Minimally-invasive versus open pancreatoduodenectomies with vascular resection: A 1:1 propensity-matched comparison study.

Background: Minimally invasive pancreatic pancreatoduodenectomy (MIPD) is increasingly adopted worldwide and its potential advantages include reduced hospital stay and decrease pain. However, evidence supporting the role of MIPD for tumours requiring vascular reconstruction remains limited and requires further evaluation. This study aims to investigate the safety and efficacy of MIPD with vascular resection (MIPDV) by performing a 1:1 propensity-score matched (PSM) comparison with open pancreatoduodenectomy with vascular resection (OPDV) based on a single surgeon's experience.
Methods: This is a retrospective review of 41 patients who underwent PDV between 2011 and 2020 by a single surgeon. After PSM, the comparison was made between 13 MIPDV and 13 OPDV.
Results: Thirty-six patients underwent venous reconstruction (VR) only and 5 underwent arterial reconstruction of which 4 had concomitant VR. The types of VR included 22 wedge resections with primary repair, 8 segmental resections with primary anastomosis and 11 requiring interposition grafts. Post-operative pancreatic fistula (POPF) occurred in 3 (7.3%) patients. Major complications (>Grade 2) occurred in 16 (39%) patients, of which 7 were due to delayed gastric emptying requiring nasojejunal tube placement. There was 1 (2.4%) 30-day mortality (OPDV). Of the 13 MIPDV, there were 3 (23.1%) open conversions. PSM comparison demonstrated that MIPDV was associated with longer median operative time (720 min vs. 485 min (P = 0.018). There was no statistically significant difference in other key perioperative outcomes such as intra-operative blood loss, overall morbidity, major morbidity rate, POPF and length of stay.
Conclusion: Our initial experience with the adoption MIPDV has demonstrated it to be safe with comparable outcomes to OPDV despite the longer operation time.

INTRODUCTION

At present, there is increasing adoption of minimally invasive pancreatic surgeries (MIPS) internationally since it was first reported two decades ago.[1] The uptake of MIPS has been slow owing to its technical complexity. However, improved knowledge, sharing of experiences, standardisation of techniques[2] and advancements in technology used in MIPS have expanded the role and adoption of MIPS globally.[34] The indications for MIPS have hence increased from less complex procedures such as pancreatic enucleation and distal pancreatectomies (DP)[5] to now even more complicated ones such as central pancreatectomies and pancreatoduodenectomies (PD)[6] and increasingly, for pancreatectomies with vascular resections and reconstruction.[78910]

Despite evidence from numerous retrospective case–control studies as well as randomised trials in DP and PD[51112] demonstrating superior short-term operative outcomes, the conventional open approach for pancreatic surgery remains the most common approach adopted by pancreatic surgeons today.[11131415] The main concerns preventing further increased adoption of MIPS remains the long and steep learning curve required from surgeons to learn and attain proficiency.[1617] This is even more so when the conventional open approach is an already complex operation with high morbidity and complication rates as with MIPS. Robotic pancreatic surgery has been proposed as an alternative to traditional laparoscopic surgery, with reportedly increased advantages such as improved dexterity and stability potentially resulting in lower open conversion and post-operative pancreatic fistulas (POPF), especially during the early learning phase.[161819]

In this study, we aimed to evaluate our experience with minimally invasive pancreatoduodenectomy with vessel resection (MIPDV) versus open PD with vessel resection (OPDV) in 41 patients who underwent PD with vessel resection from 2011 to 2020.

METHODS

Retrospective review of 41 consecutive patients who underwent PDV by a single surgeon between 2011 and 2020 were identified from a prospectively maintained database. The need for patient consent for this study was waived by the Singhealth Institution Review Board.

Clinicopathological data including patient demographics and relevant pre-operative, intra-operative and post-operative outcomes were obtained retrospectively from patient records. Clinical data were collected from a prospective computerised clinical database (Sunrise Clinical Manager version 5.8, Eclipsys Corporation, Atlanta, Georgia) and patient clinical charts while operative data was obtained from another prospective computerised database (OTM 10, IBM, Armonk, New York, USA).

