Rates and Patterns of Recurrence after Microwave Ablation of Colorectal Liver Metastases: A per-Lesion Analysis of 416 Tumors in the Era of 2.45GHz Generators

Abstract

Background:

In patients with colorectal liver metastases (CRLM), surgical resection of all visible disease can lead to long-term survival and even cure in some patients. When complete resection is not feasible, microwave ablation (MWA) can help achieve hepatic disease control. As modern, 2.45GHz MWA generators gain popularity, the characteristics of tumors most likely to benefit from this modality remain unclear. The aim of this study was to evaluate local recurrence (LR) rates, patterns of recurrence and factors contributing to treatment failure following 2.45GHz MWA of CRLM.

Methods:

Patients with CRLM who underwent operative 2.45GHz MWA between 2011 and 2019 were identified in a prospectively maintained single-institution database. Recurrence outcomes were ascertained for each lesion by imaging review. Factors associated with LR were analyzed.

Results:

184 patients bearing 416 ablated tumors were included. Most patients (65.8%) had high clinical risk scores (3-5), and 165 (90%) underwent concurrent liver resection. Median tumor size was 10mm. After median follow-up of 28.8 months, LR was observed in 45 tumors and cumulative incidence of LR at 24 months was 10.9% [95%CI:8.0%-14.3%]. LR was the first recurrence site in 7%, often combined with recurrence elsewhere. The cumulative incidence LR at 24 months was 6.8% [95%CI:3.8%-11.0%], 12.4% [95%CI:7.8%-18.1%], and 30.2% [95%CI:14.2%-48.0%] for tumors ≤10mm, 11-20mm, and >20mm, respectively. On multivariable analysis, tumors >20mm and with subcapsular location were significantly associated with increased LR risk.

Conclusions:

2.45GHz MWA of CRLM offers excellent local control at 2 years and is most successful for small tumors deep within the parenchyma.

Introduction

In patients with metastatic colorectal cancer confined to the liver, resection of all visible disease can lead to prolonged survival and, for a significant minority, cure.^{1, 2} Even though resection is the most reliable approach to achieve complete cytoreduction and disease control, it is not always feasible or advisable. In certain cases, complete resection would result in a borderline or inadequate future liver remnant, while in others, resection of a small tumor in a deep location would potentially necessitate removing a large volume of normal parenchyma. This has led to the development of ablative techniques that aim to destroy the tumor in situ, obviating the need to remove its surrounding parenchyma and the potentially critical structures contained therein.

Over the years, several techniques have been used to ablate liver tumors, based on heat-induced destruction delivered with radiofrequency (RF Ablation) or microwave (MW Ablation) antennas, cold-induced destruction (cryoablation), irreversible electroporation (IRE) and high-dose stereotactic ablative radiotherapy (SABR).^3-9 While each has distinct advantages, MWA has gained popularity over the past decade, with higher rates of energy delivery offering more reliable and time-efficient ablations, and, more recently, the single probe configuration allowing substantially simplified lesion targeting.^{10, 11} The second generation of these devices is currently in use, with high-energy generators (2.45GHz frequency, with 90+ W of energy output), allowing for faster tissue heating, more efficient ablation and promising even better outcomes.^{12, 13}

To establish the scenarios in which MWA is a reasonable alternative to resection of colorectal liver metastases (CRLM), it is critical to define outcomes for this modality. Recurrence estimates have been previously reported by our group and others. These early results have been encouraging, with local recurrence rates of 3-19%,^{4, 7, 10, 14} which compare favorably to those of RFA and cryoablation.^{3, 11} However, the wide range of recurrence rates reported betrays the limitations of these studies. Early in the MWA adoption curve, studies had limited sample sizes and duration of follow-up, incompletely elucidating recurrence patterns and predictors of success. In addition, the technology used was typically mixed, with 915MHz and 2.45GHz systems used interchangeably. More importantly, even though tumor size is the most consistently reported predictor of MWA success, prior studies have typically grouped tumors into broad groups (e.g. larger or smaller than 3 cm). This belies the variability in outcomes within each group and does not allow the definition of size and location conditions where recurrence may be comparable to that of resection (2-12%^{15, 16}) or, at the other end of the spectrum, unacceptably high.

