UvA-DARE (Digital Academic Repository) Optimizing ... · All analyses were performed using STATA 14.2 (StataCorp LP, College Station (TX), United States), supervised by a biostatistician

UvA-DARE is a service provided by the library of the University of Amsterdam (https://dare.uva.nl)

UvA-DARE (Digital Academic Repository)

Optimizing strategies in pancreatic and hepato-biliary surgery

Mungroop, T.H.

Publication date2020Document VersionOther versionLicenseOther

Link to publication

Citation for published version (APA):Mungroop, T. H. (2020). Optimizing strategies in pancreatic and hepato-biliary surgery.

General rightsIt is not permitted to download or to forward/distribute the text or part of it without the consent of the author(s)and/or copyright holder(s), other than for strictly personal, individual use, unless the work is under an opencontent license (like Creative Commons).

Disclaimer/Complaints regulationsIf you believe that digital publication of certain material infringes any of your rights or (privacy) interests, pleaselet the Library know, stating your reasons. In case of a legitimate complaint, the Library will make the materialinaccessible and/or remove it from the website. Please Ask the Library: https://uba.uva.nl/en/contact, or a letterto: Library of the University of Amsterdam, Secretariat, Singel 425, 1012 WP Amsterdam, The Netherlands. Youwill be contacted as soon as possible.

Download date:19 Jun 2021

https://dare.uva.nl/personal/pure/en/publications/optimizing-strategies-in-pancreatic-and-hepatobiliary-surgery(bf6fa380-7eb2-4ace-9655-42ac3f2f5729).html

Updated Alternative Fistula Risk Score (a-FRS) to include Minimally-Invasive Pancreatoduodenectomy:

Pan-European Validation

7CHAPTER

Timothy H. Mungroop,* Sjors Klompmaker,* Ulrich F. Wellner,

Ewout W. Steyerberg, Andrea Coratti, Mathieu D’Hondt,

Matteo de Pastena, Safi Dokmak, Igor Khatov, Olivier Saint-Marc,

Uwe Wittel, Mohammed Abu Hilal, David Fuks, Ignasi Poves,

Tobias Keck, Ugo Boggi, Marc G. Besselink for the European

consortium on Minimally Invasive Pancreatic Surgery (E-MIPS)

* Shared first authorship

Annals of Surgery. 2019 Sep 4. [Epub ahead of print]

104 | Chapter 7

ABSTRACT

Objective: To validate and optimize the alternative Fistula Risk Score (a-FRS) for patients undergoing minimally invasive pancreatoduodenectomy (MIPD) in a large pan-European cohort.

Background: MIPD may be associated with an increased risk of postoperative pancreatic fistula (POPF). The a-FRS could allow for risk-adjusted comparisons in research and improve preventive strategies for high-risk patients. The a-FRS, however, has not yet been validated specifically for laparoscopic, robot-assisted and hybrid MIPD.

Methods: A validation study was performed in a pan-European cohort of 952 consecutive patients undergoing MIPD (543 laparoscopic, 258 robot-assisted, 151 hybrid) in 26 centers from 7 countries between 2007 and 2017. The primary outcome was POPF (ISGPS grade B/C). Model performance was assessed using the area under the receiver-operating-curve (AUC; discrimination) and calibration plots. Validation included univariable screening for clinical variables that could improve performance.

Results: Overall, 202 of 952 patients (21%) developed POPF after MIPD. Before adjustment, the original a-FRS performed moderately (AUC 0.68) and calibration was inadequate with systematic underestimation of the POPF risk. Single-row pancreatojejunostomy (OR 4.6, 95-CI 2.8-7.6) and male sex (OR 1.9, 95-CI 1.4-2.7) were identified as important risk factors for POPF in MIPD. The updated a-FRS, consisting of BMI, pancreatic texture, duct size, and male sex, showed good discrimination (AUC 0.75, 95-CI 0.71-0.79) and adequate calibration. Performance was adequate for laparoscopic, robot-assisted, and hybrid MIPD as well as open pancreatoduodenectomy (OPD).

Conclusions: The updated a-FRS (www.pancreascalculator.com) now includes male sex as a risk factor and is validated for both MIPD and OPD. The increased risk of POPF in MIPD was related to single-row pancreatojejunostomy, which should therefore be discouraged.

