Which statistical test is most appropriate for an individual‑based profiling (IBP) strategy comparing multiple pediatric groups with obesity status categorized into more than three levels?

Multinomial Logistic Regression is the Best Statistical Test for This Study

For a pediatric obesity study examining multiple groups with more than three obesity categories, multinomial logistic regression (Option D) is the required statistical approach. 1

Why Multinomial Regression is Mandatory

• Multinomial logistic regression is expressly designed to model polytomous (multiple-category) outcomes and is the recommended statistical method when pediatric obesity research involves more than three weight classifications. 1

• The American Heart Association specifies that comprehensive pediatric obesity assessment employs several BMI-percentile categories: underweight, normal weight, overweight, obesity, and severe obesity—creating four to five distinct diagnostic groups that cannot be analyzed with simpler methods. 2, 1

• By simultaneously modeling all outcome categories, multinomial regression avoids the inflation of Type I error that occurs when performing multiple separate binary analyses. 1

• Preserving the full gradient of risk across weight categories—rather than collapsing them into binary outcomes—maintains critical clinical information and enhances statistical power. 1

Why the Other Options Are Incorrect

T-test (Option A) is Wrong

• T-tests only compare means between two groups and cannot handle multiple categorical outcomes. 1
• With more than three obesity categories, you would need multiple pairwise t-tests, which inflates Type I error and discards clinically relevant information. 1

Linear Regression (Option B) is Wrong

• Linear regression assumes a continuous dependent variable, but obesity categories are discrete, ordered classifications (normal, overweight, obese, severely obese). 1
• Treating categorical outcomes as continuous violates fundamental statistical assumptions and produces invalid results. 1

Logistic Regression (Option C) is Wrong

• Standard binary logistic regression only handles two-category outcomes (e.g., obese vs. non-obese). 1
• Using multiple binary logistic regressions for each pairwise comparison inflates Type I error and fails to model the polytomous outcome structure correctly. 1

Clinical Justification for Multiple Categories

• Severe obesity (BMI ≥99th percentile or ≥120% of the 95th percentile) exhibits a distinct risk-factor profile compared with lower weight categories, requiring separate analytical treatment. 2, 1

• Children with severe obesity have higher prevalence of metabolic syndrome clustering, with Hispanic and non-Hispanic Black youth showing elevated prevalence across all obesity definitions. 2, 1

• Approximately 4-6% of U.S. children aged 2-19 years have severe obesity, representing a clinically distinct population from those with standard obesity (BMI 95th-99th percentile). 2, 1

Empirical Evidence Supporting This Approach

• Multiple published studies have successfully applied multinomial logistic regression to predict weight categories from BMI percentiles and demographic factors in pediatric cohorts, confirming its practical utility. 3, 4, 5

• Researchers should avoid collapsing multiple obesity categories into binary outcomes or conducting separate binary logistic regressions for each pairwise comparison, as these practices inflate Type I error and discard clinically relevant information. 1