Abstract
758 Background: Predicting the risk of radiation therapy (RT) induced normal tissue complications is important during RT field design. A widely employed formalism for predicting complication, though sparsely validated with clinical data, is the Kutcher normal tissue complication probability (NTCP) model. We herein examine the validity of the NTCP model, in comparison to other statistical models, as a predictor of left ventricle perfusion defects caused by tangential RT to the left-breast/chest wall. Methods: Since 1998, we have been conducting an IRB-approved prospective clinical trial in patients receiving left-sided tangential photon therapy (46–50 Gy at 1.8–2 Gy/fx) to assess RT-induced changes in cardiac function. Patients had pre-RT and 6-month post-RT SPECT (single photon emission computed tomography) cardiac perfusion scans. The datasets of 68 patients with normal pre-RT SPECT scans form the basis of this report. Of these patients, 19 had SPECT perfusion defects at 6 months. The left ventricle dose-volume histograms (DVH) of all patients were used to generate the NTCP model and two other statistical discriminants: generalized equivalent uniform dose (GEUD), and linear discriminant analysis (LDA). GEUD converted the nonuniform dose distribution into an equivalent uniform dose. LDA used a linear combination of step-wise selected DVH features to optimally separate the groups with and without defects. To compare the three models, we used receiver operating characteristic (ROC) curves that plot sensitivity and specificity as a function of discriminant value. A higher ROC curve, and hence higher area under the curve, implies a better model. Results: The areas under the curve for NTCP, GEUD, and LDA were 0.81, 0.81, and 0.91, respectively. The ROC curves for NTCP and GEUD were exactly coincident. LDA was a better predictor than both (p = 0.03). Conclusions: In the breast cancer patient population studied, the widely used NTCP model was a worse predictor of RT induced cardiac toxicity than the simpler linear discriminant analysis. (Supported by DOD award BC010663. Thanks to UNC for PLUNC software.) No significant financial relationships to disclose.
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Disclaimer: All third-party content on this website/platform is and will remain the property of their respective owners and is provided on "as is" basis without any warranties, express or implied. Use of third-party content does not indicate any affiliation, sponsorship with or endorsement by them. Any references to third-party content is to identify the corresponding services and shall be considered fair use under The CopyrightLaw.