7

Updated Alternative Fistula Risk Score to Include Minimally-Invasive PD | 105

INTRODUCTION

Postoperative pancreatic fistula (POPF) remains the most important cause of morbidity and mortality after pancreatoduodenectomy (PD).1–3 Accurate risk prediction of POPF may enable surgeons to selectively use preventive measures (e.g., drain placement, use of somatostatin or hydrocortisone, or proactive monitoring). It also enables risk-adjusted comparisons in (large) cohort studies and stratification in randomized controlled trials (RCTs). The alternative Fistula Risk Score (a-FRS)4 is an externally validated, online tool to predict POPF, using BMI, pancreatic duct size, and texture. The a-FRS omitted perioperative blood loss, one of the predictors in the original Fistula Risk Score.5 Indeed, perioperative blood loss was not a significant predictor in several external validation studies.6, 7

Minimally invasive pancreatoduodenectomy (MIPD) is becoming increasingly popular because of its potential to enhance postoperative recovery compared to open PD.8–19 Some studies, however, have associated MIPD with increased rates of POPF, which may offset the potential benefits.20–23 The a-FRS could help to perform risk-adjusted comparisons of outcomes in multicenter studies and identify the most optimal MIPD technique. In addition, the a-FRS can help to identify high-risk patients who could benefit from preventative measures.24 The a-FRS has, however, not yet been validated for laparoscopic, robot-assisted, or hybrid MIPD. Therefore, the purpose of this study was to validate the a-FRS in a large international cohort of patients undergoing MIPD.

METhODS

This is a post-hoc analysis of a pan-European multicenter cohort of 952 consecutive patients undergoing MIPD (543 laparoscopic, 258 robot-assisted, 151 hybrid) in 26 centers from 7 countries between 2007 and 2017.22 The project was initiated by the European Consortium on Minimally Invasive Pancreatic Surgery (E-MIPS) and supported by the Scientific and Research Committee of the European-African Hepato-Pancreato-Biliary Association (E-AHPBA). The TRIPOD guidelines for multivariable prediction models were followed for the design, validation and reporting of this clinical prediction model.25 Patient consent was waived by the Institutional Review Board at the Amsterdam UMC, location AMC (the Netherlands), because of the observational nature of this study.

Patients and data collectionWe included adult patients with an indication for elective MIPD (laparoscopic, robot-assisted, or hybrid). No other exclusion criteria were applied. Methods for data collection and the definitions used have been described elsewhere.22 All data were collected via an

106 | Chapter 7

International Conference on Harmonization Good Clinical Practice (ICH-GCP) compliant on-line electronic case report form (eCRF) and data storage environment (CASTOR®, CIWIT B.V., Amsterdam, the Netherlands).

Outcome, predictors and definitionsThe primary outcome of the study was grade B/C POPF (according to the 2005 ISGPS classification2) within 30 days postoperatively. Due to the retrospective nature of the data, the 2016 ISGPS definition could not be applied in the MIPD dataset. The assessed potential preoperative predictors were type of approach (laparoscopic, robot-assisted, hybrid), annual MIPD volume, sex, age, BMI, Eastern Cooperative Oncology Group (ECOG) performance status26, American Society of Anesthesiology (ASA)-physical status27, Charlson comorbidity index28, diabetes mellitus, prior abdominopelvic surgeries, preoperative diagnosis, malignant suspicion, pancreatic duct diameter, and vascular- or organ involvement on preoperative CT. Laparoscopic PD was defined as laparoscopic resection and laparoscopic reconstruction and robot-assisted PD as robot-assisted resection and robot-assisted reconstruction, regardless of conversion to open. Hybrid PD was defined as laparoscopic resection with planned open anastomoses.29

Statistical analysisAll analyses were performed using STATA 14.2 (StataCorp LP, College Station (TX), United States), supervised by a biostatistician (ES). Multiple imputation (by chained equations, five permutations) was applied to address missing data and improve overall accuracy.30 Categorical variables were presented as counts and proportions and continuous variables as medians with interquartile range or means with standard deviations, as appropriate. For continuous variables, differences between groups were tested with Student’s t test, applying the central-limit theorem. Fisher’s exact test was used for proportions in all cases. All confidence intervals were 95% and an alpha of 0.05 was used for statistical significance.

Model performance was assessed according to the area under the receiver-operating-curve (AUC; discrimination) and calibration plots, in line with current recommendations.31, 32 The calibration plot presents the predicted versus the observed risk of POPF per risk decile, based on the a-FRS. Potential patient- and intraoperative characteristics that could improve the model were identified using a univariable screen. All variables with a significant (P

7

Updated Alternative Fistula Risk Score to Include Minimally-Invasive PD | 107

discrimination was also assessed separately in laparoscopic-, robot-assisted-, and hybrid MIPD. The updated a-FRS was also externally validated (cross-validation) in the original a-FRS design dataset consisting of 1,924 open pancreatoduodenectomy (OPD) patients collected according to the 2005 ISGPS definition (Dutch Pancreatic Cancer Audit and University Hospital Southampton NHS Foundation Trust) and in the validation dataset consisting of 555 OPD patients collected according to the 2016 ISGPS POPF definition (Pancreas Institute, University of Verona Hospital Trust).4 Finally, a post-hoc validation of the original-FRS5 was also performed in the current dataset.

