Abstract Objective To compare BMI with abdominal skinfold thickness (ASF), waist circumference and waist-to-height ratio in the prediction of insulin resistance (IR) in prepubertal Colombian children. Design We calculated age- and sex-specific Z -scores for BMI, ASF, waist circumference, waist-to-height ratio and three other skinfold-thickness sites. Logistic regression with stepwise selection ( P = 0·80 for entry and P = 0·05 for retention) was performed to identify predictors of IR and extreme IR, which were determined by age- and sex-specific Z -scores to identify the ≥ 90th and ≥ 95th percentile of homeostasis model assessment (HOMA IR ), respectively. We used receiver operating characteristic curves to compare the area under the curve between models. Setting Bucaramanga, Colombia. Subjects Children ( n 1261) aged 6–10 years in Tanner stage 1 from a population-based study. Results A total of 127 children (seventy girls and fifty-seven boys) were classified with IR, including sixty-three children (thirty-three girls and thirty boys) classified with extreme IR. Only ASF and BMI Z -scores were retained as predictors of IR by stepwise selection. Adding ASF Z -score to BMI Z -score improved the area under the curve from 0·794 (95 % CI 0·752, 0·837) to 0·811 (95 % CI 0·770, 0·851; P for contrast = 0·01). In predicting extreme IR, the addition of ASF Z -score to BMI Z -score improved the area under the curve from 0·837 (95 % CI 0·790, 0·884) to 0·864 (95 % CI 0·823, 0·905; P for contrast = 0·01). Conclusions ASF Z -score predicted IR independent of BMI Z -score in our population of prepubertal children. ASF and BMI Z -scores together improved IR risk stratification compared with BMI Z -score alone, opening new perspectives in the prediction of cardiometabolic risk in prepubertal children.