A systematic review and meta-analysis of studies comparing laparoscopic and open distal pancreatectomy

Abstract

Objectives

Currently, laparoscopic distal pancreatectomy (LDP) is regarded as a safe and effective surgical approach for lesions in the body and tail of the pancreas. This review compares outcomes of the laparoscopic technique with those of open distal pancreatectomy (ODP) and assesses the efficacy, safety and feasibility of each type of procedure.

Methods

Comparative studies published between January 1996 and April 2012 were included. Studies were selected based on specific inclusion and exclusion criteria. Evaluated endpoints were operative outcomes, postoperative recovery and postoperative complications.

Results

Fifteen non-randomized comparative studies that recruited a total of 1456 patients were analysed. Rates of conversion from LDP to open surgery ranged from 0% to 30%. Patients undergoing LDP had less intraoperative blood loss [weighted mean difference (WMD) −263.36.59 ml, 95% confidence interval (CI) −330.48 to −196.23 ml], fewer blood transfusions [odds ratio (OR) 0.28, 95% CI 0.11–0.76], shorter hospital stay (WMD −4.98 days, 95% CI −7.04 to −2.92 days), a higher rate of splenic preservation (OR 2.98, 95% CI 2.18–3.91), earlier oral intake (WMD −2.63 days, 95% CI −4.23 to 1.03 days) and fewer surgical site infections (OR 0.37, 95% CI 0.18–0.75). However, there were no differences between the two approaches with regard to operation time, time to first flatus and the occurrence of pancreatic fistula and other postoperative complications.

Conclusions

Laparoscopic resection results in improved operative and postoperative outcomes compared with open surgery according to the results of the present meta-analyses. It may be a safe and feasible option for patients with lesions in the body and tail of the pancreas. However, randomized controlled trials should be undertaken to confirm the relevance of these early findings.

Introduction

Laparoscopic surgery is now widely accepted and recognized as a standard technique in many surgical procedures.^1,2 Initially, the laparoscopic approach was not commonly used in pancreatic resection; however, increasing experience means laparoscopic distal pancreatectomy (LDP) is now performed more frequently in the surgical management of benign, non-invasive and even malignant lesions in the body and tail of the pancreas.³ Some studies have reported LDP to be associated with decreased intraoperative blood loss, a higher rate of splenic conservation, shorter hospital stay and less morbidity compared with open distal pancreatectomy (ODP).^4–6 By contrast, other studies report findings in favour of ODP.^7,8 Because these various reports indicate a discrepancy in the published literature, the present authors considered it necessary to summarize and analyse the published data to provide evidence to determine whether the literature supports the use of a laparoscopic approach as an alternative to open surgery in the resection of the distal pancreas.

Materials and methods

Study selection

Major databases including PubMed (MEDLINE), EMBASE, the Science Citation Index Expanded and the Cochrane Central Register of Controlled Trials (CENTRAL) in the Cochrane Library were searched for studies comparing outcomes in LDP and ODP, published in English from January 1996 to April 2012 (the first LDP was described in 1996). The medical search headings (MeSH) ‘laparoscopy’, ‘pancreatectomy’, ‘comparative study’ and combinations of these were used, as were the keywords ‘laparoscopic’, ‘open distal pancreatic resection’, ‘left pancreatic resection’, ‘pancreatic surgery’, ‘distal pancreatectomy’ and ‘minimally invasive surgery’. The reference lists of articles identified were examined to find relevant studies that had not been identified by the database searches. Only comparative clinical trials with full-text descriptions were included. The final inclusion of articles was determined by consensus between authors TJ and KA; when this failed, a third author (JJX) adjudicated.

Inclusion and exclusion criteria

Two authors (TJ and KA) identified and screened the search findings for potentially eligible studies. Inclusion criteria required the studies to: (i) be written in English and published in peer-reviewed journals; (ii) be human studies; (iii) examine at least one of the predetermined outcomes, and (iv) provide clear documentation of the operative techniques as ‘laparoscopic’ or ‘open’. In contexts in which multiple studies were published from the same institution and/or by the same authors, either the higher-quality study or the most recent publication was included in the analysis.

