Failure of the Hong Kong criteria to predict the severity of acute pancreatitis

Summary Conclusion The results of the present study demonstrate that the HK criteria do not provide effective prediction of severity. Background Fan et al. (1) have reported previously that a blood urea (BU)>7.4 mmol/L and/or glucose (BG)>11 mmol/L at the time of admission to hospital detects a severe attack of acute pancreatitis with a sensitivity of 76% and specificity of 75%. However, a similar study conducted in the West of Scotland did not confirm these findings (sensitivity 33% and specificity 83%). The reason underlying this discrepancy in prediction is unclear, but it may be because of differences in the nature of acute pancreatitis between Asian and Western populations. Aims In this study we examined the predictive ability of the Hong Kong (HK) criteria in a patient population similar to that studied by Fan et al. Patients and Methods A consecutive series of 130 patients experienced 135 attacks of acute pancreatitis. One-hundred-and-four (77%) attacks were mild and 31 (23%) severe (including 12 [9.0%] deaths). Eighty-nine (66%) episodes had a biliary etiology. In 19 (14%) of these episodes, the gallstones had a primary ductal origin being associated with recurrent pyogenic cholangitis. Results Median admission BU concentrations were 5.2 mmol/L (range 3.6–32.1 mmol/L) for the mild group and 7.6 mmol/L (range 3.6–28.8 mmol/L) for the severe group. Corresponding values of BG were 7.1 mmol/L (range 2.1–17.9 mmol/L) and 8.4 mmol/L (range 3.6–28.8 mmol/L), respectively. Differences in admission BU concentrations between patients with mild and severe episodes were significant (p=0.0001). However, differences in BG concentrations were not (p=0.16). In the severe group, 14 patients had BU and four patients BG concentrations above the cut-off values. The HK criteria predicted severe acute pancreatitis with a sensitivity of 52% and specificity of 80%. These results compare with values of 79 and 56% for the Ranson criteria and 83 and 60% for the Glasgow score. The best prediction was provided by the APACHE II score 24 h post admission (sensitivity 79%, specificity 82%).