Mode of progression after radioembolization in patients with colorectal cancer liver metastases

Background Radioembolization is an established treatment modality in colorectal cancer patients with liver-dominant disease in a salvage setting. Selection of patients who will benefit most is of vital importance. The aim of this study was to assess response (and mode of progression) at 3 months after radioembolization and the impact of baseline characteristics. Methods Three months after radioembolization with either yttrium-90 resin/glass or holmium-166, anatomic response, according to RECIST 1.1, was evaluated in 90 patients. Correlations between baseline characteristics and efficacy were evaluated. For more detailed analysis of progressive disease as a dismal clinical entity, distinction was made between intra- and extrahepatic progression, and between progression of existing metastases and new metastases. Results Forty-two patients (47%) had extrahepatic disease (up to five ≥ 1 cm lung nodules, and ≤ 2 cm lymph nodes) at baseline. No patients showed complete response, 5 (5.5%) patients had partial response, 16 (17.8%) had stable disease, and 69 (76.7%) had progressive disease. Most progressive patients (67/69; 97%) had new metastases (intra-hepatic N = 11, extrahepatic N = 32; or both N = 24). Significantly fewer patients had progressive disease in the group of patients presenting without extrahepatic metastases at baseline (63% versus 93%; p = 0.0016). Median overall survival in patients with extrahepatic disease was 6.5 months, versus 10 months in patients without extrahepatic disease at baseline (hazard ratio 1.79, 95%CI 1.24–2.57). Conclusions Response at 3-month follow-up and survival were heavily influenced by new metastases. Patients with extrahepatic disease at baseline had a worse outcome compared to patients without.


Background
Approximately 45% of colorectal cancer patients develop metastases [1,2]. Without treatment, the median overall survival for colorectal cancer patients with hepatic metastases (mCRC) is only 4.5 months [3]. The liver is the most common site of metastasis: up to 30% of mCRC patients develop hepatic metastases [4,5]. Radioembolization is a loco-regional treatment option for unresectable, systemic therapy-refractory patients with liver-only or liver-dominant disease [6,7]. Intra-arterial administration of radioactive microspheres is proven to be safe and effective [8]. Microspheres (approximately 30 μm) are loaded with the radioactive isotope yttrium-90 ( 90 Y) or holmium-166 ( 166 Ho) and injected through a microcatheter in the hepatic artery [9]. For the treatment of metastatic colorectal cancer, 90 Y-resin microspheres (SIR-Spheres®, Sirtex) are FDA-and CE-approved. 90 Y -glass microspheres (TheraSphere®, BTG/Boston Scientific) and 166 Ho microspheres (QuiremSpheres®, Quirem) are CE-approved for this indication, not FDA-approved. The injected microspheres embolize the microvasculature surrounding the tumor and emit high-energy betaradiation. The normal liver parenchyma is largely spared since healthy liver tissue is mainly supplied by the portal vein [10][11][12].
Although assessment of metabolic response has proven added benefit over anatomic response, not being hampered by, i.e., the presence of intra-tumoral necrosis and cystic changes after treatment [13,14], response of radioembolization in mCRC patients is still mostly evaluated by the Response Evaluation Criteria in Solid Tumors (RECIST) [15][16][17]. When using these criteria, the results of most clinical studies in metastatic (liver) disease are modest, with many patients experiencing early progressive disease [18][19][20][21]. Optimized treatment planning could improve response rates [22,23], but selecting patients who will benefit most is another vital aspect. An important criterion in patient selection is the definition of liverdominant disease. The extent of extrahepatic disease we are willing to accept is under constant debate at tumor board meetings in our center, but clear guidance is currently missing, due to the lack of data on this matter. Other prognostic factors that are known to influence response after treatment with radioembolization are (among others) KRAS status, primary tumor location, percentage tumor involvement, and pre-treatment CEA level [19,24,25]. These factors could possibly be used in patient selection as well.
The aim of this study was to assess the impact of baseline characteristics on changes in intra-and extrahepatic mCRC disease from baseline to 3 months after radioembolization, across all currently available radioembolization treatment modalities.

