Background and Objectives: Anthropometric equations are widely used to estimate body composition however, are only appropriate for use in populations in which they have been developed and validated. The present study developed anthropometric prediction equations for Indonesian adults and cross-validated them with selected equations used in this population.

Methods and Study Design: Six hundred Indonesian adults aged between 18-65 years (292 males and 308 females) were divided equally into development and validation groups. Stature, body weight, skinfold thickness at eight sites, girth at five sites, and bone breadth at four sites were measured. Stepwise multiple regression analysis was used to propose percentage body fat (%BF) prediction equations using measured variables and %BF from the deuterium oxide dilution technique as the reference. The proposed prediction equations were then cross-validated using the validation group and %BF estimated from several existing equations.

Results: Proposed prediction equations showed r ranged from 0.82 to 0.86 and Standard Error of the Estimate (SEE) from 4.7 to 5.4%. Cross-validation analysis showed bias with the reference %BF between 0.2 and 3.3% and Pure Error (PE) between 2.8 and 4.0%. Among the existing equations, the Durnin and Womersley equation was applicable in females whilst the equation by Davidson et al. underestimated %BF by 6.3–6.6% and the equation by Gurrici et al overestimated by 2.0–3.4% in both genders (p < 0.01).

Conclusion: The proposed prediction equations provide better options for accurate prediction of %BF in Indonesian adults.