Comparing g-computation, propensity score-based weighting, and targeted maximum likelihood estimation for analyzing externally controlled trials with both measured and unmeasured confounders: a simulation study

Table 1 Summary of potential confounders between treatment groups

Scenario/Variable	Control (A = 0)	Treatment (A = 1)	Standardized mean difference
Scenario 1
U (mean ± SD)	0.45 ± 0.99	0.56 ± 1.01	0.08
C (mean ± SD)	46.97 ± 6.12	54.39 ± 6.06	0.86
L (%)	17.96	33.98	0.42
Scenario 2
U (mean ± SD)	0.40 ± 0.97	0.62 ± 1.02	0.16
C (mean ± SD)	46.74 ± 6.08	54.24 ± 5.93	0.88
L (%)	20.63	40.38	0.44
Scenario 3
U (mean ± SD)	0.27 ± 0.95	0.76 ± 1.00	0.36
C (mean ± SD)	46.66 ± 6.14	53.84 ± 6.05	0.83
L (%)	17.38	42.92	0.55

Scenario 1, 2 & 3 stands for small and blocked confounding paths, medium and blocked confounding paths, and one large unblocked confounding path, respectively. The standardized mean difference expresses the size of the intervention effect in each study relative to the variability observed in that study
SD Standard deviation, U Unmeasured confounder, C & L Measured confounders

ISSN: 1471-2288