Article Text

Download PDFPDF

23 A validated model for the prediction of adverse cardiac outcome in patients with fabry disease derived from readily available clinical data
  1. Chris Orsborne1,
  2. Peter Woolfson2,
  3. Joshua Bradley3,
  4. Ana Jovanovic2,
  5. Anna Reid4,
  6. Matthias Schmitt5,
  7. Christopher Miller2,
  8. Thomas Anderson2,
  9. Nik Abidin2,
  10. Laura Bonnett6
  1. 1The Univeristy of Manchester
  2. 2SRFT
  3. 3University of Manchester
  4. 4MCMR
  5. 5North West Heart Centre
  6. 6Department of Health Data Science, University of Liverpool


Background The cardiac manifestations of Fabry disease are common and are the leading cause of death. Recently, we developed a prognostic model for the prediction of adverse cardiac outcome in Fabry disease which incorporated cardiac magnetic resonance (CMR) imaging and performed excellently. However, CMR is not accessible or pragmatic in all patients.

Objective To develop, internally validate, and evaluate the performance of a prognostic model, incorporating readily available clinical data, to generate an individualised risk estimate for adverse cardiac outcome for all patients with Fabry disease.

Methods A retrospective cohort of 406 patients with Fabry disease attending a tertiary Fabry clinic. Median follow-up 5.2 years. Prognostic models were developed using Cox proportional hazards modelling. Outcome was a composite of adverse cardiac events. Model performance was evaluated using discrimination and calibration.

Results Age, interventricular septal dimension, left atrial dimension, atrial fibrillation and a history of transient ischemic attack or cerebrovascular accident, were included in the parsimonious model. The median optimism adjusted c-statistic across the 5 imputed datasets was 0.85 (95% confidence interval 0.82–0.88) and the model-fitting optimism-adjusted calibration slope was 0.91. Model calibration at 3-, 5- and 7-years was poor in moderate and high-risk patients which precluded the development of a risk estimator for future use in clinical practice.

Conclusions Future prognostic models and risk estimators for adverse cardiac outcome in Fabry disease may wish to incorporate CMR imaging biomarkers if they are to guide more standardised, personalized and cost-effective care.

Abstract 23 Table 1

Final optimism-adjusted pooled model coefficients for the parsimonious multivariable for time to the composite outcome

Abstract 23 Figure 1

Survival free of the composite outcome according to predicted probability in the model. Strata = cohorts were divided into four quartiles according to predicted probability.

Abstract 23 Figure 2

Calibration plots for the optimism-adjusted parsimonious multivariable at 1- and 5-years The black line is the observed calibration, and the blue line is the bootstrap optimism- (overfitting-) corrected calibration, both estimated by adaptive linear spline hazard regression. The grey line is the line of identity and represents perfect calibration. Mean |error| is equivalent to the Integrated Calibration Index (ICI) and 0.9 quantile is equivalent to E90.17 Rug plots of the distribution of predicted outcome probabilities are at the top of the panels. 1-year (A), 5-years (B).

Conflict of Interest CO has received research support from Amicus Theraputics

  • Fabry disease
  • Cardiomyopathy
  • Prognostic Models

Statistics from

Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.