Minimally invasive (MI) distal pancreatectomy has become the gold-standard [1,2,3,4,5], for the treatment of left-sided pancreatic tumors, either benign or malignant [4, 6]. Minimally invasive pancreatoduodenectomy (MIPD) is, however, still awaiting further studies before gaining similar implementation [7,8,9]. The first MIPD was reported in the early 1990s by Gagner et al. [10], only to be rapidly suspended due to limitations in the availability of laparoscopic technology at the time including lack of advanced energy source devices, adequate laparoscopic instruments, optimal vision, and laparoscopic ultrasound [10]. MIPD was later re-visited and optimized by a second wave of pioneers demonstrating its feasibility and safety [11,12,13,14,15], encouraging early adopters (EAs) to spread it further across the globe. This phase was marked by an assessment of MIPD compared to open pancreatoduodenectomy (OPD) via case-controlled studies, randomized controlled trials, and meta-analyses [7,8,9, 16,17,18].

Initially published in 2009 [19], the IDEAL framework stands for Idea (phase 1), Development (phase 2a), Exploration (phase 2b), Assessment (phase 3), and (4) Long-term study (phase 4). According to the IDEAL framework, Innovators fall in phase 1, Pioneers in phase 2 and EAs in phase 3. Using this paradigm, a recent study analyzed the experience of Pioneers of laparoscopic liver resection and compared their results with Early Adopters (EA) [19, 20]. However, it is still unknown if MIPD has already reached phase 4.

In 2007, the first minimally invasive (MI) surgical hepatic-pancreatic and biliary (HPB) fellowship (sanctioned by the International Hepato-Pancreato-Biliary Association) was created at the Institut Mutualiste Montsouris in Paris, France [21]. Unlike the pioneers, MI-HPB fellowship-trained (FT) surgeons had the benefit of learning from the trial and errors of the experts themselves. As opposed to their predecessors, this new generation of HPB surgeons may have a different approach to surgery of the pancreas. As opposed to their predecessors, they may feel equally or potentially more comfortable with MI approaches to the pancreas. As a result, we herein analyze and compare the overall results of both MI and OPD performed by a FT surgeon. The primary endpoint was to assess postoperative mortality and morbidity. Secondary endpoints included operative parameters, length of hospital stay, and survival analysis. Furthermore, we address any practice pattern changes for a HPB surgeon straight out of training with no previous experience of independent surgery.

Methods

A single surgeon (AG) reviewed his 10-year experience with MIPD and OPD, from December 2007-February 2018. Procedures started with MI techniques, including hand-assisted variants, were analyzed in the MI arm on an intention to treat basis. All resectable peri-ampullary tumors were preferentially considered for MIPD. IRB approval was not required, and informed written consent was obtained for all patients. Borderline resectable (BR) tumors as defined by the American Hepato-Pancreato-Biliary Association, Society of Surgical Oncology and Society of the Alimentary Tract [22], treated with neoadjuvant chemoradiation were not absolute contraindications to the MI approach.

However, patients with large tumors (> 5.5 cm) or deemed preoperatively to have a high likelihood of requiring vascular reconstruction were not approached minimally invasively. Previous upper abdominal surgery, obesity, and history of pancreatitis were not considered contraindications to the MI approach. However, patients who requested an open approach or with cancer who underwent double bypass followed by successful neoadjuvant treatments were considered for open resection [23,24,25]. A subgroup analysis on all patients who underwent PD for malignant peri-ampullary tumors (PT) was done and labeled as MIPD-PT and OPD-PT, respectively. PTs were defined as pancreatic ductal cancer (PDCs), distal cholangiocarcinomas (DCCs), ampullary cancers (ACs) and duodenal cancers (DCs). Patients with Klatskin’s tumors, pancreatic endocrine cancers or with distant metastatic disease were excluded.

The MI posterior approach to PD used in this study has been previously reported [12, 14, 26,27,28]. All procedures were done with the assistance of a robotically controlled laparoscope holder [14, 29]. Frozen section analysis was done on all patients and results were obtained prior to beginning the reconstruction. All margins were inked prior to histopathological examination. Pathological staging was done according to the 8th edition of the American Joint Committee on Cancer and all margins were inked prior to histopathological examination [30]. Complications were graded according to the Dindo–Clavien classification system [31]. All complications greater than Grade 2 were considered as major. Pancreatic fistulas and Delayed Gastric Emptying (DGE) were graded according the International Study Group of Pancreatic Surgery [32, 33].

