Several cross-sectional studies have demonstrated an age-related decrease in serum 25-hydroxyvitamin D [25(OH)D] concentrations. No prospective studies, however, have been conducted to examine this correlation to date. The objectives of this study were to measure age-related changes in serum 25(OH)D concentrations and to identify predictors of change in serum 25(OH)D concentrations in the frail elderly during a 2-year follow-up period. Eighty elderly subjects (48 women and 32 men) were selected from people utilizing the long-term care insurance system in a community in Japan. All subjects participated in both the baseline and follow-up (2 years later) medical examinations. Baseline measurements included age, height, body mass index, and weight. Additionally, levels of activities of daily living (Barthel index), grip strength, lifestyle, serum 25(OH)D, intact parathyroid hormone (PTH), albumin, total protein, and creatinine concentrations were also determined at baseline. The average age of the subjects was 82.1 years (SD, 8.8). The 2-year decrease in serum 25(OH)D concentrations was calculated to be 6 nmol/l. Multiple linear regression analyses found that the 2-year change (Delta) in the log-transformed 25(OH)D was associated with Deltaalbumin (beta = 0.503, R (2) = 0.288, P < 0.0001), and Deltalog-transformed intact PTH was associated with baseline creatinine (beta = 0.453, R (2) = 0.142, P = 0.0006) and Deltalog-transformed 25(OH)D (beta = -0.512, R (2) = 0.103, P = 0.0037). In conclusion, serum 25(OH)D concentrations decreased in the 2-year follow-up. This information is useful for the maintenance of vitamin D status and prevention of vitamin D insufficiency in the frail elderly.