Metabolic benefits, 24 months after replacing a protease inhibitor with abacavir, efavirenz or nevirapine, NEFA Study
AIDS: Volume 19(9) 10 June 2005
Fisac, Cesara; Fumero, Emiliob; Crespo, Manuelc; Roson, Beatrizb; Ferrer, Elenab; Virgili, Nuriaa; Ribera, Estebanc; Gatell, Jose Mariad; Podzamczer, Danielb
From the aNutrition Section
bInfectious Disease Service, Hospital Universitari de Bellvitge, L'Hospitalet de Llobregat
cInfectious Disease Service, Hospital Universitari Vall d'Hebrón, Barcelona
dInfectious Disease Service, Hospital Clinic, Barcelona, Spain.
Objective: To evaluate the 24-month metabolic and morphological benefits obtained from replacing the protease inhibitor (PI) in a regimen with nevirapine, efavirenz or abacavir.
Design and methods: NEFA was a randomized study designed to compare the efficacy of nevirapine, efavirenz or abacavir as substitutes for PI. A subset of 90 patients [abacavir (n = 29), efavirenz (n = 32), nevirapine (n = 29)] formed the metabolic study. Fasting total cholesterol (TC), high density lipoprotein cholesterol (HDL-c) and triglycerides levels were determined. Glucose homeostasis parameters were also collected. Lipodystrophy was evaluated by clinical examination and morphological measurements.
Results: Treatment simplification led to overall lipid profile improvements. At 24 months, the two non-nucleoside reverse transcriptase inhibitors produced similar lipid benefits: HDL-c levels increased [efavirenz, 15% (P = 0.001); nevirapine, 21% (P < 0.001)] and TC to HDL-c ratios decreased [efavirenz, 14% (P < 0.001); nevirapine, 19% (P < 0.01)], an effect not observed in the abacavir arm. Non-HDL-c levels decreased by 10% in both the abacavir (P = 0.001) and efavirenz (P < 0.05) arms. Significant decreases in the levels of triglycerides occurred for the first year in all treatments; however, at 24 months most of the initial loss had been regained. Patients with baseline moderate or severe lipodystrophy obtained less-pronounced lipid benefits. Several insulin resistance markers showed a trend towards improvement. Conversely, no improvements in morphological abnormalities were observed.
At month 24, the proportion of patients in the substudy with plasma HIV-1 RNA levels < 200 copies/ml did not differ among the arms (ABC, 93%; EFV, 88%; NVP, 96%). Likewise, the median CD4 cell count increase was comparable between the study treatments: 62 (IQR, -40 to +174), 118 (IQR, 21-121) and 93 × 106 cells/l (IQR, 29-206) for ABC, EFV and NVP, respectively.
Conclusions: Replacing PI with efavirenz, nevirapine or abacavir improved the lipid profile, with more marked results in non-lipodystrophic patients. In contrast, this strategy does not seem to be effective for reversing body fat abnormalities.
Lipid metabolism outcomes
At entry, the overall median fasting serum levels of TC, non-HDL-c and triglycerides were above the desirable upper limits of 5.17 mmol/l (200 mg/dl), 4.14 mmol/l (160 mg/dl) and 1.69 mmol/l (150 mg/dl), respectively, as defined by the National Cholesterol Education Program . There were no differences in the baseline lipid profile among the groups.
Treatment simplification led to an overall lipid metabolism improvement in the entire study population and in each randomization arm. At 24 months, median TC levels had decreased significantly in the ABC arm (14%; P < 0.001), but not in the EFV or NVP arms. With respect to HDL-c, significant increases occurred in both non-nucleoside reverse transcriptase drug (NNRTI)-containing regimens at 12 months [EFV, 17% (P < 0.001); NVP, 21% (P < 0.001)] and 24 months [EFV, 15% (P = 0.001); NVP, 21% (P < 0.001)] after switching; no variations in HDL-c levels were observed in the ABC arm. Median percentage increases in HDL-c levels from baseline in the two groups switching to NNRTI differed from that of the ABC group at both follow-up time points.
