18 F-FDG accumulation in less-affected lung area predicts postoperative interstitial lung disease acute exacerbation in lung cancer: A case control study CURRENT STATUS:

Background: Pneumonectomy for lung cancer with interstitial lung disease (ILD) has been shown to cause postoperative acute exacerbation (AE) of the ILD. The accumulation of 18 F-flurodeoxyglucose (FDG) on normal parenchymal or less-affected lung fields in 18 F-FDG-positron emission tomography (PET)/computed tomography (CT) has been reported to be related to ILD disease activity and prognosis. To determine whether 18 F-FDG accumulation in normal parenchymal or less-affected lung fields on 18 F-FDG-PET/CT can predict postoperative AE of ILD in non-small cell lung cancer (NSCLC) patients with ILD. Methods: This retrospective study included 36 NSCLC patients with ILD, who underwent 18 F-FDG-PET/CT at 2 institutions before pulmonary surgery. A single volume-of-interest (VOI) was placed to measure the mean standardized uptake value (SUV mean ) in normal or less-affected lung fields at 12 areas on the ventral and dorsal locations of both lungs, in each level of the aortic arch, tracheal bifurcation, and the orifice of the right lower pulmonary vein into the left atrium. The region to which the target VOI was set corresponded to no or minimally increased attenuation on high resolution CT. The SUV mean was defined as the mean SUV of the target VOI, SUV tissue fraction (TF) as the corrected SUV mean by using TF and mean computed tomography density (CTD mean ) as the mean attenuation of the corresponding target VOI on HRCT. We performed a phantom study to optimize SUV difference among 2 institutions. The corrected SUV mean (cSUV mean ) and corrected SUV TF (cSUV TF ) were calculated based on the phantom study result. Results: Among 36 NSCLC patients with ILD who underwent pulmonary surgery, 8 patients developed postoperative AE of ILD. The cSUV mean values in the ventral and dorsal locations at the aortic arch level, and in the ventral location at the tracheal bifurcation level in the group with postoperative AE were higher than in

Conclusion: 18 F-FDG accumulation in the normal or less-affected lung fields can potentially predict postoperative AE of ILD in NSCLC patients with ILD.

Background
Interstitial lung disease (ILD) results not only from parenchymal lung diseases of known or unknown etiology, but also in lung regions with both inflammation and fibrotic changes (1,2). ILDs are recognized as a risk factor for lung cancer (3)(4)(5). Since therapy of lung cancer, such as surgery, chemotherapy, and chemoradiotherapy sometimes provoke acute exacerbation (AE) of ILD, treatment should be chosen carefully (6)(7)(8). Pneumonectomy poses a risk of postoperative AE for patients with lung cancer-associated ILD and has a reported mortality rate between 33.3% and 100% (9)(10)(11)(12)(13). In a retrospective study conducted in Japan, the incidence of postoperative AE and the mortality rate in lung-cancer-associated ILD was 9.3% and high as 43.9%, respectively (13). Moreover, a high Krebs von den Lungen-6 (KL-6) value, male sex, history of AE, preoperative steroid use, and the surgical procedure were identified as possible risk factors for postoperative AE (13). 18 F-flurodeoxyglucose (FDG) positron-emission tomography (PET) has been proven to be useful for the detection and follow-up of neoplastic lesions (14). Recently, several reports have stated that 18 F-FDG accumulation may be related to disease activity in patients with ILDs (15)(16)(17). Furthermore, increased 18 F-FDG uptake in apparently normal lung parenchyma on high-resolution computed tomography (HRCT) was observed in patients with idiopathic pulmonary fibrosis (IPF), and was related to the prognosis (18)(19)(20). These reports suggested that a high uptake of 18 F-FDG in the less-affected lung parenchyma may be an indicator of the disease activity and progression in ILD.
This retrospective study was conducted to investigate whether 18 F-FDG accumulation in the lessaffected lung parenchyma on HRCT can predict postoperative AE in non-small cell lung cancer (NSCLC)-associated ILD.

