Surg Endosc (2014) 28:950–960 DOI 10.1007/s00464-013-3254-3
and Other Interventional Techniques
Long-Term Outcome of Laparoscopic Versus Open Liver Resection for Hepatocellular Carcinoma: A Case-Controlled Study with Propensity Score Matching Hyeyoung Kim • Kyung-Suk Suh • Kwang-Woong Lee • Nam-Joon Yi • Geun Hong • Suk-Won Suh • Tae Yoo • Min-Su Park • YoungRok Choi • Hae Won Lee
Received: 6 May 2013 / Accepted: 1 October 2013 / Published online: 23 October 2013 Ó Springer Science+Business Media New York 2013
Abstract Background Laparoscopic liver resection (LR) for hepatocellular carcinoma (HCC) is usually applied to superficial and left-side small lesions. Therefore, well designed comparative studies about the results of LR versus open liver resection (OR) for HCC are difficult and still uncommon. The aim of this study was to compare the perioperative and long-term oncologic outcomes of LR versus OR for HCC between well-matched patient groups. Methods Between January 2000 and March 2012, 205 patients (43 with intent-to-treat with LR, 162 OR) underwent primary liver resection of less than three segments for HCC in our center. To select a comparison group, propensity score matching (PSM) was used at 1:1 ratio with covariates of baseline characteristics, including tumor characteristics. Outcomes were compared between the matched groups. Results The two groups were well balanced by PSM and 29 patients were matched respectively. In LR, there was more non-anatomical resection (65.5 vs. 34.5 %; p = 0.012), less postoperative ascites (0.0 vs. 17.2 %; p = 0.025), and shorter hospital stay (7.69 ± 2.94 vs. 13.38 ± 7.37 days; p \ 0.001). With the exception of these, there were no significant differences in perioperative and long-term outcomes.
H. Kim K.-S. Suh (&) K.-W. Lee N.-J. Yi G. Hong S.-W. Suh T. Yoo M.-S. Park Y. Choi H. W. Lee Department of Surgery, Seoul National University College of Medicine, 101 Daehak-ro, Jongno-gu, Seoul 110-744, Republic of Korea e-mail: [email protected]; [email protected] H. Kim H. W. Lee Department of Surgery, Seoul Metropolitan Government—Seoul National University Boramae Medical Center, Seoul, Republic of Korea
123
The 1-, 3- and 5-year survivals were 100, 100 and 92.2 % in LR, and 96.5, 92.2 and 87.7 % in OR (p = 0.267), respectively. The 1-, 3- and 5-year disease-free survivals were 81.7, 61.7 and 54.0 % in LR, and 78.6, 60.9 and 40.1 % in OR, respectively (p = 0.929). Conclusions The outcome of LR for HCC was technically feasible and safe in selected patients, and LR showed similar perioperative and long-term oncologic outcomes when compared with OR matched with PSM. Keywords Laparoscopic liver resection Hepatectomy Hepatocellular carcinoma Laparoscopy Propensity score matching Case-controlled study
Hepatocellular carcinoma (HCC) is the sixth most common neoplasm and the third most frequent cause of cancerrelated death in the world [1, 2]. In most cases, HCC develops within an established background of chronic liver disease (70–90 % of all patients) [3], and is the leading cause of death among patients with cirrhosis [2]. Potentially curative therapy for HCC includes liver resection, liver transplantation, and local ablation [2, 4, 5]. Resection remains the standard treatment for resectable HCC [6]. Since laparoscopic liver resection (LR) for HCC was first reported in 1995 by Hashizume et al. [7], the use of LR for HCC has increased steadily. LR is now well established and considered standard practice in left lateral sectionectomy [6]. LR is feasible and safe for HCC when performed by experienced surgeons [6, 8, 9], and several studies have reported comparative results of LR versus open liver resection (OR) for HCC [10–16]. However, LR for HCC has mainly applied to left lateral sectionectomy or small superficial lesions. A well-matched study is needed
Surg Endosc (2014) 28:950–960
for comparison with OR. Nevertheless, to the best of our knowledge, to date there has been no reported prospective, randomized controlled study of LR versus OR for HCC [9, 17–19], and there have been few studies of long-term oncologic outcomes for HCC with liver cirrhosis [10, 13–15, 17, 20–22]. Propensity score was first introduced by Rosenbaum and Rubin in 1983 [23], and has become increasingly used in observational studies in medical research to reduce the impact of treatment-selection bias in the comparison of treatment to a non-randomized control using observational data [24–26]. The aim of the present study was therefore to compare the perioperative and long-term oncologic outcomes of LR versus OR for HCC with well-compensated liver cirrhosis using propensity score matching (PSM).