Data analysis was performed using the Social Science Statistics software (www.socscistatistics.com) and the Prism 8: GraphPad software (https://www.graphpad.com/scientific-software/prism/) was used to generate Kaplan–Meier curves. Categorical data are expressed as percentages and quantitative data are presented as mean–standard deviation (SD) with median and range. Data was analyzed by using the Mann–Whitney U-test for continuous variables and the Fisher exact or Chi Square test for categorical values. Statistical significance was defined as p < 0.05. Survival curves were done for all cases by tumor type for overall quality control and then by technique (MIPD vs. OPD) when possible.

Results

Table 1 summarizes the preoperative characteristics of the patients. From December 2007 to February 2018, a total of 100 PDs (57 MI, 43 Open) were performed. The American Society of Anesthesiology (ASA) score was not different among the 2 groups, with most patients having an ASA class of 3. In the MIPD group, 43 patients (75.4%) underwent PD for malignant disease compared to 31 (72.1%) in the open group (p = 0.8). There were 39 patients in the MIPD-PT and OPD-PT groups, respectively. The age and sex ratio was similar across groups. More BR pancreatic ductal cancers (PDCs) were done via laparotomy compared to the MIPD-PT group, 32.0% (8 of 25 patients) compared to 8.0% (2 of 25 patients), but this did not achieve significance (p = 0.07). A total of 9 patients in the OPD-PT group compared to 2 in the MIPD-PT group underwent neoadjuvant therapy; however, this only tended to be significant (p = 0.06). Although results from the entire series and the PT group are presented in the Tables, the rest of the Results section will only focus on the PT groups below (MIPD-PT vs. OPD-PT).

Table 1 Preoperative characteristics
Full size table

Operative parameters are highlighted in Table 2. Operating time tended to be highly significantly longer in the MIPD-PT group, 456 compared to 371 min in the OPD-PT group, respectively (p < 0.001). Estimated blood loss was highly significantly less after MIPD-PT compared to OPD-PT, 345 mL compared to 971 mL, respectively (p < 0.001). Twelve patients (30.8%) who underwent MIPD-PT, had hand-assisted resection. Seven patients (17.9%) who underwent MIPD-PT required vascular repair as compared to 12 (38.7%) in the open cohort; however, this difference only approached significance (p = 0.06). The vascular repairs in 2 out of 7 BR patients who had MIPD were done via laparotomy after conversion. All patients in the OPD group underwent classic resections. However, 5 patients in the MIPD group underwent pylorus-preserving resections. Length of hospital stay averaged 12.0 days after MIPD-PT compared to 16.0 days after OPD-PT, p = 0.007.

Table 2 Perioperative results
Full size table

Two patients (5.1%) in the MIPD-PT group were converted to an open approach (Table 2). One conversion occurred due to portal vein invasion requiring partial resection with patch angioplasty. The second patient was converted because of superior mesenteric vein invasion that required resection and reconstruction (this was discussed in an earlier paper) [26]. Two patients (5.1%) required reoperation after MIPD-PT. The first patient required reoperative laparoscopic placement of a T-tube for bile leak. The second patient, re-operated on for a gastro-duodenal leak, eventually died on postoperative day #6. One patient (3.2%) required reoperation after OPD-PT. This patient had surgery for a DCC and bled from a pseudoaneurysm that was ligated in the operating room on operative day #7.

Table 3 summarizes the mortality and morbidity. The 30-day mortality rate was 2.6% after MIPD-PT and 0% after OPD-PT, (p = 1). As mentioned above, 1 patient died after MIPD-PT secondary to a gastro-duodenal anastomotic leak. Ninety-day mortality increased to 5.1% and 3.2% in the MIPD-PT and OPD-PT groups, respectively (p = 1). One patient in the MIPD-PT group suffered an anoxic brain injury during an endoscopy on postoperative day #34 and one patient in the open cohort died on postoperative day #77 due to portal vein thrombosis, this patient initially had BR disease, and was the only BR patient to suffer a mortality at either 30 or 90 days.

