Abstract 457P
Background
Existing prediction models for gastric cancer (GC) mostly faced a high risk of bias and showed restricted applicability. Current GC screening guideline in China gave region-specific high-risk classifications using a fixed cut-off value of local GC incidence, which might not be appropriate considering the constantly decreasing GC burden. We aimed to develop and evaluate a questionnaire-based GC risk assessment tool while fully taking regional disease burden in account, and compare it with guideline recommendations among Chinese population.
Methods
A multicenter prospective cohort in China was used to evaluate relative and attributable risks of GC. A total of 74,307 participants aged 40-69 years were recruited during 2015-2017 and followed up to Dec 31, 2021. Cox models were used to identify GC risk factors and build a relative risk model. Absolute GC risks were calculated by incorporating local age-, sex- and residence-specific GC incidence and GC-specific, and overal mortality rates. Model discrimination was evaluated by areas under the receiver operating curve (AUC), and the RRs (relative risks) of participants classified by predicted risk. Model calibration was assessed by Estimated/Observed (E/O) ratio and calibration plot by the deciles of predicted risk.
Results
During a median of 6.28 years of follow-up, 383 GC cases were observed. Stratified by age, sex, and region, six risk factors including pack-year of smoking, alcohol drinking, BMI, salty-food intake, family history, and fruit intake showed independent associations with GC incidence and were included in thee final model. The model was well-calibrated, yielding an E/O ratio of 0.91 (95%CI, 0.82-1.00). The AUCs of the absolute risk model, relative risk model, and guideline recomendation were 0.70 (0.66-0.73), 0.60 (0.57-0.63) and 0.52 (0.49-0.55), respectively. Individuals with highest 20% risk-scores showed significantly higher risks than the lowest group (RR=2.96, 1.02-8.60).
Conclusions
Our model has good calibration and discrimination performance compared with current guideline recommendation. It may serve as an effective tool to assist risk-stratified screening among regions with various GC burden.
Legal entity responsible for the study
The authors.
Funding
National Natural Science Foundation of China.
Disclosure
All authors have declared no conflicts of interest.