Using estimated probability of pre-diagnosis behavior as a predictor of cancer survival time: an example in esophageal cancer

Abstract

Background: Information on the associations between pre-diagnosis health behavior and post-diagnosis survival time in esophageal cancer could assist in planning health services but can be difficult to obtain using established study designs. We postulated that, with a large data set, using estimated probability for a behavior as a predictor of survival times could provide useful insight as to the impact of actual behavior. Methods: Data from a national health survey and logistic regression were used to calculate the probability of selected health behaviors from participant's demographic characteristics for each esophageal cancer case within a large cancer registry data base. The associations between survival time and the probability of the health behaviors were investigated using Cox regression. Results: Observed associations include: a 0.1 increase in the probability of smoking 1 year prior to diagnosis was detrimental to survival (Hazard Ratio (HR) 1.21, 95% CI 1.19,1.23); a 0.1 increase in the probability of hazardous alcohol consumption 10 years prior to diagnosis was associated with decreased survival in squamous cell cancer (HR 1.29, 95% CI 1.07, 1.56) but not adenocarcinoma (HR 1.08, 95% CI 0.94,1.25); a 0.1 increase in the probability of physical activity outside the workplace is protective (HR 0.83, 95% CI 0.81,0.84). Conclusions: We conclude that probability for health behavior estimated from demographic characteristics can provide an initial assessment of the association between pre-diagnosis health behavior and post-diagnosis health outcomes, allowing some sharing of information across otherwise unrelated data collections.

Imputing pre-diagnosis health behaviour in cancer registry data and investigating its relationship with oesophageal cancer survival time

“…In previous publications we have described two different approaches to imputing SEER cancer cases’ behaviour from BRFSS health survey data via common demographic variables. One approach was to develop a logistic regression model predicting behaviour from demographic variables in the BRFSS health survey data and then apply that model to each SEER cancer case to estimate their probability of having the behaviour [ 13 ]. The second approach was to stratify both data sets into age by sex by race by marital status by State of residence by year subgroups, and then within each strata randomly assign one BRFSS health survey data record to donate their behaviour to each SEER cancer registry data record [ 9 ].…”