A systematic review was conducted by searching databases including MEDLINE (Pubmed), Embase, Cumulative Index to Nursing and Allied Health Literature (CINAHL) (EBSCO host), and WHO Global Index Medicus, through March 18th, 2022. The search algorithm used was: “(‘BRAF’ OR ‘RAS’ OR ‘HER2’ ‘SMAD4’ OR ‘TP53’ OR ‘P53’ OR ‘APC’ OR ‘PI3K’ OR ‘EGFR’ OR ‘MACC1’) and (‘colon’ or ‘colorectal’ or ‘rectal’ or ‘rectum’) and (‘metastasis’ or ‘metastatic’ or ‘metastases’ or ‘mets’ or ‘metastasectomy’) and (‘hepatic’ or ‘liver’)”. Titles and abstracts of identified records were screened for relevance. Since MEDLINE included all the articles returned by the other two databases, we referred only to MEDLINE in Fig. 1.

Eligible studies included those detailing tumor mutation status and oncological outcomes in CRC patients who underwent liver metastasectomy. The inclusion criteria comprised studies performed in metastatic CRC patients undergoing liver metastases resection; reporting genetic test results for RAS, BRAF, SMAD4, or PIK3CA mutation status from resected colorectal liver metastasis (CRLM) specimens; providing outcomes of RFS and/or OS based on mutational status; and offering RFS and OS hazard ratios (HR) from multivariate analyses.

Exclusion criteria included reviews, meta-analyses, editorials, letters to the editor, case reports, studies on primary cancers other than colorectal, studies with primary metastatic sites other than liver, studies lacking specific oncological outcome variables or genetic data, non-English language studies, and studies omitting residual disease data from liver metastasectomy.

Three subgroups of eligibility were identified: “group 1” included studies with no or microscopic post-metastasectomy residual disease (R0 and R1) and without or resected extrahepatic disease; “group 2” encompassed studies with macroscopic post-metastasectomy residual disease (R2) and without or resected extrahepatic disease; “group 3” comprised studies where the status of extrahepatic disease was unspecified (either resected or not).

A predefined protocol was employed for data extraction. Extracted variables included: first author’s name, publication year, journal name, enrollment years, country of enrollment, sample size, study design, demographic profile, genes analyzed, genotyping method, tumor mutational status, chemotherapy scheme and schedule, percentage of patients with residual disease post-metastasectomy, percentage with extra-hepatic disease and oncological outcomes (OS, overall response rate (ORR), disease-free-survival (DFS), or RFS) + including main findings and 95% confidence interval (CIs). Four authors (MF, FDG, CLP, EC) independently verified inclusion criteria and executed data extraction. Adjusted estimates from original studies for relevant confounding factors were utilized.

The standard error of the log HR was deduced from the log CIs. Pooled HR and corresponding 95% CI were calculated using the DerSimonian and Laird random-effects models, accounting for within-study and between-study variabilities. In cases of a low number of studies, pooled HR was also estimated via the Hartung-Knapp-Sidik-Jonkman (HKSJ) method for a conservative approach. Statistical heterogeneity among studies was assessed using the I² and Q statistics. Influence analysis was conducted, recalculating pooled HR by sequentially omitting each study. Publication bias was evaluated through funnel plot.

Meta-analysis results were graphically presented, depicting HRs as black squares (size inversely proportional to standard error) and 95% CIs. In some studies, 95% CIs were slightly modified from those in original papers due to estimated study variances. Pooled HRs for all studies and various study designs were represented by diamonds, indicating the HR at the center and 95% CIs at the extremes. Statistical significance was set at p < 0.05 (two-sided).

Out of 47 studies eligible for the meta-analysis, 43 reported on the association between RAS mutation status and patients’ OS and RFS and included a total of 24,121 patients, of whom 8,258 had RAS mutations. The pooled KRAS mutation rate was 34.2%, which is consistent with previous reports in the literature of resected CRC LM. Specifically, 36 studies reported the association with OS and 20 studies with RFS.

The meta-analysis of OS included 36 studies with 15,766 patients, of whom 5,361 had RAS mutations.

