Abstract Background Gemcitabine and erlotinib have shown a survival benefit in the first-line setting in metastatic pancreatic cancer (mPC). The aim of this study was to assess whether combining capecitabine (C) with gemcitabine + erlotinib (GE) was safe and effective versus GE in patients with mPC. Patients and methods Previously untreated mPC patients were randomised to receive G (1000 mg/m2 , days 1, 8, 15) + E (100 mg/day, days 1–28) + C (1660 mg/m2 , days 1–21) or GE, q4 weeks, until progression or unacceptable toxicity. Primary end-point: progression-free survival (PFS); secondary end-points: overall survival (OS), response rate, relationship of rash with PFS/OS and safety. Results 120 patients were randomised, median age 63 years, ECOG status 0/1/2 33%/58%/8%; median follow-up 16.5 months. Median PFS in the gemcitabine–erlotinib–capecitabine (GEC) and GE arms was 4.3 and 3.8 months, respectively (hazard ratio HR: 0.88, 95% confidence interval CI: 0.58–1.31; p = 0.52). Median OS in the GEC and GE arms was 6.8 and 7.7 months, respectively (HR: 1.09, 95% CI: 0.72–1.63; p = 0.69). Grade 3/4 neutropenia (GEC 43% versus GE 15%; p = 0.0008) and mucositis (GEC 9% versus GE 0%; p = 0.03) were the only statistically significant differences in grade 3/4 adverse events. PFS and OS were significantly longer in patients with rash (grade ≥1) versus no rash (grade = 0): PFS 5.5 versus 2.0 months (HR = 0.39, 95% CI: 0.26–0.6; p < 0.0001) and OS: 9.5 versus 4.0 months (HR = 0.51, 95% CI: 0.33–0.77; p = 0.0014). Conclusion PFS with GEC was not significantly different to that with GE in patients with mPC. Skin rash strongly predicted erlotinib efficacy. The study was registered with ClinicalTrials.gov: NCT01303029.