Exclusion criteria excluded: (i) abstracts, letters, editorials, expert opinions, case reports, reviews and studies lacking control groups; (ii) studies that included only patients undergoing spleen-preserving LDP or ODP, and (iii) studies in which patients converted to open surgery were included in the ODP group.

Outcomes of interest and definitions of complications

The following outcomes were used to compare patients undergoing LDP with those undergoing ODP. Operative outcomes included operative time, intraoperative blood loss, blood transfusion and splenic preservation. Postoperative recovery comprised time to oral intake, time to first flatus and duration of postoperative hospital stay. Postoperative complications included pancreatic fistula, clinically significant fistula [International Study Group on Pancreatic Fistula (ISGPF) grades B and C⁹], mortality, surgical site infection, intra-abdominal abscess, intra-abdominal fluid collection, postoperative haemorrhage, reoperation and readmission.

Data extraction and quality assessment

Data were extracted by two independent observers (WH and MAJ) using standardized forms. Data recorded included patient and study characteristics, pathological characteristics of resected specimens, and operative and postoperative outcomes. The quality of studies was assessed using the modified Newcastle–Ottawa Scale, modified to reflect the needs of this study.¹⁰ The maximum numbers of points awarded in the selection, comparability and outcome categories were three, four and two, respectively. Studies achieving six or more points were considered to be of high quality.¹¹ Only these studies were included in the final analyses. Subgroup analyses included all studies (high and low quality) in order to obtain a cumulative result.

Statistical analysis

Meta-analysis was performed using Review Manager Version 5.0 (Cochrane Collaboration, Oxford, UK). For continuous variables, treatment effects were expressed as the weighted mean difference (WMD) with the corresponding 95% confidence interval (CI). For categorical variables, treatment effects were expressed as the odds ratio (OR) with the corresponding 95% CI. Heterogeneity was evaluated using the chi-squared test and a P-value of <0.1 was considered to indicate statistical significance.¹² The fixed-effects model was initially calculated for all outcomes,¹³ but if the test rejected the assumption of homogeneity of studies, random-effects analysis was performed.¹⁴ If data in the included studies were considered to be inappropriate for meta-analysis, some outcomes were presented descriptively. Sensitivity analyses were performed by removing individual studies from the dataset and analysing the effect on the overall results to identify sources of significant heterogeneity. Subgroup analyses were undertaken by including high-quality studies to present cumulative evidence. Funnel plots were constructed to evaluate potential publication bias¹⁵ based on the major complication of pancreatic fistula.

Results

Description of trials included in the meta-analysis

The search strategy initially generated 427 relevant clinical reports. Finally, 26 full-text articles^{4–6,16–38} were identified for further investigation. Of these, two studies^16,17 were excluded for various reasons: one study included patients in whom LDP had been converted to open surgery in the ODP group,¹⁶ and the other included patients who had undergone only spleen-preserving and splenic vessel-preserving LDP or ODP.¹⁷ Three studies^18–20 reported from the same institution described overlapping patient populations. According to the inclusion criteria, the highest-quality study¹⁸ was included in the present meta-analysis. Similarly, of two studies published by Baker et al.,^21,22 the higher-quality study²¹ was included. Six studies were low-quality studies, which should be excluded. Finally, 15 high-quality studies were identified for inclusion; these included one prospective non-randomized study and 14 retrospective studies. Figure 1 shows the process by which comparative studies were selected for inclusion in the present meta-analysis.

Figure 1.

Flow diagram depicting the process of identifying and selecting studies for inclusion. LDP, laparoscopic distal pancreatectomy; OPD, open distal pancreatectomy

Study and patient characteristics

The study characteristics, quality and comparability assessments are shown in Table 1. Details of outcome measures are listed in Table 2. Definitions of pancreatic fistula are shown in Table 3. A total of 1456 patients were included; this number included 561 and 895 patients in the LDP and ODP groups, respectively. Five studies^5,6,26,29,32 looked at benign tumours only. The remaining 10 studies^{4,19,21,27,28,30,31,33,34,38} looked at both benign and malignant lesions. Most of the studies conducted a matched comparative analysis. The results of the analyses are summarized in Table 4.