RESULTS

Cohort characteristicsIn total, 202 of 952 patients (21%) developed a grade B/C POPF after MIPD (Figure 1). Half of the patients were male, and the mean age was 64 years, with a mean BMI of 25. Most patients (82%) had ECOG status 0 or 1, and 29% had prior abdominal surgery, with the most common procedures being appendectomy (11%), hysterectomy/adnex extirpation (6.4%), or cholecystectomy or bile duct resection (6.3%). The most common indications for MIPD were pancreatic ductal adenocarcinoma (39%), intraductal papillary mucinous neoplasm (IPMN)/mucinous cystic neoplasm (11%), distal cholangiocarcinoma (9%), pancreatic neuroendocrine tumors (PNET) (5.9%), or intestinal adenoma (4.6%). Intra-operatively, 61% of patients were found to have a soft pancreas, 5% had vascular involvement, and the mean pancreatic duct diameter was 4 mm. Mean estimated blood loss was 390 mL, an abdominal drain was used in 92%, and octreotide was prescribed perioperative in 57% of patients.

Initial model validation The original a-FRS had moderate discrimination (AUC 0.68, 95-CI 0.59-0.76) for grade B/C POPF after MIPD. Calibration was inadequate, with systematic underestimation of the risk of POPF. Initial univariable screening revealed that patients with POPF (n=202) were operated in centers with lower annual case volume (mean 16 vs. 19, P=0.002), were more often male (62% vs. 48%, P

108 | Chapter 7

Fig

ure

1. S

tudy

Flo

w-C

hart

All

fistu

lae

wer

e gr

aded

acc

ordi

ng to

the

ISG

PS 2

005

defin

ition

.2 *

Out

com

e (P

OPF

B/C

fist

ula,

ISG

PS 2

005)

imp

uted

7

Updated Alternative Fistula Risk Score to Include Minimally-Invasive PD | 109

laparoscopic MIPD compared to other MIPD techniques (16% laparoscopic, 13% robot-assisted, 1% hybrid).

Table 1. Univariable Screen of Preoperative Variables of 921 patients after MIPD

No POPF(n=719)

POPF(n=202)

P

Hospital

Annual center volume *, mean (SD) 18.9 (12.4) 16 (8.7) 0.002

Center cohort volume **, mean (SD) 38.1 (34.5) 34.8 (28.5) 0.217

Baseline - - - - -

Male sex, no. (%) 346 (48) 126 (62)

110 | Chapter 7

Table 2. Univariable Screen of Intraoperative Variables of 921 patients after MIPD

No POPF(n=719)

POPF(n=202)

P

MIPD Approach, no. (%)

Laparoscopic MIPD 399 (56%) 113 (56%) 0.936

Robot-assisted MIPD 206 (29%) 52 (26%) 0.478

Hybrid MIPD 114 (16%) 37 (18%) 0.392

Conversion, no. (%) 114 (16%) 29 (14%) 0.661

Operative time, mean (SD), minutes 430 (±120) 433 (±139) 0.812

Estimated blood loss, mean (SD), mL 375 (±457) 466 (±515) 0.022

Perioperative transfusion, no. (%) 83 (12%) 31 (16%) 0.229

Additional resections, no. (%) - - - - -

Venous 28 (3.9%) 6 (3.0%) 0.675

Multivisceral 16 (2.3%) 4 (2.0%) >0.99

Type of anastomosis, no. (%) - - - - -

PJ - non-single row* 468 (65%) 109 (54%) 0.005

PJ - single row 80 (11%) 44 (22%) 0.99

Other 43 (6.0%) 13 (6.4%) 0.868

Soft pancreas, no. (%) 344 (56%) 132 (79%) 0.99

* This included the following techniques for PJ: Blumgart (transpancreatic U-suture), duct-to-mucosa, and dunking/intussusception. Abbreviations: PG indicates Pancreatogastrostomy; PJ, pancreatojejunostomy.

Model update and cross-validationUpon testing all univariable significant variables, only male sex (OR 1.9, 95-CI 1.4-2.7), ECOG status ³ 2 (OR 2.4, 95-CI 1.2-4.7), and single-row pancreatojejunostomy (OR 4.6, 95-CI 2.8-7.6) were associated with POPF after adjusting for the original a-FRS predictors (Table 3). Addition of these factors led to better model discrimination (AUC 0.75 vs. 0.68, P

