15O-Water Dynamic Positron Emission Tomography in Patients with Non-Small Cell Lung Cancer: Early Decrease in Blood Flow After Anti-Angiogenic Treatment is Correlated with a Shorter Time to Tumour Progression

Purpose: To evaluate tumour blood ow in patients with non-small cell lung cancer (NSCLC) who underwent chemotherapy with bevacizumab before and after administration of chemotherapy using 15 Owater positron emission tomography (PET), and to investigate the effects of bevacizumab on tumour blood ow changes and progression-free survival (PFS). Methods: Twelve patients with NSCLC were enrolled. Six patients had chemotherapy with bevacizumab, and the other six had chemotherapy without bevacizumab. 15 O-water dynamic PET scans were performed within 1week before the start of chemotherapy and within 1 week after the rst day of chemotherapy. Tumour blood ow was analysed quantitatively using a single one ‐ tissue compartment model with the correction of pulmonary circulation blood volume and arterial blood volume via an image-derived input function. Results: In the bevacizumab group mean tumour blood ow had reduced statistically signicantly post-chemotherapy (pre-chemotherapy 0.27 ± 0.14 mL/cm 3 /min, post ‐ chemotherapy 0.18 ± 0.12 mL/cm 3 /min). In the no bevacizumab group there was no signicant difference between mean tumour perfusion pre-chemotherapy (0.42 ± 0.42 mL/cm 3 /min) and post-chemotherapy (0.40 ± 0.27 mL/cm 3 /min). In the bevacizumab group there was a positive correlation between the blood ow ratio (tumour blood ow post-chemotherapy/tumour blood ow pre-chemotherapy) and PFS (correlation coecient 0.94). Conclusion: Mean tumour blood ow decreases within 1–2 days after bevacizumab administration. Greater reductions in blood ow were associated with shorter PFS.


Introduction
Vascular endothelial growth factor (VEGF) is a major contributor to angiogenesis, which plays important roles in local tumour progression and metastatic growth. Overexpression of VEGF has been observed in a variety of cancers including non-small cell lung cancer (NSCLC). Bevacizumab is a monoclonal anti-VEGF antibody that inhibits angiogenesis by preventing circulating VEGF from binding to its receptors (Ferrara et al., 2004). The addition of bevacizumab to cytotoxic chemotherapy improves overall survival and progression-free survival (PFS), and is an accepted component of care for NSCLC [1]. Notably however, the mechanism by which the combination of bevacizumab and cytotoxic chemotherapy improves survival in cancer patients remains unclear.
Positron emission tomography (PET) with 15 O-labelled water is an established method for measuring tissue blood ow quantitatively. 15 O-water is an ideal tracer for quantifying blood ow because it is distributed to tissue freely and cannot be metabolized. Compared to other PET tracers such as 82 Rb, the correlation between 15 O-water-derived values and actual blood ow is very high [2]. Previously 15 O-water PET was an invasive examination because it required continuous arterial blood sampling during the scan to obtain the input function. Now the image-derived input function is commonly used however, which is a noninvasive and accurate alternative to arterial sampling [3]. PET using 15 O-water is regarded as the gold standard for brain perfusion and it is now used in other areas such as myocardial perfusion and tumour perfusion [4,5].
Imaging is commonly and widely used as a non-invasive procedure to evaluate treatment responses. The Response Evaluation Criteria in Solid Tumours (RECIST) are widely used to evaluate treatment responses, but in many cases the criteria are not appropriate for predicting the outcomes of novel molecularly targeted cancer treatments because they are based on the reduction rate of morphological size [6]. PET facilitates the visualization of metabolic activity in tumours, resulting in more informative evaluations of treatment responses after chemotherapy. 15 O-water PET enables the measurement of tumour perfusion de ned as blood ow (mL/min) per cm 3 . We hypothesized that when assessing treatment responses to bevacizumab it may be possible to determine therapeutic effects more accurately by evaluating changes in tumour blood ow than by evaluating changes in tumour size identi ed via computed tomography (CT).
The purpose of the current prospective study was to evaluate tumour blood ow in patients with NSCLC who underwent chemotherapy with or without bevacizumab before and after that chemotherapy using 15 O-water PET, and to investigate the effects of bevacizumab on tumour blood ow and PFS.