Table 1.

Characteristics of included studies

Author(s)	Year	Country	Study design	LDP, n	ODP, n	Inclusion/exclusion criteria	Matching/comparable factorsa	Study qualityb(point scoring scale)
Velanovich4	2006	USA	Retro	15	15	Not specified	2, 4	*******
Teh et al.5	2007	USA	Retro	12	16	Benign pancreatic disease (I)	1, 2, 3, 4	******
Matsumoto et al.6	2008	Japan	Retro	14	19	Benign or borderline malignant pancreatic tumour (I)	1, 2, 3	*********
Kim et al.26	2008	Korea	Retro	93	35	Benign pancreatic disease (I)	2, 3, 4	*******
Kooby et al.18	2008	USA	Retro	23	189	Ductal adenocarcinoma (I)	1, 2, 3	******
						Background IPMN/cystadenocarcinoma (E)
Baker et al.21	2009	USA	PNR	27	85	Not specified	1, 3, 4	******
Jayaraman et al.27	2010	USA	Retro	100	100	Additional organ resection (E)	None	*******
DiNorcia et al.28	2010	USA	Retro	71	168	Laparoscopic-assisted DP (E)	1	******
						DP as part of a completion pancreatectomy (E)
						Concomitant portomesenteric venous resection and reconstruction (E)
						DP secondary to debridement for necrotizing pancreatitis (E)
						Non-pancreatic primary neoplasms or pancreatic injury (E)
Casadei et al.29	2010	Italy	Retro	22	22	Endocrine and cystic pancreatic tumours (I)	1, 2	*******
						Ductal adenocarcinoma (E)
Waters et al.30	2010	USA	Retro	18	22	Urgent surgery (E) Concurrent major surgery an indication for surgery of acute or chronic pancreatitis (E)	1, 4	******
Mehta et al.31	2012	France	Retro	30	30	DP for non-pancreatic pathologies (E)	2, 3, 4	******
						Resections amounting to less than a DP or total pancreatectomy (E)
Butturini et al.32	2012	Italy	Retro	43	73	Benign and borderline neoplasms (I)	2, 3	********
						Ductal cancer or other malignant tumours (E)
Abu et al.38	2011	UK	Retro	35	16	Additional organ resections (E)	3, 4	*****
Fox et al.33	2012	Canada	Retro	42	76	Additional organ resection (E)	1, 3, 4	*****
Limongelli et al.34	2012	Italy	Retro	16	29	Only tumour enucleation was accomplished (E)	1, 2, 4	******
						Additional organ resection (E)

^aMatching/comparable factors are: 1, American Society of Anesthesiologists (ASA) status; 2, type of pancreatic pathology; 3, mean size of tumour, and 4, presence of chronic pancreatitis.

^bBased on Newcastle–Ottawa Scale with maximum of **** for selection, ** for comparability, and ** for outcome.

LDP, laparoscopic distal pancreatectomy; ODP, open distal pancreatectomy; Retro, retrospective; PNR, prospective non-randomized; I, inclusion criteria; E, exclusion criteria; IPMN, intraductal papillary mucinous neoplasm; DP, distal pancreatectomy.

Table 2.

