Back to 2015 Annual Meeting Posters
Cure Predictability During Parathyroidectomy: Validation of a Novel Cure Predictability Model
*Glenda Callender1, *Roy Lirov2, *Paul Gauger2, Robert Udelsman1
1Yale University School of Medicine, New Haven, CT;2University of Michigan School of Medicine, Ann Arbor, MI
Objective: To validate a mathematical model yielding intraoperative parathyroid hormone (IOPTH) cure predictability curves for patients with primary hyperparathyroidism (PHPT) using data from a second institution with a different IOPTH protocol.
Design: Retrospective review
Setting: Tertiary care academic center
Patients: 102 consecutive patients with sporadic PHPT and normal renal function underwent parathyroidectomy between 1/1/2012 and 5/31/2013.
Interventions: IOPTH data analyzed using cure predictability software; predicted cure ≥98% considered intraoperative biochemical cure.
Main Outcome Measures: Intraoperative cure predictions; postoperative cure rates.
Results: 100(98%) patients experienced biochemical cure. 26/102(25.5%) patients had IOPTH samples drawn from the internal jugular vein (IJV), a method not utilized in the original institution; 24/102(23.5%) did not have samples drawn at time of gland resection, a data point always available during model formation. The model correctly predicted cure in 94/102(92.2%) patients. Image-guided targeted parathyroidectomy was planned in 84 patients; the model correctly predicted cure in 83/84(98.8%). Of these, 75/84(89%) had a single adenoma; all had predicted and actual cure. The remaining 9/84(11%) patients underwent additional exploration and all were cured; of these, the model accurately predicted residual disease after initial resection in 8 and incorrectly predicted cure in 1 patient with a markedly elevated baseline IJV IOPTH level. Planned bilateral exploration was performed in 18 patients; the model incorrectly predicted cure in 1 and incorrectly predicted failure in 3 patients. The model correctly predicted cure in 73/76(96%) patients with peripheral blood samples and in 21/26(81%) patients with IJV samples (p=0.024).
Conclusions: The cure predictability model performs well utilizing data from a second institution with an IOPTH protocol for which the model was not designed. IJV sampling led to the majority of model inaccuracy.
Back to 2015 Annual Meeting Posters
