Impact of time to ablation on survival outcomes in patients with colorectal liver metastases: an inverse probability-weighted cohort study

Abstract

Background

The optimal timing for thermal ablation (TA) in patients with colorectal liver metastases (CRLM) eligible for local curative therapy remains undefined. This study aimed to evaluate the impact of time to ablation (TTA) on long-term survival outcomes.

Methods

We retrospectively analyzed 220 patients with ablatable CRLM (≤ 5 liver metastases, maximum diameter ≤ 3 cm) who underwent TA between January 2015 and December 2022. Based on the TTA, patients were stratified into three groups: upfront TA (UTA, ≤ 3 months, n = 113), short-term delayed TA (DTA, 3-6 months, n = 59), and long‑term DTA (> 6 months, n = 48). Inverse probability of treatment weighting (IPTW) was applied to reduce baseline selection bias. The primary endpoint was progression‑free survival (PFS), and the secondary endpoint was overall survival (OS).

Results

After IPTW adjustment, the difference in PFS among the three groups was statistically significant (P = 0.012). The 1‑, 3‑, and 5‑year PFS rates were 61.5%, 31.2%, and 30.2% in the UTA; 47.7%, 15.9%, and 4.7% in the short‑term DTA; and 53.4%, 8.5%, and 3.1% in the long‑term DTA. Multivariable Cox regression identified both short-term DTA and long‑term DTA as an independent risk factor for worse PFS (HR = 1.80. 95% CI: 1.20−2.70, PP = 0.005); and HR = 1.99, 95% CI: 1.20–3.31, P = 0.008). No significant difference in OS was observed among the three groups. Subgroup analyses showed that DTA was associated with poorer PFS than UTA in patients with CEA > 5 ng/mL, synchronous metastases, N1−2 stage, or lymphovascular invasion. Moreover, among patients receiving UTA, adding neoadjuvant chemotherapy did not provide additional PFS or OS benefit compared with ablation alone (P > 0.05).

Conclusion

In patients with ablatable CRLM, UTA significantly improves PFS compared with DTA, and DTA is an independent risk factor for worse PFS. Therefore, for eligible patients, UTA is recommended in preference to prolonged systemic therapy.

Supplementary Information

The online version contains supplementary material available at 10.1186/s12885-026-16333-3.

Introduction

Statistics indicate that over 50% of colorectal cancer (CRC) patients develop liver metastases during the disease course, with approximately 20%−30% presenting with synchronous metastases at initial diagnosis [1, 2]. Colorectal liver metastases (CRLM) are one of the leading causes of death in CRC patients, significantly reducing overall survival (OS) [3]. For patients with untreated and unresectable CRLM, the median OS is less than one year; even when subjected to first-line chemotherapy, the 5-year OS remains consistently below 20% [4, 5].

Although surgical resection remains the standard of care for CRLM [6], thermal ablation (TA) has gained wide acceptance as a parenchyma-sparing alternative with curative intent. In recent years, international clinical guidelines, including those from the European Society for Medical Oncology (ESMO) and the National Comprehensive Cancer Network (NCCN), have recommended TA as a treatment option for patients with small-sized, oligometastatic CRLM that are technically amenable to ablation with adequate safety margins [7, 8]. The EORTC 2025 consensus further emphasizes that for patients with ≤ 3 cm resectable CRLM, TA should be considered equally effective as surgical resection [9]. Several randomized trials have shown that in selected patients, TA can provide long-term tumor control and survival rates comparable to surgery, while being less invasive, repeatable, and associated with fewer complications [10, 11].

At present, there is still no clear clinical evidence to support the optimal timing for TA in patients with CRLM, and the decision mainly relies on the comprehensive assessment of the multidisciplinary team [9]. In practice, the timing of ablation is not uniform: some eligible patients are treated early, whereas others, due to high tumor burden or the need for an initial systemic treatment, have their ablation postponed [12, 13]. This variation may reflect differences in tumor biology, treatment sensitivity, and clinical decision-making preferences. For patients with CRLM undergoing curative-intent TA, how the timing of the procedure influences long-term survival remains unclear. A critical and unresolved question is whether delaying ablation, as opposed to performing it promptly, negatively impacts patient progression-free survival (PFS) and OS.

Thus, this study sought to evaluate the impact of the interval between the initial diagnosis of CRLM and the first curative-intent TA on long-term survival outcomes.

Materials and methods

Patients

This study was approved by the Institutional Review Board of two hospitals (IRB No. 25/104–5050), and the requirement for informed consent was waived due to its retrospective design.

From January 2015 to December 2022, patients were consecutively enrolled in this study based on the following inclusion criteria: (1) histologically confirmed colorectal adenocarcinoma with primary tumor R0 resection; (2) diagnosis of CRLM based on biopsy or typical imaging features on contrast-enhanced imaging. (3) eligibility for curative-intent TA as determined by a multidisciplinary team (MDT), specifically defined as having ≤ 5 nodules with a maximum diameter of ≤ 3 cm; (4) absence of extrahepatic metastases as confirmed by systemic staging (contrast-enhanced CT of the chest, abdomen, and pelvis, supplemented by 18 F-FDG PET/CT or bone scintigraphy when clinically indicated); and (5) preserved liver function categorized as Child-Pugh class A or B with an Eastern Cooperative Oncology Group (ECOG) performance status (PS) of 0−2. The exclusion criteria were as follows: (1) prior history of local liver-directed therapies (hepatic resection, transarterial chemoembolization, or radiotherapy) prior to ablation; (2) history of other concurrent primary malignancies; (3) absolute contraindications to percutaneous ablation (e.g., severe cardiopulmonary dysfunction, uncorrectable coagulopathy, or active infection); and (4) incomplete medical records or a follow-up duration of less than 12 months.