The purpose of this study was to review the outcomes of patients with CRLM who underwent MWA using modern, high-power generators, at a single institution, determine the rates and patterns of recurrence after treatment, and identify predictors of local failure.

Material and Methods

Patient Population

After Institutional Review Board (IRB) approval was obtained, a prospectively maintained single-institution hepatobiliary surgery database was queried to identify patients who had undergone operative MWA for CRLM between January 2011 and December 2019. Patient characteristics, pathologic features, operative and survival data were collected from the database and complemented by retrospective clinical chart review as necessary. The clinical risk score (CRS) ¹ was calculated for each patient giving 1 point for each of the following: disease free interval <12 months, node-positive primary, preoperative CEA >200 μg/L, largest metastasis >5cm, >1 liver metastasis. Tumor and treatment parameters for each MWA were previously abstracted in the database; however, all operative reports were reviewed individually to confirm location of the tumors, completeness of ablation and the device and settings used for MWA. Only tumors that were treated with high-power MWA generators with output ≥90W were included in the study. In addition, tumors that had been previously treated with another modality (RFA or MWA) or those treated with a combination of MWA and Irreversible Electroporation (IRE) were excluded from the study. The last preoperative CT or MRI imaging (within 60 days from the MWA) was reviewed for each patient to determine the maximum dimension, segment location, and proximity to the liver capsule and major vessels for each tumor. Subcapsular tumors were defined as those within 10mm of the liver capsule. Close proximity to a major vessel was defined as tumor located within 10mm of a hepatic vein or major tributary or sector pedicle. Tumor characteristics for lesions that could not be identified on preoperative imaging were recorded based on intraoperative ultrasound findings detailed in the operative report. A prospectively maintained surgical complications database was reviewed to identify post-operative complications. The classification of complications used by this system has been reported previously. All imaging from the post-operative period through last follow up was reviewed to determine local recurrence, new intrahepatic recurrence, and extrahepatic recurrence on a per-tumor basis. “First recurrence” was defined as the first one (or several) of these to be detected. To ensure consistency, patients who had residual intrahepatic (in the context of planned 2^nd stage liver surgery) or extrahepatic disease were not included in “first recurrence” analyses. Date of first recurrence was recorded as the date of first appearance of suspicious findings on imaging, even if recurrence was confirmed at a later date by biopsy or further imaging. Local recurrence was defined as presence of tumor within 10mm of the edge of ablation zone on any follow-up imaging, even if the center of the recurrent tumor was further away. New intrahepatic recurrence was defined as any recurrence within the liver that did not meet criteria for local recurrence. The dataset was locked in March 2021.

Microwave Ablation Technique

MWA for CRLM is used at our institution both in the operating room and as a percutaneous procedure by interventional radiology. Because of the obvious differences in delivery setting, image guidance and, frequently, patient population, this study is focused exclusively on patients undergoing intraoperative MWA for CRLM in the context of a surgical procedure (open or minimally invasive). At our institution, the priority is to resect CRLM whenever possible. MWA is used in conjunction with resection and hepatic artery infusion pump (HAIP) chemotherapy to optimize parenchymal preservation and treat lesions in locations that make resection challenging or require extensive resection of normal parenchyma. Only ablations performed with the Covidien/Medtronic Emprint^™, Ethicon NeuWave^™, or Microsulis Accu2i pMTA^™ ablation devices (2.45 GHz / 90 W) were included. All ablations were performed operatively by experienced hepatopancreatobiliary surgeons. Ablation duration was determined by the surgeon at the time of ablation based on the tumor characteristics and manufacturer protocols with a goal margin of 1cm. Intraoperative ultrasound (IOUS) was routinely used for guidance of probe placement and monitoring ablation success and margin. Only tumors deemed to be completely ablated, as determined by surgeon judgement during the surgery and documented in the operative report, were included in the study.

