Effect of unplanned conversion to open surgery on resection margins and postoperative complications in minimally-invasive resection of colorectal liver metastases: A systematic review and meta-analysis with meta-regression

Abstract

Backgrounds/Aims

To evaluate the impact of unplanned conversion to open surgery on resection margin status and postoperative complications in patients undergoing minimally-invasive resection of colorectal liver metastases (CRLM).

Methods

This study performed a proportion meta-analysis and meta-regression using random-effects modelling in accordance with PRISMA guidelines. Studies with at least 15 patients that reported conversion to open surgery in individuals receiving minimally-invasive CRLM resection were included. The association of unplanned conversion with postoperative outcomes was analyzed.

Results

Eighty-six studies encompassing 18,138 patients were analyzed. The overall conversion rate was 5.8% (95% CI 5%–6.6%). Conversion was associated with improved R0 resection rates (coefficient: 2.167, p < 0.001) but was also linked to increased postoperative mortality (coefficient: 7.585, p = 0.001) and morbidity (coefficient: 1.737, p = 0.003); there was no significant impact on 5-year overall survival (coefficient: 0.700, p = 0.989) or 5-year disease-free survival (coefficient: –72.900, p = 0.157). Specifically, conversion due to oncological concern was associated with higher rates of R0 resection (coefficient: 0.638, p = 0.005); conversion resulting from iatrogenic injuries was associated with lower R0 resection rates (coefficient: –1.478, p < 0.001); conversion for technical difficulties was associated with lower postoperative morbidity (coefficient: –0.380, p = 0.006).

Conclusions

Unplanned conversion to open may carry prognostic and oncological implications for minimally-invasive resection of CRLM. Although conversion due to bleeding and iatrogenic injury is routinely considered, conversion prompted by technical difficulties or oncological concerns should not be considered failure, as it may be associated with improved patient outcomes.

INTRODUCTION

Advances in diagnostic modalities, the introduction of modern chemotherapy and biological agents, improvements in perioperative care, and progress in operative techniques have expanded the boundaries of surgical resection for colorectal liver metastasis (CRLM) [1]. As a result, the indications for minimally-invasive approaches to CRLM resection are concurrently broadening [2]. Evidence indicates that minimally-invasive techniques do not compromise oncological outcomes and may lower postoperative morbidity [3]. Securing negative resection margins during CRLM resection has a well-recognized effect on oncological and survival outcomes [4], and continues to be the most significant prognostic factor influenced by surgical technique and quality [5]. Therefore, current research is focused on identifying risk factors for positive resection margins in minimally-invasive CRLM resections [6].

During minimally-invasive resection of CRLM, there may be a need for unexpected conversion to open surgery due to bleeding, technical challenges, oncological considerations, or iatrogenic injuries [7]. Reported conversion rates to open surgery range from 2% to 13% [8]. The prognostic implications of unplanned conversion to open surgery during minimally-invasive CRLM resection, particularly its effects on postoperative morbidity, mortality, resection margins, and survival, are not yet thoroughly established. Therefore, we undertook a comprehensive meta-analysis with meta-regression to assess the prognostic impact of unplanned conversion to open surgery during minimally-invasive CRLM resections, focusing on the reasons for conversion.

MATERIALS AND METHODS

Methodological and reporting compliance

The methodology for this study was guided by the Cochrane Handbook for Systematic Reviews (version 6.4) [9], and the reporting was in accordance with the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) 2020 statement standards [10]. A predefined protocol was registered on PROSPERO (registration number: CRD42025649612), an open-access international database for prospectively registered systematic reviews.

Eligibility criteria

Study design

Eligible studies included all randomized controlled trials, cohort studies, and case series with a minimum enrollment of 15 patients. Reviews, meta-analyses, scoping reviews, case reports, and correspondence articles were excluded.