Materials and Methods Patient Selection and Data Collection We retrospectively reviewed patients who underwent primary liver resection for HCC with liver cirrhosis at Seoul National University Hospital in Korea, from January 2000 to March 2012. Selection criteria for LR in our center included the following: Child’s class A cirrhosis, located in the peripheral segments of the liver and curatively treatable by limited resection (\3 segments) and usually solitary, small (B5 cm) HCC, and no history of previous major laparotomy at upper abdomen. The diagnosis of liver cirrhosis was based on preoperative imaging, including abdominal computed tomography or magnetic resonance imaging. During the selection of patients for this study, we excluded the following: history of previous operation for HCC other than percutaneous radiofrequency ablation or transarterial chemoembolization (TACE), preoperative non-curative intent and grossly positive resection margin after operation, and combined HCC on pathologic results. Among 1,566 liver resections for HCC during our study period, 205 patients were gathered, among whom 43 patients were in the intent-to-treat with LR, while the remaining 162 patients were in the intent-to-treat with OR. This study was approved by the independent ethics committee (that is, Institutional Review Board) of Seoul National University Hospital. Surgical Procedure The types of liver resections were defined by Brisbane 2000 terminology [27]. Anatomical resections were preferred over non-anatomical resections, and non-anatomic wedge resection or tumorectomy were performed for small
951
peripheral lesions. Vascular inflow was controlled selectively in anatomical resection and hilar clamping was applied as necessary in non-anatomical resection. Among 43 patients who underwent trial of LR, conversion to laparotomy occurred in 23.3 % (n = 10). The most common cause of open conversion was difficulty of approach (n = 6, four cases due to tumor location and two for severe adhesion), followed by bleeding (n = 2), hypotension after carbon dioxide inflation (n = 1) and seeding risk (n = 1). Almost all conversions were performed at the beginning after gas inflation and laparoscopic exploration of the lesion, or in early laparoscopic manipulation. Laparoscopic procedures included total or pure LR (including one case of thoracoscopic transdiaphragmatic liver resection) in 48.5 % (n = 16), hand-assisted LR in 15.2 % (n = 5) and hybrid LR in 36.4 % (n = 12). The patients were placed in the supine position with right tilt for the left-sided HCC in LR. For the right-sided HCC, patients were placed in the supine position in the laparoscopy-assisted technique, or placed in the left lateral decubitus position in the totally laparoscopic technique. We used three to five ports (5–12 mm in diameter) according to the tumor location, and the first periumbilical port for laparoscopic camera was inserted by open technique. The procedures were performed under carbon dioxide pneumoperitoneum, and intra-abdominal pressure was electronically maintained below 12 mmHg. A 30° laparoscope was used. The liver was examined by vision and, as necessary, intraoperative ultrasonography confirmed the location of the lesions and identified their relationship with the intrahepatic vascular structures. Parenchymal transection was performed with a harmonic scalpel (UltraCision, Ethicon, USA) and a laparoscopic Cavitron ultrasonic surgical aspirator (Valleylab, Ireland) and a thermofusion device (LigaSure, Tyco Healthcare, USA). At the end of the hepatic transection, fibrin glue was applied to the cut surface. The resected specimen was placed in a plastic bag and extracted, without fragmentation as possible, through a slightly enlarged periumbilical port site, or additional minilaparotomy of midline or unilateral transverse incision. With OR, right subcostal or upper midline or inverted L-shaped incisions were used according to tumor location. Intraoperative ultrasonography (Aloka, Tokyo, Japan) was performed routinely. Parenchymal transection was done with Cavitron ultrasonic surgical aspirator (Valleylab, Ireland), and fibrin glue was applied to the cut surface after hepatectomy. Propensity Score Matching A biostatistician in the Medical Research Collaborating Center in Seoul National University Hospital Biomedical
123
952
Research Institute performed 1:1 matching from the primarily selected 195 patients using the propensity scoring method. The openly converted patients (n = 10) were excluded in the process of PSM. Propensity scores were generated with the baseline characteristics, including sex, age, American Society of Anesthesiology (ASA) score, body mass index, underlying liver disease (such as hepatitis B or C virus-related liver cirrhosis or alcoholic liver cirrhosis), preoperative laboratory data related to liver function (total bilirubin, aspartate transaminase, alanine transaminase, platelet, albumin, prothrombin time, international normalized ratio), variables related to liver cirrhosis (preoperative Child’s class score, preoperative ascites, gastric or esophageal varix, cirrhosis on postoperative pathologic report), variables related to tumor characteristics (history of preoperative procedure for HCC such as TACE or radiofrequency ablation, preoperative alphafetoprotein, tumor number and size, microvascular invasion on postoperative pathologic report) and range of resection (\2 or C2 segments), and location of tumor (right or left or caudate lobe). Among three ways of using the propensity score for comparisons (matching, stratification, multivariable adjustment [28]), we used the matching method. The logistic regression model of propensity score from two groups and the above covariates was appropriately based on the assessment of goodness-of-fit statistics proposed by Lemeshow and Hosmer (p = 0.659) [24]. Using these propensity scores, 33 patients of LR were individually matched to 162 patients of OR with SAS version 9.2 (SAS Institute, Cary, NC, USA) at a 1:1 ratio. After PSM, in each group, 29 patients were selected for this study. The others were excluded because no match could be found. After matching, we compared the matched groups for examining covariate balance and whether there were statistical differences in baseline covariates between groups or not. Finally, after we confirmed that the PSM in our study was appropriate, each of 29 pairs of patients in the two groups was compared with respect to perioperative and long-term oncological outcomes. Postoperative Management and Follow-Up Patients were transferred to the intensive care unit or general surgical ward. Postoperative care was done as usual, such as is given for general liver resections in patients with liver cirrhosis, and included limited fluid administration, sodium restriction, perioperative antibiotics prophylaxis, anti-acids or proton pump inhibitors and early mobilization. Liver function tests were monitored at postoperative day 1, 2, 3, 5 and 7, and then as clinically needed. Timing of discharge depended on the condition of the
123
Surg Endosc (2014) 28:950–960
patient. Follow-up included liver function tests, alphafetoprotein, and computed tomographic or magnetic resonance imaging (or sometimes, ultrasonography) at every 3 or 4 months. Study Criteria for Comparing the Two Matched Groups The study criteria for comparing the two matched groups included the following: (i) clinicopathologic data including baseline characteristics of covariates used in PSM; (ii) intraoperative and surgical results; (iii) postoperative outcomes; and (iv) long-term oncologic outcomes in aspect of survival and recurrence. Postoperative morbidity and mortality were defined as events occurring during the same hospital stay or within 3 months of resection. Complications were graded by Clavien–Dindo classification [29]. Postoperative ascites was defined as clinically detectable and needing treatment (such as diuretics) or as abdominal drainage output (when present) of 500 ml or more per day continuously. Statistical Analysis PSM was performed by SAS version 9.2 (SAS Institute Inc., Cary, NC, USA). The other statistical analyses after PSM were performed with SPSS version 19.0 (SPSS Inc., an IBM Company, Chicago, IL, USA). In the process of PSM, Student’s t tests or Wilcoxon rank sum test for continuous variables, and v2 test or Fisher’s exact test for categorical variables were used in analyses and comparisons of baseline covariates. After PSM, baseline characteristics and perioperative and longterm oncologic outcomes of the matched data were compared using paired t test or the Wilcoxon signed rank test for continuous variables and McNemar’s test for categorical variables. All categorical data were expressed as number or frequency (%), and all continuous data were given as mean ± standard deviation. Survival rates were estimated using the Kaplan–Meier method and compared using the Log rank test or Breslow test according to the proportional or non-proportional hazard respectively. All p-values are two-sided and p \ 0.05 was considered statistically significant.