Patient selection and study design
A total of 129 chemorefractory, unresectable mCRC patients were treated with radioembolization at our institution between August 2009 and January 2017, predominantly as part of the HEPAR-2 (Holmium Embolization Particles for Arterial Radiotherapy II) [26], or RADAR trial (RADioembolization: Angiogenic factors and Response) [22]. The studies were conducted in accordance with the institutions' Medical Ethical Committee and informed consent was obtained from the patients treated in the HEPAR-2 and RADAR studies before inclusion. For the other patients that were treated in routine clinical practice and also included in the current retrospective analysis, the need for informed consent was waived. Inclusion criteria for all patients regarding the presence of extrahepatic metastases or the primary tumor were similar: liver-dominant disease with a maximum of five lung nodules < 1 cm and lymph nodes < 2 cm. The presence of the primary tumor was not a contra-indication to treatment. Patients were included for response analysis in case CT and/or MRI scans were available at baseline and at (around) 3-month follow-up; all patients were included for survival analysis. Patients were treated with 166 Ho-microspheres (n = 24) (all as part of the HEPAR 2 study), glass 90 Y-microspheres (n = 20), or resin 90 Y-microspheres (n = 46). Imaging was performed 3 months after treatment (i.e., whole-liver or lobar treatment in one session). In case of sequential lobar treatment, imaging was performed 3 months after the last lobar treatment.

Radioembolization
The prescribed activity for the patients that were treated with glass 90 Y-microspheres was calculated according to the Medical Internal Radiation Dose (MIRD) method, with a desired absorbed dose of 80-120 Gy, according to the instructions for use [34][35][36]. Visual and quantitative assessment of 99m Tc-MAA distribution is weighted in this decision, also considering whole liver treatment in one session or sequentially. For the patients that were treated with resin 90 Y-microspheres, the body surface area (BSA) method was used. The injected activity for 166 Ho-microspheres was calculated based on the MIRD method with an aimed whole-liver absorbed dose of 60 Gy [37].

Response assessments
Two blinded readers independently performed measurements for tumor diameter on abdominal contrastenhanced CT or MRI at baseline and 3-month follow-up, using the same modality at both time points, according to RECIST version 1.1 [17]. In case no consensus was reached, a third reader gave the final call. Finally, inter-observer variability between the two raters was assessed.
Response at 3 months was dichotomized as disease control (i.e., complete or partial response (CR or PR) and stable disease (SD)) or progressive disease (PD). For a more detailed assessment of mode of progressive disease, a further subdivision was made in four categories: growth of intrahepatic metastases, growth of extrahepatic metastases, new intrahepatic metastases, and new extrahepatic metastases. All extrahepatic metastases were taken into account, regardless of their size.

Statistical analyses
Standard descriptive statistics were used to display patient demographics and summarize response measures. Cohen's kappa was used to determine agreement. Chisquare was used to test for differences in whole body response classification. Firth's logistic regression was used to explore associations between baseline characteristics and mode of progression. This type of analysis was chosen to correct for small-sample bias [38]. The analysis for the association between extrahepatic disease at baseline and disease progression was adjusted for the following possible confounders: time from diagnosis of metastases to treatment, primary tumor in situ, KRAS mutation vs wild type, and number of lines of previous systemic treatment (one versus two or more). The analysis for the association between type of microsphere used and disease progression was adjusted for the following possible confounders: age, time from diagnosis of metastases to treatment, primary tumor in situ, KRAS mutation vs wild type, number of lines of previous systemic treatment (one versus two or more), and presence of extrahepatic disease. Univariable survival analysis by the Kaplan-Meier method was used to estimate median overall survival (OS) in all treated patients. A Cox proportional hazards model with Firth's correction was used to test for differences in survival between patients with and without extrahepatic disease at baseline. All analyses were performed using R version 3.6.2 for Windows. We report effect estimates with associated 95%CIs and corresponding two-sided p values.

Patient demographics
Of the total cohort of 129 treated patients in our institution, 39 patients (30%) did not have 3-month follow-up imaging available because of the following reasons: follow-up imaging in other hospitals (n = 5), only follow-up imaging at 1 month post-treatment (n = 21), only response evaluation using 18 F-FDG PET (with no accompanying contrast-enhanced CT) (n = 5), clinical progression (n = 5), no follow-up imaging available (n = 2), and RFA artifacts (n = 1). The remaining 90 patients had either CT (n = 67, 74%) or MRI (n = 23, 26%) images available at baseline and 3-month follow-up.