Patient Selection
In the patient selection process, 388 patients with NSCLC who underwent pulmonary surgery in Shiga University of Medical Science between January 2011 and December 2016 were initially enrolled.
Among them, 38 patients with NSCLC complicated with ILD who underwent 18 F-FDG PET/CT before pulmonary surgery in our hospital or Oumikusatsu Tokushukai Hospital were identified. 2 patients with glucose metabolism abnormalities were excluded. Thus, finally, 36 patients with ILD were included.
To investigate the difference in 18 F-FDG accumulation in the apparently normal or less-affected lung area between NSCLC patients with or without ILD, we additionally included 50 consecutive NSCLC patients without ILD who underwent both 18 F-FDG PET/CT before pulmonary surgery and surgery in our hospital between August 2015 and December 2016.
Available patient characteristics, clinical laboratory data, and pulmonary functional tests, including age, sex, smoking status, Eastern Cooperative Oncology Group (ECOG) performance status, histology of NSCLC, surgical procedure, stage of NSCLC, lactate dehydrogenase, C-reactive protein, krebs von den lungen-6 (KL-6), and surfactant protein-D were obtained from medical records. Whether ILD was associated with NSCLC or not was estimated by chest HRCT underwent within 3 months before pulmonary surgery. Based on HRCT image findings the NSCLC patients with ILD were classified into having the usual interstitial pneumonia (UIP) pattern and non-UIP pattern. The UIP pattern was assigned when abnormal shadows, including subpleural basal predominance, reticular abnormalities, and honeycombing with or without traction bronchiectasis were present and features of an inconsistent UIP pattern were absent, according to the International Consensus Statement of IPF (21).
Otherwise, patients were diagnosed as having a non-UIP pattern.
This study was approved by our institutional review board (approved number: 29-190, November 2, 2017). The need to obtain informed patient consent was waived because of the retrospective nature of this research.

Criteria for Postoperative Exacerbation of ILD
The diagnosis of postoperative AE of ILD was based on the definition of the International Working Group Report for AE of IPF (22). The diagnostic criteria were as follows: acute worsening or development of dyspnea, typically of < 1 month duration after pulmonary surgery; new bilateral ground-glass opacity and/or consolidation on computed tomography (CT); deterioration not fully explained by cardiac failure or fluid overload (22). 18 F-FDG PET scanning was performed using a combined PET/CT scanner (Discovery PET/CT 710, General Electronics, Fairfield, CT, USA) in our institution, or a Discovery PET/CT ST (General Electronics) in Oumikusatsu Tokushukai Hospital). All patients were instructed to fast for 5 hours or longer before 18 F-FDG administration. Blood glucose level was measured before 18 F-FDG injection to confirm a level of less than 150 mg/dL (23). Since the blood glucose level influenced 18 F-FDG accumulation in brain (24,25), patients with unknown blood glucose level were excluded if accumulation of 18 F-FDG in their brains was judged to be insufficient. There were no cases whose blood glucose level before 18 F-FDG-PET imaging exceeded 150 mg/dl. However, among 10 patients for whom blood glucose level values were unavailable, 2 patients were excluded from analysis because of faint 18 F-FDG accumulation demonstrated in the brain. Three-dimensional PET data were acquired from the head to the thigh 60 min after the injection of a dose of 185-330 MBq/kg of 18 F-FDG.

Image Analysis
PET images were analyzed on a dedicated workstation (Advantage Workstation, version 2.0; General Electronics, Fairfield, CT, USA). On PET/CT images, a single cubic target volume-of-interest (VOI) of approximately 18 cm 3 in volume (26 mm × 26 mm × 26 mm) was carefully placed to include normalappearing regions while avoiding the mediastinum, chest wall, central bronchus, and blood vessels, in 12 areas (Figure 1): the ventral and dorsal locations in both lungs at each of the 3 predefined levels: aortic arch (AA), tracheal bifurcation (TB), and the orifice of the right lower pulmonary vein into the left atrium (RLPV), while referring to non-increased attenuation on HRCT images. Standardized uptake values (SUVs) were defined as follows: SUV mean was measured in a target VOI. SUV max was measured as the highest SUV in ILD lesion (18,19). SUV tissue fraction (TF) was analyzed to adjust for metabolic condition, regardless of differences in the degree of region-based aeration (26, 27). Mean CT density (CTD mean on PET/CT) as well as SUV mean was simultaneously obtained. CTD mean on HRCT were measured using Image J (Version 1.51. National Institutes of Health, Bethesda, ML, USA), by placing an identical-sized VOI at essentially the same location where the VOI was set on PET/CT. SUV mean , SUV TF , SUV max , and CTD mean on HRCT were independently assessed by 1 chest physician and 1 thoracic radiologist with 15 and 20 years of experience, respectively, to determine interobserver agreement.

