In the parent project 105 overweight and mildly obese individuals were recruited from the faculty, staff, and student populations of each of the three participating institutions (University of Tennessee, Purdue University, and the USDA, ARS, Western Human Nutrition Research Center at the University of California, Davis). Seventy nine subjects completed the trial and 63 met the parent study's a priori weekly compliance criteria , however due to missing measurements, the majority of data are reported for 61 subjects. Intent-to-treat analysis was not conducted because there were only two time points: baseline and 12 weeks. Subjects were included in the study if they were: 18-35 years of age, had an initial BMI of 25-34.9 kg/m2, consumed a low-calcium diet at enrollment (< 600 mg calcium/d) from non-calcium-fortified foods and < 800 mg total calcium/d, did not gain or lose more than 3 kg of body weight during the past three months, and did not recently (4 wk) change exercise intensity or frequency. Subjects were excluded from participation of the study if they smoked; required the use of oral anti-diabetic agents or insulin; used obesity pharmacotherapeutic agents and/or herbal or other preparations intended for use in obesity or weight management within the previous 12 wk; used calcium supplements within the previous 12 wk; had a history of significant endocrine, hepatic, or renal disease; were pregnant or lactating; had a recent (past 12 wk) initiation of or change in oral contraceptive; suffered any active form of malabsorption syndrome; or had a history of eating disorders. The project was approved by the Institutional Review Board of each of the three participating institutions. Written informed consent was obtained from all subjects, and the research was conducted in accordance with the ethical standards outlined in the Helsinki Declaration.
This study was designed to determine whether dairy products or calcium supplementation would accelerate weight and fat loss induced by energy restriction in otherwise healthy overweight and obese adults. After enrollment, subjects were studied for a 2-wk lead-in period to establish their current energy requirements and provide an opportunity for baseline dietary and physiological assessment, and then randomized to the following dietary regimens for 12 wk: 1) a control diet providing a 2093 kJ/d deficit (500 kcal/d), 0-1 serving of dairy products/d, 500 mg calcium/d, and a daily placebo supplement; 2) a calcium-supplemented diet identical to the control diet, with the placebo replaced by 900 mg calcium in the form of calcium carbonate; or 3) a high-dairy diet (placebo-supplemented) providing a 2093 kJ/d deficit (500 kcal/d) and containing three daily servings dairy products (milk, cheese, and/or yogurt) to bring the total calcium intake from 500 to 1400 mg/d. The calcium supplemented and placebo arms of the study were conducted in a placebo-controlled, blinded fashion, and the dairy product arm was unblinded by necessity. However, subjects on the high-dairy diet also received a placebo supplement and all groups were treated as active-treatment arms, with pill counts serving as a compliance measurement.
Following instructions from a trained nutritionist during the 2-wk lead-in period, baseline 7-d dietary records were completed by each subject and reviewed for completeness by the project dietitian at each site. Diet records were analyzed by Nutritionist Pro software and were used to provide an initial estimate of an energy intake. The estimate of energy intake was then refined by calculating energy needs using World Health Organization equations for calculating basal metabolic rate, adjusted for Physical Activity Level (PAL) set at low-active (1.4) to provide an estimate of total daily energy expenditure, and reported elsewhere . Based on this initial estimate of energy needs, a food exchange-based diet was prescribed to produce an energy deficit of approximately 2093 kJ/d. The diets for the treatment arms were constructed to provide comparable levels of macronutrient and fiber, to approximate the average consumption in the U.S. (fat, ~35% of total energy, carbohydrates ~49%, ~protein 16%, fiber 2-3 g/1,000 kJ/day). Nutritional supplements were not permitted, and caffeine intake was maintained at a constant level (individualized for each patient, based on baseline assessment). Diets were prescribed and monitored as noted above. Subjects in the high dairy group were permitted to utilize both full-fat and low-fat milk, cheese and yogurt, with the fat accounted for in exchange lists given with each individual diet prescription. Subjects were provided individual instruction, counseling, and assessment from the study dietitian regarding dietary adherence and the development and reinforcement of strategies for continued success. Subjects maintained daily food records throughout the 12 wk intervention period, and compliance was assessed by weekly subject interview and review of the diet diary and pill counts. Compliance criteria are reported elsewhere [30, 31].
