Background and purpose:Mir-21 has been demonstrated high expressed in many kinds of tumor tissues and cell lines. High expressed miR-21 leads to chemoresistance. Circulating miRNA is a novel biomarker for chemosensitivity prediction. The aim of our study was to investigate the role of plasma and tumor miR-21 levels as predictive biomarkers for irinotecan in gastric cancer. Methods: The histoculture drug response assay (HDRA) was used to determine irinotecan and cisplatin sensitivity on 35 freshly removed gastric tumor specimens. miR-21 expressions in tumor and plasma were determined by quantitative reverse transcription polymerase chain reaction. Results:Plasma miR-21 was closely correlated with corresponding miR-21 mRNA level in tumor tissues (rho=0.736, P<0.001) and was not correlated with sex, age, pathology type, differentiation, lymph node metastasis, TNM stage or tumor location. Patients with stage Ⅱ-Ⅲhad higher miR-21 levels. The tumor miR-21 expressions in cisplatin-sensitive group and resistant group were 1.32 (95%CI:0.73-1.90) and 4.06 (95%CI:1.71-6.41)(P=0.004), respectively. The plasma miR-21 expressions in cisplatin-sensitive group and resistant group were 5.25 (95%CI:0.14-10.64) and 5.82 (95%CI: 2.27-9.37)(P=0.19). The tumor miR-21 expressions in irinotecan-sensitive group and resistant group were 1.09 (95%CI:0.65-1.54) and 4.94 (95%CI:2.44-7.44)(P<0.001), respectively. The plasma miR-21 expressions in irinotecan-sensitive group and resistant group were 1.86 (95%CI:1.08-2.64) and 12.42 (95%CI:3.14-21.70)(P=0.001). Conclusion:A signiifcant correlation was observed between plasma and tumor miR-21 level. The low plasma miR-21 level could benefit from treatment with irinotecan. And low tumor miR-21 expression correlated with increased irinotecan and cisplatin response rate.