Although we appreciate the interest of Dr Armstrong and Dr Strachan for our paper on the pool chlorine/asthma risk, we cannot really take on board their reasoning concerning the statistical analyses. When questioning the strength of the associations found in our studies, they seem indeed to attribute much importance to the p values of the associations emerging between cumulated pool attendance and indicators of asthma or lung epithelium permeability. The p values, however, are not reliable indicators to judge of the strength of associations found in epidemiology since they are highly dependent on the number of observations. Our assessment of the strength of these associations was therefore based more on the values of r² and on the fact that the associations found with pool chlorine exposure were much stronger than those emerging (and a fortiori not emerging) with other variables classically presented as possible contributors to asthma and lung damage in children (for example, environmental tobacco smoke, pets, outdoor pollution).

With regard to the third study linking asthma prevalence and pool attendance, we agree of course that this is a retrospective ecological study carried out by aggregating data from each school, which was made possible due to the fact that pool attendance is a compulsory activity in Belgian primary schools. However, since this study was not specifically designed to assess the effects of pool chlorine, in our opinion, its major weakness lies less in this school based aggregation than in the fact that we could not quantify the cumulated pool chlorine exposure of these children on an individual basis, some of them having certainly attended a chlorinated pool with their parents (recreational, baby swimming, etc) or as part of a sport activity. This is the reason why we cautiously concluded our paper by recommending further studies to test this chlorine hypothesis. We have now just completed such a study exploring the links between asthma, lung inflammation, atopy, and cumulated exposure of children to pool chlorine. The results clearly show that the associations published in OEM were far from having been overestimated (Bernard et al, manuscript in preparation).