Body Mass Index Cut-off Points for Tunisian Children and Adolescents: Derivation and Comparison with International Standards

Abstract

Background: Population-specific body mass index references account for genetic and environmental variations affecting pediatric growth patterns. Tunisia established BMI reference curves in 2018 and 2021 but lacked validated clinical cut-off points for routine health assessments.

Aim: The study aimed to (i) derive BMI cut-off points for Tunisian children and adolescents aged 6-17.99 years using the Lambda-Mu-Sigma method, and (ii) evaluate agreement between Tunisian cut-offs and International Obesity Task Force and World Health Organization standards.

Methods: We analyzed cross-sectional anthropometric data from 3,533 participants (1,760 girls, 1,773 boys; age 6.00-18.99 years) collected during 2012-2013 through stratified random sampling across northern, central, and southern Tunisia. The LMS method identified percentiles corresponding to adult BMI thresholds of 18.5 kg/m² (thinness), 25 kg/m² (overweight), and 30 kg/m² (obesity) at age 18 years. Agreement between classification systems was quantified using Cohen's kappa coefficients and chi-square tests.

Results: Cut-off points corresponded to the 18.41st percentile (thinness), 88.69th percentile (overweight), and 98.81st percentile (obesity) for girls. Boys showed cut-offs at the 17.88th, 86.65th, and 96.86th percentiles respectively. Maximum deviations from IOTF reached 1.08 kg/m². Deviations from WHO reached 3.12 kg/m². Thinness frequency using Tunisian cut-offs (15.57%) exceeded IOTF (10.4%) and WHO (2.8%) estimates (P<0.001). Agreement with IOTF was substantial (kappa=0.772 for thinness) but fair with WHO (kappa=0.274).

Conclusion: Tunisia-specific BMI cut-offs differ from international standards, particularly for thinness classification. External validation in independent samples should precede widespread clinical adoption. These findings support development of population-tailored growth assessment tools for North African contexts.