Current controversies: Null hypothesis significance testing

Area of Science:

• Medical Statistics
• Clinical Research Methodology
• Obstetrics and Gynecology

Background:

• Null hypothesis significance testing (NHST) with a 0.05 significance level is standard in healthcare, particularly in obstetric and gynecological research.
• NHST's application for inferring clinical significance from statistical significance is a common misunderstanding of its original purpose.
• Limitations of NHST, including sensitivity to sample size and susceptibility to Type I and II errors, are frequently overlooked.

Purpose of the Study:

• To critically examine the historical context and inherent limitations of traditional null hypothesis significance testing (NHST).
• To address the controversy surrounding the interpretation and application of NHST in clinical decision-making, especially in obstetric and gynecological research.
• To propose a reconsideration of current statistical practices and advocate for a statistics reform concerning NHST.

Main Methods:

• This commentary reviews the historical development and conceptual underpinnings of null hypothesis significance testing (NHST).
• It analyzes the statistical and practical limitations of NHST as applied in contemporary health research.
• The commentary synthesizes arguments against the misinterpretation of statistical significance for clinical importance.

Main Results:

• NHST, particularly the p < 0.05 threshold, is often misapplied to infer clinical significance, leading to potential misinterpretations.
• The reliance on NHST can result in erroneous claims regarding intervention effectiveness or the importance of risk factors.
• Overlooking NHST's limitations, such as sample size sensitivity and error types, compromises the reliability of research findings.

Conclusions:

• The current use of NHST in healthcare decision-making, especially in obstetric and gynecological research, is problematic due to widespread misinterpretation.
• A significant gap exists between statistical significance (p-values) and clinical relevance, which NHST fails to adequately address.
• There is a compelling need for a reform in statistical methodologies used in research to ensure more accurate and meaningful interpretation of results.