The impact of indirect notification of a cancer diagnosis and a risk model based on it to predict the prognosis of postoperative stage T3 esophageal cancer patients.

Abstract

Chinese doctors are required to inform patients' direct relatives of a cancer diagnosis rather than the patients themselves. The disease may be hidden from patients by their family members, which could result in severe outcomes. We selected postoperative T3 esophageal cancer (EsC) patients hospitalized from June 2015 to December 2019 as research subjects. The patients were divided into a direct-notification group and an indirect-notification group. Several variables were used to evaluate both groups' 36-month progress-free survival (PFS). A risk prediction model of prognosis based on the risk score was established, which was assessed using the area under the curve (AUC) of the receiver operating characteristic curve. One hundred and thirteen patients were enrolled in the training group and forty-eight in the validation group. Cox multivariate regression analysis revealed that males, late stage, poor pathological differentiation, and indirect notification were independent worse risk factors for postoperative T3 stage EsC patients at 36-month PFS (hazard ratio (HR) = 0.454, 95% confidence interval (CI): 0.254-0.812, P = .008; HR = 1.560, 95% CI: 1.006-2.420, P = .047; HR = 0.595, 95% CI: 0.378-0.936, P = .025; HR = 2.686, 95% CI: 1.679-4.297, P < 0.001, respectively). The type of notification was the best correlation factor. The risk score was calculated as follows: risk score = 0.988 × cancer notification (indirect = 1, direct = 0)-0.790 × sex (female = 1, Male = 0) + 0.445 × stage (IIIB = 1, IIA + IIB = 0)-0.519 × pathological differentiation (moderately + well = 1, poorly = 0). The model had a sensitivity of 64.8% and specificity of 81.8%, with the AUC at 0.717 (95% CI: 0.614-0.810) in internal verification, and a sensitivity of 56.8% and specificity of 100%, with the AUC at 0.705 (95% CI: 0.651-0.849) in external validation. The model had good internal and external stability. The model showed a Brier score of 0.18. Indirect notification of a cancer diagnosis was an important negative predictor of postoperative EsC patients' PFS. The model displayed good accuracy and stability in the prediction of risk for cancer progression.