Agreement of four metabolic equations for estimating submaximal cycling oxygen uptake with two portable indirect calorimetry systems in untrained young women

Abstract Background Metabolic prediction equations provide a low-cost means of estimating submaximal cycling oxygen uptake (VO₂), yet equations derived from maximal exercise are often applied to steady-state fixed-load protocols without validation. Portable analysers including the COSMED K5 and CORTEX MetaMax 3B permit field-based VO₂ measurement, but neither is error-free. We assessed agreement between four prediction equations and both devices in untrained young women, partitioning total error into bias-related and residual components. Methods Nineteen untrained young women completed four constant-power cycling stages (30, 40, 50, 60 W). VO₂ was estimated from four equations (one submaximal cycling formula, three maximal-test equations) and measured concurrently with two portable indirect calorimeters. Linear mixed-effects models accounted for repeated within-participant observations. Agreement was evaluated using Lin’s concordance correlation coefficient (CCC), mean absolute percentage error (MAPE) and Bland–Altman analysis with proportional bias testing; 95% confidence intervals came from participant-level cluster bootstrap, with false discovery rate correction. Results Every equation–device combination showed significant negative bias (all P < 0.01), and underprediction worsened at higher intensities. The ACSM submaximal equation with CORTEX MetaMax 3B showed the closest agreement (CCC = 0.69, MAPE = 12.9%, bias = − 1.71 mL·kg⁻¹·min⁻¹), with residual error accounting for 68.5% of total disagreement. The three maximal-test equations paired with COSMED K5 yielded the largest errors, 83–91% of which was bias-related. Even the best combination showed growing underprediction as power increased. Conclusions The ACSM submaximal cycling equation showed the closest relative agreement among the eight tested pairings. Maximal-test equations agreed poorly with device-measured values during low-to-moderate fixed-load cycling. Without a laboratory gold standard, these findings reflect relative cross-device agreement rather than absolute accuracy, and disagreement cannot be attributed solely to equation error. The intensity-dependent bias suggests that a universal correction factor would be inadequate under the tested conditions. Any correction approach would require population- and device-specific external validation before clinical use.