7

Updated Alternative Fistula Risk Score to Include Minimally-Invasive PD | 111

Table 3. External Validation and Model Adjustment of the a-FRS

a-FRS Model extension Updated a-FRS

OR 95% CI P OR 95%CI P OR 95%CI P

Area under the curve (AUC) 0.68‡ 0.59-0.76 - 0.75‡ 0.72-0.79 - 0.75§ 0.71-0.79 -

Soft pancreatic texture 2.58 1.80-3.69 - 2.58 1.80-3.69 - 2.58 1.80-3.69 -

Duct size, per mm increase* 0.68 0.61-0.76 - 0.68 0.61-0.76 - 0.68 0.61-0.76 -

BMI, per kg/m2 increase 1.07 1.04-1.11 - 1.07 1.04-1.11 - 1.07 1.04-1.11 -

Single row¶ sutures in LPD - - - 4.58 2.75-7.62

112 | Chapter 7

Figure 3. Predicted versus Observed Risk of Pancreatic Fistula Based on the a-FRS with and without Adjustment for Anastomotic Technique, per Risk Decile (10%)

Calibration plot of the a-FRS with (solid line) and without (dashed line) adjustment for anastomotic technique. The circles and triangles indicate the observed frequencies by deciles of predicted probabilities. Perfect calibration would be displayed by a slope being exactly on the diagonal (45 degrees) reference line. This LOESS curve is a locally weighted polynomial regression.

The ua-FRS formula is:

𝑃= exp (−2.36+0.95 [Soft texture] −0.39[𝑃𝐷𝑠𝑖𝑧𝑒] +0.07[𝐵𝑀𝐼] +0.64[𝑚𝑎𝑙𝑒 𝑠𝑒𝑥] )

1+exp (−2.36+0.95 [Soft texture] −0.39[𝑃𝐷𝑠𝑖𝑧𝑒] +0.07[𝐵𝑀𝐼] +0.64[𝑚𝑎𝑙𝑒 𝑠𝑒𝑥] )

Upon (external) cross-validation on OPD in the original a-FRS design dataset4 discrimination remained stable for the 2005 ISGPS definition (AUC 0.73, 95CI 0.69-0.78) and the 2016 ISGPS definition (AUC 0.76, 95CI 0.72-0.81). Of note, discrimination of the original-FRS (Fistula Risk Score with blood loss) was similar to the a-FRS before the update (AUC 0.69, 95CI 0.61-0.76), see supplement 4.

The ua-FRS was made available on www.pancreascalculator.com. Given the increased risk of POPF associated with single-row pancreatojejunostomy in laparoscopic MIPD (OR 5.0, 95CI 3.0-8.2), the a-FRS predictions do not apply to this combination of approach and anastomotic technique.

7

Updated Alternative Fistula Risk Score to Include Minimally-Invasive PD | 113

DISCUSSION

This study describes the update and validation of the a-FRS to predict POPF, in a large pan-European cohort of 952 patients undergoing MIPD. After addition of male sex as a predictor, besides pancreatic texture, duct size and BMI, the updated a-FRS performed equally well in the subgroups of laparoscopic, robot-assisted, and hybrid MIPD as well as OPD. In addition, we found single-row pancreatojejunostomy, as compared to other anastomotic techniques, to be independently associated with a substantially increased risk (OR 4.6) of POPF in laparoscopic MIPD. This technique should be discouraged until proven safe.

The higher rate of POPF in MIPD has also been reported in previous studies20, 21 and could potentially be explained (i.e. confounded) by the use of single-row pancreatojejunostomy in MIPD.22 The inferiority of single-row pancreatojejunostomy in laparoscopic, but not in robot-assisted MIPD, could be caused by the increased technical difficulty. This hypothesis is supported by the fact that such effects of single-row sutured anastomoses are not reported for OPD. Furthermore, in a recent RCT on laparoscopic MIPD vs. open PD (using a dunking PJ)19, 35 and a large cohort study on robot-assisted MIPD vs. open PD (using modified Blumgart pancreatojejunostomy) no differences in POPF rates were seen.36 The heterogeneity in pancreatojejunostomy techniques in MIPD could explain why the initial a-FRS performed suboptimal. At this time, 17 out of 19 (89%) participating E-MIPS centers have moved away from using single-row PJ in laparoscopic MIPD. The remaining 2 centers only apply it in patients with very low risk of POPF.

The significant association between male sex and higher risk of POPF was already found in the a-FRS design study, as well as in a large national registry analysis.4, 37 However, sex was eventually not included in the a-FRS as it did not improve prediction accuracy (discrimination and calibration).38 In the current study, male sex was indispensable in order to achieve adequate performance in all subgroups of MIPD. Others have suggested that sex could be a proxy for the risk of a fatty pancreas, due to the different abdominal fat distribution between males and females, which influences the risk of POPF.39 Because abdominal fat distribution is also related to the difficulty of the reconstruction phase during MIPD40, sex may disproportionally increase POPF risk in MIPD compared to OPD, which may explain its predictive value in this study. This is also supported by the univariate analysis of this study in hydrid PD. These patients do undergo open reconstruction, but male sex was not a significant predictor in this subgroup. In the future, direct measurement of abdominal fat distribution or fatty pancreas could potentially improve the prediction of POPF.41 The inclusion of measures for abdominal fat distribution or fatty pancreas in prediction models is impractical at

114 | Chapter 7

this time.42 Poor ECOG performance status was also a predictor for POPF but did not improve the a-FRS in a meaningful way.