Table 3 Morbidity and mortality at 30 days
Full size table

Overall complication rates at 30 and 90 days were similar and occurred in 41.0% and 43.6% in the MIPD-PT group compared to 35.5% and 41.9% in the OPD-PT group, respectively (Tables 3 and 4). Pancreatic leak was noted at 30 and 90 days after 7.7% of patients in the MIPD-PT group compared to 6.5% and 9.7% in the OPD-PT group, respectively. Bile leaks were diagnosed in 2.6% after MIPD-PT compared to 0% after OPD-PT at both 30 and 90 days. Delayed gastric emptying (DGE) rates were also the same at 30 and 90 days and equaled 5.1% and 12.9% in the MIPD-PT and OPD-PT groups, respectively. None of the differences in complication rates were statistically significant.

Table 4 Morbidity and mortality at 90 days
Full size table

Pathology parameters are listed in Table 5. Tumor diameter averaged 2.2 cm in the MIPD-PT group compared to 2.7 cm in the OPD-PT group respectively, p = 0.3. The average number of lymph nodes harvested was 17 in the MIPD-PT group compared to 15 in the OPD group (p = 0.3). Final pathological stages including R0 resections were similar in both groups. When BR tumors were analyzed, only 1 of 10 patients had positive margins (R1 resection after OPD), the rest having all clear margins (R0). Pathological staging was listed for all cases of pancreatic ductal cancer (PDCs), distal cholangiocarcinomas (DCCs), ampullary cancers (ACs) and duodenal cancers (DCs). There were no statistical differences between MIPD and OPD by stage (p = 0.8). When AC were analyzed histologically, 3 patients had pancreatobiliary, 2 patients intestinal and 2 patients had mixed subtypes in the MIPD-PT group compared to 1 mixed subtype in the OPD-PT group.

Table 5 Pathological characteristics for PT = peri-ampullary tumors: pancreatic ductal cancer, distal cholangiocarcinoma, ampullary cancer and duodenal cancer: minimally invasive pancreatoduodenectomy (MIPD) vs. open pancreatoduodenectomy (OPD)
Full size table

When PDs for all malignant PDCs were combined, the 1, 3 and 5-year and median survivals were 61.7%, 32.8%, 25.5% and 23 months; when patients with node negative PDC R0 resections were analyzed survival increased to 84.2%, 54.7%, 41.1% and 37 months (Fig. 1). When all distal cholangiocarcinomas were analyzed, the 1, 3 and 5-year and median survivals were 82.5%, 49.5% and 33% and 35 months (Fig. 1). There weren’t enough ACs or DCs in either group to create separate curves so they were combined into peri-ampullary tumors (PT). Overall 1, 3 and 5-year survival rates for MIPD-PT vs. OPD-PT, and median survival were 82.5%, 59.6% and 46.3% and 38 months as compared to 52.5%, 15.7% and 10.5% and 13 months, respectively (p = 0.013) (Fig. 2). In the MIDP-PT group, recurrence free survival (RFS) at 1, 3 and 5 years and median RFS were 69.1%, 41.9% and 33.5% and 26 months as compared to 50.4%, 6.3% and 6.3% and 13 months, in the OPD-PT group, respectively (p = 0.026) (Fig. 3).

Fig. 1
figure 1

Overall survival (OS) of all pancreatoduodenectomies (minimally invasive + open), PDC pancreatic ductal cancers, R0 N0 negative margins and lymph node negative, DCC distal cholangiocarcinoma

Full size image
Fig. 2
figure 2

Overall survival (OS) for PT = peri-ampullary tumors (pancreatic ductal cancer, distal cholangiocarcinoma, ampullary cancer and duodenal cancer): MIPD minimally invasive pancreatoduodenectomy, OPD open pancreatoduodenectomy

Full size image
Fig. 3
figure 3

Recurrence Free Survival (OS) for PT = peri-ampullary tumors (pancreatic ductal cancer, distal cholangiocarcinoma, ampullary cancer and duodenal cancer): MIPD minimally invasive pancreatoduodenectomy, OPD open pancreatoduodenectomy

Full size image

Discussion

About 75.4% percent of MIPDs were done for cancer, indicating that even malignancies can be approached early in one’s experience. Although blood loss, transfusion rates and hospital stays tended to be lower after MIPD, surgical times were found to be longer, findings that are consistent with multiple reports including a meta-analysis from 2017 [17]. Interestingly one study analyzing outcomes after OPDs found a higher complication rate when operations took longer, a finding that does not seem to be observed after MIPD [34]. Although hand-assistance was used with decreasing frequency as experience was gained, it seems a useful adjunct during MIPD and may represent a potential advantage when compared to MIPD with complete robotic systems, where the surgeon at the console cannot physically use hand-assistance.

