(Sunday, 7/14/2019)
Room: ePoster Forums
Purpose: To validate model parameter estimation of normal tissue complication probability (NTCP) based on the sigmoid function for late rectal bleeding after intensity-modulated radiotherapy (IMRT) for prostate cancer.
Methods: Clinical records of 330 patients with prostate cancer who underwent IMRT at Keio University Hospital were analyzed. Patients who experienced rectal bleeding of grade ≥2 were considered bleeders. Three NTCP models (Lyman–Kutcher–Burman (LKB), logit, and Poisson) were utilized, and the model parameters were fit to the relation between the dose and complications using maximum likelihood analysis.
Results: Of the 330 patients, 19 developed late rectal bleeding. The model parameters simply estimated using the Migrad algorithm without restricting the search range were TD50 = 102.6 Gy, m = 0.27, and n = 0.11 for LKB; TD50 = 109.7 Gy, k = 4.53, and n = 0.11 for logit; and TD50 = 112.8 Gy, γ = 1.07, and n = 0.11 for Poisson. TD50 exhibited higher values compared to those in other studies. We considered that this was because most of the dose information in our study was based on the early “tail� region in the sigmoid function because the IMRT technique of radiation delivery was developed to improve normal tissue sparing. In addition, the definition of the rectum used in our study was also different from those used in other reports. In some cases, the log-likelihood values improved by changing the algorithm or restricting the search range, but TD50 tended to exhibit high values. Therefore, a further study is needed to show clinical validity.
Conclusion: We estimated the model parameters of NTCP based on the sigmoid function for late rectal bleeding. Because the data were mostly regarding low doses in the sigmoid function, the parameters derived for IMRT should be carefully validated before use in clinical practice.
Keywords
NTCP, Prostate Therapy, Intensity Modulation
Taxonomy
TH- Radiobiology(RBio)/Biology(Bio): RBio- LQ/TCP/NTCP/outcome modeling
Contact Email