Inter-observer variability
Discordant conclusions were drawn in five patients, for whom the third rater gave the final call. The level of agreement in RECIST categories was adequate with a Cohen's kappa of 0.895 (95% CI 0.805-0.985), p < 0.001.
There was no significant difference in response between the three types of microspheres used: compared to 90 Y resin microspheres, the odds ratios for progressive disease with 90 Y glass and 166 Ho were 1.11 (95%CI 0.32-4.53) and 0.67 (95%CI 0.22-2.14), respectively ( Table 2).

Discussion
This study shows that a large proportion of end-stage mCRC patients have progressive disease after radioembolization due to the development of new metastases, and to a lesser extent due to the growth of existing metastases. The presence of extrahepatic disease at baseline significantly increases the chance of early progressive disease at 3 months, especially the development of new metastases. Moreover, patients with extrahepatic metastases at baseline had a significantly worse overall survival.
At baseline, 48% of our study population was diagnosed with extrahepatic metastases. This is in line with other studies in which 35-77% of the included patients had extrahepatic metastases at baseline [18][19][20][39][40][41][42][43][44]. We found a difference in median OS with and without the presence of extrahepatic metastases at baseline, respectively 7 versus 10 months (p = 0.0018). Several other studies with a comparable patient population also found that extrahepatic disease was a predictor of survival after radioembolization [24,[45][46][47][48][49]. Other known prognostic factors are tumor load, baseline CEA level, and location (left-versus right-sidedness) of the primary tumor [24,49,50]. In our study, only location of the primary tumor showed a clear trend for significance, with the odds ratio for progressive disease being 3.88 (95%CI 1.00-25.75) for patients with a right-sided primary tumor versus patients with a left-sided primary tumor.
Genetics and biomarkers are more and more recognized as prognostic factors. We investigated the possible role of CEA, since this was associated with poorer survival after radioembolization in multiple studies [19,24,51]. However, just as in the study of Sofocleus et al., in our study, no significant correlation between pretreatment CEA level and disease progression was found [19]. Patients with KRAS mutation generally have a worse prognosis after radioembolization than patients with KRAS wild-type status [19,24,25,52]. In our study, although not significant, the odds ratios for all types of progressive disease showed a clear trend for a worse prognosis for patients with KRAS mutation versus patients with KRAS wild type (Table 2).
In The Netherlands, indications for radioembolization include liver-dominant, irresectable, systemic therapyrefractory disease. Patients with significant extrahepatic metastases are not considered eligible, but patients with stable, limited extrahepatic disease (defined by the Dutch  National Healthcare Institute as a maximum of 5 lung nodules < 1 cm and lymph nodes < 2 cm) are eligible [53]. This criterion was also used in the patients in this study. The SIRFLOX, FOXFIRE, and FOXFIRE-Global (studying the added value of radioembolization to chemotherapy in first-line mCRC patients) used similar inclusion criteria with respect to extrahepatic disease [54]. In these studies, no difference in OS or overall progression-free survival (PFS) was observed [55]. One may argue that the large percentage of patients with extrahepatic disease in these studies (i.e., 36%) clouded the potential clinical benefit of radioembolization in a more stringent selected subset. In a subgroup of patients with right-sided primary tumors, the presence of extrahepatic metastases at baseline indeed proved to be a negative prognostic factor for OS, with a HR of 1.351 (95%CI 0.96-1.91) [50]. Importantly, these studies were performed in first-line refractory disease. This limits comparison with our study in a more advanced-stage population. Objective response (CR or PR) at 3 months after treatment was obtained in only 6% of our patients. This is in line with other studies in salvage mCRC patients, with reported response ranges of 6-24% [24,56,57]. Median OS in our study was 10 months, which is also in line with other studies in a comparable patient population [22,24,58].
A reason for the modest treatment results in our study might be the dosimetric models that were used: the BSA and MIRD methods. These methods can lead to underdosing [59,60]. A personalized treatment approach, as was used in the DOSISPHERE study in HCC patients, could have led to a much higher response rate [61]. The results of earlier studies on the dose-response relations in mCRC patients treated with 90 Y-resin or 166 Ho prove this point: a significant dose-response relationship was found in both studies [22,62]. Implementing the results of these studies in future patients, using an individualized treatment approach, likely will lead to a higher treatment accuracy.
In our study, response was evaluated using the anatomic criteria as defined by the RECIST guidelines. However, this can be hampered by the presence of necrosis, hemorrhage, and cystic changes [63]. Response assessment based on changes in functional metrics as determined on [ 18 F]-FDG PET/CT would be a better The added value of the present study to the existing knowledge on radioembolization in mCRC patients is the fact that the development of new metastases is the primary cause for progressive disease after treatment. Furthermore, the study shows that the development of new lesions, as well as progressive disease in general, is more common in patients with extrahepatic disease at baseline.
The current study also has several limitations. First of all, the sample size was small. Secondly, the retrospective setting was prone to selection bias. Since radioembolization was used in a salvage setting, outcome was likely muddled by the effect of other, previous therapies (Table  1). However, since patients were selected for radioembolization based on their chemo-resistant tumors, the contribution of this variation in our patient population on the outcome of our study was considered minimal. Third, all patients were discussed in a multidisciplinary tumor board before treatment. Based on available imaging, the primary tumor was assessed for stability and the extrahepatic disease load was assessed for extent, however, not for stability. Also, although radioembolization is nowadays often performed in a lobar approach, a large fraction of patients that we studied received wholeliver treatment. Whole-liver treatment was in large part dictated by study protocols. Furthermore, three types of microspheres were used in our dataset. The differences with regard to the embolic nature of the treatment, the specific activity of the microspheres, the administered activities, and the absorbed doses may have influenced the incidence of early progressive disease, and potentially also the mode of progression, although our analyses did not show a significant difference between microsphere types. Last, KRAS status was unknown in 42% of the patients, making the number of patients for the subgroup analyses for KRAS rather small. Proper selection of patients seems fundamental for the cost-effectiveness of radioembolization treatment. Future prospective studies in the salvage setting should therefore be conservative with regard to the acceptance of extrahepatic disease. Accurate baseline imaging, including FDG-PET, may aid patient selection [66]. This will avoid futile treatments and unnecessary toxicity. However, the effect of radioembolization in patients with extrahepatic disease should be evaluated in prospective studies comparing radioembolization with best supportive care, before a firm statement can be made about the exclusion of patients with extrahepatic disease from treatment. Also, considering the development of new lesions as the major cause of progressive disease, a study in the third line, comparing TAS-102 or regorafenib with and without radioembolization, would be interesting. The study of Hendlisz et al. showed that radioembolization combined with chemotherapy was safe and effective [58]. Based on the results of this study, chemotherapy in addition to radioembolization was therefore recommended in the refractory setting.
Proper selection and individualized dosimetry-based treatment planning should ultimately lead to improved treatment accuracy in mCRC patients.

