Impact of Meal Frequency on Anthropometric Outcomes: A Systematic Review and Network Meta-Analysis of Randomized Controlled Trials

Abstract

The relation between meal frequency and measures of obesity is inconclusive. Therefore, this systematic review and network meta-analysis (NMA) set out to compare the isocaloric effects of different meal frequencies on anthropometric outcomes and energy intake (EI). A systematic literature search was conducted in 3 electronic databases (Medline, Cochrane Library, Web of Science; search date, 11 March 2019). Randomized controlled trials (RCTs) were included with ≥2 wk intervention duration comparing any 2 of the eligible isocaloric meal frequencies (i.e., 1 to ≥8 meals/d). Random-effects NMA was performed for 4 outcomes [body weight (BW), waist circumference (WC), fat mass (FM), and EI], and surface under the cumulative ranking curve (SUCRA) was estimated using a frequentist approach (P-score: value is between 0 and 1). Twenty-two RCTs with 647 participants were included. Our results suggest that 2 meals/d probably slightly reduces BW compared with 3 meals/d [mean difference (MD): −1.02 kg; 95% CI:−1.70,−0.35 kg) or 6 meals/d (MD:−1.29 kg; 95% CI:−1.74,−0.84 kg; moderate certainty of evidence).We are uncertain whether 1 or 2 meals/d reduces BW compared with ≥8 meals/d (MD1 meal/d vs. ≥8 meals/d: −2.25 kg; 95% CI: −5.13, 0.63 kg; MD2 meals/d vs. ≥8 meals/d: −1.32 kg; 95% CI: −2.19, −0.45 kg) and whether 1 meal/d probably reduces FM compared with 3 meals/d (MD: −1.84 kg; 95% CI: −3.72, 0.05 kg; very low certainty of evidence). Two meals per day compared with 6 meals/d probably reduce WC (MD: −3.77 cm; 95% CI: −4.68, −2.86 cm; moderate certainty of evidence). One meal per day was ranked as the best frequency for reducing BW (P-score: 0.81), followed by 2 meals/d (P-score: 0.74), whereas 2 meals/d performed best for WC (P-score: 0.96). EI was not affected by meal frequency. In conclusion, our findings indicate that there is little robust evidence that reducing meal frequency is beneficial