Statistical Methods

Patients and tumor characteristics are summarized using frequency and percentage for categorical covariates, and median and range for continuous covariates. Cumulative incidences of local recurrence from the time of tumor ablation were estimated using competing risks methods where the unit of interest was at the lesion level. Lesions without local recurrence at the time of patient death are considered competing events. Factors associated with time to local recurrence were done using the extension of Fine and Gray regression for cluster data to account for the correlation within multiple lesions from the same patient.¹⁷ Variables significantly associated with time to local recurrence on univariable analysis at p-value <0.2 were entered into a multivariate model. Tumor size was examined as both a categorical and continuous variable. As there were not many ablated tumors >3 cm within the cohort, tumor size was categorized as ≤10mm, 11-20mm, and >20mm. To explore the potential cut-off for the size of maximum lesion, a cubic spline model with 3 knots was used initially. The plot with the spline against hazard ratio was created and the relationship appeared to be linear. In the multivariable model, size of maximum lesion was modeled as a continuous covariate.

All statistical analyses were performed using SAS Version 9.4 (SAS Institute, Inc., Cary, NC, USA) or R Version 4.0.4 (R Foundation for Statistical Computing, Vienna, Austria). All p-values were two-sided and p-values of <0.05 were considered to indicate statistical significance.

Results

Patient Population

A total of 184 patients met inclusion criteria for this study. The clinical characteristics of these patients are summarized in Table 1. The primary tumor was in the colon for 72% of patients, and in 76% of patients it was associated with regional lymph node metastases (N1 or N2). The disease-free interval between the diagnosis of the primary tumor and the liver metastasis was less than 12 months for the majority (86%) of patients. In addition, most patients (65.8%) had a high CRS (3-5) and received preoperative (86%) and postoperative chemotherapy (95%). 43 (23%) patients received preoperative, and 125 (68%) patients received postoperative HAIP chemotherapy. Among 162 patients (88%) with available RAS sequencing, 70 (43%) had an alteration present. Finally, 165 (90%) of the patients underwent liver resection at the time of MWA.

Table 1.

Patient Characteristics

Total patients	184 (100)
CRLM treated with MWA per patient, N, median (IQR)	2 (1, 3)
Age, yr, median (IQR)	54 (45, 63)
Female gender	76 (41)
Race
White	154 (84)
Asian	10 (5)
Black	7 (4)
Hispanic	5 (3)
Other	4 (2)
Unknown	4 (2)
Primary tumor in colon	132 (72)
Disease free interval < 12 months	159 (86)
Preoperative CEA, median (IQR)a	5 (3, 15)
Largest hepatic tumor size, cm, median (IQR)	2.5 (1.5, 3.5)
Total number of CRLMs, median (IQR)	5 (3, 9)
N1 or N2 primary tumorb	134 (76)
Fong clinical risk score
0	1 (1)
1	6 (3)
2	56 (30)
3	103 (56)
4	17 (9)
5	1 (1)
RAS Alterationc	70 (43)
Preoperative systemic chemotherapyd	158 (86)
Postoperative systemic chemotherapyd	175 (95)
Preoperative HAIP chemotherapyd	43 (23)
Postoperative HAIP chemotherapyd	125 (68)
Concomitant liver resection and ablation	165 (90)

Data are n (%) unless otherwise noted; IQR: interquartile range; HAIP: Hepatic artery infusion pump; ^a Seven missing values ^b Seven patients with unknown nodal status ^c Twenty-two missing values ^d Within 3 months of ablation date

Tumor Characteristics

416 CRLM treated with MWA were included in this study. The characteristics for these tumors are detailed in Table 2. The median tumor size was 10mm. There were 34 tumors larger than 20mm. Tumor size was determined based on preoperative axial imaging. Thirty-three tumors could not be identified on preoperative imaging. The size for these tumors was recorded based on IOUS measurement. A minority of tumors were subcapsular (35%) or located within 10mm of a major vessel (33%).

Table 2.

Tumor Characteristics of Ablated Lesions

Total tumors	416 (100)
Tumor size, ablated lesion, mm, median (IQR)	10 (7, 14)
Segment location of tumor
I	8 (2)
II	35 (8)
III	30 (7)
IVA	50 (12)
IVB	33 (8)
V	42 (10)
VI	66 (16)
VII	70 (17)
VIII	82 (20)
Subcapsular locationa	139 (35)
Close proximity to vesselb	130 (33)
RAS Alteration	165 (44)
Preoperative systemic chemotherapyc	360 (94)
Postoperative systemic chemotherapyc	393 (94)
Preoperative HAIP chemotherapyc	129 (31)
Postoperative HAIP chemotherapyc	303 (73)