Patients
Between April 2012 and July 2015 patients with stage IV NSCLC who were scheduled to undergo chemotherapy at the Osaka University Hospital in Osaka, Japan were recruited by respiratory physicians before the start of treatment. Of the 76 patients encountered during the study period, 13 agreed to participate in the study (Fig. 1). One patient withdrew from participation after providing consent. The nal sample of subjects included 8 men and 4 women, with a mean age of 60 years (range 42-73 years). Ten patients had adenocarcinoma, 1 had pleomorphic carcinoma, and 1 had poorly differentiated carcinoma.
The patient characteristics are summarized in Table 1. 15 O-water PET is not covered by insurance in Japan. Approval from the internal Ethics Review Board was obtained before the initiation of the study. Informed consent was obtained from each patient. Chemotherapy Initial chemotherapy was administered to all the patients in the study in hospital. The chemotherapy regimens-including the use of bevacizumab-depended on the respiratory physicians treating each individual patient. Six patients (4 men and 2 women, mean age 62 ± 10 years) underwent chemotherapy that included bevacizumab, and the other 6 (4 men and 2 women, mean age 59 ± 11 years) had chemotherapy without bevacizumab. All patients in the bevacizumab group were diagnosed with adenocarcinoma histologically. Four of them were administered bevacizumab with carboplatin and paclitaxel, and 2 of them were administered bevacizumab with carboplatin and pemetrexed. The histological types in the no bevacizumab group included 4 adenocarcinomas, 1 pleomorphic carcinoma and 1 poorly differentiated carcinoma. The chemotherapy regimens are shown in Table 1 PET/CT imaging protocol 15 O-water dynamic PET scans were obtained as a prospective study designed to evaluate tumour blood ow before and after bevacizumab administration. Baseline PET was performed within 1 week before the start of chemotherapy (mean 3.8 days, range 1-6 days). Post-chemotherapy PET was performed within 1 week after the rst day of administration (mean 1.7 days, range 1-6 days). PET imaging was acquired on a SET-3000 GCT/X scanner (Shimadzu Corp., Kyoto, Japan). This scanner has an axial eld of view of 26 cm, divided into 99 contiguous planes. The intrinsic spatial resolution is 3.5 mm full width at half maximum (FWHM) in-plane, and 4.2 mm FWHM axially.
Patients were positioned supine in the scanner bed with both the tumour and the aortic arch or heart in the centre of the axial eld of view. For attenuation correction a 5-min transmission scan was performed using a 137 Cs point source. After that transmission scan a 10-min list mode scan was started simultaneously with an intravenous injection of 185 MBq of 15 O-water (total amount 18.5 mL, injection speed 0.5 mL/s). The emission scan was reconstructed into 22 frames (1 frame × 10 s, 8 frames × 5 s, 4 frames × 10 s, 2 frames × 15 s, 3 frames × 20 s, 2 frames × 30 s and 6 frames × 60 s) using the twodimensional dynamic row-action maximumlikelihood algorithm after three-dimensional Gaussian smoothing with a 6-mm FWHM. The voxel size was 4.7 × 4.7 × 2.6 mm.
After the emission scan, reference CT was performed. The CT acquisition parameters were breath hold at shallow inspiration, from the apex of the lung to the base of the lung, no intravenous media, 120 kVp and 50 effective mAs, 52 slices, and 5.0-mm slice thickness. Clinically baseline CT scanning was performed up to 1 month before the start of chemotherapy. Follow-up CT scans were performed after 2 courses of chemotherapy and every 2-4 months thereafter. Brain magnetic resonance imaging and bone scintigraphy were performed as needed. PFS was de ned as the time from the start of treatment to the appearance of local recurrence or new metastatic lesions.

Quantitative analysis for tumour blood ow
Tumour perfusion was calculated via the following equation (Eq. 1), using a single one-tissue compartment model with the correction of pulmonary circulation blood volume and arterial blood volume [3]. The image-derived input function was used. Volumes of interest (VOIs) with a diameter of 1 cm were drawn over the ascending aorta in approximately 10 consecutive image planes of the frame in which the rst pass of the bolus was best visualized. Projection of the resulting VOI onto all image frames yielded the arterial time-activity curve or image-derived input function C A (t). Applying a similar approach to the right ventricular cavity provided a time-activity curve for the pulmonary circulation C V (t). Parametric images of perfusion were generated using the basis function method [7]. In the present study 300 logarithmically spaced precomputed basis functions with F/VT values ranging from 0.0 to 0.4 min − 1 were used. Tumour VOIs were de ned on the reference CT and projected onto the parametric perfusion images using PMOD 3.6 (PMOD Technologies, LLC, Zürich, Switzerland). The averaged tumour perfusion over the VOI in the parametric perfusion images was used for statistical analyses.

Statistical analysis
Tumour blood ow before and after chemotherapy were compared in the bevacizumab group and the no bevacizumab group. In the bevacizumab group tumour blood changes and PFS were examined. All statistical analyses were performed using commercially available software (MedCalc version 19.2.1, Frank Schoonjans, Mariakerke, Belgium). Differences in mean tumour blood ow before and after chemotherapy in each group were analysed using the Wilcoxon test. In the bevacizumab group regression analysis was performed on the blood ow ratio before and after chemotherapy and PFS. p < 0.05 was considered statistically signi cant.