Study design

Patient clinical and tumor variables collected from the hospital information system included: sex, age, administration of neoadjuvant chemotherapy (NAC) and adjuvant chemotherapy, modified clinical risk score (mCRS) [14], carcinoembryonic antigen (CEA) level, ECOG, primary tumor TNM stage, histological grade, primary tumor location, lymphovascular invasion, RAS mutation status, as well as the location, size, and number of CRLMs.

Based on previous studies [15, 16], where the cycles of NAC typically did not exceed 7 and were recommended to be completed within 1−3 months, patients were stratified into three cohorts based on the time interval from the initial diagnosis of CRLM to the definitive ablation (Time to Ablation, TTA). The upfront TA (UTA) group (TTA ≤ 3 months) primarily included patients who underwent ablation immediately following diagnosis or after a short-course NAC. The short-term delayed TA (short-term DTA) group (TTA 3−6 months) represented a cohort that typically completed a standard course of NAC or experienced minor delays due to reversible clinical factors, such as the management of transient chemotherapy-induced hepatotoxicity. The long-term DTA group (TTA > 6 months) accounted for patients with more complex clinical trajectories. This included those requiring prolonged systemic therapy to achieve an adequate therapeutic response, treatment interruptions necessitated by severe comorbidities (e.g., cardiovascular events), or significant external disruptions to healthcare delivery (e.g., the COVID-19 pandemic), as well as protracted decision-making processes during intricate multi-stage management. Details of systemic treatment regimens are provided in Supplementary materials.

Thermal ablation procedure

Based on the patient’s overall clinical condition and specific tumor characteristics (including size, number, and anatomical adjacency to major blood vessels or critical organs), the MDT conducted a comprehensive evaluation to confirm eligibility for curative-intent ablation. All procedures, including radiofrequency ablation (RFA) and microwave ablation (MWA), were performed percutaneously under ultrasound or CT guidance by three senior interventional radiologists (each with ≥ 10 years of experience in liver ablation). Interventions were conducted under either local anesthesia with intravenous sedation or general anesthesia, depending on the lesion location and patient tolerance. Energy delivery typically lasted 5–20 min per session, maintaining target temperatures between 80 °C and 110 °C to ensure adequate coagulative necrosis. To mitigate thermal injury to the liver capsule or adjacent vulnerable structures, hydrodissection was routinely employed to create a protective fluid buffer. For tumors ≥ 2 cm, we implemented a multi-needle or multi-point overlapping ablation technique to ensure complete volumetric coverage.

In accordance with the standardized terminology, initial technical success was defined as the complete coverage of the target tumor by the ablation zone, achieving a minimum circumferential safety margin of at least 5 mm. This was evaluated using intraprocedural or immediate post-procedural contrast-enhanced CT or contrast-enhanced US. If the safety margin could not be reliably assessed immediately, confirmation via contrast-enhanced CT or MRI within 24 h was mandatory. When the initial assessment indicated inadequate safety margins (< 5 mm) or suggested residual viable tumor, supplementary ablation was performed either within the same session or within one week. Secondary technical success was defined as the achievement of complete tumor coverage confirmed by follow-up imaging following such supplementary interventions. A representative example is shown in (Fig. 1).

Fig. 1.

A 52‑year‑old man with biopsy‑proven colorectal liver metastasis (CRLM) treated with radiofrequency ablation (RFA). A Axial and B sagittal portal‑venous‑phase MR images of the liver obtained before treatment show a 1.6‑cm nodule (arrowhead) in segment VI, abutting the liver capsule, with rim enhancement. C Axial and (D) sagittal portal‑venous‑phase MR images acquired 4 weeks after ablation demonstrate complete necrosis of the tumor within the ablation zone, surrounded by a safety margin exceeding 5 mm (arrow)

Follow-up and study outcomes

A standardized follow-up protocol was strictly implemented [17, 18]. Technique efficacy was evaluated at 4−6 weeks post-ablation and was defined as the complete macroscopic ablation of the target tumor, characterized by the absence of residual active tumor or nodular enhancement on contrast-enhanced imaging. Subsequently, patients underwent routine follow-up examinations every 3 months for the first 2 years and every 6 months thereafter. Standard follow-up evaluations included physical examinations, monitoring of serum tumor markers (specifically CEA and CA 19-9), contrast-enhanced CT or MRI. If recurrence was suspected, PET/CT or biopsy was performed for confirmation. The primary outcome was PFS, defined as the interval from the date of the first technically successful ablation to the date of any disease recurrence (including local tumor progression [LTP], intrahepatic distant recurrence, or extrahepatic metastasis or death from any cause). LTP was defined as the reappearance of enhancing tumor tissue within or at the edge of the ablation zone (within a 10-mm proximity). The secondary outcome was OS. To minimize potential lead-time bias associated with the timing of the ablative intervention, OS was defined as the interval from the date of initial CRLM diagnosis to death from any cause or the last follow-up. Post-procedural complications were documented and graded according to the Clavien-Dindo classification system [19].

