Study selection

We included RCTs with either a parallel or a crossover design that were conducted in adults aged ≥18 y. Dairy products should have been used as the main intervention but not as part of a multicomponent dietary supplement in either experimental or control groups, the intervention in control groups was not soy milk, and participants had to consume dairy products for ≥4 wk. Furthermore, data on changes of body weight or fat mass in both experimental and control groups were extracted from the report or obtained from the authors. We included English-language articles only.

Data extraction and quality assessment

We extracted the following information from each study: authors, publication year, geographic location, funding sources, sample size and attrition, intervention and control regimens, study duration, study design (crossover or parallel), analysis strategy (intent-to-treat or per-protocol analysis), and participant information (age, sex, and baseline intake). We used the Jadad score to assess study quality (12). Trials scored one point for each item addressed in the study design, including random assignment, blinding, description of withdrawals and dropouts, methods of random assignment, and double-blinding status, which generated a scale from 0 to 5. Higher numbers represented the better quality of a given study. Because double blinding was almost not possible in this type of trial, studies with a Jadad score ≥3 were defined as high quality, and the rest of the studies were classified as low quality.

Most of the RCTs showed the intervention dose in servings per day. In 6 studies (13–18) in which dairy interventions were not shown in servings per day, we standardized units to servings per day, which was equivalent to 240 mL or 8 oz in volume or 452 mg Ca or 8.4 g protein in nutrient content per day according to the USDA national nutrient database for standard reference (19). In 2 studies (17, 20), participants were randomly assigned to a control group or medium- or high-dairy groups; thus, we combined medium- and high-dairy groups as the intervention arm with relevant data provided by the authors.

We extracted the means and SDs of changes from baseline to endpoint (both intervention and control arms) from all studies. SDs were calculated from SEs or CIs when necessary; for articles with missing SDs for measurements of change (4, 13, 16), change-from-baseline SDs were imputed by using the correlation coefficient method referenced in the Cochrane Handbook for Systematic Reviews of Interventions (21). We used the average correlation coefficient of 0.96 between baseline and endpoint measurements, which was estimated from all other included studies with available data. A sensitivity analysis indicated that results were not substantially altered by removing the 3 studies with imputed values.

Characteristics of studies

Primary characteristics of the 29 included trials are shown in Table 1. A total of 2441 participants were randomly assigned to intervention or control regimens in these trials with a completion rate of 84.4% (n = 2060). The sample size varied from 20 to 265 participants with a mean age range from 20.0 to 62.0 y (median: 41.4 y). The mean baseline BMI ranged from 20.2 to 43.0, and 24 trials enrolled overweight or obese participants. Of the 29 trials, 14 trials were conducted exclusively in women (5 trials conducted in postmenopausal women, 8 trials conducted in premenopausal women, and 1 trial that did not mention menopausal status), 1 trial was conducted in men, and the other 14 trials were conducted in both sexes, with 2 trials that reported results separately by sex (4, 14).

Intervention regimens were diverse. Low-fat fluid milk was evaluated in 7 trials (4, 17, 25–29); skimmed-milk powder was used in 2 trials (30, 31); yogurt was used in 2 trials (7, 32); yogurt, cheese, and milk were freely chosen as the intervention in 11 trials (6, 8, 9, 14, 17, 33–38); a yogurt and milk combination was used in 2 trials (5, 24, 39); and the remaining 5 studies (13, 15, 16, 20, 40) did not specify the intervention regimen. The intervention dose (total dairy intake) varied from 1 to 6.5 servings/d. A habitual diet or a calorie-restricted diet with lower dairy consumption was chosen as the control regimen in most studies. An isocaloric sucrose beverage or fruit juice was provided to participants in control groups in 2 studies (32, 39).

The trials lasted from 1 to 24 mo, with a median duration of 4 mo. Most of the trials used parallel study designs, and 2 trials (15, 39) used crossover designs. The quality of selected trials was diverse, with 19 studies classified as high quality (Jadad score ≥3) and 10 studies classified as low quality.

In 17 weight-loss trials, participants were instructed to reduce their daily energy intakes to a certain amount below their energy requirements. Caloric requirements were estimated individually on the basis of resting metabolic rates and physical activity levels, and energy deficits from 100 to 600 kcal/d were subtracted from estimated caloric requirements in 16 of 17 studies. The remaining one study (14) administrated isocaloric low-energy diet plans for all participants regardless of their individual energy requirements. Participants were instructed to keep physical activity levels constant in the majority of trials except in 2 studies, one of which (27) had a 2-by-2 factorial design that included an exercise component, and in the other study (28), milk was given in addition to exercise, whereas participants in the control arm received only an exercise intervention. Most studies were conducted in the United States (n = 17), and other studies were conducted in Europe (n = 2), Canada (n = 3), Australia (n = 2), Mexico (n = 1), Asia (n = 3), and Puerto Rico (n = 1).

Changes in weight and body fat

In 5 studies (26, 28, 29, 33, 34), data were analyzed by using the intention-to-treat principle, whereas the other studies analyzed data in completers only. For this reason, a total of 2101 participants in 29 studies were included in the analysis of weight change, and 1536 participants in 22 studies were included in the analysis of changes in body fat.