Operative and postoperative outcomes of included comparative studies

Author	Conversion	Spleen preservation		Operation time, min, mean/median (range)		Hospital stay, days, mean/median (range)		PF/clinically significant PF		Mortality
	n (%)	n (%)							n (%)		n (%)
		LDP	ODP	LDP	ODP	LDP	ODP	LDP	ODP	LDP	ODP
Velanovich4	3 (20)	NA	NA	NA	NA	5 (3–9)	8 (6–23)	2 (13)	2 (13)	NA	NA
Teh et al.5	2 (16.7)	NA	NA	212 (60–360)	278 (180–420)	6.2 (3–16)	10.6 (7–19)	1 (8.3)	1 (6.2)	0	0
Matsumoto et al.6	1 (7.1)	NA	NA	290.7 ± 53.2a	213.8 ± 84.6a	12.9 ± 4.8a	23.8 ± 11.8a	0	2 (10.5)	0	0
Kim et al.26	NA	38 (40.8)	2 (5.7)	195 (82–453)	190 (88–482)	10 (5–36)	16 (8–65)	8 (8.6)	5 (14.3)	0	0
Kooby et al.18	20 (12.6)	43 (30)	24 (12)	230 ± 97a	216 ± 100a	5.9 ± 3.8a	9.0 ± 6.0a	37 (26)/16 (11)	64 (32)/36 (18)	0	1 (1)
Baker et al.21	1 (3.7)	NA	NA	236 ± 82a	253.2 ± 292.3a	4.0 ± 0.3a	8.6 ± 0.7a	6 (22)/4 (14.8)	12 (14)/12 (14.1)	0	1 (1.2)
Jayaraman et al.27	32 (30)	14 (42.4)	33 (14.0)	195	160	5	6	8 (10.8)	13 (5.5)	0	2 (0.8)
DiNorcia et al.28	24 (25.3)	11 (15.5)	26 (15.8)	191 (163–214)	192 (157–236)	5 (4–6)	6 (5–8)	8 (11.3)	25 (14.9)	0	1 (0.6)
Casadei et al.29	0	4 (18.2)	4 (18.2)	225 ± 83b	145 ± 49b	8.0 ± 1.3b	11.0 ± 3.0b	2 (9.1)	4 (18.2)	0	0
Waters et al.30	2 (11)	5 (28)	3 (14)	224 (100–346)	234 (136–437)	6 (3–34)	8 (3–25)	2 (11.1)	4 (18.2)	0	0
Mehta et al.31	0	21 (70)	9 (30)	188 ± 72a	226 ± 87a	8.7 ± 4.2a	12.6 ± 8.7a	5 (16.7)/5 (16.7)	4 (13.3)/4 (13.3)	0	1 (3.3)
Butturini et al.32	0	19 (44.2)	8 (11)	180	180	8	9	12 (27.9)	10 (13.7)	0	0
Abu et al.38	0	14 (40)	3 (19)	200 (120–420)	225 (120–460)	7 (3–25)	11 (5–46)	10 (29)	7 (44)	NA	NA
Fox et al.33	5 (11.91)	15 (35.7)	17 (22.4)	304 (265–348)	281 (247–333)	5 (4–6)	7 (6–9)	12 (28.6)/0 (0)	10 (13.2)/4 (5.3)	NA	NA
Limongelli et al.34	1 (6)	5 (31)	3 (14)	204 ± 31a	160 ± 35a	6.4 ± 2.3a	8.6 ± 1.7a	3 (18)/1 (6)	6 (20)/5 (17.2)	0 (0)	1 (3)

^aMean ± standard deviation.

^bMedian ± standard deviation.

PF, pancreatic fistula; LDP, laparoscopic distal pancreatectomy; ODP, open distal pancreatectomy; NA, not available.

Table 3.

Definitions of postoperative pancreatic fistula in the included studies

Author	Definition of postoperative pancreatic fistula
Velanovich4	Amylase-rich fluid after PoD 3
Teh et al.5	Amylase >1000 after PoD 3
Matsumoto et al.6	Fluid amylase >5000 U/l after PoD 7
Kim et al.26	Drainage of >30 ml with amylase level >5-fold more than serum level ≥5 days after surgery
Kooby et al.18	Fluid amylase >350 mg/dl or need for postoperative percutaneous fluid collection
Baker et al.21	ISGPF definition
Jayaraman et al.27	ISGPF definition
DiNorcia et al.28	ISGPF definition
Casadei et al.29	ISGPF definition
Waters et al.30	Not defined
Mehta et al.31	ISGPF (grades B and C)
Butturini et al.32	ISGPF definition
Abu et al.38	ISGPF definition
Fox et al.33	ISGPF definition
Limongelli et al.34	ISGPF definition

PoD, postoperative day; ISGPF, International Study Group on Pancreatic Fistula.

Table 4.

