A biochemical prognostic model of outcome in paracetamol-induced acute liver injury.

Background: The aim of this study was to develop a prognostic model of outcome for patients with paracetamol induced acute liver injury based on admission parameters

Methods: We used a cohort of 97 patients admitted to the Scottish Liver Transplant Unit between 1997 and 1998 to identify biochemical prognostic markers of outcome and thus create a prognostic model. Blood samples were taken on admission for analysis. The model was subsequently validated by testing it on a second cohort of 86 patients admitted between 1999 and 2000.

Results: The following were identified as independent variables of poor prognosis (death/ transplant); phenylalanine, pyruvate, alanine, acetate, calcium, haemoglobin and lactate. A prognostic model was then constructed by stepwise forward logistic regression analysis: (400xPyruvate mmols/L)+(50xPhenylalanine (mmols/L)-(4 x Hemoglobin (g/dL). A value of <16 had an accuracy of 93% in predicting death correctly. When applied to the validation cohort this model had a positive predictive value of 91%, a negative predictive value of 94%, a sensitivity of 91%, and a specificity of 94%. On the same population overall, the positive and negative predictive value of the King's criteria were 94% and 93% respectively, whereas their sensitivity and specificity were 88% and 96% respectively.

Conclusions: Using admission characteristics our model is able to identify patients who die from paracetamol overdose fulminant hepatic failure as accurately as King's College criteria, but at a much earlier stage in their condition.