Statistical analysis

Continuous variables were expressed as mean ± standard deviation and compared using one-way ANOVA. Categorical variables were presented as frequencies (percentages) and compared using the Chi-square test or Fisher’s exact test. To minimize selection bias and balance baseline clinical characteristics across the three groups, inverse probability of treatment weighting (IPTW) was implemented. Propensity scores were estimated using a multinomial logistic regression model that included 18 clinically relevant covariates. To improve model stability, we used the average treatment effect among the overlap population (ATO) weighting scheme. The ATO approach down-weights patients with extreme propensity scores in a smooth manner, naturally restricting weights to values between 0 and 1, thereby preventing variance inflation caused by extreme weights. Covariate balance before and after weighting was assessed using standardized mean differences (SMD). An SMD ≤ 0.1 was considered indicative of good balance, 0.1−0.2 indicated mild imbalance still within acceptable limits, and > 0.2 indicated substantial imbalance. To visually confirm covariate balance and the absence of extreme weights, we generated a Love plot and weight histograms using the WeightIt and cobalt packages. Survival outcomes were estimated using the Kaplan-Meier method. Differences between curves were compared using the log-rank test, with pairwise comparisons performed to identify specific inter-group differences. A two-step variable selection procedure was used for prognostic analyses. Univariate Cox regression was first performed to screen each candidate variable. To avoid prematurely excluding clinically important confounders due to a conventional significance threshold (e.g., P < 0.05), we adopted a more liberal cutoff of P < 0.1 for inclusion at this screening stage. Variables meeting this criterion were then entered into a multivariable Cox proportional hazards model to identify independent prognostic factors. Subgroup analyses were conducted across different clinical strata, and forest plots were used to visually assess interaction effects. Statistical analyses were performed using SPSS 25.0 (IBM Corp., Armonk, NY, USA) for descriptive statistics and Cox regression, and R software (version 4.4.2; Vienna, Austria) for propensity score weighting, subgroup analyses, and visualization. All statistical tests were two-sided, and a P-value < 0.05 was considered statistically significant.

Results

Patient characteristics

A total of 588 patients diagnosed with CRLM were initially screened for eligibility. After applying the refined inclusion and exclusion criteria, 220 patients were ultimately enrolled in the study. These participants were stratified into UTA (n = 113), short-term DTA (n = 59), and long-term DTA (n = 48) groups for subsequent analysis (Fig. 2). Significant differences were observed among the three groups in terms of synchronous liver metastasis, NAC, ECOG PS (P < 0.05). Specifically, the UTA group exhibited higher rates of synchronous metastasis compared to the two DTA groups, while the proportion of patients receiving NAC was significantly lower. Regarding ECOG PS, both the UTA and short-term DTA had a higher proportion of patients with a score of 0 compared to the long-term DTA group. To reduce potential bias resulting from these baseline imbalances, IPTW was applied to balance the three groups. After weighting, all variables achieved good balance, with most covariates showing SMDs < 0.1. Only four variables (T stage, N stage, RAS status, and number of CRLMs) had SMDs slightly above 0.1 (range, 0.107−0.117) (Table 1; Fig. 3). Neoadjuvant chemotherapy response and ablation outcomes.

Fig. 2.

Patient selection flow chart. UTA, upfront thermal ablation, DTA, delay thermal ablation

Table 1.

