Predictive nomograms of repeat intrahepatic recurrence and overall survival after radiofrequency ablation of recurrent colorectal liver metastases

Abstract

Objectives

This study was conducted to develop nomograms for predicting repeat intrahepatic recurrence (rIHR) and overall survival (OS), after radiofrequency ablation (RFA), treatment in patients with recurrent colorectal liver metastases (CLMs) after hepatectomy based on clinicopathologic features.

Methods

A total of 160 consecutive patients with recurrent CLMs after hepatectomy who were treated with ultrasound-guided percutaneous RFA from 2012 to 2022 were retrospectively included. Patients were randomly divided into a training cohort and a validation cohort, with a ratio of 8:2. Potential prognostic factors associated with rIHR and OS, after RFA, were identified by using the competing-risks and Cox proportional hazard models, respectively, and were used to construct the nomogram. The nomogram was evaluated by Harrell’s C-index and a calibration curve.

Results

The 1-, 2-, and 3-year rIHR rates after RFA were 58.8%, 70.2%, and 74.2%, respectively. The 1-, 3- and 5-year OS rates were 96.3%, 60.4%, and 38.5%, respectively. In the multivariate analysis, mutant RAS, interval from hepatectomy to intrahepatic recurrence ≤ 12 months, CEA level >5 ng/ml, and ablation margin <5 mm were the independent predictive factors for rIHR. Mutant RAS, largest CLM at hepatectomy >3 cm, CEA level >5 ng/ml, and extrahepatic disease were independent predictors of poor OS. Two nomograms for rIHR and OS were constructed using the respective significant variables. In both cohorts, the nomogram demonstrated good discrimination and calibration.

Conclusions

The established nomograms can predict individual risk of rIHR and OS after RFA for recurrent CLMs and contribute to improving individualized management.

Keywords:

• Radiofrequency ablation
• colorectal cancer liver metastases
• nomogram
• recurrence
• RAS mutation status

Introduction

Colorectal cancer (CRC) is the second leading cause of cancer death worldwide [1], with up to 50% of patients developing liver metastases [2]. Hepatectomy is the standard management for resectable colorectal liver metastases (CLMs), and the 5-year overall survival (OS) rate of postoperative patients can reach 40–60% [3]. However, recurrence rates exceed 50% within two years after hepatectomy, and the principal site of recurrence is the liver [4]. Repeat hepatectomy in selected patients may be a reasonable therapeutic strategy to achieve long-term survival, but is challenging because of postsurgical adhesions and limited liver reserves [5]. Percutaneous thermal ablation such as radiofrequency ablation (RFA) is a curative-intent locoregional therapy for CLM, which can be used alone or with surgical resection as long as all visible CLMs is eradicated [6, 7]. Unfortunately, only 25% of patients with recurrent CLMs achieve long-term intrahepatic recurrence-free survival (ihPFS) after RFA [8]. Therefore, there is a clear need to obtain better prognostic information and establish risk prediction models to facilitate individualized clinical management.

Nomograms are a visual model for statistical predictions that utilize clinically important independent predictors, and have a better performance than staging or scoring systems [9]. Several nomograms have been constructed for predicting ihPFS and OS in CLM patients after thermal ablation, which are primarily based on features of the primary tumor and initial local treatment [9–11]. However, prognostic factors may also change dramatically with disease recurrence. To date, no predictive tools have been specifically designed for recurrent CLM patients treated with RFA.

Compared to morphological factors, molecular markers represent a better and more direct measurement of tumor biology [12]. RAS mutation status is a well-known prognostic biomarker in patients undergoing CLM ablation [13], and has been associated with worse OS and a high risk of recurrence [14, 15]. More accurate prognostic methods may be achieved by including this genetic biomarker in nomograms. In the current study, we aimed to develop nomograms that incorporate the RAS mutation status to predict repeat intrahepatic recurrence (rIHR) and OS after RFA for recurrent CLMs.

