ORIGINAL ARTICLE

Imaging Tumor Response and Tumoral Heterogeneity in Non–Small Cell Lung Cancer Treated With Antiangiogenic Therapy Comparison of the Prognostic Ability of RECIST 1.1, an Alternate Method (Crabb), and Image Heterogeneity Analysis Connie Yip, FRCR,*w Nunzia Tacelli, MD,zy Martine Remy-Jardin, PhD,zy Arnaud Scherpereel, PhD,y8 Alexis Cortot, PhD,y8 Jean-Jacques Lafitte, PhD,y8 Frederic Wallyn, MD,y8 Jacques Remy, PhD,zy Paul Bassett, MSc,z Musib Siddique, PhD,* Gary J.R. Cook, FRCR,* David B. Landau, FRCR,*# and Vicky Goh, FRCR***

Purpose: We aimed to assess computed tomography (CT) intratumoral heterogeneity changes, and compared the prognostic ability of the Response Evaluation Criteria in Solid Tumors (RECIST) 1.1, an alternate response method (Crabb), and CT heterogeneity in non–small cell lung cancer treated with chemotherapy with and without bevacizumab. Materials and Methods: Forty patients treated with chemotherapy (group C) or chemotherapy and bevacizumab (group BC) underwent contrast-enhanced CT at baseline and after 1, 3, and 6 cycles of chemotherapy. Radiologic response was assessed using RECIST 1.1 and an alternate method. CT heterogeneity analysis generating global and locoregional parameters depicting tumor image spatial intensity characteristics was performed. Heterogeneity parameters between the 2 groups were compared using the Mann-Whitney U test. Associations between heterogeneity parameters and radiologic response with overall survival were assessed using Cox regression. Results: Global and locoregional heterogeneity parameters changed with treatment, with increased tumor heterogeneity in group BC.

From the *Division of Imaging Sciences and Biomedical Engineering, King’s College London; Departments of #Clinical Oncology; **Radiology, Guy’s & St Thomas’ NHS Foundation Trust, London; zStatsconsultancy Ltd, Buckinghamshire, United Kingdom; wDepartment of Radiation Oncology, National Cancer Centre, Singapore, Singapore; zDepartment of Thoracic Imaging, Hospital Calmette; yFaculty of Medicine, Henri Warembourg; and 8Department of Pulmonary and Thoracic Oncology, University of Lille Nord de France, Lille, France. Supported financially by the Department of Health via the National Institute of Health Research Biomedical Research Centre award to Guy’s and St Thomas’ NHS Foundation Trust in partnership with King’s College London and King’s College Hospital NHS Foundation Trust; and from the Comprehensive Cancer Imaging Centre, funded by the Cancer Research UK and Engineering and Physical Sciences Research Council in association with the Medical Research Council and Department of Health. Connie Yip receives funding support from the National Medical Research Council, Singapore. The authors declare no conflicts of interest. Correspondence to: Connie Yip, FRCR, Imaging 2, Level 1, Lambeth Wing, St Thomas’ Hospital, Lambeth Palace Road, London SE1 7EH, United Kingdom (e-mail: [email protected]). Supplemental Digital Content is available for this article. Direct URL citations appear in the printed text and are provided in the HTML and PDF versions of this article on the journal’s Website, (www.thoracicimaging.com). Copyright r 2015 Wolters Kluwer Health, Inc. All rights reserved. DOI: 10.1097/RTI.0000000000000164

Entropy [group C: median 0.2% (interquartile range  2.2, 1.7) vs. group BC: 0.7% ( 0.7, 3.5), P = 0.10] and busyness [ 27.7% (62.2, 5.0) vs. 11.5% ( 29.1, 92.4), P = 0.10] showed a greater reduction in group C, whereas uniformity [1.9% ( 8.0, 9.8) vs. 5.0% (13.9, 5.6), P = 0.10] showed a relative increase after 1 cycle but did not reach statistical significance. Two (9%) and 1 (6%) additional responders were identified using the alternate method compared with RECIST in group C and group BC, respectively. Heterogeneity parameters were not significant prognostic factors. Conclusions: The alternate response method described by Crabb identified more responders compared with RECIST. However, both criteria and baseline imaging heterogeneity parameters were not prognostic of survival. Key Words: non–small cell lung cancer, bevacizumab, Response Evaluation Criteria in Solid Tumors, response, heterogeneity

