Two non-parametric methods for derivation of constraints from radiotherapy dose-histogram data.

Area of Science:

• Radiation Oncology
• Medical Physics
• Biostatistics

Background:

• Dose constraints derived from dose-volume histograms (DVHs) are crucial for radiotherapy planning.
• Current methods for deriving these constraints are often inadequately described.
• Optimizing dose constraints requires identifying dose-specific cut-points that correlate with complication incidence.

Purpose of the Study:

• To investigate and compare two non-parametric methods for deriving radiotherapy dose constraints.
• To determine optimal dose-specific cut-points using rectal toxicity data from a prostate radiotherapy trial.
• To evaluate the performance of receiver operating characteristic (ROC) analysis and a maximally-selected standardized rank sum method.

Main Methods:

• Comparison of ROC analysis and rank sum methods for deriving dose constraints.
• Utilized rectal toxicity data from a prostate radiotherapy trial.
• Applied multiple test corrections using a free step-down resampling algorithm to identify optimal cut-points.

Main Results:

• Both ROC and rank sum methods yielded consistent and significant dose-specific cut-point values.
• The rank sum method showed some sensitivity to the specific data used.
• The ROC method is easily implemented and can incorporate complication atlases.

Conclusions:

• Both investigated methods effectively identify significant dose-specific cut-points for radiotherapy planning.
• The rank sum method offers an advantage by accommodating all complication grades without dichotomization.
• These findings contribute to more refined and data-driven radiotherapy treatment planning.