Objective:
Radon exposure is a proven cause of
lung cancer and is a possible cause of other diseases. Recently, several ecologic studies explored the correlation of county-wide incidence rates for non-
lung cancers with residential
radon levels, using
radon data reported by a commercial laboratory. However, the validity of the commercial
radon data, i.e., whether they are an accurate representation of the
radon levels in the counties from which they were drawn, is unknown. Methods: We compared county-wide
radon data from the commercial laboratory with corresponding measurements from the same counties reported previously by the Environmental Protection Agency (EPA). Matching data were available for four states, Iowa, North Dakota, Texas, and Wisconsin, and were compared by paired t-tests. Criterion validity of the commercial tests, i.e., how well the commercial data predicted the EPA data, was tested using non-parametric methods, Kendall’s tau, Lin’s concordance, and Passing−Bablok regression. Results: The commercial and EPA data pairs from the four states were significantly positively correlated, although the size of the correlations was modest (tau = 0.490, Lin = 0.600). Passing−Bablok regression indicated that the commercial
radon values were significantly higher than their EPA pairs and significantly overestimated
radon at low levels (<4 pCi/L, p < 0.001). Conclusions: The commercial laboratory data were moderately predictive of EPA
radon levels at the county level but were significantly biased upwards at low levels. The disagreement likely has several causes, including selection bias from homes that were tested voluntarily. Ecologic studies that employ
radon data obtained from commercial laboratories should be interpreted with caution.