Some authors have criticized studies comparing MIPD and OPD by taking issue with perceived long operative times during OPD [8, 35, 36]. In particular, they question the quality of OPDs that take longer than 240 min. Interestingly, the largest series in the world, reporting on 2,000 PDs, reported that the average was 355 min in the 2000s down from almost 540 in the earlier part of their experience [37, 38]. Our operative times are within the longer range noted by the group from John’s Hopkins (Table 2). Notably, the group that argues that shorter operative times are an indication of operative quality report an R0 resection rate of only 80% after PD, this compares to our R0 rate of 97%. One wonders if a higher R0 resection rate warrants the increased operating time in our series.

In 2015 Sosa et al. [39] reported on national data of 6,078 patients from 2010 to 2011 who underwent OPD compared to 983 who underwent MIPD. Thirty-day mortality was found to be statistically significantly higher in the MIPD cohort at 5.1% compared to 3.1% in the open one [39]. However, it may be irrelevant to compare the preliminary experience of a new procedure with a 6 times larger series of a confirmed one. Furthermore, over the same time period, another group reported similar 30-day mortality rates of 5.2% and 3.7% that were not significantly different [40]. Mortality after PD for peri-ampullary tumors in the open literature ranges from 3 to 5%, which is comparable to our own rate (Table 4) [7, 41].

Advances in postoperative care have taken giant strides since the 1980s when the John’s Hopkins group reported a mortality rate of 24% for the first 41 cases and then 21% for the following 47 [42]. In fact, if the previous author’s incorporated these cases in their large series published in 1997 the mortality rate would have increased from 1.4% to 3.9% [43]. In our series of a FT surgeon’s initial experience, 30-day mortality rate was 2.6% in the MIPD-PT cohort compared to 0% in the OPD-PT one (p > 0.05) (Table 3). The 30-day mortality rate after OPD in the modern open literature ranges from 1.4% to 12.4%, including 1.4% to 5.9% in high-volume centers [38, 43,44,45]. Interestingly, excluding high-risk patients from analysis as recommended by some studies based on benchmarking would have lowered our MIPD-PT mortality rate to 0% [46].

A randomized controlled trial (LEOPARD-2) reported a 10% 90-day mortality rate after MIPD compared to 2% after OPD [9]. However, it is unlikely that any of the surgeons participating in the trial was FT before independent practice. The LEOPARD-2 trial was prematurely closed because of the higher mortality rate in the MIPD arm. Notably a recent national trial of 12,670 patients from France noted a 9.2% mortality rate at 90 days after OPD [47]. Mortality rates have often been a subject of criticism of MIPD. After MIPD 90-day mortality rate ranges from 0 to 10% [8, 9, 48]. Ninety-day mortality in this series was similar at 5.1% and 3.2%, after MIPD-PT and OPD-PT. In one of the largest studies to publish 90-day mortality rates after OPD, 1,463 cases from Holland were found to have a 90-day mortality rate of 6.3% [45]. In patients over 75, MIPD seems to have particular benefit with 90-day mortality decreased when compared to OPD in a 3 year national database of 1,768 patients, 6% to 10.4%, respectively [49].

Morbidity has always been significant ranging between 32.5 and 68.2% after OPD in high-volume centers [43, 50, 51]. In our series, morbidity was similar in both cohorts at 30 and 90 days at 41.0% and 43.6% and 35.5% and 41.9%, for the MIPD-PT and OPD-PT cohorts, respectively. A recent National Surgical Quality Improvement Program (NSQIP) analysis noted a pancreatic fistula rate of 9.1% or 23.6% after MIPD depending on the definition used. Our pancreatic fistula rates were 7.7% and 6.5% in the MIPD-PT and OPD-PT, respectively (Table 3) [52]. Two prospective studies addressed MIPD for peri-ampullary tumors, including one randomized from India and one non-randomized from France [7, 41]. The later found a Grade C fistula rate of 24% after MIPD and argued that the MI approach should only be done in cases with a low risk of pancreatic fistula. On the other hand, the group from India only found a grade C fistula rate of 3%.

