skip to main content
OSTI.GOV title logo U.S. Department of Energy
Office of Scientific and Technical Information

Title: Random Forests to Predict Rectal Toxicity Following Prostate Cancer Radiation Therapy

Abstract

Purpose: To propose a random forest normal tissue complication probability (RF-NTCP) model to predict late rectal toxicity following prostate cancer radiation therapy, and to compare its performance to that of classic NTCP models. Methods and Materials: Clinical data and dose-volume histograms (DVH) were collected from 261 patients who received 3-dimensional conformal radiation therapy for prostate cancer with at least 5 years of follow-up. The series was split 1000 times into training and validation cohorts. A RF was trained to predict the risk of 5-year overall rectal toxicity and bleeding. Parameters of the Lyman-Kutcher-Burman (LKB) model were identified and a logistic regression model was fit. The performance of all the models was assessed by computing the area under the receiving operating characteristic curve (AUC). Results: The 5-year grade ≥2 overall rectal toxicity and grade ≥1 and grade ≥2 rectal bleeding rates were 16%, 25%, and 10%, respectively. Predictive capabilities were obtained using the RF-NTCP model for all 3 toxicity endpoints, including both the training and validation cohorts. The age and use of anticoagulants were found to be predictors of rectal bleeding. The AUC for RF-NTCP ranged from 0.66 to 0.76, depending on the toxicity endpoint. The AUC values for the LKB-NTCP were statisticallymore » significantly inferior, ranging from 0.62 to 0.69. Conclusions: The RF-NTCP model may be a useful new tool in predicting late rectal toxicity, including variables other than DVH, and thus appears as a strong competitor to classic NTCP models.« less

Authors:
 [1];  [2];  [3];  [1];  [4];  [4];  [2];  [5];  [6];  [5];  [7];  [8];  [9];  [10];
  1. LTSI, Université de Rennes 1, Rennes (France)
  2. (France)
  3. (Colombia)
  4. (China)
  5. Département de Radiothérapie, Centre Eugène Marquis, Rennes (France)
  6. Département de Radiothérapie, Institut Gustave-Roussy, Villejuif (France)
  7. Département de Radiothérapie, Centre Alexis Vautrin, Nancy (France)
  8. Département de Radiothérapie, CRLCC Henri Becquerel, Rouen (France)
  9. Département de Radiothérapie, Hôpital Henri Mondor, Créteil (France)
  10. Escuela de Estadística, Universidad Nacional de Colombia Sede Medellín, Medellín (Colombia)
Publication Date:
OSTI Identifier:
22420389
Resource Type:
Journal Article
Resource Relation:
Journal Name: International Journal of Radiation Oncology, Biology and Physics; Journal Volume: 89; Journal Issue: 5; Other Information: Copyright (c) 2014 Elsevier Science B.V., Amsterdam, The Netherlands, All rights reserved.; Country of input: International Atomic Energy Agency (IAEA)
Country of Publication:
United States
Language:
English
Subject:
62 RADIOLOGY AND NUCLEAR MEDICINE; ANTICOAGULANTS; NEOPLASMS; PATIENTS; PROSTATE; RADIATION DOSES; RADIOTHERAPY; RECTUM; TOXICITY

Citation Formats

Ospina, Juan D., INSERM, U1099, Rennes, Escuela de Estadística, Universidad Nacional de Colombia Sede Medellín, Medellín, Zhu, Jian, Laboratory of Image Science and Technology, Southeast University, Nanjing, Department of Radiation Physics, Shandong Cancer Hospital and Institute, Jinan, Centre de Recherche en Information Biomédical Sino-Français, Rennes, Chira, Ciprian, Bossi, Alberto, Delobel, Jean B., Beckendorf, Véronique, Dubray, Bernard, Lagrange, Jean-Léon, Correa, Juan C., and and others. Random Forests to Predict Rectal Toxicity Following Prostate Cancer Radiation Therapy. United States: N. p., 2014. Web. doi:10.1016/J.IJROBP.2014.04.027.
