Development and External Validation of a Prognostic Model for Disease-free Survival After Curative Resection of Duodenal Gastrointestinal Stromal Tumours

Gastrointestinal stromal tumour (GIST) is the most common mesenchymal neoplasm of the digestive tract, originating from the interstitial cells of Cajal. GIST can arise anywhere along the gastrointestinal tract, but most commonly occurs in the stomach (50%-60%) and small intestine (20%-30%), with the colorectal region accounting for approximately 10%. Duodenal GIST is a relatively rare subtype, comprising approximately 5% of all GISTs. Due to its low incidence, clinical understanding of duodenal GIST has long been limited, and its clinical presentation is non-specific, varying considerably with tumour size, location, and growth pattern. The most common symptoms are related to gastrointestinal bleeding, including melaena, haematemesis, haematochezia, or symptomatic anaemia; abdominal pain is also a frequent presentation. With regard to anatomical distribution, the second portion of the duodenum (descending part) is the most commonly involved site.

Surgery is the first-line treatment for localised GIST, and R0 resection (negative margins) is the key to a favourable prognosis. However, because the duodenum is in close proximity to the pancreas, biliary tract, and major vascular structures, surgical decision-making for GIST at this site is far more complex than at other locations. Surgeons must carefully balance oncological radicality against preservation of pancreaticoduodenal function. Despite advances in surgical technique, the risk of recurrence after surgery for duodenal GIST remains substantial. Studies have shown that compared with GIST at other sites, duodenal GIST has a significantly poorer prognosis, and even among patients with the same NIH risk classification, clinical outcomes differ. This suggests that duodenal GIST may exhibit distinct biological behaviour and should not be equated with GIST at other sites. Accurate risk stratification is therefore critical for determining individualised follow-up intervals, the intensity of adjuvant therapy, and informed patient decision-making.

The most widely used risk stratification tool for GIST is the modified National Institutes of Health (NIH) classification, which is based on tumour size, mitotic count, and primary site. However, this classification was developed primarily from data on gastric GIST, and its performance in non-gastric GIST remains highly controversial. Previous studies have reported a C-statistic of 0.73 for the modified NIH criteria in gastric GIST, but only 0.62 in non-gastric GIST; the TNM staging system has demonstrated superior prognostic discrimination compared with the modified NIH criteria in intestinal GIST. Given the substantial differences in surgical complexity, recurrence risk, and treatment decision-making between duodenal GIST and GIST at other sites, the application of generic risk stratification tools to this anatomical subgroup is clearly inadequate.

In addition to conventional prognostic factors such as tumour size, mitotic count, and primary site, the prognosis of duodenal GIST may also be influenced by molecular biomarkers and dynamic clinical factors such as perioperative complications. At the genomic level, KIT exon 9 mutations occur at a slightly higher frequency in duodenal GIST than in GIST overall, and have been confirmed as an independent risk factor for overall survival in patients with duodenal GIST. Compared with KIT exon 11 mutations, exon 9 mutations are associated with poorer response to imatinib targeted therapy and a higher risk of metastasis. This molecular difference may partly explain the poorer clinical prognosis observed in duodenal GIST.

Given the limitations of existing risk stratification tools for duodenal GIST, together with the unique clinicopathological features, surgical complexity, and molecular biological background of tumours at this site, the development of a prognostic assessment tool specifically for duodenal GIST is of considerable clinical importance. In recent years, nomogram models have been widely applied to individualised prognostic prediction in various tumours and have demonstrated good predictive performance in the field of GIST. However, an externally validated prediction model specifically for disease-free survival after radical resection of duodenal GIST is currently lacking.

This study, based on multicentre data, aims to develop and externally validate a generalisable nomogram model for predicting disease-free survival (DFS) after radical resection of duodenal GIST, thereby providing a quantitative reference for individualised follow-up and treatment decision-making in clinical practice.