Our relatively low pancreatic fistula rate is tempered by a higher biliary fistula rate, 7.0% and 4.7% in the overall MIPD and OPD groups, respectively (Table 3). Bile leaks have been reported in up to 6% in the OPD literature [52, 53]. It is possible that some of our bile leaks were in fact pancreatic leaks. Additionally, extensive dissection during lymph node harvesting may have compromised the vascularization of the common bile duct. Regardless of these hypotheses, our combined leak rate appears well within the range noted in the open literature, even when we look at centers testing the preventive effect of the newest somatostatin analog, Pasireotide [55, 54, 55]. Nevertheless, we can glean that pancreatic fistula rates still ranges between 13.3% and 32% after OPD in modern high-volume centers [54, 55]. Notably when we looked at just the MIPD-PT and OPD-PT groups, the biliary fistula rates decreased to 2.6% and 0%, respectively (Table 3). Our rate of delayed gastric emptying of 5.1% and 12.9% in the MI and open cohorts, respectively, is within the wide range of rates reported in the literature of up to 10–61% after open PD [33, 51, 53].

We initially started to use indocyanine green fluorescence-guided imaging for all pylorus-preserving resections and then broadened it for all PDs. Although it has been used in an effort to reduce bile duct injury and biliary fistula, we found this technology particularly useful to identify ischemia during creation of duodenojejunostomy or gastrojejunostomy [56], possibly avoiding some postoperative leaks.

Lymph node retrieval averaged 17 in the MIPD-PT group and 15 in the OPD-PT group, respectively. A recent large multi-center international trial reported an R1 rate of 27%, which is considerably higher than our overall rate of 7.1% (Table 5) [46]. When only MIPDs are analyzed the positive margin rate decreased to 2.6%. Published positive margin rates after MIPD range from 0 to 22.2% [15, 57,58,59]. The discrepancy between R0 rates among the MIPD and OPD literature can possibly be explained by the enhanced visualization with the MI approach.

In addition to improved R0 rates, another potential benefit of MIPD may be that time to adjuvant chemotherapy could be decreased when compared to OPD. Kendrick et al., noted a statistically significant difference favoring the MIPD group, with chemotherapy being started on postoperative day (POD) 48 compared to POD 59 after OPD [15]. This implies that the potential benefits of the MI approach are indeed multifactorial, with time to initiation of adjuvant chemotherapy and the ability to complete treatment as perhaps the main benefits [40, 60, 61]. In our series time to chemotherapy was not recorded. However, initiation of chemotherapy delayed by more than 90 days or not given at all was recorded. Although the MIPD-PT group had fewer patients whose chemotherapy was delayed in our series, 4.5% compared to 14.3% of patients in OPD-PT cohort, this difference was not statistically significant (Table 5). In the Mayo Clinics series, these rates were similar, but notably, this difference was found to be statistically significant [62].

Survival rates after PD for peri-ampullary tumors have improved over the last several decades. Peri-ampullary tumors encompass 4 different tumor types: pancreatic ductal cancer (PDC), distal cholangiocarcinoma (DCC), ampullary cancers (AC) and duodenal cancers (DC). Consequently, there is a wide disparity in reported survival rates. In the open literature, 1, 3 and 5-year overall survival (OS) for PD for PDC 64–88%, 17–33% and 10–24% and 70.0%, 10–27%, respectively, which is within the range when all of our PDs including when BR cases were combined (Fig. 1) [37, 38, 50, 63,64,65]. When DCCs are analyzed, our 5-year survival rate of 33% favorably compares to the range of 22% to 27% reported in the open literature (Fig. 1) [37, 38, 66]. Importantly, ACs and DCs have improved survival rates when compared to PDCs and DCCs, specifically 45–60% and 43–82.5%, respectively [38, 66]. As a result, when peri-ampullary tumors are taken as a whole, 5-year survival range from 17–70% in the open literature [38, 43, 66,67,68]. When MI and OPDs for peri-ampullary tumors (PT) were analyzed, overall 5-year survival was 46.3% compared to 10.5%, and this difference was statistically significant (p = 0.01) (Fig. 2). RFS rates were also higher when the MIPD-PT and OPD-PT groups were compared, and this difference was statistically significant (p = 0.03) (Fig. 3).

