Cluster for Public Health Nutrition, Boden Institute of Obesity, Nutrition, Exercise and Eating Disorders, The University of Sydney, Sydney, New South Wales, Australia
Associate Professor TP Gill, Cluster for Public Health Nutrition, Boden Institute of Obesity, Nutrition, Exercise and Eating Disorders, Level 2, K25 Medical Foundation Building, The University of Sydney, Sydney, NSW 2006, Australia. E-mail: firstname.lastname@example.org
Associate Professor TP Gill, Cluster for Public Health Nutrition, Boden Institute of Obesity, Nutrition, Exercise and Eating Disorders, Level 2, K25 Medical Foundation Building, The University of Sydney, Sydney, NSW 2006, Australia. E-mail: email@example.com
A comprehensive literature search was undertaken to examine the relationship between dairy consumption and overweight/obesity in prospective cohort studies. A literature search from 1980 through to April 2010 was conducted. Nineteen cohort studies met all the inclusion criteria and were included in the systematic review. Of the 19 cohort studies, 10 were among children and adolescents (aged 2 to 14 years, n = 53 to 12 829, follow-up 8 months to 10 years) and nine among adults (aged 18 to 75 years, n = 248 to 42 696, follow-up 2 years to 12 years). A range of dairy food exposure measures were used. Eight studies (three out of 10 studies involving children and five out of nine studies involving adults) showed a protective association against increasing weight gain (measured in various ways); one reported a significant protective association only among men who were initially overweight; seven reported no effect; one reported an increased risk (among children), and two reported both a decreased and increased risk, depending on the dairy food type. The evidence from prospective cohort studies for a protective effect of dairy consumption on risk of overweight and obesity is suggestive but not consistent, making firm conclusions difficult.
The prevalence of obesity among children, adolescents and adults in developed countries has increased at an alarming rate with the associated complications placing a substantial burden on healthcare costs (1). Obesity, however, is a complex issue (2) and thus a wide range of strategies are required to tackle the problem. The potential for particular aspects of the diet to be involved in preventing obesity has been considered.
The common perception that the consumption of dairy foods, particularly the non-reduced fat options, leads to excessive weight gain has led to a number of recent studies exploring the association between the intake of dairy foods and weight status. However, much confusion remains about this relationship. Some of the studies have suggested that dairy products may help in the prevention of weight gain and promote weight loss (3–10); however, most of these studies, particularly interventional studies, are of small scale (7–10). A recent review of larger-scale (n > 50) randomized trials of the effect of dairy products and/or calcium supplementation on weight status did not support this hypothesis (11). A possible reason for apparent inconsistencies in findings could be that dairy supplements only impact on weight loss in overweight or obese subjects whose baseline dairy intake is low (equivalent calcium intake <600 mg d−1) or that dairy supplements do not displace other foods from the original diets of the participants, but add extra energy (8). Other studies have looked at the cross-sectional or longitudinal association between habitual dairy consumption and the risk of developing overweight/obesity and found that there are conflicting results (12).
The aim of this systematic review was to investigate the longitudinal association, if any, between dairy consumption and overweight/obesity, using data from prospective cohort studies.
A literature search using MEDLINE, EMBASE, CINAHL, Cochrane Library, Science Direct, ISI Web of Science, EBM Review, Google Scholar and Conference Proceedings databases, from 1980 through April 2010 was conducted with the Medical Subject Headings (MeSH) overweight, obesity, dairy products, milk, cheese, yoghurt and other relevant terms (see Appendix S1 for complete search strategy). The search was restricted to human studies with no restrictions on gender, age or ethnicity. Only articles published in English were included and a manual search of references cited by the identified studies was also undertaken.
To determine the eligibility of the identified studies, three investigators (AMR, JCYL and VMF) independently screened the abstracts of the 1200 identified studies, and the full text of the article was reviewed when the abstract did not provide enough information. Only prospective cohort studies that included dairy products as the exposure variable and change in body weight or body mass index (BMI) or % body fat or skin-fold thickness as the primary outcome variable were included. Data involving fortified dairy products of any type, i.e. those that have a required dietary factor added that dairy food does not normally contain, were excluded for a number of reasons. Specifically excluding fortified dairy allows the effect seen to be attributable to dairy in its natural form, not from the fortificants. Also, consumers of fortified dairy products are likely to be more health-conscious thereby biasing the results. However, studies which involve dairy products that were universally fortified were not excluded.
