Impact of surgery for colorectal cancer liver metastasis: Systematic review and meta-analysis

Abstract

BACKGROUND

Colorectal cancer is a major public health issue, with liver metastasis marking a critical and prognostically significant pathway for disease progression. This meta-analysis evaluated the association between surgical intervention for colorectal liver metastases and survival outcomes to inform multidisciplinary treatment decisions. We performed a thorough search of MEDLINE, EMBASE, and the Cochrane Central Register of Controlled Trials from their inception until February 1, 2024. Included studies enrolled at least 20 patients and compared overall survival (OS) and/or progression-free survival (PFS) between surgical and non-surgical management, reporting hazard ratios (HRs) with 95% confidence intervals (CIs) or providing sufficient time-to-event data to estimate HRs. We pooled HRs using a random-effects model. Of 2935 records, 67 studies involving 368380 patients were included, most of which were retrospective cohorts. Surgical resection was associated with improved OS (pooled HR = 0.38, 95%CI: 0.34-0.43) and PFS (pooled HR = 0.46, 95%CI: 0.31-0.66). Subgroup analyses indicated a consistent direction of association across several study-level characteristics. Given substantial heterogeneity and the predominance of observational data, residual confounding by indication cannot be excluded; therefore, pooled estimates should be interpreted as associations in selected patients. Prospective studies incorporating contemporary systemic therapy and tumor biology are warranted to refine patient selection and treatment sequencing.

AIM

To provide an aggregate prognostic value for surgery for liver metastases, incorporating HRs with 95%CIs from multivariate or univariate analyses available in the included studies.

METHODS

Sensitivity analysis was conducted even with meta-regression based on participant ethnicity (Asian vs non-Asian), number of patients, median follow-up, publication year (pre-2015 vs 2015-2024), paper quality (high vs low), and study design (retrospective vs prospective). Heterogeneity among studies was assessed using Cochran’s Q test, with P < 0.05 or I² > 50% indicating significant heterogeneity, in which case a random-effects model (Der Simonian-Laird method) was applied. Otherwise, a fixed effects model was used. HR < 1 indicated improved survival in patients undergoing resection of liver metastases. Data were analyzed using the Review Manager (RevMan) software, version 5.4, The Cochrane Collaboration, 2020. Publication bias and small-study effects were assessed by visual inspection of funnel plots, Egger’s regression test, and rank-correlation testing; Duval and Tweedie’s trim-and-fill method was applied as a sensitivity analysis. Clinically, the attenuation of the pooled effect estimate after trim-and-fill adjustment suggests that the magnitude of survival benefit associated with surgery may be partially overestimated due to small-study effects or selective publication. However, the direction of the association remained consistent, supporting an association between surgical resection and improved survival in carefully selected patients.

RESULTS

Of the 2935 records identified, 67 studies with data from 368380 patients (ranging from 21 to 72376) were included in the meta-analysis. Most of the included studies (55 out of 67) were retrospective series, whereas 12 out of 67 were prospective (either clinical trials or prospective cohorts). The treatment strategies in the included studies consisted of upfront surgery followed by adjuvant therapy or preceded by neoadjuvant or conversion therapy (including three studies in which patients received hepatic artery infusion chemotherapy). Data regarding systemic therapies were unavailable for 17 studies. Data on the overall resection rate were available for 62 of the 67 studies. Resection rates ranged from 8% to 100% (median, 45%), whereas R0 resections ranged from 2% to 87% (median, 14%). The median follow-up period ranged from 4 months to 120 months (median, 37 months); however, it was not available in 42% of the papers.

CONCLUSION

Among the evaluable studies, the publication quality was classified as low (36%), moderate (46%), or high (18%). The association between surgery for colorectal liver metastases and survival outcomes is described in subsequent sections. Outcomes were analyzed using multivariate analysis in 90% of the cases.

Key Words: Colorectal cancer; Liver metastases; Surgery; Survival; Meta-analysis

Core Tip: The survival benefit of surgical resection for colorectal liver metastases has been debated, particularly in heterogeneous clinical settings. This systematic review and meta-analysis synthesizes available evidence and shows that surgery is associated with improved overall and progression-free survival compared with non-surgical approaches, although heterogeneity is substantial and most data are observational. Future studies should focus on defining predictive factors for benefit and identifying patient subgroups in whom alternative strategies may be more appropriate.