Results Baseline Characteristics Table 1 summarizes the baseline characteristics of the two groups. Both groups were well balanced for all variables, including covariates (sex, age, ASA score, body mass index, underlying liver disease, preoperative laboratory
Surg Endosc (2014) 28:950–960 Table 1 Comparison of baseline characteristics
953
Characteristics
LR (n = 29)
OR (n = 29)
p value
Age (years)
54.62 ± 9.16
53.90 ± 10.08
0.804
Male
22 (75.86)
19 (65.52)
Female
7 (24.14)
10 (34.48)
23.88 ± 2.61
22.37 ± 4.69
0.220
ASA score
1.90 ± 0.49
1.83 ± 0.47
1.000
Child-Pugh class A
28 (96.55)
29 (100.0)
Sex
Body mass index (kg/m2)
0.549
Underlying liver disease
0.317 0.261
Hepatitis B virus
24 (82.76)
27 (93.10)
Hepatitis C virus
3 (10.34)
1 (3.45)
Alcoholic hepatitis
2 (6.90)
1 (3.45)
Preoperative laboratory values
Data are means ± standard deviations or number (%) unless otherwise indicated LR laparoscopic liver resection, OR open liver resection, INR international normalized ratio, ICG indocyanine green, TACE transarterial chemoembolization, RFA radiofrequency ablation, PIVKA-II protein induced by vitamin K absence or antagonist-II
Total bilirubin (mg/dL)
0.99 ± 0.41
0.86 ± 0.31
0.168
Aspartate transaminase (U/L)
38.83 ± 21.93
43.17 ± 22.37
0.277
Alanine transaminase (U/L) Platelets (9103/lL)
44.55 ± 27.11 156.34 ± 54.19
41.62 ± 25.89 134.72 ± 37.83
0.885 0.094
Albumin (g/dL)
3.88 ± 0.38
3.98 ± 0.43
0.328
Prothrombin time (INR)
1.09 ± 0.09
1.11 ± 0.07
0.232
ICG-R15 (%)
15.3 ± 12.4
10.5 ± 4.6
0.407
Gastric/esophageal varix
5 (17.2)
5 (17.2)
1.000
Preoperative ascites
0 (0.00)
3 (10.34)
0.083
Previous laparotomy
3 (10.34)
7 (24.14)
0.103
Previous TACE/RFA
4 (13.8)
5 (17.2)
1.000 0.368
Alpha-fetoprotein (ng/mL)
1054.01 ± 4114.09
4077.16 ± 10683.53
8 (27.6)
10 (34.5)
0.754
409.30 ± 1055.67
531.29 ± 913.16
0.534
Right lobe
10 (34.48)
21 (72.41)
Left lobe
18 (62.07)
8 (27.59)
Caudate lobe
1 (3.45)
0 (0.0)
C200 ng/mL PIVKA-II (nAU/mL) Tumor location
values, preoperative Child’s class score, preoperative ascites, gastric/esophageal varix, preoperative TACE or radiofrequency ablation, alpha-fetoprotein, location of tumor) that were considered in the PSM, as previously described. In the LR group, there were 22 males (75.9 %) and 7 females (24.1 %), with a mean age of 54.6 years. The mean body mass index was 23.9 kg/m2, and ASA score 1.90. Almost all patients (96.6 %) were Child’s class A and accompanied with hepatitis B virus-related liver cirrhosis (82.8 %). Mean ICG-R15 was 15.3. Mean serum alpha-fetoprotein was 1.54 ng/Ml, and serum protein induced by vitamin K absence or antagonist-II was 409.3 nAU/mL. To summarize, there were no significant differences of baseline data between groups. In tumor location, even though HCC at the left lobe (n = 18) was more than at the right lobe (n = 10) in the LR group and vice versa in the OR group, there was no statistical significance (p = 0.088).
