Comparison of PD20 with two alternative measures of response to histamine challenge in epidemiological studies

Abstract

Bronchial responsiveness to histamine or methacholine provides a useful objective measure for epidemiological studies of airways disease, but most people in a community population do not have a 20% fall in forced expiratory volume in one second (FEV1) with the highest dose administered. Histamine challenge data were analysed to compare the repeatability. Normality and separation of symptom groups of the early dose-response slope with provocative dose producing a 20% fall in FEV1 (PD20). Tests were continued until a 20% fall in FEV1 occurred, or 4 μmol had been given. Data were available for 510 randomly selected subjects, and for an additional 283 with wheeze. A repeat test was obtained in 104 individuals. PD20 was estimated by curve fitting, with extrapolation to 8 μmol. Least-squares slope of percentage decline in FEV1 on histamine dose was calculated, using all the measured points and two-point slope as the fall from the post-saline measurement to the maximum cumulative dose divided by the maximum dose. Log transformation of PD20 and shifted reciprocal transformations of slope produced constant variance. Over all subjects the three measures had similar repeatability; in subjects with PD20 >8 μmol the intraclass correlation for two-point slope was only 0.26, but was 0.66 for least-squares slope. Neither measure of slope was normally distributed, but the distribution of log(PD20) was consistent with a censored normal distribution. In conclusion, least-squares slope is preferable to two-point slope for epidemiological studies. Either PD20 or least-squares slope can be analysed; the former requires methods for censored data, but the latter requires transformation with an arbitrary constant, +10% μmol-1 being recommended, and methods requiring normality should be used with caution.