Fig. 1

Adjusted risk estimates and model performance for predicting MetS among OW/OB adolescents. (A) Forest plot of adjusted odds ratios (ORs) with 95% confidence intervals (CIs) from the final multivariable logistic regression model predicting MetS among adolescents with OW/OB. The vertical dashed line denotes OR = 1; dots indicate point estimates and horizontal bars the 95% CIs. Variables retained in the final model included: WHtR ≥ 0.46, unwillingness to exercise, sleep < 8 h/night, CAP ≥ 248 dB/m, watching TV > 2 h/day, milk-tea consumption, HbA1c ≥ 5.7%, household snack availability, eating while watching TV/phone, and eating out ≥ 3 times/week; exclusive breastfeeding for 6 months was protective. (B) Receiver-operating-characteristic (ROC) curve for the 10-variable model. Model performance: AUC = 0.77; optimal probability threshold = 0.481, sensitivity = 70.3%, specificity = 70.1%, PPV = 70.0%, NPV = 70.1%. The ORs shown in the figure were derived from the adjusted ORs (aORs) reported in Supplementary Tables S2, S3, and S4; variables with statistically significant associations were entered into a final multivariable logistic regression model to predict MetS. All ORs were adjusted for covariates as described in the Methods section. See Methods for details on variable coding and multiple-imputation procedures. Two-sided tests were used for all analyses. ROC, receiver operating characteristic; AUC, area under the curve; PPV, positive predictive value; NPV, negative predictive value; WHtR, waist-to-height ratio; CAP, controlled attenuation parameter; HbA1c, glycated hemoglobin.