Ablation versus resection for resectable colorectal liver metastases - Health care related cost and survival analyses from a quasi-randomised study

Background: The aim of this study was to compare healthcare related costs and survival in patients treated with microwave ablation (MWA) versus surgical resection for resectable colorectal liver metastases (CRLM), in patients from a quasi-randomised setting. Methods: The Swedish subset of data from a prospective multi-centre study investigating survival after percutaneous computer-assisted Microwave Ablation VErsus Resection for Resectable CRLM (MAVERRIC study) was analysed. Patients with CRLM (cid:1) 3 cm amenable to ablation and resection were considered for study inclusion only on even calendar weeks, while treated with gold standard resection every other week, creating a quasi-randomised setting. Survival and costs (all inpatient hospital admissions, outpatient visits, oncological treatments and radiological imaging) in the 2 years following treatment were investigated. Results: MWA (n ¼ 52) and resection (n ¼ 53) cohorts had similar baseline patient and tumour characteristics and health care consumption within 1 year prior to CRLM treatment. Treatment related morbidity and length of stay were signi ﬁ cantly higher in the resected cohort. Overall health care related costs from decision of treatment and 2 years thereafter were lower in the MWA versus resection cohort (mean ± SD USD 80 0 964 ± 59 0 182 versus 110 0 059 ± 59 0 671, P < 0.01). Five-year overall survival was 50% versus 54% in MWA versus resection groups (P ¼ 0.95). Conclusions: MWA is associated with decreased morbidity, time spent in medical facilities and healthcare related costs within 2 years of initial treatment with equal overall survival, highlighting its bene ﬁ ts for patient and health care systems. © 2022 Published by Elsevier Ltd.


Introduction
Thermal ablation using radiofrequency or microwave techniques is a low morbidity treatment alternative to surgical resection for colorectal liver metastases (CRLM). Its parenchyma-sparing aspect and easy applicability by minimally invasive treatment accesses lower major complication rates to around 8%, and shortens or eliminates in-hospital stay [1]. This further allows multiple retreatments in case of hepatic CRLM recurrence, occurring in 71% of patients treated for CRLM, regardless of the type of initial treatment [2]. Regarding oncological outcomes, data from wellpowered randomised clinical trials (RCT's) investigating noninferiority compared to surgical resection are still lacking, with ongoing trials being delayed in patient inclusion. Alternative trial designs such as population-based analyses adapting for selection bias confirm similar survival rates of around 70% and 50% after 3 and 5 years [3e5], and thermal ablation was incorporated into current guidelines as a first line treatment for oligometastatic CRLM [6].
Decreased morbidity, shortened length of hospital stay and similarity in oncological outcomes suggest that thermal ablation for CRLM might be beneficial to the health care system also in terms of costs. Previous studies investigating cost effectiveness of thermal ablation compared to surgical resection based on model parameters suggest that thermal ablation might indeed be a cost effective [7e9]. However, data comparing health care related costs after thermal ablation versus surgical resection from clinical data with patients amenable to both ablation and resection are currently lacking.
The aim of this study was to analyse health care related costs and overall survival in patients treated with thermal ablation versus surgical resection for CRLM amenable to both ablation and resection, in a Swedish patient cohort arising from a quasirandomised study design within a larger international multicentre study (MAVERRIC).

Study population
The MAVERRIC trial (Microwave Ablation VErsus Resection for Resectable Colorectal liver metastases) is an ongoing European multicenter prospective cohort study aiming to prove oncological non-inferiority (primary outcome: 3-year overall survival) of microwave ablation (MWA) versus surgical resection for curativeintent treatment of CRLM [10]. In this trial, patients with 5 CRLM of 30 mm, amenable for both surgical resection and percutaneous computed tomography (CT)-guided stereotactic MWA as decided by the local liver multidisciplinary team (MDT) conferences, were deliberately treated with percutaneous MWA. Patients with previous MWA for CRLM were excluded. Outcomes from the study group will be compared with results from a cohort of patients after surgical resection for CRLM extracted from a population-based patient registry [11]. Between December 2015 to November 2018, three recruiting tertiary liver centres in Stockholm Sweden, Groningen Netherlands and Bern Switzerland prospectively included 98 patients.
For the current analysis, the subgroup of patients included in the MAVERRIC trial in Stockholm Sweden was extracted and analysed separately. In this specific subgroup, patients were prospectively considered for inclusion into the MAVERRIC study only every second week, i. e. on even calendar weeks (MWA cohort). On odd calendar weeks, patients were not considered for inclusion into the MAVERRIC trial and were treated with surgical resection as per the current gold standard for treatment of CRLM (resection cohort). Patients theoretically meeting the inclusion criteria for the MAV-ERRIC trial on these weeks (i.e., both amenable to percutaneous MWA and surgical resection) and treated with resection were nevertheless prospectively registered in an anonymised data sheet. This particular inclusion pattern led to the formation of two prospective cohorts with a quasi-randomised setting, opening opportunities for outcome analyses. Fig. 1 shows the inclusion process and cohort selection. Ethical approval was obtained for the spin-off study from the MAVERRIC trial from the local ethics review board (2015/1453-31/4 and 2020-00787).