Population

All patients aged 18 years or older undergoing minimally-invasive resection of CRLM were eligible. The minimally-invasive modalities of interest included laparoscopic or robotic approaches. We included patients who underwent combined liver and colorectal resections, staged liver resection, or redo surgery for CRLM. Patients managed solely by ablation techniques without liver resection were excluded.

Prognostic factor

Unplanned conversion to open surgery for any reason was considered the prognostic factor of interest. The reasons for conversion were categorized as bleeding, technical difficulties, oncological concern, and iatrogenic injuries. The conversion rationale was ascertained based on intraoperative indications as described in the included studies.

Outcomes

Attainment of R0 resection was considered the primary outcome measure. Postoperative mortality (death due to any cause within 30 days of operation), postoperative morbidity (any complications within 30 days of operation), 5-year overall survival (OS), and 5-year disease-free survival (DFS) were defined as secondary outcomes.

A study was included as eligible only if, according to the above criteria, it reported the risk of conversion to open in at least 15 adult patients who underwent minimally-invasive resection of CRLM.

Information sources and search strategy

Two independent authors developed a search strategy using search keywords, limits, thesaurus headings, and operators, which was implemented in the following electronic databases: MEDLINE, Scopus, CENTRAL, the ISRCTN registry, the ICTRP registry, and ClinicalTrials.gov (Supplementary Table 1). No language restrictions were applied, and the final search was conducted on 05 January 2025. Reference lists of all included studies were additionally reviewed to identify further eligible studies.

Study selection, data collection, and data items

Titles and abstracts of the retrieved articles were independently reviewed by two authors, and eligible studies were selected after complete text assessment. If the first two authors disagreed on study selection, a third author was consulted to resolve conflicts. At the protocol development stage, authors defined the data items (bibliometric parameters, study design, baseline characteristics of the included population, follow-up duration, laparoscopic or robotic approach, simultaneous or staged liver resection, conversion to open, reasons for conversion, R0 resection, postoperative mortality and morbidity), which were finalized after pilot testing using randomly chosen studies. Data extraction was conducted by two independent authors.

Study risk of bias and evidence certainty assessment

Risk of bias was appraised for the included studies by two authors utilizing the Quality In Prognosis Studies (QUIPS) tool [11]. Certainty of evidence was assessed with the Grading of Recommendations Assessment, Development and Evaluation (GRADE) system [12]. In cases of discordance between the assessments of the initial reviewers, a third author provided an opinion. A third author was involved if disagreements persisted.

Effect measures and synthesis methods

OpenMeta [Analyst] software and Comprehensive Meta-Analysis Version 2.0 software were employed for proportion meta-analysis and meta-regression, respectively. The DerSimonian-Laird random-effects method (confidence level: 95%; correction factor: 0.5) was implemented in the proportion meta-analysis model to estimate the weighted pooled risks of conversion to open surgery, R0 resection, postoperative mortality, and postoperative morbidity. Meta-regression was conducted to evaluate the impact of conversion and its reasons on R0 resection, postoperative mortality, and postoperative morbidity. Individual patient data served as the unit of analysis, and intention-to-treat data were applied in all analyses. Statistical heterogeneity was assessed as I², utilizing Cochran’s Q test (χ²), with heterogeneity categorized as low for I² values of 0%–25%, moderate for 25%–75%, and high for 75%–100%. Sensitivity analyses included separate evaluation of studies with low overall risk of bias and leave-one-out analysis. Subgroup analyses were conducted based on patients who underwent laparoscopic resection, robotic resection, simultaneous liver and colorectal resection, and staged liver resection.

Deviation from the registered protocol

There was no deviation from the registered protocol.

Ethics consideration

Patient consent and approval from Research Ethics Committees were not required in this study because the study design was a meta-analysis which did not have direct involvement of patients and used non-identifiable data.