As a consequence of the these changes, median TC/HDL-c ratios decreased significantly in all the treatment arms at 12 months [ABC, 14% (P < 0.01); EFV, 16% (P < 0.001); NVP, 19% (P < 0.001)] but only in the NNRTI-containing arms at 24 months [EFV, 14% (P < 0.001); NVP, 19% (P < 0.01)]. Median TC/HDL-c ratios percentage changes from baseline in the two groups switched to NNRTI differed from that of the ABC group at the end of study (P < 0.05 for both). Regarding the non-HDL-c fraction, significant decreases were seen in all three treatment arms at the 12-month assessment [ABC, 17% (P < 0.01); EFV, 7% (P < 0.01); NVP, 9% (P < 0.05)]. At 24 months, these improvements were maintained in the ABC group (-10%; P = 0.001) and the EFV group (-11%; P < 0.05). Within each treatment arm, fasting triglycerides levels followed a similar pattern, falling significantly up to the 12-month time point [ABC, -20% (P < 0.05); EFV, -33% (P < 0.01); NVP, -29% (P < 0.01)] and then increased throughout the second year, although they did not exceed the initial values. No differences in the median changes or absolute values for non-HDL-c or triglycerides were found among the study treatments at any follow-up time point. In each treatment group, apo A-I and apo B levels changed in parallel with HDL-c and non-HDL-c levels, respectively. The ratios of apo A-I to HDL-c and apo B to non-HDL-c also remained unvaried throughout the study (data not shown).
Lipid profile responses were studied separately within each lipodystrophy category. In general, patients who had moderate to severe lipodystrophy at entry showed smaller improvements than the respective groups of remaining patients. The difference in response was significant (P < 0.05) for HDL-c levels for both comparisons and for TC/HDL-c ratios for the comparison with patients who had moderate to severe lipoaccumulation at entry.
Glucose metabolism outcomes
Glucose metabolism parameters were comparable among treatments at all time points. At the end of study, small but significant increases in the median fasting glucose levels had occurred in the overall cohort (4%; P < 0.01) and in the EFV group (7%; P < 0.01). However, these increases were of no clinical relevance since the levels did not exceed the upper normal limit in any patient with optimal baseline values (< 7.0 mmol/l).
With respect to fasting insulin, 24-month levels showed significant decreases overall (13%; P < 0.01) and in the EFV group (19%; P < 0.05) compared with the initial values. Although median insulin levels also decreased in the other two study regimens, the reductions did not reach statistical significance (ABC, 2%; NVP, 9%; P = 0.24 for both). As a result of these changes, the median HOMA-IR index had decreased in the overall group (2%) and in each treatment arm (ABC, 2%; EFV, 20%; NVP, 5%) at the 24-month visit; these changes, however, did not reach a significant level.
The levels of soluble tumour necrosis factor receptors 1 and 2 and adiponectin were determined at baseline and at the 24-month visit in a representative subset of 54 patients who had maintained the allocated treatment during the study period (data not shown). Overall, the median adiponectin levels increased from 5.24 to 5.71 μg/ml (P = 0.034), and the median soluble tumour necrosis factor receptor 2 levels decreased from 6.74 to 6.30 ng/ml (P = 0.014). Baseline plasma levels of soluble tumour necrosis factor receptors 1 were similar to those of the 24-month visit.
Lipodystrophy and morphological outcomes
At baseline, patients with lipodystrophy had lower percentages of body fat mass (P < 0.01) as assessed by bioelectrical impedance analysis, higher waist-to-hip ratios (P < 0.05) and trends towards smaller femoral girths (P = 0.07) than the remaining patients. Lipoatrophy and lipoaccumulation were both categorized as either moderate or severe in most cases (74% for both). Of the lipodystrophic patients, 84% were taking stavudine as part of their backbone treatment at study entry.
The percentages of patients with lipoatrophy or lipoaccumulation, as determined by physical examination, did not change over time (Table 3). In addition, the percentage of patients with moderate or severe lipodystrophy, and the severity rating within each lipodystrophy category had not varied at the end of the study. Among those who initiated with lipoatrophy and lipoaccumulation, 81% and 57%, respectively, either maintained their status or showed a certain degree of worsening. In two patients, the onset of peripheral fat loss and in two others the onset of central fat accumulation occurred during the study period.
In the overall study population, linear regression showed no significant changes in the anthropometric variables after 2 years of follow-up (data not shown). However, change-in-weight adjustment revealed a 7% significant decrease (P < 0.05) in total body fat as assessed by bioelectrical impedance analysis, a finding usually observed in the course of peripheral fat loss. The variations in anthropometric measurements, with the exceptions of tricipital skinfold and waist-to-hip ratio, were partly explained by changes in the total body weight (data not shown). No differences were observed among treatments.