Data are n (%) unless otherwise noted; IQR: interquartile range; HAIP: Hepatic artery infusion pump; ^a Eighteen missing values ^b Thirty-seven missing values ^c Within 3 months of ablation date

Complications

Post-operative complications were noted in 100 (54%) patients. 32 (17%) patients experienced a serious complication (grade 3-5) requiring surgical or procedural intervention. The most common serious complications were infections (37%), bilomas or bile leaks (22%) and bleeding (19%). There were no vascular injuries resulting from MWA. Only 3 of the 12 patients who had isolated MWA (no concurrent liver resection or HAIP placement) had a complication. Only 1 (8%) of these was serious but could not be directly attributed to the MWA (delirium tremens requiring intubation).

Follow-Up and Patient Death

Among the 184 patients, 76 deaths were observed at the time of analysis. Median follow-up among surviving patients was 28.8 months (range 2-103.6 months), and all patients had at least one follow-up scan available for review. One hundred eight patients were alive and had follow-up of at least 24 months.

Tumor Recurrence

Over the entire follow-up, 45 local recurrences were observed. Cumulative incidence of local recurrence at 12 and 24 months was 6.6% [95%CI: 4.5%-9.3%] and 10.9% [95%CI: 8.0%-14.3%], respectively (Fig. 1). Among these, only 22 (6.9%) presented with local recurrence as the first site of disease recurrence, often in combination with other sites of recurrence (Table 3). 5 of the 45 local recurrences were found at the first CT following MWA, demonstrating a technical failure rate of 1.2% (5 out of 416 tumors ablated). Of note, three patients with a total of 3 treated tumors had diffusely recurrent disease in the liver (>20 small liver tumors) and at least one within 10mm of the ablation zone; these were categorized as local recurrence to adhere to the strictest definition for the primary outcome, even though the proximity to the ablation zone may have been coincidental.

Figure 1.

Cumulative Incidence of Local Recurrence of Colorectal Liver Metastases Treated with Microwave Ablation

Table 3.

Patterns of First Recurrence per Tumor

First recurrence site	n = 316a
No recurrence	50 (16)
Local recurrence only	13 (4)
New intrahepatic recurrence only	137 (43)
Extrahepatic recurrence only	70 (22)
Local recurrence plus new intra- and/or extrahepatic recurrence	9 (3)
Extrahepatic plus new intrahepatic recurrence	37 (12)

^aFirst recurrence site not applicable to 100 ablated tumors in the study due to residual disease at the time of microwave ablation (41 with extrahepatic disease and 59 in the context of a planned two-stage liver intervention); Local recurrence: Recurrence within 10mm of the ablation zone; New intrahepatic recurrence: intrahepatic recurrence not meeting criteria for local recurrence

Clinical Predictors of Local Recurrence

Time to local recurrence for the 416 ablated tumors was significantly associated with ablated tumor size and subcapsular location on both univariable and multivariable analyses (Table 4). Larger tumors were more likely to have local recurrence, with cumulative incidence of local recurrence at 12 months of 3.4% [95%CI: 1.5%-6.6%], 7.7% [95%CI: 4.3%-12.3%], and 21.1% [95%CI: 9.1%-36.4%] for tumors ≤10mm, 11-20mm, and >20mm, respectively (Fig. 2). At 24 months, the rates were 6.8% [95%CI: 3.8%-11.0%], 12.4% [95%CI: 7.8%-18.1%], and 30.2% [95%CI: 14.2%-48.0%] for tumors ≤10mm, 11-20mm, and >20mm, respectively. The univariate results suggested tumor size was associated with increased risk of local recurrence, specifically, tumors >20mm had a significantly increased hazard of local recurrence after microwave ablation compared to tumors ≤10mm (HR 4.22, 95% CI 1.76-10.12, p=0.005). There was also no apparent optimal cutoff as the hazard of local recurrence increases linearly with tumor size (Fig. 3), with a 5% increase in risk for each 1mm increase in size of tumor (HR 1.05, 95% CI 1.02-1.08, p=0.002). This remained statistically significant on a multivariable regression model that also included subcapsular tumor location, proximity to major vessels, and use of adjuvant HAIP therapy within 3 months. There was no significant association between cumulative incidence of recurrence and type of MWA device used (p=0.23, data not shown). As 86% of patients received preoperative and 95% received postoperative chemotherapy, the effect of perioperative chemotherapy on local recurrence could not be adequately assessed. Subcapsular tumors also had increased risk of local recurrence on both univariable (HR 1.95, 95% CI 1.07–3.55, p=0.029) and multivariable analysis (HR 2.00, 95% CI 1.09-3.65, p=0.02). A potential association between tumor proximity to a major vessel and risk of local recurrence was investigated but did not reach statistical significance (HR 1.76, 95% CI 0.97-3.20, p=0.06). Finally, a sensitivity analysis was performed to explore the association between RAS mutation and time to local recurrence. In the final multivariable model, RAS mutation was not significantly associated with time to local recurrence (data not shown). Increased size of lesion and subcapsular tumor remained significantly associated with time to local recurrence.