Nutrient assessment was computed by averaging 7-d diet records for the 2-wk run-in period with each day of the week represented twice. During the intervention weeks 0-11, subjects filled out 7-day diet records but nutrient intake was assessed from averaging data from 3 of the 7 days (2 progressive weekdays and 1 weekend). The 3 day selection from each week (0-11) started with the first weekday and alternated between Saturday and Sunday. Each 3 days rotated through the week such that by the end of the 12wk intervention, a total of 36 days were included with each day of the week counted 3 times. For example, during week 0, nutrient intake data were selected from Monday, Tuesday, and Saturday and during week 1, nutrient intake data were selected for Wednesday, Thursday and Sunday. Nutrient analysis was completed using Nutritionist Pro Food Processor Plus software. Baseline micro- and macronutrient intakes were determined by averaging the pooled nutrient data from the 2-wk run-in period. Micro- and macronutrient intake during the 12 wk intervention period was determined by averaging the pooled nutrient data from weeks 2-11. Data were not included for baseline or week 1 to allow subjects a 2-wk adaptation period to accurately log dietary intake data.
Measurement of body weight was done during the 2-wk run-in period and weekly thereafter; height was measured at baseline, and waist circumference (WC) at baseline and 12 wk. BMI was calculated as kg/m2. Total fat and lean mass (LM) were assessed via dual energy X-ray absorptiometry (Lunar Prodigy instrument; GE Medical Instruments) at baseline and week 12 of the study. Statistical models were developed to predict the following anthropometric outcomes: 1) percent change in weight; 2) percent change in WC; 3) percent change in % LM and 4) percent change in % body fat (BF).
Physical activity assessment
Participants were instructed not to change their usual physical activity habits during the study and to ensure that physical activity remained unchanged; 3-d physical activity records were collected from all subjects at baseline and week 12. Participants were asked to report any changes to their usual physical activity.
Plasma glucose, TG, LDL cholesterol, and HDL cholesterol were measured at each site's clinical medical laboratory with a Beckman Lxi-725 auto-analyzer. Fasting plasma insulin was measured using a commercially available radioimmunoassay kit (Linco Research, St. Charles, MO) at each site's clinical medical laboratory. Insulin resistance at baseline and 12 wk was calculated using the Homeostasis Model Assessment (HOMA-IR) .
Analysis of the plasma lipidome
Blood was collected from all participants into evacuated tubes containing EDTA, centrifuged immediately (1300 × g, 10 min, 20°C), portioned into aliquots, and stored at -80°C until analyzed. Fatty acid analyses of circulating lipid classes were measured by high-throughput methods developed by Lipomics Technologies, Inc. (West Sacramento, CA) . In brief, the lipids from plasma (200 μl) were extracted using a modified Folch extraction in chloroform:methanol (2:1 v/v) . Each extraction was performed in the presence of a panel of quantitative authentic internal standards. Extracted lipids were dried under N2 gas for for separation by chromatography. Individual phospholipid classes within the extract were separated by high performance liquid chromatography . Individual neutral lipid classes were separated from the extract by thin layer chromatography. Each separated lipid class was collected and dried under nitrogen in preparation for trans-esterification. Lipid classes were trans-esterified in 3 N methanolic HCl in sealed vials under a nitrogen atmosphere at 100°C for 45 min. The resulting fatty acid methyl esters were extracted from the mixture with hexane and prepared for automatic injection into a gas chromatograph by sealing the hexane extracts under nitrogen. Fatty acid methyl esters were separated and quantified by capillary GC using an Agilent 6890 gas chromatograph equipped with a 30-m DB-88 capillary column (Agilent Technologies, Santa Clara, CA) and a flame-ionization detector. Fatty acids of each lipid class were determined quantitatively (μmol/L) and expressed as a percentage of total fatty acids within that class (mol %). Total lipid classes were calculated as the sum of individual fatty acid species for each lipid class fraction. Fatty acids with missing values at 20% or greater were dropped from the analyses and considered not determined (ND).
