Subjects
The VITAL study is a prospective cohort of community-dwelling adults [6]. Women and men were eligible if they were aged 50 to 76 and lived in the area covered by the Seattle-Puget Sound Surveillance, Epidemiology, and End Results (SEER) cancer registry. Using a commercial list, we mailed 364,418 questionnaires from October 2000 to December 2002. 77,719 (21.3%) participants returned questionnaires and passed eligibility and quality control checks. Baseline data were obtained from a sex-specific 24-page self-administered questionnaire that included items on medication use, diet, medical history, personal characteristics, and cancer risk factors. The Institutional Review Board of the Fred Hutchinson Cancer Research Center approved the protocol.
Participants were followed for lung cancer occurring from baseline through December 31, 2007, by linking the cohort to the Seattle-Puget Sound SEER registry. Cases and information on tumor characteristics, including histology and stage at diagnosis, are ascertained through all hospitals in the area, offices of pathologists, oncologists, and radiotherapists, and from state death certificates. The SEER registry has been shown to have accurate and complete data collection [7] and is reliable for lung cancer histology [8]. If a subject had multiple diagnoses of lung cancer, we used the stage of the first primary diagnosis.
We excluded participants with a previous diagnosis of lung cancer reported at baseline (n = 376) or for whom this datum was missing (n = 211). We also excluded subjects whose lung cancer was identified on a death certificate only or whose lung cancer morphology was classified as lymphoma (n = 10). We elected to only include cases of NSCLC since growth rates, staging, and symptomatology markedly differ from small cell lung cancer. After these exclusions, 612 participants developed NSCLC within a mean followup time of 5.9 years (SD 1.2 years).
Outcome Assessment: Stage at Diagnosis
Based on SEER data, we dichotomized stage at diagnosis into in-situ and local (early stage) versus regional, distant, or unknown stage (advanced stage). Unknown stage was combined with the latter group because it has a comparable survival rate [1].
Covariates
Sociodemographic & Health History
Subjects reported demographic and socioeconomic factors that included age, race/ethnicity, marital status, and education. Self-report of physician-diagnosed chronic obstructive pulmonary disease (COPD), including emphysema and chronic bronchitis, and previous history of cancer were recorded. We categorized family history of lung cancer as none or at least one first degree relative with lung cancer. A comorbidity scale was created based on self-report of the following conditions, categorized as yes or no for each response: coronary artery disease, heart failure, stroke, chronic pulmonary disease, rheumatoid arthritis, cirrhosis or other chronic liver disease, kidney disease (other than kidney stones), diabetes, and history of cancer other than non-melanotic skin cancer.
Subjects were asked detailed questions about their exercise habits. Exercise is calculated as usual metabolic equivalent of task (MET) hours per week for each activity averaged over the previous 10 years as follows: [Frequency of activity per week * minutes per session * years in the past 10 * MET for that activity] / [(60 minutes/hour) * 10 years] [9]. We then summed the MET hours for all activities to calculate total 10-year average MET hours per week. We categorized exercise into quartiles. Body mass index (BMI) was calculated from the respondent's self-reported current weight and height, measured as kg/m2, and categorized. Daily servings of fruit were assessed by a food frequency questionnaire (FFQ) that was an adaptation of FFQ's developed for the Women's Health Initiative and other studies [10–12], with the addition of highly supplemented foods. The measurement properties of earlier versions of this questionnaire have been published [10].
Tobacco
Smokers were defined as individuals who smoked at least one cigarette per day for at least a year. Smoking status was classified as never, current, quit 10 years or more or quit less than 10 years ago, at the date of questionnaire completion. Duration of smoking was estimated by the reported number of years smoked, intensity by the usual number of cigarettes smoked per day, and pack-years was computed as years smoked × cigarettes per day/20.
Screening Activities
All subjects were asked if they had a sigmoidoscopy or colonoscopy in the ten years prior to baseline. Men reported if they had a prostate specific antigen (PSA) test in the two years prior to baseline. Women reported if they had a mammogram in the two years prior to baseline.
Statistical Analysis
All statistical analyses were performed using Stata SE-11 (StataCorp, College Station, TX). For the univariate analyses, the association between each factor and early versus advanced stage at diagnosis was measured through logistic regression using robust standard errors. To evaluate our primary hypothesis that screening activities would be associated with stage at diagnosis, we used multivariable adjusted logistic regression using robust standard errors. A priori, we decided to adjust for age at baseline (continuous), gender, smoking status (never, current, quit < 10 years ago, quit ≥ 10 years ago), income (classified as greater or less than $40,000/year and missing), education (dichotomized as greater or less than college graduate), race/ethnicity (dichotomized as white versus other), comorbidity (modeled continuously), and self-reported COPD (dichotomized as yes versus no). Overall, there was less than 5% missing information for all variables except income (21% missing this information) and BMI (6% missing this information). P values less than 0.05 were considered statistically significant.
We evaluated whether other factors individually confounded the association of screening activities with NSCLC stage at diagnosis in the adjusted model. These included: years of smoking, packyears, history of cancer, exercise, BMI, marital status, and servings of fruit. None changed the point estimates of the screening activity variables by more than 10% or the 5% level of significance so they were not included in the final model. We performed sensitivity analyses by not including in-situ and unknown stages in the outcome. As not including these subjects in the screening analyses did not substantively change the OR's, we decided to include all stages in the final model.
We examined whether the association between sigmoidoscopy/colonoscopy receipt and NSCLC stage at diagnosis differed by age, sex, and smoking status. These models were adjusted as above with the exception of not including age, sex, or smoking status, respectively, in the stratified models. Since there were few never smokers who developed lung cancer, we did not include this group in the stratified smoking status analyses. Likelihood ratio tests were conducted to assess the interaction between sigmoidoscopy/colonoscopy and the subgroups. P values for interaction were obtained to compare the fit of the models with the interaction terms and without them.
Role of the Funding Source
This work was completed with grant support from the CHEST Foundation of the American College of Chest Physicians and the LUNGevity Foundation to C.S. This work was also supported by the National Institutes of Health [K05CA154337 to EW]. Drs. Slatore, Au, and Deffebach were supported by the Department of Veterans Affairs. This study is the result of work supported by resources from the Portland VA Medical Center, Portland, OR, and VA Puget Sound Health Care System, Seattle, Washington. The study sponsors had no role in the conduct of the study, in the collection, management, analysis, or interpretation of data, or in the preparation, review, or approval of the manuscript.