Comparative Analysis of V̇O2 Prediction Equations Using a Novel Web-Based Application: An Illustrative Example in Formerly Deployed Military Veterans

Thomas Alexander^1*Michael Falvo¹Daniel P. Wilhite¹John J. Osterholzer^2,3Bradley W. Richmond^4,5Steven Cassady⁶Daniel J. Schneider^2,3Silpa Krefft⁷Danielle Glick⁶Anays M. Sotolongo¹Stella Hines⁶Mehrdad Arjomandi⁸

• ¹Airborne Hazards and Burn Pits Center of Excellence, VA New Jersey Health Care System, East Orange, United States
• ²Research Service and Pulmonary Section Medical Service, VA Ann Arbor Healthcare System, Ann Arbor, United States
• ³Division of Pulmonary and Critical Care Medicine, Department of Internal Medicine, University of Michigan, Ann Arbor, United States
• ⁴Medical Service, VA Tennessee Valley Healthcare System, Nashville, United States
• ⁵Department of Medicine, Division of Allergy, Pulmonary, and Critical Care Medicine,, Vanderbilt University Medical Center, Nashville, United States
• ⁶Department of Medicine, VA Maryland Health Care System, Baltimore, United States
• ⁷National Jewish Health; Medical Service,, VA Eastern Colorado Health Care System, Aurora, United States
• ⁸Department of Medicine, Medical Service, San Francisco VA Health Care System, San Francisco, United States

Background: Cardiopulmonary exercise tests (CPET) use clinical-algorithms for interpretation by classifying exercise capacity based on a fixed threshold (e.g., oxygen consumption percent-predicted ≥80% [V̇O2peak-pp]). Impact of prediction equation selection on V̇O2peak-pp values and subsequent classifications have not been thoroughly examined in Veterans undergoing specialty evaluation for post-deployment concerns. We developed an application (https://tom26alex-cpx-comparison.share.connect.posit.cloud/) offering a direct comparison of multiple prediction equations for V̇O2 with data visualizations to better contextualize the individuals achieved V̇O2peak. Methods: We retrospectively reviewed CPET records from U.S. veterans undergoing evaluation for post-deployment concerns and calculated V̇O2peak-pp using six separate commonly used prediction equations. Exercise capacity was classified as normal using a fixed threshold (V̇O2peak-pp≥80%). Friedman’s test was employed for overall comparison of peak predicted across equations, followed by Cohen’s kappa (κ) to evaluate agreement in exercise capacity classification. The influence of demographic and anthropometric factors on inter-equation differences was examined using regression analysis. Results: Significant variability was noted in V̇O2peak-pp between prediction equations (Friedman’s χ2 =936.0, P<0.01, Kendall effect size=0.6). In pairwise analysis, 53% of Veterans in the study were re-classified at least once resulting in significant discordance between all pairs of equations (κ=0.24 to 0.78). Regression analysis identified body mass index (BMI) as the most significant contributor to differences in V̇O2peak-pp. Given these results the app created focuses on the effects of BMI on equations by providing a visual aid to interpret the effect of BMI changes on the predicted V̇O2peak. Interpretation: Classification of exercise capacity varies considerably as a function of prediction equations, and this variation appears most influenced by anthropometric factors. Clinicians should be aware of this variability and consider alternatives to relying on a single prediction equation approach, such as, utilizing the developed app to visualize and calculate a range of V̇O2peak-pp values derived from multiple equations.