Receiver operating characteristic analysis of prediction for gastric cancer development using serum pepsinogen and Helicobacter pylori antibody tests

Abstract

Background: Chronic Helicobacter pylori infection plays a central role in the development of gastric cancer as shown by biological and epidemiological studies. The H. pylori antibody and serum pepsinogen (PG) tests have been anticipated to predict gastric cancer development.

Methods: We determined the predictive sensitivity and specificity of gastric cancer development using these tests. Receiver operating characteristic analysis was performed, and areas under the curve were estimated. The predictive sensitivity and specificity of gastric cancer development were compared among single tests and combined methods using serum pepsinogen and H. pylori antibody tests.

Results: From a large-scale population-based cohort of over 100,000 subjects followed between 1990 and 2004, 497 gastric cancer subjects and 497 matched healthy controls were chosen. The predictive sensitivity and specificity were low in all single tests and combination methods. The highest predictive sensitivity and specificity were obtained for the serum PG I/II ratio. The optimal PG I/II cut-off values were 2.5 and 3.0. At a PG I/II cut-off value of 3.0, the sensitivity was 86.9% and the specificity was 39.8%. Even if three biomarkers were combined, the sensitivity was 97.2% and the specificity was 21.1% when the cut-off values were 3.0 for PG I/II, 70 ng/mL for PG I, and 10.0 U/mL for H. pylori antibody.

Conclusions: The predictive accuracy of gastric cancer development was low with the serum pepsinogen and H. pylori antibody tests even if these tests were combined. To adopt these biomarkers for gastric cancer screening, a high specificity is required. When these tests are adopted for gastric cancer screening, they should be carefully interpreted with a clear understanding of their limitations.

Keywords: Cancer screening; Gastric cancer; Helicobacter pylori antibody; Receiver operating characteristic analysis; Serum pepsinogen.

Fig. 1

Flow chart for target group selection. In a large-scale cohort study conducted in Japan (i.e., Japan Public Health Center Study [JPHC study]), a population-based cohort of 61,009 men and 62,567 women was identified and followed from January 1, 1990 to December 31, 2004. Of 97,644 eligible subjects in the JPHC study cohort, 13,467 men and 23,278 women who donated blood samples at baseline were included. Cases of gastric cancer were extracted from the cancer registry for the JPHC study based on site. Among 1,681 cases with a histologically proven diagnosis made from 1990 to 2004, plasma at baseline was obtained from 512 cases. One case with a technical error of H. pylori antibody measurement and the matched control were excluded. The dataset included 511 case subjects and 511 matched control subjects. For this analysis, 14 case subjects who were diagnosed as having gastric cancer before the blood donation date and 14 matched control subjects were excluded. Thus, a total of 994 subjects (497 case subjects having gastric cancer and 497 healthy control subjects) were used for the ROC analysis in this study

Fig. 2

Comparison of AUCs for gastric cancer development among single tests using serum PG status and H. pylori antibody. The AUCs were compared among PG I, PG II, PG I/II, and H. pylori antibody titer. The AUCs of the following tests were compared with that of PG I/II (AUC = 0.649 ± 0.017, 95% CI: 0.615-0.683): a, the AUC of PG I (0.561 ± 0.011, 95% CI: 0.526-0.597) was significantly lower than that of PG I/II (p < 0.001); b, the AUC of PG II (0.434 ± 0.018, 95% CI: 0.400-0.469) was significantly lower than that of PG I/II (p < 0.001); c, the AUC of H. pylori antibody (0.574 ± 0.018, 95% CI:0.538- 0.610) was significantly lower than that of PG I/II (p < 0.001)

Fig. 3

Comparison of AUCs for gastric cancer development among PG I/II, and combined tests using serum PG status and H. pylori antibody. Compared with PG I/II, the AUCs of gastric cancer using the combined method with the standard cut-off values (PG I/II = 3.0, PG I = 70.0 ng/mL, and H. pylori antibody = 10.0 U/mL) were nearly equal