Yttrium-90 radioembolization for colorectal cancer liver metastases: a prospective cohort study on circulating angiogenic factors and treatment response

Patients with unresectable and chemorefractory colorectal cancer liver metastases were enrolled in this prospective cohort study, the RADAR study (RADioembolization: Angiogenic factors and Response). This study was approved of by the institutional review board (medical ethical committee NL34970.041.11 protocol number 11-172/E). All patients gave written informed consent to participate in the study and for the study results to be published.

An overview of all pre-treatment and follow-up procedures is given in Fig. 1. All patients referred to our centre for ⁹⁰Y-RE were screened for eligibility with full medical history, standard laboratory evaluation (including but not limited to liver function tests) and ¹⁸F-FDG-PET/CT imaging (including diagnostic multiphasic contrast enhanced abdominal and thoracic CT). Consensus on treatment was reached by a tumour board consisting of physicians from the department of medical oncology, department of interventional radiology and department of nuclear medicine. If there was any doubt about possible resectable liver disease (e.g. single lesion or multiple lesions but limited to one liver lobe), a liver surgeon was consulted additionally.

Yttrium-90 radioembolization was performed by experienced interventional radiologists in concordance with international guidelines [15‐18]. Vascular anatomy was mapped during pre-treatment angiography. Subsequently, technetium-99 m macro aggregated albumin (^99mTc-MAA) particles were infused and the intrahepatic and possible extrahepatic distribution of those was evaluated with SPECT/CT imaging. In a second angiographic procedure the yttrium-90 microspheres (SIR-Spheres, SIRTeX Medical Limited, Sydney, Australia) were infused with the catheter in a similar position as during pre-treatment angiography. The injected activity was calculated using the body surface area (BSA) method in all patients. Distribution of yttrium-90 microspheres after treatment was evaluated with ⁹⁰Y-PET imaging.

Baseline blood sample measurements were performed prior to treatment. Follow-up blood sampling for angiogenic factors was performed at days 0 (immediately after injection of the microspheres), 1, 3, 7 and 30 after ⁹⁰Y-RE treatment. Blood samples consisted of EDTA and citrate samples. Platelet free plasma was obtained from the EDTA sample prior to rapid storage at −80 °C (within 1 h from sample collection).

Treatment response of liver lesions was evaluated with RECIST 1.1 on magnetic resonance imaging of the liver, including post-gadolinium series and diffusion-weighted imaging. MRI was scored by an independent radiologist specialized in abdominal imaging. In addition, whole body FDG-PET/CT was performed and read by an independent experienced nuclear medicine physician. Tumour response was evaluated for target lesions only, as well as for all liver lesions and for the whole body.

The following angiogenic factors were analysed: vascular endothelial growth factor (VEGF), hepatocyte growth factor (HGF), angiopoietin-2, basic fibroblast growth factor (FGF-b), platelet-derived growth factor (PDGF-BB), stromal cell-derived factor 1 (SDF-1a) and thrombospondin-1. Measurements were performed using an in-house developed and validated multiplex immunoassay based on Luminex technology (xMAP, Luminex Austin TX USA). Samples were incubated with antibody-conjugated MagPlex microspheres (Bio-Rad) for 1 h at room temperature with continuous shaking, followed by 1-h incubation with biotinylated antibodies, and 10-min incubation with phycoerythrin-conjugated streptavidin diluted in high-performance ELISA buffer (HPE, Sanquin, the Netherlands [19, 20]). Acquisition was performed with the Bio-Rad FlexMAP3D (Bio-Rad laboratories, Hercules USA) in combination with xPONENT software version 4.2 (Luminex). Data was analysed by 5-parametric curve fitting using Bio-Plex Manager software, version 6.1.1 (Bio-Rad). Controls were analysed to assure intra-temporal consistency of the results.

Overall survival was computed as the number of days from ⁹⁰Y-RE treatment until date of death. Seven patients were still alive at study closure and survival was censored for these patients at a common date (April 8, 2015). Progression-free survival of the liver was calculated as the number of days from ⁹⁰Y-RE until documented disease progression on imaging. Kaplan Meier estimates for survival intervals were calculated using SPSS. Comparisons of survival distributions between strata were done using a log rank test.

To estimate the association between the plasma levels of angiogenic factors, tumour response and days after treatment (longitudinal data or repeated measures), a linear mixed model was used on the logarithmic transformed values. Time points of blood sampling were entered into the model as a factor, liver response to treatment at 1-month follow-up was entered as a fixed effect (dichotomous as response: complete response (CR), partial response (PR) or stable disease (SD); no response: progressive disease (PD)), and random intercepts were used for the individual patients (independent data). An additional analysis was performed to analyse the influence of the received liver dose on the levels of angiogenic factors by adding this as a factor in the models. Liver dose was calculated as \( \frac{\mathrm{administrated}\kern0.5em \mathrm{activity}\kern0.5em \left(\mathrm{MBq}\right)}{\mathrm{treated}\kern0.5em \mathrm{liver}\kern0.5em \mathrm{volume}\kern0.5em \left(\mathrm{ml}\right)}\times 50 \). Model performance was evaluated with the Akaike information criterion (AIC values). Significant contribution of separate factors was tested with ANOVA (χ ²).

A total of 49 patients were included in our study. Seven patients were not suitable for ⁹⁰Y-RE due to uncorrectable extrahepatic deposition of ^99mTc-MAA on SPECT/CT (n = 4), unsuitable hepatic artery configuration (n = 1), rapid progression of disease (n = 1) and inadequate haemoglobin level (n = 1). A total of 42 patients were treated with ⁹⁰Y-RE and completed study follow-up.

