BackgroundStudies comparing upfront surgery with neoadjuvant treatment in pancreatic cancer may report only patients who underwent resection and so survival will be skewed. The aim of this study was to report survival by intention to treat in a comparison of upfront surgery versus neoadjuvant treatment in resectable or borderline resectable pancreatic cancer.MethodsMEDLINE, Embase and the Cochrane Library were searched for studies reporting median overall survival by intention to treat in patients with resectable or borderline resectable pancreatic cancer treated with or without neoadjuvant treatment. Secondary outcomes included overall and R0 resection rate, pathological lymph node rate, reasons for unresectability and toxicity of neoadjuvant treatment.ResultsIn total, 38 studies were included with 3484 patients, of whom 1738 (49·9 per cent) had neoadjuvant treatment. The weighted median overall survival by intention to treat was 18·8 months for neoadjuvant treatment and 14·8 months for upfront surgery; the difference was larger among patients whose tumours were resected (26·1 versus 15·0 months respectively). The overall resection rate was lower with neoadjuvant treatment than with upfront surgery (66·0 versus 81·3 per cent; P < 0·001), but the R0 rate was higher (86·8 (95 per cent c.i. 84·6 to 88·7) versus 66·9 (64·2 to 69·6) per cent; P < 0·001). Reported by intention to treat, the R0 rates were 58·0 and 54·9 per cent respectively (P = 0·088). The pathological lymph node rate was 43·8 per cent after neoadjuvant therapy and 64·8 per cent in the upfront surgery group (P < 0·001). Toxicity of at least grade III was reported in up to 64 per cent of the patients.ConclusionNeoadjuvant treatment appears to improve overall survival by intention to treat, despite lower overall resection rates for resectable or borderline resectable pancreatic cancer.PROSPERO registration number: CRD42016049374.