Comparison of rectal dose-wall histogram versus dose-volume histogram for modeling the incidence of late rectal bleeding after radiotherapy.

PURPOSE To compare the fits of normal-tissue complication probability (NTCP) models based on rectal dose-wall histograms (DWHs) vs. dose-volume histograms (DVHs) when the two are used to analyze a common set of late rectal toxicity data. METHODS AND MATERIALS Data were analyzed from 128 prostate cancer patients treated with 3-dimensional conformal radiotherapy (3D-CRT) at The University of Texas M.D. Anderson Cancer Center (UTMDACC). The DVH for total rectal volume, including contents, was obtained for each patient from the treatment-planning system. A DWH was also computed, using the outer rectal contour plus an autogenerated inner contour that corresponds to an assumed 3-mm rectal wall thickness. The endpoint for analysis was Grade 2 or higher late rectal bleeding within 2 years of treatment; all patients had at least 2 years of follow-up. Four different NTCP models were fitted to the response data by using either the DVH or the DWH to describe the dose distribution to rectum or rectal wall, respectively. The 4 models considered were the Lyman model, the mean dose model, the parallel-architecture model, and a model based on the volume of a organ receiving more than a specified dose (the "cutoff-dose" model). RESULTS For each of the models, the fit to the late rectal bleeding data was slightly improved when the analysis was based on the rectal DWH instead of on the DVH. In addition, the results of the cutoff dose and parallel architecture models were consistent with one another for the DWH data but not for the DVH data. For the DWH data, both models predict a 50% or higher incidence of Grade 2 or worse late rectal bleeding within 2 years if 80% or more of the rectal wall is exposed to doses greater than 32 Gy. A 50% or higher incidence of rectal bleeding is also predicted if the mean dose to rectal wall exceeds 53.2 Gy. CONCLUSIONS A consistent, although modest, improvement occurs in the fits of NTCP models to the UTMDACC 2-year late rectal bleeding data when the fit is based on the rectal dose-wall histogram instead of on the dose-volume histogram for entire rectum, including contents.