Irinotecan 65mg/m² D1 and D8 q21 days plus Cisplatin 60mg/m² D1 q 21 days, until disease progression or unacceptable toxicity, with standard hydration and antiemetics.

Gemcitabine 1000mg/m² D1 and D8 every 21 days plus cisplatin 25mg/m² D1 and D8 every 21 days, until disease progression or unacceptable toxicity, with standard hydration and antiemetics.

Irinotecan 65mg/m² D1 and D8 q21 days, until disease progression or unacceptable toxicity.

Given in association with standard hydration and anti-emetics.

The overall response rate will measure the number of subjects with complete or partial response as best response during the entire treatment, over the total number of subjects, for each arm. The response will be evaluated according to RECIST 1.1.

We will measure the number of subjects without progressive disease (complete response, partial response or stable disease) or any-cause death, whichever came first. For each patient, PFS will be calculated from randomization until progressive disease. Disease progression means radiologic progression per RECIST 1.1 or any-cause death. PFS will also be analysed with log-rank test, and reported as HR with respective 95% CI and p value, and median PFS will be estimated with kaplan-meyer method. All calculations will be performed 6 months after the last patient recruited. Also, the PFS rate at one and two years will be calculated.

The following biomarkers will be tested by immunohistochemistry: human equilibrative nucleoside transporter (hENT), X-ray repair cross-complementing protein 1 (XRCC1), Excision repair cross-complementing rodent repair deficiency, complementation group 1 (ERCC1) e mLH1. Then we will correlate them with efficacy end-points using non-parametric test.