Dural ectasia re-definition in Marfan syndrome

Abstract

Dural ectasia is a major criterion for Marfan syndrome (MFS) diagnosis. However, its definition remains fuzzy with studies on small sample size reporting different criteria to differentiate between Marfan and controls. All MFS patients with a mutation in the Fibrillin-1 gene who underwent a complete CT scan in our center were included in the study. They were matched 1:1 for age and sex with controls and all the measurements of dural ectasia were assessed. Between 2010 and 2016, 96 MFS patients and their matched controls were included. Mean age was 38 ± 14 years with 44% of women. According to the 3 validated definitions of the literature, the area under the curve (AUC) of ROC curves for Marfan diagnosis in our population, were 0.65, 0.61 and 0.71 respectively, with a good sensitivity (72–96%) but a poor specificity (25–64%). We performed an ANOVA followed by a logistic regression on all measurements to build a composite criterion for dural ectasia. Based on this method, we found 6 simple parameters to be measured on CT-scan to allow for MFS diagnosis: sensitivity 87%, specificity 94% and AUC 0.91. We also used a modern approach of machine learning using Decision Tree Algorithm, which showed excellent performance: Sensitivity 97%, specificity 98% and AUC 0.97. Given the importance of an early diagnosis in MFS and the delay for genetic diagnosis (2–3 years), other criteria are mandatory. We offer here both a classic and modern approaches to help for MFS diagnosis based on a CT-scan.