IntroductionIn Japan, many previous studies about nurses ' health habits, such as smoking, have been conducted. However, most of them were cross sectional studies, and few longitudinal studies have been done. Fluctuations in smoking habits, such as cessation of smoking or restarting of smoking after quitting, were not surveyed.In a cohort study, information on an individual is taken at baseline, and it usually remains fixed thereafter. However, if health habits change over time, there is some possibility of misclassifying exposure. In prospective cohort studies of health behaviors and chronic diseases, it is important to know the stability of health habits. Objective The purpose of this study was to examine changes in smoking, drinking, and dietary habits using a questionnaire that was repeated in the same subjects, as well as to determine how the behavioral changes were affected by age and reproductive events.Subjects and Methods The Gunma Nurses' Health Study GNHS is an ongoing prospective study of a cohort of 699 female nurses. They were followed using a self reported biennial questionnaire dealing with lifestyle and healthcare practices on women's health, reproductive history, and medical history. The 450 women who responded to all three questionnaires baseline, second, and third surveys were included in the analysis. We evaluate the reproducibility of smoking habits and food habits, and examine the impact of these variables on the degree of reproducibility.Results The prevalence of current smokers decreased slightly over the 4 years from 14.7 to 11.3 . For smoking habit, the total agreement was 95.0 , and the kappa coefficient was 0.788.Spearman's rank correlation coefficient for food items ranged from 0.545 for milk and dairy products to 0.821 for alcoholic beverages. Age and pregnancy affected the reproducibility of smoking and dietary habits, but menopausal status had no effect on these health habits. Epidemiological studies should consider long term changes in health habits to assess habitual exposures.