The effects of sinusoidal linear drifts on the estimation of cardiorespiratory dynamic parameters during sinusoidal workload forcing: a simulation study

This study aimed at investigating whether: 1) different sinusoidal linear drifts would affect the estimation of the dynamic parameters amplitude (A) and phase lag (φ) of minute ventilation (V̇E), oxygen uptake, carbon dioxide production and heart rate (HR) sinusoidal responses when the frequency analysis technique (F) is performed; 2) the Marquardt-Levenberg non-linear fitting technique (ML) would provide more precise estimations of A and φ of drifted sinusoidal responses compared to F. For each cardiorespiratory variable, fifteen responses to sinusoidal forcing of different sinusoidal periods were simulated by using a first-order dynamic linear model. A wide range of linear drifts were subsequently applied. A and φwere computed for all drifted and non-drifted responses by using both F (AF and φF) and ML (AML and φML). For non-drifted responses, no differences between AF vs AML and φF vs φML were found. Whereas AF and φF were affected by the sinusoidal linear drifts, AML and φML were not. Significant interaction effects (technique x drift) were found for A (P < 0.001; ƞP2 > 0.247) and φ (P < 0.001; ƞP2 > 0.851). Higher goodness of fit values were observed when using ML for drifted V̇E and HR responses only. The present findings suggest ML as a recommended technique to use when sinusoidal linear drifts occur during sinusoidal exercise, and provide new insights on how to analyse drifted cardiorespiratory sinusoidal responses.