The flow of study analysis is shown in Fig. 1. A total of 19 prospective cohort studies were included in this systematic review.
All data in the 19 studies were independently extracted by four investigators (VMF, AMR, DJH and TPG), using a custom-built database. Any discrepancies were discussed and resolved. The following study characteristics were recorded: surname of the first author, year of publication, source of publication, country of origin, language, funding source, study outcome, definition of metabolic syndrome, mean age and range, sex, ethnicity, education levels, mean BMI of group, response rate, dietary assessment tool and number of items, mode of administration of dietary assessment tool, validation method and result, number of dietary intake assessments, duration of follow-up, method of segregation according to dairy intake (tertiles, quartiles, quintiles, deciles and other), adjustments for potential confounders in basic, intermediate and final models, stratification used, and energy and macronutrient intakes.
Of the 19 prospective cohort studies examined in this systematic review, 10 studies were conducted among children and adolescents and nine were conducted among adults. Eight studies showed a protective effect of dairy consumption on weight status; one study reported a significant protective effect only among men who were initially overweight; seven studies reported no association; one study reported an increased risk of overweight and obesity among children, and two studies reported differing associations depending on the dairy food type.
Children and adolescents
Table 1 shows the characteristics of the 10 studies (4,13–21) examining the association between dairy consumption and obesity in children and adolescents. The number of participants in each study ranged from 53 to 16 679, and their age ranged from 2 to 14 years. Dairy consumption was assessed by food frequency questionnaires (FFQs) (4,16,17,21), 24-h recall (20) or food records (13–15,18,19). All FFQs had been validated against dietary calcium intake (4,21) or milk intake (16,17) obtained from food records, with the correlation coefficient ranging from 0.20 to 0.68. The unit of measure of dairy consumption was inconsistent among the studies, with some (16,19) reporting weight/volume of dairy consumed while others (4,13–15,17,18,20,21) reported servings of dairy per day, with varying definitions of serving size used. For example, Moore et al. (15) reported one cup of milk as a serving of dairy, while Johnson et al. (14) reported a serving of milk as 180 g. The type of dairy food assessed varied between studies, with six studies (4,14,16,18–20) using total milk and three studies (13,15,17) using total dairy, while one study (21) used both. Most studies (4,16–19,21) reported BMI, BMI z-score or change in BMI (per year) as the outcome measure, while others (13–15,17) reported fat mass/% body fat and/or sum of skin-folds as the outcome. The duration of follow-up ranged from 8 months to 10 years, with only two included studies (16,21) having a follow-up period of less than 2 years. The adjustment for potential confounders also varied greatly between studies, with only four studies (15,16,18,21) adjusting for total energy intake (see Table 1).
Table 1. Characteristics and outcome measures of studies examining the association between dairy consumption and obesity in children and adolescents
Full model adjustments
BMI, body mass index; FFQ, food frequency questionnaire; OB, obese; OW, overweight.
n = 2371 Aged 9–10 years 100% girls Duration of follow-up: 10 years Dairy type: total milk intake Dietary assessment: 3-day food records
Adjusted for: total kJ intake ethnicity site visit other beverages
A weak protective association was found between milk intake and change in BMI. An increase of 100 g of dairy intake per day was associated with a 0.002 (SE = 0.006) point decrease in BMI, although not statistically significant.
n = 12 829 Aged 9–14 years 94.7% Caucasian 43.3% boys Duration of follow-up: 4 years Dairy type: milk only Categorized into full fat, 2% fat, 1% fat and skim Total milk intake used for continuous analysis Dietary assessment: semi-quantitative FFQ FFQ validity: good
Stratified by gender, as well as whether milk type has changed during BMI change Adjusted for: physical activity ethnicity height growth menstrual history Tanner stage prior BMI z-score inactivity
ΔBMI per year
High milk consumption was positively associated with weight gain, which was speculated to be due to the extra energy from dairy foods. Compared to participants who consumed ≤0.5 servings of milk per day, boys and girls who consumed >3 servings of milk per day has a 0.081 (SE = 0.048) and 0.093 (SE = 0.034) higher BMI points than their counterparts respective (both P < 0.05).
