Partial Alignment of Electrocardiographic Subspaces Associated With Obesity and Heart Failure in the Third National Health and Nutrition Examination Survey.

Background Obesity and heart failure are each associated with electrocardiographic remodeling, but it is unknown whether their multidimensional electrical phenotypes share common directions within a feature space. Methods We conducted a cross-sectional analysis of historical Third National Health and Nutrition Examination Survey data (1988-1994) from 8,324 adults aged 40 years or older with complete data on obesity, self-reported physician-diagnosed heart failure, survey design variables, and resting 12-lead electrocardiography. Forty-seven features were adjusted for age, sex, and race/ethnicity, standardized, and transformed using covariance whitening relative to nonobese participants without heart failure. Across 30 repeated analyses that preserved the survey design and used one half of the sample, we estimated two-dimensional subspaces based on excess second moments. The obesity subspace was derived from obese participants without heart failure, and the heart failure subspace was derived from nonobese participants with heart failure. The primary statistic was the mean squared canonical cosine. Inference used a permutation test that repeated the full analytical pipeline in 5,000 iterations stratified by survey design. Results The cohort included 5,691 reference participants, 2,216 obese participants without heart failure, 287 nonobese participants with heart failure, and 130 participants with both conditions. The mean squared canonical alignment across the survey design splits was 0.205, compared with a permutation null mean of 0.088 and a 99th percentile of 0.120 (empirical p = 0.0002). Alignment was 0.216 after clinical adjustment, 0.219 when restricted to QRS durations below 120 ms, and 0.132 after excluding prior myocardial infarction and major conduction abnormalities. Full sample descriptive principal angles were 32.0 and 81.8 degrees. The leading shared direction was dominated by T-wave amplitudes and frontal T-axis features. Conclusions The electrocardiographic subspaces associated with obesity and self-reported physician-diagnosed heart failure demonstrated statistically nonrandom but incomplete alignment. One shared direction was dominated by repolarization features, while substantial structure remained unique to each survey-defined phenotype. These subspaces represent surface electrophysiological signatures of survey phenotypes, and the calculated principal angles are mathematical representations of historical data (1988-1994), rather than physiologically confirmed disease entities. They should not be interpreted as evidence of true physiological or subclinical heart failure mechanisms. These findings identify a target for longitudinal validation.