Background: Marfan syndrome is a heritable connective tissue disease. Definitive diagnosis is complex, and requires sequencing of a large gene, FBN1.
Aim: We aimed to develop a simple model to estimate the pre-test probability of Marfan syndrome.
Design: Prospective cross-sectional study.
Methods: We applied diagnostic standards for definitive diagnosis or exclusion of Marfan syndrome in 329 consecutive persons. In 208 persons with random assignment to our derivation group, we performed multivariate logistic regression to assess 14 clinical variables for inclusion in a prediction model with derivation of score points from the estimated coefficients. We created cut-offs to classify low, moderate and high probability of Marfan syndrome. For validation, we applied the model to the remaining 121 persons.
Results: We identified seven variables for inclusion in the final model, where we assigned four score points to ectopia lentis, two points to a family history of Marfan syndrome, and one point to previous thoracic aortic surgery, to pectus excavatum, to a wrist and thumb sign, to previous pneumothorax, and to skin striae. In the derivation group 12, 42 and 92% of persons with low (≤1 point), moderate (>1-3.5 points) or high pre-test probability (>3.5 points) had Marfan syndrome, compared to 12, 57 and 91%, respectively, in the validation group. Positive likelihood ratios were 13.96 and 8.54 in the high probability group of the derivation and validation group, respectively.
Conclusion: A simple prediction model provides evidence for Marfan syndrome. This model can be used to identify patients who require definitive diagnostic work-up.