Methods

Subjects

This retrospective cohort study was approved by our institutional review board (Ethics approval number: 2020KT64) and data was collected from a prospectively maintained database and additional information was gathered from medical records. Between January 2012 and January 2022, a total of 281 consecutive patients underwent ultrasound-guided percutaneous RFA for recurrent CLMs after hepatectomy at our hospital. The inclusion criteria were as follows: (1) The number of CLM tumors is less than or equal to 6, and the size of CLM tumors is less than or equal to 5 cm; (2) Data indicating patient RAS mutation status is available; (3) Patient underwent RFA with curative intent for CLMs; (4) The primary lesion of colorectal cancer has undergone curative resection. The exclusion criteria were as follows: (1) Residual unablated tumor after the first RFA session not retreated with an additional session of RFA; (2) Patient was lost to follow-up. Liver metastases were diagnosed based on radiologic findings of contrast-enhanced computed tomography (CT) and/or magnetic resonance imaging (MRI) in the context of a previous metastatic CRC diagnosis.

Radiofrequency ablation procedure

All RFA procedures were performed under general anesthesia using real-time ultrasound (US) or contrast-enhanced ultrasound (CEUS) guidance by a radiologist team with more than 10 years of experience. The RFA equipment used in this study included a CelonLab system (Olympus, Germany), Valleylab system (Tyco Healthcare, North Haven, CT, USA), RITA 1500X system (AngioDynamics, Latham, NY, USA) or LDRF-120S RFA system (LeadElectron, Mianyang, China). US guidance devices included the Aloka ultrasound systems (Alokaa-10, Tokyo, Japan) or GE systems (E9, GE, USA). All ablations were performed with the intent to completely cover the tumor plus a safety margin of at least 5 mm in all directions. CEUS was performed immediately following RFA to determine whether the tumor was completely covered by the ablation zone. The ablation margins were assessed on the first post-RFA imaging follow-up using contrast-enhanced CT/MRI. Perioperative chemotherapy was prescribed following consensus among the multidisciplinary team.

Imaging follow-up

Contrast-enhanced CT/MRI was performed 1 month after RFA and every 3 months thereafter over the next 2 years. If no recurrence was detected, the period between imaging sessions was extended to every 6 months. The last follow-up was completed in November 2023. Standardized terminology and reporting standards have been adopted to describe the ablation endpoint [16, 17]. Technique efficacy was defined as complete tumor ablation demonstrated by imaging at the 1-month follow-up after RFA. Residual unablated tumor was defined as the presence of residual viable tumor at the ablative margin during the 1-month follow-up after RFA. The presence of newly enhanced lesions within or adjacent to the complete ablation zone was defined as local tumor progression (LTP). Newly enhanced lesions in the liver outside the ablation zone were considered intrahepatic distant recurrence (IDR). Intrahepatic recurrence included LTP and IDR.

Ablation margins and RAS gene analysis

One month after RFA, ablation margins were evaluated by two radiologists with more than 5 years of experience, as previously described [18]. Using a PACS workstation, pre- and post-RFA contrast-enhanced CT/MR images were displayed side by side to measure distances from the tumor boundary to the selected anatomic landmarks on three orthogonal planes. The margin at each landmark was calculated by subtracting the post-RFA distance from the pre-RFA distance; the smallest value was recorded as the ablation margin. In the case of multiple tumors, the smallest margin was recorded as the ablation margin. Genomic DNA was extracted from primary tumors or resected liver metastases and subjected to either next-generation sequencing or polymerase chain reaction (PCR) to screen for mutations in KRAS/NRAS (RAS) codons 12, 13, and 61 [19].

Candidate predictor variables

Candidate predictors included age at ablation; sex; primary tumor characteristics (American Joint Committee on Cancer (AJCC) T category, node-positive primary tumor, and location (right colon, left colon, or rectum)); initial CLM diagnosis (synchronous or metachronous); RAS mutation status; characteristics of hepatectomy (largest tumor size (≤3 cm or >3 cm), number of tumors, concurrent RFA at the time of hepatectomy, margin status (R0, defined as a pathologically negative margin ≥1 mm, or R1)); interval from hepatectomy to intrahepatic recurrence (≤12 months or >12 months); and characteristics of RFA treatment (extrahepatic disease at ablation, carcinoembryonic antigen (CEA) level at ablation (≤5 ng/ml or >5 ng/ml), pre-ablation chemotherapy (none, cytotoxic, or cytotoxic + biological agent), largest tumor size (≤3 cm or >3 cm), number of tumors, ablation margin (<5 mm or ≥5 mm), and receipt of post-ablation chemotherapy). For patients with multiple hepatectomies, data from the last operation were assessed. Tumor size and number at hepatectomy were based on pathological results from resection specimens.

