Dynamic biomarker and imaging changes from a phase II study of pre‐ and post‐surgical sunitinib

To explore translational biological and imaging biomarkers for sunitinib treatment before and after debulking nephrectomy in the NeoSun (European Union Drug Regulating Authorities Clinical Trials Database [EudraCT] number: 2005‐004502‐82) single‐centre, single‐arm, single‐agent, Phase II trial.


Introduction
Renal cell carcinoma (RCC) accounts for~3% of adult malignancies [1]. Antiangiogenic therapies have improved the prognosis of patients with metastatic RCC (mRCC), and tyrosine kinase inhibitors (TKIs) such as sunitinib, may now be integrated with surgical treatment and other systemic treatments such as immunotherapy to optimise outcomes for patients with RCC [2,3].
Several trials of neoadjuvant therapy for RCC and mRCC have recently been completed or are ongoing [4][5][6][7] (Table 1), and there is increasing evidence of the safety, effectiveness and utility of this approach [4,8]. Recently, the Cancer du Rein Metastatique Nephrectomie et Antiangiogéniques (CARMENA) and Immediate Surgery or Surgery After Sunitinib Malate in Treating Patients With Metastatic Kidney Cancer (SURTIME) [2,4] trials have shown that patients with International Metastatic RCC Database Consortium (IMDC) intermediate-and high-risk RCCs are best treated with upfront TKIs rather than up-front cytoreductive nephrectomy. It is hypothesised that this approach allows the biology of the disease to be unveiled, sparing patients with aggressive RCCs from surgery and its associated morbidity [9].
One purported advantage of neoadjuvant therapy in the context of mRCC is to reduce tumour volume and facilitate surgical resection, and in cases of caval tumour thrombus (TT) this may be particularly beneficial to potentially downstage the TT to facilitate less extensive surgery. This is being explored currently in the NAXIVA trial (NCT03494816, a study of preoperative axitinib to reduce extent of tumour venous thrombus in renal cancer with venous invasion). Neoadjuvant treatment may also allow imaging and molecular assessment of response to guide future systemic therapy.
In the present study, sunitinib was chosen as the targeted agent because it was the most extensively investigated targeted agent at the time of the study, is approved as firstline treatment for mRCC, and is considered a suitable agent for neoadjuvant treatment given its ability to shrink primary tumours [2,10]. A short duration of preoperative sunitinib was planned to devascularise the tumour without having a major impact on tumour volume. The objectives of the present study were to evaluate the treatment's efficacy, safety, and tolerability, and to identify potential markers of response using imaging and translational studies.

Study Design
The NeoSun (European Union Drug Regulating Authorities Clinical Trials Database [EudraCT] number: 2005-004502-82) was a single-centre, single-arm, single-agent, Phase II trial. National research ethics committee approval was obtained (REC ref: 09/HO304/69) and written informed consent was provided by the study participants. The original design was a two-stage study with 18 patients in the first stage and 17 patients in the second stage. The study was terminated due to slow recruitment in December 2013 after 16 patients had been recruited. Data collection continued until December 2015. The trial schema is shown in Fig. S1.

Trial Treatment
Sunitinib was administered orally at 50 mg once daily starting 14 days before nephrectomy was scheduled and taken for 12 days, with a 2-day break before surgery. Patients were recommenced sunitinib when medically fit (minimum 15 days postoperatively), continuing on a 4-week on, 2-week off repeating cycle. This was continued until progressive disease

Study Objectives
Primary objective: • To determine the anti-cancer activity of sunitinib when given before and after nephrectomy to previously untreated patients with mRCC.
Secondary objectives: • To describe the toxic effects of sunitinib given before and after nephrectomy to previously untreated patients with mRCC.
• To evaluate, pathologically, evidence of sunitinib activity in the primary RCC lesion.
• To correlate pathological and clinical response. • To correlate changes in blood and tumour biomarkers with treatment response.
• To evaluate changes in structural and dynamic contrastenhanced MRI (DCE-MRI) of the tumour before and after 12 days of Sunitinib therapy.
Exploratory objectives: • To identify molecular and imaging biomarkers of response.