n = 53 Mean age: 2 years 54.7% boys Caucasian US children Duration of follow-up: 3 years and 10 months Dairy type: total dairy (given as calcium equivalents of 240-mL milk) Dietary assessment: 3-day food records
Adjusted for: gender BMI fat intake other dietary factors
Δbody fat (as gram or %)
A protective association was found. Each serving of dairy products consumed corresponds to a 907.06 g (SE = 284.60) decrease in body fat (P = 0.003).
n = 178 Aged 8–12 years 100% girls 74% Caucasian, from Massachusetts Duration of follow-up: 6 years Dairy type: total dairy (skim/low-fat milk, whole milk, cream, sherbet or ice milk, ice cream, ice cream sundaes, milkshakes, yogurt, cottage or ricotta cheese, cream cheese and other cheeses) Dietary assessment: semi-quantitative FFQ FFQ validity: validated in adults only
Adjusted for: physical activity age other dietary factors parental overweight
Δ% body fat ΔBMI z-score
No association between dairy consumption on BMI z-score and % body fat was found.
n = 362 at 5 years; 461 at 7 years Mean age: 5.16 years Recruited from Avon, UK Duration of follow-up: 2–4 years Dairy type: milk only Dietary assessment: 3-day food records
Adjusted for: education level fat intake fibre intake gender baseline characteristics other dietary factors height at 9 years baseline BMI misreporting TV watching
Δfat mass (in kg)
A protective association was found. Each serving of milk at 5 years and 7 years of age was associated with a −0.51 (95% CI: −0.86 to −0.16; P < 0.01) and −0.35 (95% CI: −0.51 to −0.14; P < 0.01) change in body fat mass at 9 years respectively.
n = 92 Aged 3–6 years From Framingham, USA ∼60% boys Duration of follow-up: 8 years Dairy type: total dairy (milk, yoghurt and cheese) Dietary assessment: 3-day food records
Adjusted for: physical activity education level total kJ intake fat intake age baseline characteristics baseline anthropometry
Σfour skin-folds at age 10–13 years BMI at age 10–13 years
A protective association was found. Higher consumption of dairy products at 3–6 years old was associated with a lower BMI at 10–13 years (mean ± SE; T1 vs. T3: 21.1 ± 0.6 vs. 19.3 ± 0.6; Ptrend = 0.046). It is also associated with a lower sum of four skin-fold measures (mean ± SE; T1 vs. T3: 82.4 ± 5.8 vs. 57.2 ± 5.8; P = 0.005).
n = 852 Age: 2 years Duration of follow-up: 1 year Dairy type: total dairy, milk only, full-fat milk, reduced-fat milk Dietary assessment: semi-quantitative FFQ FFQ validity: good
Adjusted for: baseline BMI z-score age sex race/ethnicity maternal BMI maternal education paternal BMI total kJ intake serves of non-dairy beverages television viewing
BMI z-score BMI z-score ≥ 85th percentile
Total dairy or total milk intake at age 2 years was not associated with BMI z-score or incident overweight at age 3 years. Intake of milk at age 2 years, whether full- or reduced-fat, was not associated with risk of incident overweight at age 3 years. Higher intake of whole milk at age 2, but not reduced-fat milk, was associated with a slightly lower BMI z-score (0.09 unit per daily serving at age 3 years). However, when restricted to children with a normal BMI at age 2 years, the association was null.
