Sixty subjects from 15 ACTG study sites were enrolled in this multicenter trial from February to December 1999. Of the 59 subjects randomly assigned to this double-blind study (there was one inadvertent enrollment of a subject who did not meet entry criteria), 29 subjects received the supplement containing whey protein and 30 received the control supplement. The 2 study groups were well balanced with respect to demographic characteristics at baseline (). The median age for the 7 women and 52 men was 41 y. Most of the subjects were white (66%), and 52 (88%) reported no prior use of intravenous drugs. The 2 groups were also balanced at baseline with respect to CD4 count, Karnofsky performance status, and 3-mo prior antiretroviral treatment, although one subject in the control group was not receiving antiretroviral therapy. The CD4 counts ranged from 33 to 1263 cells/mm3 (median = 344 cells/mm3), and 85% of subjects had a Karnofsky score of 90 or 100. Overall, the median HIV-1 RNA level was 400 copies/mL (range: <50 to 3196 copies/mL).
Baseline characteristics of the study population1
There were 35 subjects (15 treated with whey protein and 20 treated with the control supplement) who had extensively documented weight histories covering up to 50 wk before baseline. In this group, there was no difference (P = 0.67) in the median change in weight during the pretreatment period from the respective whey protein group [0.2 kg; interquartile range (IQR):−0.6, 0.7) and control group (0.5 kg; IQR: −0.9, 1.0), providing evidence of weight stability.
Of 59 eligible subjects, 41 subjects completed the study treatment: 17 in the whey protein arm and 24 in the control group. There was no evidence to suggest that the duration of treatment differed between the 2 study arms (P = 0.17). Five subjects in the whey protein group discontinued the study treatment prematurely because of protocol-defined toxicities. An equal number of subjects in the 2 treatment arms discontinued the treatment because of nonprotocol-defined, low-grade toxicities or clinical events. One subject in the whey protein group stopped the treatment because of the use of a prohibited medication, and another subject stopped treatment because they disliked the taste. Finally, in the whey protein group, 15 subjects had at least grade II gastrointestinal symptoms (eg, nausea, vomiting, bloating, cramps, or diarrhea) compared with 7 subjects in the control group (P = 0.03).
Body-composition measures at baseline and study weeks 6 and 12 are shown in . For the primary outcome, there was no difference in the change in LBM between the whey protein (0.2 ± 1.4 kg) and control (0.3 ± 1.3 kg) groups after 6 wk (P = 0.76, Wilcoxon’s rank-sum test). After 12 wk of study therapy, the respective changes were 0.3 ± 1.4 and 0.3 ± 1.5 kg (P = 0.61) in the 2 groups. Furthermore, there was no change in LBM across the time points by repeated-measures analyses. Similarly, the respective changes in total weight of 0.7 ± 1.8 and 0.7 ± 1.5 kg for the 2 groups at study week 6 were not significantly different (P = 0.96) nor were the changes of 0.8 ± 2.4 and 0.7 ± 2.4 kg at study week 12 (P = 0.63). For fat, the respective changes were 0.4 ± 1.2 and 0.3 ± 0.8 kg for the 2 groups at study week 6 (P = 0.33) and 0.5 ± 1.6 and 0.4 ± 1.7 kg at study week 12, respectively (P = 0.40). Data for 54 subjects were available for the evaluation of the primary endpoint in the intention-to-treat analysis. For 2 of the 54 subjects who did not have LBM data available from week 12, week 6 results were carried forward to week 12. Similar results were obtained when week 6 data were extrapolated by using a constant slope. The as-treated analysis of the 41 subjects who completed the full 12 wk of study therapy showed no significant changes in LBM between the 2 groups (P = 0.26). These outcomes were also analyzed after women were excluded to determine whether there was a sex effect; all changes remained nonsignificant (all P > 0.10; data not shown) without trends suggesting any benefit in terms of LBM, total weight, or fat. Furthermore, baseline BMI was not related to changes in weight in either group by Spearman correlation (high protein: −0.18, P = 0.36; placebo: 0.32, P = 0.12).
Outcome measures with study intervention
Changes in the variables listed in over time were also analyzed by using a repeated-measures analysis with treatment group and time as model terms (PROC MIXED in SAS, with each response variable analyzed separately; ). To examine the presence of interaction, a time-by-treatment interaction term was included in each of the models, but there was no statistically significant interaction in any of the models, possibly because of insufficient statistical power based on the sample size of the groups. Additional analyses showed that there was no statistically significant effect of the treatment groups, but there were significant effects of time for weight, total fat mass, and waist-to-thigh ratio.
