External validation of a prediction model for estimating fat mass in Arab children and adolescents

Abstract

Abstract Background/Aims Current childhood fat mass (FM) assessment techniques are not suitable for clinical and population‐level adiposity assessment. A prediction model, which accurately estimates childhood FM using predictor variables of weight, height, age, sex and ethnicity, requires validation in Arab populations. We evaluate the model's performance in Kuwaiti, Lebanese and Moroccan children/adolescents. Methods Data from three cross‐sectional studies on 471 individuals, aged 6–15 years, were obtained with complete information on predictors and the outcome of log transformed fat‐free mass assessed by reference standard deuterium dilution (lnFFM). Country‐specific predictive performance statistics of R 2 , calibration slope and calibration‐in‐the‐large (measures the calibration/agreement between observed and predicted lnFFM with ideal values of 1 and 0, respectively) and root mean square error (RMSE) were quantified and pooled across countries via random‐effects meta‐analysis. FM estimates from bioimpedance were also available for Lebanese children and were compared to the reference standard. Results The model showed strong predictive ability in all populations. Pooled R 2 calibration slope and calibration‐in‐the‐large values on the original lnFFM scale were 87.73% (95% CI: 77.20, 98.26%), 0.95 (95% CI: 0.83, 1.08) and −0.03 (95% CI: −0.16, 0.11), respectively. Model intercepts were recalibrated in each country to improve accuracy; updated country‐specific equations are provided. After recalibration, RMSEs on the FM scale were 1.3, 1.6 and 2.8 kg in Kuwait, Lebanon and Morocco, respectively. The RMSE from the model was lower than bioimpedance (2.4 kg) amongst Lebanese children. Interpretation The model explained a large proportion of the variance in FM, produced well‐calibrated predictions and relatively low RMSEs in Arab settings. It predicted FM more accurately than bioimpedance, indicating its potential for implementation in clinical‐ and population‐level settings, particularly in low‐ and middle‐income countries.