Comparing the burden of pancreatic fistulas after pancreatoduodenectomy and distal pancreatectomy

Pancreas Comparing the burden of pancreatic fistulas after pancreatoduodenectomy and distal pancreatectomy Matthew T. McMillan, BA,a John D. Christein, MD,b Mark P. Callery, MD,c Stephen W. Behrman, MD,d Jeffrey A. Drebin, MD, PhD,a Robert H. Hollis, MD,b Tara S. Kent, MD,c Benjamin C. Miller, BA,a Michael H. Sprys, MS,a Ammara A. Watkins, MD,c Steven M. Strasberg, MD,e and Charles M. Vollmer, Jr, MD,a Philadelphia, PA, Birmingham, AL, Boston, MA, Memphis, TN, and St. Louis, MO

Background. Differences in the behavior of postoperative pancreatic fistulas (POPF) have been described after various pancreatic resections. Here, we compare POPFs after pancreatoduodenectomy (PD) and distal pancreatectomy (DP) using the average complication burden (ACB), a quantitative measure of complication burden. Methods. From 2001 to 2014, 837 DPs and 1,533 PDs were performed by 14 surgeons at 4 institutions. POPFs were categorized by International Study Group on Pancreatic Fistula standards as biochemical (grade A) or clinically relevant (CR-POPF; grades B and C). ACB values were derived from fistula severity scores based on the Modified Accordion Severity Grading. The ACB of POPFs was compared between PD and DP. Results. POPFs were more common after DP compared with PD (34.5 vs 27.2%; P < .001); however, the incidence of any complication was greater after PD (64.9 vs 53.2%; P < .001). When POPFs occurred, they were more likely to be the highest-graded complication after DP compared with PD (65.1 vs 51.6%; P < .001). ACB significantly varied between PDs and DPs for grade C POPFs (0.804 vs 0.611; P < .001). POPFs accounted for 31.2% of the overall complication burden after DP compared with 17.5% of the burden after PD. ACB differed significantly across both institutions and surgeons in terms of POPFs, nonfistulous complications, and overall complications (all P < .05). Conclusion. Although POPFs occur less frequently after PD, they are associated with a greater complication burden compared with DP. ACB varies significantly between health care providers, suggesting the need for risk-adjusted comparisons of complication severity. Using ACB to evaluate a distinct morbidity has the potential to aid in assessing the impact of procedure-specific complications. (Surgery 2016;159:1013-22.) From the Department of Surgery,a University of Pennsylvania Perelman School of Medicine, Philadelphia, PA; the Department of Surgery,b University of Alabama School of Medicine, Birmingham, AL; the Department of Surgery,c Beth Israel Deaconess Medical Center, Harvard Medical School, Boston, MA; the Department of Surgery,d University of Tennessee Health Sciences Center, Memphis, TN; and the Department of Surgery,e Washington University in Saint Louis School of Medicine, St. Louis, MO

Funding: None. Presented at the Society of University Surgeons/9th Annual Academic Surgical Congress, San Diego, CA, February 6, 2014. Accepted for publication October 23, 2015. Reprint requests: Charles M. Vollmer Jr., MD, Department of Surgery, University of Pennsylvania Perelman School of Medicine, Philadelphia, PA 19104. E-mail: Charles.Vollmer@uphs. upenn.edu. 0039-6060/$ - see front matter Ó 2016 Elsevier Inc. All rights reserved. http://dx.doi.org/10.1016/j.surg.2015.10.028

CLINICAL EXPERIENCE and the literature suggest there is a difference in the behavior of postoperative pancreatic fistulas (POPF) after various pancreatic resections1,2; however, these differences have been described traditionally in terms of their frequency rather than the actual burden that they incur. The utility-based Modified Accordion Severity Grading System classifies complications into groups based on severity,3 and this has facilitated the development of the postoperative morbidity index (PMI), which is a quantitative measure of SURGERY 1013

