Table 3 Parameter values and trial characteristics initially chosen for the simulation-based power calculations of the IPD meta-analysis of pregnancy trials

No. of trials

14

Number of studies included in a previous aggregate data meta-analysis that had promised their IPD

Sample sizes

50, 931, 125, 85, 235, 327, 45, 12, 39, 142, 105, 84, 15, 124

Total sample size: taken from original trial publications (could breakdown further into the number in control and treatment groups if unequal)

α _i

13.30, 5.00, 9.68, 10.60, 11.50, 15.24, 14.20, 5.20, 5.00, 12.40, 13.80, 8.00, 15.70, 15.40

Mean weight gain in the control group: used values as stated in original trial publications

β

−0.28

Prognostic effect of BMI on weight gain: used estimate from a meta-regression of mean weight gain against mean baseline BMI in the control group

θ

−0.84

Mean treatment effect across trials: used summary estimate from random effects meta-analysis of published estimates from the 14 trials

λ

Various: −0.5, − 0.4, − 0.3, − 0.2, − 0.1, −.05, − 0.025, − 0.01

Magnitude of interaction: used range from extremely large to extremely small interaction effect

\( {\sigma}_i^2 \)

43.25, 15.57, 119.26, 52.69, 16.98, 37.37, 33.89, 6.63, 12.93, 12.63, 15.22, 12.93, 13.78, 40.53

Residual variance: used unweighted average of the variance values for treatment and control groups as stated in original trial publications

\( {\tau}_{\beta}^2 \)

0

Between-study variance of the prognostic effect of baseline BMI: set to zero for parsimony

\( {\tau}_{\theta}^2 \)

1.1

Between-study variance of overall treatment effect: used estimate from random effects meta-analysis of published estimates from the 14 trials

\( {\tau}_{\lambda}^2 \)

0

Between-study variance of interaction effect: set to zero for parsimony

Distribution of baseline BMI values

Study 1: N(34.75, 12.5)

Study 2: N(30.15, 25.51)

Study 3: N(37.95, 0.49)

Study 4: N(33.63, 14.77)

Study 7: N(24.55, 27.45)

Study 8: N(35.1, 12.25)

Study 10: N(23.85, 0.25)

Study 12: N(25.45, 13.45)

Study 13: N(23.65, 18.13)

Other studies had a mean drawn from N(30, 2.5) and within-trial standard deviation set at 3.5

Distribution of key covariate of interest: assumed normality, with means and variances as stated in original trial publication, or if unavailable, values based on those observed from within and between other trials