The usefulness of drain data to identify a clinically relevant pancreatic anastomotic leak after pancreaticoduodenectomy?

Pancreatic anastomotic leak (leak) remains a persistent problem after pancreaticoduodenectomy (PD). Recent reports indicate a mean occurrence of 10% with a range of 2%-28% of patients. However, valid comparisons for these studies cannot be made because the definition of leak is variable, and many patients deemed to have a leak are not sick. The aim of this study was to determine the meaning of the volume and amylase content of the effluent from surgical drains by comparing these values to actual clinical outcomes. From January 1996 to July 2002, 207 consecutive patients underwent PD. We considered a leak to be present if greater than 30 ml/day of drainage was observed from drains and if that drainage contained an amylase-rich fluid (greater than 5X serum) on or after postoperative day (POD) 5. Cases were then divided into three groups-no leak, chemical leak only (leak but asymptomatic), and a clinical leak group (leak that required therapeutic intervention, reoperation, readmission, or prolonged length of stay). Then the drainage volume and its amylase concentration for every postoperative day were compared between the three groups. There were no operative or hospital deaths, and the mean length of stay (LOS) was 11.2 +/- 6.1 days. Prolonged LOS was set at greater than 17 days (one standard deviation beyond the mean LOS for all cases). Leak was observed in 14% of cases (n = 29) and the patients were subsequently divided into these groups: no leak (n = 178), chemical leak only (n = 12), and clinical leak (n = 17). Surprisingly, the daily drain amylase values did not differ between the chemical leak group and the clinical leak group. The daily volume of drainage on POD 5-8 for the clinical leak group was significantly greater than the volumes of the other two groups, so that a combination of greater than 200 ml/day of drainage on POD 5 with an amylase greater than 5X serum had a positive predictive value (PPV) of 84% and a negative predictive value (NPV) of 99% for a clinically relevant leak. We used broad criteria from drainage effluent to include as many potential leaks as possible. This broad definition of leak selected 14% of the PD patients as having a leak; within this group, all of the clinical complications of leak occurred. By increasing the volume criteria from greater than 30 ml per day to greater than 200 ml per day, the PPV was increased from 59% to 84% while keeping NPV at 99%. Drain data based on the volume and amylase criteria of this study may be useful for early detection of a leak that will have clinical impact. This study's criteria for leak may be a good definition to design a clinical trial.