The results of numerous studies suggest that body weight is regulated within a specific range. When lean or obese weight-stable subjects were underfed to achieve a 7.5–10% change in body weight, total energy expenditure (TEE) decreased by ≈12–15% (5–8). Other physiologic changes induced by energy restriction include increased hunger and down-regulation of neuroendocrine functions (9–12). These adaptations together may act to antagonize further weight loss. Genetic factors are known to influence the range in which body weight is regulated (13). However, the increasing prevalence of obesity among genetically stable populations (14) indicates that environment must also play an important role in the regulation of body weight.

The experimental protocol followed a randomized, crossover design involving two 9-d admissions to the GCRC separated by a washout period of 2–6 wk. During the first 2 d of each admission (day −1 and day 0), subjects consumed a self-selected diet ad libitum and baseline measurements were obtained. During the next 6 d (days 1–6), subjects consumed either a high-GI or a low-GI energy-restricted test diet. On days 7 and 8, subjects continued to eat the high- or low-GI diets but were allowed to eat additional foods ad libitum from a private food bar. The study was concluded at dinnertime on day 8, allowing a total of 34 h for ad libitum food consumption. Subjects were required to exercise daily on a treadmill until 10% of TEE had been reached. TEE was determined for each subject by the Harris-Benedict equation multiplied by an activity factor of 1.4 (28). Before and between admissions to the GCRC, subjects were instructed to follow their usual diet and exercise patterns and to avoid caffeine, alcohol, and tobacco products.

Resting energy expenditure (REE) was measured by fasting indirect calorimetry (Vmax 29; SensorMedics Corporation, Yorba Linda, CA) for a total duration of 20–30 min while the subjects rested in bed. Blood samples were obtained after measurement of REE and before breakfast. In addition, blood samples were obtained on day 1 every 30 min by indwelling venous catheter for measurement of plasma glucose and serum insulin responses to breakfast, lunch, and dinner to verify predicted differences in GI (see below).

The macronutrient composition of the high-GI diet was (as percent of energy) 67% carbohydrate, 15% protein, and 18% fat and that of the low-GI diet was 43% carbohydrate, 27% protein, and 30% fat. The energy content of the 2 diets on days 1–6 was 50% of predicted TEE. The high-GI diet was predicted to have a GI approximately double that of the low-GI diet (15). For the average subject (aged 27.9 y, measuring 179.3 cm, and weighing 98.4 kg), the energy content of the diet was 6247 kJ, including breakfast, lunch, and dinner (each 1874 kJ) and a bedtime snack (625 kJ). Each meal was calculated to meet the high- or low-GI target macronutrient distribution. The high- and low-GI diets had similar fiber contents (25.6 and 29.0 g/d, respectively) and energy densities (2.95 and 2.96 kJ/g, respectively). Two rotating menus were used for each diet. Representative menus are shown in Table 1. The foods eaten ad libitum on days 7 and 8 included cold meats (turkey and roast beef), cheese, vegetables (celery and carrots), and condiments (mayonnaise and mustard). To maintain differences in GI, there were also slight differences between the high- and low-GI diets in terms of bread (white and pumpernickel), cereal (granola and bran cereal), fruit (banana and apple), milk (lactase-treated skim milk and 2%-fat milk), fruit drinks (sweetened with glucose and saccharine and sweetened with fructose), and gelatin (sweetened with glucose and saccharine and sweetened with fructose). Palatability of the ad libitum foods was assessed by using a 10-cm visual analogue scale. All meals were prepared under the supervision of the GCRC research dietitians at Brigham and Women’s Hospital, Boston. Nutrient compositions were measured by using FOOD PROCESSOR (version 7.0, 1996; ESHA Research, Salem, OR).

Blood samples were analyzed with the following instruments or kits: plasma glucose, Glucose Analyzer (APEC, Inc, Peabody, MA); serum insulin, Abbott IMx (Abbott Laboratories, Abbott Park, IL); and serum leptin, radioimmunoassay kit (Linco Research, Inc, St Charles, MO). Twenty-four–hour urine collections were analyzed for total nitrogen by using the Kjeldahl method (29) and were also analyzed for creatinine to assess adequacy of collection. Nitrogen balance, reported as the average of values obtained on days 3 and 6, was computed by using the following equation (30):

Serum leptin (Figure 3) decreased more rapidly and to a greater extent during the low-GI diet than during the high-GI diet (P = 0.03). By day 6, leptin had decreased by 50.0 ± 5.5% from baseline with the low-GI diet, compared with 38.3 ± 3.5% with the high-GI diet.

