No abstract

No abstract

Aims: Aim of this project was to establish a statistical model predicting the risk of death after 6 months based on the data from the one of the largest European adult HLH patient cohorts. Methods: Data of 97 adult (≥ 18 years of age) patients with newly-diagnosed HLH (from the HLH in Adults Database affiliated with PALG -Polish Acute Leukemia Group) were utilized to construct a prognostic model predicting the risk of death within 6 months from diagnosis. First, two alternative approaches were used: LASSO regression (allowing to select variables important for prognosis) and random forest (finding non-linear relationships). Model performance was evaluated by repeated cross-validation. Then, model explanation methods were used to build an improved model. R software (R Core Team, Vienna, Austria, version 3.5.1) was utilized. Results: Among 97 patients, a slight male predominance was observed (62%, 60/97). Median age at diagnosis was 39 (18-82). 43 (44%) patients survived at least 6 months from diagnosis, while 48 patients (49%) died in this period. HLH was triggered by the following factors: malignancy in 39, infection in 20, and autoimmune disease (MAS syndrome) in 19 patients. In another 19 patients cause of HLH could not be determined. LASSO regression and random forest achieved 76% and 70% cross-validated AUC (Area under the ROC curve), respectively. With variables chosen by LASSO and dependencies discovered by random forest, a logistic regression model based on dichotomized features was created. This model achieved 76% cross-validated AUC with improved accuracy. According to the model, the most important prognostic factor protecting from death is MAS syndrome (OR = 0.04 [95% CI: 0.01-0.38]). Among the other variables included in the model platelet count >50 G/l (OR = 0.35 [0.13-0.96]), total bilirubin <1 mg/dl (OR = 0.36 [0.13-1]) and Hb > 9 g/dl (OR = 0.42 [0.14-1.24]) are the most noteworthy. Summary/Conclusion: Constructed model can help in predicting prognosis of the adult HLH patients. The most important established prognostic factors are MAS syndrome as a triggering factor, total bilirubin, and platelet count with hemoglobin concentration (reflecting the severity of bone marrow failure).

Le syndrome d'anémie mégaloblastique thiamine dépendante est une affection autosomique récessive rare qui associe un diabète sucré, une anémie mégaloblastique et une surdité. Nous rapportons les observations de deux nourrissons, âgés, respectivement, de quatre et de cinq mois, hospitalisés pour acidocétose diabétique. Les examens complémentaires ont montré une anémie mégaloblastique et une thrombopénie. Les taux sériques de la thiamine étaient normaux. Les explorations neurosensorielles ont mis en évidence une surdité de perception bilatérale et une atteinte oculaire. L'évolution sous thiamine orale a été marquée par la correction des anomalies hématologiques et la diminution des doses d'insuline. Aucun effet n'a été noté sur l'atteinte neurosensorielle.