Search strategy flow diagram. LHR indicates laparoscopic hepatic resection; OHR, open hepatic resection.
Difference in operative blood loss (in milliliters). CI indicates confidence interval; LHR, laparoscopic hepatic resection; OHR, open hepatic resection; and WMD, weighted mean difference. Weights are from random-effects analysis.
Risk of resection margin less than 1 cm. CI indicates confidence interval; LHR, laparoscopic hepatic resection; OHR, open hepatic resection; and RR, relative risk. Weights are from random-effects analysis.
Difference in duration of hospital stay (in days). CI indicates confidence interval; LHR, laparoscopic hepatic resection; OHR, open hepatic resection; and WMD, weighted mean difference. Weights are from random-effects analysis.
Risk of overall postoperative complications. CI indicates confidence interval; LHR, laparoscopic hepatic resection; OHR, open hepatic resection; and RR, relative risk. Weights are from random-effects analysis.
All-cause mortality for malignant tumors. CI indicates confidence interval; HR, hazard ratio; LHR, laparoscopic hepatic resection; and OHR, open hepatic resection.
Risk of recurrence for malignant tumors. CI indicates confidence interval; HR, hazard ratio; LHR, laparoscopic hepatic resection; and OHR, open hepatic resection.
Laparoscopic vs Open Hepatic Resection for Benign and Malignant TumorsAn Updated Meta-analysis. Arch Surg. 2010;145(11):1109-1118. doi:10.1001/archsurg.2010.227
Author Affiliations: Harvard School of Public Health, Boston, Massachusetts (Drs Croome and Yamashita); Department of Surgery, University of Western Ontario, London, Ontario, Canada (Dr Croome); and Division of Cardiovascular Surgery, University of British Columbia, Vancouver, British Columbia, Canada (Dr Yamashita).
To compare laparoscopic hepatic resection (LHR) with open hepatic resection (OHR) for benign and malignant tumors.
MEDLINE, CENTRAL, and EMBASE databases were searched for relevant studies published between January 1, 1998, and May 1, 2009.
Studies clearly documenting a comparison of LHR with OHR for benign and malignant neoplasms were selected.
Operative and postoperative measures, resection margins, complications, and survival outcomes were evaluated. Weighted mean differences, relative risks, and hazard ratios (HRs) were calculated using a random-effects model.
Twenty-six studies were included in the meta-analysis. The HR of death for malignant tumors was significantly lower in the LHR group compared with the OHR group (HR, 0.64; P = .04). The HR of recurrence for malignant tumors was not significantly different between the 2 groups (HR, 0.79; P = .37). The LHR group had a lower operative blood loss (weighted mean difference, −161 mL; P < .001) and relative risk of total postoperative complications (relative risk, 0.40; P < .001). Duration of hospital stay, days of intravenous narcotic use, and days until oral intake were all significantly lower in the LHR group compared with the OHR group. Operative time between LHR and OHR was not significantly different.
Laparoscopic hepatic resection for malignant tumors is associated with a long-term survival that is at least comparable, if not superior, to OHR with no difference in disease recurrence. The use of LHR for benign and malignant tumors is a safe alternative to OHR with potential operative and postoperative benefits.
Laparoscopic hepatic resection (LHR) has gained in popularity because of its minimally invasive approach compared with open hepatic resection (OHR), potentially resulting in a reduction in perioperative morbidity and mortality combined with shorter recovery times. Barriers to widespread acceptance of LHR have been its technically challenging nature and concerns about the ability to achieve clear resection margins when it is performed for malignant lesions. If clear resection margins are not achieved, it is speculated that LHR will result in increased disease recurrence and decreased long-term survival. For LHR to become more ubiquitously accepted, there must be clear evidence that it has long-term outcomes similar to OHR, particularly in the setting of malignant tumors.
