(Unlearned) lessons from John Graunt and Kenneth Rothman: a "CLASSic" example

Area of Science:

• Epidemiology
• Biostatistics
• Medical Research Methodology

Background:

• Historical contributions of John Graunt and Kenneth J. Rothman to statistical reasoning.
• The prevalent reliance on statistical significance in interpreting study results.
• The critical need for predefined study endpoints in research.

Purpose of the Study:

• To review the foundational work on statistical significance and prespecification.
• To critique the overreliance on statistical significance for causal inference.
• To advocate for a more rational approach to interpreting research findings.

Main Methods:

• Commentary synthesizing historical and contemporary perspectives.
• Analysis of the dangers of substituting statistical significance for causal inference.
• Illustrative example using the Celecoxib Long-term Arthritis Safety Study.

Main Results:

• Statistical significance is a flawed sole criterion for evaluating study outcomes.
• Prespecification of study endpoints is essential for robust causal inference.
• The Celecoxib Long-term Arthritis Safety Study exemplifies the pitfalls of post-hoc analysis.

Conclusions:

• Researchers must move beyond a simplistic reliance on statistical significance.
• Adopting prespecified endpoints enhances the validity of causal inference.
• A rational approach to causal inference is paramount in scientific research.