Ospina, Juan D., INSERM, U1099, Rennes, Escuela de Estadística, Universidad Nacional de Colombia Sede Medellín, Medellín, Zhu, Jian, Laboratory of Image Science and Technology, Southeast University, Nanjing, Department of Radiation Physics, Shandong Cancer Hospital and Institute, Jinan, Centre de Recherche en Information Biomédical Sino-Français, Rennes, Chira, Ciprian, Bossi, Alberto, Delobel, Jean B., Beckendorf, Véronique, Dubray, Bernard, Lagrange, Jean-Léon, Correa, Juan C., & and others. Random Forests to Predict Rectal Toxicity Following Prostate Cancer Radiation Therapy. United States. doi:10.1016/J.IJROBP.2014.04.027.
Ospina, Juan D., INSERM, U1099, Rennes, Escuela de Estadística, Universidad Nacional de Colombia Sede Medellín, Medellín, Zhu, Jian, Laboratory of Image Science and Technology, Southeast University, Nanjing, Department of Radiation Physics, Shandong Cancer Hospital and Institute, Jinan, Centre de Recherche en Information Biomédical Sino-Français, Rennes, Chira, Ciprian, Bossi, Alberto, Delobel, Jean B., Beckendorf, Véronique, Dubray, Bernard, Lagrange, Jean-Léon, Correa, Juan C., and and others. Fri . "Random Forests to Predict Rectal Toxicity Following Prostate Cancer Radiation Therapy". United States. doi:10.1016/J.IJROBP.2014.04.027.
@article{osti_22420389,
title = {Random Forests to Predict Rectal Toxicity Following Prostate Cancer Radiation Therapy},
author = {Ospina, Juan D. and INSERM, U1099, Rennes and Escuela de Estadística, Universidad Nacional de Colombia Sede Medellín, Medellín and Zhu, Jian and Laboratory of Image Science and Technology, Southeast University, Nanjing and Department of Radiation Physics, Shandong Cancer Hospital and Institute, Jinan and Centre de Recherche en Information Biomédical Sino-Français, Rennes and Chira, Ciprian and Bossi, Alberto and Delobel, Jean B. and Beckendorf, Véronique and Dubray, Bernard and Lagrange, Jean-Léon and Correa, Juan C. and and others},
abstractNote = {Purpose: To propose a random forest normal tissue complication probability (RF-NTCP) model to predict late rectal toxicity following prostate cancer radiation therapy, and to compare its performance to that of classic NTCP models. Methods and Materials: Clinical data and dose-volume histograms (DVH) were collected from 261 patients who received 3-dimensional conformal radiation therapy for prostate cancer with at least 5 years of follow-up. The series was split 1000 times into training and validation cohorts. A RF was trained to predict the risk of 5-year overall rectal toxicity and bleeding. Parameters of the Lyman-Kutcher-Burman (LKB) model were identified and a logistic regression model was fit. The performance of all the models was assessed by computing the area under the receiving operating characteristic curve (AUC). Results: The 5-year grade ≥2 overall rectal toxicity and grade ≥1 and grade ≥2 rectal bleeding rates were 16%, 25%, and 10%, respectively. Predictive capabilities were obtained using the RF-NTCP model for all 3 toxicity endpoints, including both the training and validation cohorts. The age and use of anticoagulants were found to be predictors of rectal bleeding. The AUC for RF-NTCP ranged from 0.66 to 0.76, depending on the toxicity endpoint. The AUC values for the LKB-NTCP were statistically significantly inferior, ranging from 0.62 to 0.69. Conclusions: The RF-NTCP model may be a useful new tool in predicting late rectal toxicity, including variables other than DVH, and thus appears as a strong competitor to classic NTCP models.},
doi = {10.1016/J.IJROBP.2014.04.027},
journal = {International Journal of Radiation Oncology, Biology and Physics},
number = 5,
volume = 89,
place = {United States},
year = {Fri Aug 01 00:00:00 EDT 2014},
month = {Fri Aug 01 00:00:00 EDT 2014}
}
