Dubé, Michael P MD*; Komarow, Lauren MS†; Mulligan, Kathleen PhD‡; Grinspoon, Steven K MD§; Parker, Robert A ScD†; Robbins, Gregory K MD∥; Roubenoff, Ronenn MD, MHS¶; Tebas, Pablo MD#
The term lipodystrophy refers to the changes in fat distribution that are frequent among HIV-positive individuals.1 There is no uniformly accepted definition of lipodystrophy, but it is generally accepted that body fat alterations are composed of 2 components: lipoaccumulation and lipoatrophy.2 Lipoaccumulation occurs with deposition of excess adipose tissue around the neck, dorsocervical spine (buffalo hump), upper torso, or intra-abdominal region. Lipoatrophy is the loss of subcutaneous fat, often from the arms, legs, buttocks, and face, and is the most characteristic physical change associated with HIV disease and its treatment.3 The contributions of HIV infection, host genetics, inflammation, and the individual drugs and their combinations used for the treatment of HIV to the development of lipodystrophy remain controversial because of the paucity of comprehensive prospective, longitudinal, randomized trials.
Antiretroviral therapy, particularly stavudine-containing regimens,4-8 is associated with lipoatrophy, a common and serious problem that is associated with significant esthetic and metabolic derangements.1 Newer antiretroviral therapy regimens seem to be less frequently associated with this complication.8 There are conflicting data from observational studies with regard to the independent contribution of protease inhibitors to developing lipoatrophy or trunk fat accumulation.2,9,10 We recently reported that 16 to 32 weeks after the initiation of antiretroviral treatment, there are significant gains in trunk fat and limb fat.4 This likely represents a “return to health” phenomenon, similar to what is seen in certain lipid fractions after antiretroviral drug initiation.11 After 32 weeks of therapy, individuals, particularly those on stavudine-containing regimens, tended to lose fat in the extremities. Although the contribution of protease inhibitors to the development of this problem still remains controversial, in our study, those assigned to the protease inhibitor nelfinavir had greater limb fat loss compared with efavirenz over the initial 64 weeks of antiretroviral therapy. Individuals assigned to efavirenz had greater trunk fat increases.4 The group that was assigned to nelfinavir and efavirenz was not included in those intent-to-treat analyses.
The purpose of this report is to describe the long-term (up to 144 weeks) dual x-ray absorptiometry (DXA) results from this randomized prospective study of initial antiretroviral therapy. One issue with long-term follow-up antiretroviral studies is that many individuals change antiretroviral regimens during follow-up. To exclude an effect of treatment switching, our primary analysis included only those participants who continued on their original randomized regimen. To avoid biasing the study by including only individuals with DXA scans at later time points, we included all scans available until the time of drug switching. We were particularly interested in the long-term results among participants who received zidovudine and lamivudine, the most tolerable nucleoside combination from the Adult Clinical Trials Group (ACTG) 384 study.12,13 This combination remains one of the preferred components of initial antiretroviral therapy14 and is still among the most frequently used nucleoside combinations in antiretroviral-naive individuals in the world.15
HIV-infected individuals were eligible for entry into ACTG 38412,13 if they had <7 days of prior antiretroviral experience and an HIV-1 RNA level of >500 copies/mL. This substudy, A5005s, enrolled a total of 334 of the 980 ACTG 384 participants at 23 participating sites in the United States from 1998 to 1999; study follow-up continued through 2001.4 All participants provided written informed consent per the guidelines of each site's institutional review board.
