A Prospective Study on Predicting Recurrence Risk of Postoperative High-risk GISTs in NED State Based on MRD Detection Via Liquid Biopsy

A Prospective Study on Predicting Recurrence Risk of Postoperative High-risk Gastrointestinal Stromal Tumors in NED State Based on Minimal Residual Disease Detection Via Liquid Biopsy

The goal of this single-center, prospective, observational cohort study is to learn whether tumor-informed minimal residual disease (MRD) testing in blood can predict recurrence in 66 patients with high-risk gastrointestinal stromal tumor (GIST) who are not currently receiving postoperative adjuvant therapy. The main questions it aims to answer are:

Can changes in MRD during follow-up predict tumor recurrence? How closely do MRD results agree with contrast-enhanced computed tomography (CT) findings, and can MRD identify recurrence earlier than CT? Can tumor genomic features, changes in MRD, and clinicopathological features be combined to develop a model for predicting recurrence risk? Participants will have their surgical tumor tissue analyzed, provide blood samples for MRD testing, and undergo regular clinical follow-up and contrast-enhanced CT examinations for 3 years. Participants may choose to receive and discuss their MRD results. This study will not assign or change treatment based on MRD results. Recurrence and treatment decisions will remain based on routine clinical assessment, primarily contrast-enhanced CT and multidisciplinary review.

This is a single-center, prospective, single-cohort, non-interventional observational study. The prospective cohort will include 66 patients with high-risk gastrointestinal stromal tumor (GIST) who have undergone R0 resection of the primary tumor. Eligible patients will either have no plan to receive postoperative adjuvant therapy or meet protocol-defined eligibility criteria after discontinuing such therapy.

Baseline tumor tissue will undergo whole-exome sequencing (WES) and next-generation sequencing (NGS). Fifty patient-specific variants associated with tumor development, progression, and individualized treatment will be selected to construct a personalized testing module. Cell-free DNA from peripheral blood will be analyzed using unique molecular identifier-based ultra-deep targeted NGS at a sequencing depth of approximately 100,000×. MRD positivity is defined as the detection of at least two significant loci with P ≤ 0.005.

Baseline contrast-enhanced computed tomography (CT) will confirm no evidence of disease before enrollment. Contrast-enhanced CT will be repeated every 3 months during 3 years of follow-up. Magnetic resonance imaging or positron emission tomography-CT may be performed when needed. Two experienced radiologists will independently review the CT images. Disagreements will be adjudicated by a third radiologist. Imaging recurrence will be evaluated according to Response Evaluation Criteria in Solid Tumors version 1.1 and confirmed by a multidisciplinary team.

Blood-based MRD testing will be performed at baseline and at protocol-specified follow-up visits. The study will compare MRD results with CT-based recurrence assessments and evaluate the predictive performance of MRD. It will also estimate the interval between the first MRD signal of recurrence and CT-confirmed recurrence.

The study will not assign or modify treatment. Participants who choose to receive their MRD results will receive a standardized explanation that the results are investigational. MRD results cannot independently establish a diagnosis of recurrence or guide treatment decisions. Participants who initiate a tyrosine kinase inhibitor (TKI) before imaging-confirmed recurrence may remain in follow-up. The primary MRD performance analysis will include only MRD, imaging, and clinical data collected before TKI initiation. Post-TKI recurrence status will not be used to reclassify pre-TKI MRD results in the primary MRD performance analysis.

After 3-year outcomes become available, data from the 66 prospective participants and 49 retrospective cases will be used to develop a recurrence-risk prediction model. The combined cohort will be randomly divided into training and internal test sets using an 8:2 stratified split. Candidate variables will first be screened using area under the receiver operating characteristic curve and missingness criteria. Least absolute shrinkage and selection operator Cox regression will then be used to select key predictors.

An extreme gradient boosting model will be optimized using five-fold cross-validation. Internal validation will be performed using 200 bootstrap resamples. An earlier cohort of 45 patients with high-risk GIST from the same center will initially serve as an external validation set. Model performance will be evaluated using the area under the curve, concordance index, calibration curves, accuracy, sensitivity, and specificity.

Inclusion Criteria:

I. Inclusion Criteria for the Prospective Study on Postoperative Minimal Residual Disease (MRD) Surveillance in High-Risk Gastrointestinal Stromal Tumors (GIST) Without Adjuvant Therapy

1. Aged 18-75 years.
2. Patients suspected of having high-risk GIST based on preoperative imaging, or patients with biopsy-confirmed high-risk GIST who have not received preoperative neoadjuvant therapy.
3. Treatment-naive patients who have not previously received radiotherapy, chemotherapy, surgery, or any other anticancer treatment.
4. Adequate hepatic, renal, and other major organ function, with the patient considered medically fit to undergo surgery.
5. Postoperative pathological confirmation of gastrointestinal stromal tumor.
6. Tumor genotyping demonstrating a genotype associated with primary resistance to imatinib, such as a PDGFRA D842V mutation, or a clear decision by the patient to decline postoperative adjuvant imatinib therapy.
7. A definite contraindication to imatinib or another targeted therapy.
8. Patients who underwent R0 resection of the primary tumor, completed long-term postoperative adjuvant therapy, and discontinued treatment at least 3 months previously.
9. The patient and family members are able to understand the study protocol; the patient voluntarily agrees to participate and provides written informed consent.

II. Inclusion Criteria for the Development of a Genomics-Based Recurrence Risk Prediction Model for High-Risk GIST and Identification of the Optimal Target Population for MRD Surveillance a. The inclusion criteria are identical to those of the prospective study on postoperative MRD surveillance in high-risk GIST without adjuvant therapy.

Exclusion Criteria:

I. Exclusion Criteria for the Prospective Study on Postoperative Minimal Residual Disease (MRD) Surveillance in High-Risk Gastrointestinal Stromal Tumors (GIST) Without Adjuvant Therapy

1. A history of another malignant tumor or the presence of another concurrent malignancy.
2. Emergency surgery required because of intestinal obstruction, perforation, bleeding, or a similar condition.
3. Pregnancy or breastfeeding.
4. A history of severe psychiatric illness.
5. Any contraindication to surgical treatment.
6. Failure to achieve R0 resection.
7. Low- or intermediate-risk disease according to postoperative pathological risk stratification.
8. Presence of distant metastasis.
9. Any other condition that, in the investigator's judgment, renders the patient unsuitable for enrollment.

II. Exclusion Criteria for the Development of a Genomics-Based Recurrence Risk Prediction Model for High-Risk GIST and Identification of the Optimal Target Population for MRD Surveillance

a. The exclusion criteria are identical to those of the prospective study on postoperative MRD surveillance in high-risk GIST without adjuvant therapy.