monitoring trial,¹⁵ the results of which are to be presented elsewhere after 3 years of follow-up. Eligible clients were all provided with ART and cotrimoxazole prophylaxis and with treatment for tuberculosis and other illnesses when indicated. After enrollment, no routine clinic visits were scheduled. Rather, clients were visited weekly in their homes by trained lay field officers who delivered ART and cotrimoxazole and performed general health assessments. Medical officers evaluated patients at the program's facility-based clinic on an as-needed basis as determined by patient self-referrals or protocol-driven referrals by field officers for illness or suspected drug toxicity. In addition to weekly home delivery of ART, antiretroviral adherence support included group education, a medicine companion, and personal adherence plans developed with trained counselors.¹⁵

Study Population

From among the 987 HBAC Program participants (antiretroviral-naive adults enrolled during their first ART eligibility screening between July 1, 2003 and May 31, 2004), we selected a subset for the lipid (and liver function) substudy, based on sample size calculations that indicated the need to survey 150 persons with baseline and follow-up measurement to achieve 85% power to detect a difference in means of 15 mg/dL in TC using a paired t test with a 2-sided significance level or α of 0.05. A larger sample size would permit detecting smaller changes in mean cholesterol levels over time in the overall and stratified analyses. At the inception of the substudy, we identified 854 participants (645 women and 209 men) for whom the necessary serum specimens had been collected at baseline, 3 months, 6 months, 12 months, and 18 months. From among these persons, we selected a random subset of 200 women and all 209 men (n = 409) to have adequate statistical power to evaluate lipid changes separately by gender, assuming that up to 25% of selected persons could have inadequate stored sera for the necessary analyses. We also examined baseline lipid levels for an additional subset of patients who died during the first 12 months after starting HAART (n = 72).

Data Collection

Study physicians at Tororo District Hospital collected clinical information during screening using standardized instruments. We double-entered clinical and questionnaire data using Epi Info 2004 (US Centers for Disease Control and Prevention [CDC], Atlanta, GA). As part of the randomized monitoring trial, all clients had baseline and quarterly HIV viral load and CD4 cell count enumerations performed and quarterly repository blood plasma and serum specimens collected. Specimens were drawn in participants' homes by field phlebotomists and transported to the CDC laboratory at the Uganda Virus Research Institute in Entebbe for processing, testing, and storage. We measured HIV viral loads using the Cobas Amplicor HIV-1 Monitor version 1.5 (Roche, Branchburg, NJ) and counted CD4 T cells using a FACScan flow cytometer (Becton-Dickinson, Franklin Lakes, NJ). Serum concentrations of TC, HDL-c (direct), LDL-c (direct), and TG were assessed using enzymatic kits (Randox, Crumlin, United Kingdom), with reagent formulations designed to maximize the accuracy of HDL-c and LDL-c measurements in persons with elevated TG.

Statistical Analyses

For the purpose of comparing the prevalence of lipid abnormalities in our study population with data from other study populations, we used the criteria proposed in the Third Report of the US National Cholesterol Education Program (NCEP) to define the following values as abnormal: TC ≥200 mg/dL, LDL-c ≥130 mg/dL, HDL-c <40 mg/dL, and TG ≥150 mg/dL.¹⁷ We defined as abnormally elevated a TC/HDL-c ratio of ≥5. We compared baseline lipid levels among groups defined by various baseline sociodemographic and clinical characteristics using the Kruskal-Wallis test. To examine changes in lipid levels over time, we compared mean lipid levels at baseline, 12 months, and 24 months and computed associated 95% confidence intervals (CIs). Medians and interquartile ranges (IQRs) were also presented to show the variability in lipid levels among patients. We calculated within-individual actual and percentage changes in lipid levels from baseline to 12 months and from baseline to 24 months. We compared the proportions of patients with each lipid abnormality at baseline versus 24 months using the McNemar test for correlated proportions. We used generalized linear models (GLMs) to identify factors independently associated with changes in lipid levels from baseline to 24 months. To improve the fit and statistical inferences from the GLM analyses, we excluded extreme outlying lipid values (ie, those that were 3 or more IQRs above the third quartile or 3 or more IQRs below the first quartile),¹⁸ which resulted in dropping 2 TC, 3 LDL-c, 8 HDL-c, and 15 TG measurements from the analyses. Analyses were performed using SAS software version 8.02 (SAS Institute, Cary, NC).

