PURPOSE: nab-Paclitaxel plus gemcitabine was superior to gemcitabine alone for patients with metastatic pancreatic cancer (MPC) in the phase III MPACT trial. This study evaluated the association of secreted protein acidic and rich in cysteine (SPARC) levels with efficacy as an exploratory endpoint. EXPERIMENTAL DESIGN: Patients with previously untreated MPC (N = 861) received nab-paclitaxel plus gemcitabine or gemcitabine alone. Baseline SPARC level was measured in the tumor stroma and epithelia (archival biopsies) and plasma. Experiments were performed in pancreatic cancer mouse models in which SPARC was intact or deleted. RESULTS: SPARC was measured in the tumor stroma of 256 patients (30%), the tumor epithelia of 301 patients (35%), and plasma of 343 patients (40%). Stroma-evaluable samples were from metastases (71%), from the pancreas (11%), or of unidentifiable origin (insufficient tissue to determine; 17%). For all patients, stromal SPARC level [high (n = 71) vs. low (n = 185)] was not associated with overall survival (OS; HR, 1.019; P = 0.903); multivariate analysis confirmed this lack of association. There was no association between stromal SPARC level and OS in either treatment arm. Neither tumor epithelial SPARC nor plasma SPARC was associated with OS. Results from a SPARC knockout mouse model treated with nab-paclitaxel plus gemcitabine revealed no correlation between SPARC expression and tumor progression or treatment efficacy. CONCLUSIONS: SPARC levels were not associated with efficacy in patients with MPC. This exploratory analysis does not support making treatment decisions regarding nab-paclitaxel plus gemcitabine or gemcitabine alone in MPC based on SPARC expression.