Bonferroni-Holm and permutation tests to compare health data: methodological and applicative issues

Area of Science:

• Health Research
• Biostatistics
• Gastroenterology

Background:

• Statistical methodologies are crucial in health research but often misused, especially with multiple comparisons.
• Controlling Type I errors and false positive rates is essential when analyzing multiple variables.
• Permutation testing offers a robust method for assessing simultaneous significance and identifying key factors.

Purpose of the Study:

• To apply permutation tests to differentiate between Crohn's Disease (CD) and Ulcerative Colitis (UC) within Inflammatory Bowel Diseases.
• To identify specific clinical and demographic variables that significantly distinguish CD from UC patients.
• To address the challenge of multiplicity control in statistical comparisons between the two disease groups.

Main Methods:

• Utilized permutation tests, specifically the Non-Parametric Combination (NPC) test, for analyzing differences between CD and UC patient groups.
• Employed the Sequentially Rejective Multiple Test (Bonferroni-Holm procedure) to manage multiplicity and identify significant partial tests.
• Compared patient characteristics including diagnosis age, smoking status, treatment history, re-hospitalization rates, and disease severity (Charlson score).

Main Results:

• Permutation testing revealed significant differences between CD and UC patients across several variables.
• Nine variables were identified as statistically significant discriminators after applying the Bonferroni-Holm procedure: Smoking habit, Immunosuppressive therapy, Surgery, Biological Drug, Diabetes, Adverse Events, Re-hospitalization, Gender, and Duration of Immunosuppressive Therapy.
• UC patients generally presented with older diagnosis age and higher disease severity, while CD patients showed higher rates of re-hospitalization and surgery.

Conclusions:

• The study successfully identified significant variables that effectively discriminate between Crohn's Disease and Ulcerative Colitis.
• The selected nine variables (Smoking habit, Immunosuppressive therapy, Surgery, Biological Drug, Diabetes, Adverse Events, Hospitalization, Gender, Duration of Immunosuppressive Therapy) robustly distinguish the two inflammatory bowel diseases.
• The application of permutation testing and the Bonferroni-Holm procedure addressed multiplicity issues, providing reliable findings for differentiating CD and UC.