Phantom Study for Adjustment of Measured Values between the Institutions
The SUV values measured for images obtained at the 2 institutions were adjusted according to a method described in a previous study (28). Phantom experiments were performed with modification of the dedicated guideline issued by the Japan Nuclear Medicine Society, by using the National Electrical Manufacturers Association (NEMA) body phantom. The NEMA body phantom, with spherical containers of 6 different diameters: 10, 13, 17, 22, 28, and 37 mm, filled with 18 F-FDG, for insertion into different body parts, was scanned for 30 minutes in list-mode. The radioactivity levels were 2.65 kBq/mL for spherical containers filled with 18 F-FDG arranged and 0.66 KBq/mL for background parts, over 60 minutes at the time of data acquisition. SUV max values of the image slice with the most highly integrated radioisotope and the 2 antero-posterior neighboring image slices were averaged as the actual SUV max for the spherical containers. The gauged mean SUV in the phantom background was determined as the mean of the SUVs of 12 ROIs of 37-mm diameter in the image slice with the most highly integrated radio-isotope and its 4 anteroposterior neighboring image slices for the 10-mm spherical container.
The calibration factor was determined by dividing the actual SUV max by the gauged mean SUV of the phantom background to reduce inconsistencies between 2 institutes. This adjustment for interinstitutional variability in SUV values shrunk the range from 0.69-0.89 to 0.74-0.97 when the SUV max ratio was described as the SUV max of the one institution to the SUV max of the control institution. The original measured SUV max values for both institutions were corrected by multiplication with the calibration factors derived from the phantom studies to minimize inter-institutional SUV variability; these values were defined as cSUV.

Statistical Analysis
Univariate analyses were conducted to identify the difference in geographic data, physiological parameters, and laboratory data between patients with NSCLC with ILD and without ILD, and between the NSCLC patients with and without postoperative exacerbation of ILD. All continuous and categorical variables were analyzed with Wilcoxon's test and the chi-square test. For locations with difference in cSUV mean , cSUV TF , and cSUV max measured values based on PET/CT between NSCLC patients with and without postoperative exacerbation of ILD, receiver operating characteristic (ROC) curve analysis were performed. Two-sided P values of less than 0.05 were considered statistically significant. The normality of SUV mean , SUV TF , and CTD mean on HRCT measured by the 2 observers was evaluated with the Shapiro-Wilk test. Inter-observer variance was evaluated by Spearman's correlation coefficient for the measured values. All statistical analyzes were performed using JMP 9.0 software (SAS Institute Inc., Cary, NC, USA). Table E1 shows the characteristics of the NSCLC patients with ILD and those without. The median age was 73.5 years (range: 68-77 years) in NSCLC patients with ILD and 69 years (range: 63.8-72.3 years) in those without. Most NSCLC patients with ILD were men, with good ECOG performance status, and were smokers. The frequencies of adenocarcinoma and squamous cell carcinoma were higher in NSCLC patients with than in those without ILD. There was no significant difference in the surgical procedure used and stage between the 2 groups. The %VC, the %FVC, the %TLC, the DLco, and the %DLco in NCSLC patients were lower in those with than in those without ILD. Table 1  There was no significant difference in histology, surgical procedure, and stage between those with and those without postoperative AE. In terms of interstitial pneumonia classification, 4 (50%) and 12 cases (43%) were diagnosed with UIP in the AE group and the non-AE group, respectively. No significant difference was found in terms of pulmonary function test parameters between the 2 groups, except for a slightly larger value of surfactant protein D in patients with postoperative AE.

Inter-observer Agreement
Values measured by the 2 observers were not all normally distributed. Spearman correlation coefficients for inter-observer differences in measured SUV mean , SUV TF ranged from 0.86 to 0.97, from 0.80 to 0.92, from (Table E2-E4) respectively. The value in SUV max of patients with ILD (n=36) was 0.874 (P <0.001). Thus, there were small inter-observer variances.

Comparison of Corrected SUV mean , Corrected SUV TF, and CTD mean on HRCT between NSCLC patients with and without ILD
cSUV mean was higher in NSCLC patients with than in those without ILD, in all locations except for both locations at the AA level, the ventral location at the TB level, and the ventral location at the RLPV level in the left lung (Table E5). cSUV TF was higher in all locations in NSCLC patients with than in those without ILD (Table E5). CTD mean on HRCT was similar between NSCLC patients with and those without ILD at all locations (Table E6).

Comparison of Corrected SUV mean , Corrected SUV TF , Corrected SUV max , and CTD mean on HRCT between NSCLC patients with and those without postoperative ILD exacerbation
In NSCLC patients with postoperative ILD exacerbation, the value of cSUV mean in the right ventral and dorsal locations at the AA level, and the right ventral location at the TB level were higher than in those without postoperative AE (P = 0.01, P = 0.02, P = 0.02, respectively) ( Table 2). Receiver operating characteristic (ROC) curve analysis at the 3 locations with a greater cSUV mean for NSCLC patients with AE also showed that the area under curve (AUC) for the right ventral location at the AA level in the right lung was greatest (AUC = 0.82), with an optimal cut-off value of 0.46. This cut-off value yielded a sensitivity of 75.0% and a specificity of 56.8% ( Figure E1).
The value of cSUV TF , cSUV max , and CT mean on HRCT showed no significant difference between the NSCLC patients with and those without postoperative AE at any of the locations (Tables 2, 3, E7).