REE (Figure 4) decreased from baseline to the end of the period of energy restriction (mean of days 5 and 6) by 4.6% with the low-GI diet (P = 0.03) and by 10.5% with the high-GI diet (P = 0.005). Mean REE was lower on days 5 and 6 of the high-GI diet (7.38 ± 0.39 MJ/d) than on days 5 and 6 of the low-GI diet (7.78 ± 0.36 MJ/d; P = 0.04).

Nitrogen balance was positive with the low-GI diet and negative with the high-GI diet (25.7 ± 14.1 and −9.7 ± 5.5 mg N · kg⁻¹ · d⁻¹, respectively; P = 0.06).

Weight loss did not differ significantly between the high-GI diet and the low-GI diet, as would be expected from the identical energy contents of the diets. However, serum leptin decreased to a greater extent with the low-GI diet. This difference may be explained by the lower insulin concentrations associated with this diet, because insulin is a leptin secretagogue (31, 32), or by decreased adipocyte glucose metabolism (33). This observation is consistent with that of Jenkins et al (34), who showed a positive association between carbohydrate consumption and leptin concentration during energy restriction. Interestingly, the lower leptin concentration with the low-GI diet occurred without evidence of increased hunger (ad libitum food intake was actually lower with this diet), suggesting a functional improvement in the leptin resistance associated with obesity (35, 36). Further investigation into the significance of this phenomenon is warranted.

REE decreased to a lesser extent with the low-GI than with the high-GI diet. Consistent with this finding, Whitehead et al (37) showed that sleeping metabolic rate fell less rapidly with a high-protein than a low-protein energy-restricted diet. In the present study, the low-GI diet did contain more protein than did the high-GI diet. We note, however, that the present results were not caused by the increased thermic effect of protein relative to that of the other macronutrients because metabolic rate was measured 10 h after the last meal. The preservation of REE during energy restriction with the low-GI diet, together with the differences in voluntary food intake and satiety shown by us and others (21–26), suggests that low-GI diets may be better tolerated than high-GI diets.

Although both diets were protein sufficient (the average subject received 56 or 100 g protein/d), analysis of nitrogen balance suggested that fat tissue was oxidized to a lesser degree, and muscle to a greater degree, with the high-GI than the low-GI diet. This finding is consistent with that of Pasquali et al (38), who observed a more negative nitrogen balance with an isoenergetic, very-low-energy diet with a high carbohydrate-to-protein ratio than with a diet with a low carbohydrate-to-protein ratio. The possible adverse effects of the high-GI diet on body composition can be explained by hormonal responses to meals with different GIs (26). Consumption of a high-GI meal results in relatively high insulin concentrations and low glucagon concentrations. This hormonal response tends to promote uptake of glucose and triacylglycerol in the liver and adipose tissue, limit glycogenolysis and lipolysis, and, therefore, suppress concentrations of glucose and fatty acids in the postabsorptive period. The low circulating concentrations of these major metabolic fuels elicit marked increases in the counterregulatory hormones, some of which have proteolytic actions (39). Thus, decreased mobilization of the primary metabolic fuels with high-GI, energy-restricted diets may induce a series of physiologic events that favor catabolism of lean body tissue.

Two methodologic issues should be discussed. First, although the formula used for calculating nitrogen balance was validated across a variety of dietary conditions (30), the possibility remains that greater delivery of nutrients to the lower digestive tract with the low-GI than with the high-GI diet led to increased nitrogen trapping by colonic bacteria. If this were the case, we may have overestimated nitrogen balance with the low-GI diet. This possibility should be addressed in future studies. Second, previous studies by us and others sought to examine the hormonal and metabolic effects of dietary GI while controlling for other potentially confounding variables, such as macronutrient composition (21–26). However, GI is a complex dietary component that is affected by other factors, including protein and fat (15, 16, 20). Thus, one cannot simultaneously obtain maximum differences in GI and control for confounding variables. We chose to maximize differences between treatment groups because of the relative inaccuracy and imprecision of measuring metabolic rate compared with measuring hormone concentration. We believed that it would be difficult to show significant differences in resting metabolic rate in a relatively brief study using a small number of subjects without varying macronutrients. Therefore, our results cannot be definitively attributed solely to GI. Nevertheless, these results do suggest that the adaptations to energy restriction can be modified by dietary composition. Moreover, the findings are consistent with the possibility that the increased availability of metabolic fuels with a low-GI diet (26) improves tolerance to energy restriction.