Mayo Clinic Graduate School of Medicine, Division of Preventive and Occupational Medicine, 200 First Street SW, Rochester, MN 55905. E-mail: Thompson.Warren@mayo.edu
Objective: Studies suggest that high-dairy and high-fiber/low-glycemic index diets may facilitate weight loss, but data are conflicting. The effects on weight loss and body fat of a high-dairy diet and a diet high in dairy and fiber and low in glycemic index were compared with a standard diet.
Research Methods and Procedures: Ninety obese subjects were recruited into a randomized trial of three diets designed to provide a calorie deficit of 500 calories/d over a 48-week period. The study compared a moderate (not low)-calcium diet with a high-calcium diet.
Results: Seventy-two subjects completed the study. Significant weight and fat loss occurred with all three diets. A diet with 1400 mg of calcium did not result in greater weight (11.8 ± 6.1 kg) or fat (9.0 ± 6.0 kg) loss than a diet with 800 mg of calcium (10.0 ± 6.8 and 7.5 ± 6.6 kg, respectively). A diet with 1400 mg of calcium, increased fiber content, and fewer high-glycemic index foods did not result in greater weight (10.6 ± 6.8 kg) or fat (8.5 ± 7.8 kg) loss than the standard diet with 800 mg of calcium. Lipid profile, high-sensitivity C-reactive protein, leptin, fasting glucose, and insulin improved significantly, but there were no significant differences between the experimental diets and the control diet.
Discussion: We found no evidence that diets higher than 800 mg of calcium in dairy products or higher in fiber and lower in glycemic index enhance weight reduction beyond what is seen with calorie restriction alone.
A calorie is a calorie is a calorie. This dictum has been the standard for most investigators in the weight loss field. Neither the percentage of calories from fat, carbohydrate, and protein, nor the type of fat or carbohydrate seems to matter much when it comes to losing weight (1, 2, 3).
However, recent animal data suggest that dietary calcium and dairy products, in particular, may affect fat cell metabolism such that greater weight loss can occur despite an identical calorie intake. Intracellular calcium plays a critical role in the metabolism of the adipocyte (4, 5, 6). Intracellular calcium results in increases in fatty acid synthetase expression and lipogenesis (7). This process is stimulated by parathyroid hormone and 1, 25-dihydroxyvitamin D3. 1, 25-Dihydroxyvitamin D3 treatment of human adipocytes activates lipogenesis and inhibits lipolysis (8) as well as inhibiting uncoupling protein 2 (9). Feeding high-calcium diets to mice suppresses parathyroid hormone and 1, 25-dihydroxyvitamin D3, thus increasing adipose uncoupling protein 2 expression and thermogenesis and stimulating lipolysis (10). This may stimulate adipocyte apoptosis (11). Melanson et al. (12) have reported that higher calcium intakes are associated with higher rates of whole-body fat oxidation measured in a whole-room calorimeter. Thus, there is reason to believe that high dietary calcium may alter adipocyte metabolism and facilitate weight loss. However, a systematic review of randomized trials of calcium/dairy supplementation and bone density in humans showed no weight loss (although the trials were not designed for this purpose) (13).
Glycemic index is a property of foods and mixed meals that may be important in the development of obesity (14). Several studies suggest that diets that have a low glycemic index reduce voluntary food intake at the subsequent meal (15, 16, 17, 18). Two small, short-term studies have suggested that a low-glycemic index diet may enhance weight loss (19, 20). However, Raben (21) has argued that the case for low-glycemic index diets is far from proven. Dietary fiber supplementation improves glucose control but has had mixed success in weight loss (22, 23, 24, 25, 26).
These studies suggest that diet composition may make a difference in weight control independently of total calories. However, due to the conflicting literature and the short-term nature of the clinical trials, we evaluated the effect of a high-dairy diet on weight loss and whether this effect would be enhanced with a high-fiber, low-glycemic index diet in a 48-week randomized trial. Because of the potential impact of a low-glycemic index diet on lipids, glucose, and insulin (14), we also evaluated the effects of the diets on these parameters.
Research Methods and Procedures
Ninety subjects with stable weight (BMI 30 to 40) over the previous 6 months were enrolled in a 1-year randomized trial. Men and women between the ages of 25 and 70 were eligible for the study. Subjects who were lactose intolerant, were pregnant, had a history of an eating disorder, took calcium supplements (and were unwilling or unable to discontinue them for the duration of the study), took medication for diabetes or weight loss, took antidepressant agents associated with weight gain, or had digestive disorders that would prohibit following a high-fiber diet were not eligible for the study. Subjects were recruited by placing posters near employee elevators in the clinic and hospital. Subjects were recruited from October 2001 through April 2002 and were followed for 48 weeks.