Study Patients
Between October 2010 and January 2014, 22 patients were screened for eligibility, and 16 were enrolled. Patient characteristics are shown in Table 2. Key inclusion criteria were histopathologically confirmed clear cell RCC with metastases, judged by the treating clinician to potentially derive benefit from sunitinib. Key exclusion criteria included previous treatment for mRCC and contra-indication to MRI. Eligible patients underwent screening procedures before treatment with sunitinib.
Cycle 2 Day 1 (C2D1) procedures took place ≥15 days postoperatively and the patient re-started study drug that day. At the end of the study or at withdrawal for toxicity/ progression or other reasons, patient care and follow-up was as per local standard procedure. Patients who withdrew for reasons other than disease progression were assessed (including RECIST) at least 3-monthly. Survival data for all patients were provided 3-monthly.

Surgery
The operating surgeon decided on the surgical approach. Transfusion thresholds were at the discretion of the surgical team (typically haemoglobin <80 g/L unless a history of cardiovascular disease). Complications were recorded according to the Clavien-Dindo classification [12].

Evaluation of Response
Imaging (Including Imaging Biomarker Assessments) The RECIST version 1.1 [11] was used to assess tumour response using contrast-enhanced CT scans (thorax, abdomen, and pelvis), performed at C1D1 (baseline), on restarting treatment after surgery (C2D1), then 12-weekly until disease progression. For endpoints based on RECIST, measurements excluded the primary RCC target lesion (as this was resected) and included longitudinal measurements after surgery. MRI was performed at baseline and C1D12, including multiplanar anatomical T 1 and T 2 -weighted imaging, and physiological sequences including DCE-MRI using a 1.5-T Discovery MR450 system (GE Healthcare, Waukesha, WI, USA) using an eight-channel cardiac array coil.
The DCE-MRI imaging data were acquired using a threedimensional (3D) FSPGR sequence in coronal oblique orientation with the following parameters: TE 1.6 ms; TR 3.9 ms; flip angle 18°; FoV 35 × 35 cm 2 ; slice thickness 5 mm; acquisition matrix 160 × 160 × 20-30; receiver bandwidth AE41.67 kHz; 0.5 Nex (elliptical k-space coverage); parallel imaging (ASSET) factor 2; temporal resolution 4.3-6.4 s; 94-140 dynamic phases; total scan time 10 min. The number of slices (and consequently the temporal resolution) was varied in order to image the entire tumour volume. 0.1 mmol/kg of gadolinium-tetraazacyclododecane tetraacetic acid (Gd-DOTA; Dotarem, Guerbet, Paris, France) was administered during the dynamic series at a flowrate of 3 mL/s. The T 1 mapping and DCE-MRI data were processed in MIStar (Apollo Medical Imaging Technology, Melbourne, Victoria, Australia). Each dataset was pre-processed using a 3 × 3 median filter. The T 1 mapping and DCE datasets were co-registered both within and between the datasets to remove (as far as possible) spatial mis-registrations caused by motion. Registration used a mutual information metric to cope with the variations in contrast between images within the DCE and T 1 map series. Where possible a 2D rigid-body registration was applied focussed on the region of interest. In some cases the 2D approach failed and instead a 3D rigidbody registration used; occasionally neither approach worked, and a deformable registration was necessary. Registration was performed as a batch, with each image registered pairwise to the same reference image where possible, although in some cases a separate reference image was used for specific phases that failed to register correctly. In all cases the images were viewed as a movie after registration to confirm the registration success before proceeding to the DCE analysis. Together the T 1 mapping and DCE datasets were used to calculate dynamic maps of Gd concentration, which was fitted using the Tofts model 10508281 with a population average arterial input function to calculate maps of the transfer constant K trans [25].
Normal kidney, whole tumour, cystic/necrotic and solid tumour regions of interest (ROIs) were outlined by an experienced radiologist using manual segmentation using ImageSetViewer Software, version 1.7 (University Health Network, Toronto, ON, Canada) in order to generate and compare 3D renderings of tumours and surrounding normal tissues before and after treatment. Segmentation was performed on the coronal T 2 -weighted FRFSE sequence, these outlines were then directly transposed onto the coronaloblique DCE early post-contrast sequences using the ImageSetViewer Software, with minor manual adjustment or ROIs being performed as required. The segmentation was undertaken blinded to the clinical outcome and RECIST measurements.

