Aim We aimed to assess the influence of fasting insulin resistance on metabolically healthy obesity (MHO) prevalence in adolescents and to identify associated factors. Methods This retrospective, registry‐based, cross‐sectional study included 418 (51.9% girls) 10‐ to 18‐year‐old adolescents with obesity from a tertiary outpatient clinic in Sao Paulo, Brazil, between 2009 and 2013. The prevalence of MHO was estimated according to two definitions: (i) no cardiometabolic risk factors (CMRF) by the International Diabetes Federation parameters and (ii) no CMRF and homeostatic model assessment for insulin resistance <3.16. Adjusted prevalence ratios and 95% confidence intervals (CI) evaluated the association of gender, age, pubertal stages, skin colour and degree of obesity with MHO. Results Metabolically healthy obesity prevalence was lower in definition II than definition I (12.7%; 95% CI 9.1–16.3% versus 43.1%; 95% CI 38.0–48.2%, respectively). Adjusted results showed negative association between severe obesity and MHO by both definitions (p ≤ 0.01). Male and later pubertal stages were also less likely to have MHO, but neither remained significant in definition II. Conclusion Metabolically healthy obesity prevalence decreased when insulin resistance was part of the definition. Detecting pre‐clinical insulin resistance may improve the management of treatment‐seeking adolescents, especially when they present no CMRF.

indivíduo maiores (n ≥ 30 e m ≥ 16, respectivamente) para detecção do modelo correto. Consideramos dois exemplos em que ilustramos a aplicação do AIC marginal e do AIC condicional. Mostramos que a seleção de modelos por intermédio dos critérios AIC e BIC é compatível com a análise de resíduos associada.