1014 McMillan et al

complication burden.4 This system uses utility weighted scoring---ranging from 0 to 1.000---to assess the clinical impact of postoperative outcomes. A recent study established the PMI of American College of Surgeons National Surgical Quality Improvement Program (ACS-NSQIP) complications after pancreatoduodenectomy (PD), as well as their contribution to overall complication burden.5 It also introduced the concept of average complication burden (ACB), which can be calculated by summing the individual severity weights of a specific complication---when it was the highest graded complication---and dividing the aggregate sum by the total number of patients who experienced it as the highest graded complication. ACB values offer insight into the average morbidity associated with a particular complication and can range continuously between 0 and 1.000. In another application of this concept, Lee et al6 defined the PMI of ACS-NSQIP complications after distal pancreatectomy (DP). Both of these studies were advancements in surgical complication grading through demonstrating differences in overall complication burden between institutions and surgeons, yet neither reported the complication burden incurred specifically by International Study Group on Pancreatic Fistula (ISGPF) POPFs---widely considered the most common and morbid complication after both PD and DP. This shortcoming arose because the procedure-specific variable of POPF is currently not available for analysis using ACS-NSQIP. One study to date has evaluated the burden of POPF,7 but it focused exclusively on PD; furthermore, it did not compare the burden of POPF complications between surgeons and institutions. Previous studies have shown that the overall complication burden of a procedure (ie, PMI) varies between health care providers,5,6 but this variability has not been evaluated for important procedure-specific complications such as POPF. Therefore, this study sought to establish the complication burden of POPF through the concept of ACB. Additionally, it aims to compare the ACB of POPF between surgeons and institutions; variability in ACB could demonstrate that calculations of procedure-specific ACB are necessary for comparisons of surgeon/institutional performance and standardization between research studies. Secondary analyses compare the occurrence of POPF between PD and DP and evaluate the specific contribution of POPF to overall complication burden.

Surgery April 2016

METHODS Patient population and data collection. This study was approved by the Institutional Review Board at the University of Pennsylvania. Records for consecutive patients who received PD or DP for all indications between 2001 and 2014 were reviewed from prospectively collected databases at 4 high-volume pancreatic surgical practices. Postoperative complications were recorded by clinical research associates and complication grading was assigned by, or under the supervision of, attending surgeons familiar with complication severity scoring.8 Outcome assessment tools used included the ISGPF classification scheme,9 the Modified Accordion Severity Grading System,3 and the ACB.4,5 Classification of pancreatic fistulas. All pancreatic fistulas were graded in accordance with the guidelines established by the ISGPF.9,10 These standardized definitions were segregated into 2 general categories---biochemical or clinically relevant---and 3 grades (ie, A, B, C). Biochemical (grade A) fistulas are clinically asymptomatic and characterized by elevated drain amylase values (>33 the upper limit of normal serum amylase concentration) with negligible clinical consequence. Conversely, CR-POPFs (grades B and C) are more morbid and stray from the normal clinical course. Grade B POPFs are those managed with antibiotic therapy, therapeutic (nonprophylactic) somatostatin analogues, total parenteral nutrition, total enteral nutrition, transfusion, maintenance of drains for a prolonged period (>21 days), additional percutaneous drainage, or any combination thereof. Grade C POPFs necessitate a major change in clinical management and require aggressive clinical treatment; more specifically, they are characterized by $1 of the following: an operative intervention under general anesthesia, organ failure, or death. Assignment of complication severity grading. Each postoperative complication was assigned a severity grade (range, 1–6) according to the Modified Accordion Severity Grading System (Table I).3 A Modified Accordion grade of 1 signified the least deleterious complications and a 6 indicated death as a direct consequence of the complication. In the absence of any postoperative complication, patients were assigned a grade of zero. In accordance with the original tenants of complication severity assessment,8 when multiple complications occurred the patient was assigned the grade of only the most severe complication--designated as the highest grade complication.4 Other complications of lesser severity in these

McMillan et al 1015

Surgery Volume 159, Number 4

Table I. Modified Accordion Severity Grading System with severity weights Grade 1 (Mild)

2 (Moderate)

3 (Severe A)

4 (Severe B) 5 (Severe C)

6 (Death)

Description

Severity weight

Treatment of complication requires only minor invasive procedures that can be done at the bedside, such as insertion of intravenous lines, urinary catheters, and nasogastric tubes, and drainage of wound infections. Physiotherapy and antiemetics, antipyretics, analgesics, diuretics, electrolytes, and physiotherapy are permitted. Complication requires pharmacologic treatment with drugs other than those allowed for minor complications, eg, antibiotics. Blood transfusions and total parenteral nutrition are also included. No general anesthesia is required to treat the complication; requires management by an endoscopic, interventional procedure, or reoperation without general anesthesia. General anesthesia is required to treat complication. Alternately, singleorgan failure has developed. General anesthesia is required to treat complication and single organ failure has developed. Alternately, multisystem organ failure ($2 organ systems) has developed. Postoperative death occurred.

0.110

0.260

0.370

0.600 0.790

1.000

Taken from Strasberg SM, Hall BL. Postoperative morbidity index: a quantitative measure of severity of postoperative complications. J Am Coll Surg 2011;213:616-26.4