This study has some limitations. First, relevant predictors for POPF could have been missed due to the retrospective nature of the data. This may have resulted in an incomplete model or sub-optimal predictions. However, most important clinically relevant and universally available predictors were included and model coefficients remained stable across various validation and test datasets. Second, the heterogeneity in treatment and data collection among the 26 centers may have resulted in misclassification of exposures and outcome. We have no reason to believe misclassification is differential, which makes it unlikely that model discrimination was affected, although it does reduce the precision of the model coefficient estimates. Third, the ISGPS 2005 POPF definition was used in the original cohort and it was impossible to retrospectively reclassify the POPFs to the ISGPS 2016 definition in the current study.25 However, both the original a-FRS study and our own post-hoc validation found that the a-FRS model performed at least as good when the ISGPS 2016 definitions are used in OPD.4 We have no reason to believe this is different in MIPD patients. Fourth, pancreatic texture only becomes available during surgical exploration and may be influenced by the approach (open vs. minimally invasive). However, we believe surgeons are adequately equipped to estimate pancreatic texture also during the minimally invasive approach.

This study using a large, multinational cohort is the first to validate the a-FRS for MIPD. Continuous validation and coefficient updates are essential for the validity and usability of prediction models. The current validation and model update is especially justified since MIPD is a surgical approach to pancreatic resection with increasing popularity. Given the sample size of 952 patients and 202 events, the study is adequately powered for reliable assessment of prediction model performance (> 100 events are needed for adequate validation).43, 44 Moreover, well-established guidelines for risk model validation and handling of missing data were followed.30–32, 45 The multicenter setting, broad inclusion criteria, and wide time interval support the external validity of these findings for other centers and patients. Nevertheless, the authors strongly believe that MIPD should be reserved for (very) high-volume pancreas centers, as previous studies have revealed strong volume-outcome correlations for MIPD.46, 47

Future studies should investigate the apparent association between single-row anastomosis and the increased risk of POPF in MIPD. Although a causal relationship has not been (prospectively) established, in our opinion, single-row laparoscopic pancreatojejunostomy should be omitted, at least in laparoscopic MIPD, until proven safe. As with any clinical risk prediction tool, the a-FRS needs to be adjusted and re-validated to warrant adequate risk prediction in the future. When kept up-to-date, as done in this study, the updated a-FRS could offer a valuable tool to target preventive

7

Updated Alternative Fistula Risk Score to Include Minimally-Invasive PD | 115

strategies for POPF in high-risk patients. The updated a-FRS is both validated and optimized for POPF risk prediction in

MIPD, by including sex and adjusting based on anastomotic technique. The a-FRS can be used for preoperative risk prediction and post-hoc risk adjustment in both open PD as well as all subgroups of MIPD.

Acknowledgements

We acknowledge the office of the European-African Hepato-Pancreato-Biliary Association for

supporting this study.

Collaborators

We are grateful for the contributions made by Sebastiaan Festen (OLVG, Amsterdam), Mustafa

Kerem (Gazi University, Ankara), Thijs de Rooij (Amsterdam UMC, location AMC), Edward Willems

(Groeninge Hospital, Kortrijk), Régis Fara (Hôpital Européen Marseille), Patrick Pessaux (Institut

Hospitalo-Universitaire de Strasbourg), Bergthor Björnsson (Linköping University), Stefano Berti

(S. Andrea Hospital, La Spezia), Daan Lips (MSTv, Enschede), Carlo Lombardo (University of Pisa),

Misha Luyer (Catharina Hospital, Eindhoven), Alberto Manzoni (Istituto Ospedaliero, Brescia), Izaäk

Q Molenaar (UMCU, Utrecht), Eduardo Rosso (Istituto Ospedaliero, Brescia), Franky Vansteenkiste

(Groeninge Hospital, Kortrijk), Bert Bonsing (LUMC, Leiden), Bas Groot Koerkamp (Erasmus MC,

Rotterdam).

116 | Chapter 7

REFERENCES

1. Bassi C, Marchegiani G, Dervenis C, et al. The 2016 update of the International Study Group (ISGPS) definition and grading of postoperative pancreatic fistula: 11 Years After. Surgery. 2017;161(3):584-591.

2. Bassi C, Dervenis C, Butturini G, et al. Postoperative pancreatic fistula: an international study group (ISGPF) definition. Surgery. 2005;138(1):8-13.

3. Hackert T, Werner J, Büchler MW. Postoperative pancreatic fistula. Surgeon. 2011;9(4):211-217.

4. Mungroop TH, van Rijssen LB, van Klaveren D, et al. Alternative Fistula Risk Score for Pancreatoduodenectomy (a-FRS). Ann Surg. 2017;ePub:1.

