Prognostic Inflammatory–Immune Score for Non-Gastric GIST Risk Stratification

471 patient records from a multicenter retrospective study of non-gastric gastrointestinal stromal tumors (GISTs). Zhiming Cai and colleagues developed a Prognostic Inflammatory–Immune Score (PIIS) integrated with clinicopathological factors to predict recurrence-free survival. The model demonstrated superior discrimination with C-indices of 0.839 and 0.795 in training and validation cohorts, respectively.

Use Cases

• Training machine learning models for cancer recurrence prediction based on integrated inflammatory-immune and clinicopathological features.
• Validating new risk stratification systems against conventional mNIH and AFIP criteria using multicenter patient data.
• Analyzing the prognostic value of hematological ratios like platelet-to-albumin ratio (PAR) and derived neutrophil-to-lymphocyte ratio (dNLR).
• Investigating the optimal duration of adjuvant imatinib therapy for different risk subgroups defined by a nomogram.

Strengths

• Multicenter validation with an external cohort, enhancing generalizability.
• Model performance is quantified with specific C-index scores (0.839 and 0.795).
• Includes 471 patient records, providing a substantive sample for analysis.
• Feature contributions are interpreted using SHAP (Shapley Additive Explanations).

Limitations

• Column-level documentation is absent; field semantics must be inferred after download.
• Row count is unknown, which may limit suitability assessment for large-scale ML tasks.
• Data is contained in a DOCX file (20.2 MB), which may require parsing to extract structured tabular data.

Provenance

Source

figshare, author Zhiming Cai.

Collection Method

Multicenter retrospective cohort analysis.

Freshness

Last updated 2026-06-02 05:34:54; freshness should be verified.