基于SEER数据库数据的脉络膜黑色素瘤 预后预测模型构建.

Objective To construct the prognostic prediction model for choroidal melanoma. Methods The clinical data (age, gender, marital status, race, TNM stage, pathological type, bone metastasis, lung metastasis, liver metastasis, surgery information, radiotherapy and chemotherapy, survival time, and survival status, etc.) of 2 235 patients diagnosed with melanoma of choroid from 2010 to 2015 were collected from the National Cancer Institute surveillance, epidemiology, and end results （SEER） database. All data were randomly divided into the model group of 1 584 cases and verification group of 651 cases at a ratio of 7：3. First, the clinical data in the model group were analyzed using the Cox hazard proportional regression model of the "survival" package in the R4. 1. 2 software to conduct univariate and multivariate analysis to obtain independent factors affecting the prognosis of patients with melanoma of choroid. The selected influencing factors were used to construct the prognostic prediction model of melanoma of choroid using the "rms" package in the R4. 1. 2 software. The discrimination and consistency of the prediction model were evaluated by C index, ROC and calibration curve in the model group and the verification group. Results Age, pathological type, T stage, M stage, and local radio‐ therapy were independent factors affecting the prognosis of melanoma of choroid. Based on the above factors, a prognostic prediction model for melanoma of choroid was constructed. The C indices of the prognostic prediction model in the model group and the verification group were 0. 732 (95%CI：0. 710，0. 761) and 0. 713 (95%CI：0. 670，0. 756), respectively. The area under the ROC curve of the model group for 1 year，2 years, and 3 years was 0. 780，0. 766, and 0. 777 respectively； the area under the ROC curve of the verification group for 1，2, and 3 years was 0. 890，0. 755, and 0. 734, respectively. The data of both groups were higher than 0. 7, indicating good discrimination. The calibration curves of the model group and the verification group had a high degree of coincidence with the 45°straight line in the plane cartesian coordinate system, indicating good consistency. Conclusion The prognostic prediction model of melanoma of choroid is successfully established, and it has a good predictive effect on the prognosis of patients. [ABSTRACT FROM AUTHOR]