ABSTRACT This article quantifies the uncertainty reduction achievable for counterfactual estimand, and cautions against potential bias when the estimand uses Digital Twins. Posed by Neyman in 1923 (Statistical Science, 1990;5: 465–472) who showed unbiased point estimation from designed factual experiments is possible, counterfactual uncertainty quantification (CUQ) remained an open challenge for about 100 years. The counterfactual efficacy we focus on is the ideal estimand for comparing treatment with control , the expected outcome differential if each patient received both and . Enabled by our new statistical modeling principle called ETZ, we show CUQ is achievable in Randomized Controlled Trials (RCTs) with Before‐and‐After Repeated Measures, common in many therapeutic areas. The CUQ we are able to achieve typically has lower variability than factual UQ. We caution against using predictors with measurement error, which violates regression assumptions and can cause attenuation bias in estimating treatment effects. For traditional medicine and population‐averaged targeted therapy, counterfactual point estimation remains unbiased. However, in both Real Human and Digital Twin approaches, estimating effects in subgroups may suffer attenuation bias.
Wang et al. (Fri,) studied this question.