Eur J Nucl Med Mol Imaging DOI 10.1007/s00259-013-2579-4
ORIGINAL ARTICLE
Quantitative assessment of global lung inflammation following radiation therapy using FDG PET/CT: a pilot study Sarah Abdulla & Ali Salavati & Babak Saboury & Sandip Basu & Drew A. Torigian & Abass Alavi
Received: 30 May 2013 / Accepted: 12 September 2013 # Springer-Verlag Berlin Heidelberg 2013
Abstract Purpose Radiation pneumonitis is the most severe doselimiting complication in patients receiving thoracic radiation therapy. The aim of this study was to quantify global lung inflammation following radiation therapy using FDG PET/CT. Methods We studied 20 subjects with stage III non-small-cell lung carcinoma who had undergone FDG PET/CT imaging before and after radiation therapy. On all PET/CT studies, the sectional lung volume (sLV) of each lung was calculated from each slice by multiplying the lung area by slice thickness. The sectional lung glycolysis (sLG) was calculated by multiplying the sLVand the lung sectional mean standardized uptake value (sSUVmean) on each slice passing through the lung. The lung volume (LV) was calculated by adding all sLVs from the lung, and the global lung glycolysis (GLG) was calculated by adding all sLGs from the lung. Finally, the lung SUVmean was calculated by dividing the GLG by the LV. The amount of inflammation in the lung parenchyma directly receiving radiation therapy was calculated by subtracting tumor measurements from GLG.
Sarah Abdulla, Ali Salavati, and Babak Saboury contributed equally to this study. S. Abdulla : A. Salavati : B. Saboury : D. A. Torigian : A. Alavi Department of Radiology, Perelman School of Medicine, University of Pennsylvania, and Hospital of the University of Pennsylvania, Philadelphia, PA 19104, USA S. Basu Radiation Medicine Center, Bhabha Atomic Research Center, Tata Memorial Center Annexe, Bombay, India A. Alavi (*) Department of Radiology, Division of Nuclear Medicine, Hospital of the University of Pennsylvania, 3400 Spruce Street, 1 Donner Building, Philadelphia, PA 19104-4283, USA e-mail:
[email protected]
Results In the lung directly receiving radiation therapy, the lung parenchyma SUVmean and global lung parenchymal glycolysis were significantly increased following therapy. In the contralateral lung (internal control), no significant changes were observed in lung SUVmean or GLG following radiation therapy. Conclusion Global lung parenchymal glycolysis and lung parenchymal SUVmean may serve as potentially useful biomarkers to quantify lung inflammation on FDG PET/CT following thoracic radiation therapy. Keywords Radiation pneumonitis . Radiation therapy . Lung inflammation . FDG PET/CT . Global lung glycolysis . Lung parenchymal glycolysis
Introduction Radiation pneumonitis is the most severe dose-limiting complication in patients receiving thoracic radiation therapy (RT) for various types of cancer (occurring in 15 – 40 %) [1, 2]. Radiation pneumonitis normally presents clinically in the first few months after RT, with symptoms ranging in severity from mild cough to overwhelming dyspnea, fever, and respiratory failure. Diagnosis and management of respiratory symptoms after RT are complicated by the fact that many patients have preexisting lung or cardiac diseases [3–5]. As such, previous studies have shown significantly high degrees of diagnostic uncertainty (28 – 48 %) in patients with a clinical diagnosis of radiation pneumonitis [5, 6]. Several structural and functional imaging modalities have been suggested to improve the diagnosis in patients with post-radiation respiratory symptoms [7–11]. However, the major limitations of structural imaging for the detection of radiation pneumonitis are its insensitivity for the detection of early tissue injury and inflammation in the lung parenchyma, and its suboptimal specificity due to
Eur J Nucl Med Mol Imaging
imaging appearances that may overlap with those of other pulmonary disease entities. The metabolic response of normal lung tissue to radiation has been reported to begin early in the RT course and to peak within a few months following therapy, and is correlated with clinical symptoms [8, 10, 12]. The potential advantage of molecular imaging is the ability to detect disease processes early prior to the development of structural manifestations. Previous studies have shown the value of [18F]FDG PET/ CT for the management of patients with various types of thoracic malignancy [13–18]. Moreover, its value is not limited to the assessment of neoplasms, since sites of inflammation can also show increased FDG uptake due to enhanced glycolysis in inflammatory cells [19]. Therefore, FDG uptake can potentially serve as a biomarker for the underlying inflammatory process of radiation pneumonitis. In patients with locally advanced non-small-cell lung cancer (NSCLC), radiation pneumonitis has become more common due to concurrent chemoradiation therapy [1, 20], and is more complex to diagnose and treat due to the presence of preexisting lung disease. However, only a few studies using FDG PET have been reported to detect and quantify the extent of this radiation effect [2, 10, 11, 21, 22]. In this study, we therefore attempted to quantify the degree of lung inflammation following RT in patients with NSCLC using volumebased FDG PET/CT parameters.