INTRODUCTION

Colorectal cancer (CRC) is a highly relevant pathology from a public health perspective and the third leading cause of cancer-related deaths. It has shown a growing trend in recent years among a younger population than expected, with ages below 50 years[1]. Approximately 20% of patients are diagnosed with metastatic disease, with the liver being the most common site of metastasis. The hepatic location of the disease has been the subject of numerous studies aiming to clarify both the prognostic and predictive significance of responses to medical, locoregional, and surgical treatments. Regarding medical treatments, it is noteworthy that the presence of liver metastases has been associated with reduced efficacy of systemic therapies, including chemotherapy, targeted agents, and immunotherapy[2,3]. Based on this background, it is important to identify the optimal sequencing for treating hepatic disease: Upfront surgery (when feasible) followed by systemic therapy, or neoadjuvant/perioperative (including conversion) systemic therapy followed by surgery.

Not all patients are eligible for curative surgery for liver metastases, and the survival rate of patients treated with palliative surgery for metastatic disease at the onset remains very low. This percentage is significantly influenced by the treatment center’s experience and the possibility of accessing expert hepatobiliary pathology multidisciplinary teams. This allows for a reduction in postoperative mortality and a higher rate of R0 (no residual tumor) resection, with satisfactory postoperative liver function. Furthermore, sharing cases within multidisciplinary teams prevents the loss of patients who could benefit from surgery, but are mistakenly considered non-surgical if not adequately directed to the surgeon.

The opposite principle also holds true for correct patient management: Patients eligible for surgery for metastatic disease should always be discussed with the oncologist, as the resecability principle alone cannot be discerned from the knowledge of clinicopathological and molecular characteristics. It is now well-known and established that clinicopathological characteristics, such as the sidedness (left vs right side), and molecular features, such as RAS gene mutation, may negatively impact patient prognosis[4].

This meta-analysis aimed to provide an overview of the literature on the surgical treatment of liver metastases in patients with metastatic CRC. It does not claim to provide a conclusive answer on the best management of these patients but rather offers a broad look at what has been attained so far in terms of cure in the increasingly complex, personalized, and multidisciplinary management of this malignancy.

MATERIALS AND METHODS

This study adhered to the Preferred Reporting Items for Systematic Reviews and Meta-Analyses guidelines. We conducted a comprehensive search of online databases, including MEDLINE, EMBASE, and the Cochrane Central Register of Controlled Trials, for studies published from their inception to February 1, 2024. Our search strategy included the terms (‘colorectal cancer’) AND (‘liver metastasis’) AND (conversion OR resection OR surgery OR hepatectomy OR metastasectomy) AND (survival). The inclusion criteria were as follows: (1) Studies that reported outcomes of patients treated for CRC liver metastases; (2) Studies with a minimum of 20 CRC patients; (3) Studies that compared the overall survival (OS) of patients with resected liver metastases to those who did not undergo surgery; (4) Studies that reported OS and/or progression-free survival (PFS) as hazard ratios (HRs) with 95% confidence intervals (CIs) or provided sufficient time-to-event data to estimate HRs; and (5) Studies published in English. We excluded studies that (1) Only involved other ablative techniques; (2) Included surgeries for other metastatic sites in > 20% of patients; (3) Were previous versions of the same trials; and (4) Were unavailable in the full text.

Three authors (Petrelli F, Rossitto M, and Ghidini A) independently conducted the search and identification of studies, and article selection was achieved through a consensus involving a fourth author (Zaniboni A) when necessary. Data extraction was independently performed by two authors and included author/year of publication, country, patient count, study type, chemotherapy exposure rate, survival data (HRs), overall and R0 resection rates, and the median follow-up duration. When both adjusted (multivariable) and unadjusted estimates were reported, adjusted HRs were preferentially extracted. When HRs were not directly available but sufficient time-to-event information was reported (e.g., Kaplan-Meier curves and/or log-rank statistics), log (HR) and standard errors were estimated using established methods. Two independent reviewers (Petrelli F and Ghidini A) used the Newcastle-Ottawa Scale to assess the risk of bias.

