Accurate estimates of total body water and fat-free mass can be achieved by using equations with bioelectrical resistance and height (H) (1-3). Measurements of whole-body resistance (R) and reactance (Xc) at multiple frequencies in the range of 5-1000 kHz did not provide any significant improvement over measurements made at 50 kHz for the estimation of extracellular and total body water when using any combination of R and Xc (4). In patients with fluid overloading, equations validated in a healthy population produce biased estimates of body compartment volumes. If the normal range of the impedance vector were known in the reference population, the traditional prediction formulas could be applied with confidence to subjects lying in the reference interval. Following this rationale and assuming the bivariate normal distribution (5) of the vector components R/H and Xc/H, we recently calculated the bivariate, 75% and 95%, sex-specific tolerance limits of the normal impedance vector (ie, the ellipses of the healthy population, within which the vector of the individual subject falls with a probability of 75% and 95%, respectively) and we identified a bioelectrical impedance threshold for apparent edema on the lower pole of the sexspecific 75% tolerance ellipse (6). The performance of the threshold in detecting apparent edema was operating with 97% sensitivity and 81% specificity. Moreover, the frequency distributions of vectors falling out of the tolerance ellipses did not significantly differ when comparing body segments with the whole body, thus excluding operative advantages of segmental measurements in renal patients with fluid overloading (7). However, since the estimation of a normal range may be critically dependent on the sample size of the reference population, we extended the study to a population of 726 healthy subjects, 354 males [median age 48 y, range 15-85 y, body mass index 17-31 (kg/rn2)] and 372 females [median age 50 y, range 15-85 y, body mass index 16-31 (kg/m2)] living in northern Italy (Padua, Venice, and Florence). All measurements [standard tetrapolar method (1-4)] were made in the three centers by using three impedance plethysmographs of the same model (BIA-109; RJL/Akern Systems, Detroit), which emitted 800 p.A and a 50 kHz alternating sinusoidal current. The center effect (ie, the sum of operator and instrument variabilities), expressed as a CV, was 3% for R and 8% for Xc. In males, RIB (1 ± SD) was 298.6 ± 43.2 fl/rn; Xc/H was 30.8 ± 7.2 fI/m. R/H and Xc/H were significantly correlated in males (r = 0.47, P < 0.001). In females, R/H (1 ± SD) was 371.9 ± 49.0 f1/m; Xc/H was 34.4 ± 7.7 fl/rn. R/H and Xc/H were also significantly correlated in females (r = 0.41, P < 0.001). The 95% confidence ellipses (see Figure 1) of the average impedance vector did not overlap, indicating a significantly longer vector in females, which was also found in a previous pilot study with 86 healthy subjects (6). Of the 726 healthy subjects, 95 males and 91 females were aged 65-85 y, with body mass indexes of 24 (kg/rn2, range 18-30) and 25 (kg/rn2, range 16-31), respectively. Neither R/H nor Xc/H was significantly correlated with age either below or above 65 y (-0.16 < r < 0.13 in males, and -0.10 < r < 0.02 in females). Therefore, we did not plot the set of overlapping tolerance ellipses by age classes. In Figure 1, three sex-specific tolerance ellipses are drawn corresponding to the median, 75th, and 95th percentiles of the impedance vector distribution calculated from our healthy population. Interestingly, despite equal slopes, the ellipses were larger in females, which is consistent with a greater variability in female body composition (6). However, the threshold for apparent edema, ie, the lower pole of the 75% tolerance ellipse, reached similar Xc/H values in males and females, as was found in the pilot study (6). In practice, plotting direct R/H and Xc/H measurements on the population percentile distributions (ie, on the tolerance ellipses) allows one to decide whether a subject has maintained normal hydration. If estimates of volume and mass of body compartments are of interest, the traditional prediction formulas (1-3) can be safely applied with the probability level indicated by the ellipse percentile. Further studies are necessary to interpret the longer vectors falling out of the upper poles, as in dehydration, and the vectors of normal length but with extreme phase angles (ie, beyond the left or right ends of the small axes of the tolerance ellipses), as expected in extremes of lean body mass (eg, in athletes or in anorexia or wasting conditions) (6). Data from renal patients and from the literature, plotted as “RXc path graph” over the tolerance ellipses support a simple operative rule in the bedside interpretation of bioelectrical impedance analysis measurements: whereas a progressive shortening and down-sloping of the impedance vector over successive measurements indicates