The use of LHR for benign and malignant tumors has been compared with OHR in previous studies; however, the numbers of patients and follow-up times have been limited. A meta-analysis on this topic published by Simillis et al1 in 2007 compared 8 studies published between 1998 and 2005 that reported on 409 hepatic resections. It found that laparoscopy provided the benefits of shorter hospital stays and lower operative blood loss while facilitating similar oncologic clearance and no increase in adverse events. However, this meta-analysis did not address the clinically relevant issue of long-term all-cause mortality and disease-free survival. Moreover, since that time, numerous additional studies with more participants and longer follow-up times have become available. Thus, an update on the role of LHR for benign and malignant tumors is needed with an analysis of long-term outcomes. The objective of this study was to perform an updated meta-analysis on the 26 studies published between January 1, 1998, and May 1, 2009, that reported on 1906 hepatic resections to determine whether the literature supports the use of laparoscopy as an alternative to open techniques for the resection of benign and malignant liver tumors.
The MEDLINE, CENTRAL, and EMBASE databases were searched independently by 2 investigators (K.P.C. and M.H.Y.) to retrieve relevant studies published between January 1, 1998, and May 1, 2009. The Medical Subject Heading keywords laparoscopy, hepatectomy, comparative study, and combinations of them, as well as the keywords laparoscopic, open, liver resection, liver surgery, and minimally invasive surgery, were used in text word searches. The “related articles” function was used to broaden our search. Reference lists of selected articles were also examined to find relevant studies not discovered during the database searches. All foreign language studies identified were also included.
To be selected, studies had to have compared laparoscopic with open approaches in patients undergoing liver resection for benign and malignant neoplasms with clear documentation of operative technique. Also, studies had to report on 1 or more of the outcomes of interest.
Studies were excluded if there was no clear reporting on the outcomes of interest or if it was impossible to conduct statistical analyses based on the published data. Early studies published as a series of articles from the same author or institution that contained significant overlap of patient data were excluded; only the most recently published study containing the most updated data was included. Studies reporting only hepatectomies for indications other than benign or malignant tumors were excluded.
Data extraction of all variables and outcomes of interest was performed independently by 2 readers (K.P.C. and M.H.Y.). Disagreements were resolved through discussion and consensus. Non–English-language articles were translated by a physician whose native language was that of the respective article.
Operative parameters assessed include operative blood loss, risk of requiring a blood transfusion, operative time, use of portal triad clamping, and duration of portal triad clamping. Oncologic clearance was evaluated by the number of positive surgical margins and the number of surgical margins less than 1 cm. Postoperative parameters were time to first oral intake, length of hospital stay, and duration of postoperative intravenous narcotics. Morbidity was assessed by the overall number of complications stratified into liver resection–related (eg, hepatic hemorrhage, biliary leakage, postoperative ascites, and liver failure) and general complications (eg, pneumonia, urinary tract infection, incisional hernia, and bowel perforation). Long-term outcomes included all-cause mortality and disease recurrence at maximal follow-up. The quality of studies was assessed using the modified Newcastle-Ottawa Scale with changes made to reflect the needs of this study.2 The maximum numbers of stars in the selection, comparability, and outcome categories were 3, 4, and 2, respectively.
The meta-analysis was performed using the Cochrane Collaboration and the Quality of Reporting of Meta-analyses guidelines. Statistical analyses were performed using STATA statistical software, version 10.0 (StataCorp LP, College Station, Texas). Continuous variables were analyzed using the weighted mean difference, whereas categorical dichotomous variables were assessed using relative risks (RRs). In studies that only reported the mean or median, range, and size of the trial, previously described methods were used to calculate the standard deviation.3 Survival data were evaluated by using hazard ratios (HRs) until the longest point of follow-up (2, 3, or 5 years). If studies did not present survival data in the form of HRs, previously described methods were used to calculate HRs from Kaplan-Meier curves.4 The level of significance was set at .05, and corresponding 95% confidence intervals (CIs) were reported. A random-effects model was used because this method is preferable in surgical meta-analyses to account for heterogeneity in surgical technique among centers. Heterogeneity was assessed by using the I2 statistic. Funnel plots were created to visually evaluate for the presence of publication bias. Also, the Egger test was used to test for statistically significant publication bias. Sensitivity analyses were undertaken by using the following subgroups: (1) studies with a year of publication inclusive of and later than 2006, (2) studies containing more than 20 patients in each group, and (3) studies of high quality with 6 or more stars.