Table 4.

Factors Associated with Time to Local Recurrence (N=416)

	Univariable			Multivariable
Variable	HR	95% CI	p value	HR	95% CI	p value
Size, mm			0.005
1 – 10	Ref.
11 – 20	1.73	0.88 – 3.39
> 20	4.22	1.76 – 10.1
Size (per 1mm increase)	1.05	1.02 – 1.08	0.002	1.04	1.01 – 1.08	0.006
Subcapsular location			0.03			0.02
No	Ref.			Ref.
Yes	1.95	1.07 – 3.55		2.00	1.09 – 3.65
Close Proximity to Vessel			0.06			0.06
No	Ref.			Ref.
Yes	1.76	0.97 – 3.20		1.75	0.95 – 3.21
Adjuvant HAIP Chemotherapy			0.13			0.37
No	Ref.			Ref.
Yes	0.61	0.33 – 1.15		0.74	0.37 – 1.43

Bold indicates a statistically significant p value; HAIP: Hepatic Artery Infusion Pump

Figure 2.

Cumulative Incidence of Local Recurrence by Tumor Size

Figure 3.

Hazard of Local Recurrence Based on Tumor Size

Discussion

Operative, IOUS-guided MWA using modern high-power generators was associated with excellent local disease control at 2 years in this single-institution cohort of patients with surgically managed CRLM. Importantly, most local recurrences occurred either together with a recurrence at a different site or following recurrent disease elsewhere; local recurrence at the ablation site as an isolated first site of recurrence was only observed in 4% of ablated tumors. Location was associated with treatment failure, with tumors within 10mm of the liver surface recurring at twice the rate of deeper-sited tumors. As anticipated, tumor size was the strongest predictor of MWA success; larger tumors recurred at a significantly higher rate, with up to one third of those >20mm in diameter failing locally, contrasted with less than 10% for those ≤10mm.

These results validate and expand on what has been previously reported for MWA of CRLM.^{10, 11, 14} Overall, local failure is an uncommon event, and MWA remains a very useful tool for hepatobiliary surgeons managing this disease. However, unlike prior reports that grouped tumors smaller than 3 cm together and suggested that they were all appropriate for ablation,^{6, 9, 11} this report evaluates size risk on a more granular level, and suggests that even tumors >2 cm are at increased risk of local recurrence. At the other end of the spectrum, recurrence for ablated sub-centimeter tumors was 6.8% after 2 years in this series. This compares favorably to local recurrence after liver resection (reported in the range of 2-12% overall,^{15, 16} though without size specific outcomes), suggesting that MWA may offer similar local control to resection for small tumors. Ideally, as these results suggest therapeutic equipoise, the potential non-inferiority of MWA compared to resection for subcentimeter metastases should be studied prospectively.