Baseline characteristics before and after IPTW

Variable	UTA (n = 113)	Short-term DTA (n = 59)	Long-term DTA (n = 48)	Before IPTW		After IPTW
				P value	SMD	P value	SMD
Sex				0.466	0.146	0.991	0.017
Male	72 (63.7)	37 (62.7)	35 (72.9)
Female	41 (36.3)	22 (37.3)	13 (27.1)
Age, year				0.376	0.156	0.986	0.022
< 60	60 (53.1)	35 (59.3)	31 (64.6)
≥ 60	53 (46.9)	24 (40.7)	17 (35.4)
Metastasis timing				0.002	0.394	0.953	0.041
Synchronous	65 (57.5)	23 (39.0)	14 (29.2)
Metachronous	48 (42.5)	36 (61.0)	34 (70.8)
Pre-chemotherapy				< 0.001	0.449	0.883	0.060
Yes	74 (65.5)	51 (86.4)	44 (91.7)
No	39 (34.5)	8 (13.6)	4 (8.3)
Post-chemotherapy				0.123	0.220	0.986	0.019
Yes	88 (77.9)	53 (89.8)	41 (85.4)
No	25 (22.1)	6 (10.2)	7 (14.6)
T stage				0.223	0.226	0.674	0.117
T1 − 3	59 (52.2)	27 (45.8)	30 (62.5)
T4	54 (47.8)	32 (54.2)	18 (37.5)
N stage				0.051	0.273	0.711	0.115
N0	39 (34.5)	10 (16.9)	13 (27.1)
N1 − 2	74 (65.5)	49 (83.1)	35 (72.9)
Lymphovascular invasion				0.806	0.079	0.945	0.041
No	51 (45.1)	26 (44.1)	24 (50.0)
Yes	62 (54.9)	33 (55.9)	24 (50.0)
RAS status				0.236	0.309	0.947	0.108
Wild-type	63 (55.8)	30 (50.8)	34 (70.8)
Mutant	34 (30.1)	21 (35.6)	8 (16.7)
Missing	16 (14.2)	8 (13.6)	6 (12.5)
Grade of primary cancer				0.575	0.113	0.885	0.065
Low	98 (86.7)	48 (81.4)	42 (87.5)
High	15 (13.3)	11 (18.6)	6 (12.5)
Primary tumor				0.973	0.08	0.978	0.084
Right-sided	16 (14.2)	7 (11.9)	5 (10.4)
Left-sided	26 (23.0)	14 (23.7)	11 (22.9)
Rectum	71 (62.8)	38 (64.4)	32 (66.7)
No. of CRLM				0.152	0.294	0.950	0.107
1	72 (63.7)	36 (61.0)	23 (47.9)
2 or 3	35 (31.0)	17 (28.8)	17 (35.4)
4 or 5	6 (5.3)	6 (10.2)	8 (16.7)
CRLM location				0.384	0.151	0.929	0.047
Unilobar	97 (85.8)	48 (81.4)	37 (77.1)
Bilobar	16 (14.2)	11 (18.6)	11 (22.9)
Maximum tumor size, cm				0.447	0.164	0.954	0.038
≤ 1	16 (14.2)	6 (10.2)	9 (18.8)
1 − 3	97 (85.8)	53 (89.8)	39 (81.2)
BMI, kg/m²				0.429	0.145	0.930	0.049
≤ 24	64 (56.6)	30 (50.8)	22 (45.8)
> 24	49 (43.4)	29 (49.2)	26 (54.2)
ECOG PS				0.047	0.271	0.772	0.094
0	89 (78.8)	45 (76.3)	29 (60.4)
1 − 2	24 (21.2)	14 (23.7)	19 (39.6)
mCRS				0.601	0.173	0.971	0.089
0 or 1	35 (31.0)	12 (20.3)	11 (22.9)
2 or 3	72 (63.7)	44 (74.6)	34 (70.8)
4 or 5	6 (5.3)	3 (5.1)	3 (6.2)
CEA, ng/mL				0.385	0.180	0.846	0.078
≤ 5	51 (45.1)	30 50.8)	18 (37.5)
> 5	62 (54.9)	29 (49.2)	30 (62.5)

Unless otherwise specified, data are presented as n (%)

Abbreviations: ACRLM colorectal liver metastases, IPTW inverse probability treatment weighting, UTA upfront thermal ablation, DTA delay thermal ablation, SMD standardized mean difference, BMI body mass index, ECOG Eastern Cooperative Oncology Group, mCRS modified clinical risk score, CEA carcinoembryonic antigen

Fig. 3.

Assessment of covariate balance and distribution of overlap weights. A ‘Love’ plot depicting the effect of the treatment weighting (IPTW) in terms of absolute standardized mean differences (SMD) across the three groups. The vertical dashed line at 0.1 represents the threshold for an optimal balance obtained after weighting. B Histogram depicting the distribution of the overlap weights stratified by treatment group. The overlapping distributions demonstrate good common support and the absence of extreme weights. CRLM, colorectal liver metastases; CEA, carcinoembryonic antigen; DTA, delayed thermal ablation; ECOG, Eastern Cooperative Oncology Group performance status; IPTW, inverse probability of treatment weighting; mCRS, modified clinical risk score; UTA, upfront thermal ablation

Among patients treated with NAC, treatment responses were evaluated using RECIST criteria (Table 2). A significant difference was observed in the number of NAC cycles across the three groups (P < 0.001), with the long-term DTA group receiving the highest median number of cycles. In contrast, no significant difference was found in the distribution of RECIST responses among the groups (P = 0.124). The initial technical success rate of TA was 97.8% (215/220), and the secondary technical success rate reached 100% (220/220) after supplementary ablation. The technique efficacy was 99.1% (218/220). No significant differences were found among the groups with respect to technical success, ablation time, length of hospital stay, imaging guidance modality, or ablation modality. The overall complication rate was 8.6% (19/220), including perihepatic effusion (n = 6), infection (n = 3), pneumonia (n = 4), hemorrhage (n = 2), liver dysfunction (n = 2), bowel injury (n = 2), and biliary dilation (n = 1).

Table 2.

