The patterns of weight change over time, and the role of changing patterns of physical activity, especially in relation to weight, have not been studied extensively. Most studies tracking physical activity or body mass patterns used conventional growth modeling methods. These methods assume the presence of an average trajectory in the underlying population. However, considering population heterogeneity, such assumptions may be too simplistic. PURPOSE: The purpose of this study is to access the longitudinal change of cardiorespiratory fitness (CRF), physical activity (PA), and body mass index (BMI) with aging using the generalized additive model (GAM), which can account any age pattern by the non parametric structure of age function. METHODS: We studied a cohort of 1,798 women, aged 30 to 80 years, from the Aerobics Center Longitudinal Study who completed 3 to 29 health examinations from 1970 to 2006. CRF was measured by a maximal Balke treadmill exercise test. PA was self- reported and the total MET-minutes/week of PA was estimated. BMI was calculated from measured height and weight. RESULTS: The relation between age and CRF, PA, and BMI in women was nonlinear for all. The overall pattern of treadmill time was decreasing. After 60 years old, the fitness level tends to decrease rapidly compared to before 60 years. Women at age 50 always had higher PA than younger or older ages. PA tends to increase before 50 years old then to decrease after 50 years old, following a plateau after 60 years. The overall pattern of BMI was increasing. After 60 years, the rate of increasing in BMI became much slower with increasing of age. Adjusting smoking, health status, and the individual exposure of CRF, PA, and BMI did not influence the observed associations. Women who do not meet current PA recommendation or those who were low fit had a higher BMI throughout adulthood compared to their active or fit peers. CONCLUSION: This study introduces a powerful analytic tool, GAM, to track regular physical activity, fitness and body mass index over the adult lifespan, and shows that the age-related longitudinal changes in these exposures are not linear and that weight status during adulthood was associated with the initial activity or fitness levels. Supported by NIH Grants AG06945, HL62508, and DK088195.