RESULTS

Study selection and study characteristics

The search generated 1,063 articles. Screening of the titles and abstracts led to the direct exclusion of 919 articles. Full-text assessment of the remaining 144 articles resulted in the exclusion of 58 articles (18 studies had sample size <15 patients; 32 studies did not report conversion rate; four studies included overlapping populations with other eligible studies; four studies included non-CRLM cases). As a result, 86 studies encompassing 18,138 patients were included. The PRISMA flow diagram for study selection is presented in Supplementary Fig. 1. Baseline characteristics and reference lists for the included studies are detailed in Supplementary Table 2 and Supplementary Table 3, respectively.

Risk of bias in studies

Utilizing the QUIPS tool, the risk of bias for study participation was low in all studies; risk of bias due to study attrition was low in all studies; the risk in prognostic factor measurement was low across all studies; risk in outcome measurement was consistently low; bias due to confounding was low for 63 studies and unclear for 23 studies; risk related to statistical analysis was also low in all studies (Supplementary Table 4).

Unplanned conversion to open

Analysis of 18,138 patients from 86 studies indicated that unplanned conversion to open surgery occurred in 5.8% (95% CI 5.0%–6.6%; 1,269/18,138) (Fig. 1). Statistical heterogeneity was substantial (I² = 80%), and the GRADE certainty of evidence was moderate.

Fig. 1.

Forest plots for proportion meta-analysis of: (A) conversion to open; (B) conversion for bleeding; (C) conversion due to technical difficulties; (D) conversion related to oncological concern; (E) conversion for iatrogenic injuries. CI, confidence interval; Ev/Trt, events/total.

Reasons for conversion to open

Analysis of 4,577 patients across 55 studies revealed that the primary reason for conversion to open surgery was bleeding in 41.7% of cases (95% CI 28.3%–55.2%; 131/305), followed by oncological concerns in 25.4% (95% CI 15.0%–35.8%; 67/305), technical difficulties in 31.1% (95% CI 22.6%–39.5%; 87/305), and iatrogenic injuries in 4.5% (95% CI 2.4%–6.6%; 8/305) (Fig. 1). The overall GRADE certainty of evidence was assessed as moderate.

Outcomes

R0 resection

An evaluation of 12,771 patients from 71 studies indicated R0 resection was achieved in 87.3% (95% CI 84.8%–89.9%; 10,722/12,771) (Fig. 2). Considerable statistical heterogeneity was observed (I² = 97%), and the GRADE certainty was low. The Egger’s regression test identified potential publication bias (p = 0.001).

Fig. 2.

Forest plots for proportion meta-analysis of: (A) R0 resection; (B) postoperative mortality; (C) postoperative morbidity. CI, confidence interval; Ev/Trt, events/total.

Postoperative mortality

A pooled analysis of 14,192 patients from 81 studies demonstrated a postoperative mortality risk of 0.4% (95% CI 0.3%–0.5%; 90/14,192) (Fig. 2). Statistical heterogeneity was minimal (I² = 0%) and the GRADE certainty was low. The Egger’s regression test indicated a possibility of publication bias (p = 0.016).

Postoperative morbidity

Data from 14,551 patients included in 82 studies indicated a postoperative morbidity risk of 26.4% (95% CI 21.5%–31.2%; 3,399/14,551) (Fig. 2). Statistical heterogeneity was substantial (I² = 98%), while the GRADE certainty was moderate. The Egger’s regression test suggested that publication bias was unlikely (p = 0.739).

Survival outcomes

The analysis of 3,165 patients from 17 studies found the mean OS time to be 46.7 months (95% CI 39.3–54.1). Among 7,836 patients from 37 studies, the 5-year OS probability was 55.2% (95% CI 51.4%–59.0%). For 4,641 patients from 29 studies, the 5-year DFS probability was 33.1% (95% CI 28.9%–37.2%).

