Models with Weight and Height Better Predict Mortality versus BMI After Adult Heart Transplantation

Abstract

<h3>Purpose</h3> Although body mass index (BMI) is commonly used as a predictor of health outcomes, recent studies have questioned whether its use is optimal. The fixed relationship between weight and height (i.e., kg/m²) assumed in the BMI calculation may inadequately represent the clinical impact of body composition. This study investigates whether an explicit weight*height interaction effect improves prediction of mortality after heart transplant compared with BMI. <h3>Methods</h3> We included adults (>18 years old) who underwent HT included in the Scientific Registry of Transplant Recipients between January 2000 and May 2021. We developed prediction models using log BMI or log height-weight interaction for 1- and 5-year mortality. We evaluated model performance using <i>R²</i> and C-statistic, with higher values indicating improved performance. We also modeled event-free probability at 1- and 5 years across a wide range of weight-height combinations with predictors set at their median value. <h3>Results</h3> The 47,212 HT recipients had a median (25th, 75th percentile) age of 56 (46, 62) years and median BMI of 26.8 (23.6, 30.5). Compared with BMI models, the maximally flexibly weight*height models had better performance for 1-year mortality (R²=0.024, C-statistic=0.630 vs. R²=0.025, C-statistic=0.632) and similar performance for 5-year mortality (R²=0.024, C-statistic=0.604 vs. R²=0.024, C-statistic=0.605). The Figure shows heat maps displaying the pattern of predicted 1- and 5-year event-free probability as well as discrepancies between the models and a histogram of prediction differences. The BMI model predicted lower 1-year mortality (13.8% vs. 15.2%) and similar 5-year mortality (30.1% vs. 30.5%) compared to the weight*height model. <h3>Conclusion</h3> Using weight and height as independent variables improved performance of models predicting mortality after HT compared with BMI. Additional research is needed to better the most optimal method for utilizing weight and height for improving the care of adults undergoing HT.