Results
In the bevacizumab group mean tumour perfusion was statistically signi cantly reduced after chemotherapy (pre-chemotherapy 0.27 ± 0.14 mL/cm 3 /min, post-chemotherapy 0.18 ± 0.12 mL/cm 3 /min). In the no bevacizumab group there was no signi cant difference between mean tumour perfusion pre-chemotherapy (0.42 ± 0.42 mL/cm 3 /min) and post-chemotherapy (0.40 ± 0.27 mL/cm 3 /min) (Fig. 2). Table 2 shows the tumour blood ow measurements in each patient. PFS was not signi cantly correlated with tumour blood ow before or after chemotherapy in either group. In the bevacizumab group the rate of decline in tumour blood ow varied markedly in different patients. There was a positive correlation between the blood ow ratio (post-chemotherapy tumour blood ow/prechemotherapy tumour blood ow) and PFS (correlation coe cient 0.94), yielding a regression equation of y = 0.2729 + 0.001616x (p = 0.005) (Fig. 3). A smaller blood ow ratio after chemotherapy was associated with a shorter time to tumour recurrence or regrowth (Figs. 4 and 5). In the no bevacizumab group there was no signi cant correlation between the blood ow ratio and PFS (correlation coe cient 0.74, p = 0.091).

Discussion
In the present study bevacizumab was associated with signi cantly reduced tumour blood ow 1-2 days after chemotherapy. Notably however, follow-up investigation revealed that this effect was associated with rapid tumour regrowth. Patients with only a slight change in blood ow tended to exhibit longer PFS. chemotherapy regimens including bevacizumab in patients with NSCLC. The vascular normalization theory proposed by Jain [8] may explain this phenomenon. Unlike physiological angiogenesis processes such as wound healing, tumour angiogenesis continues abnormally while the tumour is growing because the tumour requires a vascular supply to provide essential nutrients and oxygen. Tumour vessels are often tortuous, disorganized and highly permeable, resulting in high interstitial pressure and reduced blood perfusion and oxygenation. Tumour cells can adapt to insu cient blood supply and hypoxia, but drug delivery is inhibited and its e cacy is reduced. Excessive antiangiogenic therapy may lead to reduced tumour blood ow and result in hypoxia and acidosis, which promote tumour progression [9]. Moderate anti-VEGF therapy may lead to 'vascular normalization', which is characterized by attenuation of hyperpermeability, increased vascular pericyte coverage, a more normal basement membrane, and a resultant reduction in tumour hypoxia and interstitial uid pressure. As a result drug delivery of cytotoxic anticancer agents is improved, and consequently chemotherapy in combination with anti-VEGF drugs improves survival.
In the current study a decline in tumour blood ow after bevacizumab administration was observed in all patients, but patients with greater reductions in tumour blood ow exhibited tumour regrowth within shorter periods. It may be that greater reduction of tumour perfusion re ects greater pruning of vessels, which leads to hypoxia and acidosis in the tumour. Heist et al. [10] reported that reduced blood perfusion after bevacizumab administration as determined via CT was associated with shorter overall survival in NSCLC patients, which is consistent with the results of the present study. They assessed tumour perfusion before bevacizumab administration and 14 days thereafter. In the current study the assessment of tumour perfusion was performed in 1-2 days after bevacizumab administration. The present study suggests that it may be possible to predict the effects of chemotherapy just days after the administration of bevacizumab using 15 O-water PET. If bevacizumab is found to be insu ciently effective at an early stage, switching to another anticancer drug could be considered earlier. In addition, 15 Owater PET can be performed at a lower dose than perfusion CT; 0.37 mSv vs. 3.5-6.5 mSv [11].
Tumour blood ow is still not well understood, so it is necessary to clarify how tumour blood ow before treatment and changes in tumour blood ow after treatment affect the e cacy of anticancer therapy, and prognoses. Accordingly, a larger number of cases needs to be examined. Tumour blood ow is associated with tumour hypoxia, which is associated with resistance to chemotherapy. An explanation for why the extent of the reduction in tumour blood ow is associated with the response to chemotherapy may be associated with the altered hypoxic region of the tumour. Further insights may be obtained by combining hypoxic imaging with a radiolabelled tracer such as 18 F-uoromisonidazol [12].
The current study had several limitations. One is the small number of participants, which resulted in low statistical power. It was also conducted at a single facility, which may have resulted in selection bias.
Each patient underwent different chemotherapy regimens, which may have in uenced blood ow and prognoses. There were also some technical limitations. The parametric images obtained are prone to noise, which slightly reduces the reliability of the quanti cation. Each VOI was manually placed over the pulmonary nodules, making it as large as possible in an effort to minimize the effects of inhomogeneity. This may have resulted in some variability due to manual measurement.
In conclusion, in the current study mean tumour blood ow diminished within 1-2 days after bevacizumab administration in NSCLC patients, and greater reductions in blood ow were associated with shorter PFS.  image after chemotherapy. The CT showed a 4.4 cm large mass in the right lung. Tumor blood ow was 0.371 ml/ml/min before treatment and slightly decreased to 0.334 ml/ml/min after chemotherapy (blood ow ratio: 0.90). He had no tumor progression for 1 year and 2 months. Tumor blood ow was markedly reduced, from 0.256 ml/cm3/min to 0.059 ml/ cm3/min (blood ow ratio 0.23). Three months later the tumor exhibited regrowth.

Supplementary Files
This is a list of supplementary les associated with this preprint. Click to download. manuscriptsupplementary.pdf