Coloured inhalers.

Suissa and colleagues raise three issues.Firstly, they suggest that the relative risk of 2-0 for fenoterol explains only a small proportion of the epidemic.Our study was based on prescribed drug treatment rather than that actually taken'; this non-differential information bias would tend to decrease the observed odds ratios towards 1.0.2Secondly, they suggest that the epidemic could have been due to overuse of all inhaled B agonists.A similar error was recently made by Staudinger and Haas.'In fact, sales of all B agonists in New Zealand increased considerably only after 1979, three years after the onset of the epidemic,4 in response to suggestions that B agonists should be used regularly rather than on demand.The large relative risks for all B agonists in the Saskatchewan study emerged only in the questionable multi- variate analysis,5 and probably tell us more about overuse of multivariate analysis than overuse of B agonists.