patients were classified as not highest graded complications.4 Concomitant with the designation of Modified Accordion Severity Grades for all complications encountered, fistula Accordion Severity Grades were assigned in cases of a POPF. Fistula Accordion Severity Grades function as a Modified Accordion analogue, which assigns severity grades directly attributable to---and only to---the sequelae of a fistula. Following ISGPF codification, fistula Accordion Severity Grades were derived by pairwise comparison of ISGPF categorization standards with the 6 levels of clinical severity established by the Modified Accordion Severity Grades. Table II displays the full spectrum of qualitative severity that a fistula may inflict upon the patient. Accordingly, only certain fistula Accordion Severity Grades align with each ISGPF tier.7 Grade A fistulas can only accrue a fistula Accordion Severity Grade of 0 or 1. The characteristics accompanying each grade A POPF fistula Accordion Severity Grade are as follows: grade 0, increased postoperative drain amylase alone (>300 IU/L) with no clinical influence; grade 1, POPF with evidence of elevated drain amylase where an operative drain is maintained on discharge from the hospital. The range of possible grade B POPF fistula Accordion Severity Grades extends from 1 to 3: grade 1, prolonged drainage persisting beyond 21 days; grade 2, application of any pharmacologic therapeutics for the fistula including, but not limited to, therapeutic octreotide, antibiotics, and total parenteral nutrition/total enteral nutrition; grade 3,

percutaneous drain placement and endoscopic or interventional management of complications from fistulas (eg, fluid collections, hemorrhage). Grade C POPF Accordion Severity Grades coincide with the most severe classifications: grade 4, reoperations or development of single organ failure; grade 5, typically multiorgan failure; and grade 6, death owing to fistula. After assessment of severity grades, each POPF was weighted using appropriate utility-derived severity weights (Table II). The ACB for each ISGPF classification of fistula was calculated for each cohort by summing the individual severity weights of highest graded POPFs and dividing the aggregate sum by the total number of patients who suffered each category of POPF. ACB values offer insight into the average morbidity associated with each type of POPF and can range continuously between 0 and 1.000. In addition to ACB, several subset comparative analyses were undertaken to elucidate the fistula complications that contribute most prominently to overall patient burden for each operation. The first subset analysis focused on the percentage of the overall burden directly attributable to ISGPF POPFs. This calculation was simply the sum of the severity weights for each ISGPF grade---when they occurred as the highest graded complication--divided by the overall complication burden experienced by each cohort. The same methodology was also used to establish the percentage of overall burden attributed to CR-POPFs (ie, grades B and C) and nonfistulous complications.

1016 McMillan et al

Surgery April 2016

Table II. Modified Accordion Severity Grading System specifically for postoperative pancreatic fistula Fistula Accordion Severity Grade None 1 2

3

4

5

6

Description

Severity weight

Elevated amylase alone, but no intervention (biochemical POPF). Discharged from hospital with original operatively placed drain, with no other interventions required. Treatment for POPF requires the use of therapeutic (not prophylactic) somatostatin analogues, antibiotics, TPN, or TEN (via pre-existing NG- or J-tubes). Any operative procedure short of general anesthesia required for POPF management: complex wound management; percutaneous drain placement or aspiration of an amylase-rich collection postoperatively; angiographic procedure for control of a pseudoaneurysm secondary to POPF; or endoscopic procedure for POPF management. Reoperation under general anesthesia for an anastomotic leak from the pancreaticojejunostomy or pancreaticogastrostomy. or Single organ failure secondary to the POPF (eg, renal failure, pulmonary failure, neurologic failure, etc). Reoperation under general anesthesia for an anastomotic leak and single organ failure secondary to the POPF event (eg, renal failure, pulmonary failure, neurologic failure, etc). or Multi-system ($2) organ failure secondary to the POPF. Death attributable to POPF.

0.000 0.110 0.260

0.370

0.600

0.790

1.000

J-tube, Jejunostomy tube; NG, nasogastric; POPF, postoperative pancreatic fistula; TEN, total enteral nutrition; TPN, total parenteral nutrition.

Last, the occurrence of fistulas and their ACB was compared between health care providers. All 4 institutions were considered, however, only surgeons who contributed >75 overall cases were included for comparative analyses of surgeon performance. The learning curve for PD has been previously reported as $50 PDs11 and >60 PDs.12 None of the surgeons who contributed #75 cases had a career PD volume of $50; however, each of the surgeons who contributed >75 cases had performed >200 PDs during their career as an attending surgeon---with their most recent experiences included in this study. Owing to this discrepancy, the present study determined that a cutoff including only surgeons who contributed >75 cases would minimize surgical experience bias when making performance comparisons. Statistical analysis. Continuous variables are expressed as mean values ± standard deviation or median (interquartile range [IQR]), and categorical variables are presented as absolute numbers and percentages. Qualitative and quantitative measures for severity of burden were analyzed using the Modified Accordion Severity Grading System and ACB, respectively. Testing for nonrandom association was conducted with chisquared analysis or Fisher’s exact test for