Six of the 10 studies in children and adolescents (16–21) reported no significant association, while three studies (13–15) reported a protective association between dairy consumption and overweight/obesity. However, one study (4) reported an increased risk, which was possibly due to excess energy intake from milk, as the risk was associated with high milk intake (≥3 servings per day). In the study by Moore et al. (15), a higher dairy consumption at 3–6 years of age was associated with a lower BMI (mean ± SE; T1: 21.1 ± 0.6 kg m−2 vs. T3: 19.3 ± 0.6 kg m−2; Ptrend = 0.046) and sum of four skin-folds (mean ± SE; T1: 82.4 ± 5.8 mm vs. T3: 57.2 ± 5.8 mm; Ptrend = 0.005) at 10–13 years of age. Carruth and Skinner (13) reported the regression coefficient between daily servings of milk consumption at age 5 years and change in fat mass (in kg) at age 9 years as −0.51 (P < 0.01), while Johnson et al. (14) also reported a significant decrease in % body fat (β ± SE = −3.54 ± 1.04; P = 0.001) and absolute amount of body fat in grams (β ± SE = −907.06 ± 284.60; P = 0.003) for each number of milk servings for 3 days.
Table 2 summarizes the characteristics of the nine studies conducted among adults (22–30). The number of participants ranged from 248 to 42 696, and their age ranged from 18 to 75 years. Most studies used FFQs (23–27,30) to collect dairy intake data, while two studies used food records (22,29) and one study (28) used multiple 24-h recalls. Most FFQs used have been validated against dietary calcium intake (23,24,30) or milk intake (25) obtained from food records, while one FFQ (27) was not validated and another (26) was validated only against nutrients not specific to dairy intake. Dairy consumption was reported mainly as ‘amount consumed (or servings) per day’ (or per week) or ‘change in consumption’. As in the studies among children and adolescents, the unit of measure of dairy consumption was inconsistent among the studies. For example, Snijder et al. (27) defined a serving of milk as 150 g, while Vergnaud et al. (28) defined that as 225 g. The types of dairy foods assessed differed between studies – some (23,24,27,28) assessed total dairy consumption, while others included either total milk consumption alone (25), full-cream dairy consumption only (26), or low-fat/skim milk and yoghurt consumption only (22). Different outcome measures have been reported, including weight change (in kg) (22,24,26–29), odds of mean weight gain ≥1 kg per year (25), change in waist circumference/waist-to-hip ratio (22,27–30), sum of skin-folds (22), BMI (27) and risk of obesity (23). The duration of follow-up ranged from 7 months to 10 years, although the majority of included studies (22–25,27,28,30) had a follow-up period of at least 5 years. Most studies (23–30) had adjusted for total energy intake.
Table 2. Characteristics and outcome measures of studies examining association between dairy consumption and obesity in adults
Full model adjustments
BMI, body mass index; FFQ, food frequency questionnaire; MAR, mean adequacy ratio; N/A, not applicable; PAL, physical activity level; WHR, waist-to-hip ratio.
n = 19 615 Mean age: 51.28 years 100% men Baseline BMI: 25.2 kg m−2 Duration of follow-up: 12 years Dairy type: total dairy (whole milk, skim milk, low-fat milk, ice cream, cheese, yogurt and cream) Categorized into low-fat and high-fat dairy Dietary assessment: semi-quantitative FFQ FFQ validity: good
Adjusted for: smoking status alcohol intake physical activity total kJ intake fat intake fibre intake age baseline characteristics other dietary factors
Δweight (in kg)
High high-fat dairy intake at baseline was protective against weight gain in 12 years (mean ± SE; Q1 vs. Q5: 3.24 ± 0.11 vs. 2.86 ± 0.11; Ptrend = 0.03). However, change in total dairy (mean ± SE; Q1 vs. Q5: 2.57 ± 0.13 vs. 3.14 ± 0.11; Ptrend = 0.001) and/or high-fat dairy (mean ± SE; Q1 vs. Q5: 2.70 ± 0.14 vs. 3.27 ± 0.11; Ptrend < 0.001) intake (i.e. increased intake) was associated with a significantly higher weight gain in 12 years.
n = 6319 Mean age: 37.03 years 42.5% men Baseline BMI: 23.4 kg m−2 Duration of follow-up: 2 years Dairy type: full-cream dairy only Dietary assessment: semi-quantitative FFQ FFQ validity: not validated for calcium or dairy
Adjusted for: smoking status alcohol intake physical activity total kJ intake age gender BMI other dietary factors
Δweight (in kg)
A protective association was found. Higher whole-fat dairy intake resulted in a lower weight gain in a 2-year period (mean ± SE; T1 vs. T3: +0.64 kg vs. +0.26 kg; Ptrend < 0.001).
