OBJECTIVE: To summarize the effects of long-term osteoporosis drug treatment (ODT) and ODT discontinuation and holidays on fractures and harms. DATA SOURCES: MEDLINE((R)), Embase((R)), and Cochrane databases from 1995 to October 2018; ClinicalTrials.gov; bibliographies of relevant systematic reviews. REVIEW METHODS: We defined long-term ODT as >3 years and ODT holidays as discontinuation for >/=1 year after >/=1 year of use. Trials were used for incident fractures and harms, and controlled observational studies were included for additional harms. Two investigators rated risk of bias. For studies with low or medium risk of bias, one investigator extracted data and a second verified accuracy. Two investigators graded strength of evidence (SOE). RESULTS: Sixty-one English-language studies were included. In women with osteoporosis, 4 years of alendronate reduced clinical fractures (hazard ratio [HR] 0.64 [95% confidence interval (CI) 0.50, 0.82]) (moderate SOE) and radiographic vertebral fractures (HR 0.50 [95% CI 0.31, 0.82]) (moderate SOE), while 4 years of raloxifene reduced clinical vertebral fractures (relative risk 0.58 [95% CI 0.43, 0.79]) (high SOE), but not hip (moderate SOE) or nonvertebral fractures (high SOE). In women with osteopenia or osteoporosis, 6 years of zoledronate reduced incident clinical fractures (HR 0.73 [95% CI 0.60, 0.90]) (moderate SOE) and clinical vertebral fractures (HR 0.41 [95% CI 0.22, 0.75]) (moderate SOE). In postmenopausal women with unknown osteoporosis or osteopenia status, both long-term oral estrogen and estrogen/progestin reduced clinical fractures (high SOE) and hip fractures (moderate SOE). After 3-5 years of prior treatment, continuation of zoledronate or alendronate versus drug holiday inconsistently reduced incident vertebral fracture outcomes (radiographic only for zoledronate [low SOE], clinical only for alendronate [moderate SOE]), but did not reduce nonvertebral fractures (low SOE). Hormone therapies increased cardiovascular events, mild cognitive impairment or dementia, and other harms. Observational studies showed that long-term bisphosphonates may increase atypical femoral fractures (AFF) (low SOE) and osteonecrosis of the jaw (low SOE in 2 comparisons, insufficient in 1). LIMITATIONS: Most data were from white, healthy, postmenopausal women, limiting generalizability. Trials often had low power for incident clinical fractures. No trials compared active treatments, sequential treatments, or different durations of drug holidays. Harms and controls were inconsistently defined. CONCLUSIONS: Long-term alendronate, zoledronate, and oral hormone therapy reduced nonvertebral fractures in older women, with oral hormone therapy also reducing hip fractures. While absolute reductions in typical fractures with long-term bisphosphonates are large relative to increases in AFF, reduced hip fracture risk with oral hormone therapy appears offset by increased risk of serious harms. Evidence is limited regarding ODT holidays for fractures and harms. Future research is needed, including randomized trials comparing ODT holiday durations and sequential treatments powered for clinical fractures, and controlled cohort studies of ODT holidays to estimate rare harms.