categorical variables and the Student’s t-test for continuous variables. Nonparametric comparisons of continuous variables were assessed by the Wilcoxon rank-sum test. Analysis of variance or Kruskal–Wallis tests were used for comparisons of multiple groups. All statistical computations were performed using IBM SPSS, version 22 (IBM Corp., Armonk, NY). RESULTS PD versus DP: Demographic, operative, and pathologic characteristics. Fourteen surgeons performed 2,370 pancreatic resections, of which 1,533 (64.7%) were PD and 837 (35.3%) DP. Numerous differences existed between the PD and DP cohorts in terms of demographic and operative characteristics. Patients undergoing DP were more likely female (59.4 vs 50.6%; P < .0001) and obese (33.5 vs 22.5%; P < .0001), whereas PD patients were significantly older (median, 64 [IQR 56–72] vs 60 [IQR 49–69] years; P < .0001). A greater proportion of PD patients had pancreatic ductal adenocarcinoma (42.5 vs 22.3%; P < .0001) and neuroendocrine tumors (15.2 vs 5.2%; P < .0001); conversely, cystic neoplasms were the indication for a greater proportion of DP patients (35.6 vs 13.4%; P < .0001). Chronic pancreatitis

McMillan et al 1017

Surgery Volume 159, Number 4

Table III. Postoperative outcomes of the overall series, as well as the PD and DP cohorts Variable, n (%)

Overall

PD

Patients 2,370 1,533 (64.7) ISGPF POPFs 706 (29.8) 417 (27.2) Grade A 409 (17.3) 243 (15.9) CR-POPFs 297 (12.5) 174 (11.4) Grade B 252 (10.6) 147 (9.6) Grade C 45 (1.9) 27 (1.8) Any complication (Accordion $1) 1,440 (60.8) 995 (64.9) Mild complication (Accordion 1–2) 997 (42.1) 691 (45.1) Severe complication (Accordion $3) 443 (18.7) 304 (19.8) ICU admit 806 (34.1) 652 (42.6) Percutaneous drain placement 161 (6.8) 113 (7.4) TPN/TEN 419 (17.7) 354 (23.1) Antibiotics 683 (28.9) 494 (32.2) Reoperation (90D) 124 (5.2) 84 (5.5) Readmission (90D) 451 (19.3) 262 (17.5) Duration of stay Mean ± SD 10.2 ± 8.1 11.6 ± 9.2 Median (IQR) 8 (7–10) 8 (7–12) Mortality (90 d) 32 (1.4) 27 (1.8) Average complication burden ISGPF POPFs 0.235 ± 0.225 0.254 ± 0.252 Grade A 0.069 ± 0.053 0.070 ± 0.053 CR-POPFs 0.383 ± 0.216 0.405 ± 0.250 Grade B 0.290 ± 0.106 0.286 ± 0.104 Grade C 0.727 ± 0.165 0.804 ± 0.171 Nonfistulous complications 0.311 ± 0.190 0.316 ± 0.193 Overall burden 439.320 312.130 Contribution to overall burden, complication burden (% of total burden) ISGPF POPFs 94.560 (21.52) 54.590 (17.49) Grade A 13.090 (2.98) 6.820 (2.18) CR-POPFs 81.470 (18.55) 47.770 (15.30) Grade B 48.760 (11.10) 26.050 (8.35) Grade C 32.710 (7.45) 21.720 (6.96) Nonfistulous complications 344.760 (78.48) 257.540 (82.51)

DP 837 289 166 123 105 18 445 306 139 154 48 65 189 40 189

P value

(35.3) (34.5) (19.8) (14.7) (12.5) (2.2) (53.2) (36.6) (16.6) (18.4) (5.7) (7.8) (22.7) (4.8) (22.6)

— <.001 .014 .019 .026 .507 <.001 <.001 .054 <.001 .13 <.001 <.001 .464 .003

7.5 ± 4.4 7 (6–8) 5 (0.6)

<.001 <.001 .019

0.213 ± 0.187 0.067 ± 0.054 0.355 ± 0.160 0.295 ± 0.109 0.611 ± 0.045 0.298 ± 0.183 127.190

.060 .710 .078 .600 <.001 .158 —

39.970 6.270 33.700 22.710 10.990 87.220

(31.43) (4.93) (26.50) (17.86) (8.64) (68.58)

— — — — — —

CR-POPF, Clinically relevant postoperative pancreatic fistula; DP, distal pancreatectomy; ICU, intensive care unit; IQR, interquartile range; ISGPF, International Study Group on Pancreatic Fistula; PD, pancreatoduodenectomy; POPF, postoperative pancreatic fistula; SD, standard deviation; TEN, total enteral nutrition: TPN, total parenteral nutrition.

was reported at a similar rate between cohorts (PD vs DP, 11.3 vs 12.1%; P = .569). In terms of perioperative management, prophylactic octreotide was administered more commonly to PD patients (25.5 vs 10.8%; P < .0001), as well as prophylactic drainage (92.6 vs 88.2%; P = .0002). All PDs were performed with duct-to-mucosa pancreaticojejunostomy (n = 1,543), whereas staplers were the primary method of closure for DP (N = 682; 82.4%). Transanastomotic stents were placed during 876 PDs (57%), with the external variety selected most frequently (n = 538; 61.3%). Both mean and median blood loss values were greater during PD (mean, 501.3 ± 631.0 vs 414.2 ± 694.1 mL [P = .002]; median, 350 [IQR 200–600] vs 250 [IQR 150–500] mL [P < .0001]).