DEXA was performed at baseline and after 12 months in a subsample of 28 patients (13 with lipodystrophy at entry) who had maintained the assigned treatment. Even though the total body weight did not vary in this subset (median change, 0.0 kg), the median fat content in the limbs decreased significantly (P < 0.05) from 3.6 kg (IQR, 2.6-5.2) to 2.9 kg (IQR, 1.9-4.0). Altogether, the analysis of body habitus changes by objective measurements indicated that there was either no recovery of lipodystrophy or a worsening of peripheral lipoatrophy during the study period.
At baseline, almost 70% of patients presented lipid profile alterations. Two years after replacing the PI, overall lipid metabolism improved in each randomization arm, although there were some distinctive features among the treatments. On the one hand, ABC seemed to exert a stronger lowering power on the total and non-HDL-c fraction, especially when compared with NVP. On the other hand, both EFV and NVP showed an evident beneficial effect on HDL-c levels and the TC/HDL-c ratio at the end of the study, which was not seen in those allocated to ABC. Finally, no differences among the treatments were detected with respect to triglycerides.
Our results are in agreement with most PI switch studies. Shorter randomized trials evaluating dyslipidaemia after substitution of ABC for a PI have shown reductions in TC and triglycerides, or lower levels when compared with the PI-continuing arms [17-21]. In one of these trials, HDL-c levels showed no differences between treatment groups after 19 months . With respect to the two NNRTI drugs, most trials have shown increases in HDL-c levels, particularly with NVP [22-25]. However, the results show greater disparity regarding TC, low density lipoprotein cholesterol and triglycerides levels, with either improvements or no changes after the replacement [8,9].
One way to address the question as to which randomization arm might have shown the most beneficial lipid profile changes is by determining whether the TC/HDL-c ratio improvement occurring in the groups switched to a NNRTI implies an additional cardiovascular disease risk benefit. Several observational studies have concluded that this ratio is the best single variable of the traditional lipid profile for predicting cardiovascular events . In an analysis of three major epidemiological studies, neither TC nor low density lipoprotein cholesterol levels conveyed additional information when the ratio was used to predict coronary heart disease . In another study designed to compare the predictive value of 11 lipid and non-lipid biomarkers as risk factors for the development of peripheral arterial disease, the strongest lipid predictor was the TC/HDL-c ratio . On this basis, we believe that the lipid changes observed in our two NNRTI-containing groups might be somewhat more cardioprotective than those seen in the ABC-containing combination.
An event not described before is the evolution seen in triglyceride concentrations. At month 12, levels had decreased significantly; this trend then inverted and baseline levels were almost recovered at 24 months. The initial lowering might be the consequence of elimination of the well-known direct effect of PI on triglyceride concentration. The elevation throughout the second year might be reflecting a worsening of lipoatrophy (as detected by DEXA in a subsample of the patients) or cumulative exposure to stavudine (78% of patients at entry) .
In keeping with several other shorter studies [17,29-31], our results indicate an improvement in glucose metabolism. Twenty-four months after the switch, overall insulin levels improved and the HOMA-IR showed a trend towards improvement. Levels of soluble tumour necrosis factor receptor 2 and adiponectin have been associated (inversely in the case of adiponectin) with insulin resistance in HIV-infected and seronegative populations [32-36]. Therefore, both the decrease in soluble tumour necrosis factor receptor 2 and the increase in adiponectin observed in our sample reinforce the likelihood of an improvement in insulin sensitivity 24 months after PI replacement.
In agreement with other studies [10,30,37], switching from a PI-containing regimen did not prove to be a good strategy for treating HIV-associated lipodystrophy. We found that the percentages of patients with lipoatrophy or with lipoaccumulation, as assessed by clinical examination, did not decrease, and the waist-to-hip ratio and other commonly used anthropometric markers did not vary. Interestingly, a mild gain in body weight masked a significant total body fat decrease, the latter being a common event in patients with the initiation or exacerbation of a peripheral fat loss process. Additionally, DEXA performed in a subsample of 28 patients showed a significant decrease in fat content of the limbs after 1 year of follow-up.
A number of cohort studies have suggested that long-term exposure to nucleoside reverse transcriptase inhibitors, particularly stavudine, results in lipoatrophy [38-40]. Additionally, replacement of stavudine or zidovudine has resulted in significant improvements in lipoatrophy, but not in the visceral fat accumulation [41-43]. Hence, maintenance of previous nucleoside reverse transcriptase inhibitor therapy, including stavudine in most cases (77% at baseline), is the most feasible explanation for why peripheral lipoatrophy did not improve in our study group.
