Résumé
Early thrombotic thrombocytopenic purpura (TTP) recognition is critical as this disease is almost always lethal if not treated promptly with therapeutic plasma exchanges. Currently, as ADAMTS13 (a disintegrin and metalloproteinase with a thrombospondin type 1 motif, member 13) activity is not widely available in emergency, scores have been developed to help differentiating TTP from other thrombotic microangiopathies (TMAs). The aim of this work was to study the accuracy of these diagnostic scores in the intensive care unit (ICU) setting. Performance of both Coppo and PLASMIC scores was studied in a cohort of adult TMA patients requiring admission to one university hospital ICU from 2006 to 2017. Receiver operating characteristic (ROC) curves were established, and confidence intervals of the area under the curve (AUC) were determined. Multivariate logistic regression analysis was performed to identify parameters specifically associated with TTP, to compare diagnostic scores and to elaborate more accurate diagnostic models. During the study period, 154 TMA patients required ICU admission, including 99 (64.2%) TTP and 55 (35.7%) non-TTP patients. AUC under the ROC curve in predicting TTP was 0.86 (95% confidence interval [CI]: 0.81-0.92) for the Coppo score, 0.67 (95% CI: 0.58-0.76) for the PLASMIC score, and 0.86 (95% CI: 0.81-0.92) for platelet count alone. Platelet count ≤20 G/L, determined as the best cut-off rate for thrombocytopenia, performed similarly to the Coppo score and better than the PLASMIC score to differentiate TTP from non-TTP patients, both using AUC ROC curve and logistic regression. In a monocentric cohort of TMA patients requiring ICU admission, the PLASMIC score had limited performance for the diagnosis of TTP. The performance of the Coppo score was good but similar to a single highly discriminant item: platelet count ≤20 G/L at admission.