Human Subjects Approval

The Uganda National Council of Science and Technology and the Institutional Review Boards of the Uganda Virus Research Institute, the University of California at San Francisco, and the US CDC approved the main study (HBAC Program). Informed consent to participate in the program was obtained from each participant in English or in 1 of 6 local languages. The Institutional Review Boards of the Uganda Virus Research Institute and the CDC also approved the lipid substudy.

RESULTS

Of the 409 patients initially selected for the substudy of lipid changes over time, sufficient remnant sera were available for 398 patients at baseline, for 407 patients at 12 months (±45 days), and for 399 patients at 24 months (±45 days). We limited our study population to the 374 persons who remained on first-line HAART (14 persons who switched to second-line HAART, which included lopinavir/ritonavir, were excluded) and had lipid measurements available at all 3 time points (baseline, 12 months, and 24 months).

Patient Characteristics

At baseline, the 374 patients evaluated (49% women) had a mean age of 39 years, CD4 count of 124 cells/mm³, log₁₀ HIV viral load of 5.2 copies/mL, and body mass index (BMI) of 19.7 kg/m². Twenty percent of patients had no education, 52% had a primary education, and 28% had a postprimary education; 38% of patients reported farming as their main source of income, whereas 26% reported trade, 19% remittances, and 17% wages/salaries. Compared with men (n = 191), women (n = 183) were younger (mean age: 37 vs. 41 years; P < 0.001), had a higher CD4 count (mean: 133 vs. 116 cells/mm³; P = 0.04), had a lower log₁₀ HIV viral load (mean: 5.1 vs. 5.4 copies/mL; P < 0.001), had a lower weight (mean: 53 vs. 57 kg; P < 0.001), and had a higher BMI (mean: 20.2 vs. 19.1 kg/m²; P < 0.001). In addition, women were less likely than men to report drinking alcohol (10% vs. 32%; P < 0.001).

The 374 patients evaluated received initial HAART composed of stavudine plus lamivudine for all patients and nevirapine for 365 (98%) or efavirenz for 9 (2%) patients. During 24 months of treatment, 99 (26%) patients had single-drug substitutions from stavudine to zidovudine and 27 (7%) had single-drug substitutions from nevirapine to efavirenz.

Baseline Lipid Values

At baseline, the mean serum lipid concentrations were as follows: 120 mg/dL for TC (range: 39 to 263 mg/dL), 53 mg/dL for LDL-c (range: 12 to 157 mg/dL), 29 mg/dL for HDL-c (range: 3 to 83 mg/dL), and 123 mg/dL for TG (range: 34 to 431 mg/dL); mean and median values were similar, except for TG (median = 113 mg/dL). TC was ≥200 mg/dL for 8 (2%) patients, LDL-c was ≥130 mg/dL for 2 (1%), HDL-c was <40 mg/dL for 330 (88%), TG were ≥150 mg/dL for 86 (23%), and the TC/HDL-c ratio was ≥5 for 111 (30%).

In univariate analyses of baseline data, the mean lipid levels were comparable by gender and age group, except that LDL-c levels were higher in women and HDL-c levels were lower and TG were higher in persons <35 year old (Table 1). Baseline HIV viral load >100,000 copies/mL was associated (P < 0.05) with lower TC, lower LDL-c, lower HDL-c, and a higher TC/HDL-c ratio, whereas baseline CD4 count <50 cells/mm³ was associated with higher TG levels. Persons with a BMI <18 kg/m² had lower TC and lower LDL-c. Alcohol use was associated with a lower TC/HDL-c ratio. Main source of income and educational level were not associated with lipid levels (data not shown).

Table 1

Compared with other patients, the 111 (30%) patients who had a baseline TC/HDL-c ratio ≥5 had lower CD4 counts (mean: 111 vs. 129 cells/mm³; P = 0.03) and higher log₁₀ HIV viral loads (mean: 5.3 vs. 5.2 copies/mL; P = 0.03). The 86 (23%) patients who had baseline TG ≥150 mg/dL were more likely to have a BMI <18 kg/m² (P = 0.05) and tended to have higher log₁₀ HIV viral loads (mean: 5.4 vs. 5.2 copies/mL; P = 0.06). Baseline TC ≥200 mg/dL and LDL-c ≥130 mg/dL were too infrequent for analogous analyses.