n = 19 352 Aged 40–55 years 100% women Baseline BMI: 23.7 kg m−2 Duration of follow-up: 10 years Dairy type: milk only, low-fat vs. full-cream Dietary assessment: FFQ FFQ validity: fair for regular fat milk; good for low-fat milk and total dairy
Stratified by baseline dairy intake (<1 serving vs. ≥1 serving) Adjusted for: alcohol intake education level total kJ intake fat intake fibre intake age baseline characteristics other dietary factors parity
Odds of mean weight gain (kg) ≥1 kg per year
A weakly protective effect was found for cheese (OR = 0.85; 95% CI: 0.73–0.99) as well as whole milk and sour milk (OR = 0.70; 95% CI: 0.59–0.84). The protective effect of cheese remained when stratified by BMI status, but that for whole milk, and sour milk only holds true for normal-weight women. The combination of constant high intake (≥1 serving per day) of these foods corresponds to a −0.99-kg difference compared with a low intake (<1 serving per day).
n = 248 Aged 18–65 years (mean age: 39.6 gyears) 45.2% men Baseline BMI: 25.3 kg m−2 Duration of follow-up: 6 years Dairy type: low-fat/skim milk and gyoghurt Dietary assessment: 3-day food grecords
Adjusted for: physical activity age baseline characteristics adiposity indicator
Δwaist circumference ΔΣsix skin-folds Δ% body fat Δbody weight
Consumption of skimmed and partly skimmed milk was associated with a weakly protective association on body weight (β ± SE = −0.20 ± 0.09, P = 0.06) and significantly negative effects on change in waist circumference (β ± SE = −0.23 ± 0.09, P = 0.02) as well as sum of 6 skin-fold thickness (β ± SE = −1.01 ± 0.40, P = 0.01). Yoghurts with <2% fat, on the other hand, was associated with a significantly positive effect of change in waist circumference (β ± SE = 0.42 ± 0.19, P = 0.02).
n = 1124 Aged 50–75 years (mean age: 60.0 years) 46.1% men Baseline BMI: 26.6 kg m−2 Duration of follow-up: 6 years and 5 months Dairy type: total dairy (milk, cheese, yogurt, curds and custard) Dietary assessment: semi-quantitative FFQ FFQ validity: good for original 75-item FFQ (38); unknown for the 92-item FFQ used in this study
Adjusted for: smoking status alcohol intake physical activity total kJ intake age gender baseline characteristics
ΔWHR Δwaist circumference Δweight ΔBMI
No significant association was found between total dairy consumption and change in body composition (BMI, weight, waist and waist-to-hip ratio).
n = 1909 Mean age: 25.07 years 45.3% men 44.05% African–American Baseline BMI: N/A Duration of follow-up: 10 years Dairy type: total dairy (milk, milk drinks, butter, cream, cheeses, yogurts, dips, ice cream, puddings and other dairy-based desserts) Dietary assessment: semi-quantitative FFQ FFQ validity: good
Adjusted for: ethnicity total kJ intake age gender baseline characteristics study centre
Obesity (BMI > 30 or WHR > 0.85 [f]/0.90[m])
A protective association was found between obesity status and total dairy (OR per 1 daily eating occasion: 0.82; 95% CI: 0.72–0.93), high-fat dairy (OR per 1 daily eating occasion: 0.84; 95% CI: 0.73–0.97) and milk and milk drinks (OR per 1 daily eating occasion: 0.83; 95% CI: 0.68–1.00) but not other types of dairy foods.