Blood-Based Biomarker Assessments (vascular endothelial growth factor [VEGF])
Blood samples were collected on C1D1 and C1D12. Whole blood (9 mL) was collected by standard venesection into Vacutainer tubes (Becton Dickinson, Franklin Lakes, NJ, USA). Blood was allowed to clot by leaving it undisturbed at room temperature for 30 min. Samples were then centrifuged at 1500g for 10 min in a refrigerated centrifuge. Serum was collected and stored in 0.5 mL aliquots at −80°C until assay. Immediately before assay, samples were thawed on ice and VEGF-A and VEGF-C proteins were measured by enzymelinked immunosorbent assay (ELISA; Human VEGF-A and VEGF-C-ELISA kits; R&D Systems, Minneapolis, MN, USA), according to the manufacturer's instructions.

Statistical Analysis
Statistical analyses were mainly descriptive on clinical endpoints (Appendix S1).

Patient Characteristics and Outcomes
Demographics A total of 16 subjects were registered between 29 Oct 2010 and 15 Jan 2014; however, two patients withdrew before starting treatment. Patient characteristics are shown in Table 2. Fig. S2. A total of 14 patients received pre-surgery cycle 1, with 71% (10/14) taking the planned 12 doses. All 14 patients underwent total nephrectomy and 13 recommenced sunitinib postoperatively. The mean (range) number of post-surgery cycles received was 12 . There was no correlation between number of pre-surgery doses and any of the primary or secondary endpoints.

Clinical Outcomes
The primary endpoint was not reached due to premature closure of the trial, but clinical outcome data are described in the Supplementary Results in Appendix S1.

Safety and Toxicity of Sunitinib
Safety and toxicity of sunitinib was consistent with previously reported outcomes (Table S1). No unexpected surgical outcomes or complications were reported (Table S2). No surgical delays occurred.

Imaging Endpoints
Changes in DCE-MRI of the primary tumour were assessed before and after 12 days of sunitinib therapy (Fig. 1). A total of 12 patients were evaluable and representative images are shown in Fig. 1A. All patients had RECIST-defined stable disease in their primary tumour after 12 days of sunitinib, measured using MRI. The mean (range) percentage change in primary tumour size measured using MRI was −6.01 (−16.6 to +4.3)%, with three patients having a change of >10% despite the short treatment duration. The mean total tumour volume reduction was 20.0% (90-1213 mL; P = 0.001) and the mean total volume reduction within the solid portions of the tumour was 24% from a mean (range) of 388 (99-903) to 295 (81-749) mL (P < 0.001). A slight increase in the mean volume of the normal ipsilateral kidney by 5.5 mL (3%; P = 0.039) was observed after 12 days of sunitinib therapy but is unlikely to be clinically significant and lies within the margin of measurement error. MRI derived K trans (a measure of capillary permeability) also significantly reduced within the solid portions of the tumour by a mean of 28% (P = 0.006; Fig. 1B). Additionally, patients with a higher percentage of solid tumour components at baseline showed a greater reduction in overall tumour volume (Spearman rank correlation coefficient ρ = 0.69, P = 0.02). There was some indication (correlation coefficient ρ = 0.4, P = 0.15) a higher baseline K trans in the solid tumour correlated to a larger solid tumour volume reduction, within the small sample size. However, none of these parameters correlated with OS or PFS or overall response rate (ORR).

