[Obesity Discussion]

Statistical Analysis

Analyses were carried out for all randomized participants using an intent-to-treat approach without carrying forward the last observation for the 2 dropouts. Data are presented as mean (SEM). SAS version 9.12 (SAS Institute, Cary, NC) was used for analysis. Changes from baseline at month 3 and month 6 were analyzed by a repeated-measures design approach with respect to treatment and time and treatment x time interactions, with baseline values included as covariates. Data were also analyzed without adjustment for baseline values. Since results by both approaches were similar, we present only the models with adjustment for baseline values. Figure 2 illustrates the weight changes in both percent of initial weight and in kilograms; however, all statistical analyses were performed on absolute changes. Linear regression at baseline (N = 48) was used to generate equations for predicting energy expenditure, and the predicted values were generated using the equation with measured FFM. Differences between predicted and measured energy expenditure were calculated and analyzed by analysis of variance. A normalizing and variance-stabilizing logarithmic transformation was applied to the calculated tail moments for the comet assay.

Figure 2. Absolute and Percentage Weight Loss by Group
Initial weight was recorded as the mean of 5 weights measured weekly during the baseline phase. The change in weight over time was significantly different between the control group and the 3 intervention groups (P<.001) and between the very low-calorie diet, calorie restriction, and calorie restriction with exercise groups (P<.001), but weight loss at week 24 was not significantly different between the very low-calorie diet, calorie restriction, and calorie restriction with exercise groups.

Power and sample size calculations were carried out for the primary end point, 24-hour energy expenditure. Sample size was calculated using different levels of baseline 24-hour energy expenditure, assuming a conservative coefficient of variation (7.5% based on previous chamber studies) and a minimal variability of means. Approximately 12 participants per treatment group were necessary to detect a 15% change in 24-hour energy expenditure from baseline in each group with an 80% power. P<.05 was considered statistically significant.

Two individuals withdrew prior to completion of the study: 1 from the control group at week 4 (personal reasons) and 1 from the very low-calorie diet group at week 5 (lost to follow-up) (Figure 1).

Baseline characteristics of the study participants are listed in Table 1. Percent weight loss from baseline to month 6 in each group was as follows: controls, –1.0% (1.1%); calorie restriction group, –10.4% (0.9%); calorie restriction group with exercise group, –10.0% (0.8%); and very low-calorie diet group, –13.9% (0.7%) (Figure 2). Fat mass was significantly reduced in all 3 intervention groups compared with baseline and compared with the controls at months 3 and 6 (month 6: calorie restriction group, –24% [3%]; calorie restriction with exercise group, –25% [3%]; very low-calorie diet group, –32% [3%]; P<.001). Fat-free mass was significantly reduced in the calorie restriction group (–5% [1%]), the calorie restriction with exercise group (–3% [1%]), and the very low-calorie diet group (–6% [1%]) compared with baseline and controls at month 6 (all P<.001).

Table 1. Baseline Screening Characteristics of Individuals Completing the Study (N = 48)

Fasting insulin levels were significantly reduced from baseline at months 3 and 6 in the calorie restriction and calorie restriction with exercise groups (both P<.01 [Figure 3]) and at month 6 in all intervention groups (all P<.01 [Figure 3]). There were no significant changes in fasting glucose or DHEAS levels in any group. Participants randomized to calorie restriction and calorie restriction with exercise had reduced mean 24-hour core body temperature (Figure 4) at month 6. There was no change in core body temperature in the control or very low-calorie diet groups.

Figure 3. Fasting Plasma Glucose, Insulin, Dehydroepiandrosterone Sulfate, and Triiodothyronine Levels at Baseline, Month 3, and Month 6
Fasting insulin was significantly reduced from baseline values at month 3 (not shown) and month 6 in the calorie restriction and calorie restriction with exercise groups. Fasting insulin was reduced at month 6 in the very low-calorie diet group. Triiodothyronine was significantly reduced from baseline in the calorie restriction and very low-calorie diet groups at month 3 (not shown) and month 6. Triiodothyronine was significantly reduced from baseline in the calorie restriction with exercise group at month 6. SI conversion factors: to convert glucose to mmol/L, multiply by 0.0555; triiodothyronine to nmol/L, multiply by 0.0154. Bars indicate mean values.

Figure 4. Change in Core Body Temperature From Baseline to Month 6 Measured Over 23 Hours Inside a Metabolic Chamber Set to a Mean (SD) Temperature of 22.2°C (0.2°C)
Values are for 7 of 11 controls, 11 of 12 participants in the calorie restriction group, 8 of 12 participants in the calorie restriction with exercise group, and 9 of 11 participants in the very low-calorie diet group. Mean 24-hour temperature and nighttime temperature (2 AM-5 AM) are shown. Average 24-hour temperature was significantly reduced from baseline in the calorie restriction and calorie restriction with exercise groups. Nighttime temperature was significantly reduced from baseline in the calorie restriction with exercise group.