Comparison of neoadjuvant chemotherapy and ablation parameters

Variable	UTA	Short-term DTA	Long-term DTA	P value
NAC cycles*	3 (0 − 12)	4 (0 − 14)	7 (0 − 22)	< 0.001
RECIST response†				0.124
CR	0/74 (0)	1/51 (2.0)	2/44 (4.5)
PR	64/74 (86.5)	41/51 (80.4)	36/44 (81.8)
SD	9/74 (12.2)	6/51 (11.8)	2/44 (4.5)
PD	1/74 (1.4)	3/51 (5.9)	4/44 (9.1)
Technical success	111/113 (98.2)	58/59 (98.3)	46/48 (95.8)	0.609
Ablation time (min)	15.8 ± 10.3	16.4 ± 9.1	18.3 ± 11.0	0.359
Hospital stay (days)	6.9 ± 2.7	7.5 ± 2.3	7.4 ± 2.9	0.241
Image guidance				0.574
Ultrasound	69/113 (61.1)	41/59 (69.5)	30/48 (62.5)
CT	44/113 (38.9)	18/59 (30.5)	18/48 (37.5)
Ablation modality				0.159
MWA	35/113 (31.0)	24/59 (40.7)	22/48 (45.8)
RFA	78/113 (69.0)	35/59 (59.3)	26/48 (54.2)
Complication				0.074
Yes	8/113 (7.1)	3/59 (5.1)	8/48 (16.7)
No	105/113 (92.9)	56/59 (94.9)	40/48 (83.3)

Unless otherwise specified, data are presented as n (%) or mean ± standard deviation

^*Number of NAC cycles is presented as median (range)

^†RECIST evaluation was performed only for patients who received neoadjuvant chemotherapy

Abbreviations: NAC neoadjuvant chemotherapy, RECIST Response Evaluation Criteria in Solid Tumors, CR complete response, PR partial response, SD stable disease, PD progressive disease, MWA microwave ablation, RFA radiofrequency ablation, UTA upfront thermal ablation, DTA delayed thermal ablation

Progression-free survival

The overall median PFS was 12 months, with 172 (78.2%) patients experiencing disease progression during follow-up. Specific details are provided in Table S1. In the unweighted cohort, PFS differed significantly among the three groups (P = 0.036). The median PFS was 16.0months for UTA, 11.0 months for short‑term DTA, and 12.0 months for long‑term DTA. The 1‑, 3‑, and 5‑year PFS rates were 58.6%, 23.1%, and 21.6% for UTA; 48.0%, 15.2%, and 7.8% for short‑term DTA; and 48.7%, 10.8%, and 5.4% for long‑term DTA. After IPTW adjustment, the differences in PFS among the three groups remained statistically significant (P = 0.012). The median PFS was 16.0months for UTA, 12.0 months for short‑term DTA, and 13.0 months for long‑term DTA. The 1‑, 3‑, and 5‑year PFS rates were 61.5%, 31.2%, and 30.2% for UTA; 47.7%, 15.9%, and 4.7% for short‑term DTA; and 53.4%, 8.5%, and 3.1% for long‑term DTA (P = 0.012). Post‑hoc pairwise comparisons indicated that UTA yielded significantly better PFS than both short‑term DTA (HR = 1.64, P = 0.022) and long‑term DTA (HR = 1.85, P = 0.005), whereas no significant difference was observed between the two delayed ablation strategies (HR = 1.18, P = 0.474) (Fig. 4, Table S2).

Fig. 4.

Kaplan-Meier curves of progression-free survival in the unweighted (A) and IPTW-weighted (B) cohorts. IPTW, inverse probability treatment weighting, UTA, upfront thermal ablation, DTA, delay thermal ablation

In the IPTW‑adjusted multivariable analysis, the timing of ablation was confirmed as an independent prognosticator: compared with UTA, both short-term DTA (HR = 1.80, 95% CI: 1.202.70, P = 0.005) and long‑term DTA (HR = 1.99, 95% CI: 1.20−3.31, P = 0.008) was associated with a significantly increased risk of disease progression. Moreover, mCRS 4−5 (HR = 1.62, 95% CI: 1.05−2.49, PP = 0.028) and CEA > 5 ng/mL (HR = 1.85, 95% CI: 1.30−2.63, PP < 0.001) were also independent risk factors for worse PFS (Table 3).

Table 3.

Univariable and multivariable Cox regression analyses for progression‑free survival

Variables	Univariable analysis		Multivariable analysis
	HR (95% CI)	P value	HR (95% CI)	P value
Time to ablation
Short-term DTA	1.60 (1.08–2.38)	0.018	1.80 (1.20–2.70)	0.005
Long-term DTA	1.68 (1.03–2.75)	0.040	1.99 (1.20–3.31)	0.008
Female sex	1.34 (0.98–.84)	0.067	1.37 (1.00–.88)	0.051
Age ≥ 60 years	0.89 (0.65–1.20)	0.440
Synchronous metastasis	1.15 (0.85–1.56)	0.350
Pre-chemotherapy, yes	1.06 (0.75–1.50)	0.741
Post-chemotherapy, yes	0.77 (0.51–1.16)	0.213
T4	0.75 (0.56–1.02)	0.063	0.74 (0.54–1.01)	0.058
N 1 − 2	1.12 (0.80–1.57)	0.518
Lymphovascular invasion	0.96 (0.71–1.29)	0.765
RAS status
Mutant	1.04 (0.74–1.47)	0.826
Missing	0.95 (0.61–1.50)	0.840
High grade	1.12 (0.75–1.68)	0.564
Primary tumor
Left-sided	0.99 (0.58–1.68)	0.962
Rectum	1.19 (0.75–1.88)	0.457
No. of CRLM
2 or 3	1.21 (0.87–1.68)	0.262
4 or 5	1.14 (0.66–1.96)	0.636
Bilobar location	1.08 (0.72–1.61)	0.703
Tumor size, 1–3 cm	1.15 (0.73–1.80)	0.542
BMI > 24 kg/m²	0.96 (0.71–1.30)	0.802
ECOG PS, 1 − 2	1.19 (0.85–1.66)	0.304
mCRS
2 or 3	.52 (0.17–1.57)	0.246	0.34 (0.08−1.50)	0.153
4 or 5	1.40 (0.96–2.04)	0.080	1.62 (1.05−2.49)	0.028
CEA > 5 ng/mL	1.59 (1.11–2.28)	0.012	1.85 (1.30−2.63)	< 0.001

