Information content of the multibreath nitrogen washout: effects of experimental error

Abstract

Methods that recover a continuous distribution of specific ventilation (ventilation-to-volume ratio, VA/V) from the multibreath N2 washout curve theoretically can resolve up to four modes of ventilation and reveal the major characteristics of the underlying distribution if experimental error is absent. This paper quantitatively assesses the effects of experimental error on resolution. The washout curves from five typical distributions were studied using linear programming and Monte Carlo methods. A measurement error of 0.1% was assumed in the mixed-expired N2 signal. Only 7 of the first 17 breaths contribute independent information about the underlying distribution, and only rough resolution of the underlying distribution is possible in the presence of error. Only two modes of ventilation plus an estimate of dead space can be confidently resolved. It is not possible to separate VA/V greater than 10 from dead space. A 10-fold reduction in experimental error will not greatly improve resolution. Experimental error, by reducing the linear independence of the defining kernels, significantly limits the information content and resolution of the multibreath N2 washout.