Subjects completed a 2-week weight maintenance run-in phase in which they kept daily food records and exercise logs that they reviewed with the dietitian once weekly. There were 28 people who did not complete this phase and were not enrolled in the trial. Patients were instructed to exercise (e.g., brisk walking, treadmill, or exercise bicycle) at least 30 minutes four times each week during the study. After completing the run-in phase, patients were randomized to one of three diets. The first diet was calculated at an energy deficit of 500 calories with 30% fat, 20% protein, and 50% carbohydrate. The diet was designed to provide an average level of calcium and fiber; two servings of dairy were prescribed. (In the maintenance phase, the average calcium intake was 932 mg, and the average fiber intake was 16.2 grams for the whole cohort.) The second diet was the same as the first, except four servings of dairy were prescribed, at least two of which were fluid milk. The third diet was the same as the second, except with an increased amount of fiber (through additional whole grains, fruits, and vegetables) and with a reduction in glycemic index (foods with a glycemic index >100 were strongly discouraged).
Basal energy expenditures based on the Harris-Benedict equation using actual weight with a 40% activity factor were calculated for each subject. Subjects were given a meal plan based on their calorie level using American Diabetes Association exchanges to achieve a 500-calorie deficit. Subjects bought and prepared food on their own. (Neither subjects nor dietitians were blinded to assigned diet.) They completed a food diary daily, which was reviewed weekly with the dietitian. (Subjects who adhered to the plan were seen biweekly in the second one-half of the study; others were seen weekly throughout the study.) Participants recorded minutes of exercise each day. The dietitians discussed problems with the subjects to enhance adherence and provided all subjects with patient education materials designed to enhance weight loss. The Harris-Benedict equation was recalculated at 12, 24, and 36 weeks, and subjects were given new caloric goals to further weight loss. If subjects had lost fewer than 6 pounds at each interval, caloric intake was adjusted downward by 200 calories more than indicated by the Harris-Benedict equation. The difference between average calories prescribed and average calories indicated by the Harris-Benedict equation was 100 calories in all three groups. Subjects followed these diets for 48 weeks.
The primary outcome of the study was weight change at 48 weeks. Secondary outcomes included change in body fat, trunk fat, waist circumference, hip circumference, fasting lipid profile, fasting glucose and insulin, 2-hour glucose and 2-hour insulin, high-sensitivity C-reactive protein (hs-CRP),1 and leptin. Oral glucose and insulin tolerance tests were performed twice at the beginning and twice at the end of the study using a 75-gram glucose load. Fasting lipid profile, hs-CRP, homocysteine, and leptin were measured twice at the beginning of the study and twice at 48 weeks. Weight and waist-to-hip ratio were determined twice at the start of the study, at 12, 24, and 36 weeks, and twice at the end of the study. Resting energy expenditure (REE) by indirect calorimetry and body fat by DXA scan were measured at the beginning, at 24 weeks, and at the end of the study. The average of two measures was used when the variable was measured twice.
Food diaries were analyzed at weeks 11, 23, 35, and 47 for 2 week days and 1 weekend day using Food Processor, version 7.9 (ESHA Research, Salem, OR). Written informed consent was obtained from all subjects, and the study was approved by the Institutional Review Board (Mayo Clinic, Rochester, MN). The funding organization played no role in collecting or analyzing data, preparing the manuscript, or deciding to submit for publication.
Laboratory methods are previously described (27). Leptin was measured by the Human Leptin double antibody radioimmunoassay kit (Linco Research, Inc. St. Louis, MO 63, 304). Waist circumference was measured at the umbilicus. Hip circumference was measured at the widest point of the hip/buttocks area with the tape measure parallel to the floor.
The statistician generated the computer randomization algorithm stratified by sex. Participants were assigned to their groups by the unit secretary of the General Clinical Research Center (otherwise uninvolved with the study). The sequence was concealed until the intervention was assigned. The participants were enrolled by the dietitians. Distributions were characterized by percentages, means, and SDs. Differences between baseline and end-of-study measurements were assessed using the Wilcoxon rank sum test. Differences among groups were determined by the Kruskal-Wallis test. Non-parametric tests were used because many of the factors measured were highly skewed. Three analyses were performed: an intention-to-treat analysis with the last available value carried forward, a completer's analysis (all those who completed the trial), and an adherer's analysis (all subjects who adhered to exercise guidelines and were rated as adherent by the dietitians 75% of the weeks in the study).
