Postoperative CRP predicts pancreatic fistula formation after Whipple procedure: An internal validation study
*Pathik Aravind MD, *Leanne Brown MD, *Wendy Feng MD, *Nicholas Peters MD, Sajid A. Khan MD, Kevin G. Billingsley MD, MBA, Ronald Salem MD, John W. Kunstman MD, MHS
General Surgery, Yale School of Medicine, New Haven, CT
Background
Pancreatic fistulae or abscesses (PF/A) are frequent causes of morbidity and readmission after pancreaticoduodenectomy (PD). Elevated C-reactive protein (CRP) is a known biomarker for anastomotic leak following gastrointestinal surgery. We have previously shown correlation of elevated CRP with PF/A formation following PD in an experimental cohort and developed a model for PF/A prediction. This study sought to assess this model in a large, representative validation cohort.
Study Design
Analysis of the previous experimental cohort (n=161) determined maximum CRP?270mg/dL between postoperative day (POD) 2-4 was most predictive of PF/A occurrence, with area under the curve (AUC) of 0.88-0.91. PF were characterized according to consensus guidelines from the International Study Group for Pancreatic Fistula. This model was retrospectively applied to a prospectively maintained database of all patients undergoing PD at a tertiary care center from 7/2016-7/2023. The primary outcome measure was PF/A occurrence. Additional correlative outcomes of interest included known PF risk factors and perioperative variables. Patients were allocated based on CRP level and statistical testing for PF/A correlation was performed.
Results
343 patients comprised the validation cohort. Of these, 26.8% (n=92) developed PF/A, with 11.6% having clinically-relevant PF (consensus grade B/C, CR-PF). 27.4% (n=94) patients had CRP?270 of which 45.7% experienced PF/A compared to 19.7% of those with CRP<270 (p<0.001). Patients with CRP?270 were more likely to have a soft pancreatic texture (79.6% vs. 50.2%, p<0.001), smaller pancreatic duct (3.3mm vs. 4.1mm, p=0.008), and less likely to receive neoadjuvant systemic therapy (38.4% vs. 19.4%, p=0.003). Multivariate regression incorporating these variables as well as age and pathology found CRP?270 to be independently predictive of PF/A occurrence (OR 2.57 [95% CI 1.27-5.21], p<0.001). Among those who developed the primary outcome measure with CRP?270, the most common event was CR-PF (70%); the most common event in those with CRP<270 was biochemical amylase leak (48.7%) without attributable change in their postoperative recovery. Sensitivity analysis limited to only CR-PF/A development found CRP?270 to be strongly predictive using the same multivariate model (OR 4.19 [95% CI 2.10-8.39], p<0.001).
Conclusion
Monitoring of CRP following PD anticipates PF/A occurrence independent of patient-specific risk factors such as gland texture and duct size. The previously-described model utilizing a cutoff of CRP?270 from POD 2-4 was proven highly predictive. In particular, CRP?270 identified patients likely to experience the most severe PF-related complications such as CR-PF. Incorporating routine CRP measurement may facilitate postoperative risk stratification in patients undergoing PD and should be further investigated in a prospective trial.
Back to 2024 Abstracts