5. Callery MP, Pratt WB, Kent TS, et al. A prospectively validated clinical risk score accurately predicts pancreatic fistula after pancreatoduodenectomy. J Am Coll Surg. 2013;216(1):1-14.

6. Shubert CR, Wagie AE, Farnell MB, et al. Clinical Risk Score to Predict Pancreatic Fistula after Pancreatoduodenectomy: Independent External Validation for Open and Laparoscopic Approaches. J Am Coll Surg. 2015;221(3):689-698.

7. Grendar J, Jutric Z, Leal JN, et al. Validation of Fistula Risk Score calculator in diverse North American HPB practices. Hpb. 2017;19(6):508-514.

8. De Rooij T, Lu MZ, Steen MW, et al. Minimally Invasive Versus Open Pancreatoduodenectomy: Systematic Review and Meta-analysis of Comparative Cohort and Registry Studies. Ann Surg. 2016;264(2):257-267.

9. Langan RC, Graham JA, Chin AB, et al. Laparoscopic-assisted versus open pancreaticoduodenectomy: Early favorable physical quality-of-life measures. Surg (United States). 2014;156(2):379-384.

10. Kim SCC, Song KBB, Jung YSS, et al. Short-term clinical outcomes for 100 consecutive cases of laparoscopic pylorus-preserving pancreatoduodenectomy: improvement with surgical experience. Surg Endosc. 2013;27(1):95-103.

11. Iglesias ÃM, Poves I, Burdı ÃF. Comparison of Perioperative Outcomes Between Laparoscopic and Open Approach for Pancreatoduodenectomy The PADULAP Randomized Controlled Trial. 2018;XX(Xx):1-9.

12. Orti-Rodriguez RJ, Rahman SH. A comparative review between laparoscopic and robotic pancreaticoduodenectomies. Surg Laparosc Endosc Percutan Tech. 2014;24(2):103-108.

13. Correa-Gallego C, Dinkelspiel HE, Sulimanoff I, et al. Minimally-invasive vs open pancreaticoduodenectomy: Systematic review and meta-analysis. J Am Coll Surg. 2014;218(1):129-139.

14. Edwin B, Sahakyan MA, Abu Hilal M, et al. Laparoscopic surgery for pancreatic neoplasms: the European association for endoscopic surgery clinical consensus conference. Surg Endosc Other Interv Tech. 2017;31(5):2023-2041.

15. De Rooij T, Klompmaker S, Abu Hilal M, et al. Laparoscopic pancreatic surgery for benign and malignant disease. Nat Rev Gastroenterol Hepatol. 2016;13(4):227-238.

16. Ricci C, Casadei R, Taffurelli G, et al. Minimally Invasive Pancreaticoduodenectomy: What is the Best “Choice”? A Systematic Review and Network Meta-analysis of Non-randomized Comparative Studies. World J Surg. 2018;42(3):1-18.

17. Kendrick ML, Cusati D. Total laparoscopic pancreaticoduodenectomy: feasibility and outcome in an early experience. Arch Surg. 2010;145(1):19-23.

18. Boone B a a, Zenati M, Hogg MEE, et al. Assessment of Quality Outcomes for Robotic Pancreaticoduodenectomy: Identification of the Learning Curve. JAMA Surg. 2015;15232:1-7.

19. Palanivelu C, Rajan PS, Rangarajan M, et al. Evolution in techniques of laparoscopic pancreaticoduodenectomy: a decade long experience from a tertiary center. J Hepatobiliary Pancreat Surg. 2009;16(6):731-740.

20. Boggi U, Napoli N, Costa F, et al. Robotic-Assisted Pancreatic Resections. World J Surg. 2016;40(10):2497-2506.

21. Lai ECH, Yang GPC, Tang CN. Robot-assisted laparoscopic pancreaticoduodenectomy versus open pancreaticoduodenectomy - A comparative study. Int J Surg. 2012;10(9):475-479.

22. Klompmaker S, Van Hilst J, Wellner UF, et al. Outcomes After Minimally-invasive Versus Open Pancreatoduodenectomy A Pan-European Propensity Score Matched Study for the European consortium on Minimally Invasive Pancreatic Surgery (E-MIPS). 2018;ePub.

7

Updated Alternative Fistula Risk Score to Include Minimally-Invasive PD | 117

23. Dokmak S, Ftériche FS, Aussilhou B, et al. Laparoscopic pancreaticoduodenectomy should not be routine for resection of periampullary tumors. J Am Coll Surg. 2015;220(5):831-838.

24. Laaninen M, Sand J, Nordback I, et al. Perioperative hydrocortisone reduces major complications after Pancreaticoduodenectomy a randomized controlled trial. Ann Surg. 2016;264(5):696-702.