Results of meta-analysis comparing outcomes in laparoscopic and open distal pancreatectomy in high-quality studies

Outcome of interest	Studies, n	Patients, n	OR/WMD	95% CI	P-value*
Operative outcomes
Operation time, min	5	343	19.71	−10.01 to 49.44	0.19
Intraoperative blood loss, ml	5	343	−263.36	−330.48 to −196.23	<0.00001
Blood transfused	5	375	0.28	0.11–0.76	0.01
Splenic preservation	10	1148	2.98	2.18–4.06	<0.00001
Postoperative recovery
Time to oral intake, days	2	161	−2.63	−4.23 to −1.03	0.001
Time to first flatus, days	2	161	−1.80	−2.14 to −1.47	<0.00001
Hospital stay, days	5	343	−4.98	−7.04 to −2.92	<0.00001
Postoperative complications
Clinically significant fistula	4	335	0.67	0.41–1.09	0.11
Postoperative haemorrhage	4	329	1.87	0.59–5.95	0.29
Intra-abdominal abscess	6	331	0.78	0.25–2.45	0.67
Intra-abdominal fluid collections	3	304	1.47	0.67–3.22	0.34
Surgical site infection	5	264	0.37	0.18–0.75	0.006
Mortality	11	1155	0.61	0.16–2.27	0.46
Reoperation	8	683	0.90	0.43–1.84	0.76
Readmission	5	772	0.70	0.41–1.21	0.20

WMD, weighted mean difference; OR, odds ratio; 95% CI, 95% confidence interval.

Operative outcomes

Five studies^{6,19,21,31,34} reported operation time. The present analysis showed no statistically significant difference between the two groups (WMD 19.71 min, 95% CI −10.01 to 49.44; P = 0.19). Similarly, findings in five studies^{6,19,21,31,34} were pooled to provide an estimation of mean blood loss in each of the LDP and ODP groups. Intraoperative blood loss was significantly lower in the LDP group than in the ODP group (WMD −263.36 ml, 95% CI −330.48 to −196.23; P < 0.00001). Additionally, fewer patients required blood transfusions in the LDP group (OR 0.28, 95% CI 0.11–0.76; P = 0.01). The rate of splenic preservation was identified as significantly higher in the LDP group than in the ODP group (OR 2.98, 95% CI 2.18–3.91; P < 0.00001). Forest plots are shown in Fig. 2.

Figure 2.

Forest plots illustrating the results of a meta-analysis comparing operative outcomes in laparoscopic and open distal pancreatectomy. Pooled odds ratios (ORs) or weighted mean differences (WMDs) with 95% confidence intervals (CIs) were calculated using the fixed-effects model or the random-effects model. (a) Operation time. (b) Intraoperative blood loss. (c) Blood transfusions. (d) Splenic preservation. SD, standard deviation

Postoperative recovery

Patients in the LDP group had a shorter postoperative hospital stay (WMD −4.98 days, 95% CI −7.04 to −2.92; P < 0.00001) and were able to resume flatus earlier (WMD −1.80 days, 95% CI −2.14 to −1.47; P < 0.000001) than their counterparts in the ODP group (Fig. 3). However, there was no significant difference between the groups in time to oral intake (WMD −2.63 days, 95% CI −4.23 to 1.03; P = 0.001).

Figure 3.

Forest plots illustrating the results of a meta-analysis comparing postoperative recovery outcomes in laparoscopic and open distal pancreatectomy. Pooled weighted mean differences (WMDs) with 95% confidence intervals (CIs) were calculated using the random-effects model. (a) Time to first flatus. (b) Time to oral intake. (c) Length of hospital stay. SD, standard deviation

Postoperative complications

There was no difference in rates of clinically significant fistula (OR 0.67, 95% CI 0.41–1.09; P = 0.11) and mortality (OR 0.61, 95% CI 0.16–2.27; P = 0.46) between the two techniques. The LDP group experienced fewer surgical site infections (OR 0.37, 95% CI 0.18–0.75; P = 0.006), but other postoperative complications, such as intra-abdominal abscesses (OR 0.78, 95% CI 0.25–2.45; P = 0.67), intra-abdominal fluid collections (OR 1.47, 95% CI 0.67–3.22; P = 0.34) and postoperative haemorrhage (OR 1.87, 95% CI 0.59–5.95; P = 0.29), were found to occur at similar frequencies in both groups. In addition, rates of reoperation (OR 0.90, 95% CI 0.43–1.84; P = 0.76) and readmission (OR 0.70, 95% CI 0.41–1.21; P = 0.20) did not differ between the groups (Fig. 4).

