Determinants of Reclassification of Exercise Performance across Cardiopulmonary Exercise Testing Prediction Equations.

Abstract

Purpose: Volume of oxygen uptake at peak exercise ( ) is a key indicator of cardiorespiratory fitness used to assess surgical risk and clinical trial outcomes. However, multiple prediction equations exist for its interpretation, often yielding inconsistent classifications of exercise performance. We aimed to determine how different prediction equations affect exercise performance classification and identify factors contributing to its variability.

Methods: We performed a post hoc analysis of cardiopulmonary exercise testing (CPET) data collected from 332 nonsmoking flight attendants who participated in our previous studies examining the health effects of occupational exposure to secondhand tobacco smoke. Percent-predicted was calculated using seven prediction equations (Fitness Registry and the Importance of Exercise National Database, Wasserman, Hansen-Wasserman, Study of Health in Pomerania, Jones, Bruce, and Cooper-Storer). Reduced exercise performance was defined as < 80 percent-predicted. Agreement across equations was assessed using Cohen's κ. Associations with dyspnea (modified Medical Research Council ≥ 1) were examined using C-statistics to assess the discriminative performance of prediction equations for this patient-centered outcome. The magnitude of change in percent-predicted when switching from one equation to another ("recalculation magnitude") was quantified for individual participants, and multivariable regression modeling was used to identify demographic and anthropometric determinants of recalculation magnitude.

Results: Participants were 56 ± 12 years old, 86% female, with body mass index = 24 ± 4 kg/m 2 . Depending on the prediction equation, mean percent-predicted varied widely (83%-113%), and the prevalence of reduced exercise performance ranged from 7% to 51%. Agreement across equations ranged from slight to substantial (κ = 0.10-0.79). At an individual level, switching between equations recalculated percent-predicted by -93% to +84%, leading to bidirectional exercise performance reclassification of 6.6%-45.5% of participants. Body mass index was the strongest determinant of recalculation magnitude, followed by age and sex. The discrimination accuracy of prediction equations for dyspnea was modest (area under the curve = 0.56-0.71), with the Study of Health in Pomerania equation showing the best performance.

Conclusions: Different prediction equations yielded materially different percent-predicted values and frequently changed exercise capacity classification in a bidirectional manner. The observed discordance appears to arise in substantial part from limitations in how demographic and anthropometric factors, particularly body size, are modeled across equations, underscoring the need for improved reference standards that better capture physiological variability.