25. Moons KGM, Altman DG, Reitsma JB, et al. Transparent Reporting of a multivariable prediction model for Individual Prognosis Or Diagnosis (TRIPOD): Explanation and Elaboration. Ann Intern Med. 2015;162(1):W1.

26. Oken M, Creech R, Tormey D, et al. Toxicity and response criteria of the Eastern Cooperative Oncology Group. Am J Clin Oncol. 1982;5(6):649-656.

27. Keats AS. The ASA classification of physical status - a recapitulation. Anesthesiology. 1978;49(4):233-236.

28. Mary E. Charlson Kathy L. Ales, Ronald MacKenzie PP. A new method of classifying prognostic comorbidity in longitudinal studies- development and validation. J Chron Dis. 1987;40(5):373.

29. Montagnini AL, Røsok BI, Asbun HJ, et al. Standardizing terminology for minimally invasive pancreatic resection. Hpb. 2017;19(3):182-189.

30. White IR, Royston P, Wood AM. Multiple imputation using chained equations: Issues and guidance for practice. Stat Med. 2011;30(4):377-399.

31. Steyerberg EW. Clinical Prediction Models.; 2008.

32. Moons KGM, Kengne AP, Grobbee DE, et al. Risk prediction models: II. External validation, model updating, and impact assessment. Heart. 2012;98(9):691-698.

33. Andrew J Vickers, Angel M Cronin, Colin B Begg. One statistical test is sufficient for assessing new predictive markers. BMC Med Res Methodol. 2011;11(1):13.

34. Janssen KJM, Moons KGM, Kalkman CJ, et al. Updating methods improved the performance of a clinical prediction model in new patients. J Clin Epidemiol. 2008;61(1):76-86.

35. Palanivelu C, Senthilnathan P, Sabnis SC, et al. Randomized clinical trial of laparoscopic versus open pancreatoduodenectomy for periampullary tumours. BJS. 2017;104(11):1443-1450.

36. McMillan MT, Zureikat AH, Hogg ME, et al. A propensity score-matched analysis of robotic vs open pancreatoduodenectomy on incidence of pancreatic fistula. JAMA Surg. 2017;152(4):327-335.

37. Kantor O, Pitt HA, Talamonti MS, et al. Minimally invasive pancreatoduodenectomy: is the incidence of clinically relevant postoperative pancreatic fistula comparable to that after open pancreatoduodenectomy? Surg (United States). 2018;163(3):587-593.

38. Posada D, Buckley TR. Model selection and model averaging in phylogenetics: advantages of akaike information criterion and bayesian approaches over likelihood ratio tests. Syst Biol. 2004;53(5):793-808.

39. Gaujoux S, Cortes A, Couvelard A, et al. Fatty pancreas and increased body mass index are risk factors of pancreatic fistula after pancreaticoduodenectomy. Surgery. 2010;148(1):15-23.

40. Mathur A, Pitt HA, Marine M, et al. Fatty pancreas: A factor in postoperative pancreatic fistula. Ann Surg. 2007;246(6):1058-1064.

41. Gaujoux S, Rebours V, Sauvanet A. Comments on “Alternative Fistula Risk Score for Pancreatoduodenectomy (a-FRS) Design and International External Validation.” Ann Surg. 2018:1.

42. Mungroop TH, Klompmaker S, Groot Koerkamp B, et al. Added Value of Body Fat Distribution in Predicting Clinically Significant Pancreatic Fistula in the a-FRS Following Pancreatoduodenectomy Currently Unclear. Ann Surg. 2018;ePub.

43. Collins GS, Ogundimu EO, Altman DG. Sample size considerations for the external validation of a multivariable prognostic model: A resampling study. Stat Med. 2016;35(2):214-226.

44. Vergouwe Y, Steyerberg EW, Eijkemans MJC, et al. Substantial effective sample sizes were required for external validation studies of predictive logistic regression models. J Clin Epidemiol. 2005;58(5):475-483.

45. Moons KGM, Kengne AP, Woodward M, et al. Risk prediction models: I. Development, internal validation, and assessing the incremental value of a new (bio)marker. Heart. 2012;98(9):683-690.

46. Adam MA, Thomas S, Youngwirth L, et al. Defining a hospital volume threshold for minimally invasive pancreaticoduodenectomy in the United States. JAMA Surg. 2017;152(4):336-342.

47. Kutlu OC, Lee JE, Katz MH, et al. Open Pancreaticoduodenectomy Case Volume Predicts Outcome of Laparoscopic Approach: A Population-based Analysis. Ann Surg. 2018;267(3).

