S Sttaattiissttiiccaall aapppprrooaacchheess ttoo tthhee iiddeennttiiffiiccaattiioonn o off llaattee aasstthhmmaattiicc rreeaaccttiioonnss

Late asthmatic reactions can be difficult to recognize because of their prolonged time course and the confounding effects of superimposed circadian rhythms of ventilatory function. Conventional methods of analysis are rather arbitrary. They depend for example on a 15 or 20% fall in forced expiratory volume in one second (FEV1) from baseline or from time-matched control measurements. We have, therefore, investigated whether statistical approaches applied to individual subjects can assist in the identification of late asthmatic reactions. In two separate series of aerosol inhalation tests, three symptomatic workers, three asthmatic controls and three nonasthmatic controls were challenged blindly with increasing doses of two chemical agents, and saline. One of the agents, sodi- um isononanoyl oxybenzene sulphonate (SINOS) was a suspected cause of occupa- tional asthma. Prior to the challenges, FEV1 was measured hourly on three control days. Cumulative late changes on both control and active challenge days were quan- tified as the area between a line extrapolated from a 10.00 h baseline and the actu- al measurements from 12.00-22.00 h (the 2-12 h area decrement). The area decrement measurements on control and active challenge days were compared using Student's t-tests. The sensitivity of this method for detecting late asthmatic reactions among potentially positive tests (SINOS challenge tests in the workers) was examined, as was its specificity. The latter was determined from the false positive rate among the negative tests. A second statistical method based on the pooled standard devi- ation of serial (hourly) FEV1 measurements was investigated in the same way. In total, the data from 220 challenge and 30 control days were available for analysis. Late responses associated with falls in FEV1 of 8-16% were statistically signifi- cant when a t-test was used to compare area decrement on each active challenge day with three control days. This approach was, therefore, potentially more sen- sitive than conventional techniques for identifying late asthmatic reactions. The false positive rate was 4%. The serial FEV1 method was more sensitive, identify- ing a further five positive tests, but was less specific, with a false positive rate of 7%. These results suggest that when the day-to-day variability of lung function has been estimated from at least three control days, statistical tests can be applied to potential late asthmatic reactions, allowing them to be identified with greater pre- cision than conventional clinical techniques.