Optimal Pancreatic Surgery: Are We Making Progress in North America?

Joal D. Beane, MD; Jeffrey D. Borrebach, MS; Amer H. Zureikat, MD; E. Molly Kilbane, RN; Vanessa M. Thompson, PhD; Henry A. Pitt, MD


Annals of Surgery. 2021;274(4):e355-e363. 

In This Article


In this study of the 2013 to 2017 ACS-NSQIP pancreatic surgery database trends over time were analyzed for PD and DP. Patient, process, procedure, and 30-day outcomes were evaluated. Optimal pancreatic surgery was defined as no mortality, serious morbidity, postoperative percutaneous drainage, or reoperation while simultaneously achieving an acceptable LOS with no readmissions. For both PD (P = 0.002) and DP (P < 0.001), optimal pancreatic surgery increased over time as preoperative, operative, and postoperative management evolved.

Changes in the perioperative and operative management of patients undergoing PD and DP were observed over the study period. These changes in processes were the result of a growing body of evidence to guide the care of patients who undergo pancreatectomy. One of the most dramatic changes included the increased use of neoadjuvant therapy (NAT) among patients with pancreatic cancer undergoing PD. In addition, increased utilization of biliary stents for patients undergoing PD over this same time period was observed and, most likely, related to the increased utilization of NAT. The survival benefits of NAT in patients with pancreatic cancer are well described.[17–20] The driver of this increase in the use of NAT is likely multifactorial. In addition to the increase in the number of publications supporting the use of NAT, a greater percentage of patients with pancreatic cancer also are being treated at high-volume centers.[21] The increased utilization of NAT in our study may be a result of the centralization of care in patients with pancreatic cancer and increased access to multidisciplinary oncology teams.

Another trend that emerged was the use of minimally invasive surgery for patients undergoing pancreatectomy. We found no significant increase in the percentage of patients who underwent minimally invasive PD, but the percentage of patients who underwent RPD statistically increased while the percentage of patients who underwent laparoscopic PD statistically decreased. Whether these trends are a result of the same surgeons changing their operative approach or the result of surgeons abandoning laparoscopy while others are adopting RPD is unknown. Unlike LPD, a validated curriculum for RPD has been developed and is being disseminated.[22] Recent studies have reported that the learning curve of RPD is surmountable and perhaps less steep compared with the laparoscopic approach.[23,24] In addition, conversion to open and the negative impact of conversion is reduced in patients who undergo RPD compared with LPD.[25,26] In contrast to PD, nearly half of all DPs in this analysis were performed using a minimally invasive approach. Almost 80% of DPs were performed using traditional laparoscopy, whereas the robotic platform was only used in 20%. However, like RPD, a statistically significant increase in the percent of DP patients who underwent a robotic approach was observed over time.

Surgeons also have changed the postoperative management of intraoperatively placed pancreatic drains. During the study period, intraoperatively placed drains were used in nearly 90% of patients, and this practice remained constant over time. A management approach that has been shown to reduce the incidence of postoperative pancreatic fistula and abdominal complications is the early removal of surgically placed drains when guided by the POD 1 DFA-1.[27–31] Over the study period, an increasing percentage of both PD and DP patients had a DFA-1 analyzed as part of their drain management strategy (P < 0.001). In addition, the percentage of patients who had their surgically placed drains removed by POD 3 increased over the time for both PD and DP (P < 0.001). However, in patients with DFA-1 less than 5000 IU/L, the percentage of patients who had their drains removed by POD 3 was only 27% and 30% for PD and DP, respectively, and these rates did not change significantly over time. Thus, more opportunities may exist to increase the percentage of patients having early drain removal as thresholds for DFA-1 become more standardized.

Significant strides have been made in reducing mortality over the past 3 decades in patients undergoing pancreatectomy.[1] For both PD and DP, 30-day mortality was low (1.7% and 0.6%), and on risk-adjusted analysis did not change over time. However, to our knowledge, this study is the first to demonstrate an improvement in postoperative morbidity in patients after PD. Over 4 years, the overall morbidity was 47% and decreased by 3.7% upon risk-adjusted analysis (P < 0.02). Factors that contributed to these improvements included statistically significant reductions in superficial and deep SSIs, and also in sepsis/septic shock. However, no changes were observed in clinically relevant postoperative pancreatic fistulas or in delayed gastric emptying. Also, overall morbidity did not change over time for DP.

