Validation of normal tissue complication probability predictions in individual patient: late rectal toxicity. Int J Radiat Oncol Biol Phys 2013 Mar 15;85(4):1103-9
Date
10/02/2012Pubmed ID
23021435DOI
10.1016/j.ijrobp.2012.07.2375Scopus ID
2-s2.0-84875231583 (requires institutional sign-in at Scopus site) 5 CitationsAbstract
PURPOSE: To perform validation of risk predictions for late rectal toxicity (LRT) in prostate cancer obtained using a new approach to synthesize published normal tissue complication data.
METHODS AND MATERIALS: A published study survey was performed to identify the dose-response relationships for LRT derived from nonoverlapping patient populations. To avoid mixing models based on different symptoms, the emphasis was placed on rectal bleeding. The selected models were used to compute the risk estimates of grade 2+ and grade 3+ LRT for an independent validation cohort composed of 269 prostate cancer patients with known toxicity outcomes. Risk estimates from single studies were combined to produce consolidated risk estimates. An agreement between the actuarial toxicity incidence 3 years after radiation therapy completion and single-study or consolidated risk estimates was evaluated using the concordance correlation coefficient. Goodness of fit for the consolidated risk estimates was assessed using the Hosmer-Lemeshow test.
RESULTS: A total of 16 studies of grade 2+ and 5 studies of grade 3+ LRT met the inclusion criteria. The consolidated risk estimates of grade 2+ and 3+ LRT were constructed using 3 studies each. For grade 2+ LRT, the concordance correlation coefficient for the consolidated risk estimates was 0.537 compared with 0.431 for the best-fit single study. For grade 3+ LRT, the concordance correlation coefficient for the consolidated risk estimates was 0.477 compared with 0.448 for the best-fit single study. No evidence was found for a lack of fit for the consolidated risk estimates using the Hosmer-Lemeshow test (P=.531 and P=.397 for grade 2+ and 3+ LRT, respectively).
CONCLUSIONS: In a large cohort of prostate cancer patients, selected sets of consolidated risk estimates were found to be more accurate predictors of LRT than risk estimates derived from any single study.
Author List
Semenenko VA, Tarima SS, Devisetty K, Pelizzari CA, Liauw SLAuthor
Sergey S. Tarima PhD Associate Professor in the Institute for Health and Equity department at Medical College of WisconsinMESH terms used to index this publication - Major topics in bold
Dose-Response Relationship, RadiationGastrointestinal Hemorrhage
Humans
Male
Models, Statistical
Organs at Risk
Probability
Prostatic Neoplasms
Radiation Injuries
Rectal Diseases
Rectum
Risk Assessment
