We demonstrate a method for differentiating tissue disease states using the intrinsic texture properties of speckle in optical coherence tomography (OCT) images of normal and tumor tissues obtained in vivo. This approach fits a gamma distribution function to the nonlog-compressed OCT image intensities, thus allowing differentiation of normal and tumor tissues in an ME-180 human cervical cancer mouse xenograft model. Quantitative speckle intensity distribution analysis thus shows promise for identifying tissue pathologies, with potential for early cancer detection in vivo.