0.088
Surgical Outcomes Table 2 summarizes the types of operation in the groups. There were no major hepatectomies (C3 segments). Table 3 shows the intraoperative data and pathologic results by group. Non-anatomical resection was significantly more frequent in LR (65.5 vs. 34.5 % in OR; p = 0.012). Intraoperative blood loss tended to be greater in LR (483.85 ± 819.90 vs. 261.15 ± 300.66 mL in OR; p = 0.065). However, the other intraoperative data including range of resection (that was one of the covariates in the process of PSM; p = 0.791), operative time (210.48 ± 82.07 vs. 203.48 ± 51.19 min; p = 0.681), hilar clamping (18.51 vs. 34.48 %; p = 0.344), total clamping time (5.93 ± 14.55 vs. 4.67 ± 9.19 min; p = 0.790) and red blood cell transfusion (3.4 vs. 0.0 %; p = 0.317) were similar between the LR and OR groups. For pathologic results, there were no significant differences. The covariates used in PSM, such as cirrhosis, tumor
123
954
Surg Endosc (2014) 28:950–960
Table 2 Comparison of operation
Table 3 Comparison of intraoperative data and pathologic results
Operations
LR (n = 29)
Left lobe
18 (62.07)
8 (27.59)
8 (27.59)
5 (17.24)
10 (34.48)
3 (10.34)
Segment 2
5 (17.24)
0 (0.00)
Segment 3
5 (17.24)
1 (3.45)
Left lateral sectionectomy Tumorectomy or segmentectomy
Segment 4
OR (n = 29)
0 (0.00)
2 (6.90)
10 (34.48)
21 (72.41)
Right anterior sectionectomy
0 (0.00)
1 (3.45)
Right posterior sectionectomy
1 (3.45)
5 (17.24)
Tumorectomy or segmentectomy
9 (31.03)
15 (51.72)
Segment 5 Segment 6
3 (10.34) 5 (17.24)
1 (3.45) 6 (20.69)
Segment 7
0 (0.00)
2 (6.90)
Segment 8
1 (3.45)
6 (20.69)
1 (3.45)
0 (0.00)
Right lobe
Caudate lobe
Data are means ± standard deviations or number (%) unless otherwise indicated LR laparoscopic liver resection, OR open liver resection
size, solitary tumor, and microvascular invasion, were similar. Between the LR and OR groups, the other variables of resection margin, including minimal length of margin (11.07 ± 6.76 vs. 10.03 ± 7.49 mm; p = 0.392), resection margin C10 mm (58.6 vs. 44.8 %; p = 0.454), positive margin (10.3 vs. 3.4 %; p = 0.500), and tumor stage according to the 7th edition of the Union for International Cancer Control TNM classification; Edmondson and Steiner grade and differentiation were also similar (Table 3). Postoperative Outcomes Table 4 summarizes postoperative outcomes. In the LR group, duration of hospital stay (7.69 ± 2.94 vs. 13.38 ± 7.37 days; p \ 0.001) showed significant differences compared with the OR group. Additionally, postoperative ascites occurred much less in the LR group (0.0 vs. 17.2 % in the OR group; p = 0.025). However, there were no significant differences in variables of intensive care unit stay and duration, overall complication rate and postoperative wound complication (including subcutaneous seroma), bile leakage from operative hepatic cut surface, and severity of complications according to the Clavien–Dindo grade between two groups. There was no in-hospital mortality in either group. Long-Term Oncologic Outcomes Table 5 and Figs. 1 and 2 show long-term oncologic outcomes. There was no significant difference in mean follow-
123
Intraoperative data Hepatectomy Non-anatomical resection Anatomical resection Range of resection Tumorectomy/ uni-segmentectomy CBi-segmentectomy Operative time (min) Hilar clamping Total clamping time (min) Estimated blood loss (mL) Red blood cell transfusion Pathologic results Cirrhosis Tumor size (cm) Tumor number No. of patients with solitary tumor Microvascular invasion Resection margin Length (mm) C10 mm Positive Tumor stage (7th edition) I II IIIa [IIIb Edmondson and Steiner grade I II III IV Differentiation Differentiated Undifferentiated
LR (n = 29)
OR (n = 29)
p value
19 (65.5)
10 (34.5)
10 (34.5)
19 (65.5)
19 (65.52)
17 (58.62)
10 (34.48) 210.48 ± 82.07 5/27 (18.51) 5.93 ± 14.55
12 (41.38) 203.48 ± 51.19 10/29 (34.48) 4.67 ± 9.19
0.681 0.344 0.790
483.85 ± 819.90
261.15 ± 300.66
0.065
1 (3.4)
0 (0.0)
0.317
18 (62.07) 3.59 ± 2.17 1.24 ± 0.58 24 (82.76)
19 (65.52) 4.28 ± 2.55 1.03 ± 0.19 28 (96.55)
0.782 0.278 0.084 0.103
7 (24.14)
11 (37.93)
0.317
11.07 ± 6.76 17 (58.6) 3 (10.3)
10.03 ± 7.49 13 (44.8) 1 (3.4)
0.392 0.454 0.500 0.599
20 (69.0) 7 (24.1) 2 (6.9) 0 (0.0)
18 (62.1) 10 (34.5) 1 (3.4) 0 (0.0)
0.012
0.791
0.736
2/25 (8.0) 13/25 (52.0) 9/25 (36.0) 1/25 (4.0)
1/25 (4.0) 10/25 (10.0) 11/25 (44.0) 3/25 (12.0)
15/25 (60.0) 10/25 (40.0)
11 (44.0) 14 (56.0)
0.481
Bold value is statistically significant (p \ 0.05) Data are means ± standard deviations or number (%) unless otherwise indicated LR laparoscopic liver resection, OR open liver resection
Surg Endosc (2014) 28:950–960
955
Table 4 Comparison of postoperative outcomes LR (n = 29)
OR (n = 29)
Intensive care unit stay