Treatment strategy and follow-up
Percutaneous stereotactic CT-guided MWA procedures were performed by multidisciplinary teams of surgeons and radiologists, with the primary responsibility being under the surgery department. The procedure has been described in detail previously [12,13]. In brief, CRLM were treated with local thermal ablation (microwave energy at 2,45 MHz), aiming to locally destroy all tumour tissue. Patients were under general anaesthesia and ventilated with high frequency jet ventilation to minimise breathing related liver motion to enhance the stereotactic treatment [14]. Ablation planning and stereotactic tumour targeting were performed using the CAS-one system (CAScination AG, Bern, Switzerland). Immediate post-interventional review overlay of pre-and post-ablation CT images allowed review of the completeness of the tumour ablation, with possibility of immediate re-ablation if necessary [13]. Patients were kept in hospital until pain levels were manageable with oral agents, food intake assured and urinary output secured. Radiological follow-up included liver CT or magnetic resonance imaging (MRI) every 3 months for the first year and every 6 months in the two following years (Fig. 1). No synchronous MWA and resection of the colorectal primary tumour was performed in the MWA cohort, since CT-guided ablations in the OR-suite was not an available option.
Surgical resection was performed as per local guidelines and standards of care at the Karolinska University Hospital in Stockholm Sweden, which is a tertiary referral centre for Hepato-Pancreatico-Biliary surgery. A laparoscopic or open access was used, with the goal of minimising surgical trauma and loss of liver parenchyma. No combined resection and ablation was performed in the studied cohorts. Patients were discharged to specialised rehabilitation units for prolonged rehabilitation care when appropriate as per the local standards. Radiological follow-up included liver CT or MRI every 6 months for the first two years and yearly imaging thereafter (Fig. 1).

Data collection
Baseline characteristics of patients, the primary colorectal cancer and of CRLM at the time of the index liver treatment (MWA or surgical resection) were extracted from patients' medical records. The appearance of CRLM was defined as synchronous if the diagnosis of CRLM was made prior to or during surgery for the primary colorectal tumour [15]. Complications related to the index treatment were classified according to Clavien-Dindo [16]. Intrahepatic recurrences and re-interventions, recurrence of primary tumour and distant metastases arising during follow-up were extracted.
To investigate baselines in patients' health care use, health care consumption (number and type of outpatient visits for any medical reason, number of inpatient hospital admissions for any medical reason) were collected for a time 12 months prior to patient inclusion at the liver MDT. Overall health care consumption was further collected from the time of patient inclusion at the liver MDT for 24 months onwards, including the index liver treatment. These included the number of primary tumour related outpatient visits to oncologists and surgeons, number of inpatient days for any hospital admissions, number of MDTs, diagnostic imaging for CRLM using CT, MRI or positron emission tomography-CT (PET-CT), number and kind of oncological treatments received (chemotherapy and radiotherapy). Cost data was collected in Swedish Crowns and converted into United States Dollars (USD) using the average conversion rate during the study period of patient inclusion.
Total costs were calculated from patient inclusion at the liver MDT and 24 months onwards, including all inpatient admissions for any medical reason during this time period, outpatient visits (surgery, oncology or palliation team) related to the colorectal cancer disease, all radiological imaging related to the colorectal cancer disease (CT, MRI and PET), oncological treatments (chemotherapy and radiotherapy) related to the colorectal cancer disease and costs for MDT conferences related to the colorectal cancer disease.
Costs for inpatient hospital admissions were obtained using Diagnosis-Related Group (DRG) and Cost Per Patient (CPP) measures. DRG is an international system for classifying patient visits to health care centres into groups of homogenous medical character and resource consumption, based on main diagnosis, secondary diagnoses, interventions performed, complications, age, time spent at the health centre and mode of discharge. CPP represent the actual cost for each contact with a health care provider in Sweden, including overhead costs. CPP was preferred in the current analyses, when available, since it gives a more precise assessment of actual costs than DRG. For all other occasions the DRG-based costs were extracted. Fixed remunerations reported by some rehabilitation units were used for assigning such inpatient costs.
For outpatient visits, DRG based costs were readily extractable from the electronic medical journal system Take Care (TakeCare, CompuGroup Medical, Stockholm, Sweden).
Costs for radiological imaging, MDT conferences and oncologic treatments were obtained by consulting economists in the different hospitals in the region of Stockholm, and calculated accordingly. The cost of oncological treatments including chemotherapy and radiotherapy was calculated from the costs associated to the administration of one treatment cycle and from the costs of actual drugs used.

