Data envelopment analysis (DEA) is a widely used method in operations research for the benchmarking and empirical assessment of productive efficiency. We have previously applied DEA for treatment plan analysis and demonstrated its ability to determine relative plan quality, however considerations regarding the optimal use of DEA were not considered in that work. In the current work we have extended the complexity of the DEA modelling to include an increased number of measures of treatment plan quality as well investigating the best method of accounting for patient geometry. Forty-one IMRT prostate treatment plans were retrospectively analysed using an input-oriented variable returns to scale DEA method. The impacts of DEA weight restrictions were analysed with reference to the ability of DEA to differentiate plan performance at a level of clinical significance. Patient geometry significantly influences plan quality and alternative methods for considering geometry in the DEA model were investigated. In this work we identify how best to use DEA for the relative assessment of prostate treatment plan quality.