First, we found that the incidence of atypical femur fracture was unrelated to BMD and only weakly related to age. Second, when we performed a subgroup analysis in the Kaiser Permanente cohort that was limited to women who were older than 65 years of age and who had a BMD T-score of less than− 2.5, we found an incidence of atypical femur fracture similar to the incidence in the overall cohort (unpublished data). The correspondents question whether we overestimated the fracture benefit of 5 to 10 years of alendronate, since the FLEX trial1 and observational data2 showed no increase in the risk of clinical fracture after alendronate was discontinued after 5 years. However, studies involving patients with 5 years of previous treatment are not relevant, since our estimate of the fractures that were prevented with alendronate was based on a comparison of women who were never treated with those who were continuously treated. We agree that our findings that show large decreases in the risk of atypical femur fracture after the discontinuation of alendronate provide support for a drug holiday after 5 years, particularly for women at lower risk. Garton questions whether an underestimation of the incidence of atypical femur fracture biases the ratio of such fractures to fragility fractures. However, we think that our adjudication, starting with all femoral-shaft or subtrochanteric fractures, identified virtually all complete unilateral and bilateral atypical femur fractures. However, incomplete fractures would not have been included unless they had been prophylactically repaired. We agree that accumulating microdamage is part of the pathophysiology of atypical femur fracture, but its effect on only a small minority of patients remains unexplained. 3 Regarding the consequences of atypical femur fracture as compared with hip fracture, several studies have suggested similar mortality4 after either type of fracture, although surgical repair of atypical femur fractures can be more challenging. 3