[en] Objective: Dose-volume histograms (DVHs) may be very useful tools for estimating probability of normal tissue complications (NTCP), but there is not yet an agreed upon method for their analysis. This study introduces a statistical method of aggregating and analyzing primary data from DVHs and associated outcomes, with the goal of exploring the dose-volume relationship for NTCP of the rectum, using long-term data of rectal wall bleeding following irradiation for prostate cancer. Comparison is made between these results and those generated by a mathematical model of NTCP. Materials and Methods: Previously published data were updated on 41 patients with stages T3 and T4 prostatic carcinoma treated with photons followed by perineal proton boost. DVHs of each patient's anterior rectal wall were used as well as data on the occurrence of post-irradiation rectal bleeding (minimum FU > 4 yrs). 128 separate dose-volume combinations were analyzed. For a given dose-volume combination, each patient's DVH was classified as low-risk or high-risk, and this classification was evaluated as an independent variable with treatment outcome as the dependent variable in a logistic regression. The odds-ratios derived from the regression coefficients of those combinations that tested significant (p<.05 by Wald Chi-Square) were compared. Combinations of these 'statistically significant discriminators' (SSDs) were analyzed for predictive power. These results were compared with results of the 'critical volume model', a mathematical model based on underlying biophysical assumptions. Results: 9 of 128 dose-volume combinations proved to be SSDs between bleeders ((15(41))) and non-bleeders ((26(41))), ranging contiguously between 60 Cobalt-Gy-equivalent (CGE) to 70% of the anterior rectal wall and 75 CGE to 30%. Calculated odds-ratios between low- and high-risk groups were not significantly different across the SSDs, a result of high correlation among groupings across SSDs. However, analysis combining SSDs allowed segregation of DVHs into three risk groups: low, intermediate, and high. Estimates of NTCPs based on these risk groups correlated strongly with observed data (p=0.003) and with biomathematical model-generated NTCPs. Conclusion: The data show there is a dose-volume relationship for rectal mucosal bleeding, at least between 60 and 75 CGE. The results demonstrate that using combinations of SSDs to stratify risk groups represents a useful means of analyzing empirical data as a function of hetereogeneous dose distributions. Strong correlation of these results with estimates of the biomathematical 'critical volume' model argue for their usefulness in estimating NTCP. Further clinical data is required to confirm these results. Further statistical refinement may be explored with Monte Carlo simulation techniques