The decision for MIPDV versus OPDV was not based on a formal protocol. The indications for MIPDV gradually expanded over the study period as the surgeon gained increasing experience with MIPS. In this early experience, cases requiring vascular reconstruction via end-to-end anastomosis and interposition graft were only planned for the hybrid procedure and open reconstruction.

Definitions

Extended pancreatectomy was defined according to the 2014 ISPGS definition.[20] Totally minimally invasive surgery (MIS) cases were defined as cases whereby both resection and reconstruction were performed laparoscopically or robotically. Hybrid procedures were cases whereby the resection was completed via MIS but the reconstruction (pancreaticojejunostomy, hepaticojejunostomy, gastrojejunostomy and vascular anastomoses) was performed via an open midline incision. Any procedure whereby the resection phase could not be completed via MIS was considered an open conversion.

Post-operative complications were classified according to the Clavien-Dindo grading system[21] and recorded regardless of the length of postoperative stay or, if the patient was discharged and readmitted, within a 30-day duration. Pancreatic fistulas were defined and graded according to the latest 2016 International Society of Grading of Pancreatic fistula.[22] Drain fluid amylase was routinely collected on post-operative day 3 and a value more than 3 times the concentration of the upper limit of serum amylase or a value more than 300 IU/L associated with a clinically relevant condition was considered a clinically relevant pancreatic fistula. Grade B pancreatic fistula included patients who had surgical drains kept in placed for more than 3 weeks or required endoscopic or percutaneous placement of new drains. Grade C pancreatic fistulas were fistulas which required reoperation, lead to organ failure or resulted in mortality. Purely asymptomatic fistulas previously referred to as a grade-A fistula were not classified as pancreatic fistula or morbidity but considered a biochemical leak.

Thirty-day mortality was defined as any death within 30 days from surgery and in-hospital mortality was defined as any death during the index hospital stay regardless of time from surgery. Thirty-day readmissions were defined as any admission into the hospital occurring within 30 days of discharge.

Statistical analysis

Propensity score matching (PSM) analysis was performed to minimise selection bias. Factors included in the PSM model were age, gender, ASA status, pre-operative biliary drainage, benign/malignant pathology, tumour size, type of pancreatectomy (pancreatoduodenectomy versus total pancreatectomy), MIS approach (totally MIS, hybrid or open conversion), presence of arterial resection and type of venous reconstruction (VR). PSM was performed by nearest neighbor matching in a 1:1 ratio without replacement, and analysis was performed between minimally invasive (MIPDV) and open (OPDV) pancreatoduodenectomy with vascular resection (PDV). Univariate analyses were performed using the Mann-U-Whitney test or Chi-squared tests as appropriate and P < 0.05 was considered statistically significant. All statistical analyses were performed using RStudio version 1.2.1335 (RStudio Team, R Studio, Boston, MA, USA).

RESULTS

The baseline demographics and operative details are summarised in Table 1. A total of 41 PDV was performed during the study period. One patient in the OPDV arm requiring arterial resection received neoadjuvant chemo-radiotherapy. There were 13 (31.7%) MIPDV (6 robotic-assisted) and 28 (68.3%) OPDV. Of the MIPDV cases, 6 cases (46.1%) were performed totally MIS (4 robotic-assisted), 4 cases (30.8%) were performed using a hybrid approach (2 robotic assisted) and 3 cases required conversion to open during the resection phase. Most of the cases involved venous resection only with 11 (84.6%) MIS cases and 25 (89.3%) open cases. Five patients underwent arterial resection with concomitant VRs. The types of VRs in the 41 patients included 9 MIPDV cases and 13 OPDV with a wedge resection of the portal vein with primary closure. One MIPDV and 7 OPDV cases underwent segmental resection with primary end-to-end anastomosis. Interposition grafts were performed in 3 MIPDV and 8 OPDV. All 6 cases of totally MIS MIPDV had wedge venous resection with primary closure. Two-hybrid cases underwent wedge resection with primary closure and 2 underwent interposition graft. Of the 3 MIPDV with open conversions, 2 required arterial reconstruction and 1 required an interposition graft.

