Introduction

The adoption of minimally invasive pancreaticoduodenectomy (MIPD) has been cautious due to the complexity of the operation, the need to perform multiple delicate anastomoses, the concern for suboptimal oncological outcomes, and the high morbidity of the operation, even when performed in an open approach.1 Recently, multiple studies have suggested that MIPD is safe and feasible, especially in high-volume centers, with inferior outcomes in low-volume hospitals.2,3,4,5,6,7,8,9,10,11,12,13,14 Most of these studies were either a case series of laparoscopic or robotic operations or observational comparative studies between MIPD and open pancreaticoduodenectomy (OPD).2,3,4,5,6,7,8,9,10,11,12,13,14 The majority of these reports showed that MIPD has at least equivalent postoperative and oncological outcomes compared to OPD, but few studies have compared laparoscopic (LPD) versus robotic pancreaticoduodenectomy (RPD) to determine which minimally invasive platform may be best to adopt more broadly.14,15,16,17,18,19

Currently the surgery field is divided between centers supporting LPD or RPD without objective evidence of the superiority of one approach over the other. The advocates of LPD propose that laparoscopy is already engrained in surgical training and thus its adoption to another operation such as pancreaticoduoedenectomy (PD) should be easier than adopting the robotic platform, which is not usually used by residents or even fellows.6 The proponents of RPD suggest that the improved visual perception and ergonomics of the robotic platform allow for easier dissemination of this platform and possibly better outcomes, yet there are no large comparative studies evaluating the impact of the minimally invasive platform utilized on postoperative outcomes.20

The primary aim of this study was to compare RPD and LPD with respect to 30-day overall complication rates using the pancreas-targeted American College of Surgeons National Surgical Quality Improvement Program (ACS-NSQIP) database.

Materials and Methods

Study Design

We used the 2014–2015 pancreas-targeted ACS-NSQIP database to perform a retrospective study comparing robotic and laparoscopic cases. The NSQIP program collects more than 150 variables from 500 participating hospitals, including preoperative, intraoperative, and 30-day postoperative mortality and morbidity outcomes.21 The pancreas-targeted component has an additional 26 variables specific to pancreatectomy in comparison to the general NSQIP database and is only available at 120 hospitals. The ACS-NSQIP database is maintained by trained and certified surgical clinical reviewers who collect and enter the data, and the web-based database is audited periodically to ensure the highest quality.21

Patient Selection

After merging the pancreas-targeted NSQIP participant user data files with the general database, we selected the following current procedural terminology (CPT) codes: 48150, 48152 (classic Whipple-type procedure with and without pancreatojejunostomy), 48153, and 48154 (pylorus-preserving PD [PPPD] with and without pancreatojejunostomy). The following patients were excluded (Fig. 1):

Fig. 1
figure 1

CONSORT diagram

Full size image

Patients who

  1. 1.

    Underwent a nonelective procedure;

  2. 2.

    Underwent a hybrid procedure. The NSQIP defines a hybrid procedure as “a combination of approaches not otherwise specified,” making it unclear how these operations were performed;

  3. 3.

    Had missing data.

The diagnosis group was divided into pancreatitis, T0–T2 malignant, T3–T4 malignant, ≤ 5 cm benign, and > 5 cm benign lesions. We defined multivisceral resection as a colonic, hepatic, and/or intestinal resection performed with MIPD. MIPD was defined as LPD or RPD.

Outcomes

The primary outcome of this study was the 30-day overall complication rate, and the secondary outcome was the conversion rate. A patient who had any of the following was considered to have a major complication: pneumonia, unplanned intubation, pulmonary embolism, on ventilator for > 48 h, deep surgical site infection, organ space surgical site infection, dehiscence, bleeding requiring transfusion within the first 72 h of surgery start time, deep vein thrombosis/thrombophlebitis, cerebrovascular accident, cardiac arrest, myocardial infarction, sepsis/septic shock, renal failure, or postoperative pancreatic fistula.

Statistical Analysis

We used SPSS version 24 to perform all statistical analysis. Categorical variables were presented as counts and proportions, and continuous variables were presented as means with standard deviations or medians. We performed a t test or univariate logistic regression for continuous variables and a chi-squared or Fischer’s exact or univariate logistic regression test, when appropriate, for categorical variables. To determine the association of the type of minimally invasive technique with overall complications, we adjusted for pre- and intraoperative factors using forward multivariable logistic regression. As a secondary analysis, we determined the predictors of conversion by using a forward multivariable logistic regression. Variables with p < 0.25 on univariate analysis were explored in both models, and only statistically significant variables were kept in the final models. Two-tailed tests were used with the significance level set at < 0.05.

Results

Baseline Characteristics of RPD and LPD

We identified 428 MIPD cases, of which 235 (55%) were performed laparoscopically and 193 (45%) were performed robotically (Fig. 1). Patients who underwent RPD were more likely to be white compared to those who underwent LPD (88.6 vs 78.7%; p = 0.024), less likely to have dyspnea (2.1 vs 6.0%; p = 0.046), and less likely to undergo preoperative radiotherapy (2.6 vs 9.4%; p = 0.004). In addition, RPD patients were less likely to undergo vascular resection (12.4 vs 23.4%; p = 0.004) and multivisceral resection (4.7 vs 12.3%; p = 0.005) but were more likely to have drains placed (99.0 vs 92.8%; p = 0.002). All other baseline characteristics were similar (Table 1).