While individual metrics of performance like mortality and morbidity are important in understanding surgical quality, they fail to provide a broader picture of the average patient's experience. Composite measures are aggregates of multiple individual quality metrics and have been created in an attempt to provide a more accurate description of surgical quality.[6,7,10,32] In addition, utilization of all-or-none composite measures improves the ability to measure variation among surgeons, hospitals, and healthcare systems.[8,11] "Textbook outcomes" are composite measures that are less robust in their scientific validation, but are "patient-centric," all-or-none measurements that conform to an accepted standard held by experts in the field regarding the most flawless hospital course.[10,32] While textbook outcomes provide important information as composite measures, the connotation of the term "textbook" asserts that the information used is dated—not unlike an actual textbook by the time it is published. Therefore, we propose the term "optimal pancreatic surgery," which means the most desirable or satisfactory outcome.[12] Also, for many years, the American College of Surgeons has utilized the term "optimal" in describing resources for trauma patients.

For the purposes of this study, an optimal outcome after PD and DP was defined as the absence of postoperative mortality, serious morbidity, postoperative percutaneous drainage, or reoperations, while maintaining an acceptable postoperative LOS and no readmission. These components were not weighted as they were developed as a patient-centered composite outcome using an "all-or-none" approach. These metrics were chosen due to their well-defined impact on the patients' postoperative course and their availability to be measured in the NSQIP PUFs. Overall, 55% of patients who underwent PD and 56% of patients who underwent DP experienced optimal pancreatic surgery. Additionally, the percentage of patients undergoing PD and DP experiencing optimal pancreatic surgery increased over time. These trends (both P < 0.01) also were statistically significant when analyzing the subset of patients with pancreatic cancer.

These improvements in optimal pancreatic surgery for patients undergoing PD and DP have been driven, in part, by a reduction over time in the need for postoperative percutaneous drainage and reductions in postoperative LOS. While data on enhanced recovery are limited from the current study, the broad dissemination and implementation of enhanced recovery pathways over the past 5 to 10 years may have contributed to the observed reduction in LOS.[33–35] In addition, 3 webinars per year and 2 in-person meetings, and also an annual symposium at the ACS Quality and Safety Conference have been provided by the NSQIP-HPB Collaborative. These programs have focused on best practices to reduce SSIs and to prevent pancreatic fistulas. With regards to the reduction in postoperative percutaneous drain placement, the earlier removal of operatively placed drains that was observed over the study time may have played a role. As above, multiple retrospective studies and a randomized controlled trial also have reported the benefits of early drain removal (by POD 3) in the setting of a low POD 1 DFA-1 level.[27–31]

The above findings must be interpreted within the context of the study limitations. Because the NSQIP PUFs do not include patient identifiers or institutional data, changes over time in the participating institutions and their respective hospital volume/surgeon experience was not available for inclusion in the risk-adjusted models. However, the participating institutions in the pilot program—the Pancreatectomy Demonstration Project—were primarily high-volume centers which subsequently became the first institutions in NSQIP to collect pancreatectomy-specific variables.[36] As such, the addition of subsequent centers is unlikely to have further improved outcomes based on volume and surgeon experience alone. Another limitation is that data on pancreas-specific variables were not available in the year 2013. Whether this information might have changed the present results is unknown. Lastly, adverse events and mortality were standardized across the study time period and amongst centers, but 90-day morbidity and mortality, and in-hospital events that occurred beyond 30 days postoperatively were not available. As such, the morbidity and mortality reported in the present study (30-day) may underestimate the true incidence after PD and DP.[37,38]

Despite these limitations, the fact that a statistically significant improvement in optimal pancreatic surgery was observed over the abbreviated study period is an important finding and is testament to the efforts by the pancreatic surgical community to improve postoperative outcomes. While the absolute difference in the percentage of patients who received optimal pancreatic surgery was 3% to 5%, the implications of these findings are broad and this trend should continue over the next decade. The impact of even small improvements in postoperative morbidity and mortality when applied over thousands of patients over many years cannot be overstated. The process changes and quality measures put into place to achieve such improvements over time have the potential to increase survivorship, improve patient quality of life, and reduce cost. In patients with pancreatic cancer, reducing postoperative morbidity has important implications with regards to receiving adjuvant chemotherapy and improving prognosis. Should the small, but incremental, improvements observed in our study continue, thousands of patients will benefit from the surgical community's ongoing efforts of process and quality improvement.