  • Purpose: To assess the impacts of patient age and comorbid illness on rectal toxicity following external beam radiation therapy (EBRT) for prostate cancer and to assess the Qualitative Analysis of Normal Tissue Effects in the Clinic (QUANTEC) normal tissue complication probability (NTCP) model in this context. Methods and Materials: Rectal toxicity was analyzed in 718 men previously treated for prostate cancer with EBRT (≥75 Gy). Comorbid illness was scored using the Charlson Comorbidity Index (CCMI), and the NTCP was evaluated with the QUANTEC model. The influence of clinical and treatment-related parameters on rectal toxicity was assessed by Kaplan-Meier and Coxmore » proportional hazards models. Results: The cumulative incidence of rectal toxicity grade ≥2 was 9.5% and 11.6% at 3 and 5 years and 3.3% and 3.9% at 3 and 5 years for grade ≥3 toxicity, respectively. Each year of age predicted an increasing relative risk of grade ≥2 (P<.03; hazard ratio [HR], 1.04 [95% confidence interval (CI), 1.01-1.06]) and ≥3 rectal toxicity (P<.0001; HR, 1.14 [95% CI,1.07-1.22]). Increasing CCMI predicted rectal toxicity where a history of either myocardial infarction (MI) (P<.0001; HR, 5.1 [95% CI, 1.9-13.7]) or congestive heart failure (CHF) (P<.0006; HR, 5.4 [95% CI, 0.6-47.5]) predicted grade ≥3 rectal toxicity, with lesser correlation with grade ≥2 toxicity (P<.02 for MI, and P<.09 for CHF). An age comorbidity model to predict rectal toxicity was developed and confirmed in a validation cohort. The use of anticoagulants increased toxicity independent of age and comorbidity. NTCP was prognostic for grade ≥3 (P=.015) but not grade ≥2 (P=.49) toxicity. On multivariate analysis, age, MI, CHF, and an NTCP >20% all correlated with late rectal toxicity. Conclusions: Patient age and a history of MI or CHF significantly impact rectal toxicity following EBRT for the treatment of prostate cancer, even after controlling for NTCP.« less
  • We seek to identify dosimetric and anatomic indicators of late rectal toxicity in prostate cancer patients treated with intensity modulated radiation therapy (IMRT). Data from 49 patients sampled from 698 patients treated for clinically localized prostate cancer at the Memorial Sloan-Kettering Cancer Center with IMRT to a dose of 81 Gy were analyzed. The end point of the study was late Grade 2 or worse rectal toxicity within 30 months of treatment. Dosimetric analysis was performed on the rectum surface in three dimensions and on two-dimensional dose maps obtained by flattening the rectum surface using a conformal mapping procedure. Severalmore » parameters including the percentage and absolute surface area of the rectum irradiated, mean dose as a function of location on the rectum, planning target volume (PTV) size and rectum size were analyzed for correlation to toxicity. Significance was set at p<0.05 for a two-sided t-test. Correlation between absolute areas irradiated and toxicity was observed on both the rectum surface and flattened rectum. Patients with toxicity also received a significantly higher mean dose to the superior 25% of the rectum surface and 15% of the flattened rectum. PTV volume, PTV height, rectum surface area and average cross-sectional area were significantly larger in patients with toxicity. The conformal mapping procedure has potential utility for evaluating dose to the rectum and risk of toxicity. Late rectal toxicity was related to the irradiation of the upper part of the rectum and also to the absolute area irradiated, PTV size, and rectum size on the planning computed tomography (CT) scan.« less
  • Purpose: To determine if differences in patient positioning methods have an impact on the incidence and modeling of grade >=2 acute rectal toxicity in prostate cancer patients who were treated with Intensity Modulated Radiation Therapy (IMRT). Methods: We compared two databases of patients treated with radiation therapy for prostate cancer: a database of 79 patients who were treated with 7 field IMRT and daily image guided positioning based on implanted gold markers (IGRTdb), and a database of 302 patients who were treated with 5 field IMRT and daily positioning using a trans-abdominal ultrasound system (USdb). Complete planning dosimetry was availablemore » for IGRTdb patients while limited planning dosimetry, recorded at the time of planning, was available for USdb patients. We fit Lyman-Kutcher-Burman (LKB) model to IGRTdb only, and Univariate Logistic Regression (ULR) NTCP model to both databases. We perform Receiver Operating Characteristics analysis to determine the predictive power of NTCP