There was no significant difference in body weight changes between the dairy intervention and control groups (−0.14 kg; 95% CI: −0.66, 0.38 kg; Figure 2A); however, a significant reduction that favored dairy products in body fat was shown (−0.45 kg; 95% CI: −0.79, −0.11 kg; Figure 2B). A significant heterogeneity was observed for both body weight (I² = 86.3%) and body fat (I² = 70.9%).

Open in a separate window

Open in a separate window

FIGURE 2.

Net change (95% CI) in body weight (A) and fat (B) associated with dairy interventions expressed as the change (kg) during the intervention with dairy products minus the change during control regimen. Horizontal lines denote the 95% CIs; solid diamonds represent the point estimate of each study. The open diamond represents the pooled estimate of the intervention effect. The dashed line denotes the point estimate of the pooled result. ID, identification; WMD, weighted mean difference.

Subgroup analysis

Meta-regression analysis showed that energy restriction (P = 0.008) and study duration (P = 0.008) significantly influenced pooled estimates. Therefore, we conducted subgroup analyses according to these variables.

A significant reduction in body weight in the dairy group was shown in studies that imposed energy restriction (−0.79 kg; 95% CI: −1.35, −0.23 kg; I² = 38.5%; Figure 3A). On the contrary, in studies without energy restriction, no significant effect of dairy intervention was observed (0.39 kg; 95% CI: −0.36, 1.13 kg; I² = 89.7%). Similarly, body fat declined significantly in energy-restricted trials (−0.94 kg; 95% CI: −1.53, −0.34 kg; I² = 59.2%; Figure 3B) but not in studies without energy restriction (−0.12 kg; 95% CI: −0.71, 0.46 kg; I² = 81.9%).

Open in a separate window

Open in a separate window

FIGURE 3.

Net change (95% CI) by energy restriction in body weight (A) and fat (B) associated with dairy interventions expressed as the change (kg) during the intervention with dairy products minus the change during the control regimen. Horizontal lines denote the 95% CIs; solid diamonds represent the point estimate of each study. The open diamond represents the pooled estimate of the intervention effect. The dashed line denotes the point estimate of the pooled result. ID, identification; WMD, weighted mean difference.

We also classified studies according to duration as short term (<1 y duration) or long term (≥1 y duration; Figure 4). Dairy products significantly reduced body weight in the short-term interventions (−0.47 kg; 95% CI: −0.90, −0.03 kg; I² = 59.2%) but moderately increased weight gain in long-term interventions (0.66 kg; 95% CI: −0.14, 1.46 kg; I² = 80.7%). Furthermore, a significant reduction in body fat in the dairy group was shown in short-term studies (−0.91 kg; 95% CI: −1.43, −0.38 kg; I² = 70.3%), whereas no significant change was observed in long-term trials (0.32 kg; 95% CI: −0.17, 0.81 kg; I² = 59.5%).

Open in a separate window

Open in a separate window

FIGURE 4.

Net change (95% CI) by study duration (short-term: <1 y; long-term: ≥1 y) in body weight (A) and fat (B) associated with dairy interventions expressed as the change (kg) during the intervention with dairy products minus the change during control regimens. Horizontal lines denote the 95% CIs; solid diamonds represent the point estimate of each study. The open diamond represents the pooled estimate of the intervention effect. The dashed line denotes the point estimate of the pooled result. ID, identification; WMD, weighted mean difference.

Stratified analysis by both energy restriction and duration indicated that dairy products significantly facilitated weight loss in short-term trials with energy restriction, whereas a marginally significant weight gain was observed in long-term trials without energy restriction. For body fat, a significant reduction was shown in short-term studies with energy restriction, whereas a marginally significant weight gain was observed in long-term trials without energy restriction. No significant difference was observed in weight or body fat change between dairy and control groups in short-term RCTs without energy restriction or long-term energy-restricted RCTs. See Figure 1 under “Supplemental data” in the online issue for forest plots.

Sensitivity analysis

We conducted several sensitivity analyses by excluding 1 RCT that provided isoenergetic low-energy diet plans (14), 2 RCTs (27, 28) that included a physical activity–intervention component, 3 RCTs (3, 13, 16) for which change-from-baseline SDs were imputed, 3 RCTs of <3-mo duration (16, 25, 39), and 10 RCTs for which the Jadad score was <3, respectively. For all analyses, results were not substantially changed (data not shown).

We also carried out sensitivity analyses to explore heterogeneity. After excluding the study of Stancliffe et al (36) from short-term trials without energy restriction, the heterogeneity in that subgroup decreased dramatically (I² = 78.4–0.0%), and a similar impact of the study of Manios et al (5) was observed on the heterogeneity of long-term studies without energy restriction (I² = 63.4–3.0%).

The authors thank Jo-Anne Gilbert, Leonie van Meijl, Cristina Palacios, and Andrea Josse for providing their unpublished data. We appreciate Eric Ding for his valuable comments.

The authors’ responsibilities were as follows—MC, AP, VSM, and FBH: designed the research; MC and AP: performed the systematic review and analyzed data; AP: reviewed data; FH: had primary responsibility for the final content of the manuscript; and all authors: wrote, reviewed, and approved the final manuscript. None of the authors had a conflict of interest.