PD versus DP: Overall outcomes. The incidence of any complication (Modified Accordion $1) was greater after PD (64.9 vs 53.2%; P < .001; Table III); PD patients also had higher rates of mild (Modified Accordion 1–2; 45.1 vs 36.6% [P < .001]) and severe (Modified Accordion $3; 19.8 vs 16.6% [P = .054]) complications. The following management options were used more frequently after PD: intensive care unit admission (42.6 vs 18.4%; P < .001), total parenteral nutrition/total enteral nutrition (23.1 vs 7.8%; P < .001), and therapeutic antibiotics (32.2 vs 22.7%; P < .001); alternatively, readmissions were more common after DP (22.6 vs 17.5%; P = .003). Median duration of hospital stay was greater for PD patients (8 [IQR 7–12] vs 7 [6–8]

1018 McMillan et al

Surgery April 2016

Fig 1. Comparisons of average complication burden between pancreatoduodenectomy and distal pancreatectomy for International Study Group on Pancreatic Fistula grades and nonfistulous complications. POPF, Postoperative pancreatic fistula; SD, standard deviation.

days [P < .001]), as well as 30-day mortality (1.4 vs 0.5%; P = .042) and 90-day mortality (1.8 vs 0.6%; P = .019). PD versus DP: Frequency and burden of pancreatic fistulas. POPFs, of any severity, were more common after DP compared with PD (34.5 vs 27.2%; P < .001); this discrepancy was true for both grade A POPFs (19.8 vs 15.9%; P = .014) and grade B POPFs (12.5 vs 9.6%; P = .026; Table III). Although DPs were characterized by higher rates of CR-POPF (14.7 vs 11.4%; P = .019), they did not have significantly higher rates of grade C POPF (2.2 vs 1.8%; P = .507). When POPFs occurred, they were more likely to be the highest graded complication after DP compared with PD (65.1 vs 51.6%; P < .001). Stratified by ISGPF grade, this was true for grade A POPFs (56.0 vs 39.9%; P < .001) but not grade B POPFs (73.3 vs 61.9%; P = .058); grade C POPFs were always the highest graded complication for both operations. Comparisons of ACB between ISGPF grades demonstrated a significant difference between PD and DP (Fig 1). PD trended toward higher ACB for overall POPFs compared with DP (0.254 ± 0.252 vs 0.213 ± 0.187; P = .060), and this was significant for grade C POPFs (0.804 ± 0.171 vs 0.611 ± 0.045; P < .001; Table III). The ACB for nonfistulous complications was similar after PD compared with DP (0.316 ± 193 vs 0.298 ± 0.183; P = .158). After both procedures, the ACB for CR-POPFs and grade C POPFs was

Fig 2. Contribution of pancreatic fistula to overall complication burden.

greater than the ACB for nonfistulous complications (both P < .001; Table III); however, the ACB for overall and grade A POPFs was less than the ACB of nonfistulous complications for both operations. The contribution of POPFs to overall complication burden differed between procedures. After DP, POPF contributed to nearly onethird of the overall complication burden (31.22%); alternatively, only 17.5% of the burden after PD was attributed to POPFs (Fig 2). Variability in fistula occurrence and burden between institutions and surgeons. Institutional and surgeon (>75 overall cases contributed) comparisons of complications after PD revealed

Surgery Volume 159, Number 4

McMillan et al 1019

Fig 3. Comparing the burden and severity of fistulas, by the International Study Group on Pancreatic Fistula nomenclature, between institutions (labeled 1–4) for (A) pancreatoduodenectomy and (B) distal pancreatectomy.

Fig 4. Comparing the burden and severity of fistulas, by the International Study Group on Pancreatic Fistula nomenclature, between surgeons who contributed >75 cases (labeled 1–7) for (A) pancreatoduodenectomy and (B) distal pancreatectomy.

significant variability in the occurrence of nonfistulous complications, overall POPFs, grade A POPFs, CR-POPFs, grade B POPFs, grade C POPFs, and overall complications (Supplemental Tables I and II). When analyzing DP, significant variability was noted for each of the same complications as PD with the exception of nonfistulous complications (Supplemental Tables I and II). There was also significant variability in the burden of complications between institutions. After both PD and DP, this was apparent for the ACB of nonfistulous complications, overall POPFs, CRPOPFs, grade B POPFs, and overall complications (Fig 3; Supplemental Table I). Interestingly, for both PD and DP, there was a negative correlation between annual institutional PD volume and the ACB of overall POPFs, CR-POPFs, grade B POPFs, and nonfistulous complications (all P # .05). For PD, there was also a negative correlation between overall institutional pancreatectomy volume and the ACB of overall POPFs, CR-POPFs, grade B POPFs, and nonfistulous complications (all P # .05). For DP, there was a negative correlation between institutional pancreatectomy volume and the ACB of overall POPFs, CR-POPFs, and grade B POPFs (all P < .002).

