The therapeutic activity of anticancer agents depends critically on their ability to penetrate through tumour tissue to reach their target cells, a requirement that is especially important for hypoxia-activated prodrugs. Here we use multicellular layers (MCL) grown in vitro from HT29 colon carcinoma cells to measure tissue diffusion coefficients (D_mcl) of 67 structurally diverse benzotriazine di-N-oxides (analogues of the hypoxia-activated prodrug tirapazamine) plus four miscellaneous compounds. An algorithm was developed to predict D_mcl from physicochemical parameters (molecular weight, octanol/water partition coefficient at pH 7.4, number of hydrogen bond donors and acceptors); the fitted multivariate relationship had an explained variance (R²) of 0.907 and predictive power (Q²) of 0.879. Using a subset of nine compounds tested as a single cassette, the algorithm was shown to apply, with some adjustment of coefficients, to MCLs from three other tumour cell lines with differing cell packing densities (SiHa, HCT8-Ea, and HCT8-Ra). The demonstrated relationships provide tools for optimizing extravascular transport of anticancer agents during lead optimization.