BACKGROUND: It was impossible to predict the development of hypocalcemia following parathyroidectomy (PTE) in patients with primary hyperparathyroidism (PHPT) until now. Hypocalcemia may be accompanied by myalgia, generalized seizures up to tetany, and arrhythmias. Hypocalcemia following PTE can be prevented by preoperative cholecalciferol supplementation. However, patients with severe hypercalcemia above 3 mmol/L do not receive vitamin D due to the risk of hypercalcemia progression. Despite the existing data showing the safety of cholecalciferol therapy in case of mild elevation of serum calcium, not all patients are prescribed vitamin D supplementation, probably due to the lack of a suitable tool to assess the postoperative hypocalcemia risks.AIM: To design a mathematical model and a software tool for predicting hypocalcemia 1–3 days post-PTE in PHPT patients using the patient’s demographic and clinical data, laboratory test results and preoperative therapy status.MATERIALS AND METHODS: This retrospective study included 478 PHPT patients diagnosed with adenomas and carcinomas of the parathyroid gland (PTG) who underwent radical PTE between 1993–2010 or 2018–2020 at the Endocrinology Research Centre. The following parameters were analyzed: sex; age; laboratory markers prior to calcimimetic and antiresorptive therapy: PTH, total calcium, phosphorus; osteocalcin (OC), alkaline phosphatase (ALP), C-terminal telopeptide of type 1 collagen. Also we analyzed 25-hydroxyvitamin D (25(OH)D); bone mineral density (BMD) measured by X-ray densitometry; medical history of low-energy fractures; preoperative therapy with denosumab, bisphosphonates, cinacalcet, cholecalciferol. Categorical gradient boosting (CatBoost) was built to predict the risk of postoperative hypocalcemia.RESULTS: The prevalence of severe osteoporosis is higher in the postoperative hypocalcemia group compared to the nonhypocalcemia group (27% vs. 15%), wherein the frequency of preoperative administration of cholecalciferol in this group is lower (8% vs. 25%). A CatBoost model was built to predict postoperative hypocalcemia using 13 predictors (sex, age, PTH, serum total calcium, phosphorus, OC, BMD reduction, 25(OH)D, administration of cholecalciferol, bisphosphonates, denosumab, and cinacalcet. The proposed model http://194.87.111.169/hypocalcemia for post-PTE hypocalcemia in PHPT patients achieved the following metrics: positive predictive value 73.3%-86.7%; negative predictive value 74.9%-89.3%.CONCLUSION: The model can be used to choose the appropriate preand postoperative approaches for patients who undergo rPTE.
Read full abstract