The initial literature search retrieved 1209 relevant articles. Subsequently, a total of 28 articles published between January 1, 1998, and May 1, 2009, that met the inclusion criteria were identified. One conference abstract5 was excluded because it did not have sufficient data to be used in the analysis. Another study6 was excluded because it compared their results of LHR to the results of 1 of the other articles already included in our study. This resulted in 26 articles that were included in the meta-analysis (Figure 1). There were a total of 871 patients in the LHR group and 1019 in the OHR group. This accounted for 1906 hepatic resections because a few patients had more than 1 resection. Three studies looked at only benign tumors, 9 studies looked at only malignant tumors, and 14 studies looked at benign and malignant tumors. All studies were observational, with most conducting a matched comparative analysis (Table 1 and Table 2).
A total of 29 conversions (3.5%) from laparoscopic to open surgery were reported. One study did not convert any patients to open surgery but converted 20 from completely laparoscopic liver resection to hand-assisted resection. Another study did not report whether the investigators had any conversions. In all but 1 study, conversions were handled in an intention-to-treat manner.
Twenty-one studies reported on operative blood loss (Table 3). Blood loss was found to be 161 mL lower in the LHR group compared with the OHR group (95% CI, −209 to −114 mL; P < .001) (Figure 2). No significant difference was found in operative time, risk of receiving a blood transfusion, risk of portal triad clamping, or duration of clamp time between the LHR and OHR groups. Significant heterogeneity among the studies was observed for operative time (I2, 92.9%; 95% CI, 90.6% to 94.7%; P < .001), operative blood loss (I2, 84.0%; 76.7% to 89.0%; P < .001), risk of portal triad clamping (I2, 88.1%; 81.8% to 92.5%; P < .001), and duration of clamp time (I2, 93.3%; 86.1% to 96.8%; P < .001). These between-study differences likely reflect different surgeon preferences and varying surgical techniques.
No significant difference was found in the risk of a positive resection margin. However, the risk of having a surgical margin smaller than 1 cm was approximately 2 times higher in the LHR group (RR, 1.99; 95% CI, 1.31 to 3.02; P = .001) (Figure 3). This result was no longer statistically significant when the most influential study, by Morino et al,12 was excluded (data not shown). No significant heterogeneity was found among the studies on these resection margin parameters.
Duration of hospital stay was 3.52 days lower in the LHR group (95% CI, −4.27 to −2.77 days; P < .001) (Figure 4). The LHR group also had a 1.14-day shorter duration until oral intake (95% CI, −1.84 to −0.43 days; P = .002) and a 2.15-day shorter duration of intravenous narcotic requirements (−3.11 to −1.20 days; P < .001). Heterogeneity of studies was statistically significant for duration of hospital stay (I2, 85.5%; 95% CI, 79.5% to 89.8%; P < .001) and time until oral intake (I2, 97.1%; 95.6% to 98.1%; P < .001). This can likely be explained by different clinical practice at the various study sites. Heterogeneity for duration of intravenous narcotic requirements was not significant (I2, 52.2%; 95% CI, 0% to 86.2%; P = .12).
Twenty-three studies reported on overall postoperative complications. The RR of having any postoperative complication (Figure 5) was 59.9% lower in the LHR group (RR, 0.40; 95% CI, 0.31 to 0.52; P < .001). The RR of bile leak, however, was not significantly different between the 2 groups. No significant heterogeneity of studies was found on these complication parameters. In terms of perioperative mortality (<30 days from surgery), 4 deaths occurred in the LHR group and 7 deaths in the OHR group.
