Testing for baseline differences in randomized controlled trials: an unhealthy research behavior that is hard to...

Area of Science:

• Clinical Trials Methodology
• Biostatistics
• Health Services Research

Background:

• Significance testing of baseline differences in randomized controlled trials (RCTs) is discouraged by CONSORT guidelines and numerous experts.
• Despite recommendations, there is persistent demand for baseline difference testing in nutrition behavior research.
• This paper aims to provide evidence against the statistical testing of baseline differences between intervention groups.

Purpose of the Study:

• To demonstrate why statistical testing of baseline differences in RCTs is superfluous.
• To provide evidence supporting the cessation of publishing significance tests for baseline differences.
• To advocate for improved covariate selection in statistical analyses of RCTs.

Main Methods:

• Review of existing literature and expert recommendations on baseline difference testing.
• Analysis of data from four supermarket trials examining pricing strategies on fruit and vegetable purchases.
• Comparison of fully adjusted analyses with analyses adjusted only for significant baseline differences.

Main Results:

• Testing baseline differences is flawed as it overlooks the prognostic strength of variables for confounding adjustment.
• Selecting covariates based on significance tests may lead to the omission of important variables or inclusion of irrelevant ones.
• Analyses adjusted for significant baseline differences only can yield results that appreciably differ from fully adjusted analyses.

Conclusions:

• Journals should cease publishing significance tests for baseline differences, adhering to CONSORT 2010 recommendations.
• Researchers should adjust for known or anticipated important prognostic variables, pre-specified in trial protocols.
• Reporting both fully adjusted and crude analyses is recommended, particularly for dichotomous and time-to-event data.