Abbreviations: HR hazard ratio, CI confidence interval, UTA upfront thermal ablation, DTA delayed thermal ablation, CRLM colorectal liver metastases, BMI body mass index, ECOG PS Eastern Cooperative Oncology Group performance status, mCRS, modified clinical risk score, CEA carcinoembryonic antigen

Overall survival

The overall median OS was 44.5 months, with 97 (44.1%) deaths occurring during the follow-up period. Before IPTW adjustment, no statistically significant difference in OS was observed among the three groups (P = 0.174). The median OS was 72.0 months for UTA, 64.0 months for short‑term DTA, and 52.0 months for long‑term DTA. The 1‑year OS rate was 100.0% in all groups. The 3‑year OS rates were 75.7% for UTA, 84.5% for short‑term DTA, and 67.7% for long‑term DTA. The 5‑year OS rates were 62.3%, 61.5%, and 39.9%, respectively. After IPTW, the difference in OS remained non‑significant (P = 0.687). The median OS was 69.0 months for UTA, 63.0 months for short‑term DTA, and 67.0 months for long‑term DTA. All groups maintained a 1‑year OS rate of 100.0%. The 3‑year OS rates were 74.5%, 84.1%, and 70.5%, respectively. The 5‑year OS rates were 56.6%, 55.7%, and 51.5%, respectively (Fig. 5, Table S2).

Fig. 5.

Kaplan-Meier curves of overall survival in the unweighted (A) and IPTW-weighted (B) cohorts. IPTW, inverse probability treatment weighting, UTA, upfront thermal ablation, DTA, delay thermal ablation

Multivariable Cox regression analysis showed that the timing of ablation (UTA vs. DTA) was not an independent prognostic factor for OS in either the unweighted or weighted cohorts. In the IPTW‑adjusted multivariable analysis, synchronous metastasis (HR = 2.13, 95% CI: 1.21−3.75, P = 0.009) and having 2−3 tumors (HR = 2.45, 95% CI: 1.07−5.63, P = 0.034) were identified as independent risk factors for OS. A high grade of primary cancer was also shown to be an independent prognostic factor (HR = 0.37, 95% CI: 0.14−0.98, P = 0.046) (Table 4).

Table 4.

Univariable and multivariable Cox regression analyses for overall survival

Variables	Univariable analysis		Multivariable analysis
	HR (95% CI)	P value	HR (95% CI)	P value
Time to ablation
Short-term DTA	1.15 (0.66–2.02)	0.616	1.18 (0.67–2.08)	0.570
Long-term DTA	1.50 (0.75–3.01)	0.251	1.53 (0.79–2.99)	0.210
Female sex	1.36 (0.72–2.56)	0.344
Age ≥ 60 years	1.27 (0.73–2.21)	0.399
Synchronous metastasis	1.75 (1.03–2.98)	0.039	2.13 (1.21–3.75)	0.009
Pre-chemotherapy, yes	1.02 (0.60–1.75)	0.939
Post-chemotherapy, yes	1.11 (0.62–1.97)	0.726
T4	1.20 (0.68–2.10)	0.527
N1 − 2	0.84 (0.47–1.50)	0.548
Lymphovascular invasion	0.75 (0.43–1.33)	0.326
RAS status
Mutant	0.73 (0.42–1.27)	0.267	0.95 (0.57–1.59)	0.857
Missing	0.47 (0.20–1.10)	0.082	0.48 (0.21–1.09)	0.078
High grade	0.52 (0.25–1.09)	0.084	0.37 (0.14–0.98)	0.046
Primary tumor
Left-sided	1.27 (0.46–3.52)	0.647
Rectum	0.93 (0.48–1.80)	0.821
No. of CRLM
2 or 3	1.95 (1.07–3.53)	0.028	2.45 (1.07–5.63)	0.034
4 or 5	1.45 (0.65–3.22)	0.366	2.18 (0.74–6.40)	0.157
Bilobar location	1.59 (0.95–2.67)	0.079	0.95 (0.43–2.11)	0.899
Tumor size, 1–3 cm	1.06 (0.50–2.25)	0.869
BMI > 24 kg/m²	0.97 (0.55–1.69)	0.909
ECOG PS, 1 − 2	1.49 (0.87–2.55)	0.145
mCRS
2 or 3	0.81 (0.49–1.33)	0.405
4 or 5	2.67 (0.68–10.41)	0.158
CEA > 5 ng/mL	1.73 (1.03–2.92)	0.040	1.40 (0.77–2.55)	0.264