(J Thorac Imaging 2015;30:300–307)

L

ung cancer is the most common cancer worldwide, of which the majority are non–small cell lung cancers (NSCLCs).1 Survival in NSCLC remains poor, with 5-year overall survival (OS) approaching 73% and 13% in early and advanced stages, respectively.2 The introduction of targeted therapies has changed the management of advanced nonsquamous NSCLC.3–7 These new agents have included antiangiogenic therapy such as bevacizumab, a monoclonal antibody against the vascular endothelial growth factor (VEGF). The combination of cytostatic antiangiogenic therapy and cytotoxic platinum-based conventional chemotherapy has been shown to improve outcomes in NSCLC.6–8 The optimal method to assess treatment response after antiangiogenic therapy remains in question. The Response Evaluation Criteria in Solid Tumors (RECIST) based on changes in tumor size is well established in the clinical trial setting to evaluate treatment response after cytotoxic chemotherapy9 but is less successful in the response assessment of targeted therapy.10 Antiangiogenic agents have been shown to induce tumor necrosis with subsequent cavitation, which may increase the actual size of the lesion.11,12 Alternative assessment criteria have been proposed to account for the different morphologic changes that occur

300 | www.thoracicimaging.com J Thorac Imaging  Volume 30, Number 5, September 2015 Copyright r 2015 Wolters Kluwer Health, Inc. All rights reserved.

J Thorac Imaging



Volume 30, Number 5, September 2015

during treatment with targeted therapy such as tumor cavitation13 and enhancement characteristics,14,15 but remain in restricted use in specialist settings. Intratumoral genetic and biological heterogeneity are increasingly recognized as a cause of differential treatment response within and between patients, as well as treatment resistance.16–18 This may be reflected by noninvasive imaging.19 Imaging assessment of intratumoral heterogeneity may be associated with biological heterogeneity20 and has the potential to provide additional prognostic and predictive information.21–24 Image heterogeneity analysis evaluates the intensity and spatial distribution of pixels within a diagnostic image.25 This can be measured using different approaches. The more commonly applied statistical methods may generate global first-order and locoregional second-order and higher-order parameters. Global first-order parameters are based on the histogram distribution and properties of each individual pixel. Locoregional second-order and higher-order parameters account for the spatial relationship between the pixels in addition to the intensities.26 Model-based methods, such as

TABLE 1. Patient and Tumor Characteristics

Characteristics Age Median (range) Sex Male Female Histology Adenocarcinoma Squamous cell carcinoma Large cell carcinoma Tumor location Central Peripheral Central and peripheral Tumor stage* IIIA IIIB IV Treatmentw Gemcitabine 1 g/m2 D1,8; Ifosfamide 3 g/m2 D1; cisplatin 50 mg/m2 D1 Docetaxel 75 mg/m2 D1 or Gemcitabine 1 g/m2 D1,8 or Vinorelbine 15 mg/m2 D1,8; Paclitaxel 175 mg/m2 D1 and Cisplatin 75 mg/m2 D1 or Carboplatin AUC6 D1 Mitomycin-C 6 mg/m2 D1; Ifosfamide 3 g/m2 D1; Cisplatin 50 mg/m2 D1 Pemetrexed 500 mg/m2 D1,8,15; Cisplatin 75 mg/m2 D1 Paclitaxel 80 mg/m2 D1,8,15 Gemcitabine 1 g/m2 D1,8; Ifosfamide 3 g/m2 D1;

Group C (N = 23) [n (%)]

Group BC (N = 17) [n (%)]