Figure 4.

Forest plots illustrating the results of a meta-analysis comparing postoperative complications in laparoscopic and open distal pancreatectomy. Pooled odds ratios (ORs) with 95% confidence intervals (CIs) were calculated using the fixed-effects model. (a) Clinically significant fistula. (b) Postoperative haemorrhage. (c) Intra-abdominal abscess. (d) Intra-abdominal fluid collections. (e) Surgical site infection. (f) Mortality. (g) Reoperation. (h) Readmission

Sensitivity and subgroup analysis

Sensitivity analyses were carried out by excluding each study from the analysis of each outcome measure. These exclusions did not alter the results obtained in cumulative analyses. In addition, subgroup analyses were undertaken for all outcome measures by including only high-quality studies. Analysis of the high-quality studies showed time to first flatus to be significantly shorter (WMD −1.80 days, 95% CI −2.14 to −1.47; P < 0.00001). However, the analyses for other outcomes did not change in comparison with previous analyses. These are summarized in Table 4.

Publication bias

A funnel plot based on the incidence of pancreatic fistula is shown in Fig. 5. None of the studies lies outside the limits of the 95% CI and hence there is no evidence of publication bias.

Figure 5.

Funnel plot based on incidences of pancreatic fistula. The funnel plot revealed no publication bias

Discussion

Gagner and Pomp³⁹ are considered to have pioneered the introduction of laparoscopy in pancreatic surgery for chronic pancreatitis and published the first description of an LDP. Initially, laparoscopy was mainly used for diagnostic purposes in pancreatology. However, laparoscopic pancreatic resections are gaining in popularity as a result of improvements in technology and increasing laparoscopic surgical experience. Distal pancreatectomy, which involves the resection of the pancreas to the left of the superior mesenteric vessels, is regarded as the standard procedure in chronic pancreatitis or benign and malignant tumours in the body or tail of the pancreas.⁴⁰ The increased use of LDP represents a paradigm shift from the practice of ODP and the former is now recognized as providing feasible, safe and effective treatment for some conditions of the pancreas.³

The results of the present meta-analysis favour the laparoscopic approach with regard to intraoperative blood loss and blood transfusion rate. Interestingly, there was no difference in operating time between the laparoscopic and open interventions, although the pooled estimate tends to favour the ODP group. This may reflect the fact that the current data were not stratified according to surgical experience, although stage on the learning curve has been shown to affect intraoperative blood loss, operation time and other intraoperative parameters.⁴ Moreover, patient-specific factors, such as localized fibrosis, inflammatory changes caused by the tumour and tumour infiltration can also prolong the duration of surgery.⁵

En bloc splenectomy is usually performed during conventional ODP. Some studies have advocated splenic preservation whenever possible as it has been found to be associated with a reduced incidence of perioperative infectious complications and severe complications, and a shorter hospital stay.^41–44 The present results illustrate that LDP favours splenic preservation. These findings are consistent with the higher rates of splenic preservation, of up to 85% in some series, reported after LDP,⁴⁵ which may be attributable to the improved vision afforded by the laparoscopic approach, which, in turn, facilitates the more accurate dissection of the pancreas from the splenic vessels and the splenic hilum.^31,32

Margin-negative resection is the only way of curing pancreatic cancer.⁴⁶ The success of any oncological operation is determined by the achievement of tumour-free margins and lymph node yield. Some studies in the past have suggested that pancreatic adenocarcinoma is a contraindication to laparoscopic resection^47,48 because the role and oncologic safety of laparoscopic resection in pancreatic cancer remain unknown. Recent studies^49,50 have shown that the laparoscopic approach to malignant pancreatic tumours is feasible and results in similar rates of morbidity and mortality as it does in benign tumours. Of the studies included in the present analyses, only a few studies^21,30 with small sample sizes reported this outcome and thus the available data are not sufficient to make a cumulative analysis. Furthermore, none of the studies provide data on longterm oncologic follow-up and therefore additional well-designed studies are needed to provide convincing evidence.