n = 76 nutrition students Mean age: 19.2 years 14.5% male Baseline BMI: 22.9 kg m−2 Duration of follow-up: 7 months Dairy type: total dairy and low-fat (≤1% milk fat) dairy Dietary assessment: 7-day food records
Total dairy adjusted for: % energy intake/energy expenditure race sex Low-fat dairy also adjusted for BMI % body fat total dairy intake
Δweight (in kg) Δwaist circumference Δtruncal fat Δ% body fat
No significant association was seen between total dairy intake and change in body weight, waist circumference, % total body fat or % truncal fat. Low-fat dairy intake had a protective effect with higher intake of low-fat dairy products being associated with lower gains in body weight and reductions in waist circumference, % truncal fat and % total body fat (note that subjects who consumed higher intakes of low-fat dairy products also consumed more healthy diets in general).
n = 2267 Age > 45 years Mean age: 51 years 54.9% male Baseline BMI: 24.4 kg m−2 Duration of follow-up: 6 years Dairy type: total dairy (milk, cheese, yoghurt); also analysed milk, cheese and yoghurt separately Dietary assessment: 6 × 24 h recalls
Stratified by: baseline weight status sex Adjusted for: age intervention group (study part of a supplementation trial) baseline weight/WC education level smoking status PAL total kJ intake MAR (index of diet quality) menopausal status (for women) alcohol
Δweight (in kg) Δwaist circumference
Milk and yoghurt consumption was protective against 6-year changes in weight (Ptrend = 0.02 for milk and 0.01 for yoghurt) and waist circumference (Ptrend = 0.02 for milk and 0.01 for yoghurt) only in men who were initially overweight. No associations were found between total dairy intake and weight/waist circumference among men, and in women significantly positive trends with yoghurt and waist circumference in normal-weight women were found.
n = 42 696 Age: 50–64 years 47.2% male Baseline BMI: 25.4 kg m−2 Duration of follow-up: 5 years Dairy type: low-fat and high-fat dairy products Dietary assessment: semi-quantitative FFQ FFQ validity: good
Stratified by sex Adjusted for: baseline waist circumference BMI age smoking status sport activities kJ from wine, beer and spirits baseline kJ
A protective association with 5-year difference in waist circumference in women was seen for high-fat dairy only (β = −0.09 cm waist circumference for each 60 kcal of dairy). Among men none of the dairy products showed significant associations with 5-year difference in waist circumference.
One study (27) showed no protective association between dairy consumption and overweight/obesity, while five studies (23,25,26,29,30) showed a protective association. One of these studies (29) found a protective effect of low-fat dairy but not total dairy. Another study (28) reported a significant protective effect only among men who were initially overweight. Two studies (22,24) reported both a decreased and increased risk – Rajpathak et al. (24) reported that high high-fat dairy intake at baseline was protective, while change in total dairy was associated with higher weight gain in 12 years; Drapeau et al. (22) found that for waist circumference, skimmed and partly skimmed milk was associated with a decreased risk, while low-fat yoghurt was associated with an increased risk.
Among the five studies which reported a protective association, the strength of the association varied. Pereira et al. (23) reported an 18% reduction in odds of being overweight/obese per daily eating occasion of total dairy products, while Rosell et al. (25) reported that baseline cheese consumption ≥1 servings per day was associated with a 30% reduction in odds of having mean weight gain ≥1 kg per year. Sánchez-Villegas et al. (26) reported similar findings in a 2-year period. Poddar et al. (29) showed that subjects who had a higher intake of low-fat dairy products had a lower increase in weight (1.41 ± 0.42 kg vs. 0.09 ± 0.41; P = 0.034), and greater reductions in waist circumference (1.10 ± 0.81 cm vs. −1.62 ± 0.78; P = 0.023), % truncal fat (0.80 ± 0.39 vs. −0.44 ± 0.38; P = 0.033) and % body fat (0.64 ± 0.35 vs. −0.44 ± 0.34; P = 0.038). Halkjær et al. (30) reported a 0.09-cm (95% CI = 0.03–0.15) decrease in waist circumference per 60 kcal increase of high-fat dairy among women in a 5-year period.
It was not possible to conduct a meta-analysis on the studies either in children or in adults, because of the high heterogeneity of the studies as well as inconsistent exposure and outcome measures, as described above.