Sample Size Calculations
The study was designed to have 80% power to detect a difference in means of 3.1 kg, assuming a common SD of 3.8 kg using one-way ANOVA and a 20% drop-out rate, resulting in 24 in each group (M.B. Zemel, unpublished data). The final drop-out rate was 20% overall, although dropouts were distributed unevenly across the groups (see Table 3). The observed common SD of weight loss was 7.0 kg, which resulted in 80% power to detect a difference of 5.7 kg.
Ninety subjects enrolled in the trial. One subject (Group 2) dropped out immediately due to difficulty obtaining a blood sample. Baseline data are shown in Table 1. The number of subjects who were exercising >100 min/wk before the start of the study was equally distributed among the three groups. There were no significant differences in any of the baseline parameters among the three groups. Two subjects (Group 2) dropped out due to pregnancy (Figure 1). One subject dropped out after week 24 (Group 2) because of weight gain. The remainder (three, four, and seven subjects from Groups 1, 2, and 3, respectively) dropped out because they were unable to comply with weekly dietitian visits and food records. There were no significant adverse events. Seventy-two subjects completed the study (completers), and 53 subjects were compliant with diet and exercise requirements >75% of the time (adherers).
One subject dropped out immediately because of difficulty obtaining blood; n = 29 for DXA and blood values.
Age (years± SD)
42.0 ± 8.8
41.2 ± 9.3
41.1 ± 8.6
Sex (female, male)
Exercise > 100 min/wk prior to study (no. of patients)
98.1 ± 13.3
98.7 ± 11.0
99.1 ± 17.0
167.2 ± 8.5
168.0 ± 7.2
169.1 ± 10.6
35.0 ± 3.1
35.0 ± 3.2
34.5 ± 3.0
Waist circumference (cm)
107.8 ± 9.6
107.1 ± 9.8
109.3 ± 10.4
0.88 ± 0.07
0.88 ± 0.07
0.90 ± 0.07
Body fat (kg; DXA)
50.6 ± 8.2
51.2 ± 9.6
49.6 ± 8.2
Trunk fat (kg; DXA)
27.2 ± 4.7
26.6 ± 5.4
26.6 ± 4.6
5.09 ± 0.83
4.91 ± 0.91
4.88 ± 0.88
2.15 ± 1.97
1.53 ± 0.56
1.62 ± 0.62
1.02 ± 0.24
1.11 ± 0.22
1.12 ± 0.29
3.23 ± 0.81
3.10 ± 0.89
3.02 ± 0.86
Fasting glucose (mM)
5.08 ± 0.39
5.17 ± 0.43
5.32 ± 0.57
Fasting insulin (pM)
63.9 ± 27.8
70.8 ± 52.1
63.9 ± 31.9
37.0 ± 33.2
34.8 ± 20.4
29.2 ± 11.2
High-sensitivity C-reactive protein (mg/L)
4.5 ± 3.5
4.7 ± 3.8
5.6 ± 5.0
Diet, exercise, and compliance data for the 72 subjects who completed the study are given in Table 2. There was no significant difference in calories prescribed or in caloric intake during the study among the three study groups. We calculated the difference between average prescribed calories over the course of the study and reported energy intake (average daily calorie intake over the course of the study) and the difference between REE (by indirect calorimetry) and reported energy intake. There were no significant differences among the three groups using either method among completers. Among adherent subjects, the difference between prescribed calories and reported calorie intake was greater (100–150 kcal/d) in the standard diet group (p = 0.034). Weight loss was similar despite a greater reported calorie deficit. However, the correlation between weight loss and the difference between prescribed calories and reported calorie intake was −0.031.
Calcium intake was significantly lower in the standard group, and fiber intake was higher and high glycemic food intake lower in the high-fiber, low-glycemic index group. Despite efforts to keep percentage fat, percentage protein, and percentage carbohydrate equal in the three groups, there were significant differences due primarily to the high-fiber group having a larger carbohydrate and a lower fat intake. There were no significant differences in exercise minutes among the three groups, nor were there differences in number of days that food intake was recorded. The dietitians also rated the food records for completeness, and there was no difference among the groups (data not shown). Results for these parameters using either the intention-to-treat analysis or the adherer's analysis were similar.