In addition to the Newcastle-Ottawa Scale, the risk of bias of the included observational studies was further evaluated using the Risk of Bias In Non-randomized Studies of Interventions (ROBINS-I) tool, which assesses potential bias across domains including confounding, selection of participants, classification of interventions, deviations from intended interventions, missing data, measurement of outcomes, and selection of reported results. Furthermore, the certainty of evidence for the primary (overall survival) and secondary (progression-free survival) outcomes was assessed using the Grading of Recommendations Assessment, Development and Evaluation (GRADE) framework. Evidence certainty was categorized as high, moderate, low, or very low based on risk of bias, inconsistency, indirectness, imprecision, and publication bias. Detailed ROBINS-I and GRADE assessments are reported in the Supplementary material.

The primary endpoint was OS with PFS as the secondary endpoint. HRs were pooled for survival analysis to provide an aggregate prognostic value for surgery for liver metastases, incorporating HRs with 95%CIs from multivariate or univariate analyses available in the included studies.

Statistical analysis

During data verification, Wu et al[5] reported an implausible hazard ratio estimate of 0.00 (95%CI: 0.00-0.01) for overall survival. Because such values are not statistically compatible with Cox proportional hazards modeling and the estimate could not be verified from the original publication, this study was excluded from the pooled analysis to preserve data integrity. Sensitivity analyses confirmed that exclusion of this study did not materially affect the pooled estimates. Sensitivity analysis was conducted even with meta-regression based on participant ethnicity (Asian vs non-Asian), number of patients, median follow-up, publication year (pre-2015 vs 2015-2024), paper quality (high vs low), and study design (retrospective vs prospective). Heterogeneity among studies was assessed using Cochran’s Q test, with P < 0.05 or I² > 50% indicating significant heterogeneity, in which case a random-effects model (Der Simonian-Laird method) was applied. Otherwise, a fixed effects model was used. HR < 1 indicated improved survival in patients undergoing resection of liver metastases. Data were analyzed using the Review Manager (RevMan) software, version 5.4, The Cochrane Collaboration, 2020. Publication bias and small-study effects were assessed by visual inspection of funnel plots, Egger’s regression test, and rank-correlation testing; Duval and Tweedie’s trim-and-fill method was applied as a sensitivity analysis.

RESULTS

Among the evaluable studies, the publication quality was classified as low (36%) moderate (46%), or high (18%). The association between surgery for colorectal liver metastases (CRLM) and survival outcomes is described in subsequent sections and summarized in Figures 1, 2 and 3. Outcomes were analyzed using multivariate analysis in 90% of the cases.

Figure 1  Flow diagram of included studies.

Figure 2 Overall survival with surgery vs no surgery for liver metastases (random-effects model; heterogeneity: Tau² = 0.16; I² = 96%). CI: Confidence interval.

Figure 3  Progression-free survival with surgery vs no surgery for liver metastases (random-effects model; heterogeneity: Tau² = 0. 20; I² = 67%).

Overall survival

OS analysis was available for all studies (n = 67). The pooled HR favored CRLM resection (HR = 0.38, 95%CI: 0.34-0.43; P < 0.01; Figure 2). Between-study heterogeneity was substantial [Tau² = 0.16; χ² = 1628.20, df = 66 (P < 0.00001); I² = 96%] (Tables 1 and 2).

Table 1 Characteristics of included studies.