Follow-Up Lipid Values and Changes From Baseline to 24 Months

During 24 months of treatment, patients' mean weight increased from 55 to 58 kg and their mean CD4 count increased from 124 to 385 cells/mm³ (for both, P < 0.001). Mean TC, LDL-c, and HDL-c levels were significantly higher at 24 months compared with the baseline levels (for all, P < 0.001); the TC/HDL-c ratio was significantly lower (P < 0.001), and the TG levels were comparable (P = 0.40) (Table 2). During 24 months of treatment, TC increased by a mean of 31 mg/dL, LDL-c by a mean of 26 mg/dL, and HDL-c by a mean of 19 mg/dL, whereas the TC/HDL-c ratio decreased by a mean of 1.2 mg/dL and TG decreased by a mean of 3.8 mg/dL. These actual changes translated to median percentage changes of 23% for TC, 51% for LDL-c, 61% for HDL-c, -13% for TG, and -22% for the TC/HDL-c ratio over the 24-month period (medians rather than means were reported because the distributions of percentage changes were skewed). The 12-month results indicated that TC and HDL-c levels increased during the first year on HAART and subsequently leveled, whereas LDL-c levels increased gradually over the 2-year period; TG initially decreased and then returned to baseline levels at 24 months (see Table 2; Fig. 1).

Table 2

Figure 1

Comparing baseline and 24-month data, the proportions of patients with TC ≥200 mg/dL (2% vs. 10%) and LDL-c ≥150 mg/dL (1% vs. 6%) were modestly but significantly higher at 24 months (P < 0.001 for both; Fig. 2). The proportions of patients with HDL-c <40 mg/dL (88% vs. 41%) and a TC/HDL-c ratio ≥5 (30% vs. 7%) were significantly lower at 24 months (P < 0.001 for both); the proportion with TG ≥150 mg/dL was similar (23% vs. 20%; P = 0.29).

Figure 2

Factors Associated With Changes in Lipid Levels From Baseline to 24 Months

In univariate analyses, baseline lipid levels were strongly associated with the subsequent changes in lipid levels over 24 months of HAART. Persons who had the lowest (ie, first quartile) baseline levels of a given cholesterol had the greatest mean increases in that cholesterol (65 mg/dL for TC, 38 mg/dL for LDL-c, and 25 mg/dL for HDL-c), whereas persons with the highest baseline levels of that cholesterol (fourth quartile) had the smallest increases or losses (-4 mg/dL, 8 mg/dL, and 13 mg/dL, respectively) (all P < 0.001). The same pattern was present for TG: persons with baseline levels in the first quartile had a mean increase of 30 mg/dL, whereas persons with baseline levels in the fourth quartile had a mean decrease of -54 mg/dL (P < 0.001). This pattern of lipid changes held regardless of patients' baseline CD4 cell count, HIV viral load, or weight.

In univariate analyses, 24-month changes in any lipid were not statistically different (P > 0.05) according to gender or age (data not shown). In addition, there was no effect modification (interaction) by gender or age in the associations between baseline lipid levels or other clinical and sociodemographic variables and lipid changes.

We performed multivariable GLM analyses by including sociodemographic and clinical factors that in univariate analyses were at least weakly associated (P < 0.20) with a 24-month change in at least 1 lipid variable (ie, TC, LDL-c, HDL-c, TG). All models thus included baseline lipid level, age, gender, baseline BMI, baseline CD4 cell count, baseline HIV viral load, alcohol use, change in weight, and change in CD4 cell count from baseline to 24 months (Table 3). Lower baseline levels of a lipid were strongly and independently predictive of greater increases in that lipid for TC, LDL-c, and HDL-c and of lesser decreases for TG and TC/HDL-c ratio over 24 months. After controlling for baseline lipid levels and other factors, there were no marked or statistically significant differences in the 24-month lipid changes according to gender or age. Likewise, the clinical baseline covariates (CD4 cell count, HIV viral load, and BMI) and change in weight and CD4 cell count over the 24-month period were not associated with or explained relatively little of the additional variability in lipid changes, after controlling for baseline lipid levels. For example, for the TC/HDL-c ratio, patients who had an HIV viral load ≥100,000 copies/mL or low BMI at baseline had more reduction in the ratio. The estimates of mean changes were similar in parsimonious models, which excluded factors with P > 0.05.