Abbreviations: HR hazard ratio, CI confidence interval, UTA upfront thermal ablation, DTA delayed thermal ablation, CRLM colorectal liver metastases, BMI body mass Index, ECOG PS Eastern Cooperative Oncology Group performance status, mCRS modified clinical risk score, CEA carcinoembryonic antigen

Subgroup analysis

Subgroup analyses revealed that DTA yielded significantly poorer PFS than UTA in several high-risk cohorts: CEA > 5 ng/mL (HR = 1.65, 95% CI: 1.10−2.48, P = 0.016), synchronous metastases (HR = 1.55, 95% CI: 1.01−2.39, P = 0.045), N1−2 stage (HR = 1.57, 95% CI: 1.10−2.24, P = 0.012), and lymphovascular invasion (HR = 1.65, 95% CI: 1.10−2.50, P = 0.017). A significant PFS benefit for upfront ablation was also observed in males (HR = 1.64, 95% CI: 1.12–2.38, P = 0.010), patients with ECOG PS 0 (HR = 1.48, 95% CI: 1.04−2.11, P = 0.030), and those receiving pre-chemotherapy (HR = 1.55, 95% CI: 1.09−2.22, P = 0.015) or post-chemotherapy (HR = 1.58, 95% CI: 1.14−2.20, P = 0.006). For OS, DTA was associated with worse outcomes solely in patients with CEA > 5 ng/mL (HR = 2.13, 95% CI: 1.18−3.84, P = 0.012) or ECOG PS 1−2 (HR = 2.44, 95% CI: 1.08−5.50, P = 0.032). No significant differences were found in other subgroups (Figures S1, S2).

Moreover, among patients undergoing UTA, we compared the strategies of NAC followed by TA versus TA alone. No statistically significant difference was found between the two approaches for either OS (P = 0.466) or PFS (P = 0.343) (Figures S3).

Discussion

TA has emerged as an established minimally invasive alternative to surgical resection for CRLM. However, the optimal timing for this intervention remains a topic of ongoing clinical debate. In this study, we examined the impact of TTA on long-term oncological outcomes in patients with ablatable CRLM. After IPTW was applied to balance baseline characteristics, our results indicated that UTA was associated with significantly superior PFS compared to DTA (P < 0.001). Both short-term DTA (HR = 1.80, P = 0.005) and long-term DTA (HR = 1.99, P = 0.008) were identified as independent risk factors for inferior PFS. Furthermore, subgroup analyses revealed that the survival disadvantage associated with DTA was more pronounced in patients with high-risk clinical features, such as elevated CEA levels and synchronous metastases.

Li et al. [12] conducted a comparative analysis between patients who received TA prior to systemic therapy and those who underwent 2−3 months of systemic therapy before ablation. The results indicated a statistically significant improvement in disease-free survival (DFS) in the early ablation group (median DFS: 1.95 years vs. 1.10 years, P = 0.007). This observation aligns with the findings of the current study, collectively reinforcing the clinical value of timely ablation in enhancing local tumor control. In contrast, the analysis by Chen et al. [20] of 281 patients who underwent surgical resection for CRLM demonstrated that while OS from initial diagnosis was similar across groups (P > 0.05), prolonged time to surgery (TTS > 6 months) was independently associated with worse postoperative OS (HR = 1.6, P < 0.01). They also suggested that an intermediate TTS of 3–6 months might yield better long-term survival. Consistent with the surgical findings reported by Chen et al. [20], our study found no significant differences in OS rates among the three groups, and UTA showed no significant survival advantage. However, UTA (≤ 3 months) provided direct benefits in PFS, while even short-term DTA significantly increased recurrence risk (HR = 1.84, P = 0.003). The discrepancy between the significant PFS benefit and non-significant OS advantage associated with UTA may be attributed to the effectiveness of subsequent salvage therapies (e.g., repeat ablation, systemic therapy) upon disease recurrence, which could mitigate the initial survival disadvantage. OS remains a complex endpoint influenced by multiple lines of treatment and underlying tumor biology beyond the initial local intervention. Therefore, this finding reinforces the premise that, in patients with ablatable CRLM, the definitive oncologic benefit lies with prompt local intervention rather than a deferred strategy. From a theoretical perspective, TA offers the capacity to preemptively eradicate detectable lesions, preventing their transition toward aggressive phenotypes or the development of occult micrometastases. Although OS remains a multifaceted endpoint shaped by intrinsic tumor biology, host immune response, and sequential therapies, the primary clinical utility of UTA is the optimization of the therapeutic window [21, 22]. By arresting disease progression early, UTA not only extends the interval to recurrence but also significantly mitigates the cumulative psychological and physiological burden on the patient.