Nine studies reported on overall long-term survival. Using the available data, it was possible to calculate HRs of all-cause mortality for 6 of the studies. All 6 of these studies reported only on malignant tumors, with 5 of the 6 reporting solely on hepatocellular carcinomas. The HR of all-cause mortality for malignant tumors until 2 to 5 years of follow-up (Figure 6) was significantly lower in the LHR group (HR, 0.64; 95% CI, 0.42 to 0.99; P = .04). No significant heterogeneity was found (I2, 0%; 95% CI, 0% to 74.6%; P = .42). The HR of recurrence for malignant tumors until 2 to 5 years of follow-up (Figure 7) could be calculated in 3 of the studies and was not significantly different between the 2 groups (HR, 0.79; 95% CI, 0.46 to 1.34; P = .37). Again, no significant heterogeneity was found (I2, 0%; 95% CI, 0% to 89.6%; P = .48). Of the 7 studies that reported on the incidence of port-site recurrences, no study identified any affected patients.
Funnel plots demonstrated no visual evidence of publication bias. Also, Egger tests did not show any statistical evidence of publication bias.
In a separate analysis of studies published since the previous meta-analysis,1 the direction and significance of all RRs and weighted mean differences did not change except in 2 instances. The RR of having a surgical margin smaller than 1 cm had the same direction of effect but was no longer significant (RR, 1.73; 95% CI, 0.62 to 4.78; P = .29). Similarly, the HR of death had the same direction of effect but was no longer significant (HR, 0.67; 95% CI, 0.39 to 1.15; P = .14).
Another analysis of studies containing more than 20 patients in each of the LHR and OHR groups was undertaken to evaluate whether centers with larger patient experiences would have significantly different results. This analysis demonstrated that the measures of effect were similar to the overall results in all cases with 2 exceptions. The RR of portal triad clamping was now lower in the LHR group (RR, 0.25; 95% CI, 0.08 to 0.76; P = .01), as opposed to not being different in the overall results. This may be an indication that more experienced centers were less likely to require clamping during LHR. As well, the HR of all-cause death was no longer significant, although the direction of effect was preserved (HR, 0.67; 95% CI, 0.39 to 1.15; P = .14).
In terms of high-quality studies, those with a score of 6 stars or greater on the Newcastle-Ottawa Scale2 were also separately analyzed. Only the HR of all-cause mortality changed; it was not statistically significant in the higher-quality studies (HR, 0.68; 95% CI, 0.44 to 1.06; P = .08).
Laparoscopic surgery for the removal of benign and malignant liver tumors has become more common because of potentially reduced operative blood losses, shorter recovery times, fewer complications, and a better cosmetic outcome. In the setting of malignant tumors, these potential benefits can only be justified if the long-term survival after LHR is noninferior to that of OHR. This meta-analysis indicates that the risk of all-cause mortality for malignant tumors until 2 to 5 years of follow-up after LHR is 36% lower than that after OHR. Although this difference is statistically significant, we would caution anyone about concluding that LHR has better long-term survival than OHR. First, only 6 studies reported long-term survival results. Although the HR of all-cause mortality still had the same direction of effect in the sensitivity analyses, it was no longer significant. Second, the ability to calculate accurate study-specific HRs was limited by the data available in the published studies, requiring some estimations from Kaplan-Meier curves. Third, despite the matching done in the studies, it is possible that those who underwent LHR were a highly selected population with favorable parameters for survival that were not captured in the matching factors. However, all the study-specific HRs indicated a protective effect of LHR. Moreover, a previous study33 on an animal model has suggested that tumor growth may be slower after laparoscopic surgery than after comparable open surgery, possibly leading to improved survival. It is speculated that this finding is a result of less immune suppression. Thus, because of the consistent direction of effect and biological plausibility, we are confident in concluding that the long-term survival after LHR is at least equivalent to that after OHR.