Table 3

Changes in Lipid Levels by Type of Antiretroviral Regimen

In the GLM analyses that compared 275 patients who remained on stavudine and 99 patients who switched to zidovudine during the 24 months (median zidovudine exposure = 17 months, range: 3 to 21 months), the baseline to 24-month changes in all lipids were not statistically different for the 2 groups after adjusting for baseline lipid levels, except for the mean change in the TC/HDL-c ratio, which was -1.0 for the stavudine group and -1.2 for zidovudine group (P = 0.03). In analogous analyses comparing 338 patients treated continuously with nevirapine with 36 patients who started on or switched to efavirenz (median efavirenz exposure = 22 months, range: 3 to 24 months), the baseline to 24-month changes in all lipids were not statistically different between the 2 groups.

Comparative Baseline Data for Additional Patients Who Died During the First 12 Months

The 374 patients selected for the primary analyses of 24-month lipid changes all, by definition, survived to 24 months. Comparative baseline lipid data were available for 63 of 72 patients who died during the first 12 months in HBAC Program. At baseline, these 63 patients (63% women) had a mean age of 39 years, CD4 count of 80 cells/mm³, log₁₀ HIV viral load of 5.5 copies/mL, and BMI of 17.4 kg/m². Compared with the baseline data for the 374 patients, the mean baseline serum lipid concentrations for 63 patients who died were somewhat lower for TC (115 mg/dL; P = 0.09) and LDL-c (49 mg/dL; P = 0.19), significantly lower for HDL-c (22 mg/dL; P < 0.001), and significantly higher for TG (163 mg/dL; P < 0.001) and the TC/HDL-c ratio (5.8; P < 0.001).

DISCUSSION

In this rural population of Ugandan patients with advanced HIV disease, we found infrequent lipid elevations at baseline and after 24 months on nonnucleoside reverse transcriptase inhibitor (NNRTI)-based HAART. Most patients had low HDL-c levels before starting treatment. TC and HDL-c levels increased mostly during the first year and remained stable during the second year, whereas LDL-c levels continued to increase through 24 months. TG levels decreased during the first year and returned to baseline levels by 24 months. The increases in HDL-c were proportionally greater than the increases in TC, resulting in a marked reduction in the TC/HDL-c ratio. There was substantial variability among patients in the magnitude of lipid changes, however, with the most prominent increases in TC, HDL-c, and LDL-c among patients with the lowest baseline levels, and thus the most opportunity to increase.

Our study was not designed to predict the risk of CVD. Although studies from industrialized countries would suggest that the lipid changes we observed (eg, substantial reduction in the TC/HDL-c ratio) are likely to reduce the risk of CVD,^1,2,7-9 findings from these populations may not necessarily be generalizable to sub-Saharan Africans because of potential differences in diet, activity levels, and genetic susceptibility to CVD. Further studies to establish the normal and abnormal lipid levels among sub-Saharan Africans^19,20 and how lipid levels relate to the CVD risk among HIV-infected and HIV-uninfected Africans would be necessary to determine if the lipid changes we observed could indeed be beneficial. Although the NCEP thresholds for abnormal lipid levels may not be clinically meaningful for HIV-infected rural Ugandans, we presented the data in this manner to contrast our findings to those from other studies in industrialized and resource-constrained settings,^14,21,22 of which 1 study suggested more atherogenic changes in lipids on the same first-line HAART.¹⁴

Studies in industrialized countries have documented that compared with HIV-uninfected persons, patients with HIV/AIDS have decreased levels of TC, LDL-c, and HDL-c and increased levels of TG;²³ the extent of lipid abnormalities correlates with the severity of disease in these populations.^24-26 The changes we observed in lipid values in this study (an increase in TC, HDL-c, and LDL-c and a decrease in TG) may therefore, at least in part, represent the return to normal lipid values when the immune status improves and HIV viremia is controlled on HAART. Indeed, after 24 months of HAART, the HIV-infected HBAC Program patients had mean TC and HDL-c levels that were approximately similar to the levels found in 2 studies of general adult populations in rural Tanzania, which presumably included mostly HIV-uninfected persons.^19,20 Antiretroviral agents may also directly affect the metabolism of lipids and resultant serum lipid levels.^27-29 NNRTI-based nevirapine-containing regimens have been associated with more favorable changes in lipid profiles (lesser increases in TC, TG, and LDL-c and proportionally greater increases in HDL-c) than PI-based regimens^4,10,30,31 or NNRTI-based regimens containing efavirenz.^4,12 By contrast, stavudine has been linked to dyslipidemia and lipodystrophy.^3,32,33 In the HBAC Program, stavudine was chosen over zidovudine because of concerns of anemia in this malaria-endemic region. It is reassuring that the stavudine-containing NNRTI-based regimens were not associated with frequent lipid abnormalities during the first 24 months of use and were associated with improvement in the TC/HDL-c ratio. It is uncertain if these changes in lipid profiles would persist beyond the 24 months of use of this regimen in HBAC Program patients.¹³