Ref.	Type of study	Country	n2	Liver metastases/other %	CH (adj/neoadj %)	Resection %	R0 resection %	Median follow up (months)	HR for OS	Type of analysis (MVA/UVA)	NOS score
Aasebø et al[19], 2020	Prospective, cohort	Scandinavia	452	64/25 (lung)/29 (nodes)/20 peritoneum)	621	-	7	13	0.32 (0.20-0.52)	MVA	6
Afshari et al[20], 2019	Retrospective, case control	Sweden	281	88/14 (lung)	67.2/13.5	44	77.5	-	0.12 (0.06-0.24)	MVA	5
Ahmed et al[21], 2014	Retrospective, cohort	Canada	1378	100/50	42.31	14.4	-	7.1	0.54 (0.45-0.65)	MVA	6
Albertsmeier et al[22], 2017	Retrospective	Germany	456	74.8/25.2	-/23.3	49.6	36.4	26.4	0.31 (0.22-0.43)	MVA	7
Allard et al[23], 2017	Retrospective	Various	12406	100/9.8	-/40.8	100	87.62	29.4	0.35 (0.26-0.48)	MVA	7
Angelsen et al[24], 2017	Retrospective	Norway	2960	100/43.3	-	18.1	-	10.9	0.54 (0.34-0.86)	MVA	6
Badic et al[25], 2022	Retrospective, cohort	France	1115	40.5/4 (lung)/16 (peritoneum)	21.2	21.2	2	4.2	0.30 (0.24-0.38)	MVA	6
Bakalakos et al[26], 1998	Retrospective	United States	301	100	100	100	39.5	-	0.60 (0.44-0.82)	MVA	5
Beppu et al[27], 2014	Retrospective	Japan	71	100	100	37	25.3	21.8	0.19 (0.09-0.40)	MVA	7
Bhatti et al[28], 2022	Retrospective	Canada	28	100/46	100	53	-	16.5	0.62 (0.40-0.96)	MVA	7
Brouquet et al[29], 2011	Prospective	United States	65	100	100	72	-	50	0.13 (0.04-0.41)	MVA	8
Canseco et al[30], 2023	Retrospective	Taiwan	2612	100/45.5	22.8/11.9	24.4	-	60	0.15 (0.11-0.20)	MVA	8
Chen et al[31], 2018	Prospective	Taiwan	161	78.9	-/84.5	8	-	84	0.76 (0.43-1.33)	MVA	8
Chen et al[32], 2022	Retrospective	China	448	100	-	39.7	-	-	0.68 (0.53-0.87)	MVA	5
Cheng et al[33], 2022	Retrospective, cohort	China	1874	81.1/18.9	-	27.5	-	-	0.37 (0.29-0.47)	MVA	5
Cremolini et al[34], 2017	Pooled analysis of prospective trials	Italy	205	100	100	44.4	36.12	36.5	0.37 (0.22-0.62)	MVA	7
Cummings et al[35], 2007	Retrospective	United States	13599	100	-	6.1	-	-	0.33 (0.30-0.36)	UVA	5
D’Angelica et al[36], 2015	Prospective, phase 2	United States	45	100	71.1	47	40	39	0.16 (0.05-0.51)	MVA	7
Datta et al[37], 2022	Prospective	United States	128	100	100 (CH + HAI); 78.1 (CH)	40	27.3	65.5	0.24 (0.12-0.48)	MVA	8
de la Fouchardiere[38], 2019	Retrospective	France	287	51.9/25.8 (lung)/37.3 (peritoneum)	57.3/89.2 (L1)	15.3	-	47.5	0.47 (0.29-0.76)	MVA	8
Dexiang et al[39], 2012	Retrospective	China	1613	100/13.9	15.5 (CH + HAI); 8.1 (CH)	32.8	-	19	0.44 (0.33-0.59)	MVA	6
Díez-Alonso et al[40], 2021	Retrospective, cohort	Spain	104	100	-/57.7	67.9	-	-	0.44 (0.22-0.90)	MVA	5
Engstrand et al[41], 2018	Retrospective, cohort	Sweden	1026	68.8	56 (periop)	38	-	63.6	0.21 (0.13-0.33)	MVA	8
Famularo et al[42], 2023	Retrospective	Italy	847	100	69.5 (periop)	25.7	-	30	0.20 (0.08-0.48)	MVA	6
Folprecht et al[43], 2014	Retrospective	Germany, Austria	111	100	-/100	100	64.2	-	0.34 (0.14-0.83)	MVA	5
Galizia et al[44], 2013	Prospective	Italy	48	100	-/100	27.1	-	-	0.14 (0.03-0.65)	UVA	5
Ge et al[45], 2019	Retrospective, cohort	China	19796	70.6	-	-	-	-	0.48 (0.44-0.51)	MVA	5
Ghiringhelli et al[46], 2014	Retrospective	France	932	50/3.2 (lung)/13.1 (peritoneum)	58.3	-	11.2	60	0.18 (0.13-0.26)	MVA	6
Grande et al[47], 2016	Retrospective	Italy	751	41.1/10.3 (lung)/34.4 (multi)	46/78 (L1)	19	-	12	0.40 (0.30-0.53)	MVA	5
Habbous et al[48], 2023	Retrospective, cohort	Canada	1498	29 (liver and lung)/60 (lung)	-	-	49.5	34	0.80 (0.66-0.95)	MVA	7
Hackl et al[49], 2014	Retrospective	Germany	5772	24.7	26/48 (periop)	26	-	120	0.44 (0.39-0.50)	MVA	9
