Statistical reassessment of the association between waist circumference and clustering metabolic abnormalities in Japanese population.

AIM

The use of ethnic-specific cutoff values of waist circumference (WC) has been recently recommended, but they were originally developed on different statistical grounds. We investigated whether different statistical procedures and clinical settings generate different WC cutoff values in one ethnic population.

METHODS

We recruited 3810 Japanese subjects and performed the following three statistical analyses: 1) search for WC cutoff points associated with the risk of clustering metabolic abnormalities, 2) calculation of WC associated with certain body mass index (BMI) levels, 3) evaluation with receiver operating characteristic (ROC) curves for the risk. We also simulated population models to evaluate whether WC cutoff values depend on the clinical settings of the study population.

RESULTS

First, in risk analysis, males and females had the same risk when females had 10 cm larger WC than males, although the risk increased almost linearly, without any clear threshold. Second, WC corresponding to BMI of 25 kg/m(2) was 87 cm in males and 85 cm in females, with a slight sex difference. Third, ROC curves showed that the male optimal cutoff value was 85 cm, larger than the female one (79 cm). However, simulated population models with various WC distributions gave different ROC curves and different WC cutoff values. Furthermore, WC cutoff values varied by age in any of the three statistical procedures.

CONCLUSIONS

Different statistical grounds could generate different WC cutoff values. If one plans to establish a unified trans-ethnic diagnostic tool of metabolic syndrome, ethnic-specific WC cutoff values should be primarily provided based on a unified statistical ground.