Comparisons of burden were also made between the seven surgeons who contributed >75 cases (Fig 4; Supplemental Table II). Similar to the findings from institutional comparisons of burden after PD and DP, there were differences in complication burden in terms of nonfistulous complications, overall POPFs, CR-POPFs, grade B POPFs, and overall complications. Bringing the concepts of complication frequency and burden together demonstrates these disparities (Figs 3 and 4). DISCUSSION This study demonstrates that, although the overall incidence of ISGPF POPFs is higher after DP compared with PD, the ACB of POPFs is typically greater after PD. Furthermore, POPFs are more likely the highest graded complication after DP, and they contribute to a greater proportion of overall complication burden after this procedure (31.4 vs 17.5%). Applying the concept of burden to the outcomes of health care providers revealed that ACB varied between institutions and surgeons for both PD and DP in terms of overall POPFs, CR-POPFs, grades B and C POPFs, nonfistulous complications, and overall complications. For DP and PD, both institutional PD and overall

1020 McMillan et al

pancreatectomy volumes negatively correlated with the ACB (ie, higher volume centers reported lower ACB) of overall POPFs, CR-POPFs, and grade B POPFs. Variability in the occurrence of these complications was also apparent on the institutional and surgeon levels. Comparative analyses between PD and DP cohorts revealed that POPFs occurred more frequently after DP. This result conflicts with several reports that exist in the literature, but many of these accounts took place before the advent of the ISGPF nomenclature for POPF.13-16 After the establishment of ISGPF guidelines, a comparative analysis---similar to the current study---established higher POPF rates after DP compared with PD.2 This could be partly explained by the higher rates of grade A (biochemical) POPF after DP. A proposed explanation for higher rates of biochemical leaks after DP is that the small branches of the main pancreatic duct continue to release pancreatic fluids at the site of the transection margin after DP.17-21 Another notable point in the current study was the higher rate of CR-POPF among DP patients (14.7 vs 11.4%; P = .019). This finding might be confounded by the greater proportion of obese (body mass index $30 kg/m3) patients in the DP cohort (33.5 vs 22.5%; P < .0001). Greater body mass index has previously been reported as a CR-POPF risk factor for both PD and DP.22 Although CR-POPFs occurred more frequently after DP, the ACB of CR-POPFs trended higher after PD (0.405 vs 0.355; P = .078) and was significantly greater when comparing the more severe grade C POPFs (0.804 vs 0.611; P < .001). One factor driving this variance was the greater incidence of mortality in the presence of grade C POPFs after PD relative to DP (25.9 vs 0%; P = .031). For grade C POPFs, the difference in ACB between procedures was equivalent to greater than an entire level of utility-weighted severity. For example, a patient with a grade C POPF after DP can expect a reoperation or single organ failure, corresponding with a Modified Accordion Grade 4 weight of 0.600. In stark contrast, the average patient with a grade C POPF after PD can expect reoperation and single-organ failure or multisystem organ failure; put another way, these characterize a Modified Accordion Grade 5 severity weight of 0.790. Contrasting the severity of POPFs between procedures demonstrated that POPFs were more likely the highest graded complication after DP (65.1 vs 51.6%; P < .001). This discrepancy could be owing to the greater frequency of severe (Modified Accordion $3), nonfistulous complications after PD

Surgery April 2016 compared with DP (14.4 vs 7.6%; P < .001). The greater prevalence of nonfistulous, severe, highest graded complications after PD was underscored by POPFs contributing to a smaller proportion of overall complication burden after PD compared with DP (17.5 vs 31.4%). This finding has been corroborated by recent studies evaluating complication burden after PD5 and DP.6 A study by Lee et al6 reported that organ space infection---an ACS-NSQIP surrogate for POPF---accounted for 26.1% of the overall complication burden after DP; conversely, an analysis of PD attributed only 18.6% of overall complication burden to organ space infections.5 POPFs were more likely to be the highest graded complication after PD; however, when grade C POPFs occurred, they were always the highest graded complication for both PD and DP. To the authors’ knowledge, this study is the first of its kind to report this finding for DP. This phenomenon was observed recently in a multiinstitutional analysis of >4,300 PDs; the study reported that, when grade C POPFs occurred, they were the highest graded complication 97.5% of the time.23 This shows that grade C POPFs are almost never surpassed in terms of complication severity. Another difference between the PD and DP cohorts was the greater incidence of readmission after DP. This discrepancy is likely related to the timing of CR-POPF incidence. In the absence of a CR-POPF, readmission rates were actually marginally higher after PD compared with DP (15.0 vs 14.1%; P = .583); however, when CR-POPFs occurred, readmissions occurred significantly more often after DP (71.5 vs 36.9%; P < .000001). This phenomenon was previously reported by Pratt et al,2 where the readmission rate for CR-POPF patients was significantly greater after DP compared with PD (88 vs 21%; P < .001). Bringing the concepts of complication frequency and burden together, this study also demonstrated disparities in outcomes between health care providers (Figs 3 and 4). The discordance between complication frequency and burden is particularly important when evaluating major procedure-specific complications such as POPF. Most performance measures to date have evaluated the frequency of complications, rather than severity; however, this study demonstrates that considerations of complication frequency are not sufficient. The ACB burden of overall POPF, CR-POPF, grade B POPF, nonfistulous complications, and overall complications differed significantly between health care providers for both PD