Materials and methods Study sample This retrospective study was performed at the Hospital of the University of Pennsylvania following Institutional Review Board approval and Health Insurance Portability and Accountability Act waiver. Included in this study were 20 subjects (11 women and 9 men; median age 60 years, range 46 – 84 years) with locally advanced NSCLC who had undergone FDG PET/CT before and after RT and who had not previously undergone pulmonary resection. Eight subjects had stage IIIA disease, and 12 had stage IIIB disease. Of the 20 subjects, 1 did not receive chemotherapy but only received RT, whereas the other 19 received chemoradiation therapy. Chemotherapy regimens included: cisplatin/etoposide (10 patients), carboplatin/paclitaxel (7 patients) and cisplatin/ docetaxel (2 patients). The radiation dose administered was 6, 873±692 cGy, and RT lasted for a median of 51 days (range 30 – 83 days). FDG PET/CT was performed a median of 14 days (range 3 – 53 days) prior to the start of RT. FDG PET/CTwas repeated a median of 89 days (range 10 – 213 days) after the completion of the RT course. The median time between the pretreatment and posttreatment FDG PET/CT scans was 150 days (range 70 – 290 days).
Image acquisition All PET/CT scans were acquired using a 16-detector row LYSO whole-body PET/CT scanner with time-of-flight capabilities (Gemini TF; Philips Healthcare, Bothell, WA). Threedimensional PET data were acquired from the skull base to the mid thighs about 60 min after intravenous administration of 444 – 555 MBq (15 mCi) of FDG for 3 min per bed position. All subjects fasted for at least 6 h before FDG injection, and serum glucose levels were verified to be ≤200 mg/dl prior to intravenous administration of FDG. The system model included time-of-flight as well as normalization, attenuation, random, and scatter corrections. Images were reconstructed using a list-mode maximum-likelihood expectation-maximization algorithm with 33 ordered subsets and three iterations. Rescaled low-dose CT images were utilized for attenuation correction of PET images. PET and CT images were reconstructed at 4 mm nominal slice thickness. Imaging and quantitative data analysis PET/CT images were analyzed using dedicated image visualization and analysis software (Extended Brilliance Workstation; Philips Healthcare). Regions of interest (ROIs) were drawn manually around the outer boundaries of the lung on every transverse slice passing through the lung on fused FDG PET/CT images from each subject. The trachea and main stem bronchi were excluded from the ROIs. Lung sectional mean standardized uptake value (sSUVmean) and the area of the lung ROI were recorded from each slice (Fig. 1). Subsequently, the sectional lung volume (sLV) was calculated from each slice by multiplying the lung ROI area (in centimeters squared) by 0.4 (slice thickness 4 mm). The sectional lung glycolysis (sLG) was calculated by multiplying sLV and lung sSUVmean from each slice. The lung volume (LV) was calculated by adding all the sLV from slices passing through the lung, and the global lung glycolysis (GLG) was calculated by adding all the sLG from slices passing through the lung. Finally, the lung SUVmean was calculated dividing the GLG by the LV. These parameters included the lung parenchyma as well as the lung tumor lesions present. To measure tumor metabolic response due to RT, we measured metabolically active tumor volumes (MTV), SUVs and total lesion glycolysis (TLG). To quantify these parameters we used an adaptive contrast-oriented thresholding algorithm [23–27], which permits delineation of the boundaries of lesions based on PET images alone. This modified adaptive thresholding delineation technique combines automatically determined background correction and local adaptive thresholding in an iterative algorithm model [23–26] (ROVER software; ABX, Radeberg, Germany). For partial volume correction (PVC) a local background PVC algorithm