Unplanned conversion to open and R0 resection

Meta-regression analysis demonstrated that conversion to open was correlated with increased likelihood of R0 resection (coefficient: 2.167, p < 0.001) (Fig. 3). Subgroup analyses by conversion cause revealed that conversion due to oncological concern was correlated with higher rates of R0 resection (coefficient: 0.638, p = 0.005), while conversion due to iatrogenic injuries was correlated with lower rates of R0 resection (coefficient: –1.478, p < 0.001). R0 resection achievement was not significantly impacted by conversion resulting from bleeding (coefficient: 0.157, p = 0.375) or technical difficulties (coefficient: 0.161, p = 0.437) (Table 1).

Fig. 3.

Meta-regression analysis results: (A) conversion to open versus R0 resection; (B) conversion to open versus postoperative mortality; (C) conversion to open versus postoperative morbidity.

Table 1.

Results of meta-regression analysis evaluating the impact of conversion to open surgery on outcomes

Independent variable	Dependent variable
	R0 resection			Mortality			Morbidity			5-year overall survival			5-year disease-free survival
	Coefficient	p-value		Coefficient	p-value		Coefficient	p-value		Coefficient	p-value		Coefficient	p-value
Conversion to open	2.167	< 0.001		7.585	0.001		1.737	0.003		0.700	0.989		–72.900	0.157
Conversion due to bleeding	0.157	0.375		0.129	0.749		0.173	0.125		Not estimated			Not estimated
Conversion due to oncological concern	0.638	0.005		–0.238	0.687		0.096	0.507		Not estimated			Not estimated
Conversion due to technical difficulties	0.161	0.437		–0.238	0.642		–0.380	0.006		Not estimated			Not estimated
Conversion due to iatrogenic injuries	–1.478	< 0.001		–0.878	0.484		–0.145	0.577		Not estimated			Not estimated

Conversion rate in each study was used as independent variable.

Unplanned conversion to open and postoperative mortality

Meta-regression demonstrated that conversion to open was correlated with increased postoperative mortality (coefficient: 7.585, p = 0.001) (Fig. 3). Subgroup analysis by reason for conversion indicated that postoperative mortality was not influenced by conversion due to bleeding (coefficient: 0.129, p = 0.749), oncological concerns (coefficient: –0.238, p = 0.687), technical difficulties (coefficient: –0.238, p = 0.642), or iatrogenic injuries (coefficient: –0.878, p = 0.484) (Table 1).

Unplanned conversion to open and postoperative morbidity

Meta-regression indicated that conversion to open was correlated with increased postoperative morbidity (coefficient: 1.737, p = 0.003) (Fig. 3). Subgroup analysis by conversion reason revealed that conversion attributed to technical difficulties was correlated with reduced postoperative morbidity (coefficient: –0.380, p = 0.006); nonetheless, conversion for bleeding (coefficient: 0.173, p = 0.125), oncological concern (coefficient: 0.096, p = 0.507), or iatrogenic injuries (coefficient: –0.145, p = 0.577) did not impact postoperative morbidity (Table 1).

Unplanned conversion to open and survival outcomes

Conversion to open did not influence 5-year OS (coefficient: 0.700, p = 0.989) or 5-year DFS (coefficient: –72.900, p = 0.157). The available data were insufficient to permit regression analysis by specific reasons for conversion (Table 1).

Subgroup analysis

Table 2 presents the outcomes of subgroup analyses for patients undergoing laparoscopic resection, robotic resection, simultaneous liver and colorectal resection, and staged liver resection. The conversion risk to open surgery was 5.2% (95% CI 4.5%–6.0%) for laparoscopic resection, 7.1% (95% CI 2.8%–11.5%) for robotic resection, 4.9% (95% CI 2.8%–7.1%) for simultaneous liver and colorectal resection, and 5.4% (95% CI 4.1%–6.7%) for staged liver resection. Conversion to open surgery was correlated with increased R0 resection rates in laparoscopic resection (coefficient: 9.620, p < 0.001) and staged liver resection (coefficient: 0.826, p < 0.001); however, it did not affect R0 resection in robotic resection (coefficient: –10.130, p = 0.273) or in simultaneous liver and colorectal resection (coefficient: 0.471, p = 0.351). Conversion to open surgery was correlated with increased mortality in laparoscopic resection (coefficient: 9.101, p < 0.001) but not in other subgroups. Morbidity remained unaffected by conversion to open surgery across all subgroups.

