Personalized tumor-informed ctDNA monitoring for recurrence in high-risk locally advanced gastrointestinal stromal tumor.

Abstract

e23514 Background: Gastrointestinal stromal tumor (GIST) is the most prevalent mesenchymal neoplasm of the gastrointestinal tract. While radical resection is the gold-standard treatment for localized GIST, the 5-year recurrence rate of high-risk GIST exceeds 50%, posing a major clinical challenge. Whether ctDNA-based detection of postoperative molecular residual disease (MRD) can reliably predict recurrence in high-risk locally advanced GIST patients remains a key question in translational oncology. Methods: This prospective study enrolled patients with high-risk locally advanced GIST who underwent R0 resection. Surgical tissue specimens were obtained for genomic profiling. Blood samples were collected at baseline (pre-surgery), at a landmark time point within one month after surgery, and serially thereafter at 3- to 6-month intervals during follow-up. Tumor-derived variants were identified through whole-exome sequencing of resected tumors. A personalized tumor-informed assay was designed by selecting up to 50 top-ranked variants with a variant allele frequency ≥3.0% to monitor MRD status. Results: As of Jan 2026 (cutoff), 44 eligible patients were enrolled. 44 tumor samples, 42 pre-surgical plasma samples, and 330 post-surgical blood samples were analyzed, with a median follow-up of 21 months. Tumor sequencing detected mutations in KIT and PDGFRA in 39 (88.6%) and 4 (9.1%) patients. Prior to surgery, MRD was detected in 57.1% (24/42) of plasma samples, and positivity correlated with larger tumor volume, higher Ki-67 index, mitotic count, and TMB (all P < 0.05). Multivariable logistic regression identified the Ki-67 index as an independent predictor of pre-surgical MRD positivity (OR, 1.20; 95% CI, 1.02 to 1.41; P = 0.024). In landmark analysis, ctDNA was detectable in 4 patients (4/41, 9.8%). These patients had worse DFS trend vs. MRD-negative patients (HR=3.48, 95% CI=0.70–17.35, P =0.11). To date, 8 patients had radiologically confirmed recurrence via CT. Among 4 patients with both MRD positivity and recurrence, one had longitudinal MRD positivity at 9 months post-op (3-month lead before CT relapse), and 2 had simultaneous detection by both methods. Notably, none of these 4 patients received regular adjuvant therapy. Additionally, multivariable Cox regression (adjusted for sex, age, TNM stage, and adjuvant therapy) showed longitudinal MRD positivity was associated with inferior DFS (HR=4.59, 95% CI=1.00–20.98, P =0.05). Patients with sustained negative MRD (n=29) had better survival than those with MRD conversion to negative (n=2), positive (n=8), or persistent positivity (n=2) (P=0.1). Conclusions: This study indicates that personalized tumor-informed ctDNA may predict recurrence in high-risk locally advanced GIST patients, particularly those not receiving regular adjuvant therapy. Clinical follow-up is ongoing. Clinical trial information: NCT05408897 .