Table 1 Patient, tumor, and operative characteristics
Full size table

Perioperative Outcomes of RPD and LPD

An unadjusted comparison showed that RPD was associated with a lower conversion rate compared to LPD (11.4 vs 26.0%; p = 0.004; Table 2). There was no difference in operative time, reoperation rate, length of stay, 30-day mortality, and overall and major complication rates. RPD was associated with increased superficial surgical site infections (9.3 vs 3.8%; p = 0.020), but there was no difference in all other complications, including postoperative pancreatic fistula and delayed gastric emptying.

Table 2 Unadjusted perioperative outcomes of robotic and laparoscopic cases
Full size table

On multivariable analysis, we found that the type of minimally invasive approach, whether laparoscopic or robotic, was not associated with overall complications (Table 3). The predictors of overall complication were higher body mass index (odds ratio [OR], 1.05; 95% confidence interval [CI], 1.02–1.09), vascular resection (OR, 2.10; 95% CI, 1.23–3.58), and operative time (OR, 1.002; 95% CI, 1.001–1.004).

Table 3 Multivariable analysis determining predictors of overall complication
Full size table

Conversion of MIPD: Predictors

The overall conversion rate for MIPD was 19.4% (Table 4). The independent predictors of conversion were dyspnea (OR, 4.56; 95% CI, 1.63–12.74), PPPD (OR, 2.42; 95% CI, 1.33–4.39), multivisceral resection (OR, 2.86; 95% CI, 1.32–6.23), and vascular resection (OR, 5.30; 95% CI, 2.97–9.45). After adjusting for these factors, robotic surgery was independently associated with a lower odds of conversion (OR, 0.46; 95% CI, 0.26–0.81).

Table 4 Multivariable analysis determining predictors of conversion of minimally invasive pancreaticoduodenectomy
Full size table

Discussion

In this large study from a national cohort of patients, we found that robotic surgery was associated with a similar 30-day overall complication rate in an intention-to-treat comparison to LPD. Furthermore, we found that RPD was associated with a lower conversion rate. The predictors of conversion were dyspnea, PPPD, laparoscopic approach, and multivisceral and vascular resection. This report provides evidence that both approaches seem to have similar 30-day outcomes, with RPD having the advantage of a lower conversion rate.

Robotic surgery is thought to be superior to laparoscopy in different disciplines, including urological,22 gynecological,23 colorectal,24 gastric,25 and distal pancreatic operations,26 but none have compared RPD to LPD in a large study,19 as few institutions perform this complex operation using minimally invasive techniques, and the majority have adopted one approach versus the other, preventing single institutional comparison. The robotic platform provides a magnified three-dimensional image, 7 degrees of freedom, and eliminates hand tremor and the fulcrum effect of rigid laparoscopic instruments—allowing for precise suturing, easier tissue handling, better control of large blood vessels, and the ability to work at angles not possible with the laparoscope. Such advantages are important, especially in a complex operation such as PD due to the need to perform intracorporal anastomoses such as the pancreatojejunostomy and hepaticojejunostomy. In this report, we have shown that there was no difference between the approaches in adjusted overall complications; however, we could not determine the severity of complications as described by the Clavien-Dindo classification as the NSQIP database lacks this variable. The robotic platform was superior to laparoscopy with respect to conversion, even after adjusting for important factors associated with conversion like concomitant vascular and multivisceral resection.

Identifying risk factors for conversion may help the surgeon to better select patients, reduce conversion, and to better counsel patients on the risks of such an event. In this study, we identified dyspnea, PPPD, laparoscopic approach, and multivisceral and vascular resection as important predictors of conversion. The advantage of using a national database is the ability to determine the adjusted odds of conversion for the different risk factors due to the presence of a relatively large number of conversions; however, we could not identify the exact causes for conversion in this dataset, such as bleeding, as that level of detail is not captured in NSQIP.27 Several factors may be involved in conversion. Vascular and multivisceral resections can pose challenges when completing the operation in a minimally invasive fashion and are usually associated with locally advanced disease, leading to a difficult dissection. Having pulmonary disease has been previously associated with a higher conversion in other minimally invasive operations and it may correlate with poor pulmonary reserve, leading to an inability to tolerate pneumoperitoneum.28,29 Interestingly, PPPD was associated with a higher conversion rate, which may be due to the need for a more challenging gastrojejunal anastomosis. The use of a robot may facilitate the performance of the complex reconstructions needed in PD and the control of bleeding due to improved ergonomics, dexterity, and better visualization in comparison to laparoscopy, explaining the lower rate of conversion.

This study was limited by its retrospective design and its relatively small sample size, preventing a more robust analysis, such as propensity matching, which would allow us to adjust for all known and clinically important confounders and subsequently let us compare all outcomes between both groups. Instead we chose two important outcomes—30-day overall complication and conversion rates—and determined their independent association with the robotic or laparoscopic surgical approach using logistic regression and adjusting for other important factors. Thus, caution should be taken when comparing both approaches with regard to the other unadjusted outcomes, as the groups were imbalanced at baseline. In addition, the NSQIP database lacks information on surgeon and hospital volumes of pancreatic surgery in general, and MIPD specifically, and therefore we could not determine where the surgeons performing these operations were in their learning curves. This is important, as MIPD is associated with a steep learning curve and requires performing at least 80 cases to reduce operative time and 20 cases to reduce conversion.30 Finally, we could not assess the oncological outcomes between both operations as the NSQIP database does not provide pathological variables or survival data; therefore, such an analysis would be best performed from other cancer-targeted prospective databases.

Conclusions

RPD was associated with a lower rate of conversion than LPD and a similar 30-day overall complication rate. This report suggests that both minimally invasive approaches have similar outcomes with regard to postoperative complications. Prospective evaluation is needed to corroborate these findings. Adopting one approach versus the other should be based on objective data, and more importantly, on the experience of the surgeon with the minimally invasive platform used.