Table 2.

Results from subgroup analyses

	Independent variable	Dependent variable
	Conversion to open (95% CI)	R0 resection		Mortality		Morbidity
		Coefficient	p-value	Coefficient	p-value	Coefficient	p-value
All patients	5.8% (5–6.6)	2.167	< 0.001	7.585	0.001	1.737	0.003
Laparoscopic resection	5.2% (4.5–6.0)	9.620	< 0.001	9.101	< 0.001	1.339	0.088
Robotic resection	7.1% (2.8–11.5)	–10.130	0.273	Not estimated		–0.879	0.673
Simultaneous liver and colorectal resection	4.9% (2.8–7.1)	0.471	0.351	–4.187	0.337	–0.313	0.814
Staged liver resection	5.4% (4.1–6.7)	0.826	< 0.001	7.518	0.124	–0.811	0.562

CI, confidence interval.

Sensitivity analyses

Leave-one-out analysis and analysis limited to studies at low overall risk of bias did not alter the main conclusions.

DISCUSSION

We performed a systematic review and meta-analysis, incorporating meta-regression, to assess the prognostic impact of unplanned conversion to open surgery during minimally-invasive resection of CRLM, specifically considering the underlying reasons for conversion. Evaluation of 18,138 patients across 86 studies demonstrated that unplanned conversion to open was correlated with a higher rate of R0 resection, but was associated with increased postoperative mortality and morbidity risk. Distinct analyses by reason for conversion indicated that conversion due to oncological concern was correlated with improved R0 resection, while conversion related to iatrogenic injuries was correlated with lower R0 resection rates; conversion prompted by technical difficulties was associated with reduced postoperative morbidity. Conversion to open surgery did not significantly influence 5-year OS or 5-year DFS outcomes. The GRADE certainty of evidence was low to moderate.

The prognostic relevance of unplanned conversion to open surgery during minimally-invasive resection of CRLM has not been previously addressed in the literature; as such, direct comparisons with earlier studies are not possible, but the external validity of our findings is supported by existing research. The conversion rate observed was 5.8% (95% CI 5.0%–6.6%), aligning with previous reports. Goh et al. [8] documented a conversion to open risk of 2%–13% in the context of laparoscopic liver resections. R0 resection was achieved in 87.3% (95% CI 84.8%–89.9%) of patients as observed in our study, which is in agreement with the network meta-analysis by Mkabaah et al. [12], demonstrating R0 resection rates of 84.6% and 82.1% following laparoscopic and robotic resection of CRLM, respectively [12]. Postoperative mortality and morbidity rates in this investigation were 0.4% (95% CI 0.3%–0.5%) and 26.4% (95% CI 21.5%–31.2%), respectively, which are in line with prior studies [8,12,13]. Furthermore, a 5-year OS of 55.2% (95% CI 51.4%–59.0%) and a 5-year DFS of 33.1% (95% CI 28.9%–37.2%) found in this study are corroborated by the results of earlier reports [3,13,14]. Collectively, the evidence from previous studies supports the external validity of our findings regarding both the prognostic factor (conversion to open) and the associated clinical outcomes.

The results of the present study addressing the effect of conversion to open surgery on outcomes after CRLM resection are original. The reasons for conversion, listed in order of frequency, were bleeding, technical difficulties, oncological concerns, and iatrogenic injuries. Conversion prompted by oncological concern resulted in an increased rate of R0 resection. This is a significant finding as it emphasizes that conversion to open should not be regarded as a failure; rather, surgeons should proceed to conversion if there is doubt that a complete resection is achievable with a minimally-invasive approach. Conversely, conversion driven by technical difficulties was associated with reduced postoperative morbidity. This relationship may be attributed to the fact that technical challenges such as adhesions, tumor size or location, and limited access to vascular pedicles can heighten the risk of intraoperative bleeding and iatrogenic injury, thereby increasing morbidity. Notably, the study found that conversion due to iatrogenic injuries was associated with a decreased rate of R0 resection. Based on this, it may be hypothesized that technical challenges could elevate the risk of iatrogenic injuries, which may subsequently compromise R0 resection; therefore, timely conversion in the presence of technical difficulties may also indirectly enhance oncological outcomes.