McMillan et al 1021

Surgery Volume 159, Number 4

and DP (Figs 3 and 4). This reinforces the notion that complication burden is necessary for comparisons of surgeon/institutional performance, and could be useful for standardization between research studies and the establishment of metrics for clinical trial design. Furthermore, a provider’s frequency of POPF does not necessarily correlate with the burden incurred. Although this study’s quantitative foundation represents an advance from the use of imprecise, qualitative terms, it has several notable limitations. First, the ACB reported in this study is not risk adjusted; thus, some of the differences in complication burden and occurrence may reflect components of patients’ endogenous and operative risk. One potential way to address this limitation would be the use of a sequential logit approach to adjust for the risk of different classes of complications conditioned on having any complication. The riskadjusted performance of surgeons and institutions could then be estimated by calculating an observed to expected ratio of complications and burden. This shortcoming must be addressed in future studies before complication burden can be adopted as a primary measure of surgeon and institutional performance. Second, data in this study were accrued from individual practice databases that, despite being completed by surgeons with proficiency in complication grading, lack the methodologic rigor of a system such as ACS-NSQIP.24 The ACS-NSQIP offers 2 primary advantages: (i) complications are assessed by trained reviewers and (ii) every complication is rigorously defined. However, a major limitation of ACS-NSQIP is their inability to accrue important, procedure-specific complications such as POPF. This may be addressed with the implementation of a HPB-NSQIP,25 which would address the occurrence of POPF, but not their severity. The importance of procedurespecific complications in defining overall complication burden was recently demonstrated by Lee et al.6 Although that study did not report an ACB for each grade of ISGPF POPF, it did show that incorporating POPF data from private databases--to augment their analysis of ACS-NSQIP complication data---caused a considerable increase (20%) in overall complication burden. Therefore, until the ACS-NSQIP incorporates procedure-specific variables such as POPF, it might be underestimating the complication burden experienced by patients. Despite these limitations, this study is the largest of its kind to establish a standard for the expected ACB of procedure-specific complications arising from PD and DP. Quantitative analysis revealed

that although POPFs occur less frequently after PD, they incur a greater complication burden compared with DP. Additionally, the ACB for individual complications varied significantly between surgeons and institutions, indicating potential differences in quality of care, and suggesting the need for risk-adjusted comparisons of complication severity. This analysis of a distinct morbidity after a high-acuity operation, using the concept of ACB, could aid in assessing the impact of specific complications associated with other operations. SUPPLEMENTARY DATA Supplementary data related to this article can be found at http://dx.doi.org/10.1016/j.surg.2015.10.028.