In clinical practice, DTA often reflects a “test of time” strategy, where clinicians defer local intervention to monitor tumor biology or response to systemic therapy [23]. In the unadjusted analysis, synchronous metastasis was associated with an increased likelihood of assignment to the DTA group. This association may reflect the clinical practice of prioritizing management of the primary tumor, together with the higher disease burden typically observed in these patients, both of which contribute to a prolonged interval before local treatment [20]. Leal et al. similarly noted that treatment delays are often multifactorial, stemming from a combination of patient factors, physician biases, and resource availability [24]. However, such inherent selection bias frequently obscures the independent prognostic value of treatment timing. To mitigate this, we applied IPTW to balance observed baseline covariates. We acknowledge that IPTW has inherent limitations. It only adjusts for measured confounders; unmeasured factors like NAC regimen, physician preferences, and patient frailty may still bias outcomes [25]. It also cannot account for the wide range of subsequent salvage therapies that strongly influence survival. This finding suggests that delaying intervention provides a window of opportunity for biologically aggressive clones to proliferate and potentially develop extrahepatic micrometastases [26, 27]. In other words, delayed ablation may be not only a cause of poorer prognosis but also the result of its selection for patients with more aggressive tumor biology.

The role of NAC in CRLM remains controversial. Some studies have shown that its combined use with TA has different effects on survival rates [28–30]. In this study, the proportion of patients receiving NAC and the number of NAC cycles were significantly higher in the DTA group compared to the UTA group (P < 0.001). Although IPTW was applied to balance baseline characteristics, residual differences in NAC utilization persisted between the groups. This is an inevitable consequence of the inherent clinical distinctions between the UTA and DTA strategies. Notably, our analysis within the UTA subgroup revealed no statistically significant survival benefit for patients receiving NAC followed by TA compared to those undergoing TA alone (P = 0.466 for OS; P = 0.343 for PFS). This reinforces the premise that within the early therapeutic window, the prompt attainment of local control itself is the paramount factor, rather than the routine integration of NAC.

It is important to emphasize that the more intensive NAC regimen in the DTA group did not translate into a survival benefit. Subgroup analyses showed that the PFS benefit associated with UTA was observed across most clinical subgroups, including those with high-risk features such as synchronous metastasis, positive primary lymph nodes, lymphovascular invasion, or CEA > 5 ng/mL. These findings suggest that the advantage of UTA is not limited to low-risk patients. For patients eligible for curative-intent ablation, proceeding with UTA rather than delaying the intervention may help reduce the opportunity for disease progression or dissemination during the waiting period [24]. This study has several limitations. First, the retrospective design introduces inherent selection bias. Although we used IPTW to adjust for measured confounders, unmeasured factors such as physician treatment preferences, socioeconomic status and subsequent salvage therapies may still have influenced treatment allocation and outcomes. Bias may not have been fully eliminated. Second, the timing of ablation was based on clinical decision-making rather than randomized assignment. For analytical purposes, our categorization of time-to-ablation into predefined intervals, may not fully capture the individualized clinical rationale behind each treatment path. Third, our study did not perform further stratification based on ablation margins. Previous studies showed that the ablation margin is a key factor affecting local tumor control [31, 32], and visual assessment based on postoperative imaging is inherently less accurate than dedicated 3D-quantitative software [33]. Therefore, future prospective studies should routinely adopt ablation margin assessment based on 3D quantitative software to achieve precise stratification and optimize local tumor control. Finally, our analysis did not account for the heterogeneity in systemic chemotherapy regimens, which could represent an unmeasured confounding variable influencing survival outcomes. Future studies should aim to include a larger patient population, adopt a prospective design, and conduct more detailed analyses of specific populations receiving delayed ablation treatment to confirm our conclusions.

Conclusion

In conclusion, for patients with ablatable CRLM, UTA significantly improves PFS compared with DTA, and DTA (both short-term and long-term) is identified as an independent risk factor for worse PFS. No significant difference in OS was observed among the three groups. Moreover, routine addition of NAC did not confer a survival benefit in patients receiving UTA. Therefore, UTA is recommended over prolonged systemic therapy for eligible patients.

Abbreviations

CEA

Carcinoembryonic antigen

CRLM

Colorectal liver metastases

DTA

Delayed thermal ablation

ECOG PS

Eastern Cooperative Oncology Group performance status

IPTW

Inverse probability of treatment weighting

mCRS

Modified clinical risk score

MWA

Microwave ablation

NAC

Neoadjuvant chemotherapy

OS

Overall survival

PFS

Progression‑free survival

RECIST

Response Evaluation Criteria in Solid Tumors

RFA

Radiofrequency ablation

TA

Thermal ablation

TTA

Time to ablation

UTA

Upfront thermal ablation

Authors’ contributions

X.Y. Conceptualization, Data Curation, Formal Analysis, Investigation, Methodology, Software, Visualization, Writing—Original Draft.Y.K. Conceptualization, Data Curation, Formal Analysis, Investigation, Resources, Writing—Original Draft.X.Z. Conceptualization, Funding Acquisition, Software, Supervision, Validation, Writing—Review & Editing.F.T. Formal Analysis, Visualization, Writing—Review & Editing.X.H. Methodology, Resources, Validation, Writing—Review & Editing.X.C. Project Administration, Writing—Review & Editing.

Funding

This study was funded by the Grant of National Natural Science Foundation of China (No. 62271509) and Research Project of Fujian Ministry of Science and Technology (No. 2019Y9053).

Data availability

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

Declarations

Ethics approval and consent to participate

This retrospective study was approved by the Ethics Committees of the two participating centers (Approval Number: 25/104–5050), and the requirement for informed consent was waived. The study was conducted in accordance with the Declaration of Helsinki.

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.