Affordable Access

Access to the full text

Dose–volume histogram parameters for predicting radiation pneumonitis using receiver operating characteristic curve

Authors
  • Wang, Dongqing1
  • Shi, Jian2
  • Liang, Shaohua3
  • Lu, Shiyong2
  • Qi, Xiangjie2
  • Wang, Qiang2
  • Zheng, Guojing2
  • Wang, Sheng2
  • Zhang, Kemin4
  • Liu, Hongfu3
  • 1 Shandong Cancer Hospital, Shandong Academy of Medical Sciences, Department of Radiation Oncology, Jinan, People’s Republic of China , Jinan (China)
  • 2 People’s Hospital of LinZi District, Affiliated to Binzhou Medical College, Department of Radiation Oncology, Zibo, People’s Republic of China , Zibo (China)
  • 3 Binzhou Medical University, 346 Guanhai Road, Laishan District, Yantai, 264003, People’s Republic of China , Yantai (China)
  • 4 Central Hospital of Binzhou, Binzhou, People’s Republic of China , Binzhou (China)
Type
Published Article
Journal
Clinical & Translational Oncology
Publisher
Springer-Verlag
Publication Date
Sep 14, 2012
Volume
15
Issue
5
Pages
364–369
Identifiers
DOI: 10.1007/s12094-012-0931-y
Source
Springer Nature
Keywords
License
Yellow

Abstract

ObjectiveTo assess the predictability of dose–volume histogram (DVH) parameters for radiation pneumonitis (RP) using receiver operating characteristic (ROC) curve.MethodsOne hundred and thirty-five cases of locally advanced non-small cell lung cancer patients treated with three-dimensional radiotherapy and chemotherapy were analyzed retrospectively. The end point of follow-up was ≥2 grade RP defined according to the National Cancer Institute Common Terminology Criteria for Adverse Events, version 3.0. The ROC curve was used to explore the predictive sensitivity (SEN), specificity (SPE), positive predictive value (PPV), and negative predictive value (NPV) for potential DVH parameters associated with RP.ResultsRelative volumes of total lungs receiving ≥5 Gy (V5), ≥10 Gy (V10), ≥13 Gy (V13), ≥20 Gy (V20), and mean lung dose (MLD), were all correlated to the development of RP (p < 0.05), among which V5 and V20 were the most important factors (p = 0.045 and 0.037; OR = 3.166 and 3.030). However, collinearity was found between V5 and V20 (Spearman’s rho 0.771, p < 0.01). The area under the ROC curve was 0.643 and 0.648 for using V5 and V20 as predictors. If predictive cut-off values were established as follows: V5 = 0.8 and V20 = 0.3, the parameters could provide predictive SEN, SPE, PPV and NPV were 0.387 and 0.581, 0.882 and 0.701, 0.444 and 0.321, and 0.855 and 0.873, respectively.ConclusionsV5 and V20 could act as predictors for RP; however, single DVH metrics did not appear to have high predictive power for RP.

Report this publication

Statistics

Seen <100 times