Tumor location has historically been considered an important factor in ablation success^{6, 11 18} and was specifically evaluated in this study. Proximity to major blood vessels has been reported as a major risk factor for recurrence,^{6, 18} attributed to a combination of physical properties (blood flow cooling the adjacent parenchyma, preventing it from reaching target temperatures for ablation, also known as the “heat-sink effect”), and surgeon hesitancy to ablate as aggressively near major structures (particularly inflow pedicles, where bile duct injury may ensue). Unfortunately, these results do not definitively address this phenomenon; proximity to blood vessel fell short of statistical significance in recurrence analyses (p=0.06), and, therefore, acknowledging the likelihood of Type II error, the question remains open. On the contrary, these data offer convincing evidence that MWA of superficially located tumors is associated with higher local recurrence rate. Even after accounting for size and other treatment characteristics, tumors within 10 mm from the surface were twice as likely to recur. This finding has been reported in some other series but is not consistently associated with local recurrence.^{6, 18} Although there is no definitive reason why this association may exist, one theory is that there is some alteration in the tissue response to ablation when there is a small or no rim of normal liver surrounding the tumor. In addition, for a tumor on the surface, it can be more challenging to place the active portion of the MWA probe in the center of the tumor.

These findings have significant implications for the use of MWA in the setting of CRLM. First, the data should reassure surgeons utilizing MWA to treat small tumors with a parenchymal preservation approach; isolated local recurrence, clinically the most critical type of failure, was observed only in 4% of cases. Additionally, these results challenge the theory that local recurrence after ablation overwhelmingly occurs within the first year. Unlike prior reports with more limited follow-up,^{6, 10, 11, 18} nearly half of local recurrences in this cohort were observed after 12 months from surgery. Finally, somewhat unexpectedly given the theoretical advantages of 2.45 GHz generators and the widespread enthusiasm for their use,^{13, 19} no decrease in the rates of local recurrence was observed compared to historical microwave ablation cohorts (albeit with shorter follow-up) from our institution and others (5-6%). ^{9, 10, 14} Indeed, the generally consistent reports of approximately 5-10% local failure after MWA for CRLM, regardless of generator technology, suggest that there may be factors related to biology (microscopic tumor spread) or technique (intraoperative targeting) that define a lower limit of treatment success in this setting.

Building on prior reports from our institution and others, several key factors allowed this study to further elucidate the role of MWA in the management of CRLM. A relatively large number of treated tumors, from a single institution with extensive experience in the surgical management of CRLM and homogeneous treatment algorithms were included. The median follow-up of nearly 29 months compares favorably to prior reports and is reasonable given the majority of recurrences occur within 12-18 months.¹⁰ Finally, and perhaps more importantly, this study was conducted at a lesion level, with each postoperative CT scan independently reviewed by the investigators to verify the presence or absence of recurrence meeting the prespecified parameters, rather than relying on previously collected data and radiology reports. This allowed capturing the outcomes of interest as accurately as possible and the classification of recurrences (local vs. not) based on uniform definitions, a common point of critique in prior studies.

Even though this series’ large and relatively homogeneous population highlights the value of this study, there are certain limitations that should be acknowledged. As a retrospective study, the treatment and post-treatment surveillance of the population was not defined a priori and varied among individuals. Routine early postoperative imaging was not consistently performed, so the effect of ablation margin on recurrence could not be studied here. As a single-center study in the context of operative MWA in conjunction with resection and HAIP therapy, the results should be interpreted in this context, and may not necessarily apply to isolated percutaneous MWA. However, the management of patients with CRLM at our institution follows well established protocols in accordance with national guidelines. In addition, prior histological confirmation of all ablated lesions was not obtained, and it is possible that some ablated lesions did not represent CRLM; this is inherent to studies of CRLM ablation, and, in the context of IOUS-guided ablation performed by experienced hepatobiliary surgeons, the expected rate of misdiagnosis is low. Apart from RAS mutation status, genomic data was unfortunately unavailable for most patients, and the intriguing possibility of molecular predictors could not be addressed. Finally, uncovering the risk factors for recurrence after MWA relies on an adequate number of observed events. With only 11% recurrence rate in this series, it is likely that there was inadequate power to evaluate the effect of vessel proximity or post-operative HAIP chemotherapy on MWA success.

In conclusion, though not demonstrably superior to prior generation devices, IOUS-guided MWA using 2.45GHz generators offers durable local control for small CRLM and is especially advantageous for tumors located deep in the parenchyma. Given the increased local failure rate compared to smaller tumors, surgeons should use caution when using operative MWA to treat CRLM >2cm.

Synopsis:

2.45GHz microwave ablation (MWA) for colorectal liver metastases resulted in durable local control for nearly 90% of tumors in this study of 416 ablated tumors. MWA is most successful for small tumors located deep within the liver parenchyma.