Abstract
B143
The potential public health benefits of individualized estimates of the probability of developing lung cancer are large. Using data from an ongoing case-control study, we recently developed separate epidemiologic risk assessment models for lung cancer in current, former, and never smokers. Exposures to second-hand smoke and to dust, family history of cancer, history of respiratory disease in the index patient, and smoking history were strong risk factors. These models had only modest discriminatory power and did not include data on diet or physical activity. We have therefore expanded the model to evaluate the added benefit of incorporating these variables. Lung cancer patients were recruited from the University of Texas M.D. Anderson Cancer Center, while healthy controls, who were matched on cases by age, gender, ethnicity and smoking status, were recruited at the Kelsey Seybold Clinics, the largest private physician group in the greater Houston area. Dietary data were collected on patients for the year prior to diagnosis and on controls for the year prior to interview using a 129-item modified National Cancer Institute -Block food frequency questionnaire, and responses were categorized according to the 2006 USDA guideline for pyramid food groups. Physical activity was reported for sports and other forms throughout adult years. We developed multiple variable models using backward elimination (P < 0.10) for diet and physical activity while retaining the original risk factors by smoking status. Among never smokers, the odds ratios (O.R.) for lung cancer were O.R. = 2.09 (95% CI 1.03, 4.21) and an O.R. = 2.15 (95% CI 1.17, 3.95) for participants who ate three servings or less than three servings of green salad per week, respectively, compared to those who consumed 4+ servings; and working in the garden once or twice per week, compared to those who did not report gardening, reduced risk by 40 to 46%. A similar model appeared for former smokers. Among current smokers, the O.R. of lung cancer was: 2.73 (95% CI 1.44,4.18) for consumption of three or fewer servings of carrots per week, compared to those who reported 4+ servings, and gardening once again reduced risk by 33 to 45%. Our new risk prediction models have a discriminatory power of 64%, 67%, and 71% respectively for never, former and current smokers, and were statistically significantly better than baseline models. This is the first lung cancer risk prediction model that examines both diet and physical activity, using the USDA pyramid serving guidelines and therefore provides easily applicable public health modalities for intervention strategies.
Sixth AACR International Conference on Frontiers in Cancer Prevention Research-- Dec 5-8, 2007; Philadelphia, PA