Development and validation of risk prediction models for permanent hypocalcemia after total thyroidectomy in patients with papillary thyroid carcinoma

Abstract

Hypocalcemia is a common complication and can be permanent in patients following total thyroidectomy (TT). The aim of this study was to identify factors associated with permanent hypocalcemia and to develop a validated risk prediction model for permanent hypocalcemia to assist surgeons in the appropriate follow-up of high-risk patients regarding supplemental therapy. We included data of 92 patients with papillary thyroid carcinoma (PTC) undergoing TT who were randomly allocated in a 7:3 ratio to a training set (n = 65) and validation set (n = 27). Univariate and multivariate logistic regression analyses revealed significant correlations of permanent hypocalcemia with parathyroid hormone (PTH) at postoperative month 1 (IM PTH), IM calcium (Ca), and IM phosphorus (P). These variables were constructed two models. Model 1 used the three indicators listed above; model 2 also included tumor, node, metastasis staging. The receiver operating characteristic (ROC) curve analysis showed that the areas under the curve (AUC) for models 1 and 2 were high for both the training set (0.905/0.913) and the validation set (0.894/0.800). Calibration curves showed good agreement between the incidence of permanent hypocalcemia estimated using the predictive models and the actual incidence. Model 1 may be more concise and convenient for clinical use.

Keywords: Hypocalcemia; Nomogram; Papillary thyroid carcinoma; Predictive model; Total thyroidectomy.

Fig. 1

Model 1 nomogram. PTH parathyroid hormone, Ca calcium, P potassium, IM postoperative month 1.

Fig. 2

Model 2 nomogram. PTH parathyroid hormone, Ca calcium, P potassium, IM postoperative month 1, T, tumor, N node.

Fig. 3

(A) Model 1 nomogram receiver operating characteristic (ROC) curves generated from the training set; (B) Model 1 nomogram ROC curves generated using the validation set. AUC, area under the ROC curve.

Fig. 4

(A) Model 2 nomogram receiver operating characteristic (ROC) curves generated from the training set; (B) Model 2 nomogram ROC curves generated using the validation set. AUC, area under the ROC curve.

Fig. 5

(A) Calibration plot for the training set of Model 1; (B) calibration plot for the validation set of Model 1.

Fig. 6

(A) Calibration plot for the training set of Model 2; (B) calibration plot for the validation set of Model 2.

Fig. 7

(A) Decision curve analysis (DCA) curves for the training set of Model 1 and 95% confidence interval (CI); (B) DCA curves for the validation set of Model 1 and 95% CI.

Fig. 8

(A) Decision curve analysis (DCA) curves for the training set of Model 2 and 95% confidence interval (CI); (B) DCA curves for the validation set of Model 2 and 95% CI.