Low-Frequency Respiratory Mechanics Using Ventilator-Driven Forced Oscillations

We evaluated the potential for using a fast Fourier transform (FFT) analysis applied to a standard ventilator waveform to estimate (< 2 Hz) frequency dependence of respiratory or lung resistance (R) and elastance (E). In four healthy humans we measured pressure and flow at the airway opening while applying sine wave forcing from 0.2 to 0.6 Hz at two tidal volumes (VT; 250 and 500 ml). We then applied a step inspiratory ventilator flow wave with relaxed expiration at the same VT and only 0.2 Hz. Step waveform data were also acquired from nine mechanically ventilated patients under intensive care unit conditions. Finally, we simultaneously measured total respiratory (rs), lung (L), and chest wall (cw) impedance data from two dogs (0.156-2 Hz) before and after severe pulmonary edema. Rrs and Ers were estimated by the FFT approach. Humans displayed a small frequency dependence in Rrs and Ers from 0.2 to 0.6 Hz, and both Rrs and Ers decreased at the higher VT. The spectral estimates of Rrs and Ers with the step ventilator wave were often qualitatively comparable to sine wave results below 0.6 Hz but became extremely erratic above the third harmonic. Conversely, in dogs the step wave produced reliable and stable estimates up to 2 Hz in all conditions. Nevertheless, Ecw and Ers still displayed clear and correlated oscillations with increasing frequency, whereas EL showed none. This suggests that nonlinear processes, most likely at the chest wall, contribute to periodic-like fluctuations in respiratory mechanical properties when estimated by applying FFT to a step ventilator wave. Moreover, in humans, but not dogs, a ventilator flow cycle contains insufficient signal energy beyond the third harmonic. We show that the amount of energy available at higher frequencies is largely governed by the mechanical time constant contributing to passive expiratory flow. In dogs the shorter time constant contributes to increased energy. In essence, the frequency content of the flow is subject dependent, and this is not a desirable situation for controlling the quality of the impedance spectra available from a standard ventilator wave.