Accumulating evidence has identified a mechanism potentially responsible for the inactivation of tumor suppressor genes, namely transcriptional silencing by aberrant methylation of CpG islands. A previous study has shown the loss of RUNX3 expression, due to aberrant methylation of its CpG island, in gastric cancer cell lines, suggesting that RUNX3 is a target for epigenetic gene silencing in gastric carcinogenesis. However, there are limited data on the methylation status of RUNX3 in the neoplastic and non-neoplastic tissues in various types of human cancers, including gastric cancer. Here, we report that 60% of gastric cancer cell lines and 64% of primary gastric carcinomas (n=75) were methylated at the RUNX3 CpG island. RUNX3 methylation was also detected in hepatocellular carcinomas (73%, n=48), larynx cancers (62%, n=37), lung cancers (46%, n=24), breast cancers (25%, n=25), prostate cancers (23%, n=44), endometrial cancers (12.5%, n=24), colon cancers (4.9%, n=61) and uterine cervical cancers (2.5%, n=40), showing that RUNX3 methylation is not restricted to gastric cancer. Interestingly, the RUNX3 methylation was especially frequent in tumors from tissues of a foregut derivative, that is, the stomach, liver, larynx and lung. Next, the methylation status of RUNX3 in various non-neoplastic tissues was examined, including the premalignant lesions of gastric carcinomas. The RUNX3 methylation was found in 8.1% of chronic gastritis (n=99), 28.1% of intestinal metaplasia (n=32), 27.3% of gastric adenomas (n=77) and 64% of gastric carcinomas (n=75), but not in chronic hepatitis B, normal prostate and colon mucosa, even though in cases of chronic hepatitis, the methylation frequency of its neoplastic tissues was very high. In conclusion, RUNX3 methylation is frequently found in human cancers, including gastric cancer, and is mostly cancer specific, with the exception of the stomach, and thus, might be useful as a potential diagnostic biomarker of cancer.