OBJECTIVES:To describe the natural history of frailty transitions in a large cohort of community-dwelling older men and identify predictors associated with progression to or improvement from states of greater frailty. DESIGN:Prospective cohort study. SETTING:Six U.S. sites. PARTICIPANTS:Community-dwelling men aged 65 and older (N = 5,086). MEASUREMENTS:Frailty was measured at baseline and an average of 4.6 years later. Frailty was defined as having three or more of the following components (low lean mass, weakness, self-reported exhaustion, low activity level, and slow walking speed); prefrailty was defined as having one or two components. Separate multivariable logistic regression models were analyzed for progression and improvement in frailty status. RESULTS:Of the 5,086 men, 8% were frail, 46% were prefrail, and 46% were robust at baseline. Between baseline and follow-up, 35% progressed in frailty status or died, 56% had no change in frailty status, and 15% of prefrail or frail participants improved, although only 0.5% improved across two levels, from frail to robust. In multivariable models, factors associated with improvement in frailty status included greater leg power, being married, and good or excellent self-reported health, whereas presence of any instrumental activity of daily living (IADL) limitations, low albumin levels, high interleukin-6 levels, and presence of chronic obstructive pulmonary disease or diabetes mellitus were associated with lower likelihood of improvement in frailty status. CONCLUSION:Improvement in frailty status was possible in this cohort of community-dwelling older men, but improvement from frail to robust was rare. Several predictors were identified as possible targets for intervention, including prevention and management of comorbid medical conditions, prevention of IADL disability, physical exercise, and nutritional and social support.