Quantitative comparisons of summary receiver operating characteristics (sROC) curves among conventional serological tumor biomarkers for predicting gastric cancer in Chinese population

The objective of this study is to quantitatively analyze the predictive strength among the conventional serological tumor biomarkers for gastric cancer in Chinese population. Thirty-three hospital-based case-control studies were searched out through Chinese databases and PubMed during 1999-2009. Gastric cancer cases and healthy volunteers or benign gastric diseases controls were detected of any of serological CA724, CA242, CEA, CA199, CA125, or CA153. Areas under the curve (AUC) and optimal Q indexes of summary receiver operating characteristics (sROC) curves were quantitatively compared. The summary positive and negative likelihood ratios (sLR + and sLR-) were pooled. Totally, 2,390 gastric cancer cases and 2,893 controls were analyzed. CA724 and CA242 both had the greatest AUCs (0.88), respectively, followed by the combination CA724 + CEA + CA199 (0.85), CA125 (0.82), CEA (0.80), and CA199 (0.76), but all of them had no statistical significance to CA153 (negative control) by Z tests, possibly due to relatively great standard errors. The results of Q index analyses were similar to those of AUCs, that CA724 and CA242 had the optimal strength. The sLR + of CA724 (16.08, 95 % confidence interval (CI) 7.86-32.86) or CA242 (11.03, 95 % CI 7.12-17.08) was strong to judge the gastric cancer status based on its positive result. The combination of CA724 + CEA + CA199 had the prior sLR- (0.33, 95 % CI 0.25-0.43) to the others. Serological CA724 or CA242 has predictive effect for screening gastric cancer and can be recommended into the screening program of population-based or symptomatic cases. However, prospective epidemiological studies are required before confirmative conclusion.