Eur J Nucl Med Mol Imaging
Fig. 1 Global lung glycolysis and global lung parenchymal glycolysis calculation. a Manually drawn ROIs (green) around the irradiated right and nonirradiated left lung on a representative transverse FDG PET/CT slice. The tumor boundary (red) was delineated by semiautomatic software. b In the irradiated right lung, global lung parenchymal glycolysis
(green area) was calculated by subtraction of tumor total lesion glycolysis (red area) from the global lung glycolysis. In nonirradiated left lung, global lung parenchymal glycolysis (green area) was equal to global lung glycolysis
was used. The accuracy and reproducibility of these methods have been verified [23, 26, 28, 29]. To quantify the amount of inflammation in the lung parenchyma directly receiving RT, the following formulas were used:
between variables, the Pearson correlation test was used. The Wilcoxon rank-sum test was used to compare ΔPVC GLPG, ΔGLPG, ΔLP PVC SUVmean and ΔLP SUVmean between groups of subjects defined based on age (suggested cut-off for age >65 years) or chemotherapy regimen (cisplatin/etoposide versus carboplatin/paclitaxel) [1, 4]. Graphs were produced using GraphPad Prism for Windows (version 6.1; GraphPad Software, San Diego, CA).
&
&
& &
ΔLung parenchymal SUVmean (ΔLP SUVmean) = [(postradiation GLG − postradiation TLG)/(LV − postradiation MTV)] − [(preradiation GLG − preradiation TLG)/ (LV − preradiation MTV)] ΔLung parenchymal PVC SUVmean (ΔLP PVC SUVmean) = [(postradiation GLG − postradiation PVC TLG)/(LV − postradiation MTV)] − [(preradiation GLG − preradiation PVC TLG)/(LV − preradiation MTV)] ΔGlobal lung parenchymal glycolysis (ΔGLPG) = (postradiation GLG − postradiation TLG) − (preradiation GLG − preradiation TLG) ΔPVC global lung parenchymal glycolysis (ΔPVC GLPG) = (postradiation GLG − postradiation PVC TLG) − (preradiation GLG − preradiation PVC TLG)
Results Tumor response In primary tumors, postradiation MTV, SUVmax, uncorrected SUVmean, PVC SUVmean, uncorrected mean TLG, and PVC TLG all significantly decreased (Table 1). The changes in TLG and tumor SUVmean became more apparent when PVC was utilized (ΔPVC TLG −556.2 cm3 versus ΔTLG −366.6 cm3; ΔPVC SUVmean−9.25 versus ΔSUVmean −5.39).
Statistical analysis Inflammation Analysis was performed using Stata 11 (StataCorp, College Station, TX). To summarize the variables of this study, descriptive statistics were calculated (means, standard deviations, 95 % confidence intervals). To compare the values before and after treatment, a paired t-test was used. To calculate correlations Table 1 Tumor metabolic response parameters
Where TLG=MTV × SUVmean, PVC TLG=MTV × PVC SUVmean
In the lung directly receiving RT, ΔLP SUVmean and ΔLP PVC SUVmean showed a significant increase (Table 2). However, the increase in ΔGLPG in the lung receiving radiation was not statistically significant but ΔPVC GLPG increased
FDG PET/CT parameter
Absolute change
Relative change (%)
p value
ΔSUVmax ΔSUVmean ΔPVC SUVmean ΔMTV ΔTLG ΔPVC TLG
−9.08 (95 % CI 6.85 – 11.32) −5.39 (95 % CI 3.90 – 6.89) −9.25 (95 % CI 6.62 – 11.87) −26.49 cm3 (95 % CI 4.89 – 48.08 cm3) −366.6 cm3 (95 % CI 139.0 – 594.2 cm3) −556.2 cm3 (95 % CI 230.5 – 881.8 cm3)
−62 −62 −65 −75 −73 −74
