IntroductionThe main emphasis of the research about adjuvant imatinib for high‐risk gastrointestinal stromal tumors (GISTs) is prolonging the treatment duration and ignores the heterogeneous that 10‐year recurrence rates ranged from about 20%–100%. Thus, this study evaluated the effect of different durations of adjuvant imatinib on outcomes in high‐risk GISTs to explore the feasibility of individual treatment.MethodsWe analyzed 855 high‐risk GIST patients from three centers who underwent macroscopically complete resection between December 2007 and September 2020. The patients were divided into training (n =564) and two validation cohorts (n = 238 and53) based on their source. Recurrence‐free survival (RFS) was the primary point. Cox multivariate analysis was used to develop the nomogram. C‐index, time‐dependent area under the curves, and calibration plots were used to assess the performance of the nomogram.ResultsUnivariate analysis showed that longer adjuvant imatinib was significantly associated with better 5‐year RFS (p < 0.0001). Further investigation identified that the same high‐risk patients with lower tumor‐associated recurrence risk benefitted little from prolonged treatment and that the recommended adjuvant imatinib duration was insufficient for those with higher recurrence risk. A nomogram for predicting 2‐, 3‐, and 5‐year RFS based on different treatment durations and four major risk factors, namely, tumor site, size, mitotic count, and rupture status, was built and validated, with a C‐index of 0.82, 0.74, and 0.70 in training and two external validation cohorts, respectively. An online dynamic nomogram was further developed for clinical applications (https://ruolinliu666.shinyapps.io/GIST/), offering predictive recurrence rates based on different treatment durations and tumor features.ConclusionsWe developed a nomogram to predict the recurrence risk for high‐risk patients according to tumor features and treatment durations of imatinib to help physicians on decision‐making for individualized treatment duration.