Huemer et al[50], 2023	Retrospective	Austria	117	68/34 (lung)/19 (peritoneum)	100	22	-	-	0.22 (0.09-0.54)	UVA	5
Imai et al[51], 2019	Prospective	Japan	163	100	47.2/55.8	100	-	38.8	0.07 (0.02-0.26)	MVA	7
Javed et al[52], 2022	Retrospective	France	105	100	-/63/87 (L1)	86	-	-	0.17 (0.08-0.34)	MVA	5
Jiao et al[53], 2023	Retrospective	China	4575	100	-	83.8	-	84	0.50 (0.39-0.64)	MVA	8
Joharatnam-Hogan et al[54], 2020	Retrospective	United Kingdom	125	44/7 (lung)/49 (multi)	18/34	-	-	54	0.47 (0.22-1.00)	UVA	8
Kataoka et al[55], 2014	Retrospective	Japan	145	100	39.1/-	52.2	-	25.2	0.25 (0.08-0.80)	MVA	6
't Lam-Boer et al[56], 2015	Retrospective	Netherlands	1617	100	62	100	-	-	0.32 (0.25-0.40)	MVA	5
Lan et al[57], 2016	Prospective	China	5239	68.2/19 (lung)/34 (peritoneum)	69/5.8	54.8	-	-	0.41 (0.32-0.51)	MVA	5
Lemini et al[58], 2019	Retrospective	United States	31172	81.3/12.2 (lung)	71.2	42.4	76.3	-	1.08 (0.95-1.23)	MVA	5
Leone et al[59], 2023	Prospective	Italy	167	100	100	23.9	-	48	0.31 (0.19-0.51)	-	7
Leporrier et al[60], 2006	Retrospective, population-based	France	358	100/18.4	100	17.3	-	26	0.79 (0.67-0.93)	MVA	6
Li et al[61], 2023	Retrospective, SEER population-based	China	2127	100	88.3	25.4	-	38	0.41 (0.23-0.73)	MVA	6
Lin et al[62], 2022	Retrospective	China	229	100	0/100	45.9	-	20	0.63 (0.33-1.21)	UVA	6
Luo et al[63], 2018	Retrospective, SEER population-based	China	20268	87.9/10 (lung)	-	11.5	-	-	0.66 (0.62-0.70)	MVA	5
Nakai et al[64], 2013	Prospective	Japan	21	100	100 (HAI)	38.1	-	-	0.08 (0.01-1.46)	MVA	5
Nogueira-Costa et al[65], 2020	Retrospective	Portugal	102	-	100	21.6	-	15	0.34 (0.15-0.77)	MVA	6
Norén et al[66], 2016	Retrospective	Sweden	3125	100	17.6	17.8	-	-	0.23 (0.19-0.27)	MVA	5
Okuno et al[67], 2020	Prospective, phase 2	Japan	34	100/23.5	100	61.7	-	72.6	0.19 (0.06-0.60)	MVA	8
Padman et al[68], 2013	Retrospective, cohort	Australia	455	100	26.6	45.7	-	16.7	0.38 (0.28-0.52)	MVA	6
Park et al[69], 2013	Retrospective, cohort	South Korea	847	100	-	16.5	-	-	0.32 (0.19-0.54)	MVA	6
Qiao et al[70], 2022	Retrospective, cohort	China	7583	68/14.9	-	34.3	-	-	0.74 (0.69-0.79)	MVA	6
Raoof et al[71], 2020	Retrospective, cohort	United States	16382	100/55.8	54.2	10	-	-	0.84 (0.79-0.89)	MVA	6
Robinson et al[72], 1999	Retrospective, cohort	United States	48	100	-	52	-	-	0.29 (0.15-0.56)	MVA	5
Rouyer et al[73], 2018	Retrospective, cohort	France	360	72.2/80	100	21	57.9	24	0.09 (0.02-0.41)	MVA	7
Rouyer et al[74], 2016	Retrospective, cohort	France	389	100/62.5	99	27.2	16.9	36	0.41 (0.19-0.88)	MVA	7
Sarkar et al[75], 2023	Retrospective, cohort	United States	72376	100/40.5	29/9.4	35.6	-	-	0.41 (0.35-0.49)	MVA	6
Shindoh et al[76], 2013	Retrospective, cohort	United States	123	100	-/61.8	70.7	-	50.2	0.33 (0.19-0.57)	UVA	8
Siebenhüner et al[77], 2020	Retrospective, cohort	Swiss	10325	92.2/20.6	-	28.1	-	19	0.73 (0.68-0.78)	MVA	6
Thomasset et al[78], 2013	Retrospective, cohort	United Kingdom	269	100	-	58	-	-	0.70 (0.53-0.92)	UVA	5
Villard et al[79], 2021	Retrospective, cohort	Sweden	100	100	-/100	31	-	-	0.54 (0.33-0.88)	MVA	5
Wang et al[80], 2010	Retrospective, cohort	China	293	100	-	49	-	-	0.40 (0.18-0.90)	MVA	5
Ye et al[81], 2021	Retrospective, cohort	China	93	76.3/23.7	-	34.4	-	26.7	0.46 (0.22-0.96)	MVA	5
Yi et al[82], 2020	Retrospective, cohort	China	27878	72.2/30.6	-	14.5	-	-	0.75 (0.72-0.78)	MVA	6
Zaydfudim et al[83], 2015	Retrospective, cohort	United States	31574	100	2.5/1.1	6.9	-	-	0.40 (0.38-0.42)	MVA	6
Zhang et al[84], 2020	Retrospective, National Cancer Database	United States	58171	70/30	-	8.7	44.6	60	0.42 (0.38-0.46)	MVA	8