Our findings stand in contrast to those from a population of 150 self-paying patients in Pune, western India,¹⁴ who were treated for a mean of 20 months with stavudine, lamivudine, and nevirapine and were roughly comparable by age and immunologic status to our study population but included more men (mean age = 38 years, 83% male, mean baseline CD4 count = 117 cells/mm³). In that study, 41% of patients developed TC levels ≥200 mg/dL and 45% developed TG levels ≥150 mg/dL.¹⁴ The contrasting findings of the Indian study and our study of Ugandan patients may have been attributable to differences in socioeconomic status, diet, and genetic makeup,³⁴ which may all affect lipid changes in patients receiving HAART. Our findings are largely consistent with those from the large multicontinental 2NN study, which recorded mean percent increases of 27% for TC, 35% for LDL-c, and 43% for HDL-c among 417 patients who received 48 weeks of lamivudine, stavudine, and nevirapine (median age = 35 years, 63% male, median CD4 count = 180 cells/mm³), with sustained increases in TC and LDL-c but not in HDL-c through 1.5 years of observation.^12,13

The HBAC Program provided a unique opportunity to compare lipid profiles among HIV-infected men and women. With the exception of higher baseline LDL-c levels among women, we did not find marked differences by gender in pretreatment lipid levels or in the subsequent changes in lipid levels on HAART, which have been documented in studies of HIV-infected patients elsewhere.^12,26 Compared with the patients in the aforementioned studies, however, HBAC Program participants had lower pretreatment lipid levels, likely resulting from advanced HIV/AIDS disease and a diet low in animal fats, and these factors may have diminished potential differences, if any, in lipid profiles according to gender.

Our findings should be interpreted in light of the following caveats. First, lipid measurements were performed on nonfasting repository sera, and although recent food intake is unlikely to have a substantial impact on the concentrations of cholesterol, HDL-c and LDL-c, it may increase the concentrations of TG.³⁵ Nevertheless, the use of nonfasting sera would most likely have led to a systematic overestimation of TG levels at all time points and would not have affected our inferences about changes in TG levels on HAART. Second, we did not collect individual-level nutritional data. An estimated 10% of HBAC Program patients were enrolled in a TASO food support program, but we do not believe that our study population experienced major changes in nutritional intake during 24 months; thus, the changes in lipids we observed over the course of the study were most likely attributable to control of HIV infection and the action of specific antiretroviral agents rather than to changes in nutritional status. Third, we recognize that although some of the differences we detected in lipid profiles by patient characteristics may be statistically significant, they may not be clinically meaningful. Fourth, because of the ethical imperative to provide antiretrovirals to all screened and eligible patients with advanced HIV/AIDS, we did not have a comparison group of HIV-infected patients followed without antiretrovirals, which might have allowed us to isolate the effect of HAART versus other factors on changes in lipid levels; a comparison group of HIV-uninfected persons from the general population of the region was also not available. Fifth, we could not assess whether variable virologic response may have affected changes in lipids among individuals, as done by van Leth et al.^12,13 Prior HBAC Program analyses found that 99% of HBAC Program patients had a pill count adherence ≥95% and that 96% had an undetectable HIV viral load after 12 months of HAART use;¹⁵ thus, differential adherence and/or virologic responses alone would not explain the variability in lipid changes that we observed.

In conclusion, in this rural population of sub-Saharan African patients with advanced HIV disease, we found that serum cholesterol levels increased during 24 months of an NNRTI-containing regimen but that relatively few patients developed high TC, LDL-c, or TG levels; the overall HDL-c levels increased; and the TC/HDL-c ratio decreased. These findings lend support to the current WHO guidelines, which do not recommend routine monitoring of lipid levels on first-line HAART regimens.¹¹

ACKNOWLEDGMENTS

The authors thank the volunteers, staff, and clients of the TASO and the staff of CDC-Uganda and the Global AIDS Program headquarters.