Indoor Radon and Lung Cancer

Abstract

The authors respond: The design of the French case-control study on lung cancer risk associated with indoor radon exposure1 was conducted within the framework of a common European program focusing on lung cancer risk linked to radon in dwellings. The same protocol and questionnaire were applied in other countries (and the same approach to data collection was used in Germany and Great Britain). Intercomparison of measurement techniques has been performed, and major collaboration also was realized in the field of statistical analysis (same logistic regression approach in Germany and France, with the same definition of time-weighted variables and discussion of log-linear and linear excess relative risk models). Our study was consistent with international recommendations and satisfied specific criteria to be part of a pooled analysis regrouping major European studies; this joint analysis was recently published by Darby et al.2 Considering the specific criticisms of Lachet,3 we agree with the fact that, in Table 3, the last column does not correspond to the category of exposure >400 Bq/m3 but presents the results of the model considering radon exposure as a continuous variable. A clear separation between the last 2 columns would have made this more evident. Other criticisms derive from a misunderstanding of the nature of the data and of the statistical approach. As stated in the Methods section of our article,1 the analysis was based on logistic regression. This approach is classically used in epidemiology for the analysis of case-control design studies,4 and was used in previous analyses of indoor radon case-control studies.5,6 This regression method relies on a log-linear model and, therefore, it is not surprising that the estimated exposure-risk relationship presented in Figure 1 is not a linear one. As also stated in the Methods section, 2 approaches were performed to analyze the relationship between radon and lung cancer risk: the first one considers radon as a categorical variable with the <50 Bq/m3 category being the reference (RR is fixed to 1.0 in this category) and the second one considers radon as a continuous variable on the whole range of exposure. The second approach does not depend on a reference category. It is therefore not surprising that the fitted model represented in Figure 1 starts from the origin. Finally, the reanalysis presented by Lachet in the last paragraph of his letter does not take into account the distribution of cases and controls in the different exposure categories. His approach reflects a misinterpretation of the nature of the data and is not scientifically justified. When Lachet is criticizing the results of our study, he seems to ignore the fact that other groups of researchers with independent data have come to the same conclusion as we have. The European study, with very large numbers of cases and controls, has been recently published in the British Medical Journal.2 There is no heterogeneity between the studies having contributed to this joint analysis. This analysis showed strong evidence of an association between residential radon and lung cancer; the risk of lung cancer increased by 8.4% per 100 Bq/m3 increase of radon exposure. The classic logistic regression approach we used1 gave results similar to those obtained with the French data by Darby et al.2 when they applied a linear excess relative risk model. Collectively, all these data allowed direct assessment of the risk of lung cancer from indoor radon exposure. The appropriate debate should move now from the question of the existence of a risk to the reduction or elimination of this risk. WHO is presently evaluating the risk linked to indoor radon, and our hope is that this international organization will help to implement the appropriate risk management and communication to the general public for this well-documented public health problem. Hélène Baysson Margot Tirmarche Klervi Leuraud Dominique Laurier Institute for Radiological Protection and Nuclear Safety (IRSN) Fontenay aux Roses, France