Early hemoconcentration is associated with pancreatic necrosis only among transferred patients

Studies evaluating hemoconcentration as a marker of necrosis in acute pancreatitis have reached different conclusions. The aim of this study was to determine the impact of transfer status on the accuracy of hemoconcentration for the prediction of pancreatic necrosis. We prospectively enrolled 339 patients in an observational cohort study from June 2005 to December 2007. Univariate and multivariate logistic regression analyses were used to evaluate the impact of transfer status on the relationship between hemoconcentration and necrosis. Accuracy for prediction of necrosis was measured by the area under the receiver operating characteristic curve. Hemoconcentration was associated with increased risk of necrosis only among transferred patients (odds ratio [95% confidence limits], 3.6 [1.2, 10.8]). The area under the receiver operating characteristic curve for admission hematocrit for prediction of necrosis was 0.78 among the transferred patients versus 0.55 among those with primary admissions (chi2, P < 0.0001). Transferred patients had greater initial severity (median bedside index of severity in acute pancreatitis, 2 vs 1; P < 0.0001), were more likely to have hemoconcentration (44% vs 18%; chi2, P < 0.0001), and experienced increased necrosis (37.5% vs 3.6%; chi2, P < 0.0001) compared with primary admissions. After adjusting for sex, disease severity, fluid resuscitation, and transfer status, hemoconcentration was not associated with necrosis (Wald chi2, P = 0.14). Transfer status is a confounder in the relationship between hemoconcentration and pancreatic necrosis.