Statistical analysis

Categorical variables were reported as frequencies with percentages, while continuous variables were reported as mean ± SD or median with IQR. The comparison of categorical variables was conducted using either the chi-square test or Fisher’s exact test, whereas the comparison of continuous variables was performed using Student’s t-test or Mann-Whitney U-test. Time to rIHR was calculated from the time of RFA to the first radiologic evidence of LTP or IDR, whichever occurred first. Death that occurred before rIHR was classified as a competing risk. The cumulative incidence function (CIF) was used to estimate the cumulative rate of rIHR. OS was measured from the time of RFA to patient death or last follow-up. The Kaplan–Meier method was used to estimate OS. A competing-risks (Fine-Gray) model was performed for univariate and multivariate analysis of LTP, IDR, and rIHR. A Cox proportional hazard model was employed for univariate and multivariate analysis of OS. Factors with a P value <0.05 in univariate analyses were selected to enter the multivariate analyses. Nomograms were developed based on the results of multivariate analysis to predict rIHR and OS rates. To evaluate model performance, the Harrell’s C-index and calibration curves were used to assess discrimination and calibration ability, respectively. Additionally, 1,000 bootstrap samples were used in the internal validation of the model. All patients were classified into low-risk and high-risk groups according to the median of the total score predicted by the nomogram. The rIHR rates of each group were estimated by the CIF method and compared by Gray’s test. The OS rates of each group were estimated by the Kaplan–Meier method and compared by a log-rank test. Statistical analyses were performed using R 4.2.2 (http://www.r-project.org). P values of <0.05 were considered significant.

Results

Patient characteristics

The study included 160 patients with 258 lesions treated in 171 sessions (Figure 1). The mean age was 58.5 ± 8.4 years, and 31.2% of the patients were female. The mean size of ablated lesion was 1.7 ± 0.7 (range, 0.6–4.9) cm. A total of 151 patients (94.4%) had undergone one round of hepatectomy, and the other 9 patients had undergone two rounds of hepatectomy. At the time of RFA, 37 (23.1%) patients had extrahepatic disease, including lung only (n = 28), lung and lymph nodes (mediastinal, abdominal, both mediastinal and abdominal) (n = 4), abdominal lymph nodes (n = 3), and bone (n = 2). Subjects were allocated randomly into two cohorts: a training cohort (n = 128) and a validation cohort (n = 32), maintaining an 8:2 ratio. There were no statistically significant differences observed between the training and validation cohorts at baseline (p > 0.05 for all comparisons). Baseline characteristics are presented in Table 1.

Figure 1. Patient inclusion flowchart.

Table 1. Characteristics of the patients.

