BACKGROUND: Several drugs are available to treat metastatic renal-cell carcinoma (MRCC), and predictive markers to identify the most adequate treatment for each patient are needed. Our objective was to identify potential predictive markers of sunitinib activity in MRCC. METHODS: We collected sequential serum samples from 31 patients treated with sunitinib. Sera of six patients with extreme phenotypes of either marked responses or clear progressions were analysed with a Human Cytokine Array which evaluates 174 cytokines before and after treatment. Variations in cytokine signal intensity were compared between both groups and the most relevant cytokines were assessed by ELISA in all the patients. RESULTS: Twenty-seven of the 174 cytokines varied significantly between both groups. Five of them (TNF-a, MMP-9, ICAM-1, BDNF and SDF-1) were assessed by ELISA in 21 evaluable patients. TNF-a and MMP-9 baseline levels were significantly increased in nonresponders and significantly associated with reduced overall survival and time-to-progression, respectively. The area under the ROC curves for TNF-a and MMP-9 as predictive markers of sunitinib activity were 0.83 and 0.77. CONCLUSION: Baseline levels of TNF-a and MMP-9 warrant further study as predictive markers of sunitinib activity in MRCC. Selection of patients with extreme phenotypes seems a valid method to identify potential predictive factors of response.