American Journal of Clinical Nutrition, Vol. 86, No. 3, 707-713,
September 2007
© 2007 American Society for Nutrition
ORIGINAL RESEARCH COMMUNICATION |
An 18-mo randomized trial of a low-glycemic-index diet and weight change in Brazilian women1,2,3
Rosely Sichieri,
Anibal S Moura,
Virginia Genelhu,
Frank Hu and
Walter C Willett
1 From the Department of Epidemiology, Institute of Social Medicine (RS), the Department of Physiology, Institute of Biology (ASM), and the Department of Clinical and Experimental Medicine (VG), State University of Rio de Janeiro, Rio de Janeiro, Brazil; and the Department of Nutrition, Harvard School of Public Health, Cambridge, MA (FH and WCW)
2 Supported by grant R03 TW005773-03 from the National Institutes of Health and grant 500404/2003-8 from the Brazilian National Research Council–CNPq
3 Reprints not available. Address correspondence to R Sichieri, Department of Epidemiology, Institute of Social Medicine, State University of Rio de Janeiro, Rua São Francisco Xavier, 524,7° andar, Bloco E, Cep 20550-012, Rio de Janeiro, RJ, Brasil. E-mail: sichieri{at}ims.uerj.br.
 |
ABSTRACT
|
|---|
Background: Despite interest in the glycemic index diets as an approach to weight control, few long-term evaluations are available.
Objective: The objective was to investigate the long-term effect of a low-glycemic-index (LGI) diet compared with that of a high-glycemic-index (HGI) diet; all other dietary components were equal.
Design: After a 6-wk run-in, we randomly assigned 203 healthy women [body mass index (in kg/m2): 23–30] aged 25–45 y to an LGI or an HGI diet with a small energy restriction. The primary outcome measure was weight change at 18 mo. Secondary outcomes included hunger and fasting insulin and lipids.
Results: Despite requiring a run-in and the use of multiple incentives, only 60% of the subjects completed the study. The difference in glycemic index between the diets was 
35–40 units (40 compared with 79) during all 18 mo of follow-up, and the carbohydrate intake from energy remained at 60% in both groups. The LGI group had a slightly greater weight loss in the first 2 mo of follow-up (–0.72 compared with –0.31 kg), but after 12 mo of follow-up both groups began to regain weight. After 18 mo, the weight change was not significantly different (P = 0.93) between groups (LGI: –0.41 kg; HGI: –0.26 kg). A greater reduction was observed in the LGI diet group for triacylglycerol (difference = –16.4 mg/dL; P = 0.11) and VLDL cholesterol (difference = –3.7 mg/dL; P = 0.03).
Conclusions: Long-term weight changes were not significantly different between the HGI and LGI diet groups; therefore, this study does not support a benefit of an LGI diet for weight control. Favorable changes in lipids confirmed previous results.
Key Words: Low-glycemic-index diet • weight change • Brazilian women
|
INTRODUCTION
|
|---|
In prospective studies, a diet with a high glycemic load, the combination of a high glycemic index (HGI) and a high carbohydrate intake, has been an important risk factor for high fasting triacylglycerol concentrations (1), type 2 diabetes (2, 3), and coronary heart disease (4). However, intervention studies in healthy persons are limited. In a nonrandomized follow-up (average: 4 mo) of children attending a program of obesity treatment, children assigned to a low-glycemic-index (LGI) diet (n = 64) experienced a greater reduction in BMI than did those assigned to a reduced-fat, HGI diet (n = 43) (5). In a randomized study in healthy overweight women aged 20–40 y, ad libitum LGI and HGI diets with similar carbohydrate contents were compared, and no differences in weight and hunger were found after a 10-wk follow-up (6). The difference in the GI of the 2 diets in this study was 20 units.