Statistical analyses
Continuous data were analysed as means with standard deviations (SD) and medians with interquartile ranges (IQR), and categorical variables as absolute numbers. Differences between groups were analysed using the Wilcoxon rank-sum test, Pearson's chi-squared test or Fisher's exact test, as appropriate. Survival was defined as the time from index treatment until time of death from any diagnosis. Overall survival (OS) curves were obtained and probability of survival estimated with the Kaplan-Meier method and differences in OS analysed using the Log-rank test and the Wilcoxon (Breslow) test. The assumption of proportional hazards was investigated (testing of Schoenfeld residuals, adding timedependent interaction term into multivariable Cox model), yielding no evidence against the proportionality of hazards. Factors potentially influencing OS, chosen based on previous knowledge, were investigated using univariable and multivariable Cox proportional hazard models. Factors yielding a p-value of <0.2 in univariable analysis were included into multivariable analyses. The threshold for statistical significance was set to a <0.05. STATA/IC version 15.0 (StataCorp, 4905 Lakeway Dr, College Station, TX 77845, USA) was used for all statistical analyses.

Results
Between December 2015 and November 2018, 681 patients with CRLM were reviewed at the Stockholm regional liver MDT conference, of which 105 fulfilled the inclusion clinical criteria for the MAVERRIC trial. This led to inclusion of 52 patients into to the MAVERRIC trial on even calendar weeks, treated with stereotactic CT guided MWA, and 53 patients treated with surgical resection on odd calendar weeks, during the same time period (Fig. 1).
In the cohort of patients undergoing surgical resection for CRLM, 12 (23%) had laparoscopic liver resections, 6 (11%) laparoscopic converted to open resections and the rest were open liver resections. Twenty-seven (51%) were atypical local resections, 14 (26%) were segmental resections, 10 (19%) were right lobectomies and 2 (4%) were left lobectomies as the main resection type. Of the eleven patients treated for bilobar CRLM, 2 had a left lobectomy plus local resections, 3 had a right lobectomy plus local resections and 6 had multiple local and/or segmental resections. Sixteen patients (30%) had synchronous resection of the colorectal primary tumour at the time of CRLM resection, including 11 right hemicolectomies, 3 left hemicolectomies and 2 low anterior resections. Resection margins of the CRLM were negative in all resected liver specimens in 43 (81%) patients. In the MWA cohort, treatment efficacy (patients with no tumour residue at site of ablation at post ablation follow-up imaging at 3 months) was 85%.
Baseline characteristics of MWA and surgical resection cohorts are summarised in Table 1. Patient characteristics including comorbidity, primary tumour and CRLM characteristics, and oncological perioperative chemotherapy regimens were similar in both cohorts. There was no difference in the number of patients undergoing a "liver first" approach, i.e. having the primary tumour in situ at the time of the index liver treatment, in both cohorts. Baseline health care consumption within one year prior to the liver MDT conference were similar, except the number of non-oncology related outpatient visits, which was higher in the MWA cohort. There was no difference in the number of patients having their primary tumour resected within one year prior to the liver MDT conference ( Table 2).
Main peri-operative outcomes, i.e. health care consumption from liver MDT conference two years onwards, and costs are summarised in Table 2. Patients who underwent surgical resection had significantly more overall complications than the MWA cohort (60% versus 15%). Major complications of Clavien-Dindo grade III in the surgical resection group included two cases of pneumothorax or pleural effusion requiring drainage, seven of perihepatic fluid collections requiring percutaneous drainage, three cases of bile leakage requiring endoscopic drainage and one case of colovesical fistulation requiring re-operation in a case of synchronous resection of the colorectal primary tumour. Of the four patients with simultaneous resection of the primary tumour developing a major complication, only one was associated with the colonic resection. In the MWA cohort, one patient required pleural drainage after a haemothorax. The mortality in the surgical resection cohort was a case of liver failure (small-for-size liver insufficiency exacerbated due to infection) and kidney failure following right hepatectomy and local resections for bilobar CRLM. In the MWA cohort, a patient died on day 21 from a cardiac event after surgical reintervention due to a liver abscess fistulating into small bowel. No patient required admissions to rehabilitation centres following hospital discharge in the MWA cohort, leading to significantly enhanced rehabilitation admissions and related costs in the surgical resection cohort. While health care consumption was similar, total health care related costs from the liver MDT conference two years onwards were significantly lower in the MWA versus surgical resection cohorts (Fig. 2) ( Table 2). The main difference in costs were related to inpatient hospital admissions, including the index treatment admission, which were significantly higher in the surgical resection versus MWA cohort ( Table 2).
Median follow-up was 46 (IQR 37e63) months in the MWA cohort and 48 (IQR 30e58) months in the surgical resection cohort. Total liver recurrences of CRLM were 69% in the MWA cohort and 45% in the surgical resection cohort (3-year recurrence-free survival 33% versus 56%, P ¼ 0.014 Log rank test). In the MWA cohort, 27 (52%) patients had 40 further treatments for recurrent CRLM within the first 2 years, of which 36 were re-ablations with percutaneous MWA and 4 were resections. In the liver resection cohort, 14 (26%) patients had 16 further treatments for recurrent CRLM within the first 2 years (two patients had 2 more treatments), of which 13 were surgical resections and 3 were percutaneous ablations.
Any recurrence, including liver, extrahepatic or at primary tumour location, occurred in 32 (60%) of patients in the surgical resection cohort and 40 (77%) in the MWA cohort. Estimated 3-year disease-free survival was 28% in the MWA and 40% in the surgical resection cohorts, 5-year disease-free survival was 22% in the MWA and 37% in the surgical resection cohort (P ¼ 0.150 Log rank test).
Overall survival after index treatment was similar in both cohorts, with estimated 3-year OS of 75% (CI 61e85%) after MWA versus 70% (56e80%) after surgical resection. Estimated 5-year overall survival was 50% (35e64%) after MWA versus 54% (40e67%) after surgical resection (Log rank test P ¼ 0.947, Wilcoxon (Breslow) test P ¼ 0.831) (Fig. 3). Multivariable Cox regression analysis yielded the number of liver metastases at the time of patient inclusion as the only factor significantly influencing overall survival with a HR of 2.4 (CI 1.3, 4.6). The type of index treatment approach (MWA versus surgical resection) had no effect on survival, with a HR of 1.0 (CI 0.6, 1.7) ( Table 3).

