Selection of Cases and ControlsLung cancer deaths were identified using municipal registration records and death certificates. 14, 15 We followed subjects until late 1999, and eligible case subjects were defined as those who died from lung cancer (International Statistical Classification of Diseases and Related Health Problems 10th Revision: C34). During about 10-year follow-up, 240 deaths from lung cancer were identified among the subjects who provided baseline serum samples. Of these, we excluded 8 subjects with a previous history of cancer, and excluded 21 who lacked suitable samples. For each case of lung cancer death (hereafter referred to as "cases"), 2 or 3 controls were selected from the survivors, matching for sex, age (as near as possible), and participating institution. Of the initial 569 matched controls, we excluded 6 with a previous history of cancer, 30 who lacked sufficient samples, and 46 whose matched cases were excluded. Thus, 211 cases and 487 controls were included in the analysis.Determination of Serum Carotenoids, Retinol, Tocopherols, and Folic Acid All samples were analyzed by trained staff blinded to case-control status. Serum total cholesterol (TC) was determined using an autoanalyzer. Serum concentrations of carotenoids, RE, and tocopherols were measured by high-performance liquid chromatography, as described elsewhere.17 All samples were stored in deep freezers at -80 for about 11 years, and serum levels of carotenoids, RE and tocopherols were measured using the same equipment for all specimen; the ranges of repeatability and dayto-day variation were 4.6% to 6.9% and 6.3% to 20.0%, respectively, for the assays of carotenoids, RE, and tocopherols. 17 We could not separately measure serum levels of zeaxanthin and lutein or -and -tocopherols, and therefore report the combined levels as ZL or -and -tocopherols (BT), respectively. We calculated total carotenes (TCR) as the sum of AC, BC, and LY, calculated total xanthophylls (TXP) as the sum of CR, ZL, and canthaxanthin (CX), and calculated total provitamin A (PVA) as the sum of AC, BC, and CR. Total carotenoids (TCA) were calculated as TCR plus TXP. Serum folic acid levels were determined using a cloned enzyme donor immunoassay.
18
Statistical AnalysisBody mass index (BMI) at baseline was calculated from reported height and weight: BMI = (weight in kg)/(height in m) 2 . Mean differences between lung cancer cases and matched controls were examined by t test after converting serum levels of carotenoids, RE, tocopherols, folic acid and TC to logarithmic values. Analysis of covariance (ANCOVA) was also performed after controlling for age, participating institutions, smoking, alcohol consumption, BMI and serum TC level.19 Two conditional logistic regression models were used to calculate odds ratio (OR) for lung cancer death. 19 Variables adjusted in the models were as follows: model 1 (OR1), age, participating institution, and smoking; model 2 (OR2), age, participating institution, smoking, alcohol con-Many epidemiological studies have shown ...