Weight loss was significant and comparable in the three groups (see Table 3). Subjects achieved a 9% weight loss over the 48-week period using the intention-to-treat method (completers lost 11% of body weight, and adherers lost 12%). Most of the weight loss in all three groups occurred in the first 24 weeks (Figure 2). In subjects who completed the trial, the average difference in weight loss between the standard diet group and the high-dairy group was 1.8 kg [95% confidence interval (CI) = −2.0 to 5.6 kg]. The difference between the standard group and the high-fiber group was 0.5 kg (95% CI = −3.7 to 4.8 kg). CIs for weight loss were narrower at 24 weeks (−2.8 to 2.7 and −2.8 to 3.1 kg, respectively). DXA scans demonstrated a significant amount of fat loss and trunk fat loss in all three groups. However, there were no differences among the groups in body fat loss, trunk fat loss, change in waist circumference, or change in hip circumference using any of the methods of analysis (only completer analysis is shown; Table 3).
Table 4 shows the results of the laboratory parameters for the 72 subjects who completed the trial. Each group showed a significant increase in high-density lipoprotein (HDL) cholesterol and significant reductions in total cholesterol, low-density lipoprotein (LDL) cholesterol, fasting glucose, fasting insulin, leptin, and hs-CRP, but there were no differences among the groups. There were no differences among diet groups in the change in 2-hour glucose or 2-hour insulin. Using the intention-to-treat method or analyzing only those subjects who were most adherent to the protocol did not change these results.
p value for differences from baseline to 48 weeks.
p value for differences among the three dietary groups.
For cholesterol, triglycerides, HDL-cholesterol, and LDL-cholesterol, there were 19, 21, and 21 subjects in each group (five subjects in the standard group, four in the high-dairy group, and three in the high-fiber group had changes in lipid medications during the study and were excluded).
This study demonstrated substantial weight loss and fat loss with calorie restriction, increased exercise, and close follow-up with a dietitian. However, high dairy consumption and high-fiber, low-glycemic index consumption did not enhance weight or fat loss relative to a moderate-dairy diet. Strengths of the study include 1-year follow-up, a low drop-out rate (20%) relative to most weight loss studies, and a randomized trial design. The design did not include a fourth group of subjects assigned a diet with a standard calcium intake plus high-fiber and low-glycemic index foods. Although this is a potential limitation, there is no reason to suspect that a high-dairy diet would have interfered with the effect of a high-fiber, low-glycemic index diet.
A number of studies have suggested a benefit of dietary calcium on weight. Analysis of the National Health and Nutrition Examination Survey III dataset found that weight was inversely related to calcium intake (7). The Quebec Family Study (28) found that women who consumed <600 mg calcium daily had significantly higher body weight and percentage body fat than woman consuming more calcium, after controlling for age, energy intake, and socioeconomic status. The differences in men were not significant. The Tromsρ study (29), however, reported no relationship of calcium intake in women with BMI and an unexpected positive relationship in men. Unlike the above cross-sectional studies, the CARDIA study (30) evaluated subjects prospectively. An inverse relationship between dairy products and the subsequent development of obesity was seen for overweight but not normal-weight individuals. However, the Iowa Women's Health Study (31) found no relation between calcium intake and BMI or waist-to-hip ratio.
A systematic review of randomized trials of calcium/dairy supplementation and bone density, some of which demonstrated weight loss, concluded that it has not been demonstrated that calcium causes weight loss (13). Other reviews (not systematic) have reached opposite conclusions (32, 33). A recent trial assigned young women who were not calorie restricted to low (calcium = 742 mg/d), medium (calcium = 1026 mg/d), and high (calcium = 1121 mg/d) dairy intakes and found no significant difference in body weight or fat mass after 1 year (34). In most of the trials, subjects were not calorie restricted; however, three studies of calcium supplementation vs. placebo in conjunction with a calorie restricted diet showed no difference in weight loss (35, 36, 37). The first studied 31 postmenopausal women for 6 months, comparing 515 mg in the placebo group and 1646 mg in the treatment group. Both groups lost 10% body weight and 18% body fat by DXA. The second study (36) evaluated 52 women for 3 months comparing 800 mg in the control group and 1800 in the treatment group. Weight loss was 6% in both groups. The third paper (37) combines data from three randomized, double-blind trials and reports that 1000 mg of calcium supplementation had no effect on weight loss in 100 obese women over 6 months.