¹First-line.

²R0-R1.

CH: Chemotherapy; HAI: Hepatic arterial infusion; HR: Hazard ratio; OS: Overall survival; MVA: Multivariate analysis; UVA: Univariate analysis; L1: First chemotherapy line; periop: Perioperative; multi: Multiple sites; NOS: Newcastle-Ottawa Scale.

Table 2 Summary characteristics of included studies.

Characteristic	Value
Included studies, n	67
Total patients, n	368380
Study design	Retrospective 55 (82%); prospective 12 (18%)
Publication years	1998-2023
Sample size (per study)	Median 448 (IQR 124-2786)
Geographic region (approximately)	Asia 22; non-Asia 44; multi-region 1
Median follow-up	37 months (not reported in 42% of studies)
Systemic therapy reporting	Not available in 17 studies

IQR: Interquartile range.

Progression-free survival

PFS analysis was performed in nine studies. The pooled HR favored CRLM resection (HR = 0.46, 95%CI: 0.31-0.66; P < 0.01; Figure 3). Heterogeneity was moderate [Tau² = 0.20; χ² = 24.26, df = 8 (P = 0.002); I² = 67%].

Sensitivity analysis and publication bias

Survival was better in prospective than in retrospective studies (HR = 0.27, 95%CI: 0.19-0.38; P < 0.01 vs HR = 0.4, 95%CI: 0.35-0.45; P < 0.01, respectively). It was similar in the last decade (HR = 0.39, 95%CI: 0.34-0.44; P < 0.01) and in older series (HR = 0.37, 95%CI: 0.3-0.46; P < 0.01); in Asian series (HR = 0.38, 95%CI: 0.32-0.46; P < 0.01) and Western countries (HR = 0.38, 95%CI: 0.33-0.45; P < 0.01). As expected, the effect size was larger in smaller studies (P for meta-regression = 0.01) but not linked to the median follow-up duration (P = 0.42). Moderate-high quality papers provided similar results to poor-quality studies (HR = 0.41, 95%CI: 0.36-0.48; P < 0.01; HR = 0.41, 95%CI: 0.37-0.46; P < 0.01).

Funnel plot inspection suggested asymmetry. While rank-correlation testing was not statistically significant (P = 0.27), Egger’s regression indicated small-study effects (P < 0.01). Duval and Tweedie’s trim-and-fill method imputed 18 potentially missing studies, attenuating the pooled OS estimate to HR = 0.48 (95%CI: 0.43-0.53). Therefore, publication bias/small-study effects cannot be excluded, although the association remained in favor of surgery after adjustment.

DISCUSSION

CRLM remain a major driver of mortality in metastatic CRC. In this systematic review and meta-analysis of 67 comparative studies including 368380 patients, liver resection (with or without systemic therapy) was associated with longer OS and PFS compared with non-surgical management[6,7].

