Elusive reproducibility

Abstract

Reproducibility remains a mirage for many biomedical studies because inherent experimental uncertainties generate idiosyncratic outcomes. The authentication and error rates of primary empirical data are often elusive, while multifactorial confounders beset experimental setups. Substantive methodological remedies are difficult to conceive, signifying that many biomedical studies yield more or less plausible results, depending on the attending uncertainties. Real life applications of those results remain problematic, with important exceptions for counterfactual field validations of strong experimental signals, notably for some vaccines and drugs, and for certain safety and occupational measures. It is argued that industrial, commercial and public policies and regulations could not ethically rely on unreliable biomedical results; rather, they should be rationally grounded on transparent cost-benefit tradeoffs.