All statistical procedures were conducted using SPSS version 16 for Windows (SPSS, Chicago, IL). Means ± SD are reported for baseline and 12 wk anthropometric measurements; reported macro- and micronutrient intake; circulating plasma clinical metabolites; circulating total lipid classes; and fatty acid composition of circulating cholesterol ester (CE), free fatty acids (FFA), phosphatidylcholine (PC), and triacylglycerol (TG). All data were examined for normality and transformed as needed prior to conducting statistical analyses. Repeated Measures ANOVA was performed to determine the effect of time, treatment and time x treatment on anthropometric outcomes, dietary intake and circulating lipids. If repeated measures ANOVA demonstrated a significant time effect between baseline and 12 wk variables, a paired sample two-tailed t-test was performed to identify the treatment group that reached significance. ANCOVA was performed to determine differences between treatment groups for both baseline and 12 wk circulating lipid metabolites, clinical metabolites, dietary intake, and anthropometric measurements. Multiple comparisons analysis with a Bonferroni adjustment was used to determine differences among the three treatment groups. Final models on dietary intake were adjusted for site, sex, and energy intake at 12 wk. Models for dependent variables that already included energy in their calculations--Protein, SFA, MUFA and PUFA (% of energy)--were not adjusted for 12wk energy intake. Stepwise regressions were generated to explore the relationships between 12 wk dietary fat composition and 12 wk circulating lipids of the same family (i.e., dietary saturated fat against circulating saturated fatty acids). Details of the criteria followed to generate stepwise regressions are reported elsewhere . Stepwise regressions were also used to examine the relationships among 12 wk diet composition; plasma clinical biomarkers; and lipid metabolites with changes in anthropometrics. The three treatments and sites were converted into two orthogonal variables for use in linear regression. For each stepwise regression, the F statistic probability was set at an alpha between 0.01 and 0.05. Normality for each stepwise regression model was determined by generating normal probability plots of the regression standardized residual. Equal variance for each regression model across each dependent variable was determined by plotting the standardized predicted dependent variable by the standardized residuals. Outlying cases that strongly influenced each stepwise regression model were tested by Cook's distance (D). Data points with larger D values than the rest of the data were considered highly influential and deleted. The models with deleted observations with large D values were re-regressed and compared to ensure the model was statistically relevant and not a product of one highly influential data point. Multi-collinearity between baseline lipids selected by stepwise regression was checked by a Variance Inflation Factor of ≤ 2.0. If variables demonstrated a Variance Inflation Factor greater than 2.0, they were dropped from the final regression model.
The following stepwise regressions were generated: reported intake of % saturated fat (SFA) (% of fat) at 12 wk as a dependent variable and 12 wk circulating plasma 14:0, 15:0, 16:0, and 18:0 of each lipid class as independent variables; reported intake of % monounsaturated fat (MUFA) (% of fat) at 12 wk as a dependent variable and 12 wk circulating plasma 16:1(n-7), 18:1(n-7), 18:1(n-9) of each lipid class as independent variables; and reported intake of % polyunsaturated fat (PUFA) (% of fat) 12 wk as a dependent variable with 12 wk circulating plasma 18:2(n-6), 18:3(n-6), 20:3(n-6), 20:4(n-6), 18:3(n-3), 20:5(n-3), and 22:6(n-3) of each lipid class as independent variables. Stepwise regressions were also generated to determine the relationships between 12 wk dietary fat composition (SFA, MUFA, and PUFA) as a percent of energy and changes in body composition.
For each statistical analysis, only fatty acids for each lipid class with mean abundances of 1.0% or greater were analyzed. Estimation of desaturase and elongase enzymatic activities involved in fatty acid metabolism as product-to-precursor ratios were also analyzed. The ratios of circulating 16:1(n-7)/16:0 and 18:1(n-9)/18:0 were used as surrogates for delta 9-desaturase activity; the ratios of 18:3(n-6)/18:2(n-6) and 20:4(n-6)/20:3(n-6) were used as surrogates for delta 6- and delta 5-desaturase activity, respectively; and the ratios of 18:1(n-7)/16:1(n-7), 18:0/16:0, and 20:2(n-6)/18:2(n-6) were used as surrogates for elongase activities .
Fatty acid data were only available for 61 subjects; total lipid class data for 60 subjects; and plasma glucose, HOMA-IR and waist circumference at 12 wk were only available for 59 subjects and reported herein. ANCOVA and regressions used to determine relationships between any independent variable and change in anthropometrics were adjusted for site, age, sex, energy, protein (g/d) intakes and physical activity at 12 wk and baseline HOMA-IR.