Our investigation offers an extensive quantitative analysis of the surgical advantages of CRC hepatic metastases over a 25-year period (1998-2023) with an accentuated focus in the last decade. These studies predominantly encompassed patients from the 2000 to 2010 era, a phase prior to the prevalent adoption of computed tomography (CT) scans and alternative ablative methodologies. Historically, randomized trials exclusively assessing surgical or other localized treatments for CRC liver metastases have been lacking, which is attributed to the multitude of efficacious interventions that render such studies ethically questionable. Meta-analyses of rigorously designed observational endeavors, such as cohort and case-control analyses, substantiate treatment efficacy, albeit with less potent evidence than randomized controlled trials[8].

We found that resection of metastases, with or without perioperative or neoadjuvant CT, may extend survival in patients with stage IV CRC, reducing the risk of death by 62% and disease progression by a similar magnitude (54%) in a series of 67 observational studies (retrospective or prospective cohorts) including more than 300000 patients. The results were derived mainly from multivariate analysis and seemed to be independent of other known prognostic factors.

The integration of systemic CT with surgical measures has garnered significant attention and recently enhanced outcomes. Advances in both systemic CT and surgical modalities have increased the survival rate of patients with operable hepatic metastases of CRC. Optimal CT scheduling, whether preoperative or postoperative, incites an ongoing debate, each bearing inherent merits and drawbacks. Considerations of CT-induced hepatic toxicity, tumor progression risks, and emergent resistant strains require meticulous deliberation. Currently, perioperative CT epitomizes the standard regimen for treatable conditions, whereas conversion therapy, deploying multi-agent CT, is hailed as the pinnacle for potentially operable metastases. Conversely, for patients with inoperable CRC liver metastases, alternative modalities like stereotactic body radiotherapy have demonstrated impressive local control and survival rates, coupled with moderate adverse effects, underlining the importance of advancing and refining non-surgical alternatives[9-11]. Discussions now increasingly consider tumor biology, potential extrinsic hepatic metastasis locations, and primary tumor laterality in evaluating resection advantages. Despite potentially diminished benefits, excluding patients from surgical options based on unfavorable prognoses is unwarranted[12-14].

Considering the overarching scenario, surgeons are inclined to recommend hepatic metastasis resection and closely collaborate with oncologists to determine the optimal timing of surgery, either before or after CT, considering the critical role of systemic therapy in these patients. Once surgical intervention is agreed upon, the preferred approach, first addressing the primary tumor, followed by metastases or simultaneous treatments-is outlined. Generally, prioritizing the primary tumor is reserved for cases in which its management is crucial for initiating systemic treatments, such as hemorrhagic conditions. However, whenever possible, simultaneous surgeries for both primary and metastatic sites are recommended to avoid multiple operations, especially in patients who have undergone preoperative CT. It is essential to recognize that medical interventions aim not only to reduce the disease burden, facilitating the achievement of surgical R0 (complete removal with no residual tumor), but also to identify patients unresponsive to medical therapies. This lack of response is a prognostic adverse factor that essentially excludes surgery[15].

Investigated scores for identifying patients potentially benefitting from surgical intervention in metastatic contexts have incorporated molecular traits of the disease. However, outside the research framework, these scores do not dictate the curative treatment decisions. They represent merely one facet within the multifaceted clinical-pathological and biological intricacies characterizing these cancers[16]. Managing CRC liver metastases necessitates a multidisciplinary strategy, where expanding surgical eligibility, comprehensive evaluations, systemic therapy integration, and novel perioperative techniques collectively enhance patient prognoses. Ongoing research is vital to refine treatment protocols and to elucidate the long-term effects of these interventions on patient survival and quality of life.

Importantly, the evidence base is dominated by retrospective observational cohorts, and candidacy for hepatic surgery is inherently selective (e.g., resectable disease, limited tumor burden, adequate performance status, favorable biology and/or response to systemic therapy, and access to experienced hepatobiliary centers). Although most studies reported multivariable-adjusted estimates, confounding by indication and residual unmeasured confounding remain unavoidable; therefore, the pooled hazard ratios should be interpreted as associations in selected patients rather than causal effects comparable with randomized evidence.

Between-study heterogeneity was substantial for OS (I² = 96%) and moderate for PFS (I² = 67%), reflecting variability in metastatic extent (liver-only vs extrahepatic disease), timing (synchronous vs metachronous), treatment intent (curative vs palliative), systemic regimens, and treatment era. While subgroup and meta-regression analyses suggested a consistent direction of association across several study-level characteristics, the magnitude of benefit likely differs across clinically distinct subgroups and settings; consequently, pooled estimates should be applied cautiously to individual patients.