5 (17.2)
5 (17.2)
1.000
Intensive care unit stay (days)
0.45 ± 1.09
0.41 ± 0.95
0.952
Hospital stay (days)
7.69 ± 2.94
13.38 ± 7.37
and Other Interventional Techniques
Long-Term Outcome of Laparoscopic Versus Open Liver Resection for Hepatocellular Carcinoma: A Case-Controlled Study with Propensity Score Matching Hyeyoung Kim • Kyung-Suk Suh • Kwang-Woong Lee • Nam-Joon Yi • Geun Hong • Suk-Won Suh • Tae Yoo • Min-Su Park • YoungRok Choi • Hae Won Lee
Received: 6 May 2013 / Accepted: 1 October 2013 / Published online: 23 October 2013 Ó Springer Science+Business Media New York 2013
Abstract Background Laparoscopic liver resection (LR) for hepatocellular carcinoma (HCC) is usually applied to superficial and left-side small lesions. Therefore, well designed comparative studies about the results of LR versus open liver resection (OR) for HCC are difficult and still uncommon. The aim of this study was to compare the perioperative and long-term oncologic outcomes of LR versus OR for HCC between well-matched patient groups. Methods Between January 2000 and March 2012, 205 patients (43 with intent-to-treat with LR, 162 OR) underwent primary liver resection of less than three segments for HCC in our center. To select a comparison group, propensity score matching (PSM) was used at 1:1 ratio with covariates of baseline characteristics, including tumor characteristics. Outcomes were compared between the matched groups. Results The two groups were well balanced by PSM and 29 patients were matched respectively. In LR, there was more non-anatomical resection (65.5 vs. 34.5 %; p = 0.012), less postoperative ascites (0.0 vs. 17.2 %; p = 0.025), and shorter hospital stay (7.69 ± 2.94 vs. 13.38 ± 7.37 days; p \ 0.001). With the exception of these, there were no significant differences in perioperative and long-term outcomes.
H. Kim K.-S. Suh (&) K.-W. Lee N.-J. Yi G. Hong S.-W. Suh T. Yoo M.-S. Park Y. Choi H. W. Lee Department of Surgery, Seoul National University College of Medicine, 101 Daehak-ro, Jongno-gu, Seoul 110-744, Republic of Korea e-mail: [email protected]; [email protected] H. Kim H. W. Lee Department of Surgery, Seoul Metropolitan Government—Seoul National University Boramae Medical Center, Seoul, Republic of Korea
123
The 1-, 3- and 5-year survivals were 100, 100 and 92.2 % in LR, and 96.5, 92.2 and 87.7 % in OR (p = 0.267), respectively. The 1-, 3- and 5-year disease-free survivals were 81.7, 61.7 and 54.0 % in LR, and 78.6, 60.9 and 40.1 % in OR, respectively (p = 0.929). Conclusions The outcome of LR for HCC was technically feasible and safe in selected patients, and LR showed similar perioperative and long-term oncologic outcomes when compared with OR matched with PSM. Keywords Laparoscopic liver resection Hepatectomy Hepatocellular carcinoma Laparoscopy Propensity score matching Case-controlled study
Hepatocellular carcinoma (HCC) is the sixth most common neoplasm and the third most frequent cause of cancerrelated death in the world [1, 2]. In most cases, HCC develops within an established background of chronic liver disease (70–90 % of all patients) [3], and is the leading cause of death among patients with cirrhosis [2]. Potentially curative therapy for HCC includes liver resection, liver transplantation, and local ablation [2, 4, 5]. Resection remains the standard treatment for resectable HCC [6]. Since laparoscopic liver resection (LR) for HCC was first reported in 1995 by Hashizume et al. [7], the use of LR for HCC has increased steadily. LR is now well established and considered standard practice in left lateral sectionectomy [6]. LR is feasible and safe for HCC when performed by experienced surgeons [6, 8, 9], and several studies have reported comparative results of LR versus open liver resection (OR) for HCC [10–16]. However, LR for HCC has mainly applied to left lateral sectionectomy or small superficial lesions. A well-matched study is needed
Surg Endosc (2014) 28:950–960
for comparison with OR. Nevertheless, to the best of our knowledge, to date there has been no reported prospective, randomized controlled study of LR versus OR for HCC [9, 17–19], and there have been few studies of long-term oncologic outcomes for HCC with liver cirrhosis [10, 13–15, 17, 20–22]. Propensity score was first introduced by Rosenbaum and Rubin in 1983 [23], and has become increasingly used in observational studies in medical research to reduce the impact of treatment-selection bias in the comparison of treatment to a non-randomized control using observational data [24–26]. The aim of the present study was therefore to compare the perioperative and long-term oncologic outcomes of LR versus OR for HCC with well-compensated liver cirrhosis using propensity score matching (PSM).
