Gene expression profiling platforms were recently shown to accurately assign cancer of unknown primary (CUP) to a primary tissue of origin, with unknown impact on patient outcome. We examined chemotherapy activity and outcome in CUP trials and in metastatic solid tumour trials in order to screen for a distinct biological behaviour of CUP.An online search for autopsy or molecular platform studies on CUP indolent primaries was followed by identification of phase II or III clinical trials enrolling at least thirty patients with poor-risk CUP from 2002 or later. Chemotherapy activity and patient survival data were narratively compared to data from phase III chemotherapy trials on patients with metastatic breast, lung, pancreatic and colon cancer, to which CUP is most commonly classified by molecular profiling.Lung and pancreatic tumours were the primaries most commonly found in CUP autopsy series, whereas microarray platforms assigned CUP to breast, colon in a third and pancreatic, lung primaries in <25% of cases. 14 phase II trials managed 918 CUP patients with platinum-based chemotherapy resulting in objective response rate (ORR) of 32%. Six trials administered anthracycline-containing or gastrointestinal-type chemotherapy in 401 CUP patients, reporting ORR of 22%. The median of quoted median survival times was nine months for platinum and seven for anthracycline or GI-type regimens. Though tumour shrinkage and median survival in CUP patients were similar to those of patients with metastatic lung and pancreatic cancer, they were vastly inferior to response rates of 40-70% and median survival of 15-24 months seen in patients with metastatic breast and bowel cancer.This systematic review hints that CUP, though accurately classified by molecular methods, may harbour molecular/genetic traits distinct from tumours of known primaries. These should be sought and the impact of molecularly classified primary site-directed therapy on patient outcome prospectively validated in trials.