To our knowledge, this is the first systematic review of prospective cohort studies to assess the longitudinal relationship between habitual dairy consumption and the risk of overweight/obesity. Among the 19 cohort studies investigated in this systematic review, eight studies, three out of 10 studies involving children and five out of nine studies involving adults, showed a protective association of dairy consumption against inappropriate weight gain (measured in various ways), and seven studies reported no effect. One study reported a significant protective association only among men who were initially overweight; one study reported an increased risk only among children with very high milk intake, and two studies in adults reported both a decreased and increased risk, depending on the dairy food type. Only one study, among adults, showed a detrimental effect of dairy consumption on weight status.
Even though there was a much higher proportion of studies among adults which showed a protective effect, the association between dairy consumption and weight status does not seem to be consistent in either children/adolescents and adults. Coupled with the differences in growth/metabolism between the two groups, it is difficult to conclude whether the relationship varies with life stage. However, we can at the very least conclude that dairy products showed no harmful effect on weight status, in both children and adults.
Contrary to popular belief, in this review we did not find low-fat dairy products to be more beneficial to weight status than regular-fat dairy products. In fact, the data suggest the reverse may be true. The majority of the included studies examined milk intake, while only a few reported using other dairy foods as the exposure variable. The weight of evidence from this review suggested milk intake was more likely to be associated with beneficial weight outcomes, while a very small number of studies showed beneficial effects of other dairy products such as yoghurt/cheese among subjects with specific baseline characteristics.
Different dietary assessment methods used in the included studies may partly explain the inconsistent findings. While a semi-quantitative FFQ with good validity could reliably rank the participants according to their usual intake of dairy foods, a non-validated FFQ (or one with poor validity) could result in gross misclassification of the subjects' intake, hence reducing the reliability of the findings. This may have occurred in some of the included studies (17,26,27). In addition, all three studies among children and adolescents that showed a protective effect of dairy on overweight/obesity collected dairy consumption data using food records rather than FFQs. Food records have been suggested to provide more accurate food consumption data than FFQs (31), thereby lending strength to the findings from these three studies that there is a protective effect of dairy consumption on weight gain.
Meta-analyses of data from studies conducted among children and/or among adults were not possible due to the variability between studies, particularly in the outcome and exposure measures. Other influences on variability included sample size, age of children, adjustment models and duration of follow-up. Although age of the children studied varied greatly (from 2 to 14 years), there was no consistent association between age and the direction of the outcomes. Sample size of the included studies varied from 53 in the study by Carruth and Skinner (13) to 42 696 in the study by Halkjær et al. (30). Overall, in studies among children and adolescents, those which found no association or a protective effect of dairy against obesity were smaller in sample size (n from 53 to 2371) compared to the study which reported a detrimental association between high milk consumption and obesity (n = 12 829) (4). This is in contrast to the studies among adults, where the three largest studies (24,25,30) showed a protective effect of dairy on weight status and weight gain. The contrasting findings depending on sample size could be accounted for by the varying adjustment models used in the analyses. Adjustment for total energy intake is important as the energy from dairy food, if consumed as an excess of the total daily energy, could have masked the true effect (if any) of dairy products on obesity risk. Berkey et al. (4) who reported a positive association between high milk consumption and change in BMI in children acknowledged this as a possible reason for their findings. Ideally, models should be computed with and without energy adjustment to determine the effect of energy intake on the outcome. Adjustment for other factors varied from study to study and included variables such as age, gender, baseline BMI or other anthropometric characteristics, ethnicity, education level, poverty level, smoking, activity and inactivity levels, total energy intake, fat intake, fibre intake, alcohol intake, diet quality, misreporting, intake of other beverages, other dietary factors, and for children height growth, menstrual history, Tanner stage, birthweight and parental obesity. Because of this large degree of heterogeneity in the adjustment variables applied to the individual studies, in particular nutrient and dietary factors, the outcomes of the studies are difficult to compare.