Materials and Methods Patient Selection and Data Collection We retrospectively reviewed patients who underwent primary liver resection for HCC with liver cirrhosis at Seoul National University Hospital in Korea, from January 2000 to March 2012. Selection criteria for LR in our center included the following: Child’s class A cirrhosis, located in the peripheral segments of the liver and curatively treatable by limited resection (\3 segments) and usually solitary, small (B5 cm) HCC, and no history of previous major laparotomy at upper abdomen. The diagnosis of liver cirrhosis was based on preoperative imaging, including abdominal computed tomography or magnetic resonance imaging. During the selection of patients for this study, we excluded the following: history of previous operation for HCC other than percutaneous radiofrequency ablation or transarterial chemoembolization (TACE), preoperative non-curative intent and grossly positive resection margin after operation, and combined HCC on pathologic results. Among 1,566 liver resections for HCC during our study period, 205 patients were gathered, among whom 43 patients were in the intent-to-treat with LR, while the remaining 162 patients were in the intent-to-treat with OR. This study was approved by the independent ethics committee (that is, Institutional Review Board) of Seoul National University Hospital. Surgical Procedure The types of liver resections were defined by Brisbane 2000 terminology [27]. Anatomical resections were preferred over non-anatomical resections, and non-anatomic wedge resection or tumorectomy were performed for small
951
peripheral lesions. Vascular inflow was controlled selectively in anatomical resection and hilar clamping was applied as necessary in non-anatomical resection. Among 43 patients who underwent trial of LR, conversion to laparotomy occurred in 23.3 % (n = 10). The most common cause of open conversion was difficulty of approach (n = 6, four cases due to tumor location and two for severe adhesion), followed by bleeding (n = 2), hypotension after carbon dioxide inflation (n = 1) and seeding risk (n = 1). Almost all conversions were performed at the beginning after gas inflation and laparoscopic exploration of the lesion, or in early laparoscopic manipulation. Laparoscopic procedures included total or pure LR (including one case of thoracoscopic transdiaphragmatic liver resection) in 48.5 % (n = 16), hand-assisted LR in 15.2 % (n = 5) and hybrid LR in 36.4 % (n = 12). The patients were placed in the supine position with right tilt for the left-sided HCC in LR. For the right-sided HCC, patients were placed in the supine position in the laparoscopy-assisted technique, or placed in the left lateral decubitus position in the totally laparoscopic technique. We used three to five ports (5–12 mm in diameter) according to the tumor location, and the first periumbilical port for laparoscopic camera was inserted by open technique. The procedures were performed under carbon dioxide pneumoperitoneum, and intra-abdominal pressure was electronically maintained below 12 mmHg. A 30° laparoscope was used. The liver was examined by vision and, as necessary, intraoperative ultrasonography confirmed the location of the lesions and identified their relationship with the intrahepatic vascular structures. Parenchymal transection was performed with a harmonic scalpel (UltraCision, Ethicon, USA) and a laparoscopic Cavitron ultrasonic surgical aspirator (Valleylab, Ireland) and a thermofusion device (LigaSure, Tyco Healthcare, USA). At the end of the hepatic transection, fibrin glue was applied to the cut surface. The resected specimen was placed in a plastic bag and extracted, without fragmentation as possible, through a slightly enlarged periumbilical port site, or additional minilaparotomy of midline or unilateral transverse incision. With OR, right subcostal or upper midline or inverted L-shaped incisions were used according to tumor location. Intraoperative ultrasonography (Aloka, Tokyo, Japan) was performed routinely. Parenchymal transection was done with Cavitron ultrasonic surgical aspirator (Valleylab, Ireland), and fibrin glue was applied to the cut surface after hepatectomy. Propensity Score Matching A biostatistician in the Medical Research Collaborating Center in Seoul National University Hospital Biomedical
123
952
Research Institute performed 1:1 matching from the primarily selected 195 patients using the propensity scoring method. The openly converted patients (n = 10) were excluded in the process of PSM. Propensity scores were generated with the baseline characteristics, including sex, age, American Society of Anesthesiology (ASA) score, body mass index, underlying liver disease (such as hepatitis B or C virus-related liver cirrhosis or alcoholic liver cirrhosis), preoperative laboratory data related to liver function (total bilirubin, aspartate transaminase, alanine transaminase, platelet, albumin, prothrombin time, international normalized ratio), variables related to liver cirrhosis (preoperative Child’s class score, preoperative ascites, gastric or esophageal varix, cirrhosis on postoperative pathologic report), variables related to tumor characteristics (history of preoperative procedure for HCC such as TACE or radiofrequency ablation, preoperative alphafetoprotein, tumor number and size, microvascular invasion on postoperative pathologic report) and range of resection (\2 or C2 segments), and location of tumor (right or left or caudate lobe). Among three ways of using the propensity score for comparisons (matching, stratification, multivariable adjustment [28]), we used the matching method. The logistic regression model of propensity score from two groups and the above covariates was appropriately based on the assessment of goodness-of-fit statistics proposed by Lemeshow and Hosmer (p = 0.659) [24]. Using these propensity scores, 33 patients of LR were individually matched to 162 patients of OR with SAS version 9.2 (SAS Institute, Cary, NC, USA) at a 1:1 ratio. After PSM, in each group, 29 patients were selected for this study. The others were excluded because no match could be found. After matching, we compared the matched groups for examining covariate balance and whether there were statistical differences in baseline covariates between groups or not. Finally, after we confirmed that the PSM in our study was appropriate, each of 29 pairs of patients in the two groups was compared with respect to perioperative and long-term oncological outcomes. Postoperative Management and Follow-Up Patients were transferred to the intensive care unit or general surgical ward. Postoperative care was done as usual, such as is given for general liver resections in patients with liver cirrhosis, and included limited fluid administration, sodium restriction, perioperative antibiotics prophylaxis, anti-acids or proton pump inhibitors and early mobilization. Liver function tests were monitored at postoperative day 1, 2, 3, 5 and 7, and then as clinically needed. Timing of discharge depended on the condition of the
123
Surg Endosc (2014) 28:950–960
patient. Follow-up included liver function tests, alphafetoprotein, and computed tomographic or magnetic resonance imaging (or sometimes, ultrasonography) at every 3 or 4 months. Study Criteria for Comparing the Two Matched Groups The study criteria for comparing the two matched groups included the following: (i) clinicopathologic data including baseline characteristics of covariates used in PSM; (ii) intraoperative and surgical results; (iii) postoperative outcomes; and (iv) long-term oncologic outcomes in aspect of survival and recurrence. Postoperative morbidity and mortality were defined as events occurring during the same hospital stay or within 3 months of resection. Complications were graded by Clavien–Dindo classification [29]. Postoperative ascites was defined as clinically detectable and needing treatment (such as diuretics) or as abdominal drainage output (when present) of 500 ml or more per day continuously. Statistical Analysis PSM was performed by SAS version 9.2 (SAS Institute Inc., Cary, NC, USA). The other statistical analyses after PSM were performed with SPSS version 19.0 (SPSS Inc., an IBM Company, Chicago, IL, USA). In the process of PSM, Student’s t tests or Wilcoxon rank sum test for continuous variables, and v2 test or Fisher’s exact test for categorical variables were used in analyses and comparisons of baseline covariates. After PSM, baseline characteristics and perioperative and longterm oncologic outcomes of the matched data were compared using paired t test or the Wilcoxon signed rank test for continuous variables and McNemar’s test for categorical variables. All categorical data were expressed as number or frequency (%), and all continuous data were given as mean ± standard deviation. Survival rates were estimated using the Kaplan–Meier method and compared using the Log rank test or Breslow test according to the proportional or non-proportional hazard respectively. All p-values are two-sided and p \ 0.05 was considered statistically significant.