The cumulative HR for OS was 1.68 (95% CI, 1.54–1.84) (Fig. 2A), indicating that RAS mutation is an independent prognostic factor associated with poorer OS in patients with metastatic CRC undergoing liver resection.

Regarding RFS, 20 studies involving 8,355 patients were analyzed, with 2,897 patients having RAS mutations. The cumulative HR was 1.46 (95% CI, 1.33–1.61) (Fig. 2B) demonstrating that RAS mutations are associated with a higher risk of relapse and a negative impact on prognosis compared to patients with wild-type genes.

An analysis of the forest plot illustrating the effect of RAS mutations on RFS in Fig. 2B reveals a notable increase in result homogeneity across different studies beginning in 2019. This uniformity is likely attributable to standardized methodologies in patient enrollment and evaluation across studies, enhancing the reliability of meta-analytic findings.

Of the 36 studies included in the meta-analysis on OS, 21 specifically investigated the association between KRAS mutation alone and patient OS. The cumulative HR for these studies was 1.63 (95% CI, 1.42–1.86) (Fig. 3A). In contrast, the remaining 15 studies that examined the association between all-RAS mutations and patients’ OS, reported a cumulative HR of 1.74 (95% CI, 1.56–1.94) (Fig. 3A). The pooled analysis indicates that the impact on OS is similar whether studies considered only KRAS mutations or all RAS mutations.

For RFS, the findings were similarly consistent. Twelve studies that focused on KRAS mutations alone reported a cumulative HR of 1.36 (95% CI, 1.24–1.49), while eight studies that evaluated the association between all-RAS mutations and RFS presented a cumulative HR of 1.54 (95% CI, 1.30–1.82) (Fig. 3B). This comparison suggests that the prognostic impact of KRAS mutations is comparable to that of all-RAS mutations on RFS.

Among the 47 studies included in our meta-analysis, 13 reported on OS after CRC LM resection, stratified by BRAF mutation status, while 8 studies reported on RFS. The studies included data on 3- or 5-years OS or both, and RFS; 3- or 5-years DFS data were also considered as part of RFS. A total of 8,969 patients were analyzed for BRAF-related OS and RFS outcomes, with 433 patients harboring BRAF mutations. The pooled BRAF mutation rate was 4.83%, consistent with previous literature on resected CRC LM.

For the OS, the analysis of 13 studies included 5,831 patients, 270 of whom had BRAF mutations. The cumulative HR was 2.62 (95% CI, 2.14–3.20) (Fig. 4A), indicating that the BRAF mutation is an independent prognostic factor negatively associated with OS in metastatic CRC patients undergoing complete liver resection.

In terms of RFS, 8 studies with 3,138 patients were analyzed, 163 of whom had BRAF mutations. The cumulative HR was 1.89 (95% CI, 1.32–2.73) (Fig. 4B) demonstrating that BRAF mutations are associated with a worse prognosis and a higher risk of relapse compared to wild-type patients.

Only 6 studies provided data on OS and/or RFS for CRC LM patients with SMAD4 mutations. These studies, which focused on SMAD4 3-year OS or RFS or DFS, included a total of 3,020 patients in the analysis for both SMAD4-related OS and RFS, with 347 harboring a SMAD4 mutation. Of these, 222 patients were analyzed for OS, and 125 for RFS. The cumulative HR for OS was 1.93 (95% CI, 1.56–2.38) (Fig. 5A), and for RFS was 1.95 (95% CI, 1.31–2.91) (Fig. 5B), confirming SMAD4 mutations as a detrimental prognostic factor in CRC patients undergoing liver metastasectomy.

For PIK3CA mutations, only 3 studies were deemed eligible for the evaluation of OS and RFS: two addressing RFS and one addressing OS. No significant association with either OS or RFS was found.

No significant differences were observed in the analyses of OS and RFS for studies evaluating RAS, BRAF, and SMAD4 mutations, even when data were subdivided according to the subgroup of eligibility (Suppl. Figs. 1–3).

Microsatellites, repeated sequences of 1 to 6 base pairs in length composed mostly of non-coding DNA [66,67], are primarily located at chromosome ends and contribute to the individual genetic fingerprint [68].