Intakes of total energy and macronutrients at baseline and at study weeks 6 and 12 are shown in for 17 subjects in the whey protein arm and 24 in the control group who completed the study. As designed, the change in dietary protein intake was greater in the whey protein group than in the control group at week 6 (68.7 ± 40.4 and −7.7 ± 35.6 g/d; P < 0.001) and at week 12 (64.4 ± 45.3 and −27.6 ± 38.5 g/d; P < 0.01), which resulted in a significantly greater total intake of protein per kilogram of body weight in the whey group at both time points. Despite the significant differences in protein intake between the groups, average total daily protein intake remained well above the RDA (). Also, and as designed, the change in dietary carbohydrate intake at week 6 was lower in the whey protein group (−3.3 ± 109.6 and 85.8 ± 137.6 g/d; P < 0.01), but at week 12 the difference in change was not significant between the groups (−34.4 ± 160.3 and 36.3 ± 108.9 g/d; P = 0.19).
Intake of total calories and macronutrients1
Changes in dietary intake over time were also analyzed by using a repeated-measures analysis, with treatment group and time as model terms (PROC MIXED, with dietary measurements as response variables analyzed separately). To examine the presence of an interaction, the time-by-treatment interaction term was included in each of the 4 models, but there was no statistically significant interaction in any of the models. Again, this result was possibly due to insufficient statistical power. Furthermore, for any of the 4 dietary measures, there was no statistically significant effect of treatment group, but there was a significant (P < 0.01) related effect of time for each of the macronutrients. The same results were reached in the analysis of ranks of the continuous variable in place of the response variable in the model.
As shown in , there were no differences in the total intake of energy per kilogram of body weight (includes both self-selected food plus supplements) between the 2 study groups at study week 6 or study week 12 (P > 0.05 for each comparison). Moreover, the change from baseline in total daily energy intake at study week 12 did not differ significantly between the whey protein (177 ± 991 kcal) and control (183 ± 831 kcal) groups (P = 0.12; ). Although self-selected intake of total energy, protein, carbohydrates, or fat (excludes energy provided by supplements) did not change at study week 6 compared with baseline, and self-selected energy intake decreased less at study week 12 in the whey protein group (−325 ± 1032 kcal/d) than in the control group (−777 ± 818 kcal/d) (P = 0.04). The lack of a significant increase in total energy consumption was due to a decrease in the self-selected energy intake at study weeks 6 and 12 ().
FIGURE 1. Total caloric intake is shown at baseline (week 0) and after 6 and 12 wk of study therapy for the 17 subjects in the whey protein arm and the 24 subjects in the control group who completed the study. Each bar shows the amount of fat, carbohydrate, and (more ...)
Several secondary outcomes were evaluated (). An evaluation of the paired data at week 6 showed that the waist-to-hip ratio decreased in the whey protein group but increased in the control group (P = 0.025). The waist-to-thigh ratio decreased to a greater extent with whey protein (P = 0.014 compared with control), whereas the thigh-to-waist-to-hip ratio increased to a greater extent in the whey protein group (P = 0.004 compared with control). At study week 12, there were no significant changes in these variables (P ≥ 0.09; ). Fasting serum triacylglycerol in subjects with paired data decreased at week 12 by 16 ± 62 mg/dL in the whey protein group and increased by 39 ± 98 mg/dL in the control group (P = 0.035). There were no significant changes in other lipids, fasting glucose, insulin, HOMA-IR, or QUICKI (data not shown).
Another secondary outcome included assessment of the effects of whey protein on immune function. From baseline to week 12, in subjects in whom paired data were available, those treated with whey protein had an average increase in CD4 lymphocytes of 31 ± 84 cells/mL (P = 0.04), whereas those who received the control supplement had a slight decrease in CD4 lymphocytes of −5 ± 124 cells/mL (P = 0.19). The difference in CD4 changes from baseline to week 12 between the 2 study groups was significant (P = 0.03). The change from baseline to week 12 in CD4 counts was also analyzed with week 0 values as a covariate in a nonparametric analysis of covariance model (PROC GLM in SAS with response variable as ranks). In this analysis, the effect of the treatment group was statistically significant (P = 0.04).