Table 1

Baseline characteristics of all patients and comparison between minimally-invasive pancreatoduodenectomy with vascular resection versus open pancreatoduodenectomy with vascular resection

	Total (n=41)	MIPDV (n=13)	OPDV (n=28)	P	MIPDV (n=13)	OPDV (PSM) (n=13)	P
Median age (IQR), years	67 (11)	65 (6)	68 (11.5)	0.465	65 (6)	65 (12)	0.679
Male	21 (51.2)	9 (69.2)	12 (42.9)	0.216	9 (69.2)	8 (61.5)	1
ASA score
1	0	0	0	0.128	0	0	0.687
2	30 (73.2)	7 (53.8)	23 (82.1)		7 (53.8)	9 (69.2)
3	11 (26.8)	6 (46.2)	5 (17.9)		6 (46.2)	4 (30.8)
Preoperative biliary drainage	21 (56.1)	6 (46.2)	15 (16.7)	0.673	6 (46.2)	9 (69.2)	0.427
Malignant tumour	39 (95.1)	11 (84.6)	28 (100)		11 (84.6)	13 (100)
Pancreatic adenocarcinoma	34 (82.9)	8 (61.5)	26 (92.9)	0.177	8 (61.5)	12 (92.3)	0.462
Others	5 (12.2)	3 (23.1)	2 (7.1)		3 (23.1)	1 (7.7)
Median tumour size (range)	3.5 (0–8.1)	3 (0–8.1)	3.8 (1.2–6)	0.232	3 (0–8.1)	3.5 (1.2–4.5)	0.504
Type of pancreatectomy
Pancreatoduodenectomy	32 (78)	9 (69.2)	23 (82.1)	0.600	9 (69.2)	10 (76.9)	1
Total pancreatectomy	9 (22)	4 (30.8)	5 (17.9)	4 (30.8)	3 (23.1)
MIS approach
Totally MIS	NA	6 (46.1)	NA	NA	6 (46.1)		NA
Hybrid		4 (30.8)			4 (30.8)
Converted open		3 (23.1)			3 (23.1)
Type of vascular resection
Venous only	36 (87.8)	11 (84.6)	25 (89.3)	1	11 (84.6)	10 (76.9)	1
Arterial only/arterial + venous	5 (12.2)	2 (15.4)	3 (10.7)		2 (15.4)	3 (23.1)
Type of venous reconstruction
Wedge/sleeve	22 (53.7)	9 (69.2)	13 (46.4)	0.313	9 (69.2)	8 (61.5)	0.239
Primary end-to-end/patch	8 (19.5)	1 (7.7)	7 (25)		1 (7.7)	4 (30.8)
Interposition graft	11 (26.8)	3 (23.1)	8 (28.6)		3 (23.1)	1 (7.7)

ASA: American society of anaesthesiology, IQR: Interquartile range, MIPDV: Minimally-invasive pancreatoduodenectomy with vascular resection, MIS: Minimally invasive surgery, OPDV: Open pancreatoduodenectomy with vascular resection, PSM: Propensity score matching

PSM was used to select for a matched number of OPDV cases. The proportion of the number of cases with VR for wedge resection and primary repair, primary end-end anastomosis with patch and interposition graft are 8 (61.5%), 4 (30.8%) and 1 (7.7%) respectively. Similarly, the proportion of cases for OPDV cases were subjected to PSM, 10 (76.9%) of cases had a PD whereas 3 (23.1%) had a total pancreatectomy. Most cases 10 (76.9%) involved venous resection.

Outcomes between minimally invasive pancreatoduodenectomy with vessel resection versus open pancreatoduodenectomy with vascular

The outcomes for patients undergoing MIPDV versus OPDV are summarised in Table 2. MIPDV patients had a statistically significant longer median operative time of 720 min whilst the median time for OPDV was 492.5 min (P = 0.005). This was similar after PSM with a median time of 485 min in propensity score-matched OPDV cases (P = 0.182). There was no statistically significant difference in median estimated blood loss and intra-operative blood transfusions. In terms of clinically significant post-operative complications classified, there was no statistically significant difference in Clavien Dindo grade 3–5 complications in the MIPDV versus OPDV groups. Likewise, patients in the MIPDV group had no significant difference in terms of median post-operative hospital stay, 30-day re-admissions, in-hospital and 30-day mortality. Five patients (13.9%) in the OPDV group had an unplanned reoperation for complications. These included 3 re-operations for post-pancreatectomy haemorrhage and 2 for vascular thromboses (1 venous thrombosis and 1 arterial thrombosis). Both patients with vascular thromboses underwent revision of the anastomoses (1 from primary venous anastomosis to interposition graft, 1 revision of the arterial anastomosis). In the MIPDV group, two patients (15.4%) needed reoperations, 1 for internal herniation of the biliopancreatic limb and the other for interposition venous graft thrombosis. Open intra-operative thrombectomy was performed followed by post-operative anticoagulation to manage the graft thrombosis.