Microsatellite instability (MSI) results from accumulated errors at microsatellite sites due to a deficient DNA mismatch repair (dMMR) system. This deficiency may arise sporadically, such as through CpG island methylator phenotype (CIMP) due to oxidative stress in MLH1 promoter hypermethylation, or through inherited conditions like Lynch syndrome or Muir-Torre syndrome [66,69–73]. The MMR system, involving four main genes and their encoded proteins, plays a critical role in DNA replication accuracy by correcting erroneous insertions, deletions, or mis-incorporated bases, thus preventing mismatches between DNA strands [71–74]. Abnormal MMR function leads to frameshift mutations, contributing to a high tumor mutational burden (TMB) and an increase in tumor infiltrating lymphocytes (TILs) [71,74–76].

These factors are pivotal for the effectiveness of immune checkpoint inhibitors in treating unresectable or metastatic tumors with such pathogenesis [77–83]. The meticulous characterization of immune infiltration has become essential for the development of the Immunoscore, a prognostic tool that enhances the TNM classification for high microsatellite instability (MSI-H) CRC [84–87]. Ongoing trials aim to refine survival predictions by introducing new biomarkers [84,88,89].

MSI is associated with various neoplasms, particularly CRC, where it is found in up to 15% of cases, but only 5% in metastatic settings [71,90–93].

Sporadic MSI-H tumors, often identified in older women and associated with the right colon, exhibit distinct features such as poorly differentiated mucinous histology and pronounced lymphocytic infiltration, and a Crohn’s-like host response [66,90,94]. While MSI is a positive prognostic factor in early-stage CRC, its benefit does not extend to metastatic disease, where outcomes are similar to microsatellite stable (MSS) tumors [91,95–99] and dMMR. Recent studies have highlighted the complex role of MSI in predicting intrahepatic recurrences post-liver resection [100], underscoring the need for further research into its prognostic significance alongside other genes. This is in contrast with a previous paper by Haddad et al., which challenged the prognostic significance of MSI, suggesting that the survival of patients with stage IV MSI-positive tumor was not influenced by the surgical resectability of liver metastases [99]. This divergence highlights the complex nature of MSI’s role in CRC and underscores the necessity for more in-depth analysis. Specifically, the relationship between MSI and other commonly examined genes warrants further investigation to clarify its prognostic implications in the context of CRC liver metastases [91,101–103].

The prognostic and predictive value of somatic genetic mutations is under investigation for selecting CRC patients with liver-only metastases for liver transplantation as a curative strategy. Despite the survival benefits observed with surgical treatment, only a minority qualify for resection, and recurrence remains a significant concern. The concept of completely removing the affected liver and performing transplantation emerged in the early eighties but faced challenges due to the high perioperative mortality rate and a significant disease recurrence rate within the first year. However, advancements in the surgical management of CRC, alongside the introduction of fluoropyrimidines-based combination chemotherapy with irinotecan and oxaliplatin (FOLFOX and FOLFIRI achieving response rates around 50%, and the triplet regimen FOLFOXIRI reaching up to 60%) as well as the advent of targeted therapies, have reopened the consideration of liver transplantation as a viable treatment option for patients with metastatic CRC [104].

A landmark in this evolving field was the Secondary Cancer (SECA) I trial, reported in 2013. This trial presented significant survival benefits in a small cohort of stage IV CRC patients with liver-only metastases, documenting 1-, 3-, and 5-year OS rates 95%, 68%, and 60%, respectively, for patients with unresectable liver-only metastases undergoing liver transplantation. The promising results have catalyzed a series of prospective studies aimed at evaluating the feasibility and clinical impact of liver transplantation in patients with surgically inoperable LM from CRC [105].

Particular emphasis has been placed on the molecular criteria for selecting patients for these studies, with the goal of excluding those whose tumors possess characteristics indicating a poorer prognosis. As a result, most ongoing trials require patients to have tumors that are wild-type for both RAS and BRAF genes.

The results of the Italian COLT trial were recently published [106]. It demonstrated that liver transplantation in metastatic CRC, coupled with improved patients selection strategies including tumor molecular characterization for RAS/RAF can give patients a 5-year OS similar to other indications for liver transplantation and a better outcome than those undergoing chemotherapy alone [106].