Participants were randomized to 1 of 6 treatment arms using a factorial design to receive nelfinavir, efavirenz, or both drugs in a blinded fashion, combined with open-label zidovudine and lamivudine or didanosine and stavudine.12,13 Participants with intolerance to zidovudine or didanosine were allowed to switch to stavudine or lamivudine, respectively, while continuing the rest of the originally assigned regimen. Drug changes attributable to lipodystrophy were not allowed by the protocol. Reasons for switching were most commonly gastrointestinal intolerance for zidovudine and gastrointestinal intolerance or low-grade lipase elevations for didanosine. Subjects assigned to a 3-drug regimen who experienced failure for virologic or toxicity reasons were switched from nelfinavir to efavirenz and switched from 1 nucleoside pair to the other. Subjects assigned to a 3-drug regimen who experienced failure of the second 3-drug regimen and those assigned to a 4-drug regimen who experienced failure for virologic or toxicity reasons had reached a primary study endpoint and discontinued the study. Further details on the ACTG 384 parent protocol are available elsewhere.12,13
Dual X-Ray Absorptiometry Scans
Whole-body scans were performed at 18 sites at entry and every 16 weeks thereafter on a subset of 157 participants enrolled in a substudy of metabolic and morphologic complications.4 Regional analysis was performed centrally at Tufts University. Total limb fat was the sum of arm and leg fat mass.
An as-treated analysis was conducted in which DXA data were censored 28 days after any antiretroviral drug change. A change was considered to have occurred when patients stopped taking at least 1 originally assigned antiretroviral drug for any reason, regardless of whether it was considered a regimen change by the parent protocol. For example, if a participant was randomized to zidovudine and lamivudine and stavudine was substituted for zidovudine because of anemia, this was considered an antiretroviral drug change for the purpose of this analysis, even though this was not considered a regimen change by the parent protocol.12,13 All data were censored at week 144.
Change in limb and trunk fat is reported as percent change and absolute change in kilograms from baseline. The Wilcoxon rank sum test was used to compare change in body fat at specific time points. Mixed model analysis of variance (MMANOVA) was used to assess differences in percent change between treatment groups over time. The heterogeneous Toeplitz (TOEPH) correlation structure, which assumes that any 2 visits an equal distance apart have the same correlation, was used for consistency with previous modeling of DXA data from the A5005s study.4 This analysis modeled percent change from baseline in limb fat (sum of upper and lower limb fat) and trunk fat. Time was modeled using piecewise variables, where one variable captures the changes from baseline to week 32. The second variable combines with the first variable to provide a slope for changes from week 32 to week 144. Percent change at week 0 was considered to be 0. Jittered data for week 0 were generated using a normal distribution with a mean of 0 and a standard deviation of 0.10.
One hundred fifty-seven participants had DXA at entry. Baseline characteristics were similar between treatment groups and are shown in Table 1. Overall, participants had a median age of 36 years and 51% were white, with a median CD4+ cell count of 260 cells/mm3. Sixty-five participants had week 144 DXA, including 32 who continued to receive their original regimen. Baseline limb fat as measured by DXA was 6.3 kg (interquartile range: 4.2 to 9.4 kg) and baseline trunk fat was 7.5 kg (interquartile range: 4.5 to 11 kg). Weight increased by a median of +2.7 kg (interquartile range: −0.9 to +6.8) for the group as a whole at week 144. Among 13 participants who continued to receive didanosine and stavudine at week 144, the median change from baseline was +4.3 kg, and with zidovudine and lamivudine, the median change from baseline was +3.4 kg (P = 0.6 between groups). Among 7 participants receiving nelfinavir without efavirenz at week 144, the median change from baseline weight was +3.6 kg; among 14 receiving nelfinavir + efavirenz, the median change was +2.3 kg; and among 11 participants on efavirenz without nelfinavir, the median change was +3.5 kg (P = 0.7 between groups).
Limb Fat Changes
For the group as a whole (intent-to-treat), limb fat increased during the first 32 weeks and subsequently declined (Fig. 1). At week 144, the median change from baseline in limb fat was −12.9% (n = 65; P = 0.08 from baseline), corresponding to a median −0.9-kg change. Median limb fat increased similarly in all treatment arms during the first 32 weeks (as-treated, Figs. 2A, B). Intent-to-treat analyses revealed similar plots over time (not shown). Among 13 participants who continued to receive didanosine and stavudine at week 144, the median change from baseline in limb fat was −32.5% (P = 0.04 from baseline), corresponding to a median −2.8-kg change, and among 19 participants on zidovudine and lamivudine, the median change was +1.4% (P = 1.0 from baseline), corresponding to a median +0.1-kg change (see Fig. 2A). After week 32, limb fat changed by −1.7% per year for zidovudine and lamivudine and by −19.0% per year for didanosine and stavudine (MMANOVA, P < 0.0001).
