Background: Several definitions for pancreatic fistula exist and the reported range of 2-50% underscores this variation. The goal of this study was to compare how well two current definitions are at predicting outcomes that are associated with a pancreatic fistula.
Methods: Participating surgeons entered de-identified data into a web-based database designed to collect pre and post Whipple data. Definitions used were the ISGPF definition: ≥3 days, amylase 3x normal) and Sarr’s definition: ≥5 days, amylase 5x normal, > 30 ml. We assessed the association between these definitions and outcomes of interest using logistic regression to obtain age and gender adjusted Odds Ratios and PPV and NPV. Missing data were excluded.
Results: There were 1,507 cases submitted from 20 institutions and 40 surgeons. The median age was 64 (15-91) with 53.8% males. A PPPD was performed in 76.2%.Median pancreatic duct size was 3 mm (0-15 mm).Duct-to-mucosa anastomosis was performed in 66.4% with internal stenting in 49.9%.Drain placement occurred in 98.0% with median removal on day 7 (0-219 days). A malignant diagnosis was reported in 67.4%. Postoperative morbidity included bile leak (3.5%), intra-abdominal abscess (6.5%), DGE (NPO > 10 days)(12.5%), intra-abdominal bleeding (3.6%), and reoperation (3.5%).Using the ISGPF definition the pancreatic fistula rate was 21.2%.By comparison the Sarr definition yielded a 9.4% fistula rate. With the ISGPF definition duct-to-mucosa anastamosis (vs dunking) was associated with a reduced odds of leak, OR 0.76 (95%CI: 0.59,0.99) but not with Sarrs,OR=1.03 (95% CI: 0.71,1.49).Both leak definitions correlated with an increased LOS, need for percutaneous drain, reoperation, and DGE. Neither was associated with an increased risk of ICU stay or 30 day mortality (Table 1). The PPVs for all outcomes were low (< 30.0%) except for LOS (Table 1).
Conclusion: The pancreatic fistula rate observed significantly differs when using 2 common definitions however the odds of pancreatic fistula related adverse outcomes were similar. The low PPV values do highlight the fact using data from drains alone is not enough to predict poor patient outcomes.
Table 1 Adjusted Odds Ratio (CI) & Predictive values of outcomes based on leak definition

	ISGPF			Sarr
	OR (CI)	PPV, %	NPV, %	OR (CI)	PPV,%	NPV,%
LOS (> 10 day median)	2.2 (1.7, 2.9)	59.8	59.6	4.7 (3.1, 7.1)	77.1	58.9
ICU LOS (>1 day median)	0.9 (0.7,1.2)	30.2	68.8	0.9 (0.6, 1.3)	29.5	68.9
Percutaneous Drain	6.7 (4.5,10.0)	21.6	95.8	8.5 (5.6,13.0)	32.6	94.6
Reoperation	3.0 (1.7,5.2)	7.3	97.5	2.7 (1.4, 5.4)	7.9	96.9
Delayed Gastric Emptying	1.7 (1.2, 2.3)	17.3	88.8	2.9 (1.9, 4.4)	26.4	89.0
30-Day Mortality	1.9 (0.7, 5.0)	2.1	98.9	0.5 (0.1, 4.1)	0.8	98.6