Discussion
This study shows that while treatment with MWA or surgical resection for patients with CRLM lead to similar overall survival, total health care related costs are significantly reduced in the two first years after MWA treatment, with reduced treatment related complications and time spent in hospital. To the best of our knowledge, this is the first study analysing survival and costs from clinical data arising from a quasi-randomised setting, where MWA or surgical resection was allocated based on calendar week number, in patients qualifying for both percutaneous CT-guided MWA and surgical resection. This led to two comparable treatment cohorts with similar baseline patient-, tumour-and health care consumption characteristics.
No data from well-powered RCT's on oncological outcomes are available to date. Several attempts to conduct RCT's on thermal ablation versus surgical resection analysing oncological endpoints have been initialised, but have either been abandoned, or not concluded [17e19]. Reasons for this are based mainly on the large number of trial participants needed when designing clinical trials on meaningful oncological endpoints. The nature of investigated treatments often requires trial designs with involvement of more complex logistics with different medical specialities and clinical units, i.e. surgeons for surgical resection and radiologists for thermal ablation. Also, a transfer of curative treatment from surgery to interventional radiology can cause challenging resource shifting issues. Facing the current lack of data from RCT's, more recent population based analyses show that when adapting for confounders leading to inclusion of healthier patients for surgical resection and more comorbid patients for thermal ablation, long term overall survival is similar [3,5]. This was confirmed in the current analysis using prospective data, with survival rates corresponding to previously published rates of around 70% at 3 years and 55% at 5 years after MWA for potentially resectable CRLM [20] and the type of treatment not influencing survival in multivariable  analyses. For liver specific recurrence rates, conflicting data exist when comparing thermal ablation to surgical resection for CRLM, with higher liver recurrence rates after ablation reported in systematic reviews [21]. In the current analysis, liver specific recurrence free survival was shorter in the MWA group, with more patients undergoing re-treatments for recurring CRLM within 2 years. These consisted mostly of repeat percutaneous thermal ablations, which has been shown to be a valid low-morbidity option for repeat treatment of recurring CRLM [22]. As the target population for curative-intended therapy becomes older with more severe comorbidity, the concept of repeat but lowmorbidity treatments associated with short in-hospital stay is gaining importance [9]. Treatment related morbidity is well known to be decreased with thermal ablation compared to surgical resection for CRLM [21], which in turn is also relevant for patient survival [23]. Decreased morbidity, length of hospital stay and requirements for rehabilitation were confirmed in the current analysis, leading to decreased overall time spent in medical facilities. This underlines the hypothesis of reduced health care related costs and thus benefit the health care system when applying local thermal ablation as a first treatment choice for CRLM. Several studies investigated cost effectiveness, i.e. including quality-adjusted life years (QALY's), after thermal ablation versus surgical resection, based on estimation models using data extracted from previously  published studies, report conflicting results [7e9]. While earlier works favoured resection over ablation, and radiofrequency ablation (RFA) over MWA [7,9], the most recent study suggests enhanced cost-effectiveness with MWA over both RFA and surgical resection [8]. Another study based on clinical data comparing combined treatment strategies of thermal ablation and resection versus two-stage hepatectomy for bilobar CRLM favoured combined strategies including ablation in terms of cost effectiveness and survival [24]. In the current analysis based on prospectively collected clinical data in a quasi-randomised setting, reduced total health care related costs associated with MWA over resection were confirmed. This was mainly caused by reduced costs related to inpatient admissions, and