Results Baseline Characteristics Table 1 summarizes the baseline characteristics of the two groups. Both groups were well balanced for all variables, including covariates (sex, age, ASA score, body mass index, underlying liver disease, preoperative laboratory
Surg Endosc (2014) 28:950–960 Table 1 Comparison of baseline characteristics
953
Characteristics
LR (n = 29)
OR (n = 29)
p value
Age (years)
54.62 ± 9.16
53.90 ± 10.08
0.804
Male
22 (75.86)
19 (65.52)
Female
7 (24.14)
10 (34.48)
23.88 ± 2.61
22.37 ± 4.69
0.220
ASA score
1.90 ± 0.49
1.83 ± 0.47
1.000
Child-Pugh class A
28 (96.55)
29 (100.0)
Sex
Body mass index (kg/m2)
0.549
Underlying liver disease
0.317 0.261
Hepatitis B virus
24 (82.76)
27 (93.10)
Hepatitis C virus
3 (10.34)
1 (3.45)
Alcoholic hepatitis
2 (6.90)
1 (3.45)
Preoperative laboratory values
Data are means ± standard deviations or number (%) unless otherwise indicated LR laparoscopic liver resection, OR open liver resection, INR international normalized ratio, ICG indocyanine green, TACE transarterial chemoembolization, RFA radiofrequency ablation, PIVKA-II protein induced by vitamin K absence or antagonist-II
Total bilirubin (mg/dL)
0.99 ± 0.41
0.86 ± 0.31
0.168
Aspartate transaminase (U/L)
38.83 ± 21.93
43.17 ± 22.37
0.277
Alanine transaminase (U/L) Platelets (9103/lL)
44.55 ± 27.11 156.34 ± 54.19
41.62 ± 25.89 134.72 ± 37.83
0.885 0.094
Albumin (g/dL)
3.88 ± 0.38
3.98 ± 0.43
0.328
Prothrombin time (INR)
1.09 ± 0.09
1.11 ± 0.07
0.232
ICG-R15 (%)
15.3 ± 12.4
10.5 ± 4.6
0.407
Gastric/esophageal varix
5 (17.2)
5 (17.2)
1.000
Preoperative ascites
0 (0.00)
3 (10.34)
0.083
Previous laparotomy
3 (10.34)
7 (24.14)
0.103
Previous TACE/RFA
4 (13.8)
5 (17.2)
1.000 0.368
Alpha-fetoprotein (ng/mL)
1054.01 ± 4114.09
4077.16 ± 10683.53
8 (27.6)
10 (34.5)
0.754
409.30 ± 1055.67
531.29 ± 913.16
0.534
Right lobe
10 (34.48)
21 (72.41)
Left lobe
18 (62.07)
8 (27.59)
Caudate lobe
1 (3.45)
0 (0.0)
C200 ng/mL PIVKA-II (nAU/mL) Tumor location
values, preoperative Child’s class score, preoperative ascites, gastric/esophageal varix, preoperative TACE or radiofrequency ablation, alpha-fetoprotein, location of tumor) that were considered in the PSM, as previously described. In the LR group, there were 22 males (75.9 %) and 7 females (24.1 %), with a mean age of 54.6 years. The mean body mass index was 23.9 kg/m2, and ASA score 1.90. Almost all patients (96.6 %) were Child’s class A and accompanied with hepatitis B virus-related liver cirrhosis (82.8 %). Mean ICG-R15 was 15.3. Mean serum alpha-fetoprotein was 1.54 ng/Ml, and serum protein induced by vitamin K absence or antagonist-II was 409.3 nAU/mL. To summarize, there were no significant differences of baseline data between groups. In tumor location, even though HCC at the left lobe (n = 18) was more than at the right lobe (n = 10) in the LR group and vice versa in the OR group, there was no statistical significance (p = 0.088).
