Exposure To Radon And Cancer Risk

Environmental Health Sciences

Investigators

Linked publications & trials

Abstract

Aims: We carried out a case-control study of lung cancer risk associated with residental exposure to radon. The study measured radon levels in all adult residences as well as in a significant childhood residences and attempted to take into account exposure variations within a home and over time. The study included men and women and both smokers and non-smokers in order to explore possible interactions between smoking and radon. Procedures and techniques: This population-based study was carried out in Utah/southern Idaho and Connecticut where living-level radon was estimated to be elevated in about 10% of homes, although this estimate proved to be high. Potential controls were identified by list-assisted telephone screening in Connecticut and by random telephone screening (under age 65) and from lists of Medicare recipients (age 65 and over) in Utah. We conducted in-person and telephone interviews to obtain residence histories and information on housing characteristics, lifestyle factors, occupational history, smoking by decade, exposure to smoke in each home and at work, lung diseases and demographics. With supplemental mail questionnaires we obtained limited diet histories and information on medical history, diagnostic and therapeutic radiation, and family medical and smoking history. An attempt was made to measure radon levels on up to three floors of all one-plus year residences between age 25 and 5 years prior to diagnosis, as well as the longest-duration childhood home. Current occupants of past homes provided data on housing characteristics and changes made to the home. Data were used to construct life-time exposure histories as well as average exposures for successive time windows. As might be imagined, missing data was a problem. Methods for dealing with missingness could affect study interpretation, motivating the exploration of the statistical properties of different approaches. Using exposure data for an index (lowest living level) level of measured control homes and other data on house characteristics and location as well as geographical data on soil type, atmospheric radiation levels, altitude, depth of the water table, altitude and population density, we used Computer Assisted Regression Tree (CART) methods to develop predictive models for radon levels in a home when a measurement was not available. To do this, it was necessary to obtain latitude/longitude coordinates (geocoding) for every address provided by study participants. We found that geologic variables, house type (single family or other), type of heat distribution (forced air or not), and the location of the detector relative to ground level were among the more important predictors of radon levels. Linear regression models relating measured radon values on pairs of floors within a home were then used to impute a value for unmeasured levels of a home. These level-specific values were used to calculate time-weighted exposure estimates for each residence which were then averaged over different time periods. We estimated the excess relative risk associated with average residential radon exposure during specific time-windows of interest. Models took into account sampling weights for randomized recruitment as well as factors such as education, residential mobility, hours spent at home, altitude, population density, decade-specific smoking data, state, and gender. We took advantage of the work done to geocode the addresses of all adult residences to estimate annual exposure to air pollutants based on monitoring data from the US EPA's AIRS database. We estimated average annual exposure to TSP and esposure during specific time-windows prior to lung cancer diagnosis or interview. Accomplishments: The study includes 1474 cases and 1811 controls. Only 129 of the lung cancer cases had never smoked, but most were nonsmokers at the time of cancer diagnosis. The average radon exposure levels were low; the average level in measured homes was 1.5 pCi/L in Utah and 0.9 pCi/L in Connecticut. Only 3% of Connecticut homes and 7% of Utah homes exceeded the US EPA's "action level" of 4 pCi/L. We calculated cumulative and average radon exposure estimates for all participants. The excess relative risk associated with radon was estimated after controlling for decade-specific smoking and other relevant variables and interactions. While our results appear to be consistent with those of other studies - including extrapolations based on miners, there is little evidence of any excess relative risk of lung cancer associated with radon in this study. Estimates were greater when data were restricted to participants who did not require proxy data, to those with complete coverage (i.e no imputed data) during the 20-year exposure window of interest, to nonsmokers, or to Connecticut, but no result is statistically significant. Our lung cancer study was one of the first designed specifically to explore the possible interaction between smoking and radon in lung cancer risk. As such, a new method for case and control selection was developed. This method has broader application and has been used in at least two studies where there was interest in balancing the recruitment of African Americans and whites without losing the ability to estimate the contribution of race to cancer risk. The radon levels found in our study are lower than what was expected at the start of the study but they are reflective of levels seen throughout much of the US. Thus, the absence of risk in our study can provide some reassurance to most of us, although the possibility that we lack the power to detect a small risk cannot be dismissed. Our study will provide ample data to anchor the low end of the dose-response curve in the pooled analysis and test the assumptions of linearity and no threshold. We found a small increase in lung cancer risk associated increasing average exposure to TSP. Risk estimates were greater in Connecticut, where TSP may be a better marker of carcinogenic agents in air pollution than in Utah. Lung cancer risk increased with increasing level of TSP. Risk appeared to be greater for exposures that occurred 15-20 years prior to diagnosis than for more recent exposures. Monitoring data are sparse more than 20 years prior to the time of diagnosis for most participants making it difficult to evaluate risks associated with exposures in the distant past.