INTRODUCTION
Minimally invasive techniques in pancreatic surgery continue to advance, and despite the recognized long learning curve, the indications for minimally invasive pancreatic surgery are increasing [1]. Safety and feasibility of minimally invasive techniques have been demonstrated in total pancreatectomy [2,3], distal pancreatectomy with or without splenectomy [4,5], and pancreatoduodenectomy [6]. The potential advantages of the minimally invasive techniques include reduced length of hospital stay and intraoperative blood loss, without comprising morbidity and oncological outcomes [2-5].
Central pancreatectomy is a parenchyma‐sparing technique which has become a popular alternative to distal pancreatectomy or pancreatoduodenectomy for benign or low-grade tumors located at the neck and proximal body of the pancreas [6]. Central pancreatectomy can reduce the risk of postoperative pancreatic endocrine and exocrine dysfunction, because it preserves pancreas parenchyma, avoids the morbidity associated with biliary and gastric anastomoses in pancreatoduodenectomy, and avoids the need for splenectomy [7-9]. However, central pancreatectomy is technically challenging, and may increase the risk of postoperative pancreatic fistula, due to the two cut surfaces of the pancreas after segmental resection [7].
Although systematic reviews of single-arm studies have demonstrated the safety and feasibility of minimally invasive and open central pancreatectomies [6,10], they have yet to be evaluated in a systematic review with a comparative meta-analysis model. Whether minimally invasive central pancreatectomy confers advantage over the open approach remains poorly understood. Consequently, we aimed to conduct a systematic review and meta-analysis to directly compare procedural outcomes of minimally invasive and open central pancreatectomy.
MATERIALS AND METHODS
We followed the standards and recommendations highlighted by the Preferred Reporting Items for Systematic reviews and Meta-Analyses (PRISMA) statement standards to develop the protocol, and to complete the study [11]. Ethical approval and consent from the included patients were not required, as the study design did not involve direct human or animal participation.
Eligibility criteria
Study design: all comparative randomized and non-randomized studies were considered for inclusion.
Population: all adult participants who underwent central pancreatectomy due to benign or malignant lesions in the neck or proximal body of the pancreas were considered eligible for inclusion. The list of pathologies of interest was not exhaustive, and included solid pseudopapillary neoplasm, intraductal papillary mucinous neoplasm (IPMN), neuroendocrine tumor, mucinous cystic neoplasm, serous cystic neoplasm, pancreatic ductal adenocarcinoma, and metastatic lesions.
Intervention of interest and comparison: minimally invasive (laparoscopic or robotic) central pancreatectomy was considered the studied intervention, while open central pancreatectomy was the studied comparison.
Outcomes: the outcomes of interest were operative time, intraoperative blood loss, need for blood transfusion, Clavien–Dindo ≥ 3 complications, grade B−C postoperative pancreatic fistula, postoperative mortality, and length of hospital stay.
Search strategy
Two separate investigators who had expertise in evidence synthesis designed a strategy for search considering relevant operators, search limits, thesaurus headings, and keywords (Appendix 1). The date that the last search was run was March 10, 2024, and the search was not restricted to a specific language. The other sources used to identify potentially eligible articles were reference lists of relevant original studies and systematic reviews.
Article screening, article selection, and data collection
Two separate investigators reviewed (through title and abstract) the articles found by the above search, and obtained the full texts that were judged to be relevant. The eligible articles were reviewed, and were pilot-tested to develop an electronic data collection sheet that contained the following information: type of minimally invasive approach, description of the included population, age, gender, body mass index (BMI), lesion size, pathology of resected lesion, outcomes, bibliographic information, study design, and sample size. Disagreements in study selection and data extraction were resolved by discussion or input from a third author.
Evaluation of methodological quality and bias
Two separate authors with expertise in evidence synthesis judged the methodological quality of randomized controlled trials using the Cochrane risk of bias tool [12], and observational studies using the Risk of Bias in Non-Randomized Studies of Interventions (ROBINS−I) tool [13]. Disagreements in methodological quality assessment were resolved by discussion or input from a third author.
Outcome synthesis and statistical analysis
Review Manager 5.4 software (Cochrane Collaboration) was used for the meta-analysis. The odds ratio (OR) and mean difference (MD) were calculated as summary measures for dichotomous and continuous outcomes, respectively. We computed the risk difference (RD) for dichotomous baseline characteristics and for the dichotomous outcomes in the case of no event in both groups in more than a third of the included studies. Random effects modeling using intention to treat data and individual patient as unit of analysis was used for analyses, and forest plots with 95% confidence intervals (CIs) were constructed to present the results. The statistical heterogeneity was quantified as I2 using Cochran Q test (χ2) ranging 0% to 100%. It was classified as low at 0%−24%, moderate at 25%−74%, and high at 75%−100%. Publication bias related to each outcome was planned to be determined using funnel plots, provided a minimum of 10 studies reported the outcome.
