About 80% of all cancers are diagnosed in elderly and more than 80% of known risk factors of cancer are considered to be potentially preventable. We present an approach to estimating the contributions of a spectrum of various measurable risk factors including behavioral/lifestyle risk factors to cancer risk in the U.S. elderly population.

The three sources of data used are 1) nationally representative National Long Term Care Survey (NLTCS) - for measuring the functional status and behavioral factors in the elderly, 2) Medicare Claim files linked to each person from the NLTCS - for cancer incidence and mortality in the U.S. population, and 3) SEER data used as a “gold standard” for comparison of SEER age patterns of selected cancers with age patterns predicted through using NLTCS-Medicare data. The 1994 and 1999 NLTCS were analyzed and, in total, more than 200 variables for each survey were selected and grouped in fourteen groups, including daily living activities, physical activities, smoking, alcohol consumption, social activities, self-reported comorbidity, health insurance, medical providers, and others. Several approaches were tested to elaborate the measures in each group and aggregated scores: 1) stepwise technique allowing to select variables in these groups which are most predictive for selected cancer incidences, 2) constructing aggregating scores by combining information from all variables in each group, 3) factor analysis methods to extract one factor for each group. Elaborated characteristics were tested for association with selected cancer incidences during the following 5 years since the date of interview. Breast, prostate, lung, and colon cancers were selected for analyses due to their high incidence rates, and because their incidence rates can be reconstructed relatively well from Medicare data.

As a result of this analysis we obtained a description of higher cancer risk groups in terms of variables measured in NLTCS and aggregated indices constructed from these variables. Groups of parameters of physical activity, tobacco consumption, comorbid conditions, demographic characteristics, and health insurance and medical care providers showed significant contributions in increasing or decreasing the risk of incidence of breast, lung, colon, and prostate cancers in the elderly. Contribution of groups of factors, and of individual factors inside the group varied depending on cancer site, e.g., smoking and pulmonary comorbidity were strongly associated with lung cancer risk, physical activity was associated with breast cancer risk. The performed sensitivity analyses included the impact of scheme for filling missing data and the comparison of the results obtained from different algorithms of cancer onset reconstruction.

Behavioral risk factors analyzed both individually and grouped in related clusters, significantly affect cancer risks in the U.S. elderly population. The most influential of the potentially preventable risk factors can be detected using this approach and applied to further deeper analyses, including other data sets with detailed risk factors description. Using information about the duration and intensity of influential risk factors, more individualized forecast of cancer risk can be made and more individualized prevention strategies can be applied.

Citation Information: Cancer Prev Res 2008;1(7 Suppl):B119.

Seventh AACR International Conference on Frontiers in Cancer Prevention Research-- Nov 16-19, 2008; Washington, DC