Objective: Most current methods for normalizing cardiac dimensions for differences in body size assume a linear isometric relationship between variables. This approach may be flawed. We therefore compared the isometric and allometric modeling approach to assess the optimal scaling method (using the left atrial dimension [LAD] as an example) in a large group of adults with normal echocardiograms.
Methods: We studied 15,667 consecutive echocardiographically normal adults. Their mean age was 40 +/- 14 years; LAD was 32 +/- 4 mm; body weight (BW) was 72 +/- 17 kg; height (HT) was 1.7 +/- 0.2 m; body surface area (BSA) was 1.8 +/- 0.2 m(2); and body mass index was 25 +/- 6 kg/m(2). The LAD in the anterior-posterior dimension was indexed by each size variable using both isometric and allometric methods.
Results: LAD correlated with all size variables (r = 0.20 to r = 0.45). Simple isometric indices (e.g., LAD/BSA), rather than removing, actually increased the correlation with all body size variables (r = -0.27 to -0.80). Normalization by the optimal allometric exponent (BW = 0.262; HT = 0.428; BSA = 0.449; body mass index = 0.266) eliminated the association of the indexed variable with body size. Indexation by BW best removed the influence of all measures of body size. The proportion of variance of LAD explained by BW was 21.4% compared with only 3.9% for HT. There were minor, but statistically significant, age-, gender-, and obesity-related differences in the indexed left atrial size (e.g., optimal allometric exponent for BW in male patients = 0.257 vs. in female patients = 0.253).
Conclusion: The standard isometric corrections for body size fail to adequately account for the associations between LAD and body size, and in fact increase the relationship of the indexed parameter to all body size measures. Normalization using the optimal allometric exponent removes the effect of that variable, with BW best removing the effect of body size.