Reference equations using segmented regressions for impulse oscillometry in healthy subjects aged 2.7–90 years

Abstract

Background Published reference equations for impulse oscillometry (IOS) usually encompass a specific age group but not the entire lifespan. This may lead to discordant predicted values when two or more non-coincident equations can be applied to the same person, or when a person moves from one equation to the next non-convergent equation as he or she gets older. Thus, our aim was to provide a single reference equation for each IOS variable that could be applied from infancy to old age. Methods This was an ambispective cross-sectional study in healthy nonsmokers, most of whom lived in Mexico City, who underwent IOS according to international standards. A multivariate piecewise linear regression, also known as segmented regression, was used to obtain reference equations for each IOS variable. Results In a population of 830 subjects (54.0% female) aged 2.7 to 90 years (54.8% children ≤12 years), segmented regression estimated two breakpoints for age in almost all IOS variables, except for R5−R20 in which only one breakpoint was detected. With this approach, multivariate regressions including sex, age, height and body mass index as independent variables were constructed, and coefficients for calculating predicted value, lower and upper limits of normal, percentage of predicted and z-score were obtained. Conclusions Our study provides IOS reference equations that include the major determinants of lung function, i.e. sex, age, height and body mass index, that can be easily implemented for subjects of almost any age.