Although gene expression profiling is a promising diagnostic technique to determine the tissue of origin for patients with cancer of unknown primary site (CUP), no clinical trial has evaluated yet site-specific therapy directed by this approach compared with empirical chemotherapy. We therefore performed a randomized study to assess whether such site-specific therapy improves outcome compared with empirical chemotherapy in previously untreated patients with CUP.

Comprehensive gene expression profiling was performed by microarray analysis, and an established algorithm was applied to predict tumor origin. Patients with CUP were randomly assigned (1:1) to receive standard site-specific therapy or empirical paclitaxel and carboplatin (PC). The primary end point was 1-year survival rate.

One hundred thirty patients were randomly assigned and had sufficient biopsy tissue for molecular analysis. Efficacy analysis was performed for 50 and 51 patients in the site-specific therapy and empirical PC arms, respectively. Cancer types most commonly predicted were pancreatic (21%), gastric (21%), and lymphoma (20%). The 1-year survival rate was 44.0% and 54.9% for site-specific treatment and empirical PC ( P = .264), respectively. Median overall and progression-free survival were 9.8 and 5.1 months, respectively, for site-specific treatment versus 12.5 and 4.8 months for empirical PC ( P = .896 and .550, respectively). Median overall survival (16.7 v 10.6 months; P = .116) and progression-free survival (5.5 v 3.9 months; P = .018) were better for predicted more-responsive than less-responsive tumor types.

Site-specific treatment that was based on microarray profiling did not result in a significant improvement in 1-year survival compared with empirical PC, although prediction of the original site seemed to be of prognostic value.