aceMachine learning prognostication in nasopharyngeal carcinoma: a european multicentre analysis of survival and risk of second malignancy
Nasopharyngeal carcinoma (NPC) is rare in Europe, and emerging data suggest poorer outcomes in Caucasian patients compared with Asian populations, highlighting the need for region-specific prognostic tools. Inflammation-based biomarkers and artificial intelligence show promise for risk stratification and prediction of survival and second primary cancers (SPCs). We conducted a retrospective multicentre study including 405 NPC patients from six European institutions. Demographic, clinicopathological, and haematolo…
"Frederick National Laboratory for Cancer Research" by National Institutes of Health (NIH) is marked with Public Domain Mark 1.0. To view the terms, visit https://creativecommons.org/publicdomain/mark/1.0/.
In a retrospective analysis of 405 patients from six European centers, investigators built machine learning models to predict 5-year overall survival and second primary cancer risk in nasopharyngeal carcinoma, a rare cancer in Europe. The cohort had a median age of 52, was 91.6% White/European ancestry, and showed 66.6% 5-year survival with 12.8% developing second primaries.
The work matters because it provides the first ML prognostic models derived from a predominantly Caucasian European cohort and suggests low-cost inflammatory markers add predictive value, especially for second cancers. Uncertainty remains about generalizability beyond six centers, clinical utility given modest AUCs of 0.66 for survival and 0.74 for second cancers, and prospective validation.
- Retrospective multicentre study of 405 NPC patients from six European institutions, median age 52 years, 91.6% White/European ancestry, 77.3% received chemoradiotherapy.
- Five-year overall survival was 66.6% and 12.8% developed second primary cancers.
- Random Forest was best for OS prediction with accuracy 0.74 and AUC 0.66 using complete feature set; SPC prediction reached accuracy 0.80 and AUC 0.74.
- Exclusion of inflammatory markers led to consistent decline in accuracy, with feature-importance analysis highlighting inflammatory ratios among strongest predictors.
Machine learning classifiers using demographic, clinicopathological and inflammatory markers predicted 5-year overall survival and second primary cancer occurrence in a European nasopharyngeal carcinoma cohort.
The rundown
Researchers assembled 405 nasopharyngeal carcinoma patients from six European institutions, predominantly White/European ancestry, to address poorer outcomes reported in Caucasian versus Asian populations and lack of region-specific tools.
They trained and tested multiple machine learning classifiers with and without systemic inflammatory ratios to predict 5-year overall survival and second primary cancer occurrence, reporting best results with Random Forest and noting improved accuracy when inflammatory markers were included.
Sources
- Peer-reviewedEuropean Archives of Oto-Rhino-Laryngology2026-07-17
How should this claim be treated?
The debate