Characteristic	Total cohort (n = 160)	Training cohort (n = 128)	Validation cohort (n = 32)	p Value
Age at ablation, mean (SD), years	58.5 (8.4)	58.5 (8.2)	58.7 (9.3)	0.942
Sex				0.136
Male	110 (68.8)	92 (71.9)	18 (56.2)
Female	50 (31.2)	36 (28.1)	14 (43.8)
Primary tumour site				0.088
Left colon	79 (49.4)	67 (52.3)	12 (37.5)
Right colon	19 (11.9)	17 (13.3)	2 (6.3)
Rectum	62 (38.8)	44 (34.4)	18 (56.2)
Primary T stage				1.000
T1-3	126 (78.8)	101 (78.9)	25 (78.1)
T4	34 (21.2)	27 (21.1)	7 (21.9)
Nodal metastases				0.643
No	52 (32.5)	40 (31.2)	12 (37.5)
Yes	108 (67.5)	88 (68.8)	20 (62.5)
RAS mutation status				0.100
Wild type	93 (58.1)	79 (61.7)	14 (43.8)
Mutant	67 (41.9)	49 (38.3)	18 (56.2)
No. of previous hepatectomy				0.385
Single	151 (94.4)	122 (95.3)	29 (90.6)
Two	9 (5.6)	6 (4.7)	3 (9.4)
Initial CLM diagnosis				0.232
Synchronous	116 (72.5)	96 (75.0)	20 (62.5)
Metachronous	44 (27.5)	32 (25.0)	12 (37.5)
Factors associated with hepatectomy of CLM
Number of tumours, median (IQR)	4.0 (2.0–6.0)	3.5 (2.0–6.0)	4.0 (2.0–7.3)	0.507
Size of largest CLM, cm
Median (IQR)	2.2 (1.5–3.3)	2.2 (1.5–3.3)	2.1 (1.5–3.3)	0.667
≤3 cm	113 (70.6)	91 (71.1)	22 (68.8)
>3 cm	47 (29.4)	37 (28.9)	10 (31.2)
Type of procedure				0.452
Resection	116 (72.5)	95 (74.2)	21 (65.6)
Resection and RFA	44 (27.5)	33 (25.8)	11 (34.4)
Resection margin				0.965
R0	113 (70.6)	91 (71.1)	22 (68.8)
R1	47 (29.4)	37 (28.9)	10 (31.2)
Factors associated with RFA of recurrent CLM
Interval from hepatectomy to intrahepatic recurrence, median (IQR), months	6.0 (3.0–10.2)	6.0 (3.0–10.0)	7.0 (4.0–11.2)	0.426
Number of tumours, median (IQR)	1.0 (1.0–2.0)	1.0 (1.0–2.0)	1.0 (1.0–2.0)	0.375
Size of largest CLM, cm
Median (IQR)	1.8 (1.4–2.4)	1.8 (1.4–2.4)	1.7 (1.3–2.2)	0.265
≤3 cm	152 (95.0)	122 (95.3)	30 (93.7)
>3 cm	8 (5.0)	6 (4.7)	2 (6.3)
CEA level at ablation	5.4 (3.2–9.9)	5.3 (3.2–9.8)	5.5 (2.4–12.3)	0.969
Extrahepatic disease at ablation				0.373
No	123 (76.9)	96 (75.0)	27 (84.4)
Yes	37 (23.1)	32 (25.0)	5 (15.6)
Preablation chemotherapy				0.328
No	99 (61.9)	82 (64.1)	17 (53.1)
Cytotoxic chemo	13 (8.1)	11 (8.6)	2 (6.3)
Cytotoxic chemo + biological agent	48 (30.0)	35 (27.3)	13 (40.6)
Postablation chemotherapy				1.000
No	94 (58.8)	75 (58.6)	19 (59.4)
Yes	66 (41.2)	53 (41.4)	13 (40.6)
Ablation margin				0.400
<5mm	67 (41.9)	51 (39.8)	16 (50.0)
≥5mm	93 (58.1)	77 (60.2)	16 (50.0)

RFA, radiofrequency ablation; CLM, colorectal liver metastases; CEA, carcinoembryonic antigen.

At a median follow-up of 52 months (range 19–136 months) after RFA, 121 patients (75.6%) experienced rIHR after RFA, including 95 patients with IDR at initial recurrence, 19 patients with LTP at initial recurrence, and 7 patients with synchronous IDR and LTP. Death occurred in 85 (53.1%) patients. The 1-, 2-, and 3-year rIHR rates after RFA were 58.8%, 70.2%, and 74.2%, respectively. The median OS after RFA was 49 months, and the 1-, 3- and 5-year OS rates were 96.3%, 60.4%, and 38.5%, respectively. Patients with mutant RAS had a higher incidence of rIHR and reduced OS than those with wild-type RAS (3-year rIHR: 83.6% vs. 67.9%, p < 0.001; 3-year OS: 43.9% vs. 72.7%, p < 0.001). Through the end of the study, 36 of 258 (14.0%) treated tumors showed evidence of LTP, and 110 of 160 (68.8%) patients showed evidence of IDR.

