Critical Appraisal of the Estimation of Body Composition Via Two-, Three-, and Four-Compartment Models R.T. WITHERS, 1 * J. LAFORGIA, 2 AND S.B. HEYMSFIELD 3 1 Exercise Physiology Laboratory, School of Education, The Flinders University of South Australia, GPO Box 2100, Adelaide 5001, Australia 2 School of Pharmacy and Medical Sciences, University of South Australia, GPO Box 2471, Adelaide 5001, Australia 3 Obesity Research Center, St Luke’s-Roosevelt Hospital, Columbia University College of Physicians and Surgeons, New York, New York 10025, USA ABSTRACT This review explores the robustness of the assumptions un- derpinning the two- (fat mass [FM], fat-free mass [FFM]), three- (FM; total body water [TBW], fat free dry solid), and four- (FM; TBW; bone mineral [BM], residual) compartment models of body composition. The measurement of body density (BD) via underwater weighing (UWW) and TBW via isotopic dilution are the two most frequently used two-compartment techniques. The former assumes that the FM and FFM have densities of 0.9007 g/cm 3 and 1.1000 g/cm 3 , respectively, while the latter uses a FFM hydration constant. Although both techniques can estimate body composition precisely (technical error of measurement [TEM]: UWW  0.4 %BF; TBW  0.6 %BF), the validity of these estimates is adversely affected by biological variability in the assumed percentages for the FFM components (TBW  73.72%; protein  19.41%; BM  5.63%; non-BM  1.24%). The three-compartment model, which incorporates measures of BD and TBW, greatly increases validity by removing errors relating to variability in TBW, which comprises the largest percentage of the FFM and is furthermore acutely variable. The four- compartment model marginally improves on the three-compartment model by additionally controlling for BM, which displays less variability than the TBW component of the FFM. The three- and four-compartment models there- fore provide more valid estimates of body composition than the two- compartment model, and this increased accuracy is not offset by propagation of errors (TEM  0.7 %BF for both models) from the combinations of multiple measurements (BD, TBW, BM). Am. J. Hum. Biol. 11:175–185, 1999. © 1999 Wiley-Liss, Inc. Hydrodensitometry dates back to the turn of the century when Stern (1901) and Spi- vak (1915) attempted to quantify corpulence by underwater weighing. The technique failed to meet expectations until almost four decades later when Behnke et al. (1942) re- fined it for density determination. Subse- quent cadaver analyses resulted in the clas- sical two-compartment hydrodensitometric model (Broz ˇek et al., 1963; Siri, 1956, 1961) which separates the body into the fat mass (FM) and fat-free mass (FFM). This has de- veloped into three (Siri, 1956, 1961: FM, to- tal body water [TBW], fat-free dry solid) and four-compartment models (Heymsfield et al., 1990; Selinger, 1977: FM, TBW, bone mineral mass, residual). While body compo- sition research has expanded to include such techniques as gamma neutron activa- tion analysis, neutron inelastic scattering, computed tomography, and magnetic reso- nance imaging, this review concentrates on the prototype two-, three-, and four-com- *Correspondence to: Dr. R.T. Withers, Exercise Physiology Laboratory, School of Education, The Flinders University of South Australia, GPO Box 2100, Adelaide 5001, Australia. E- mail: edrtw@flinders.edu.au Received 15 October 1998; Accepted 19 November 1998 AMERICAN JOURNAL OF HUMAN BIOLOGY 11:175–185 (1999) © 1999 Wiley-Liss, Inc.