Among 7 participants receiving nelfinavir without efavirenz at week 144, the median change from baseline limb fat was −23.8% (P = 0.69 from baseline), corresponding to a median −0.8-kg change; among 14 receiving nelfinavir + efavirenz, the median change was −21.4% (P = 0.05 from baseline), corresponding to a median −2.0-kg change; and among 11 participants on efavirenz without nelfinavir, the median change was +2.4% (P = 0.90 from baseline), corresponding to a median +0.2-kg change (see Fig. 2B). When all 21 nelfinavir recipients were combined, the median change in limb fat at week 144 was −23.8% (P = 0.05 from baseline), corresponding to a median −1.9-kg change. After week 32, there was an additional decrease in limb fat of −8.7% per year in the combined nelfinavir and nelfinavir + efavirenz group as compared with efavirenz alone, after adjusting for nucleoside backbone (MMANOVA, P = 0.03). Focusing only on participants receiving ZDV + 3TC (Fig. 3A), after week 32, limb fat changed by +2.7% per year with efavirenz and by −7.9% per year for the combined nelfinavir and nelfinavir + efavirenz group (MMANOVA, P = 0.03).
Trunk Fat Changes
For the group as a whole (intent-to-treat), trunk fat increased during the first 32 weeks and was maintained over time (see Fig. 1). At week 144, the median change from baseline in trunk fat was +15.5% (n = 65; P < 0.001 from baseline), corresponding to a median +1.1-kg change. Median trunk fat increased similarly in all treatment arms during the first 32 weeks (as-treated; see Figs. 2C, D). Among 13 participants who continued to receive didanosine and stavudine at week 144, the median change from baseline in trunk fat was +6.4% (P = 0.6 from baseline), corresponding to a median +0.2-kg change, and among 19 participants on zidovudine and lamivudine, the median change was +17.4% (P = 0.003 from baseline), corresponding to a median +1.6-kg change (see Fig. 2C). After week 32, trunk fat changed by +7.0% per year for zidovudine and lamivudine and by −3.4% per year for didanosine and stavudine (MMANOVA, P = 0.05).
Among 7 participants receiving nelfinavir without efavirenz at week 144, the median change from baseline trunk fat was +2.7% (P = 0.4 from baseline), corresponding to a median +0.3-kg change; among 14 receiving nelfinavir + efavirenz, the median change was +6.9% (P = 0.5 from baseline), corresponding to a median +0.6-kg change; and among 11 participants on efavirenz without nelfinavir, the median change was +32.0% (P = 0.01 from baseline), corresponding to a median +3.5-kg change (see Fig. 2D). When all 21 nelfinavir recipients were combined, the median change in trunk fat at week 144 was +6.4% (P = 0.3 from baseline), corresponding to a median +0.3-kg change. Focusing only on participants receiving ZDV + 3TC (see Fig. 3B), after week 32, trunk fat changed by 18.9% per year with efavirenz and by −3.5% per year for the combined nelfinavir and nelfinavir + efavirenz group (MMANOVA, P < 0.0001).
This long-term, 144-week, as-treated, follow-up analysis of participants who received nelfinavir, efavirenz, or both, combined with zidovudine and lamivudine or stavudine and didanosine, demonstrates findings largely consistent with the short-term intent-to-treat results previously reported.4 Importantly, for limb fat, there was a persistent advantage to receipt of zidovudine and lamivudine versus stavudine and didanosine that was maintained over time. After an initial increase in limb fat in all groups, participants receiving zidovudine and lamivudine tended to return to baseline values, whereas those receiving stavudine and didanosine tended to continue to lose limb fat over time.