The purpose of this study was to investigate the effect of a larger difference in GI of 2 diets (40 units)—all other dietary components held equal—on weight and satiety in healthy Brazilian women. Dried beans, a frequent component of the Brazilian diet, have an exceptionally low blood glucose response (7), which allows a larger contrast between diets and a longer trial. The trial also aimed to increase adherence to the diets, a major problem in obesity trials, by recruiting young overweight women instead of obese women, by including a run-in period, and by aiming for a small weight loss.
|
SUBJECTS AND METHODS
|
|---|
Subjects
From October 2003 to September 2004, 414 healthy women with a body mass index (BMI; in kg/m2) of 23–29.9, who were aged 25–45 y, not pregnant, not breastfeeding, had at least one child, and did not anticipate a pregnancy in the next year, were recruited for the study. Women with physician-diagnosed thyroid disease or diabetes or who were menopausal were not eligible to participate; we also excluded those who could not eat beans on a daily basis or who had a particular dislike for them. Recruitment was conducted in 2 primary care centers of the State University of Rio de Janeiro, Brazil. The progress of the women during the study is shown in Figure 1
.
All participants received information about the goals of the study, which aimed at a small weight loss during the follow-up. The study was approved by the Institutional Review Boards of Harvard School of Public Health and State University of Rio de Janeiro. A sample calculation made before the beginning of the study was based on a mean (±SD) difference in BMI of 1.2 ± 2.5, assuming 90% power and a 5% significance level. The needed total sample size was 148 (8). Allowing for noncompliance in both groups (9), the estimated sample size was 172; after further adjustment for an estimated 20% loss during follow-up, the total sample size was estimated to be 206.
Study design
Dietary counseling was based on a small energy restriction (ie, 100–300 kcal), and skipping the diet 1 d/wk was allowed. Individual nutritionist counseling every month with menus and exchange lists was provided. Both diets were designed with 26–28% of energy as fat. For each meal (Table 1), the LGI diets were designed to maintain an average difference of 40 GI units compared with the HGI diet. Calculations were based on published GI values for healthy individuals (10), with white bread as the standard GI of 100%. The overall GI was calculated by multiplying the carbohydrate intake of each food by its GI, summing up the products for all foods and dividing the sum by the total carbohydrate intake. Because sticky rice versus parboiled rice was one of the major determinants of the difference in GI between the 2 diets, beyond the amount of beans, we determined the hydrolysis of the most-reported brand of rice consumed by the women via vitro hydrolysis analysis (11) (Figure 2). The difference in GI between the 2 types of rice (25%) was of a magnitude similar to GI values previously reported (116 compared with 91, with white bread as reference) (10). In vitro analyses were also conducted for foods commonly used in Brazil for which no GI was available, such as okra, guava, cheese bread, and manioc bread.
View larger version (9K):
[in this window]
[in a new window]
|
FIGURE 2.. In vitro starch hydrolysis of the most reported brand of rice consumed by the women in the study. Mean (±SD) of 3 measurements. The P value is for the area under the curve.
|
|
Subjects were instructed to eat 3 meals and 3 snacks according to a 6-d menu plan. Instructions also included limiting to a minimum all candies, added sugar, and sodas, except for the weekly day free of diet. Every month, the portions of staple foods were reduced if the participants reported that they were prescribed too much food.
The initial phase of the study was a 6-wk run-in period, which consisted of 2 wk of an LGI diet followed by 4 wk of an HGI diet (Figure 1
). Of the 414 women recruited, 203 completed the 2 run-in periods and were randomly assigned to an LGI or an HGI diet. The randomization list was computer-generated with blocking.
Measurements
Hunger was measured according to a Likert scale from 1 to 10 (12). Weight and hunger were measured every month, and fasting blood samples were collected at baseline and after 3, 6, 12, and 18 mo. Blood was collected after the subjects had fasted for 10 h, and all measurements were performed in the morning. Height was measured to the nearest 0.5 cm with a wall-mounted stadiometer, and body weight was measured by using the same calibrated digital scale for all participants.