Systemic therapy remains integral to modern CRLM management, both to select patients with favorable tumor biology and to maximize the probability of complete (R0) resection. In the revised manuscript, terminology for treatment timing was standardized as follows: Neoadjuvant (preoperative therapy in resectable disease), perioperative (planned pre- and postoperative therapy), and conversion (therapy aimed at downsizing initially unresectable disease to enable resection).

Evidence for PFS was limited to nine studies and definitions and assessment schedules were inconsistently reported. Therefore, conclusions regarding disease control are less robust than OS findings and are presented cautiously.

Assessment of small-study effects suggested possible publication bias: The funnel plot was asymmetric and Egger’s regression was statistically significant (P < 0.01), whereas rank-correlation testing was not (P = 0.27). Trim-and-fill imputed 18 potentially missing studies and attenuated the pooled OS estimate to HR ≈ 0.48, indicating that the survival association persists but may be smaller than the primary pooled estimate.

An additional limitation is the limited availability of key prognostic and predictive factors across included studies (e.g., RAS/BRAF mutation status, primary tumor sidedness, number and size of liver metastases, resection margin status, and the presence and extent of extrahepatic disease), which precluded biologically informed subgroup analyses. Future registries and prospective cohorts should report these variables in a standardized manner to clarify which patient subgroups derive the greatest benefit from surgery.

Clinically, hepatic resection should be considered within a multidisciplinary framework that integrates systemic therapy, modern cross-sectional imaging, and locoregional options. For patients who are not candidates for resection, non-surgical modalities (e.g., stereotactic body radiotherapy and other ablative techniques) may offer local control and should be evaluated in comparative prospective studies.

Strengths of this study include the comprehensive search, large sample size, and consistent use of time-to-event effect measures. Limitations include the observational design of most included studies, substantial heterogeneity, possible small-study effects, and incomplete reporting of systemic therapy details and follow-up in a proportion of studies.

Surgery for liver metastases of CRC represents a crucial component of a multidisciplinary approach to metastatic CRC, offering a potential cure for an increasing number of patients. Advances in surgical techniques and perioperative care have significantly increased eligibility for liver surgery in patients with CRC. Surgical resection is now possible in patients with multiple metastases and in those previously considered inoperable due to improvements in preoperative imaging, minimally invasive surgical techniques, and enhanced recovery protocols. The integration of CT with surgical resection has further improved the outcomes of patients with liver metastases from CRC. Despite these advancements, the management of liver metastases from CRC remains challenging and a significant proportion of patients experience recurrence.

An essential avenue for future research is to explore personalized treatment modalities that can be adapted based on the patient’s genetic profile, tumor biology, and response to previous treatments. Given the heterogeneity of CRC, understanding the molecular pathways that drive liver metastasis could lead to the development of targeted therapies that improve resectability rates and postsurgical outcomes. Furthermore, advancements in liquid biopsy technology could offer noninvasive methods to monitor disease progression and response to treatment, potentially guiding the timing of surgery and the use of adjuvant therapies[17]. The evolution of surgical strategies for colorectal liver metastases, including the role of minimally invasive techniques and the integration of robotic surgery, represents a critical frontier in improving patient outcomes[18]. These techniques offer the potential for reduced postoperative morbidity, shorter hospital stays, and quicker return to systemic therapies. The development of intraoperative imaging and augmented reality technologies could further enhance the precision of metastasectomy, allowing for more extensive resection while preserving the healthy liver tissue. Moreover, the concept of oligometastatic CRC, in which patients have a limited number of metastatic sites, challenges the traditional dichotomy between operable and inoperable diseases. Investigating the outcomes of aggressive surgical approaches in this subgroup could redefine treatment paradigms, offering hope for curative outcomes in patients previously deemed unsuitable for surgery.

CONCLUSION

In summary, across predominantly observational evidence, surgical management of CRLM is associated with improved survival compared with non-surgical approaches. These findings support timely referral to experienced multidisciplinary teams to optimize patient selection and treatment sequencing. Prospective studies incorporating contemporary systemic regimens and tumor biology are needed to better estimate the magnitude of benefit and to identify patients unlikely to benefit from resection.