Current body mass index (BMI) norms for children and adolescents are developed from a reference population that includes obese and slim subjects. The validity of these norms is influenced by the observed secular increase in body weight and BMI. We hypothesized that the performance of children in health-related physical fitness tests would be negatively related to increased BMIs, and therefore fitness tests might be used as criteria for developing a more appropriate set of BMI norms. We evaluated the existing data from a nation-wide fitness survey for students in Taiwan (444 652 boys and 433 555 girls) to examine the relationship between BMI and fitness tests. The fitness tests used included: an 800/1600-m run/walk; a standing long jump; bent-leg curl-ups; and a sit-and-reach test. The BMI percentiles developed from the subgroup whose test scores were better than the ‘poor’ quartile in all four tests were compared with those of the whole population and linked to the adult criteria for overweight and obesity. The BMIs were significantly related to the results of fitness testing. A total of 43% of students had scores better than the poorest quartile in all of their tests. The upper BMI percentile curves of this fitter subgroup were lower than those of the total population. The 85th and 95th BMI percentile values of the fitter 18-year-old-students (23.7 and 25.5 kg m−2 for boys; 22.6 and 24.6 kg m−2 for girls) linked well with the adult cut-off points of 23 and 25 kg m−2, which have been recommended as the Asian criteria for adult overweight and obesity. Hence, the BMI norms for children and adolescents could be created from selected subgroups that have better physical fitness. We expect that the new norms based on this approach will be used not only to assess the current status of obesity or overweight, but also to encourage activity and exercise.