Plasma lipids and glucose were measured by using GoldAnalisa kits with an intraassay CV ranging from 0.9% to 1.2% and an interassay CV ranging from 1.9% to 2.7%. LDL- and VLDL-cholesterol concentrations were calculated according to the Friedewald equation (13), based on triacylglycerol measures. Serum insulin concentration was determined by radioimmunoassay with an ImmuChem 125/RIA kit with an intraassay CV ranging from 4.2% to 8.2% and an interassay CV ranging from 6.4% to 8.8%. Relative insulin resistance [homeostasis model assessment of insulin resistance (HOMA-IR)] was estimated according to the formula [glucose (in mmol/L) x insulin (in µIU/mL)/22.5]. A HOMA-IR > 2.5 was considered to indicate insulin resistance.
Food intake, based on a food frequency questionnaire developed for and validated in the adult Brazilian population (14), was measured at the beginning of the run-in period and after 3, 6, 12, and 18 mo of follow-up.
Data analysis
The intention-to-treat analysis included all subjects, regardless of compliance. Hunger and weight changes over time for parallel groups with repeated measurements were determined by using PROC MIXED in SAS, including the baseline measure as a covariate (version 8.2; SAS Institute Inc, Cary, NC). Baseline characteristics of the 2 groups were compared by using Student's t test or the chi-square test. Hunger scales at each main meal and the sum of the 3 scales were compared between the 2 groups. Because of the nonlinear weight change observed, the model incorporated a quadratic term (time x time) variable. Change over time was measured by the interaction between time and type of diet. Because diet x time interactions for weight change were not significantly different, models were reduced. A secondary analysis excluded the 18-mo follow-up, when women who were not actively attending returned for the last visit. Blood lipids in this secondary analysis changed linearly; therefore, the quadratic term was not incorporated in the models. When more than one measurement was available per person per period, only the first measurement was included in the analysis.
Residual plots of all models were examined, and their distribution did not show major deviations from regression assumptions. Energy intake, average GI, average glycemic load, fiber, and selected food items related to compliance were estimated by using a food-frequency questionnaires. Baseline intake was compared by using Student's t test. Statistical analysis for changes during follow-up tested the time x diet variable, with time = 0 for baseline and time = 1 for follow-up.
|
RESULTS
|
|---|
The 414 women who initiated the run-in had a race distribution not significantly different from those who were ultimately randomly assigned, but were less educated. Race distribution among those who initiated the run-in was 51% white, 30% mulatto, and 18% black; 33% had <4 y of schooling. Characteristics of those women randomly assigned to the 2 diet groups are shown in Table 2. No significant differences in any of the characteristics were observed between groups. The LGI group reported a significantly higher glycemic load and GI during follow-up (Table 3). Mean values of the other dietary components were not significantly different between the LGI and HGI groups (Table 3).
View this table:
[in this window]
[in a new window]
|
TABLE 3. Dietary changes from baseline between the low-glycemic-index (LGI) and high-glycemic-index (HGI) diet groups1
|
|
Losses to follow-up during the 18-mo period were 38% in the LGI group and 41% in the HGI group; 5 losses (6%) were due to pregnancy (Figure 1). All women not showing up at scheduled appointments were called and invited to be rescheduled. The main reason given for not returning was an overly restricted diet. The average number of appointments in the 2 dietary groups was 13. Dropouts were younger (35.6 compared with 38.2 y; P = 0.001), lost less weight during the run-in period (0.52 compared with 1.10 kg; P = 0.005), were less educated (P = 0.06), had a lower income (P = 0.07), and had a greater total hunger score at baseline (11.6 compared with 10.3; P = 0.04). However, dropouts were not significantly different from those followed in relation to race and BMI at baseline. Adherence to treatment (completing >10 appointments) was greater in women in the LGI group than in the HGI group (61% compared with 46%; P = 0.0006).
