C-Reactive Protein as a Predictor of Complicated Acute Pancreatitis: Reality or a Myth?

Introduction C-reactive protein (CRP) has been reported as a predictor of the severity of acute pancreatitis (AP). However, there is conflicting evidence in the literature. The proposed cut-off values and intervals for best prediction include an absolute value of 150 at 48 hours; an absolute value of 190 at 48 hours; and the interval change in CRP of 90 at 48 hours. The current study assesses the value of CRP at different intervals and cut-offs in predicting complicated acute pancreatitis (CAP) and compares its performance against other available predictors like neutrophil to lymphocyte ratio (NLR); Glasgow scoring system and modified CT severity index (MCTSI).  Methods Analysis of prospectively maintained data for index episodes of acute pancreatitis managed in 225 patients over a period of five years (2014-2018) was done. CAP was defined by using revised Atlanta classification and included all the AP patients with local and or systemic complications. It was used as a gold standard. Diagnostic and predictive performance of different biochemical markers and multifactorial scoring systems were determined by analyzing receiving operating curves (ROCs), the area under the curve (AUC), sensitivity, specificity, and predictive values (positive and negative).  Results Out of 225 patients, 122 were female while 103 patients were male. CAP developed in 47 patients (20.9%) while 178 (79.1%) patients had mild AP. Overall, in-hospital mortality rate was 1.8% (n=4). ROC analysis demonstrated that CRP at admission had low discriminatory value (AUC= 0.54, p-value=0.74). CRP at 48 hours had AUC of 0.70 (p-value=0.007). At a cut-off of 150, the positive predictive value (PPV) of 150 was 30 %. The PPV of CRP at 48 hours at a cut-off of 190 was 28%. Interval change in CRP at 48 hours greater than 90 had a PPV of 26 %. Further comparison of CRP with other scoring systems like Glasgow scoring system (AUC= 0.65), NL ratio (AUC=0.54), and MCTSI was performed. Among the single predictors, although, NL ratio showed good sensitivity at a cut-off value of 4.7 (87.23%), however, its discriminatory power was negligible (AUC=0.542, p-value=0.513). The overall best performance was achieved by the MCTSI scoring system at a cut-off of 3 (AUC=0.90, sensitivity=83.33 %, specificity=100%, diagnostic accuracy=94.49%).  Conclusion CRP measured at admission or at 48 hours has a very limited role in the prediction of CAP. Along with other scoring systems, its negative predictive value should be used to predict cases with mild AP which can help in clinical decision making for early discharge or management of such patients on ambulatory care basis. MCTSI scoring system can be used in cases with high suspicion of CAP.