Table 2

Perioperative outcomes of all patients and comparison between minimally-invasive pancreatoduodenectomy with vascular resection versus open pancreatoduodenectomy with vascular resection

	Total (n=41), n (%)	MIPDV (n=13), n (%)	OPDV (n=28), n (%)	P	MIPDV (n=13), n (%)	OPDV (n=13), n (%)	P
Median operation time (IQR), min	595 (266)	720 (75)	492.5 (167.5)	0.005	720 (75)	485 (160)	0.0182
Median estimated blood loss (IQR), mls	800 (1100)	700 (1350)	800 (625)	0.832	700 (1350)	600 (500)	0.7
Intra-operative blood transfusion	25 (61)	7 (53.8)	18 (64.3)	0.769	7 (53.8)	7 (53.8)	1
Median post-operative stay (IQR), days	17 (24)	24 (26)	16 (16.2)	0.354	24 (26)	10 (13)	0.158
30 days readmission	7 (17.1)	4 (30.8)	3 (10.7)	0.253	4 (30.8)	3 (23.1)	1
Post-operative complication (= all CD 1/2/3/4/5)	32 (78)	10 (76.9)	22 (78.6)	0.975	11 (84.6)	12 (92.3)	1
Major post-operative complication (CD 3/4/5)	16 (39)	5 (38.5)	11 (39.3)	0.960	4 (30.8)	7 (53.8)	0.427
Major complication excluding NJT for DGE	9 (22)	1 (7.7)	8 (28.6)	0.133	1 (7.7)	4 (30.7)	0.322
Clinically relevant pancreatic fistula
Grade B	2 (4.9)	0	2 (7.1)	0.834	0	0	NA
Grade C	1 (2.4)	0	1 (3.5)	1.000	0	0
Clinically relevant PPH
POPF related PPH	1 (2.4)	0	1 (3.5)	0.561	0	0	0.462
Other PPH	2 (4.9)	0	2 (7.1)	0	2 (15.4)
Vascular complications	4 (9.8)	2 (15.4)	2 (7.1)	0.793	2 (15.4)	1 (7.7)	1
Delayed gastric emptying	9 (22)	4 (30.8)	5 (17.9)	0.600	4 (30.8)	3 (23.1)	1
Reoperation	7 (17.1)	2 (15.4)	5 (17.9)	1	2 (15.4)	3 (23.1)	1
R1 resection	10 (24.4)	2 (15.4)	8 (28.6)	0.458	2 (15.4)	3 (23.1)	1
Mortality
30-days	1 (2.4)	0	1 (3.6)	1	0	1 (7.7)	1
In-hospital	2 (4.9)	0	2 (7.1)	0.834	0	1 (7.7)	1

DGE: Delayed gastric emptying, IQR: Interquartile range, MIPDV: Minimally-invasive pancreatoduodenectomy with vascular resection, NJT: Nasojejunal tube, OPDV: Open pancreatoduodenectomy with vascular resection, POPF: Post-operative pancreatic fistula, PPH: Postpancreatectomy haemorrhage, PSM: Propensity score matching

DISCUSSION

At present, evidence supporting the role of MIPD over OPD remains mixed. Four randomised control trials[13232425] have been published to date comparing laparoscopic PD (LPD) over OPD.[23] Three of these studies demonstrated some short-term advantages associated with LPD over OPD[132324] whereas 1 study, the LEPOARD-2 trial[25] was terminated early due to safety concerns arising from the increased mortality observed with LPD. However, these trials did not analyse patients who underwent MIPDV and data on the outcomes of MIPDV remain limited today.