Although conversion prompted by bleeding was not linked to higher rates of postoperative morbidity or mortality, it is crucial to recognize that avoiding conversion to open in instances of uncontrolled bleeding increases the risk of postoperative morbidity and mortality. The protective effect of conversion in such cases is difficult to demonstrate in clinical research because of confounding by indication; conversion to open becomes ethically and clinically necessary to avert adverse outcomes in the context of uncontrolled bleeding, and thus the consequences of not converting cannot be ethically studied.

Although the pooled analysis of all studies suggested that conversion to open may correlate with increased morbidity or mortality, subgroup analyses stratified by reason for conversion revealed no association between conversion to open and these outcomes. This discrepancy warrants discussion. One possible explanation is that the pooled analysis included a larger sample size, while subgroup analyses had smaller sample sizes because some studies lacked data on reasons for conversion; this may have resulted in a type 2 error in the subgroup analyses that failed to detect a correlation between conversion and complications. Alternatively, the reason for conversion may have acted as a confounding factor in the pooled analysis, with its effect being mitigated in the subgroup analyses.

Regarding the implications of the current study’s findings for real-world practice, it is reasonable to suggest that surgeons should maintain a low threshold for conversion to open surgery in situations involving bleeding, technical challenges, oncological concerns, or iatrogenic injuries, as this approach may enhance patient outcomes. Additionally, preoperative multidisciplinary discussions should objectively and thoroughly assess factors that could increase the risk of technical difficulty or oncological issues (including the number, size, and location of lesions, as well as their proximity to major vessels) when selecting patients for minimally-invasive resection.

The current study exhibits both strengths and limitations. The very large sample size reduces the risk of type 2 error, as evidenced by the narrow confidence intervals observed for most reported outcomes. Additionally, the findings of this study, with respect to the prognostic impact of conversion to open surgery and related outcomes, are supported by previous research and therefore possess external validity. Sensitivity analyses further demonstrated robustness of the results. However, a major limitation is that none of the studies in the existing literature are adequately powered to specifically assess conversion to open surgery, nor do they directly compare patients who underwent conversion with those who did not; consequently, a direct comparison meta-analysis could not be performed in this context. As an alternative, we implemented a meta-regression model, which is a recognized method for synthesizing evidence when evaluating prognostic factors such as conversion to open surgery. Given that the meta-regression included only one independent variable (conversion to open), residual confounding from other potential factors cannot be excluded. Statistical heterogeneity between studies was high, potentially due to variability in the proportion of patients undergoing laparoscopic versus robotic resections, as well as differences in simultaneous versus staged liver and colorectal resections; to address this, we conducted appropriate sensitivity and subgroup analyses and also downgraded the certainty of evidence to reflect this limitation. Unfortunately, insufficient data from included studies prevented subgroup analysis based on institutional volume or surgeon experience, factors that may contribute to the substantial between-study heterogeneity observed. Finally, Egger’s test indicated possible publication bias, which was accounted for by downgrading the certainty of the evidence.

In conclusion, unplanned conversion to open surgery may have both prognostic and oncological implications in minimally-invasive resection of CRLM. Although conversion due to bleeding or iatrogenic injuries is generally straightforward, conversion for technical difficulties or oncological concerns should not be regarded as an unsuccessful outcome because it may be correlated with improved prognostic results.

SUPPLEMENTARY DATA

Supplementary data related to this article can be found at https://doi.org/10.14701/ahbps.25-166.