Univariate and multivariate analysis of outcome

Analyses were performed on 128 patients in the training cohort. Univariate fine-gray analysis identified five factors that significantly affected rIHR after RFA: primary tumor location, RAS mutation status, interval from hepatectomy to intrahepatic recurrence, CEA level at ablation, and ablation margin. Multivariate fine-gray analysis showed that mutant RAS (subdistribution hazard ratio [SHR], 1.92; 95%CI, 1.29–2.84; p = 0.001), interval from hepatectomy to intrahepatic recurrence ≤12 months (SHR, 2.00; 95%CI, 1.07–3.73; p = 0.029), CEA level at ablation >5 ng/ml (SHR, 1.57; 95%CI, 1.04–2.36; p = 0.032) and ablation margin <5 mm (SHR, 1.69; 95%CI, 1.14–2.50; p = 0.009) were independently associated with higher rIHR after RFA (Table 2). Univariate Cox regression analysis identified five factors that significantly affected OS after RFA: RAS mutation status, largest CLM at hepatectomy, CEA level at ablation, extrahepatic disease, and ablation margin. Multivariate Cox regression analysis showed that mutant RAS (hazard ratio [HR], 2.15; 95%CI, 1.29–3.60; p = 0.003), largest CLM at hepatectomy >3 cm (HR, 2.10; 95%CI, 1.26–3.50; p = 0.004), CEA level at ablation >5 ng/ml (HR, 2.05; 95%CI, 1.16–3.62; p = 0.013) and extrahepatic disease (HR, 1.88; 95%CI, 1.07–3.32; p = 0.028) were independent predictors of reduced OS (Table 2).

Table 2. Univariate and multivariate analyses of risk factors for rIHR and OS after RFA.

	Repeat intrahepatic recurrence				Overall survival
	Univariate		Multivariate		Univariate		Multivariate
Variable	SHR (95% CI)	p Value	SHR (95% CI)	p Value	HR (95% CI)	p Value	HR (95% CI)	p Value
Age (≤60 as ref.)
>60	1.00 (0.68–1.46)	0.993			1.29 (0.80–2.08)	0.302
Gender (female as ref.)
Male	0.95 (0.60–1.51)	0.828			0.79 (0.48–1.32)	0.374
Primary tumour site (left colon as ref.)
Right colon	1.79 (1.03–3.11)	0.038	1.26 (0.70–2.27)	0.430	1.60 (0.80–3.19)	0.185
Rectum	1.39 (0.91–2.11)	0.120	1.16 (0.77–1.75)	0.480	1.10 (0.65–1.88)	0.720
Primary T stage (T1-3 as ref.)
T4	1.29 (0.82–2.03)	0.279			1.30 (0.75–2.24)	0.349
Nodal metastases (no as ref.)
Yes	0.95 (0.65–1.37)	0.772			1.46 (0.85–2.51)	0.171
RAS mutational status (wild type as ref.)
Mutant	2.28 (1.52–3.42)	<0.001	1.92 (1.29–2.84)	0.001	2.79 (1.71–4.53)	<0.001	2.15 (1.29–3.60)	0.003
Initial CLM diagnosis (metachronous as ref.)
Synchronous	1.37 (0.86–2.18)	0.186			1.58 (0.87–2.86)	0.131
No. of previous hepatectomy (one as ref.)
Two	1.19 (0.54–2.64)	0.668			1.45 (0.53–3.99)	0.473
Factors associated with hepatectomy of CLM
No. of CLM (solitary as ref.)
Multiple	1.49 (0.82–2.68)	0.189			0.98 (0.50–1.93)	0.961
Size of largest CLM (≤3 cm as ref.)
>3 cm	1.05 (0.65–1.69)	0.839			1.82 (1.11–3.00)	0.018	2.10 (1.26–3.50)	0.004
Type of procedure (resection as ref.)
Resection and RFA	1.39 (0.95–2.02)	0.087			0.83 (0.47–1.46)	0.512
Surgical margin (R0 as ref.)
R1	1.43 (0.94–2.17)	0.098			1.46 (0.87–2.44)	0.148
Factors associated with RFA of recurrent CLM
Interval from hepatectomy to recurrence (>12 months as ref.)
≤12 months	2.13 (1.22–3.72)	0.008	2.00 (1.07–3.73)	0.029	1.17 (0.61–2.23)	0.642
No. of CLM (unique as ref.)
Multiple	1.15 (0.78–1.70)	0.487			1.21 (0.74–1.98)	0.444
Size of largest CLM (≤3 cm as ref.)
>3 cm	1.16 (0.54–2.47)	0.704			1.30 (0.47–3.58)	0.616
CEA level at ablation (≤5 ng/ml as ref.)
>5 ng/ml	1.87 (1.25–2.79)	0.002	1.57 (1.04–2.36)	0.032	2.48  (1.47–4.16)	<0.001	2.05 (1.16–3.62)	0.013
Extrahepatic disease (no as ref.)
Yes	1.17 (0.75–1.81)	0.496			3.27 (1.98–5.41)	<0.001	1.88 (1.07–3.32)	0.028
Preablation chemotherapy (none as ref.)
Done	1.08 (0.72–1.62)	0.697			0.89 (0.54–1.48)	0.662
Postablation chemotherapy (none as ref.)
Done	0.86 (0.59–1.26)	0.441			1.30 (0.79–2.13)	0.298
Ablation margin (≥5 mm as ref.)
<5 mm	1.92 (1.31–2.81)	<0.001	1.69 (1.14–2.50)	0.009	2.05 (1.27–3.33)	0.003	1.44 (0.84–2.49)	0.188