Additional analyses
When an outcome was reported by at least four studies, sensitivity analysis was performed: 1) Studies with low likelihood of bias were analyzed separately; 2) The effect of each study on the pooled outcomes was investigated through leave-one-out analysis.
Certainty of evidence
The GRADE system was followed to evaluate the certainty of evidence for each outcome [14].
RESULTS
Electronic search results
The search resulted in 1,447 articles; of these, 1,438 articles were not relevant, and were excluded directly. Further review of the remaining nine articles through their full text led to the exclusion of two more articles, because they were review articles (Fig. 1). Consequently, seven comparative studies including a total of 289 patients were included [15-21]. Table 1 details the included studies.
Baseline characteristics of the included patients
Among the included 289 patients, 147 patients underwent minimally invasive central pancreatectomy, while 142 underwent open operation. The included populations in both groups were similar in age (MD: −0.09 year, 95% CI: −4.00, 3.82, p = 0.96), male sex (RD: −0.06, p = 0.30), BMI (MD: 0.54 kg/m2, p = 0.41), and lesion size (MD: −0.31 cm, p = 0.14) (Fig. 2).
Pathology of resected lesions
The pathology of the resected lesions was similar between the two groups (Fig. 3): IPMN (RD: 0.01, 95% CI: −0.07, 0.09, p = 0.85); neuroendocrine tumor (RD: −0.05, 95% CI: −0.14, 0.03, p = 0.24); solid pseudopapillary neoplasm (RD: −0.03, 95% CI: −0.13, 0.07, p = 0.52); mucinous cystic neoplasm (RD: 0.02, 95% CI: −0.07, 0.11, p = 0.66); serous cystic neoplasm (RD: −0.00, 95% CI: −0.04, 0.03, p = 0.91); other pathology (RD: 0.02, 95% CI: −0.05, 0.09, p = 0.55).
Evaluation of methodological quality and bias
Supplementary Fig. 1 shows the outcomes of methodological quality and bias evaluation based on the tools mentioned above.
Outcomes
Operative time
Data related to the 289 patients from seven studies were pooled, which showed no difference in operative time between minimally invasive and open approaches (MD: 60.17 minutes, 95% CI: −57.03 to 177.38, p = 0.31) (Fig. 4). The heterogeneity was high (I2 = 99%, p < 0.00001), while the GRADE certainty was moderate (Supplementary Table 1).
Intraoperative blood loss
Data related to the 289 patients from seven studies were pooled, which showed the minimally invasive approach resulted in less intraoperative blood loss, compared with the open approach (MD: −153.13 mL, 95% CI: −238.33 to −67.93, p = 0.0004) (Fig. 4). The heterogeneity was high (I2 = 90%, p < 0.00001), while the GRADE certainty was moderate (Supplementary Table 1).
Need for blood transfusion
Data related to 218 participants from five studies were pooled, showing no difference in need for blood transfusion between the minimally invasive and open groups (OR: 0.30, 95% CI: 0.09 to 1.03, p = 0.06) (Fig. 4). The heterogeneity was low (I2 = 0%, p = 0.59), while the GRADE certainty was moderate (Supplementary Table 1).
Clavien–Dindo ≥ 3 complications
Data related to 258 participants from six studies were pooled, showing no difference in the risk of Clavien–Dindo ≥ 3 complications between the minimally invasive and open groups (OR: 1.11, 95% CI: 0.52 to 2.37, p = 0.78) (Fig. 4). The heterogeneity was low (I2 = 0%, p = 0.56), while the GRADE certainty was high (Supplementary Table 1).
Grade B−C postoperative pancreatic fistula
Data related to the 289 patients from seven studies were pooled, which showed the minimally invasive approach resulted in lower risk of grade B−C postoperative pancreatic fistula, compared with the open approach (OR: 0.54, 95% CI: 0.31, 0.94, p = 0.03) (Fig. 4). The heterogeneity was low (I2 = 0%, p = 0.77), while the GRADE certainty was moderate (Supplementary Table 1).
Postoperative mortality
Data related to the 289 patients from seven studies were pooled, showing no difference in the risk of postoperative mortality between the minimally invasive and open groups (RD: −0.00, 95% CI: −0.03 to 0.03, p = 0.81) (Fig. 4). The heterogeneity was low (I2 = 0%, p = 0.93), while the GRADE certainty was high (Supplementary Table 1).
Length of hospital stay
Data related to the 289 patients from seven studies were pooled, showing no difference in the length of stay in hospital between the two approaches (MD: −3.77 days, 95% CI: −7.97 to 0.43, p = 0.08) (Fig. 4). The heterogeneity was moderate (I2 = 73%, p = 0.001), while the GRADE certainty was high (Supplementary Table 1).
