Latest Research
All publications from the Cancer3.AI database, newest first.
Construction of an XGBoost-SHAP-based malignant transformation risk prediction model for gallbladder polyps.
Luo W, et al
Researchers developed and validated a machine learning model to predict which gallbladder polyps are at risk of becoming cancerous, a clinical challenge that currently lacks reliable tools. Using data from 1,027 surgical patients across two Chinese hospitals, the team applied the XGBoost algorithm combined with SHAP (Shapley Additive Explanations) technology to build an interpretable predictive framework. LASSO regression analysis identified five key risk factors for malignant transformation: the presence of concomitant cholecystitis, polyp count, polyp base width, patient age, and maximum polyp diameter. The model achieved a strong area under the ROC curve (AUC) of 0.862 in the training cohort and 0.777 in an independent validation cohort, indicating good predictive accuracy. The use of SHAP technology allowed clinicians to understand exactly how each factor contributed to individual risk scores, enhancing the model's practical transparency. This tool could help physicians make more informed, personalized decisions about when to operate versus monitor patients with gallbladder polyps, potentially improving early cancer detection outcomes.
Annals of medicine
Source →Optimizing Lymph Node Staging in Gallbladder Cancer: A Cutoff of 11 Nodes for Accurate Staging and Improved Survival.
Yoon SJ, et al
Researchers investigated the optimal number of lymph nodes that surgeons should remove and examine during gallbladder cancer surgery to ensure accurate disease staging, studying 601 patients across two independent cohorts who underwent curative-intent operations between 2008 and 2017. Using a statistical modeling approach, the team found that retrieving at least 11 lymph nodes reduced the rate of false-negative nodal staging — cases where cancer spread to lymph nodes goes undetected — to approximately 11 to 13 percent across all tumor stages. Spline-based survival analyses identified an inflection point around 11 retrieved lymph nodes, suggesting that this threshold may correspond to a meaningful shift in patient outcomes, though formal survival comparisons between groups did not reach statistical significance. These findings provide clinicians with a practical benchmark: aiming to examine at least 11 lymph nodes during gallbladder cancer surgery may help ensure that nodal involvement is not missed, which is critical because incorrect staging can lead to undertreated disease. The study recommends that this 11-node threshold be considered a reference standard in surgical and pathological practice to improve the reliability of gallbladder cancer staging worldwide.
Journal of hepato-biliary-pancreatic sciences
Source →The effect of tumor budding index on clinical parameters in early-stage oral cavity cancers.
Toprak F, et al
Researchers in Turkey followed 87 patients surgically treated for oral cavity cancer between 2010 and 2021 to determine whether tumor budding — the presence of small clusters of cancer cells at the invasive front of a tumor — could predict outcomes in early-stage oral cavity squamous cell carcinoma. The tongue was the most common tumor site, accounting for 70 percent of cases, and patients were monitored for a median of 46.5 months. The study found that tumor budding was an independent predictor of overall survival, disease-free survival, and lymph node metastasis, even when standard staging factors were taken into account. Patients with high tumor budding had a significantly greater risk of cancer spreading to nearby lymph nodes and experienced shorter survival times. These findings indicate that routinely assessing tumor budding in surgical pathology specimens could meaningfully improve postoperative risk stratification and guide more individualized surveillance or treatment planning for oral cavity cancer patients.
Acta oto-laryngologica
Source →Mutational signature-based classification uncovers emerging oral cancer subtypes with distinct molecular patterns.
Deneuve S, et al
A new study published in the International Journal of Oral Science investigated the molecular basis of oral cavity squamous cell carcinoma (OCSCC) in patients with no identified traditional risk factors — non-smokers, non-drinkers, and HPV-negative individuals — a group whose cancer incidence has been rising at an unexplained rate. Researchers analyzed publicly available multi-omics data from 253 OCSCC cases and 94 laryngeal cancer cases, using mutational signature analysis to classify tumors and reveal biologically distinct subtypes. The study uncovered two novel molecular subgroups among patients lacking traditional risk factors: one enriched in tongue cancers driven by heightened endogenous clock-like mutagenesis linked to natural cellular aging, and a second group characterized by elevated APOBEC enzyme-associated mutations, both harboring unique cancer driver gene mutations and distinct epigenetic methylation patterns not observed in tobacco- or alcohol-related cancers. These risk-factor-free tumors also displayed pronounced immune evasion mechanisms and antimicrobial transcriptomic responses, pointing to potential roles of the oral microbiome in tumor development. These findings provide the first comprehensive genomic and molecular characterization of this emerging oral cancer subtype and carry significant implications for early detection strategies, personalized treatment approaches, and cancer prevention in patients who do not fit conventional risk profiles.
International journal of oral science
Source →Distinct mechanisms of otitis media with effusion before/after treatment in nasopharyngeal carcinoma.
Kumai T, et al
A new study published in Auris, Nasus, Larynx investigated the mechanisms underlying otitis media with effusion (OME) — a condition in which fluid accumulates in the middle ear — in patients with nasopharyngeal carcinoma (NPC), specifically comparing cases occurring before and after cancer treatment. Among 31 patients with newly diagnosed NPC, pre-treatment OME (present in 66% of patients) was significantly associated with tumor-related factors, including advanced primary tumor classification and MRI-confirmed involvement of the tensor veli palatini muscle, the structure responsible for normal middle ear ventilation. Post-treatment OME (present in 45% of patients), however, showed no meaningful association with tumor extent or lymph node involvement; instead, it was significantly linked to poor mastoid pneumatization, an anatomical host characteristic of the bone behind the ear. Critically, 35% of patients experienced persistent OME both before and after treatment, demonstrating that successful tumor control does not reliably resolve middle ear fluid accumulation. These findings establish that pre- and post-treatment OME in NPC are biologically distinct clinical entities driven by different mechanisms, which carries important implications for how clinicians monitor and manage hearing-related complications in NPC patients throughout and beyond their cancer treatment.
Auris, nasus, larynx
Source →