Box-Jenkins transfer function modelling for accurate determination of VO2 kinetics in COPD patients

Abstract

VO2 kinetics provide information about the ability to respond to the increased energetic demand during a constant work rate test (CWRT). Typically, kinetic features that describe the phase II VO2 response (see figure), such as time constant (TC) and mean response time (MRT), are estimated using exponential models. Box-Jenkins transfer function (BJ-TF) models can extract the same features without being affected by unwanted noise contributions (e.g. phase I or measurement noise). This is realised by separately modelling phase II (system model) and unwanted contributions (noise model). We compared model fits and estimated kinetic feature values between exponential models and BJ-TF models. Both models were used to extract kinetic features from CWRT data of 18 COPD patients (14 male; age: 63 ± 7 years; FEV1: 48 ± 14 %predicted). The BJ-TF full models (illustrated in the figure for one patient, black dashed line) could accurately model the total VO2 response for all patients. Furthermore, the BJ-TF system models (orange full line), in contrast to the exponential models (grey dotted line), were able to surpass the unwanted phase I contribution, leading to more accurate TC (54 ± 23s vs 80 ± 59s; p