Inter-criteria reproducibility of RECIST, WHO, Japanese response criteria for gastric cancer or esophageal cancer: From the response dataset of Japan Clinical Oncology Group (JCOG) phase III trials

Abstract

4065 Background: For assessments of tumor response in upper gastrointestinal cancers, Japanese criteria (JPN), which needs X-ray or endoscopic findings for primary lesion, has been generally used in Japan instead of international RECIST or WHO criteria. To compare the inter-criteria reproducibility of these radiological assessments, we analyzed the datasets of tumor responses from two phase III trials. Methods: A total of 115 gastric cancer patients in two treatment arms of JCOG9205, 5-FU and 5-FU plus cisplatin, and 103 esophageal cancer patients in a neoadjuvant 5-FU plus cisplatin arm of JCOG9907 were eligible for this analysis. Overall response rates (ORR) were calculated according to each of three criteria. The tumor responses, furthermore, were divided into four categories (CR, PR, SD, PD), following estimation of kappa statistics between two criteria as agreement measure. Results: As shown in the Table , ORR according to RECIST and WHO were very similar in both cancers. JPN tended to estimate higher ORR comparing to RECIST or WHO, particularly in esophageal cancer. In gastric cancer, kappa for concordance to RECIST were 0.92 for WHO and 0.86 for JPN. While in esophageal cancer, kappa were 0.72 for WHO and 0.40 for JPN. Conclusions: In these trials for gastric cancer and esophageal cancer, there was fairly consistent reproducibility between unidimensional RECIST and bidimensional WHO criteria. JPN criteria, depending on assessment of primary lesion, may give higher ORR particularly in esophageal cancer, and cautious interpretation may be necessary when evaluating ORR from Japanese trials using JPN criteria (Table). Overall response rates of respective treatment arms calculated according to RECIST, WHO and JPN criteria. [Table: see text] No significant financial relationships to disclose.