RFA, radiofrequency ablation; CLM, colorectal liver metastases; CEA, carcinoembryonic antigen; rIHR, repeat intrahepatic recurrence; OS, overall survival; SHR, subdistribution hazard ratio; HR, hazard ratio.

A separate analysis of risk factors associated with LTP and IDR was conducted based on the total cohort. The univariate fine-gray analysis revealed several prognostic factors associated with LTP and IDR (Supplementary Table 1). Multivariate fine-gray analysis identified mutant RAS (SHR, 2.02; 95% CI, 1.05–3.88; p = 0.036) and ablation margin <5 mm (SHR, 9.33; 95% CI, 4.26–20.4; p < 0.001) as risk factors of LTP. Mutant RAS (SHR, 1.84; 95%CI, 1.27–2.68; p = 0.001), interval from hepatectomy to intrahepatic recurrence ≤12 months (SHR, 2.35; 95% CI, 1.34–4.12; p = 0.003), and CEA level at ablation >5 ng/ml (SHR, 1.55; 95%CI, 1.05–2.28; p = 0.028) were independent risk factors of IDR.

Construction and validation of nomograms

Nomograms were constructed based on the multivariate model to estimate the rIHR and OS rates after RFA for recurrent CLMs (Figures 2 and 3). The C-index of the rIHR nomogram through 1 and 2 years was 0.675 and 0.669 for the training cohort, and 0.722 and 0.704 for the validation cohort. The C-index of the OS nomogram through 3 and 5 years was 0.719 and 0.702 for the training cohort, and 0.714 and 0.665 for the validation cohort. In both cohorts, nomograms also demonstrated good calibration for rIHR and OS predictions (Figures 4 and 5).

Figure 2. Nomograms for predicting 1- and 2-year rIHR rates after RFA for patients with recurrent CLMs. RFA, radiofrequency ablation; CLM, colorectal liver metastases; CEA, carcinoembryonic antigen; rIHR, repeat intrahepatic recurrence.

Figure 3. Nomograms for predicting 3- and 5-year OS rates after RFA for patients with recurrent CLMs. RFA, radiofrequency ablation; CLM, colorectal liver metastases; CEA, carcinoembryonic antigen; OS, overall survival.

Figure 4. Calibration curves for rIHR after RFA based on nomogram prediction and actual observation. (A–B) 1- and 2-year rIHR rates in the training cohorts. (C–D) 1- and 2-year rIHR rates in the validation cohorts. RFA, radiofrequency ablation; rIHR, repeat intrahepatic recurrence; OS, overall survival.

Figure 5. Calibration curves for OS after RFA based on nomogram prediction and actual observation. (A–B). 3- and 5-year OS rates in the training cohorts. (C–D) 3- and 5-year OS rates in the validation cohorts. RFA, radiofrequency ablation; OS, overall survival.