After establishing at least similar long-term survival, it is prudent to examine the other operative and postoperative outcomes. The results of this meta-analysis provide additional support to the many claimed advantages of laparoscopic surgery. In this study, LHR was shown to be superior to OHR with regard to operative blood loss and risk of total postoperative complications. It has been speculated that a higher incidence of bile leak may be present after LHR, but these results did not show a significant difference. Furthermore, duration of hospital stay, days of intravenous narcotic use, and days until oral intake were all significantly lower in the LHR group. Operative time between LHR and OHR was not significantly different. These operative results are consistent with those previously reported by Simillis and colleagues.1
Another concern of laparoscopic resection for malignant tumors is oncologic clearance owing to the loss of tactile sensation, which can hinder the achievement of sufficient surgical margins. This meta-analysis did not find any difference in the risk of having a positive resection margin between LHR and OHR. A 1-cm free surgical margin is generally considered desirable in hepatic resections.34,35 This study demonstrated a larger risk of having a resection margin smaller than 1 cm in the LHR group. Although this finding is concerning, it did not result in higher long-term all-cause mortality or higher local recurrence for LHR.
Since this is a meta-analysis of observational studies, with many of them retrospective, there are several limitations that must be considered when interpreting these results. All meta-analyses of observational studies are subject to the biases inherent in the studies used, such as selection bias, measurement error, and confounding. Selection bias is probably the most concerning in this analysis because patients with more comorbidities would likely have been referred for conventional OHR rather than the novel, more difficult LHR. However, almost all the studies matched on pertinent clinical factors and potentially strong confounders, thereby minimizing selection bias and residual confounding. In addition, a sensitivity analysis of higher-quality studies showed results consistent with our overall analysis. Therefore, the effect of a higher level of comorbidities in the OHR group was probably minimal, although selection bias remains a significant limitation of this meta-analysis. We therefore recommend LHR as a safe alternative to OHR in carefully selected patients. Although it might have been optimal to include randomized controlled trials, we are not aware of any such published studies, although there is a randomized controlled trial based in South Korea registered with ClinicalTrials.gov.36 Another limitation of this meta-analysis was the heterogeneity seen in several of the parameters. As previously stated, the heterogeneity in the operative parameters likely exists because of different surgical techniques and preferences. Operative time is frequently surgeon dependent, as is use of portal triad clamping. The heterogeneity in the postoperative parameters is also not surprising given the variability of postoperative care at different institutions in different countries. Thus, it is not unexpected that the time until oral intake, duration of intravenous narcotic use, and length of hospital stay were shown to be heterogeneous across studies. Last, not all the individual studies presented their survival and recurrence data in the form of HRs. As a result, some HRs were calculated using the Kaplan-Meier curves and some estimations were made. This may have led to bias, although this would have been nondifferential across operative groups. However, all the published Kaplan-Meier curves showed a trend toward improved overall survival for the LHR patients.
In conclusion, LHR is a safe alternative to conventional OHR for carefully selected patients. For malignant tumors, LHR has a long-term survival rate that is at least comparable, if not superior, to OHR. Also, LHR leads to lower operative blood loss, shorter need for intravenous narcotics, faster time to oral intake, and shorter hospital stays. These results combined with no difference in postoperative adverse events and disease-free survival make the laparoscopic approach an attractive alternative for treating benign and malignant tumors.
Correspondence: Michael H. Yamashita, MDCM, MPH, CPH, 1081 Burrard St, Room 484, Vancouver, BC V6Z 1Y6 Canada (firstname.lastname@example.org).
Accepted for Publication: November 24, 2009.
Author Contributions: Drs Croome and Yamashita, as co–first authors, had full access to all the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis. Study concept and design: Croome and Yamashita. Acquisition of data: Croome and Yamashita. Analysis and interpretation of data: Croome and Yamashita. Drafting of the manuscript: Croome and Yamashita. Critical revision of the manuscript for important intellectual content: Croome and Yamashita. Statistical analysis: Croome and Yamashita. Administrative, technical, and material support: Croome and Yamashita.
Financial Disclosure: None reported.
Additional Contributions: Christie Jeon, MSc, and Russell de Souza, RD, MSc, provided assistance with the statistical analysis; Felix Akinbami, MBBS, MS, helped with the literature search; and Douglas Hanto, MD, PhD, provided suggestions during the critical revisal of the manuscript.