Our previous studies did find a significantly greater weight loss (4.47 kg at 24 weeks) or fat loss with a diet higher in dairy calcium (38, 39). There are several possible explanations for the discrepancy between these studies and the present data. Since both studies were relatively small, our earlier studies (n = 32 and 34 patients) could have been false positive trials, or the present study could be a false negative trial. Studies of this nature cannot be double-blind, so dietitian bias could potentially have affected the results of these studies. The differences in calcium intake between groups (700 and 600 mg daily) were similar. However, the daily calcium intake in the control group in the earlier study was 430 mg as compared with 800 mg in the present study (38). This suggests the possibility of a threshold effect. That is, if the calcium intake is <600 mg daily [such a threshold is suggested by the Quebec Family Study (28)], then weight loss might be diminished. In our earlier study (38), the control group lost 6.4% body weight, 9.1% body fat, and 5.2% trunk fat, whereas the high-dairy group lost 10.9% body weight, 14.1% fat loss, and 14% trunk fat loss. However, in the present study, the control group lost 10% body weight, 14.8% body fat, and 17.6% trunk fat. Thus, weight loss and fat loss in the control group in the present study are virtually the same as those seen in the high-dairy groups in both earlier studies. This would suggest that a daily calcium intake of 800 mg or two dairy servings per day may be sufficient to enhance weight loss. The present study did not have a sufficient number of subjects with a calcium intake < 600 mg daily in the standard group to evaluate the possibility of a threshold effect. It should be noted that three studies (35, 40, 41) have not found a difference in weight loss between a low-calcium (500 to 600 mg) and a high-calcium diet. A randomized trial (42) with 1471 postmenopausal women (40% of whom were not overweight) showed that 1 gram of supplemental calcium did not result in weight or fat loss. However, there was a trend (not significant) toward a slightly greater weight loss in those whose baseline calcium intake was <600 mg daily. Thus, further studies are warranted to test this hypothesis. Another possibility for the discrepancy between this study and Zemel et al.'s (38) is the greater difference seen in the present study between calories prescribed and calories reported in the standard group. However, this was seen only in the adherent subjects. Furthermore, if calories ingested were accurately reported, there should have been a significant correlation between the difference (between calories prescribed and calories reported) and weight loss. The correlation was not significant (r = −0.031).
Our study showed no difference in weight loss in subjects following a diet designed to be lower in glycemic index and higher in fiber and dairy products. In a series of double-blind trials, Ryttig and colleagues (22, 23, 24, 25) conducted a number of studies using fiber supplements, which demonstrated a modest increase in weight loss (1 to 3 kg over 6 months). However, another study did not show that fiber supplements resulted in weight loss (26). The present study did not provide fiber supplements, but subjects in the high-fiber group consumed significantly more high-fiber foods than those in the other two groups (Table 2), and the difference was greater than the amount of fiber supplementation in the studies by Ryttig. A prospective study (43) followed 74, 091 women for 12 years and demonstrated that those with the greatest increase in fiber gained 1.52 kg less than those with the smallest increase in fiber. Our study did not have sufficient power to detect such small differences, so it is possible that fiber may result in small increases in weight losses. Larger, long-term randomized studies are indicated to examine this issue.
A small study (n = 14) in adolescents (20) demonstrated a modest improvement in BMI with a low glycemic load diet. A larger (n = 45) 10 week study (44) that was not calorie restricted showed no difference in weight between a low-glycemic index diet and a high-glycemic index diet. Finally, a 4-month trial (n = 24) that was not calorie restricted showed a slightly greater weight loss with a high-glycemic index diet (45). Acarbose (a medication used in the treatment of diabetes that slows glucose absorption in the gut), which should have similar effects to a low-glycemic index diet, has generally not been found to result in weight loss in long-term studies (46, 47, 48). Furthermore, in a comprehensive review, Raben (21) found little evidence for counseling obese subjects to follow low-glycemic index diets. We demonstrated that a diet in which fewer high glycemic foods, bread and starch, and more fruits and vegetables were consumed (Table 2) did not result in additional weight loss. However, glycemic index of individual foods can differ from glycemic index of mixed meals. We could not test the glycemic index of the meals our subjects ingested. Thus, in a free-living study such as ours, it is impossible to be certain that the glycemic index was truly lower in the experimental group. Studies where meals with known glycemic index are provided to subjects over several months would be the ideal way to evaluate the effects of glycemic index on weight loss.
In summary, the results of this study do not refute the dictum that a calorie is a calorie. A diet high in dairy products (whether or not it was also high in fiber and lower in glycemic index) did not enhance weight loss. Further work needs to be done to compare diets low in dairy products (400 to 600 mg daily calcium) with diets that are moderate (800 to 1000 mg) and high (1200 to 1400 mg) in dairy products. This study indicates that high-dairy (vs. moderate dairy) and low-glycemic index diets should not be accepted as being effective without further evidence from long-term randomized trials.
This study was funded by the National Dairy Council. Additional support was provided by Grant M01RR00585 to the Mayo General Clinical Research Center and by the Division of Preventive and Occupational Medicine. N.R.H. is currently employed by General Mills, which makes yogurt; The National Dairy Council has supported a number of studies by M.B.Z., and he has served on speaker panels for the National Dairy Council.