significantly also due to the index treatment admission of MWA versus surgical resection. Despite using top-end computer-assisted navigation technology for percutaneous MWA treatment in a multidisciplinary team and a more expensive follow-up protocol. Hence, despite necessitating more repeat interventions related to liver specific recurrences, patients with CRLM treated with MWA spent less time in medical facilities including rehabilitation centres. This reduced time spent as inpatients, a decreased treatment related morbidity and similarity in overall survival support the hypothesis that quality of life might be enhanced with minimally invasive thermal ablation as the initial intervention, especially in an older patient cohort [25e27]. Data on quality of life was not collected as part of the MAVERRIC study and hence not available for analysis in the current patient cohorts. No high quality prospectively collected data on quality of life after thermal ablation of CRLM are available to date, which is why we decided against the estimation of QALY's based on data from published literature and thus a formal cost-effectiveness analysis. The comparison of quality of life data from ongoing RCT's are desperately awaited and will yield important information on patient reported outcomes in the initial treatment of patients with CRLM [18,19]. In the meantime, it might be assumed that the initial treatment of CRLM likely does not influence quality of life significantly in the longer term, since patients suffer the same chronic colorectal cancer disease. This leads to similar oncology related health care consumption and diseasefree survival as shown herein, and overall recurrences in around 70e75% of patients regardless of the initial type of treatment [2]. Hence, despite higher liver specific recurrences after MWA in the current analysis, shorter recurrence-free survival calculated after the first intervention probably reflects only marginally on the overall patient quality of life outcome [2].
The current study has several limitations. Even though the quasi-randomised setting led to two cohorts with similar baseline characteristics, a degree of inclusion bias certainly remained, e.g. due to inclusion of patients at the liver MDT conference and potential differences in the understanding of resectability and ablatability between different individuals. This element might exist even in true RCTs depending on inclusion criteria. While a discrepancy in the number of synchronous colorectal primary resection and liver treatments existed (n ¼ 16 in the resection cohort, n ¼ 0 in the MWA cohort), only one patient's complications were clearly related to the colorectal primary resection in the resection group. Also, complications rates have previously been shown to be similar with simultaneous liver and colorectal primary resections [28]. To allow combined treatments and reduce bias related to logistics, treatment options for entities like CRLM should optimally all be available in one facility. Lastly, due to the current analysis being a "spin- off" study from the MAVERRIC trial, the sample size was not predetermined for overall and recurrence free survival, potentially opening up for type 2 errors.
In conclusion, findings from this quasi-randomised trial suggest that treatment with MWA for patients with CRLM leads to similar long term overall survival rates as surgical resection, while being associated with significantly decreased morbidity, time spent in medical facilities and health care related costs within 2 years of treatment. Awaiting results from studies with prospectively collected data on quality of life, this supports the preferred initial use of thermal ablation as a minimally invasive treatment with curative intent, with regard to benefits to patients and health care systems.
CRediT authorship contribution statement

Declaration of competing interest
All authors certify that they have NO affiliations with or involvement in any organization or entity with any financial interest (such as honoraria; educational grants; participation in speakers' bureaus; membership, employment, consultancies, stock ownership, or other equity interest; and expert testimony or patent-licensing arrangements), or non-financial interest (such as personal or professional relationships, affiliations, knowledge or beliefs) in the subject matter or materials discussed in this manuscript.