CONFOUNDING BY SEVERITY DOES NOT EXPLAIN THE ASSOCIATION BETWEEN FENOTEROL AND ASTHMA DEATH

Three recent case-control studies from New Zealand, and one from Saska tchewan, Canada, have found that fenoterol increases the risk of death in patients with severe asthma. It has been suggested that these find ings may be due to confounding by severity, if fenoterol was selective ly prescribed to more severe asthmatics. This 'confounding by severity ' hypothesis has now been investigated in further analyses of data fro m the New Zealand case-control studies. This analysis found that among patients whose asthma was severe enough to require hospital admission (the population in whom the case-control studies were conducted), fen oterol was not preferentially prescribed to the more severe asthmatics . There was greater co-prescribing of other drugs with fenoterol (comp ared with salbutamol) during the later years of the epidemic, but thes e differences did not explain the excess risk associated with fenotero l, and there was little evidence of greater co-prescribing during the earlier years of the New Zealand epidemic of asthma deaths. There was no association between the prescription of fenoterol and markers of ac ute asthma severity or psychosocial problems. Patients were not select ively changed to fenoterol as a result of a severe attack resulting in a hospital admission. Most importantly, in the case-control studies o f asthma deaths, the inhaled fenoterol relative risk increased when th e analysis was restricted to sub-groups defined by markers of chronic asthma severity; whereas the relative risk would have decreased toward s 1.0 in these sub-group analyses if the overall elevated risk for fen oterol was due to confounding by severity. We conclude that in patient s whose asthma is severe enough to require hospital admission, there i s little evidence that fenoterol was selectively prescribed to the mor e severe patients and that the findings in the most severe sub-groups effectively exclude the 'confounding by severity' hypothesis as an exp lanation for the recent case-control findings.