For the group as a whole, limb fat loss seemed to level off from week 80 onward. This is in contrast to the results of Mallon and colleagues,10 who observed a continued decline from weeks 96 through 144 in their participants, 65% of whom were receiving stavudine. This could be a result of fewer DXA observations in their study, which numbered a total of 40 from week 96 onward as compared with 303 in the current study, which includes 150 in the on-treatment analysis. For the group as a whole, the median change in total limb fat was modest, approximately 900 g over 3 years, and was not statistically significant from baseline.
Our long-term results provide additional evidence that the protease inhibitor nelfinavir can contribute to limb fat loss. In our original intent-to-treat 64-week analysis, randomization to nelfinavir as opposed to efavirenz resulted in greater limb fat loss over time,4 although the magnitude of this effect was small in comparison to the effect of nucleoside assignment. Participants who received nelfinavir + efavirenz were not included in that analysis. In the current long-term as-treated analysis, to investigate a persistent nelfinavir effect, we included all participants who received nelfinavir and compared them with those only receiving efavirenz and found similar results. When focusing only on the participants receiving zidovudine and lamivudine (see Fig. 3A), receipt of efavirenz alone was superior to receipt of nelfinavir (with or without efavirenz). Notably, participants who received the 3-drug combination of zidovudine, lamivudine, and efavirenz had well-maintained limb fat over time. A recent study found greater absolute amounts of limb fat with a tenofovir-emtricitabine-efavirenz regimen at week 48 as compared with a zidovudine-lamivudine-efavirenz regimen16 but did not include baseline measures of limb fat or longer term results, thus leaving open the question of whether tenofovir has a lesser tendency to promote limb fat loss over time than zidovudine. It is important to note that the percentage or absolute amount of limb fat loss that equates to a clinically relevant change has not yet been defined.
Trunk fat increased over time in the group as a whole. In particular, receipt of zidovudine and lamivudine resulted in marginally greater increases over time compared with stavudine and didanosine, and receipt of efavirenz resulted in greater increases in trunk fat than receipt of nelfinavir (with or without efavirenz). Because DXA does not distinguish between visceral trunk fat and subcutaneous trunk fat, our results cannot address the partitioning of these fat changes or their potential metabolic and cardiovascular implications.
A limitation of our study is its relatively small sample size, an issue aggravated by the frequent antiretroviral changes. Nevertheless, it is still the largest longitudinal follow-up of a randomized trial with baseline information that has been reported. Another significant limitation is the relevance of the antiretroviral regimens evaluated in the ACTG 384 study in the current therapeutic era, particularly in the developed world. This is an unavoidable issue when more than 20 new antiretroviral medications have been approved in the past decade, many of which have a better therapeutic index than previously introduced agents. Our study addresses the relative differences between the drug regimens used, which is only one factor associated with this complex syndrome. In addition to drug effects on adipose tissue, it is possible that the differences seen could be related to drug effects on appetite and dietary composition or attributable to pharmacokinetic or pharmacodynamic influences.
As the HIV therapeutics field matures and less toxic antiretroviral combinations with more durable effects are developed, it should be easier to maintain patients on the same regimen for longer periods of time and still have relevant results using contemporary regimens at the end. Our findings are still particularly relevant for the developing world, however, where zidovudine- and stavudine-based regimens are the most frequently recommended nucleoside backbones.15 The combination of stavudine and didanosine in combination with nelfinavir or lopinavir remains the second-line regimen in some sub-Saharan countries such as Botswana. This combination was associated with significantly greater virologic failure12 and demonstrated unfavorable metabolic and body composition outcomes in our study.
We conclude that with long-term use, there are persistent advantages with regard to limb fat with the use of zidovudine and lamivudine compared with stavudine and didanosine as well as a persistent advantage with the use of regimens containing efavirenz plus nucleosides compared with regimens containing the protease inhibitor nelfinavir plus nucleosides (with or without efavirenz). In particular, the 3-drug combination of zidovudine, lamivudine, and efavirenz seems to maintain limb fat well during long-term follow-up. Long-term studies that include baseline DXA evaluations are needed with more contemporary regimens.