There are some specific limitations to this study. The statistical power to detect changes over time in the anthropometric measurements and in the proportions of patients with lipoatrophy or lipoaccumulation proved to be limited. Hence, we cannot rule out the possibility of true changes in these variables not detected by our analyses. Also, we cannot exclude that changes in diet or exercise habits may have accounted for some of the changes in lipid levels. Data from recent studies suggest that not all PI may have the same effects on lipid and carbohydrate metabolism . Hence, it is possible that our results might have shown some variation if prior treatment regimens had contained a different PI. Finally, caution is necessary when interpreting our results in terms of cardiovascular risk benefit, as a number of unmeasured risk factors might also have been affected.
In conclusion, our analysis suggests that replacing a PI with NVP, EFV or ABC results in an improved metabolic profile, especially in non-lipodystrophic patients. It is, therefore, a reasonable approach for patients with good virological control at high risk for cardiovascular disease, such as those with a prior coronary event or equivalent risk. In contrast, PI replacement did not prove to be a good strategy for stopping the peripheral fat atrophy process. Full interpretation of the metabolic effects of switching is difficult for many reasons. However, the specific improvement in TC/HDL-c ratio exerted by the two NNRTI drugs might imply a further benefit in terms of cardiovascular risk.
Population characteristics and treatment response
Except for the distribution of the prior PI (P < 0.001), the overall baseline characteristics of substudy patients were comparable with those of the main study population. Baseline characteristics of the 90 selected patients did not differ among the treatment arms. At the end of the study, five (6%) patients had been lost to follow-up, 13 (14%) patients had discontinued because of adverse effects and four (4%) patients because of antiretroviral failure. Hence, 69 (77%) out of the 90 initially selected patients maintained the randomization treatment for 24 months. The adverse effect that led to discontinuation in one patient randomized to EFV was severe hypertriglyceridaemia (18.8 mmol/l), and the reason for discontinuation in one patient randomized to ABC was persistent complaints about body shape changes. Baseline characteristics of the patients who discontinued the allocated treatment did not differ from those who maintained it to the end of study (data not shown). Two patients in each treatment group received lipid-lowering therapy at some time in the study. In addition, two of these six patients (one on EFV and one on NVP) also received antihyperglycaemic agents. These data were excluded from analysis, as appropriate.
Current antiretroviral therapies for HIV infection are associated with an increasing prevalence of metabolic side effects and morphological abnormalities . Lipid profile alterations and insulin resistance are the most frequently reported metabolic derangements [2,3]. The morphological alterations are mostly characterized by subcutaneous fat depletion in peripheral areas and, in some patients, fat accumulation around the dorsocervical region and abdomen . Although the combination of these morphological and metabolic manifestations have been designated the unique denomination of HIV-associated lipodystrophy syndrome, they may also appear separately, questioning whether they are parts of a single phenomenon . Among currently available antiretroviral agents, the protease inhibitor (PI) drugs are most highly associated with these metabolic alterations and at least one form of morphological abnormality [1-4].
Hyperlipidaemia and insulin resistance are well-known risk factors for cardiovascular disease in the general population [5,6], and body changes resulting from abnormal fat distribution might have severe psychological repercussions that could affect adherence to antiretroviral treatment . Therefore, strategies to prevent or revert metabolic and body habitus abnormalities in these patients are urgently needed. Several trials have been conducted to evaluate the clinical benefits of changing to PI-sparing regimens [8,9]. However, differences in the drugs used, number of patients included and follow-up time, as well non-randomized designs and the use of confusing or equivocal methods for assessment have led to divergent results.
The NEFA study was a randomized trial conducted to evaluate the overall clinical benefit of a treatment-simplification strategy . Patients were randomized to nevirapine (NVP), efavirenz (EFV) or abacavir (ABC) as substitutes for the PI part of their regimen. After 1 year, a trend toward a higher rate of virological failure and a lower incidence of adverse effects leading to discontinuation was observed among those given ABC. In a subsample of NEFA patients, data were compiled to assess the metabolic and morphological benefits obtained from the PI switch. Herein, we present the results of that study after 24 months of follow-up.
The NEFA study was a multicentre, randomized, open-label clinical trial conducted at 15 centres in Spain and designed to compare the efficacy of NVP, EFV or ABC as substitutes for PI in a group of successfully treated patients with HIV-1 infection. Briefly, patients were randomly assigned to receive NVP, EFV or ABC in place of the PI used in their current antiviral therapy while continuing to take their nucleoside reverse transcriptase inhibitors. The study details are described elsewhere .