118 | Chapter 7

Supplement 1. Univariable Screen of Preoperative Variables for Laparoscopic MIPD

No POPF(n=399)

POPF(n=113)

P

Hospital

Annual center volume *, mean (SD) 21.8 (±15.0) 16 (±10.0)

7

Updated Alternative Fistula Risk Score to Include Minimally-Invasive PD | 119

Supplement 1. Univariable Screen of Preoperative Variables for Laparoscopic MIPD (continued)

No POPF(n=399)

POPF(n=113)

P

Type of anastomosis, no. (%)

PJ - non-single row 262 (66%) 50 (44%)

120 | Chapter 7

Supplement 2. Univariable Screen of Preoperative Variables for Robot-assisted MIPD (continued)

No POPF(n=206)

POPF(n=52)

P

Preoperative diagnosis, no. (%)

Ductal adenocarcinoma 84 (41%) 15 (29%) 0.151

Neuroendocrine tumor 15 (7.3%) 3 (5.8%) >0.99

IPMN/ MCN 24 (12%) 9 (17%) 0.351

Intestinal adenoma 3 (1.5%) 3 (5.8%) 0.098

Cholangiocarcinoma 13 (6.3%) 8 (15%) 0.045

Other 32 (16%) 9 (17%) 0.832

Unknown 35 (17%) 5 (9.6%) 0.282

Malignant, no. (%) 118 (57%) 33 (64%) 0.436

Pancreatic duct diameter, no. (%) 3.7 (1.3%) 3.5 (1.3%) 0.28

Intraoperative

Conversion, no. (%) 12 (5.8%) 2 (3.8%) 0.742

Operation time, mean (SD), minutes 496 (±120) 544 (±181) 0.024

Estimated blood loss, mean (SD), mL 346 (±407) 331 (±348) 0.806

Perioperative transfusion, no. (%) 21 (11%) 7 (14%) 0.618

Type of anastomosis, no. (%)

PJ - non-single row 138 (67%) 35 (67%) >0.99

PJ - single row 31 (15%) 3 (5.8%) 0.106

PG 35 (17%) 13 (25%) 0.23

Other 2 (1.0%) 1 (1.9%) 0.492

Soft pancreas, no. (%) 117 (61%) 33 (79%) 0.033

Intra-abdominal drain, no. (%) 205 (100%)

Octreotide perioperative, no. (%) 115 (56%) 27 (52%) 0.642

Supplement 3. Univariable Screen of Preoperative Variables for Hybrid MIPDNo POPF(n=114)

POPF(n=37)

P

Hospital

Annual center volume *, mean (SD) 15 (±6.3) 18 (±5.8) 0.011

Center cohort volume **, mean (SD) 30.6 (±29.4) 41.5 (±30.0) 0.055

Baseline

Male sex, no. (%) 59 (52%) 21 (57%) 0.705

Age, mean (SD), years 63.3 (±13) 63.7 (±13) 0.875

BMI, mean (SD), kg/m2 24.9 (±3.5) 26.9 (±3.9) 0.004

7

Updated Alternative Fistula Risk Score to Include Minimally-Invasive PD | 121

Supplement 3. Univariable Screen of Preoperative Variables for Hybrid MIPD (continued)

No POPF(n=114)

POPF(n=37)

P

ECOG performance status, no. (%)

ECOG 0-1 113 (99%) 35 (95%) 0.149

ECOG 2 1 (0.9%) 2 (5.4%) 0.149

ECOG 3-4 - -

Unknown - -

Charlson comorbidity index, mean (SD) 0.5 (±1.0) 0.4 (±0.5) 0.63

Diabetes mellitus, no. (%) 18 (16%) 7 (19%) 0.621

Prior abdominopelvic surgeries, no. (%) 30 (26%) 8 (22%) 0.826

ASA-Classification, no. (%)

ASA 1 25 (22%)

ASA 2 58 (51%) 25 (68%) 0.089

ASA 3-4 30 (26%) 12 (32%) 0.528

Preoperative diagnosis, no. (%)

Ductal adenocarcinoma 42 (37%) 15 (41%) 0.7

Neuroendocrine tumor 7 (6.1%) 2 (5.4%) >0.99

IPMN/ MCN 22 (19%) 7 (19%) >0.99

Intestinal adenoma 11 (9.6%) 3 (8.1%) >0.99

Cholangiocarcinoma 10 (8.8%) 3 (8.1%) >0.99

Other 10 (8.8%) 6 (16%) 0.223

Unknown 12 (11%) 1 (2.7%) 0.188

Malignant, no. (%) 58 (51%) 21 (57%) 0.574

Pancreatic duct diameter, no. (%) 4.4 (1.2%) 4.5 (1.0%) 0.852

Intraoperative

Conversion, no. (%)

Operation time, mean (SD), minutes 397 (±116) 452 (±92) 0.01

Estimated blood loss, mean (SD), mL 344 (±300) 607 (±455) 0.001

Perioperative transfusion, no. (%) 1 (1.0%) 7 (19%)

122 | Chapter 7

Supplement 4. Discrimination Including the Original Fistula Risk Score