Despite advances in the medical management of secondary hyperparathyroidism, parathyroidectomy remains necessary in some end-stage renal disease patients. Observational studies may help with the design of intervention trials. We linked the retrospective Waves 1, 3, and 4 Dialysis Morbidity and Mortality Study datasets to Medicare claims data to identify incident parathyroidectomy in 10,588 Medicare patients receiving hemodialysis in the United States on December 31, 1993. The mean age was 60.0 years, and the mean follow-up 3.6 years. De novo parathyroidectomy incidence was 14.2/1000 patient-years. Considered as quintiles (Q), higher levels of standard bone metabolism variables were associated (p<0.0001) with parathyroidectomy stepwise, such that adjusted hazards ratios (AHR) for Q5 (vs. Q1) were, for calcium (>10.3 mg/dL), 5.09 (3.64-7.10); for phosphorus (>7.5 mg/dL), 2.92 (2.06-4.15); for calcium-phosphorus product (>71 mg2/dL2), 3.32 (2.27-4.85); and for parathyroid hormone (PTH; >480 pg/mL), 13.81 (7.47-25.55). Other antecedent associations included younger age, lower hemoglobin, and longer dialysis vintage, while transplantation, as a time-dependent covariate, was associated with lower hazards ratios. Using interval Poisson analysis, parathyroidectomy was associated with higher mortality risk ratios in the first year, and progressively lower risk ratios subsequently. Demographic variables may modify the risk of parathyroidectomy. Younger patients on long-term hemodialysis may be at a special risk. Parathyroidectomy risk increases stepwise with alterations in bone metabolism variables, suggesting that a single-threshold management approach may not be ideal.