The authors wish to thank the following A5005s study team members who contributed to the design and conduct of the trial: Robert A. Zackin, ScD (Statistical and Data Analysis Center, Harvard University, deceased), Thomas A. Buchanan, MD (University of Southern California), Kevin Yarasheski, PhD (Washington University), Sally Snyder, BS (Social and Scientific Systems), and Jeff Taylor (ACTG Community Constituency Group representative, San Diego, CA). They are grateful to all the participants who volunteered for this study. The authors acknowledge the invaluable assistance of Kathy L. Flynn and Gina-Bob Dubé with managing the references and of Abby Shevitz, MD (deceased), and Jodi L. Forand, BS, of the Friedman School of Nutrition Science and Policy at Tufts University for DXA scan management. The following ACTG investigators and sites participated in this study: Barbara M. Gripshover, MD, and Kathleen Burgner, RN, BSN (Case Western Reserve University [A2501]), Ian Frank, MD, and Isabel Matozzo, RN (University of Pennsylvania, Philadelphia [A6201]), Laura Laughlin, RN, and Diane Gochnour, RN (Ohio State University [A2301]), Grant U01 AI025924; Tammy Powell, RN, ACRN, and Pamposh Kaul, MD (University of Cincinnati [A2401]), Grant AI 25897; Debra deMarco, RN, and John Stoneman, RN (Washington University, St. Louis [A2101]), Grants AI 25903 and RR-00036; Connie A. Funk, RN, MPH, and Kathleen E. Squires, MD (University of Southern California [A1201]), Grant U01 A127673; Margaret A. Fischl, MD, and Leslie Thompson, RN, BSN (University of Miami [A0901]), Grant AI27675; Mitch Goldman, MD, and Helen Rominger, RN, MSN, FNP (Indiana University Hospital [A2601]), Grants U01 AI25859 and RR-00750; Linda Meixner, RN, and Tari Gilbert, NP (University of California, San Diego [A0701]), Grant AI27670; Christine Fietzer and Kathy A. Fox (University of Minnesota [A1501]), Eileen Chusid, PhD, and Donna Mildvan, MD (Beth Israel Medical Center) (Mount Sinai Medical Center, NY [A1801]), Grant AI 46370; University of North Carolina (A3201), Howard University (A5301), Lynn Dumas, RN (Beth Israel-Deaconess Hospital), Betsy Adams, RN (Boston Medical Center), Theresa Flynn, NP (Massachusetts General Hospital), Harvard University (A0101), Mallory Witt, MD, and Tomasa Maldonado, RN, BS (Harbor General/UCLA School of Medicine [A0601]), Juan J.L. Lertora, MD, PhD, and Rebecca Clark, MD, PhD (Tulane University [A1701]), Grants U01 AI3844 and RR-05096; Pat Cain, RN, and Sylvia Stoudt, RN (Stanford University [A0501]), New York University/Bellevue (A0401), Harold A. Kessler, MD, and Ruth M. Davis, RN (Rush-Presbyterian/St. Lukes [A2702]), Grant UO1 AI025915; Santiago Marrero, MD, and Irma Torres, RN (University of Puerto Rico [A5401]), Grant A134832-12; Diane Havlir, MD, and Jody Lawrence, MD (San Francisco General Hospital [A0801]), Grant RR-00083; University of Hawaii (A5201), University of Rochester Medical Center (A1101), Clifford Gunthel, MD, and Ericka Patrick, RN (Emory University [A5802]), Michael J. Borucki, MD, and Gerianne Casey, RN (University of Texas, Galveston [A6301]), Grant AI32782; Ilene Wiggins, RN, and Dorcas Baker, RN (Johns Hopkins University [A0201]), Grants RR-00052 and AI27668; Carol Dukes-Hamilton, MD, and Shelia Tedder, RN (Duke University Medical Center [A1601]), Valery Hughes, FNP, and Todd Stroberg, RN (Cornell University [A2201]), Grant AI 46386; Sally Canmann, BSN, and Cathi Basler, MSN (University of Colorado Health Sciences Center [A6101]), Beck Royer, PA-C, and N. Jeanne Conley, RN (University of Washington [A1401]), Grant AI 27664.
© 2007 Lippincott Williams & Wilkins, Inc.