Identifying high-risk candidates for prolonging progression-free survival in primary gastric carcinoma subject to “double invasion”: an analytical approach utilizing lasso-cox regression

• Published in: BMC cancer Vol. 25; no. 1; pp. 381 - 16
• Main Authors: Wang, Liwei, Chang, Yu, Ma, Jinfeng, Qu, Wenqing, Li, Yifan
• Format: Journal Article
• Language: English
• Published: England BioMed Central Ltd 28.02.2025; BioMed Central; BMC
• Subjects: Accuracy; Adult; Aged; Biomarkers; Calibration; Cancer invasiveness; Cancer therapies; Carcinoma; Chemotherapy; Complications and side effects; Datasets; Development and progression; Female; Gastrectomy; Gastric cancer; Gastrointestinal surgery; Health aspects; Hospitals; Humans; Lymph nodes; Lymphatic system; Male; Medical records; Middle Aged; Neoplasm Invasiveness; Neoplasm Staging; Nomogram; Nomograms; Nomography (Mathematics); Patients; Prognosis; Progression-Free Survival; Proportional Hazards Models; PTEN protein; Retrospective Studies; Risk Assessment; Risk Factors; Risk groups; Risk stratification; ROC Curve; Statistical analysis; Stomach cancer; Stomach Neoplasms - mortality; Stomach Neoplasms - pathology; Stomach Neoplasms - surgery; Surgical outcomes; Variables; Vascular invasion and neural invasion; China; Gastric cancer; Progression-free survival; Nomogram; Risk stratification; Vascular invasion and neural invasion
• DOI: 10.1186/s12885-025-13810-z
• ISSN: 1471-2407

To identify high-risk gastric carcinoma patients with concurrent vascular and neural invasion ("double invasion") who are at heightened risk of progression-free survival (PFS) decline, enabling personalized clinical management. In this multi-center retrospective study, 559 patients with double invasion who underwent curative gastrectomy between May 2002 and December 2020 were analyzed. Prognostic factors for PFS were identified using Lasso-Cox regression. Model validation included internal bootstrapping, calibration plots, and comparison against the American Joint Committee on Cancer(AJCC) 8th edition TNM staging system via Harrell's C-index, decision curve analysis (DCA), and time-dependent receiver operating characteristic (ROC) curves. The nomogram integrated gender, positive lymph node count, surgical gastrectomy method, PTEN/FHIT expression levels, and maximum tumor diameter. It demonstrated superior predictive accuracy to AJCC staging, with a C-index of 0.651 (95% CI: 0.612-0.691) versus 0.543 (95% CI: 0.517-0.569). Calibration plots showed strong agreement between predicted and observed outcomes. The area under the curve(AUC) for 3- and 5-year PFS predictions were 0.719 (95% CI: 0.655-0.771) and 0.767 (95% CI: 0.670-0.841), respectively. DCA confirmed clinical utility across decision thresholds, and risk stratification effectively differentiated low- and high-risk groups. In the training cohort, the model significantly outperformed AJCC staging (NRI: 0.218, p < 0.01; IDI: 0.085, p < 0.01). However, this superiority was not statistically significant in the validation cohort (NRI: 0.141, p = 0.08; IDI: 0.031, p = 0.239). We developed a Lasso-Cox regression-based nomogram to stratify PFS risk in gastric carcinoma patients with double invasion. While the model outperformed AJCC staging in training, validation cohort results highlight the need for further refinement. This tool holds potential for guiding tailored therapeutic strategies, though broader validation is warranted to confirm clinical applicability.