Conclusions
In conclusion, response at 3-month follow-up and survival were heavily influenced by new intra-and extrahepatic metastases. Patients with extrahepatic disease at baseline had a worse outcome compared to patients without extrahepatic disease at baseline. Based on the results of this observational, retrospective study, extrahepatic disease may be considered a contraindication for treatment with radioembolization. Authors' contributions CR was involved in conception and design of the study, performed the analyses, wrote the manuscript, and did the response assessments. JJ was involved in the design of the study, helped write the manuscript, and also did the response assessments. MS was involved in the design of the study, helped write the manuscript, and was involved in a critical revision before it was decided to publish the article. SE was involved in the design of the study, helped perform the analyses, and was involved in a critical revision before it was decided to publish the article. MK was involved in the design of the study and was involved in a critical revision before it was decided to publish the article. OK was involved in the design of the study and was involved in a critical revision before it was decided to publish the article. IBR was involved in the design of the study and was involved in a critical revision before it was decided to publish the article. ML was involved in conception and design of the study, helped write the manuscript, and was involved in a critical revision before it was decided to publish this article. He was also consulted in consensus meetings regarding response assessment. The authors read and approved the final manuscript.

Funding
This study was supported by the Dutch Cancer Society (grant no. UU2013 −5865 for J.J.). The Department of Radiology and Nuclear Medicine of the University Medical Center Utrecht has received royalties and research support from Quirem Medical. The HEPAR I and II studies were sponsored by a grant from the Dutch Cancer Society and the Technology Foundation STW.

Availability of data and materials
The datasets used and/or analyzed during the current study are available from the corresponding author on reasonable request.

Ethics approval and consent to participate
All procedures performed in this study were in accordance with the ethical standards of the institutional and national research committee and with the 1964 Helsinki Declaration and its later amendments or comparable ethical standards.

Consent for publication
The institutions' Medical Ethical Committee provided a waiver for informed consent for this retrospective analysis.