Four RCTs (n=2,115) were included in the review. Sample sizes ranged from 150 to 1,029 patients. Three of the included trials were published as abstracts and one trial was published in a peer-reviewed journal. The overall risk of bias was unclear in the fully reported trial. The abstracts appeared to meet some of the inclusion criteria but none of them reported allocation concealment, blinding or source of funding.
There were benefits observed in overall survival with S-1-based group to with the 5-FU-based therapy (HR 0.87, 95% CI 0.79 to 0.96, three trials Ι²=0%), although there were no significant differences between treatment groups in overall response rate (OR 1.25, 95% CI 0.31 to 5.09, two trials; random-effects model). For this latter finding there was significant statistical heterogeneity observed across the trials (Ι²=90.5%).
Grade 3 to 4 neutropenia (three trials) was more likely to occur in patients treated with 5-FU-based therapy compared with the patients treated with S-1-based therapy, although statistically significant heterogeneity was observed, (OR 0.37, 95% CI 0.29 to 0.48, Ι²=70.2%). With the omission of one trial, heterogeneity was no longer statistically significant. There were no statistically significant differences observed between the treatment groups in grade 3-4 anaemia, leucopenia, nausea, diarrhoea in non-haematological toxicity (Ι²=88.9%), stomatitis (Ι²=90.8%) and treatment-related deaths.
The results of the Begg and Egger tests did not show any evidence of publication bias.