Using the nomogram, each prognostic variable gets a point score (Supplementary Tables 2 and 3). The cohort was divided into low-risk (≤167) and high-risk (>167) groups according to the median nomogram-predicted rIHR and low-risk (≤100) and high-risk (>100) groups according to the median nomogram-predicted OS (Figure 6). The 1- and 2-years rIHR rates were 35.1% and 51.9% for the low-risk groups, and 79.1% and 86.0% for the high-risk groups, respectively (p < 0.001). The 3- and 5-years OS rates were 79.3% and 58.4% for the low-risk groups, and 40.9% and 16.9% for the high-risk groups, respectively (p < 0.001).

Figure 6. The risk score was closely related to the prognosis of recurrent CLM patients treated with RFA. (A) Kaplan–Meier curves of OS. (B) CIF curves of rIHR. RFA, radiofrequency ablation; CLM, colorectal liver metastases; rIHR, repeat intrahepatic recurrence; OS, overall survival; CIF, cumulative incidence function.

Discussion

The recent expansion of surgical indications has increased the number of patients with potentially resectable disease, but it has also increased the risk of recurrence [20]. In this situation, thermal ablation is a valid treatment option for small recurrent CLMs according to the guidelines of the European Society for Medical Oncology [6]. This single-center cohort study reviewed the data on RFA in patients with recurrent CLMs from the last 10 year at our institution, and identified mutant RAS, interval from hepatectomy to intrahepatic recurrence ≤ 12 months, CEA level >5 ng/ml, and ablation margin <5 mm as independent predictors for rIHR. Mutant RAS, largest CLM at hepatectomy >3 cm, CEA level >5 ng/ml, and extrahepatic disease were independent predictors of poor OS. Furthermore, two nomograms constructed with respective prognostic factors to predict the rIHR and OS of an individual after RFA showed good performance in terms of discrimination and calibration. To our knowledge, our study is the first to develop nomograms for patients with recurrent CLMs who were treated with RFA.

In this study, the 1-, 2-, and 3-year rIHR rates were 58.8%, 70.2%, and 74.2%, respectively. High rIHR rates after RFA have been reported by prior studies. Valls et al. [21] reported that of 59 recurrent CLM cases treated with RFA, the 1- and 2-year liver disease-free survival rates were 54% and 24%, respectively. Zimmermann et al. [8] analyzed 23 patients treated with RFA for recurrent CLMs after major hepatectomy and reported a 1-year rIHR rate of 65%. While approximately 70% of patients develop rIHR within 2 years, RFA can be easily repeated to treat small-size recurrent CLM [5] and in turn, an improved survival [20]. Compared with rIHR, OS provides a more direct evaluation of prognosis. In this cohort, the 5-year OS after RFA was 38.5% and the median OS was 49 months, which was consistent with a recent study [22].

Nomograms are statistical tools that use line graphs for prediction. The OS nomogram can assist patients with evaluating the potential benefits of RFA and making personalized decisions. Patients who are predicted as low-risk by the OS nomogram have a good prognosis, and RFA is a reasonable therapy for such patients. The rIHR nomogram helped determine the surveillance intensity after RFA. Identification of a high-risk may prompt clinicians to enhance the frequency of abdominal radiological surveillance. Early detection of tumors significantly increases the probability of complete eradication, as smaller tumors are generally more amenable to complete ablation.

RAS mutations occur in approximately 40–50% of patients undergoing CLM resection [23] and have been associated with lower effectiveness of preoperative chemotherapy [24] and poor OS in patients with resectable CLMs [19]. Many studies [11,15,25] have reported that mutant RAS patients have a higher risk of intrahepatic recurrence after local treatment. In this study, patients with mutant RAS had a higher incidence of rIHR and reduced OS rates than those with wild-type RAS. Multivariate analysis also confirmed mutant RAS was an independent adverse prognostic factor after RFA for recurrent CLM. In a study of 97 patients undergoing CLM ablation, OS and time to new liver metastases was significantly shorter in mutant RAS patients. Upon multivariate analysis, mutant RAS was an independent predictor for worse outcomes after ablation of CLMs, including OS (p = 0.009, HR: 2.0) and new liver metastases (p = 0.037, SHR: 2.0) [15]. Additionally, many studies [15,26,27] have reported a higher LTP rate in tumors with mutant RAS after ablation, which may be attributed to more aggressive tumor biology.