Sensitivity analyses
The results remained consistent through sensitivity analyses for most of the outcomes, except postoperative pancreatic fistula, and the need for blood transfusion. Removal of individual studies from the analyses shifted the pooled effect in favor of the open technique for postoperative pancreatic fistula, and in favor of minimally invasive technique for need for blood transfusion. However, separate analyses based on methodological quality and risk of bias did not shift the pooled effects.
DISCUSSION
We compared the procedural outcomes of minimally invasive and open central pancreatectomy in this meta-analysis, which analyzed the data related to 289 patients from seven studies. The analyses showed that minimally invasive central pancreatectomy may reduce the risks of intraoperative bleeding (moderate certainty) and postoperative pancreatic fistula (moderate certainty), compared with the open approach; however, the two approaches were similar in terms of operative time (moderate certainty), need for blood transfusion (moderate certainty), Clavien–Dindo ≥ 3 complications (high certainty), postoperative mortality (high certainty), and length of hospital stay (high certainty).
While open and minimally invasive techniques in performing central pancreatectomy have not previously been assessed in a comparative meta-analysis, our results may be compared with the findings of previous meta-analyses evaluating the outcomes of other types of pancreatectomy. In a meta-analysis of 4,275 patients undergoing total pancreatectomy, Chen et al. [2] concluded that minimally invasive total pancreatectomy may result in lower risks of major morbidity, intraoperative blood loss, and the need for intraoperative transfusion, compared with open total pancreatectomy. Wei et al. [3] demonstrated similar results in the setting of total pancreatectomy. In a meta-analysis of 4,346 patients undergoing distal pancreatectomy, Cucchetti et al. [4] concluded that laparoscopic distal pancreatectomy may result in shorter length of hospital stay and less intraoperative blood loss, compared with open distal pancreatectomy.
While there was no systematic review with comparative meta-analysis model on the outcomes of open and minimally invasive central pancreatectomy prior to the current study, Farrarons et al. [6] and Rompianesi et al. [10] evaluated the outcomes of each approach using a proportional meta-analytical model of single-arm studies. Farrarons et al. [6] concluded that minimally invasive central pancreatectomy is safe in selected patients, and in experienced hands. Moreover, Rompianesi et al. [10] concluded that robotic central pancreatectomy is safe, and associated with low perioperative mortality. Consistent with the findings of Farrarons et al. [6] and Rompianesi et al. [10], the findings of the current study support the safety and feasibility of minimally invasive central pancreatectomy. Regarding the benefits of the minimally invasive approach over the open approach, although our results showed lower risks of intraoperative bleeding and postoperative pancreatic fistula in the minimally invasive group, we should highlight that less intraoperative bleeding did not translate into less need for blood transfusion, and our finding about postoperative pancreatic fistula did not remain consistent through sensitivity analyses. Consequently, whether minimally invasive central pancreatectomy confers advantage over the open approach remains the subject of debate. We are mindful about the risk of type 2 error, hence adequately powered studies are required for definite conclusions; future studies may use our data for power analysis.
Among the included studies, the technique used for pancreas-enteric anastomosis was pancreaticojejunostomy in the laparoscopic arms, and pancreaticojejunostomy or pancreatogastrostomy in the robotic or open arms. We believe that the choice of surgical approach and technique depends on various factors that include the surgeon’s experience and preference, characteristics of the patient and tumor, and available resources. Although a robotic approach would facilitate the fine dexterity and motor skills required for pancreas-enteric anastomosis reconstruction, it may be associated with disadvantages, such as longer operation time, and inability to provide an immediate response to unexpected intraoperative events. On the other hand, a laparoscopic approach would allow better tactile feedback and faster response with each hand movement; however, the reconstruction phase may be challenging. Consequently, a hybrid laparoscopic–robotic approach combining laparoscopic resection and robotic reconstruction may be ideal.
The current study has inherent limitations. Six studies had non-randomized design, which would subject their results to inevitable selection bias. The effect of location of tumor (neck versus proximal body) on outcomes could not be evaluated based on the available data, and confounding by indication cannot be excluded. Moreover, the included studies had relatively small sample sizes, increasing the likelihood of type 2 error. The sensitivity analyses suggested inconsistent results for postoperative pancreatic fistula and the need for blood transfusion. The risk of publication bias could not be assessed formally, due to the number of included studies. Taking the limitations into account, the current study has some strengths. The included studies were homogenous in terms of baseline characteristics of the included studies; this was highlighted by detailed meta-analysis of the baseline characteristics, which suggested low risk of confounding bias. We evaluated the certainty of the evidence using the GRADE system, and downgraded the evidence accordingly. The moderate to high certainty for the outcomes suggests that our results are robust enough to be used for hypothesis synthesis in future research.