Notably, several limitations exist for this study. Importantly, we did not have enough ACs or DCs to do independent survival curves. Furthermore, the majority of ampullary cancers were removed minimally invasively; however, this difference did not reach significance (p = 0.2) (Table 1). As mentioned above, there were statistically more vascular resections in the overall OPD group (p = 0.02); however, when only the PTs were analyzed this difference only tended to be significant (p = 0.06). More BR patients and patients undergoing neoadjuvant treatments tended to be in the OPD-PT when compared to the MIPD-PT group; however, these tendencies did not reach statistical significance, p = 0.07 and 0.06, respectively.

The learning curve for MIPD is generally believed to be about 50 cases [69]. However, a group from China evaluated the learning curve for MIPD using both cumulative sum (CUSUM) and risk-adjusted CUSUM (RA-CUSUM) methods [20, 70]. They observed three phases of the learning curve, (1) the initial learning phase, (2) the technical competence phase and (3) the challenging period [70]. The first phase was defined as the first 17 cases, followed by the second phase from the 18th until the 38th with the final phase being cases after 38. A similar trend was noted by Speicher et al., who found a significant decrease in operative time after the first 10 cases, with a total of 50 cases being needed to fully overcome the learning curve [69].

Unlike the steady improvement noticed with laparoscopic liver resection, Wang et al. noticed a phenomenon in phase 3 where the conversion rate remained high [71]. They realized that they began to do more complex cases after approximately 40 MIPDs, specifically, larger tumors and more often malignant. Although we did not notice an increase in conversions, we did notice an increase in overall mortalities. Specifically, no mortalities were in the Initial period, but 1 each noted in the technical competency and challenging periods. This highlights the importance of continued vigilance even after the initial phase of the learning curve is overcome. Notably, the surgeon in the Chinese study had already done 30 OPDs and approximately 300 complex laparoscopic procedures prior to embarking on MIPD. Additionally, 22.8% of cases were done for pancreatic ductal cancer compared to 38.6% in our series.

Conclusions

Hopefully as happened with MI liver resection, difficulty scoring systems can be devised, which will help future MI-HPB surgeons better determine which tumors should be brought to the operating room for PD [72]. Unfortunately, there is still a wide variability between R0 rates after either MI or OPD of 51–100% even though an R0 resection is fundamental to a durable 5-year OS rate [7, 8, 37, 41, 70]. Additionally new devices like the Modified Frailty Index may help to limit the number of high-risk patients going to the operating room for PD [73]. National registries are being increasingly utilized that will better reveal unadulterated outcomes after pancreatic surgery and not only selected results [74]. Lastly, a large group of international experts recently convened in Miami and developed evidenced-based guidelines to MI pancreatectomy [75]. These added with increasing numbers of FT HPB surgeons fluent in minimally invasive surgery may enable increasing victories in the fight against peri-ampullary cancers.

As opposed to either Pioneers or Early Adopters of minimally invasive pancreatoduodenectomy, the minimally invasive hepatic-pancreatic and biliary fellowship-trained surgeon in this study opted to approach all cases that did not clearly need vascular reconstruction or have a contraindication to laparoscopy via minimally invasive techniques early in his experience even in cases of malignancy. Potential short-term benefits of minimally invasive pancreatoduodenectomy include decreased blood loss, fewer transfusions and shorter hospital stays at the expense of longer operative times. The superior R0 resection rates and decreased time to adjuvant chemotherapy may ultimately result in improved survival rates. It is believed that increasing numbers of MI-HPB Surgery or traditional HPB Fellowships with equal exposure to MIPDs should be created to allow new surgeons to get the skill set necessary to safely perform these technically demanding procedures.

Disclosures

Dr. Andrew A. Gumbs, Professor Elie Chouillard, Professor Mohamed Abu-Hilal, Professor Roland Croner, Professor Brice Gayet and Professor Michel Gayet have no financial ties or disclosures to report.