The number of women followed up at specific visits and crude mean changes in weight loss and the hunger scale from baseline are shown in Table 4. Mean weight loss and reduction in hunger (sum of Likert scale ratings completed before all main meals) were not significantly different between the LGI and HGI groups. Similar findings were seen for the reduction in hunger at each meal (data not shown). Estimated changes based on the crude data in Table 4 during the 18 mo of follow-up are shown in Figure 3. The P values in Figure 3 for the time x diet variable indicate changes over time, whereas the P values in Table 4 reflect differences at the 18-mo time point. Both analyses indicated that the effects of diet were not significantly different. Exclusion of those who were dropouts but were weighed at the last visit did not change the results substantially (weight change before exclusion: 0.31 kg compared with 0.21 kg, P = 0.18; weight change after exclusion: 0.68 kg compared with 0.96 kg, P = 0.10). Thus, in models that excluded the time x diet interaction (P > 0.30) and excluded women who were not actively attending, the constant difference over time between the LGI and HGI groups was –0.013 kg (P = 0.94).
View this table:
[in this window]
[in a new window]
|
TABLE 4. Weight changes and mean changes in hunger scale during follow-up, according to low-glycemic-index (LGI) and high-glycemic-index (HGI) diet
|
|
View larger version (13K):
[in this window]
[in a new window]
|
FIGURE 3.. Estimated changes in weight and hunger scales based on a repeated-measurement analysis with baseline values as covariables. Models also included age, center, time, and time x time and time x diet interactions. The P values represent the time x diet interaction.
|
|
The LGI diet reduced triacylglycerol at all measurement time points until 12 mo, but the only statistically significant effect of the diet was the lower VLDL-cholesterol concentration with the LGI diet (P = 0.03; Table 5). After the last observations were excluded, these effects were even stronger (Table 5). When fiber intake at 3 mo was included in the model and the data analysis was restricted to 12 mo, the reductions in total cholesterol and LDL cholesterol became statistically significant (P = 0.009 for total cholesterol and P = 0.01 for LDL cholesterol).
View this table:
[in this window]
[in a new window]
|
TABLE 5. Fasting serum blood lipids during the low-glycemic-index (LGI) and high-glycemic-index (HGI) diets in women1
|
|
At baseline, 3.7% of the women had insulin values >20 µU/mL. No significant differences in fasting serum glucose, insulin, and HOMA-IR were observed between dietary interventions at 3 mo (Table 6).
View this table:
[in this window]
[in a new window]
|
TABLE 6. Serum fasting insulin, glucose, and homeostasis model assessment of insulin resistance (HOMA-IR) with the low-glycemic-index (LGI) and high-glycemic-index (HGI) diets
|
|
|
DISCUSSION
|
|---|
In the nonobese women in the present study, an LGI diet did not facilitate long-term weight loss compared with an HGI diet. After an initial small weight loss, both dietary groups began to regain weight by 12 mo, and the LGI group had regained almost all of the weight lost by the end of the study. The magnitude of weight loss was small, possibly because the study aimed at a small slow weight loss, with the rationale that a small long-term negative energy balance would not elucidate metabolic changes for weight regain, and, as a consequence, compliance would be facilitated. However, compliance in our study was only slightly greater than adherence rates observed in trials with popular diets such as Atkins (carbohydrate restriction), Zone (macronutrient balance), Weight Watchers (calorie restriction), and Ornish (fat restriction) (15). For all 4 of these diets, self-rated adherence after 4 mo of follow up was <50%. Also, only 65% of those on the Zone and Weight Watchers diet and only 50% of those on the other 2 diets finished 1 y of follow-up. In our study, 60% of the subjects finished the 18-mo follow-up for both diets, and adherence (completing >10 appointments) was greater in women in the LGI diet group.
A major limitation of the study was the high rate of losses to follow-up, which was greater than expected in this selected population. We anticipated a higher adherence rate because a weight-loss program was not readily available at public primary health clinics, and the diets were based on commonly used foods. After 3 mo of follow-up we incorporated group activities and rewards for those who kept appointments, but even these strategies had no further effect. Losses to follow-up may not explain the lack of long-term effect of the LGI diet given that results were unchanged when we excluded the 18-mo follow-up, which included women who had stopped active participation. Also, the difference in the GI of diets in the 2 arms of the trial was >35 units over time, according to the food-frequency questionnaires, and favorable changes in serum lipids confirmed previous results (6), which indicated that the pattern of dietary intervention proposed was maintained in both arms of the trial. Weight regain after the 12-mo follow up was associated with an increase in energy intake. This finding has been seen repeatedly in weight-loss trials.