Laparoscopic surgery has the advantage of requiring smaller incisions and less bowel manipulation than does open surgery and thereby reduces pain and analgesic requirements, and facilitates the earlier recovery of bowel function and ambulation.^5,6 The results of this present meta-analysis were found to be consistent with published data in terms of indicating a shorter time to oral intake and shorter hospital stay in LDP patients than in ODP patients. When only high-quality studies were included in the analysis, time to first flatus was found to be greatly reduced in the LDP group. Additionally, a low rate of conversion to open surgery was observed, which highlights the feasibility of the laparoscopic approach.

The present analysis found that surgical site infections occurred less frequently in the LDP group compared with the ODP group. This may reflect the association between laparoscopic surgery and reduced surgical trauma, which results in a less acute phase response compared with open surgery, and the fact that local (peritoneal) immune function is affected by carbon dioxide as a result of pneumoperitoneum.⁵¹ The present analyses failed to demonstrate any differences between the LDP and ODP groups with regard to postoperative mortality, haemorrhage, reoperation rate, readmission rate, intra-abdominal fluid collections and intra-abdominal abscesses. It is of note that the sample sizes for reporting these complications were small; studies with larger sample sizes are needed to facilitate an accurate summary. It was not possible to undertake cumulative analyses to assess overall morbidity, as reported in many studies, because the criteria used to define such complications varied among the studies. Criteria for assessing morbidity included the DeOliveira scoring system, the Martin scoring system, the National Cancer Institute's common toxicity criteria and the Clavien classification system; it is inappropriate to pool results obtained using such varied systems in order to make a cumulative analysis.

Pancreatic fistula remains the most challenging complication in pancreatic surgery as it can lead to intra-abdominal abscess, delayed gastric emptying, haemorrhage, sepsis and electrolyte imbalances.⁵² Reported rates of pancreatic fistula in distal pancreatectomy vary between 23% and 26%.^53,54 In the present analysis, the pooled results show no significant difference in the rate of pancreatic fistula between the LDP (17%) and ODP (18%) groups and no difference in the rate of clinically significant fistula. It is important to note that although the majority of the reports included in the present analysis used the ISGPF definition of pancreatic fistula, variation exists in this regard and therefore, in order to achieve homogeneity and increase the reliability of the present results, only studies that used the ISGPF definition were included in the analysis.

This meta-analysis of non-randomized studies may have several limitations, which must be taken into account when considering the results. Firstly, all of the studies included were non-randomized in nature and therefore the results provide only a possible estimate. This remains the biggest limitation of this study and results in a weak level of evidence. However, the present study made a strong attempt to select the best evidence available by selecting high-quality studies and pooling their findings. This does not resolve the problem, but it does improve the quality of the synthesized data and adds reliability to the results. It also highlights the need for better designed randomized studies that are able to resolve all of the relevant questions. It is hoped that this will also provide guidelines for the design of future trials on the subject. Secondly, it is important to note that the results were not stratified according to whether the underlying pathology was benign or malignant in nature as the studies included patients with different types of pancreatic disease. This may have an effect on the outcome measures. Thirdly, significant heterogeneity was seen among the included studies for some outcome measures. This may well reflect differences in adjuvant treatment measures and medical insurance systems. However, investigation of this heterogeneity through meta-regression was not possible because of the small number of studies and the unavailability of relevant data. Additionally, it will be of crucial importance to ascertain the financial implications of these procedures in order to make recommendations for their specific indications.

Conclusions

In conclusion, this is the most comprehensive review to date to compare outcomes of LDP and ODP. The present results indicate that LDP is a safe and feasible technique in comparison with ODP. The current findings are reliable and the pooled estimates enable the resolution of some of the discrepancies in data among individual studies in the literature. However, these results need to be validated in large, well-designed randomized controlled trials.

Conflicts of interest

None declared.