REFERENCES 1. Fernandez-del Castillo C, Rattner DW, Warshaw AL. Standards for pancreatic resection in the 1990s. Arch Surg 1995;130:295-9; discussion 9–300. 2. Pratt W, Maithel SK, Vanounou T, Callery MP, Vollmer CM Jr. Postoperative pancreatic fistulas are not equivalent after proximal, distal, and central pancreatectomy. J Gastrointest Surg 2006;10:1264-78; discussion 78–9. 3. Porembka MR, Hall BL, Hirbe M, Strasberg SM. Quantitative weighting of postoperative complications based on the accordion severity grading system: demonstration of potential impact using the American College of Surgeons National Surgical Quality Improvement Program. J Am Coll Surg 2010;210:286-98. 4. Strasberg SM, Hall BL. Postoperative morbidity index: a quantitative measure of severity of postoperative complications. J Am Coll Surg 2011;213:616-26. 5. Vollmer CM Jr, Lewis RS, Hall BL, Allendorf JD, Beane JD, Behrman SW, et al. Establishing a quantitative benchmark for morbidity in pancreatoduodenectomy using ACS-NSQIP, the Accordion Severity Grading System, and the Postoperative Morbidity Index. Ann Surg 2015; 261:527-36. 6. Lee MK, Lewis RS, Strasberg SM, Hall BL, Allendorf JD, Beane JD, et al. Defining the post-operative morbidity index for distal pancreatectomy. HPB 2014;16:915-23. 7. Miller BC, Christein JD, Behrman SW, Callery MP, Drebin JA, Kent TS, et al. Assessing the impact of a fistula after a pancreaticoduodenectomy using the Post-operative Morbidity Index. HPB 2013;15:781-8. 8. Clavien PA, Sanabria JR, Strasberg SM. Proposed classification of complications of surgery with examples of utility in cholecystectomy. Surgery 1992;111:518-26. 9. Bassi C, Dervenis C, Butturini G, Fingerhut A, Yeo C, Izbicki J, et al. Postoperative pancreatic fistula: an international study group (ISGPF) definition. Surgery 2005;138: 8-13. 10. Hashimoto Y. The Pancreas Club ISGPS calculator. Pancreas Club, Inc; 2012. [4/16/2014]; Available from: http:// pancreasclub.com/calculators/isgps-calculator/. 11. Schmidt CM, Turrini O, Parikh P, House MG, Zyromski NJ, Nakeeb A, et al. Effect of hospital volume, surgeon experience, and surgeon volume on patient outcomes after pancreaticoduodenectomy: a single-institution experience. Arch Surg 2010;145:634-40.

1022 McMillan et al

12. Tseng JF, Pisters PW, Lee JE, Wang H, Gomez HF, Sun CC, et al. The learning curve in pancreatic surgery. Surgery 2007;141:456-63. 13. Knaebel HP, Diener MK, Wente MN, Buchler MW, Seiler CM. Systematic review and meta-analysis of technique for closure of the pancreatic remnant after distal pancreatectomy. Br J Surg 2005;92:539-46. 14. Fahy BN, Frey CF, Ho HS, Beckett L, Bold RJ. Morbidity, mortality, and technical factors of distal pancreatectomy. Am J Surg 2002;183:237-41. 15. Balzano G, Zerbi A, Cristallo M, Di Carlo V. The unsolved problem of fistula after left pancreatectomy: the benefit of cautious drain management. J Gastrointest Surg 2005;9: 837-42. 16. Kuroki T, Tajima Y, Kanematsu T. Surgical management for the prevention of pancreatic fistula following distal pancreatectomy. J Hepatobiliary Pancreatic Surg 2005;12:283-5. 17. Konishi T, Hiraishi M, Kubota K, Bandai Y, Makuuchi M, Idezuki Y. Segmental occlusion of the pancreatic duct with prolamine to prevent fistula formation after distal pancreatectomy. Ann Surg 1995;221:165-70. 18. Abe N, Sugiyama M, Suzuki Y, Yamaguchi Y, Yanagida O, Masaki T, et al. Preoperative endoscopic pancreatic stenting for prophylaxis of pancreatic fistula development after distal pancreatectomy. Am J Surg 2006;191: 198-200. 19. Truty MJ, Sawyer MD, Que FG. Decreasing pancreatic leak after distal pancreatectomy: saline-coupled radiofrequency

Surgery April 2016

20.

21.

22.

23.

24.

25.

ablation in a porcine model. J Gastrointest Surg 2007;11: 998-1007. Kitagawa H, Ohta T, Tani T, Tajima H, Nakagawara H, Ohnishi I, et al. Nonclosure technique with saline-coupled bipolar electrocautery in management of the cut surface after distal pancreatectomy. J Hepatobiliary Pancreatic Surg 2008;15:377-83. Sudo T, Murakami Y, Uemura K, Hayashidani Y, Hashimoto Y, Nakashima A, et al. Distal pancreatectomy with duct-to-mucosa pancreaticogastrostomy: a novel technique for preventing postoperative pancreatic fistula. Am J Surg 2011;202:77-81. McMillan MT, Vollmer CM Jr. Predictive factors for pancreatic fistula following pancreatectomy. Langenbeck Arch Surg 2014;399:811-24. McMillan MT, Vollmer CM Jr, Asbun HJ, Ball CG, Bassi C, Beane JD, et al. The characterization and prediction of ISGPF grade c fistulas following pancreatoduodenectomy. J Gastrointest Surg 2015. Sanford DE, Woolsey CA, Hall BL, Linehan DC, Hawkins WG, Fields RC, et al. Variations in definition and method of retrieval of complications influence outcomes statistics after pancreatoduodenectomy: comparison of NSQIP with non-NSQIP methods. J Am Coll Surg 2014; 219:407-15. Pitt HA, Kilbane M, Strasberg SM, Pawlik TM, Dixon E, Zyromski NJ, et al. ACS-NSQIP has the potential to create an HPB-NSQIP option. HPB 2009;11:405-13.