The Role Of Classification And Regression Tree On Assesssing The Predictors For Parathyroid Crisis

Objective: To analyze the clinical features of patients with parathyroid crisis(PC),and use classification and regression tree(CART)analysis to create a predictive model of PC risk factors,so as to intervene in high-risk population timely and reduce the occurrence of adverse outcomes of PC.Methods: The clinical data of 102 cases of primary hyperparathyroidism(PHPT)treated in the first affiliated hospital of Chongqing Medical University from Jan.2013 to Mar.2019 were consecutively collected and these patients were divided into crisis group and non-crisis group according to whether serum calcium level was equal or greater than 3.5mmol/L and whether there was multiple organ involvement.Then comparison of general data,clinical manifestations,laboratory tests,complications and surgical data between the two groups were conducted,and finally,the predictors for PC were revealed by CART analysis.Results: 1)16 cases(15.7%)were enrolled into crisis group,while86 cases(84.3%)were included into non-crisis group.2)Compared with the non-crisis group,the patients in the crisis group were more likely to be men and had lower body mass index(BMI);the incidence of nonspecificsymptoms,digestive symptoms,neuromuscular symptoms and hypercalcemia-related symptoms score(HRSS)were higher in crisis group;3)Serum calcium concentration,parathyroid hormone(PTH)level,serum creatine level in crisis groups were higher than those in non-crisis group,while serum albumin level,estimated glomerular filtration rate(eGFR),serum magnesium,potassium and 24-hour urinary calcium levels were lower in crisis group compared with non-crisis group;however,although the incidence of urinary calculi level,left ventricular hypertrophy and arrhythmia was higher in crisis group,there was no significant difference between the two groups;4)Tumor size was larger in crisis group,and the main pathological type was more likely to be adenoma,but there was also no significant difference between the two groups,while the crisis group was more likely to develop hungry bone syndrome after parathyroidectomy;5)A decision tree model was created by CART analysis,which suggested PTH level,eGFR and serum magnesium concentration could be predictors of PC,and PTH level was the most important predictor.Conclusion: PC is relatively uncommon,but has some unique clinical characteristics.Once the diagnosis of PHPT is established,the high-risk patients for PC may be identified through the decision tree model created by this study,and should be treated with surgery as soon as possible to reduce adverse outcomes of PC.