Absolute 24-hour energy expenditure and sleeping energy expenditure were significantly reduced from baseline in the calorie restriction, calorie restriction with exercise, and very low-calorie diet groups (all P<.001 [Table 2]). At baseline, FFM accounted for 86% of the variance in sedentary 24-hour energy expenditure (24-hour energy expenditure [kcal/d] = 596 + 26.8 x FFM; r2 = 0.86, P<.001), whereas fat mass, age, and sex did not statistically account for any additional variance. Compared with predicted 24-hour energy expenditure values, measured daily 24-hour energy expenditure at months 3 and 6 were unchanged in controls and significantly reduced in the calorie restriction, calorie restriction with exercise, and very low-calorie diet groups (Table 2). More specifically, after adjustment for changes in body composition, sedentary 24-hour energy expenditure was unchanged in controls (–18 kcal/d [52 kcal/d]; P>.05), but decreased in the calorie restriction (–135 kcal/d [42 kcal/d]), calorie restriction with exercise (–117 kcal/d [52 kcal/d]), and very low-calorie diet (–125 kcal/d [35 kcal/d]) groups (all P<.008). These data are shown in Table 2 as actual 24-hour energy expenditure minus predicted energy expenditure. Individual data points at month 6 and the baseline regression line for 24-hour energy expenditure vs FFM are presented in Figure 5. When participants from the 3 intervention groups were pooled, adjusted 24-hour energy expenditure values were statistically lower than controls at months 3 and 6 (P<.05).

Table 2. Absolute Energy Expenditures (24-Hour Sedentary and Sleeping) Measured in a Metabolic Chamber At Baseline, Month 3, and Month 6*

Figure 5. Measured 24-Hour Energy Expenditure, Sleep Energy Expenditure, and Fat-Free Mass at Month 6
Correlation between measured 24-hour energy expenditure and fat-free mass at month 6 (24-hour energy expenditure [kcal/d] = 596 + 26.8 x fat-free mass, r2 = 0.86, P<.001) (left) and measured sleep energy expenditure and fat-free mass at month 6 (sleeping energy expenditure = 501 + 20.2 xfat-free mass, r2 = 0.76, P<.001) (right); fat-free mass was the major determinant of sleep energy expenditure. Regression lines are derived from data at baseline in all participants (n = 48) and data markers indicate individual's values at month 6 in the calorie restriction, calorie restriction with exercise, and very low-calorie diet groups.

Since the predicted 24-hour energy expenditure data were derived from just 48 participants, we also compared the 24-hour energy expenditure data from each group to 865 individuals (510 men; 355 women; mean age, 32 years; mean weight, 88.5 kg) measured in a similar metabolic chamber at the National Institute of Diabetes and Digestive and Kidney Diseases in Phoenix, Ariz.36 Importantly, 24-hour energy expenditure was not different between the reference population and the calorie restriction, calorie restriction with exercise, or very low-calorie diet groups at baseline or at any time point in the controls. However, adjusted 24-hour energy expenditure was significantly lower at months 3 and 6 in the calorie restriction, calorie restriction with exercise, and very low-calorie diet groups (all P<.01). Similar to 24-hour energy expenditure, measured sleeping energy expenditure was lower than predicted at months 3 and 6 in the calorie restriction and calorie restriction with exercise groups (Table 2 and Figure 5). There were no significant changes from baseline in the level of spontaneous physical activity or in the thermic effect of food expressed as percentage of energy intake.

Plasma T3 levels were reduced from baseline in the calorie restriction (–10.2 ng/dL [0.15 nmol/L]) and very low-calorie diet (–18.9 ng/dL [0.29 nmol/L]) groups at month 3 (both P<.01) and in the calorie restriction (–8.9 ng/dL [0.13 nmol/L]), calorie restriction with exercise (–4.52 ng/dL [0.07 nmol/L]), and very low-calorie diet (–23.24 ng/dL [0.36 nmol/L]) groups at month 6 (all P<.02). A significant treatment effect for plasma T3 (P = .001; Figure 3) with only a tendency for a time effect (P = .07) was observed. Similar results were found for change in plasma T4 level in response to treatment (P<.05). When the participants in the 3 treatment groups were combined, we observed significant linear relationships between the change in plasma thyroid hormones and deviations in measured 24-hour energy expenditure from predicted values at month 3 only (T3: r = 0.40, P = .006; T4: r = 0.29, P = .05).

Serum protein carbonyl concentrations were not changed from baseline to month 6 in any group (Figure 6). DNA damage was reduced from baseline in the calorie restriction (–0.56 AU [0.11 AU]), calorie restriction with exercise (–0.45 AU [0.12 AU]), and very low-calorie diet (–0.35 AU [0.12 AU]) groups at month 6 (all, P<.005), but not in the controls (Figure 6). This decrease was not statistically different compared with the controls when the 3 treatment groups were combined. We found no significant relationships between the changes in DNA damage and changes in adjusted energy expenditure, fat mass, or body weight.