Fourier analysis versus multiple linear regression to analyse pressure-flow data during artificial ventilation.

Respiratory resistance (Rrs) and elastance (Ers) are commonly measured in artificially-ventilated patients or animals by multiple linear regression of airway opening pressure (Pao) versus flow (V') and volume (V), according to the first order model: Pao = P0 + Ers.V + Rrs.V', where P0 is the static recoil pressure at end-expiration. An alternative way to obtain Rrs and Ers is to derive them from the Fourier coefficients of Pao and V' at the breathing frequency. A potential advantage of the second approach over the first is that it should be insensitive to a zero offset on V' and to the corresponding volume drift. The two methods were assessed comparatively in six tracheotomized, paralysed and artificially ventilated rabbits with and without adding to V' an offset equal to 5% of the mean unsigned flow. The 5% flow offset did not modify the results of Fourier analysis, but increased Rrs and Ers from linear regression by 15.8 +/- 4.6% and 4.55 +/- 0.64%, respectively. Without additional offset, differences between the two methods averaged 30.2 +/- 14.0% for Rrs and 9.3 +/- 6.2% for Ers. The differences almost completely disappeared (2.47 and 0.61%, respectively) when the flow signal was zero-corrected using the assumption that inspired and expired volumes were the same. After induced bronchoconstriction, however, Ers was still slightly larger by linear regression than by Fourier analysis, which may result from nonlinearities and/or frequency dependence of the parameters. We conclude that the regression method requires zero flow correction and that Fourier analysis is an attractive alternative.