This substudy (LipNEFA) comprised 90 patients enrolled in two of the main study centres (Bellvitge Hospital and Vall d'Hebrón Hospital, Barcelona, Spain). Anthropometric and biochemical determinations were performed at baseline and at 3-month intervals up to 24 months after the initiation.
The presence and evolution of fat distribution alterations were evaluated at each visit by physical examination and anthropometric measurements. All the measurements and assessments were performed by the same examiner at each of the two participating centres. The examiners were not involved in the patient's primary care.
Anthropometric measurement followed standard techniques described elsewhere  and included height, weight, triceps' skinfold, and circumferences at five body sites (waist, hip, mid-arm, mid-thigh and calf). The body mass index (weight in kilograms divided by height in metres squared) and waist-to-hip ratio (waist circumference/hip circumference) were also calculated. Single frequency bioelectrical impedance analysis (model 310e, Biodynamics Corp., Seattle, Washington, USA) was used to estimate total body fat mass. Additionally, total and regional body composition was determined by dual-energy X-ray absorptiometry (DEXA; Lunar DPX-L equipment, Madison, Wisconsin, USA) in a subsample of 23 patients. DEXA measurements were performed at baseline and at 12 months and analysed by an independent observer.
Participants underwent an extensive physical examination to detect clinical signs of central adiposity (lipoaccumulation) and peripheral fat wasting (lipoatrophy), as previously described . Central adiposity features included breast enlargement, dorsal fat accumulation and increased abdominal girth, whereas peripheral lipoatrophy features consisted of decreased subcutaneous fat tissue in the face, buttocks and extremities. For the purpose of the analysis, body-fat abnormalities were categorized as lipoatrophy or lipoaccumulation. Therefore, patients with both fat loss and fat accumulation were included in both categories. The extent of both lipoatrophy and lipoaccumulation was scored 0 when considered absent, 1 (mild) when noticeable only on close inspection, 2 (moderate) when readily noticeable by patient and physician, and 3 (severe) when noticeable to a casual observer.
At all examinations, blood specimens were obtained after a 12-14 h overnight fast. Serum levels of total cholesterol (TC), high density lipoprotein cholesterol (HDL-c), triglycerides, apolipoproteins A-I (apo A-I) and B (apo B), glucose and insulin were measured with the routine assays at each participating centre. The non-HDL-c (TC minus HDL-c) was used instead of low density lipoprotein cholesterol for the analysis because it may better express the atherogenicity of the lipid profile, particularly in a population with above optimal levels of triglycerides . Insulin resistance was estimated from fasting glucose and insulin values, according to the homeostasis model assessment (HOMA-IR) . Plasma concentrations of soluble tumour necrosis factor receptors 1 and 2 and adiponectin were measured at a single laboratory from samples stored at -80°C, as described elsewhere .
The primary study endpoint was the evolution of the lipid parameters from baseline to month 24. Secondary endpoints were the changes in glucose metabolism parameters, and the changes in the prevalence and severity of lipoatrophy and lipoaccumulation as assessed by clinical examination and morphological measurements. All endpoints were analysed by using both the intention-to-treat and on-treatment approaches. In the intention-to-treat analysis, the last observation was carried forward in the case of those patients who were lost to follow-up. Since both analyses led to similar conclusions, the intention-to-treat analysis has been used to show the data because it may better express the results from a clinical perspective.
Data at baseline and at 12 and 24 months were used for the analyses. The results of the continuous variables are presented as the median absolute value or median percentage change from baseline, when appropriate, and the interquartile ranges (IQR). The chi-square test or Fisher exact test were used to analyse the associations in two-way frequency tables. The Wilcoxon rank sum test was used to compare continuous values between two independent groups and the Kruskal-Wallis test when three independent groups were compared. The Wilcoxon matched-pairs signed-rank test was used for comparisons with baseline values. To analyse whether a change in the anthropometric parameters depended on a change in weight, Spearman's correlation coefficients were estimated. Changes in lipodystrophy were assessed by the McNemar chi-square test in the case of dichotomous characteristics and by the Wilcoxon signed-rank test in the case of ordinal characteristics. Simple comparisons were made with a two-sided alpha level of 0.05. Each of the three pairwise comparisons used a two-sided significance level of 0.0167.