Aim: To assess the validity of the measurement of pepsinogen I and II as a screening test for gastric cancer and pre-malignant lesions, namely low-grade dysplasia, both in the general population and in selected groups of patients.
Methods: A meta-analysis of sensitivity and specificity results from individual papers on the use of the pepsinogen test. An intrinsic cut-off effect was assumed and a random effect model was used for pooling.
Results: Forty-two data sets were included: 27 (64%) population-based screening studies (n=296,553) and 15 (36%) sets of selected individuals (n=4385). Homogenous sensitivity and diagnostic odds ratio (DOR) estimates were found in studies using both pepsinogen I levels and pepsinogen I/II ratio calculations. Pooled pairs of sensitivity and false positive rates (FPr) for pepsinogen I < or = 70; pepsinogen I/II ratio < or = 3, pepsinogen I < or =50; pepsinogen I/II ratio < or = 3, and pepsinogen I < or = 30; pepsinogen I/II ratio < or = 2, were sensitivity 77%/FPr 27%, sensitivity 68%/FPr 31%, and sensitivity 52%/FPr 84%, respectively. Positive predictive values (PPV) varied between 0.77% and 1.25%, and negative predictive values (NPV) varied between 99.08% and 99.90%. In selected groups, pooling was only possible when considering pepsinogen I < or = 70; pepsinogen I/II ratio < or = 3: giving sensitivity 57%, specificity 80%, PPV 15% and NPV 83%. As for the diagnosis of dysplasia, studies considering pepsinogen I <50; pepsinogen I/II ratio <3 obtained sensitivity 65% and specificity ranging from 74%-85%, both with NPV >95%.
Conclusion: Pepsinogen test definition should include pepsinogen I/II ratio as consistency was obtained, both in population based studies and in selected groups for those studies that used pepsinogen I serum levels together with pepsinogen I/II ratio for screening for gastric cancer in high-incidence regions other than Japan. Further studies of this test in the management of high-risk patients seem to be worthwhile.