As expected, the mean reported consumption of women in the HGI diet group had a GI and a glycemic load only slightly higher than the GI and the glycemic load at baseline. The LGI diet appeared to be more beneficial than did the HGI diet with regard to weight loss and appetite, measured by the Likert scale only in the first 2 mo of follow-up. Hunger increased in the HGI diet group in the first month, and, curiously, there was a statistically significant reduction in hunger over time with both diets, even after 12 mo when energy intake increased. A possible explanation is that women felt less hungry because they were eating more.
Few studies that manipulated the GI or glycemic load had isocaloric meals differing only in the GI, as in the present study. One of these isocaloric studies was a 10-wk randomized study of 45 subjects with ad libitum intake. Subjects in this study lost 1.9 kg with the LGI diet and 1.3 kg with the HGI diet, but the difference was not statistically significant. However, large favorable changes in lipids were found and insulin and HOMA showed a nonsignificant decline with both the LGI and HGI diets (6). Our results from a much larger sample indicated no significant change in insulin or HOMA. The characteristics of our study population may explain the lack of change in insulin concentrations. Obese women were excluded, and only 3.7% of women had high insulin values at baseline. Wolever and Mehling (16) showed that an LGI diet increased insulin secretion in subjects with impaired glucose tolerance, and weight loss was greater with an HGI diet after a 4-mo follow-up. These findings may explain why the initial greater change in weight observed with the LGI diet was followed by a resistance to further weight loss. Insulin is an anabolic hormone, and its increase poses an extra difficulty for weight loss.
The low frequency of insulin resistance at baseline in our study population also may explain the lack of efficacy of the LGI diet. In a small clinical trial (17), participants with high baseline insulin concentrations lost more weight with the LGI diet, and the reverse was observed in those with low insulin concentrations at baseline.
Our results do not support the hypothesis that an LGI diet enhances weight-loss success, and existing evidence of other benefits was confirmed. The possibility that LGI diets would be effective for weight control mainly among insulin-resistant individuals could not be tested and will require further study with a greater percentage of insulin-resistant individuals.
|
ACKNOWLEDGMENTS
|
|---|
The authors' responsibilities were as follows—RS and WCW: designed the study, interpreted the results, and wrote the paper; ASM: coordinated the storage and measurement of the biochemical samples and interpreted the results; VG: provided clinical support and interpreted the results; FH: helped design the study. All authors declared that they participated in the study and that they saw and approved the submitted version of the manuscript.
All authors agreed to sign a transfer of copyright agreement and had no conflicts of interest.
|
REFERENCES
|
|---|
- Liu S, Manson JE, Stampfer MJ, et al. Dietary glycemic load assessed by food-frequency questionnaire in relation to plasma high-density lipoprotein cholesterol and fasting plasma triacylglycerols in postmenopausal women. Am J Clin Nutr 2001;73:560–86.[Abstract/Free Full Text]
- Salmeron J, Ascherio A, Rimm EB, et al. Dietary fiber, glycemic load, and risk of NIDDM in men. Diabetes Care 1997;20(4):545–50.[Abstract]
- Salmeron J, Manson JE, Stampfer MJ, Colditz GA, Wing AL, Willett WC. Dietary fiber, glycemic load, and risk of non-insulin-dependent diabetes mellitus in women. JAMA 1997;277(6):472–7.[Abstract/Free Full Text]
- Liu S, Willett WC, Stampfer MJ, et al. A prospective study of dietary glycemic load, carbohydrate intake, and risk of coronary heart disease in US women. Am J Clin Nutr 2000;71(6):1455–61.[Abstract/Free Full Text]
- Spieth LE, Harnish JD, Lenders CM, et al. A low-glycemic index diet in the treatment of pediatric obesity. Arch Pediatr Adolesc Med 2000;154(9):947–51.[Abstract/Free Full Text]
- Sloth B, Krog-Mikkelsen I, Flint A, et al. No difference in body weight decrease between a low-glycemic-index and a high-glycemic-index diet but reduced LDL cholesterol after 10-wk ad libitum intake of the low-glycemic-index diet. Am J Clin Nutr 2004;80:337–47.[Abstract/Free Full Text]
- Jenkins DJ, Wolever TM, Taylor RH, Barker HM, Fielden H. Exceptionally low blood glucose response to dried beans: comparison with other carbohydrate foods. Br Med J 1980;281(6240):578–80.[Abstract/Free Full Text]
- Pocock, SJ. Clinical trials. A practical approach. Brisbane, Australia: Wiley & Sons, 1989.