models. Results: The incidence of grade >= 2 acute rectal toxicity in IGRTdb was 20%, while the incidence in USdb was 54%. Fits of both LKB and ULR models yielded predictive NTCP models for IGRTdb patients with Area Under the Curve (AUC) in the 0.63 – 0.67 range. Extrapolation of the ULR model from IGRTdb to planning dosimetry in USdb predicts that the incidence of acute rectal toxicity in USdb should not exceed 40%. Fits of the ULR model to the USdb do not yield predictive NTCP models and their AUC is consistent with AUC = 0.5. Conclusion: Accuracy of a patient positioning system affects clinically observed toxicity rates and the quality of NTCP models that can be derived from toxicity data. Poor correlation between planned and clinically delivered dosimetry may lead to erroneous or poorly performing NTCP models, even if the number of patients in a database is large.« less
  • Purpose: To examine a potential correlation between the in vitro apoptotic response of lymphocytes to radiation and the risk of developing late gastrointestinal (GI)/genitourinary (GU) toxicity from radiotherapy for prostate cancer. Methods and Materials: Prostate cancer patients formerly enrolled in a randomized study were tested for radiosensitivity by using a radiation-induced lymphocyte apoptosis assay. Apoptosis was measured using flow cytometry-based Annexin-FITC/7AAD and DiOC{sub 6}/7AAD assays in subpopulations of lymphocytes (total lymphocytes, CD4+, CD8+ and CD4-/CD8-) after exposure to an in vitro dose of 0, 2, 4, or 8 Gy. Results: Patients with late toxicity after radiotherapy showed lower lymphocyte apoptoticmore » responses to 8 Gy than patients who had not developed late toxicity (p = 0.01). All patients with late toxicity had apoptosis levels that were at or below the group mean. The negative predictive value in both apoptosis assays ranged from 95% to 100%, with sensitivity values of 83% to 100%. Apoptosis at lower dose points and in lymphocyte subpopulations had a weaker correlation with the occurrence of late toxicity. Conclusions: Lymphocyte apoptosis after 8 Gy of radiation has the potential to predict which patients will be spared late toxicity after radiation therapy. Further research should be performed to identify the specific subset of lymphocytes that correlates with late toxicity, followed by a corresponding prospective study.« less
  • Purpose: To update the incidence of late toxicity of RTOG 9406, a three-dimensional conformal radiation therapy (3DCRT) dose escalation trial for prostate cancer. Methods and Materials: A total of 1,084 men were registered to this Phase I/II trial of 3DCRT (eligible patients, 1,055). The dose for level I was 68.4 Gy; 73.8 Gy for level II; 79.2 Gy for level III; 74 Gy for level IV; and 78 Gy for level V. Patients in levels I to III received 1.8 Gy/fraction, and those in levels IV to V received 2.0 Gy/fraction. Disease group I patients were treated at the prostatemore » only, group 2 patients were treated at the prostate and at the seminal vesicles with a prostate boost, and group 3 patients were treated at the prostate and seminal vesicles. The median follow-up period for surviving patients was 6.1 y (level V) to 12.1 y (level I). Results: The incidence rates of RTOG grade 3 or less gastrointestinal or genitourinary toxicity were 3%, 4%, 6%, 7%, and 9% in group 1 and 6%, 2%, 6%, 9%, and 12% in group 2 at dose levels of I, II, III, IV, and V, respectively. In group 1, level V patients had a higher probability of grade2 late or greater gastrointestinal or genitourinary toxicity than those in levels I, II, and III (hazard ratio [HR] = 1.93, p = 0.0101; HR = 2.29, p = 0.0007; HR = 2.52, p = 0.0002, respectively). In group 2, dose level V patients had a higher probability of grade 2 or greater late gastrointestinal or genitourinary toxicity than those in dose levels II, III, and IV (HR = 2.61, p = 0.0002; HR = 2.22, p = 0.0051; HR = 1.60, p = 0.0276, respectively). Conclusions: Tolerance to high-dose 3DCRT remains excellent. There is significantly more grade 2 or greater toxicity with a dose of 78 Gy at 2 Gy/fraction than with 68.4 Gy to 79.2 Gy at 1.8 Gy/fraction and with 74 Gy at 2 Gy/fraction.« less