Several studies had also explored the longitudinal relationship between calcium and weight status. Gonzalez et al. (32) reported an inverse relationship between dose of calcium supplementation and 10-year weight change in 10 591 middle-aged (53–57 years) men and women. Rajpathak et al. (24), on the other hand, did not find any association between calcium (both baseline and change in intake) and 12-year weight change. In the same study, however, the authors reported that increasing dairy (and hence calcium) consumption was associated with increasing weight gain. A possible explanation for the contrasting finding was suggested by Major et al. (33) that the impact of increased dairy intake on weight status is only apparent in those individuals who had a low intake of calcium at baseline. Compared to subjects with higher calcium intake, those who had very low baseline levels of calcium intake (<600 mg) had significantly improved weight reduction when calcium supplements were given. There is, however, insufficient data from Rajpathak et al. (24) to allow for such a comparison.
The weight status of the sample population at baseline may be another factor that influences the relationship between dairy food consumption and weight change over the period of follow-up. One study found that dairy consumption was inversely associated with the development of obesity in those individuals who were overweight at baseline (BMI ≥ 25 kg m−2) but not in those whose weight was below this level at baseline (28). The higher protection offered by adequate dairy intake among this group may be a result of the potentially higher susceptibility to weight gain among overweight/obese individuals.
Duration of follow-up period has also been suggested as an important factor to consider when interpreting the results from these studies, as weight gain is time-dependent, hence sufficient time should be allowed for an effect to be seen (34). This may be responsible for the lack of findings in the study by Newby et al. (16), which only followed the participants for 8 months. However, most studies included in this systematic review had a longer follow-up period of more than 2 years, which is more suited to identifying weight changes.
While we have excluded studies involving fortified dairy products from the review, it is possible that dairy products in some included studies were universally fortified (35). Among many of the possible nutrients that could be fortified to milk, only calcium (36) and protein (37) have been associated with weight status. They have been adequately adjusted for in the included studies. Therefore, we believe this alone was insufficient to explain the inconsistencies seen in the studies.
This investigation highlights the enormous difficulty in conducting meta-analyses of food groups in relation to an outcome which is reported in varying ways. Despite BMI now being accepted as the standard measure of weight status with specific criteria for overweight and obesity, considerable variation was found in the way changes in weight status are reported within studies. An outcome which requires adherence to specific diagnostic criteria, e.g. diabetes, cancer, myocardial infarct, is likely less difficult to assess. In addition, investigation of a food group as opposed to a nutrient is also difficult as it can be defined and described differently, making it difficult to compare research findings. It would be valuable if future research in this area includes a set of defined exposure and outcome variables, measured using the same methods and reported in a consistent manner. For example, weight change in children should be reported in a consistent format such as BMI change per year or change in BMI z-score, while for adults the weight change in kg or the BMI change in kg m−2 should be reported. A standard base adjustment model would also allow for direct comparison between studies. In addition, standardized adiposity measures such as % body fat or waist circumference should be considered for use as outcome measures in addition to BMI measures. Similar consideration needs to be applied to reporting of intake of dairy foods and composition of the dairy foods group.
There is some indication of a protective effect of dairy food consumption on weight status in adults, although the currently available evidence from prospective cohort studies is inconclusive. The small number of studies and the inconsistencies in the way the dairy consumption, as well as weight outcomes, have been defined preclude meta-analysis. The small magnitude of the impact on weight status reported in those studies which demonstrated a protective effect suggests that if such an effect does exist, the magnitude is likely to be small, although these differences over many years can be clinically meaningful and of public health significance. In summary, there is currently insufficient evidence to conclude that increased dairy consumption, particularly of regular-fat varieties, is associated with weight status.
Conflict of Interest Statement
This systematic review was funded by Dairy Australia. The authors declare that Dairy Australia had no influence on the review process or the conclusion drawn.
Associate Professor Victoria Flood and Dr Debra Hector were affiliated with the Cluster for Public Health Nutrition, Boden Institute of Obesity, Nutrition, Exercise and Eating Disorders, The University of Sydney, at the time the review was conducted.
All authors were involved in the development of the search strategy. JCYL designed the purpose-built database for data extraction. VMF, DJH and AMR screened the abstracts and extracted the data from the original literature search. TPG and AMR extracted the data from the additional literature search. JCYL drafted the manuscript. All authors were involved in the subsequent edits of the manuscript, and read and approved the final manuscript.