Abstract Background CUP are heterogeneous tumors that share the unique characteristic of metastases with no identifiable origin. The outcome of patients (pts) with CUP is poor despite empiric chemotherapy that has activity against a wide variety of neoplasms such as the cisplatin-gemcitabine combination (Culine S, JCO 2002). Molecular tests may identify primary sites in up to 80% of pts, and results suggest that at least 1/3 of identified primaries may not be sensitive to empiric chemotherapy used in CUPs (Gross-Goupil G 2012). In the GEFCAPI 04 phase III trial, we hypothesized that tailored treatment will improve outcomes. Methods Eligible pts had pathologically-confirmed metastatic CUPs and were treatment naive. Pts belonging to pre-defined favorable subsets were excluded. After relevant workup had identified no primary site, pts were randomized 1:1 to either Arm A (Cisplatin 100 mg/m² d1+ Gemcitabine 1250 mg/m², day 1 and 8, q3w) or Arm B (gene expression test followed by a la carte treatment according to the suspected primary). The test consisted of the Tissue Of Origin (Pathwork, n = 21) or CancerTYPE ID (Biotheranostics, n = 222). The primary endpoint was PFS (HR = 0.625, power=80%, 5% bilateral test). Stratification was on site, PS and LDH level. Secondary endpoints were PFS in pts with pre-defined cancers likely insensitive to cisplatin-gemcitabine and OS. Results From 03/12 to 02/18, 243 pts from 4 EU countries were randomized (Arm A: 120, Arm B: 123). Primary cancers most often reported by tests were pancreatico-biliary cancer (19%), squamous cell carcinoma (11%, kidney cancer (8%), and lung cancer (8%). Treatment was tailored by molecular test results in 91/123 arm B pts (74%). PFS by central review was similar: HR = 0.95 (0.72-1.25); p = 0.7; medians: 5.3 m arm A vs 4.6 m arm B. PFS by local review also showed no significant difference: HR = 0.80 (0.60-1.06); p = 0.12; medians 5.8 vs 6.4 m. OS was also similar in the overall population (HR: 0.92 (0.69-1.23), medians: 10 vs 10.7 m) and in 60 pts with suspected cancers likely insensitive to GC. Conclusions In GEFCAPI 04, using a molecular test followed by tailored systemic treatment did not improve outcomes of pts with CUP. Clinical trial identification 2011-A01202-39. Legal entity responsible for the study Institut Gustave Roussy. Funding Programme Hospitalier de Recherche Clinique (PHRC) from the French Ministry of Health. Disclosure K. Fizazi: Advisory / Consultancy: Astellas; Advisory / Consultancy: AAA; Advisory / Consultancy: Bayer; Advisory / Consultancy: Clovis; Advisory / Consultancy: Curevac; Advisory / Consultancy: Incyte; Advisory / Consultancy: Janssen; Advisory / Consultancy: MSD; Advisory / Consultancy: Orion; Advisory / Consultancy: Sanofi. R. Morales Barrera: Honoraria (self), Advisory / Consultancy, Speaker Bureau / Expert testimony: MSD; Honoraria (self), Advisory / Consultancy, Speaker Bureau / Expert testimony: Sanofi; Honoraria (self), Advisory / Consultancy, Speaker Bureau / Expert testimony: Astrazeneca; Honoraria (self), Advisory / Consultancy, Speaker Bureau / Expert testimony: Asofarma; Honoraria (self), Advisory / Consultancy, Speaker Bureau / Expert testimony: Johnson and Jonhson; Honoraria (self), Advisory / Consultancy: Roche. C.A. Schnabel: Full / Part-time employment: bioTheranostics. All other authors have declared no conflicts of interest.