Mean age was 62 years (female n = 13, male n = 29) (Table 1). Most patients (n = 30/42 i.e. 71 %) had liver only disease, while 12 patients had one or more tumour localisations outside the liver, mainly local lymphadenopathy (mainly in the hepatoduodenal ligament). Any measurable disease according to RECIST 1.1 outside the liver was termed extrahepatic disease (EHD). In 3 patients, the primary colorectal tumour had not been surgically resected, but treated with chemotherapy and/or radiation and was no longer metabolically active on ¹⁸F-FDG-PET imaging. All patients had received at least one regimen of systemic treatment. Systemic treatment was stopped upon progressive disease in 25 patients and due to toxicity in 5 patients. Twelve (out of 42) patients had declined further systemic treatment after having received at least one regimen. None of the patients were using systemic anti-tumour treatment (chemotherapy or monoclonal antibodies) while treated with ⁹⁰Y-RE. A limited number of patients had previously undergone treatment of liver lesions, mainly surgical resection. At the time of ⁹⁰Y-RE, tumour load in the liver was <25 % in 36 patients. ⁹⁰Y-RE was performed in one session for right and left liver lobe in 38 patients, in two session (right and left liver lobe consecutively) in two patients, and two patients had only the right lobe treated (n = 1 only tumour in right lobe, n = 1 infusion of microspheres stopped prematurely due to severe pain).

Proportions of responders were calculated for target lesions, whole liver and whole body (Table 2). Even though target lesions showed disease control in all patients at 1-month follow-up (either partial response (PR) or stable disease (SD)), 10/42 (24%) of these patients had progressive disease of the liver. This was due to the appearance of new lesions in 8 patients and due to unequivocal progression of non-target lesions in 2 patients, while all target lesions showed stable disease. A total of 5/42 patients (12%) were lost to imaging follow-up, mainly due to clinically progressive disease, hampering hospital visits for study purposes. At 3-month follow-up 17 patients showed disease control of target lesions, but considering the entire liver disease control was seen in 11 patients. Partial response of the liver at 6 months after treatment was seen in 3 patients. At the level of extrahepatic lesions, disease control rates were clearly lowest, i.e. 61, 21 and 2% at 1, 3, and 6-months of follow-up, respectively.

Median overall survival was 9.2 months (95% confidence interval (CI) 6.1–12.4). Median time to liver progression was 3.0 months (95% CI 2.8–3.3). A significant difference in time to liver progression was observed between patients with (n = 12) versus without EHD (n = 30) at baseline (1.4 versus 3.5 months; log rank test p = 0.001) (Table 3).

A total of 214 blood samples were analysed in 42 patients over 6 time points. In two patients, the blood sample directly after injection of the microspheres (+0 days) was not collected, due to symptoms of postembolization syndrome directly following ⁹⁰Y-RE treatment. Most missing values occurred at three and seven days after treatment (respectively 16 and 12 missing), as patients were at that moment often not able to visit the hospital due to complaints of fever and fatigue, as part of the postembolization syndrome. At 30-day follow-up, blood samples were not collected in 4 patients, 2 due to severe illness from probable radiation induced liver disease (REILD) and 2 were lost to follow-up.

Following ⁹⁰Y-RE, three of the investigated angiogenic factors demonstrated an overall increase in plasma levels, i.e. VEGF, HGF and Ang-2. Figure 2 shows the plasma values for these three angiogenic factors, stratified for patients with response (CR, PR or SD, n = 32) and those with progressive disease (n = 10) of the liver at 1-month follow-up. The linear mixed model analysis showed that time as a factor did not contribute to the model performance for FGF-b and SDF-1a, therefore excluding a significant change over time.

For VEGF, a slight overall rise in plasma level was observed at 7 days post treatment, though more pronounced in patients with progressive disease at 1 month after ⁹⁰Y-RE. This time effect did contribute to the model performance; however, there was no statistically significant difference between the model with and without response as a fixed effect (p = 0.28). Liver dose did not contribute to the model performance either.

Plasma levels of HGF demonstrated a rise promptly after injection of the microspheres (blood drawn from the still indwelling sheath in the femoral artery). This holds true for both responders and non-responders. Yet, plasma levels of non-responders showed another peak, i.e. at 3 and 7 days after treatment, which was significantly different from responders (p = 0.007). Angiopoietin-2 levels were also statistically significantly higher at 3 and 7 days after treatment for patients with rapidly progressive disease, i.e. liver progression at 1-month follow-up, compared to those with disease control (p = 0.01). Model performances for HGF and angiopoietin-2 did not improve on adding received liver dose as a factor.

In addition, time variation of angiogenic factors for two patient subgroups were analysed, i.e. (1) patients with extrahepatic disease versus patients without extrahepatic disease (patients with liver only disease) (Fig. 3) and (2) patients previously treated with bevacizumab versus those who had not received bevacizumab (Fig. 4). There were no statistically significant differences between the models with and without respectively EHD or use of bevacizumab as a fixed effect.

Furthermore, a statistically significant difference in median overall survival was demonstrated between early responders and non-responders, of 341 (224–458) days versus 160 (122–198) days respectively (p value 0.003) (Table 3). There was no difference in baseline value of HGF and Ang-2 between early responders and non-responders. No statistically significant differences were seen between the 10 patients with rapidly progressive disease at 1-month follow-up and the 32 patients with CR, PR or SD at 1-month follow-up with respect to a number of characteristics that can be assumed to be important influences on prognosis and response, i.e. liver tumour involvement, performance score, presence of extrahepatic disease and received liver dose (Table 4). The difference in Ang-2 levels at 3 and 7 days after treatment between responders and non-responders is still statistically significant when comparing patients based on response at three months post treatment.