Many publications have shown that the ablation margin is a critical predictor of LTP [14, 15, 26, 27]. The current study revealed that an ablation margin of <5 mm was associated with higher incidence of LTP and rIHR after RFA. In studies using a conventional manual margin assessment, an ablation margin of >10 mm was associated with the best local tumor control, while a 5 mm margin is the minimum requirement [18, 28–30]. However, the accuracy of conventional manual measurement may be limited by the misalignment of the liver. Several studies [31–33] have reported using 3D volumetric quantification technology to enhance the accuracy of measuring ablation margins. The metabolic information provided by PET can also further improve the assessment of the ablation zone [34–36]. Recent papers [32, 37] have shown that immediate intraprocedural 3D margin assessment is superior to margin assessment at 4–8 weeks post-ablation for predicting LTP. Intraoperative identification of lesions with suboptimal margins can guide decisions to perform additional ablations during the same treatment session. Additionally, performing a biopsy of the ablation zone intraprocedurally can further optimize thermal ablation as local curative therapy for CLM [38, 39]. Therefore, to achieve adequate local tumor control, standardized intraprocedural evaluation and reproducible methods for margin quantification are urgent.

Currently, CEA is the most widely used biomarker in CRC and is a predictor for the benefit of perioperative chemotherapy [40]. In this study, patients with CEA levels >5 ng/mL exhibited a higher rIHR rate and a worse OS rate than those with lower levels. Therefore, elevated CEA levels might reflect the presence of micro-metastases and aggressive tumor biology [40]. A short interval from hepatectomy to recurrence usually reflects unidentified metastases or microscopic remnant disease [41]. In our study, patients who presented with intrahepatic recurrence within 12 months of hepatectomy experienced rIHR more frequently. Therefore, several publications recommend RFA as a “test-of-time” strategy for patients with early recurrence [42]. Local tumor control by RFA provides time for the development of unresectable multifocal disease, allowing these patients to avoid unnecessary surgery. Additionally, a previous study showed that approximately 10% of patients had concomitant extrahepatic disease at the time of intrahepatic recurrence after hepatectomy [41]. Although the involvement of extrahepatic organs indicates poor outcomes, patients with limited and treatable extrahepatic disease can also benefit from ablation for CLMs [43].

In general, tumor size and number at RFA have been considered key predictors for recurrence and survival, but these indicators failed to show statistical significance in our data. This may be explained by the fact that recurrent CLMs are detected early after hepatectomy due to the close follow-up regimen, meaning that usually only small-sized and solitary recurrent CLMs are observed [22]. Notably, the present study suggests that pre- and post-ablation chemotherapy are less significant predictors of rIHR and OS after RFA. This finding is in line with a previous report showing that adding neoadjuvant chemotherapy before repeated local treatment did not improve OS and progression-free survival [44].

There are some limitations to this study. First, using conventional manual measurement, comparing pre- and post-RFA two-dimensional CT/MR images, may introduce operator bias and fail to consider biomechanical changes in the liver related to the ablation procedure, potentially limiting the accuracy of ablation margin assessment. Second, although inclusion of RAS mutation status in the rIHR nomogram led to improved prognostic accuracy, the discrimination of the rIHR prediction nomogram remained modest; future studies are needed to incorporate additional somatic gene mutations to update the nomogram. Third, since this was a retrospective study, information and selection bias may be present. Thus, external validation in a multicenter prospective study is still needed to confirm the performance of our nomograms.

In conclusion, these established nomograms may predict the individual risk of rIHR and OS in recurrent CLM patients treated with RFA and may contribute to better individualized management.

Disclosure statement

No potential conflict of interest was reported by the author(s).

Data availability statement

The data presented in this study are available on request from the corresponding authors.

Additional information

Funding

This work was supported by Capital’s Funds for Health Improvement and Research (2020-2-2152).