Pathological Response
A total of 13/14 patients had evaluable tissue for analysis of necrosis using IHC before treatment with 12 days of sunitinib therapy, and 14/14 patients had evaluable tissue after therapy. The mean (range) necrosis before and after treatment was significantly increased from 9.6 (0-60)% to 29.6 (0-75)%, respectively (P = 0.024; Fig. 2A   these significantly correlated with OS or ORR. Larger percentage tumour volume reduction after 12 days of sunitinib therapy (measured using MRI) correlated with smaller baseline percentage necrosis (coefficient = −0.51, P = 0.03).

Ki67
A total of 13/14 patients had evaluable tissue for analysis of Ki67 using IHC before and after treatment with sunitinib for 12 days. The mean (range) percentage of cells expressing Ki67 significantly decreased after treatment from 21.2 (4.0-45.3)% to 3.6 (1.5-8.2)% (P = 0.037; Fig. 2B). The mean (range) change in the percentage of cells expressing Ki67 before and after treatment was 53.7 (15.8-91.7)%. Neither of these significantly correlated with OS or ORR.

Blood-Based Endpoints
A total of 12/14 patients had evaluable serum samples for VEGF-A and VEGF-C measurement both before and after 12 days of sunitinib therapy. The mean (range) VEGF-A levels before and after treatment were 539 (122-1175) and 885 (229-1590) pg/mL, respectively. The mean (range) difference between VEGF-A levels before and after treatment was 346 (−302 to −1270) pg/mL. Neither baseline nor C1D12 VEGF-A levels, nor the difference between them, significantly correlated with OS.
The VEGF-C levels were significantly reduced after 12 days of sunitinib therapy, at a mean (range) of 880 (353-1219) vs 638 (136-1038) pg/mL (P = 0.042; Fig. 2E). The mean (range) difference between VEGF-C levels before and after treatment was −242 (−808 to −28) pg/mL. Whilst the change in VEGF-C levels did not significantly correlate with OS,  Fig. 2 The molecular effect of pre-surgical sunitinib treatment (for 12 days). Intratumoral necrosis (A), proliferation measured using Ki67 (B), microvessel density measured using CD31 (C), and VEGF-C levels (F) were significantly different before and after treatment with 12 days of sunitinib. Pre-treatment microvessel density also correlated with objective response (measured on CT using RECIST version 1.1) (D). Pre-treatment VEGF-C levels greater than the median correlated with improved overall survival (F). CR, complete response; PR, partial response; SD, stable disease.

250
© 2021 The Authors. BJU International published by John Wiley & Sons Ltd on behalf of BJU International median OS was significantly improved in patients with VEGF-C levels ≥median compared with VEGF-C levels ≤median levels (29.3 months vs not reached, respectively; P = 0.020; Fig. 2F). This was the same whether levels of VEGF-C were taken before treatment or after treatment.

