Context: Data are needed to guide therapeutic decisions about stopping bisphosphonates after an initial treatment period.
Objective: To define significant predictors of fracture and quantify fracture incidence in risk factor-defined subgroups of women who discontinue zoledronic acid (ZOL) after 3 years of treatment. To determine if continuing ZOL reduces fracture risk in subgroups.
Design: This study is based on data from the 3 year extension of HORIZON.
Setting: Subjects were in the ZOL arm of the Multicenter HORIZON trial.
Participants: One thousand two hundred thirty three women who previously received 3 ZOL treatments during the Core trial.
Intervention: Randomization to three additional annual ZOL (Z6, n = 616) or placebo infusions (Z3P3, n = 617).
Main outcomes: The risk of morphometric vertebral fractures (MorphVertFx) and clinical nonvertebral fractures (NVF).
Results: The incidence of MorphVertFx in Z3P3 was predicted by femoral neck (FN) t-score ≤-2.5 [OR 3.3 (1.4, 8.0), p = .008], total hip (TH) t-score ≤-2.5 [OR 4.0 (1.8, 9.0), p = .0007], and incident MorphVertFx during Core [OR 4.75 (1.4, 16.8), p < .015]. Incidence of NVF was predicted by TH t-score [for 1 decline, HR 1.7 (1.2, 2.6), p = .008], incident NVF during Core [HR 2.5 (1.2, 5.3), p = .014], and prevalent vertebral fracture [HR 3.0 (1.4, 6.3), p = .005]. For MorphVertFx, there were no significant treatment subgroup interactions; absolute fracture reductions with continued ZOL were greatest in high-risk subgroups. For NVF, there were no significant treatment reductions overall or in subgroups and no significant interactions.
Conclusions: After 3 years of ZOL, in women who have a TH t-score above -2.5, no recent incident fracture and no more than one risk factor (almost 55% of the population), risk for subsequent fracture (over three additional years) is low if treatment is discontinued (for MorphVertFx, average risk 3.2% and for NVF, average risk 5.8%). In these patients, discontinuation for up to 3 years is reasonable.