Development and validation of a predictive model for follow-up non-adherence in intermediate- and high-risk gastric gastrointestinal stromal tumors

For patients with intermediate- or high-risk gastric gastrointestinal stromal tumors (gGISTs), postoperative surveillance is vital for early recurrence detection and improving survival. However, non-adherence to scheduled follow-up remains a notable clinical challenge. This study aimed to develop and validate a predictive model for identifying patients at high risk of non-adherence. This multi-center retrospective study (2015–2022) included intermediate-/high-risk patients with gGISTs who underwent surgery at five hospitals. The training cohort (TC, n = 449) and the internal validation cohort (IVC, n = 193) came from two hospitals, and the external validation cohort (EVC, n = 268) came from three others. Adherence was defined as guideline-recommended abdominopelvic contrast-enhanced CT follow-up. Multivariate logistic regression identified independent risk factors for follow-up non-adherence, informing a scoring system. Model performance was assessed via the area under the receiver operating characteristic curve (AUC), sensitivity, specificity, positive predictive value (PPV), negative predictive value (NPV), and accuracy. The incidence of non-adherence in the TC, IVC, and EVC was 47.7%, 47.2%, and 52.2%, respectively. Multivariate analysis identified three independent predictors: unmarried status (OR = 2.91, 95% CI 1.86–4.60, P < 0.001), unemployment (OR = 3.35, 95% CI 1.83–6.27, P < 0.001), and imatinib-related severe adverse events (OR = 6.91, 95% CI 4.48–10.81, P < 0.001). A scoring system was developed, assigning 1 point each for unmarried status and unemployment, and 2 points for adverse events. The model showed acceptable performance in the TC (AUC = 0.786, accuracy = 73.1%), IVC (AUC = 0.794, accuracy = 75.1%), and EVC (AUC = 0.783, accuracy = 72.8%). Patients were stratified into low-risk (scores 0–2) and high-risk (scores 3–4) groups for non-adherence, with non-adherence rates of 34.3% and 84.0% in the IVC, and 38.6% and 84.8% in the EVC, respectively. We developed and externally validated a scoring system based on marital status, employment status, and imatinib-related adverse events to predict follow-up non-adherence in intermediate- and high-risk gGISTs. This model might help identify patients at increased risk of non-adherence who may require more intensive follow-up support; however, its impact on recurrence detection, survival, and quality of care was not evaluated in this study.