Serum albumin levels are a practical marker of general health status in the elderly and have been used to determine the severity of underlying diseases and the risk for death. This longitudinal study evaluated the relationship between serum albumin concentrations and tooth loss over 10 years in elderly subjects, after controlling for confounding factors. A sample of 554 dentate subjects among enrolled subjects (n = 600) was involved in this planned longitudinal study with follow-up examinations after 5 and 10 years. At the 5-year follow-up, 373 (67·3%), subjects were available for re-examination. In addition, 331 (59·7%) were available at the 10-year follow-up. Multiple Poisson regression analysis was conducted to evaluate the relationship between the number of missing teeth over 5 or 10 years and serum albumin levels at baseline after adjusting for 10 variables: gender, serum markers levels at baseline, dental status, smoking habits and educational years and oral health behaviour. The number of missing teeth over 5 or 10 years was significantly negatively associated with serum albumin levels at baseline [incidence rate ratios (IRR) = 0·373, P < 0·0001 for 5 years; IRR = 0·570, P < 0·0001 for 10 years]. We conclude that elderly subjects with hypoalbuminemia are at high risk for 5- and 10-year tooth loss.