62 (46-78)

57 (39-73)

22 (96) 1 (4)

12 (71) 5 (29)

18 (78) 3 (13) 2 (9)

17 (100) 0 0

3 (13) 6 (26) 14 (61)

3 (18) 5 (29) 9 (53)

2 (9) 2 (9) 19 (82)

2 (12) 1 (6) 14 (82)

1 (4) 11 (48)

10 (44)

0 3 (18)

0

0

6 (35)

0 1 (4)

8 (47) 0

*Tumors staged according to the AJCC Cancer Staging Manual seventh edition. wAll chemotherapy doses given every 3 weeks for a total of 4 to 6 cycles in both groups, and bevacizumab was given until disease progression or unacceptable toxicities in group BC.

Copyright

r

Imaging Tumor Response and Tumoral Heterogeneity

fractal analysis, examine the underlying global structural geometry of an image.27 A radiomics approach has the potential to provide additional information in response assessment but has yet to be studied in NSCLC. We hypothesized that changes in contrast-enhanced computed tomography (CT) heterogeneity may occur during treatment and may provide additional predictive and prognostic information to current response assessment. We aimed to compare the prognostic value of RECIST 1.1, an alternative response assessment method described by Crabb et al,13 and CT tumoral heterogeneity in advanced and metastatic NSCLC treated with bevacizumab in combination with cytotoxic chemotherapy.

MATERIALS AND METHODS A post hoc analysis was performed using data from a previously reported single-institution prospective study, which evaluated CT perfusion changes in locally advanced and metastatic NSCLC patients treated by chemotherapy combined with or without bevacizumab.28 The previous study evaluated tumor vasculature changes that occur with chemotherapy with or without bevacizumab and evaluated whether these early radiologic perfusion changes could predict clinical response as assessed by RECIST 1.0 and overall clinician’s evaluation. In the current study, we assessed the changes in CT tumoral heterogeneity during treatment, and compared the prognostic ability of RECIST 1.1, an alternate method, and tumoral heterogeneity. As CT heterogeneity evaluates the pixel distribution and intensity within the images, it may provide further information related to other biological processes such as tumor oxygenation and cellular density, in addition to perfusion characteristics within a tumor.

Patients Forty consecutive patients with histologically confirmed inoperable locally advanced or metastatic NSCLC were prospectively enrolled between March 2008 and June 2011 in a previous study.28 Patients with asymptomatic stable brain metastases were eligible for this study. Inclusion criteria were at least 1 unidimensionally measurable lesion as per RECIST criteria, performance status 0 to 2, and adequate hematologic, renal, and hepatic function. Patients were treated with combination cytotoxic chemotherapy (group C, n = 23). In the absence of contraindications to anti-VEGF monoclonal antibodies such as predominantly squamous histology, history of grade 2 or greater hemoptysis, radiologic tumor invasion of major blood vessels, use of aspirin >325 mg/d, full-dose anticoagulation, and/or uncontrolled hypertension, other patients were also given bevacizumab in addition to chemotherapy (group BC, n = 17). Table 1 summarizes the patient, tumor, and treatment characteristics in both groups. All chemotherapy regimens were given every 3 weeks for a total of 4 to 6 cycles in both groups, and bevacizumab was given until disease progression or unacceptable toxicities in group BC. None of the patients underwent radiotherapy during the study.

Imaging All patients underwent dynamic contrast-enhanced CT (DCE CT) of the chest on a dual-source 64-slice multidetector CT (Somatom Definition; Siemens Medical Solutions, Forchheim, Germany). The same protocol was used for baseline and subsequent imaging with 11 consecutive spiral acquisitions covering the entire tumor. The first CT data set was obtained before intravenous contrast injection.

2015 Wolters Kluwer Health, Inc. All rights reserved.

www.thoracicimaging.com |

Copyright r 2015 Wolters Kluwer Health, Inc. All rights reserved.

