Table 2.

Stratified analyses of estimated causality between dairy intake and BMI in dominant model.

SubgroupDairy (outcome), serving/dayBMI (outcome), kg/m2Estimated causality
Studies SNP (IV)Studies SNP (IV)
No.β (95% CI)P valueNo.β (95% CI)P valueβ (95% CI)P value
Age, years
    ≥50190.19 (0.12–0.26)<0.001190.12 (0.05–0.19)<0.0010.64 (0.22–1.07)<0.001
    <5050.35 (0.17–0.53)<0.00160.06 (−0.13 to 0.24)0.5130.16 (−0.38 to 0.69)0.371
BMI, kg/m2
    ≥25190.20 (0.14–0.27)<0.001190.11 (0.06–0.17)0.0010.55 (0.23–0.88)<0.001
    <2550.21 (0.15–0.27)<0.00160.15 (−0.17 to 0.47)0.1470.68 (−0.83 to 2.20)0.269
Follow-up, years
    ≥5130.20 (0.13–0.27)<0.001140.11 (0.03–0.19)0.0010.55 (0.09–1.00)<0.001
    <5110.24 (0.13–0.36)<0.001110.13 (−0.01 to 0.26)0.1120.51 (−0.07 to 1.10)0.074
Sample size
    ≥1000200.23 (0.17–0.29)<0.001210.12 (0.05–0.18)<0.0010.51 (0.19–0.83)<0.001
    <100040.21 (0.15–0.27)<0.00140.04 (−0.35 to 0.43)0.9120.18 (−1.63 to 2.00)0.615
Ethnic group
    European190.21 (0.15–0.28)<0.001200.13 (0.08–0.18)<0.0010.62 (0.30–0.93)<0.001
    Non-European50.22 (0.04–0.41)<0.0015−0.18 (−0.42 to 0.07)0.611−0.79 (−2.07 to 0.49)0.662
Study design
    Cohort190.20 (0.17–0.22)<0.001200.12 (0.06–0.17)<0.0010.60 (0.32–0.88)<0.001
    Cross-sectional50.1 (0.13–0.19)<0.00150.11 (−0.01 to 0.22)0.2310.69 (−0.04 to 1.41)0.06
HWE
    ≤0.05200.23 (0.16–0.28)<0.001210.12 (0.07–0.19)<0.0010.52 (0.18–0.82)<0.001
    >0.0540.21 (0.16–0.28)<0.00140.05 (−0.34 to 0.45)0.7530.19 (−1.57 to 1.76)0.714
  • SNP, single-nucleotide polymorphism; HWE, Hardy–Weinberg equilibrium.

  • Linear regression was used to test the association of MCM6 variant rs4988235 with dairy intake or BMI after adjustment of age, sex, ethnicity, region, total energy, and principal component for population stratification, as appropriate.

  • We pooled β coefficients across studies using random-effect metaanalysis because of the heterogeneity between studies (I2 > 50%; P < 0.001).

  • We used the IV estimators to quantify the strength of the causal association of dairy intake and BMI in each study. The IV estimator that is identical to that derived by the widely used 2-stage least-squares method will be calculated as the β of the regression coefficients MCM6 rs4988235-BMI and MCM6 rs4988235-dairy.