Existing scoring systems highly effective at predicting organ failure in pancreatitis
Nine clinical scoring systems were able to comparably and accurately predict persistent organ failure in patients with acute pancreatitis, according to recent results.
Researchers evaluated data from training (n=256 patients) and validation (n=397) cohorts in two concurrent prospective studies on acute pancreatitis. Clinical scores for each participant were calculated upon admission and after 48 hours according to nine systems to determine their efficacy in predicting the development of persistent organ failure. Investigators then used the combined scores to establish a series of 12 increasingly complex rules to optimize predictive accuracy.
In the training cohort, 62 patients (24%) developed persistent organ failure, as did 34 patients (9%) in the validation cohort. Ten patients (3.9%), all of whom had organ failure, died while hospitalized in the training cohort. In the validation cohort, there were 14 deaths (3.5%), including 10 patients who developed organ failure.
Scores resulted in areas under the curve (AUC) ranging from 0.62 to 0.84 in the training cohort and from 0.57 to .074 in the validation cohort at admission. Of the nine study systems, the Glasgow score was determined the best classifier in both cohorts at admission. After 48 hours, the AUCs for each of the systems had a similar range (0.64-0.84 for the training cohort; 0.57-0.79 for the validation cohort), and Japanese severity score (JSS) was determined the best classifier.
The 12 predictive rules included six for admission and six for 48 hours following admission. In each set of six the initial two rules used Systemic Inflammatory Response Syndrome, Harmless Acute Pancreatitis Score and Panc 3 values; the following two rules incorporated Bedside Index for Severity in Acute Pancreatitis, Pancreatitis Outcome Prediction, and Ranson scores, and the final rules incorporated Glasgow, Acute Physiology and Chronic Health Examination-II and JSS scores. When investigators applied these rules, accuracy increased to 0.92 in the training cohort and to 0.84 in the validation cohort.
Georgios I. Papachristou, MD, PhD
“Clinical scoring systems performed with moderate accuracy in predicting severe acute pancreatitis on admission, [and] all clinical scores performed with similar accuracy,” Georgios I. Papachristou, MD, PhD, assistant professor of medicine at the University of Pittsburgh Medical Center, told Healio.com. The researchers also noted that, while the rule combination proved more accurate, it was more difficult to use and consequently not as clinically useful.
“We therefore believe that the development of new clinical scores is no longer needed because our ability to predict severe disease in patients with acute pancreatitis cannot be expected to improve significantly unless we change our approach,” the researchers concluded.