Discussion
Unfortunately, the primary endpoint of the present study was not reached due to the study closing prematurely as a result of changes to the standard of care treatment for mRCC. However, the study revealed important exploratory data that warrants further study and, within the small sample size, presurgical treatment of patients with mRCC with sunitinib was safe. Whilst the study population is small, the demographics of the population are similar to other neoadjuvant studies in mRCC [4][5][6]. Importantly, consistent with other studies, presurgery TKIs did not delay surgery or increase complications intra-or postoperatively [4][5][6].
Our present study differs from many others in that we gave a short TKI pre-surgical treatment course (12 days vs 6-18 weeks [4,5,13,14]). Given our short treatment course, chosen to devascularise the tumour rather than have any major effect on tumour size, it is interesting that MRI showed a tumour size decrease in all patients, albeit of <10% in the majority of patients. In comparison, others have shown a median tumour volume reduction of 14.4% with 12-14 weeks of preoperative pazopanib [13], and 32% with a 6-12 week course of preoperative sunitinib [5]. This suggests that a prolonged preoperative treatment course may be of limited additional downstaging benefit, although further work is required to explore this given the small numbers in our present study.
However, in our present study, it should be noted that 58.3% and 91.7% of patients treated had a CR/PR and clinical benefit respectively, which is higher than the reported response rates to sunitinib (~17% and 76%, respectively) [10]. A recent pooled study of 461 treatment-naïve patients with their primary tumour in situ reported primary tumour response rates of~28% [15]. Our present higher response rate is likely to reflect patient selection within the small sample size. Notably, newer TKIs such as axitinib, which are more potent inhibitors of the VEGF receptor (VEGFR) may have higher response rates, with one study assessing pre-surgical axitinib reporting a PR rate of 45.8% [16].
Of note, tumours with a higher percentage of solid tumour components at baseline showed a greater reduction in overall tumour volume. Larger percentage tumour volume reduction after 12 days of sunitinib therapy was also correlated with a smaller baseline percentage necrosis on pathology. Additionally, pre-treatment microvessel density was significantly higher in patients whose best response was CR or PR. These findings are important as a previous study using 4-6 weeks pre-surgical sunitinib found a modest reduction in tumour size in 83% patients, allowing a partial nephrectomy in some tumours where this was not initially an option [5]. Our present study suggests that patients more likely to have tumour volume reductions with sunitinib could potentially be selected based on baseline pathological and MRI characteristics, although this hypothesis requires validation.
The NeoSun aimed to identify early molecular biomarkers of response to sunitinib treatment. We observed that VEGF-C levels above the median (at both baseline and after 12 days treatment) correlated with survival, which may indicate that tumours secreting higher VEGF levels are more likely to respond to sunitinib, consistent with sunitinib's mechanism of action. Others have also found changes in stromal cellderived factor-1 and soluble VEGFR-1 during neoadjuvant sunitinib treatment to be significantly associated with OS [17]. These biomarkers are regulated by hypoxia and may have a role in guiding sunitinib therapy; tumours with a hypoxic signature are known to have a more aggressive clinical course [18].
Tumour imaging and molecular changes observed were consistent with sunitinib's known mechanism of action. Reduction in K Trans and decreased size of the tumours, were consistent with the reductions in microvessel density (measured by CD31) Ki67, and the increase in necrosis. Decreased microvessel density has also been demonstrated with neoadjuvant sorafenib [19]. One previous study has assessed very early treatment response after 3 and 10 days and found no change in K Trans compared to the baseline examination [26]. However, the significant reduction in K Trans seen in the Neosun after 12 days is in agreement with the reduction observed by Hudson et al. [27] after 14 days. Reduction in K Trans has also been observed after 4 and 10 weeks of either sunitinib or pazopanib treatment and change of K Trans between weeks 4 and 10 was associated with the risk of disease progression after 6 months [28].
Baseline K Trans trended towards a positive correlation with volumetric response. However, K Trans was not correlated with PFS in our present study. This may be due to the small sample size in our study. Previous studies have shown a longer PFS in patients treated with sorafenib with a baseline K Trans above the median [29]. Similar findings have been reported for sunitinib, sorafenib and pazopanib, respectively, after longer treatment periods than in the Neosun [27,30,31]. Thus, K Trans may be a useful non-invasive biomarker to confirm biological mechanism of action at an early stage of treatment but requires investigation in larger studies to determine its utility at very early time points as a predictor of longer-term response.
Overall, the NeoSun confirmed that pre-surgical treatment with sunitinib may allow early assessment of treatment response. Further work is required to establish if any molecular or imaging parameters may robustly predict response to sunitinib (or other more potent TKIs such as axitinib), but both imaging and blood-based biomarkers have been identified that might have potential to aid clinical decision making. We are seeking to validate them within the NAXIVA clinical trial assessing the effect of a short course of neoadjuvant axitinib in locally advanced RCC (NCT03494816).

Supporting Information
Additional Supporting Information may be found in the online version of this article:   .  Table S1. Summary of toxicities. Table S2. Surgical outcomes and complications. Appendix S1. Supplementary methods and results.