In the present study, we have demonstrated that MIPDV can be adopted safely with no statistically significant difference in terms of key post-operative outcomes including morbidity and mortality compared to OPDV. Concordant with the literature, while MIPDV was associated with significantly longer median operative times as compared to OPDV,[26] in our experience this did not result in higher morbidity rates. An early series by Croome et al. from the Mayo Clinic[8] evaluated MIPDV compared to OPDV which demonstrated that MIPDV was associated with a lower intra-operative blood loss, shorter length of hospital stay and no significant difference in the total number of or severe complications and 30-day mortality compared to OPDV. Furthermore, the investigators[8] also found that the oncologic outcomes in terms of R0 resection was higher in the laparoscopic versus open group. This was attributed possibly due to better visualisation and magnification with laparoscopy allowing for more detailed lymphadenectomy and tumour resection.[8] A more recent study by Shyr et al demonstrated that robotic PD with vascular resection and the reconstruction resulted in acceptable vessel patency rates with good oncologic and clinical outcomes.[9] In the present study, although MIPDV was associated with longer median operative times, it is expected that the operative times would decrease with increasing experience as we progress along the learning curve.[27] Of note, there was no statistically significant difference in the estimated blood loss and intra-operative blood transfusion between MIPDV and OPDV. Overall, while our study did not demonstrate the benefit of the decreased intra-operative blood loss as reported by Croome et al., we surmised that this was likely due to this representing our initial experience whereby the surgeon was mounting the steep learning curve associated with MIPDV. Nonetheless, we have demonstrated that MIPDV can be adopted safely and is not associated with an increased risk of intra-operative bleeding or post-operative complications.

POPF is the most important and one of the most frequent complications after PD and remains the Achilles heel of the procedure. Construction of the pancreatic anastomoses is one of the major obstacles for surgeons to embark on MIPD. However, in the present study, there were no cases of clinically-significant POPF associated with MIPDV. This was consistent with the results by Wang et al.[24] which demonstrated that MIPD was not inferior to open surgery in terms of Grades B and C POPF. One of the major reasons for the low POPF rate is that MIPDV is usually performed in the setting of a firm pancreas with a dilated pancreatic duct resulting in a low risk of developing POPF.

More recently, some authors have proposed the use of robotic surgery for reconstruction after pancreatoduodenectomy[28] due to the theoretical superiority of the robot in terms of dexterity and increased degrees of freedom possible with robotic surgery.[18] This was supported by the study by Watkins et al.[28] which demonstrated that robotic reconstruction of the pancreatic anastomoses during the learning curve did not result in an increase in anastomotic complications such as POPF. The recent study by Kauffman et al.,[29] further demonstrated the feasibility of robotic-assisted PD with vascular resection and reconstruction. The authors reported low rates of open conversion and intra-operative blood transfusions. It has been postulated that with the better ergonomics and improved 3D visualisation from robotic surgery,[30] this would result in many advantages to the operating surgeon, especially for highly complicated procedures such as, robotic-assisted PDV resection. Nonetheless, it is important to highlight that MIPDV should be undertaken with caution only by surgeons with extensive prior experience with both open and MIS pancreatic surgery due to the complexity of the procedure including vascular reconstruction. Surgeons should have ideally already mounted the learning curves for standard MIPD. It is important to highlight that in this early experience, all 6 cases of totally MIPDV were performed for wedge resections with primary venous repair and all 7 MIPDV cases requiring venous and arterial anastomoses in this series were performed via the open technique either via the hybrid approach or after open conversion. Although previous authors in small case series (8–10) have reported the feasibility of performing vascular anastomoses via laparoscopic or robotic surgery, we elected to perform the vascular anastomoses via the open technique to ensure the safety of patients during this early experience.

CONCLUSION

Our initial experience with MIPDV demonstrated that it can be adopted safely by experienced surgeons with no significant difference in short-term perioperative outcomes compared to OPDV. Further studies preferably in the form of a multi-centre randomised control trial are needed to determine the role and potential advantages of MIPDV.

Financial support and sponsorship

Nil.

Conflicts of interest

There are no conflicts of interest.