Study on the predictive value of preoperative CT features for the mitotic index of GIST based on the nomogram

Abstract

Abstract This study aimed to construct a Nomogram based on preoperative CT features to predict the mitotic index in gastrointestinal stromal tumors and to establish preoperative risk stratification. The constructed nomogram prediction model is targeted towards guiding preoperative risk stratification, facilitating the provision of rational drug administration regimens, and tailoring appropriate surgical plans for personalized treatment. The imaging and pathological data of 250 patients with gastrointestinal stromal tumors in Shanxi Provincial hospital from January 2019 to January 2024 were retrospectively analyzed. According to the pathological data, the patients were divided into high mitotic index and low mitotic index, and were divided into a training group (n = 176) and a validation group (n = 74) according to a stratified sampling ratio of 7:3. In the training group, statistically significant variables were screened out by univariate analysis for multivariate logistic regression analysis, and independent risk factors were screened out and a Nomogram prediction model was constructed. The receiver operating characteristic (ROC) was used to evaluate the model discrimination, and the predicted probability risk was stratified by the optimal cutoff value. The Hosmer-Lemeshow test (HL test) was performed, and the calibration curve was drawn by Bootstrap repeated sampling 1000 times to evaluate the model consistency. Finally, the clinical application value of the prediction model was evaluated by the decision curve analysis (DCA). There were no significant differences in the distribution of clinical characteristics and CT features between the training group and the validation group ( P>0.05). Univariate analysis showed that the differences in tumor size, tumor site, boundary, calcification, liquefaction/necrosis, morphological characteristics, growth pattern, and ulceration were statistically significant (P<0.05). Multivariate logistic regression analysis screened out tumor size (GIST ≤ 2 cm, P = 0.018; GIST 2–5 cm, p = 0.009; GIST 5–10 cm, P = 0.017), liquefaction/necrosis (P = 0.002), and morphological characteristics (P = 0.002) as independent risk factors for high mitotic index. The Nomogram was established based on these three factors. The area under the curve (AUC) of the training group and the validation group of the model were 0.851 (95%CI: 0.793–0.91) and 0.836 (95%CI: 0.735–0.937), the specificity was 0.696 and 0.735, and the sensitivity was 0.869 and 0.760, respectively. The HL test had good calibration (training group P = 0.461, validation group P = 0.822), indicating that the predicted risk was consistent with the actual risk. The DCA also showed good clinical practicality. The Nomogram prediction model that incorporates preoperative CT features of tumor size, liquefaction/necrosis, and morphological characteristics can effectively predict the number of mitotic figures in gastrointestinal stromal tumors, and can perform effective preoperative risk stratification to guide clinical decision-making and personalized treatment.

Keywords

• Gastrointestinal stromal tumor
• Mitotic index
• CT features
• Prediction model