301

Yip et al

After injection of 108 mL of 300 mg/mL iodinated contrast agent (Ultravist 300, iopromide; Schering, Berlin, Germany), 10 successive postcontrast data sets were then acquired. Each spiral acquisition was reconstructed with overlapping 3-mm-thick transverse mediastinal sections that were spaced 2 mm apart. DCE CT was performed before (Time0) and after 1 (week 3; Time1), 3 (week 9; Time2), and 6 (week 18; Time3) cycles of chemotherapy. All patients underwent Time0 and Time1 scans, with 83%/88% and 61%/71% of patients in group C/group BC having had Time2 and Time3 scans, respectively (Table A1 in Appendix, Supplemental Digital Content 1, http://links.lww.com/ JTI/A57).

Imaging Assessment Treatment response assessment was performed by a single observer (C.Y.). Primary and/or any dominant thoracic lesions and enlarged lymph nodes >1.5 cm in short axis were included in the response evaluation and heterogeneity analysis. Morphologic tumor response was assessed as per RECIST 1.19 as well as the alternate method13 that subtracts the longest diameter of any intratumoral cavity from the longest total diameter of the target lesion, with all measurements taken in the same plane. Classification of response [partial response (PR), stable disease (SD), progressive disease (PD)] is identical using either method. Parenchymal lung tumors were evaluated using lung windows, whereas soft tissue and lymph nodes were assessed using mediastinal window images. The best response was pragmatically determined as the best thoracic radiologic response obtained at any time point (Time1, Time2, or Time3), but no confirmation of response or duration of response was required in the trial. Image heterogeneity analysis was performed using inhouse software implemented under MATLAB (The MathWorks Inc., Natick, MA) to derive global and locoregional heterogeneity parameters (Appendix, Supplemental Digital Content 1, http://links.lww.com/JTI/A57). For this analysis, images from the same CT phase of contrast agent enhancement were used for all patients. Regions of interest were delineated on the axial images to generate a volume of interest, which included the entire target lesion(s). Heterogeneity analysis was performed using statistical (first-order, second-order, and higher-order parameters) and modelbased methods. Global first-order parameters derived from the histogram distribution of pixel intensities included mean, median, skewness, kurtosis, entropy, and uniformity. Locoregional second-order and higher-order parameters included those based on the gray-level co-occurrence matrix (autocorrelation, cluster shade, contrast, entropy), graylevel run-length matrix (short run emphasis, gray-level nonuniformity, run percentage, intensity variability, run length variability), gray-level difference matrix (entropy), gray-level size zone matrix (short zone emphasis, intensity nonuniformity, zone percentage, intensity variability, size zone variability), and neighborhood gray-tone difference matrix (contrast and busyness). Gray-level co-occurrence matrix describes the frequency of various combinations of gray values within a region of interest.29 Gray-level runlength matrix evaluates the number of runs of consecutive pixels with the same intensity.29 Gray-level difference matrix describes the intensity differences between vectors and are based on the absolute differences between pairs of pixels.30 Gray-level size zone matrix measures the size of

J Thorac Imaging



Volume 30, Number 5, September 2015

cluster of voxels within a region,31 and neighborhood graytone difference matrix describes the local intensity differences between each voxel and its direct 26 neighbors.32,33 Global fractal parameters included mean fractal dimension and fractal lacunarity. Fractal dimension describes the degree of irregularity or roughness of an image, whereas lacunarity indicates the gaps or holes observed in the pixel distribution.33 The definition of each parameter is given in the Appendix (Supplemental Digital Content 1, http:// links.lww.com/JTI/A57). In patients with >1 target lesion, the average or mean values were obtained for each heterogeneity parameter. The reproducibility of statistical and model-based heterogeneity features has been previously evaluated.34–38 The average concordance correlation coefficient for statistical-based heterogeneity features extracted from contrast-enhanced CT was 0.67.34 Good reproducibility was found in 3-dimensional fractal analysis derived from contrast-enhanced CT with coefficients of variation