Discussion
We investigated whether SUV in the less-affected lung parenchyma could predict postoperative AE of ILD in patients with NSCLC and ILD. The SUV mean in NSCLC patients with postoperative AE of ILD was significantly higher than in those without AE, in several regions. ROC analysis showed that the AUC was greatest in the right ventral location at the AA level. Increased 18 F-FDG uptake in the dorsal and lower lung field has been shown to be associated with a gravitational effect and blood flow increase in some previous studies that used 18 F-FDG PET imaging (29,30). Since respiratory movement in the lower lung field is generally regarded as larger than in the upper lung field, FDG accumulation tended to be overestimated in 18 F-FDG PET imaging in the lower lung field under free-breathing conditions. In an additional visual evaluation on HRCT before pulmonary surgery, with regard to the laterality of the interstitial lung shadow, the interstitial lung shadow in the group with postoperative AE was predominant in the right lung (P = 0.06). Therefore, while taking into account the factors influencing FDG accumulation, we consider that the higher SUV mean in the ventral region of the upper lung field in those with postoperative AE than in those without AE is meaningful, and that SUV mean measurement in normal-appearing ventral regions may be useful to predict postoperative AE.
On the other hand, SUV TF was not significantly different between NSCLC patients with and those without postoperative AE. This may be because the mean CT value of 18 F-FDG-PET/CT in the right upper and middle lung field was higher in individuals with than in those without postoperative AE of ILD (data not shown). In contrast, the mean CT value of HRCT obtained at peak-inspiration was similar between those with and those without postoperative AE of ILD. Since 18 F-FDG-PET/CT image data were obtained under free breathing conditions, the breathing level could not be kept constant. Data acquisition might sometimes have occurred during the expiratory phase on FDG-PET/CT in some patients. Moreover, the average CT value may be affected by the degree of pulmonary regional collapse as well as the increased blood flow. Therefore, SUV might be overcorrected based on regional CT density, due to the predominant focal collapse in NSCLC patients with postoperative AE of ILD, which could have obscured differences in SUV TF between NSCLC patients with and those without postoperative AE of ILD. These results paradoxically imply that conventional CT obtained at peakexpiration may be useful for differentiating patients at increased risk of postoperative AE from those not at. As SUV TF needs to be calculated by measuring the CT value of soft tissue and air (26), more time is required for imaging, which should be addressed as a drawback of SUV TF calculation.
As in previously reports (18,19), both SUV mean and SUV TF in the less-affected lung fields of NSCLC patients with ILD were significantly higher than in those patients without ILD. These results suggest that accumulation of 18 F-FDG in less-affected lung fields may reflect inflammatory conditions in the pulmonary interstitium, which cannot be visually detected with conventional HRCT. However, the mechanism underlying accumulation of 18 F-FDG in less-affected lung fields has remained incompletely understood. Disorders of the alveolar epithelium, migration of inflammatory cells, and release of inflammatory mediators were involved in the process of lung fibrosis in a bleomycinhamster model (31). 18 F-FDG migrates into cells via glucose transporter 1 (GLUT-1), which was shown to be expressed in fibrocytes, bronchial epithelial cells, and inflammatory cells (32)(33)(34). These mechanisms may contribute to accumulation of 18 F-FDG in ILD. Accumulation of 18 F-FDG in lessaffected lung fields in cases with ILDs would therefore show disease activity before morphological changes become detectable on HRCT.
There were several limitations in this study. First, this is a retrospective analysis with a small sample size. Therefore, the usefulness of SUV mean measurement for less-affected regions in ventral upper lung field should be re-assessed in a larger number of cases in future. Second, regions where emphysematous changes exceeded 50% of the area were excluded from the target VOI setting in this study; however, the effect of emphysema could not be excluded completely. In a previous report, accumulation of 18 F-FDG in emphysema was elevated in chronic obstructive pulmonary disease patients as compared to healthy control (35).
In conclusion, the accumulation of 18

Ethics approval and consent to participate
This study was approved by our institutional review board (approved number: 29-190, November 2, 2017). The need to obtain informed patient consent was waived because of the retrospective nature of this research.

Consent for publication
Not applicable.

Availability of data and materials
All data generated or analysed during this study are included in this published article and its supplementary information files.

Supplementary Files
This is a list of supplementary files associated with this preprint. Click to download. Figure E1.pptx