0.088
Surgical Outcomes Table 2 summarizes the types of operation in the groups. There were no major hepatectomies (C3 segments). Table 3 shows the intraoperative data and pathologic results by group. Non-anatomical resection was significantly more frequent in LR (65.5 vs. 34.5 % in OR; p = 0.012). Intraoperative blood loss tended to be greater in LR (483.85 ± 819.90 vs. 261.15 ± 300.66 mL in OR; p = 0.065). However, the other intraoperative data including range of resection (that was one of the covariates in the process of PSM; p = 0.791), operative time (210.48 ± 82.07 vs. 203.48 ± 51.19 min; p = 0.681), hilar clamping (18.51 vs. 34.48 %; p = 0.344), total clamping time (5.93 ± 14.55 vs. 4.67 ± 9.19 min; p = 0.790) and red blood cell transfusion (3.4 vs. 0.0 %; p = 0.317) were similar between the LR and OR groups. For pathologic results, there were no significant differences. The covariates used in PSM, such as cirrhosis, tumor
123
954
Surg Endosc (2014) 28:950–960
Table 2 Comparison of operation
Table 3 Comparison of intraoperative data and pathologic results
Operations
LR (n = 29)
Left lobe
18 (62.07)
8 (27.59)
8 (27.59)
5 (17.24)
10 (34.48)
3 (10.34)
Segment 2
5 (17.24)
0 (0.00)
Segment 3
5 (17.24)
1 (3.45)
Left lateral sectionectomy Tumorectomy or segmentectomy
Segment 4
OR (n = 29)
0 (0.00)
2 (6.90)
10 (34.48)
21 (72.41)
Right anterior sectionectomy
0 (0.00)
1 (3.45)
Right posterior sectionectomy
1 (3.45)
5 (17.24)
Tumorectomy or segmentectomy
9 (31.03)
15 (51.72)
Segment 5 Segment 6
3 (10.34) 5 (17.24)
1 (3.45) 6 (20.69)
Segment 7
0 (0.00)
2 (6.90)
Segment 8
1 (3.45)
6 (20.69)
1 (3.45)
0 (0.00)
Right lobe
Caudate lobe
Data are means ± standard deviations or number (%) unless otherwise indicated LR laparoscopic liver resection, OR open liver resection
size, solitary tumor, and microvascular invasion, were similar. Between the LR and OR groups, the other variables of resection margin, including minimal length of margin (11.07 ± 6.76 vs. 10.03 ± 7.49 mm; p = 0.392), resection margin C10 mm (58.6 vs. 44.8 %; p = 0.454), positive margin (10.3 vs. 3.4 %; p = 0.500), and tumor stage according to the 7th edition of the Union for International Cancer Control TNM classification; Edmondson and Steiner grade and differentiation were also similar (Table 3). Postoperative Outcomes Table 4 summarizes postoperative outcomes. In the LR group, duration of hospital stay (7.69 ± 2.94 vs. 13.38 ± 7.37 days; p \ 0.001) showed significant differences compared with the OR group. Additionally, postoperative ascites occurred much less in the LR group (0.0 vs. 17.2 % in the OR group; p = 0.025). However, there were no significant differences in variables of intensive care unit stay and duration, overall complication rate and postoperative wound complication (including subcutaneous seroma), bile leakage from operative hepatic cut surface, and severity of complications according to the Clavien–Dindo grade between two groups. There was no in-hospital mortality in either group. Long-Term Oncologic Outcomes Table 5 and Figs. 1 and 2 show long-term oncologic outcomes. There was no significant difference in mean follow-
123
Intraoperative data Hepatectomy Non-anatomical resection Anatomical resection Range of resection Tumorectomy/ uni-segmentectomy CBi-segmentectomy Operative time (min) Hilar clamping Total clamping time (min) Estimated blood loss (mL) Red blood cell transfusion Pathologic results Cirrhosis Tumor size (cm) Tumor number No. of patients with solitary tumor Microvascular invasion Resection margin Length (mm) C10 mm Positive Tumor stage (7th edition) I II IIIa [IIIb Edmondson and Steiner grade I II III IV Differentiation Differentiated Undifferentiated
LR (n = 29)
OR (n = 29)
p value
19 (65.5)
10 (34.5)
10 (34.5)
19 (65.5)
19 (65.52)
17 (58.62)
10 (34.48) 210.48 ± 82.07 5/27 (18.51) 5.93 ± 14.55
12 (41.38) 203.48 ± 51.19 10/29 (34.48) 4.67 ± 9.19
0.681 0.344 0.790
483.85 ± 819.90
261.15 ± 300.66
0.065
1 (3.4)
0 (0.0)
0.317
18 (62.07) 3.59 ± 2.17 1.24 ± 0.58 24 (82.76)
19 (65.52) 4.28 ± 2.55 1.03 ± 0.19 28 (96.55)
0.782 0.278 0.084 0.103
7 (24.14)
11 (37.93)
0.317
11.07 ± 6.76 17 (58.6) 3 (10.3)
10.03 ± 7.49 13 (44.8) 1 (3.4)
0.392 0.454 0.500 0.599
20 (69.0) 7 (24.1) 2 (6.9) 0 (0.0)
18 (62.1) 10 (34.5) 1 (3.4) 0 (0.0)
0.012
0.791
0.736
2/25 (8.0) 13/25 (52.0) 9/25 (36.0) 1/25 (4.0)
1/25 (4.0) 10/25 (10.0) 11/25 (44.0) 3/25 (12.0)
15/25 (60.0) 10/25 (40.0)
11 (44.0) 14 (56.0)
0.481
Bold value is statistically significant (p \ 0.05) Data are means ± standard deviations or number (%) unless otherwise indicated LR laparoscopic liver resection, OR open liver resection
Surg Endosc (2014) 28:950–960
955
Table 4 Comparison of postoperative outcomes LR (n = 29)
OR (n = 29)
Intensive care unit stay
5 (17.2)
5 (17.2)
1.000
Intensive care unit stay (days)
0.45 ± 1.09
0.41 ± 0.95
0.952
Hospital stay (days)
7.69 ± 2.94
13.38 ± 7.37
