Optimal cutoff value of the serum pepsinogen level for prediction of gastric cancer incidence: the Hisayama Study

Background. Serum pepsinogen (sPG) levels have been established as a good marker of chronic atrophic gastritis and the sequential occurrence of gastric cancer. However, there have been few prospective investigations which investigated the predictive performance of sPG for future gastric cancer incidence. Subjects and methods. We prospectively followed-up a total of 2446 community-dwelling Japanese aged >= 40 years for 10 years and used the Youden's index to determine the cutoff values of the pepsinogen I level and pepsinogen I/II ratio to accurately discriminate gastric cancer events. Predictive performance of sPG was assessed by ROC curve. Results. During the follow-up, 69 subjects developed gastric cancer. The most predictive sPG test criteria were determined to be a pepsinogen I level <= 59 ng/ml and pepsinogen I/II ratio <= 3.9. The sensitivity and specificity of these criteria to discriminate the actual occurrence of gastric cancer were 71.0% and 69.2%, respectively. The area under the ROC curve for gastric cancer occurrence increased significantly by adding the sPG test to the model that included the status of Helicobater pylori infection and other potential risk factors (from 0.742 to 0.809; p for difference in the area < 0.001). Conclusions. This study determined the optimal sPG test criteria for predicting gastric cancer occurrence over 10 years in a general Japanese population. These criteria would be effective to screen for individuals at high risk of this disease.