- Sato T. Sample size calculations with compliance information. Stat Med 2000;19:2689–97.[Medline]
- Foster-Powell K, Holt SHA, Brand-Miller JC. International table of glycemic index and glycemic load values: 2002. Am J Clin Nutr 2002;76:5–56.[Abstract/Free Full Text]
- Granfeldt Y, Bjorck I, Drews A, Tovar J. An in vitro procedure based on chewing to predict metabolic response to starch in cereal and legume products. Eur J Clin Nutr 1992;46(9):649–60.[Medline]
- Haber GB, Heaton KW, Murphy D, Burroughs LF. Depletion and disruption of dietary fibre. Effects on satiety, plasma-glucose, and serum-insulin. Lancet 1977;2(8040):679–82.
- Friedewald WT, Levy RI, Fredrickson DS. Estimation of the concentration of low-density lipoprotein in plasma, without the use of preparative ultracentrifuge. Clin Chem 1972;18:499–502.[Abstract]
- Sichieri R, Everhart JE. Validity of a Brazilian food frequency questionnaire against dietary recalls and estimated energy intake. Nutr Res 1998;18:1649–59.
- Dansinger ML, Gleason JA, Griffith JL, Selker HP, Schaefer EJ. Comparison of the Atkins, Ornish, Weight Watchers, and Zone diets for weight loss and heart disease risk reduction: a randomized trial. JAMA. 2005;293(1):43–53.[Abstract/Free Full Text]
- Wolever TM, Mehling C. High-carbohydrate-low-glycaemic index dietary advice improves glucose disposition index in subjects with impaired glucose tolerance. Br J Nutr 2002;87:477–87.[Medline]
- Pittas AG, Das SK, Hajduk CL, et al. A low-glycemic load diet facilitates greater weight loss in overweight adults with high insulin secretion but not in overweight adults with low insulin secretion in the CALERIE Trial. Diabetes Care 2005;28(12):2939–41.[Free Full Text]
Received for publication November 6, 2006.
Accepted for publication April 13, 2007.
This article has been cited by other articles:
|
|
|
|
 
A. A. Aziz, L. S. Kenney, B. Goulet, and E.-S. Abdel-Aal
Dietary Starch Type Affects Body Weight and Glycemic Control in Freely Fed but Not Energy-Restricted Obese Rats
J. Nutr.,
October 1, 2009;
139(10):
1881 - 1889.
[Abstract]
[Full Text]
[PDF]
|
|
|
|
|
|
|
C. E. Smith, K. L. Tucker, N. Yiannakouris, B. Garcia-Bailo, J. Mattei, C.-Q. Lai, L. D. Parnell, and J. M. Ordovas
Perilipin Polymorphism Interacts with Dietary Carbohydrates to Modulate Anthropometric Traits in Hispanics of Caribbean Origin
J. Nutr.,
October 1, 2008;
138(10):
1852 - 1858.
[Abstract]
[Full Text]
[PDF]
|
|
|
TOP
ABSTRACT
INTRODUCTION
