Article
The prediction of adverse cardiovascular events in chronic kidney disease patients with Group LASSO Cox proportional hazard regression
Search Medline for
Authors
-
Sahar Ghasemi
- Department of Psychiatry and Psychotherapy, University Medicine Greifswald, Greifswald, Germany
-
Michael Altenbuchinger
- Computational Biology, Institute of Biology, University of Hohenheim, Stuttgart, Germany
-
Johannes Raffler
- Institute of Computational Biology, Helmholtz Center Munich, Neuherberg, Germany
-
Fruzsina Kotsis
- Institute of Genetic Epidemiology, Faculty of Medicine and Medical Center, University of Freiburg, Freiburg, Germany
-
Inga Steinbrenner
- Institute of Genetic Epidemiology, Faculty of Medicine and Medical Center, University of Freiburg, Freiburg, Germany
-
Peggy Sekula
- Institute of Genetic Epidemiology, Faculty of Medicine and Medical Center, University of Freiburg, Freiburg, Germany
-
Kai-Uwe Eckardt
- Charité-Universitätsmedizin Berlin, Berlin, Germany
-
Matthias Schmid
- Department of Medical Biometry, Informatics and Epidemiology, University of Bonn, Bonn, Germany
-
Hans Grabe
- Department of Psychiatry and Psychotherapy, University Medicine Greifswald, Greifswald, Germany
-
Anna Köttgen
- Institute of Genetic Epidemiology, Faculty of Medicine and Medical Center, University of Freiburg, Freiburg, Germany
-
Wolfram Gronwald
- University of Regensburg, Regensburg, Germany
-
Peter Oefner
- University of Regensburg, Regensburg, Germany
-
Ulla Schultheiß
- University of Freiburg, Freiburg, Germany
-
Helena Zacharias
- Department of Psychiatry and Psychotherapy, University Medicine Greifswald, Greifswald, Germany
| Published: | February 26, 2021 |
|---|
Outline
Text
Chronic kidney disease (CKD) patients are at high risk of experiencing major adverse events, including progression to end-stage kidney disease (ESKD), cardiovascular (CV) events, and death. Timely identification of CKD patients at risk of experiencing future adverse events is a prerequisite for the initiation of targeted treatments, thus lowering patient mortality and morbidity, as well as associated health care costs. The risk prediction for individual CKD patients could be facilitated by employing adverse event risk equations specifically optimized for the CKD setting. We explore the potential of machine learning algorithms to develop novel risk equations for CKD patients. Our study cohort comprises 5,215 CKD patients enrolled in the German Chronic Kidney Disease (GCKD) study, who have been prospectively followed-up for four years. Patient parameters assessed at baseline were used as possible predictors. End-points included a major nonfatal or fatal CV event. To facilitate easy transfer into clinical practice, our set of possible predictors were restricted to subsets of easily accessible variables like phenotype, lifestyle, clinical chemistry, and disease history variables readily available from routine CKD patient examinations. The adverse event risk equations were developed employing the Group LASSO Cox proportional-hazards algorithm and were subsequently tested in a rigorous subsampling approach. Our machine-learning-based adverse event risk equations showed overall good predictive performances, assessed by concordance indices (C statistics) ranging from a minimum of 0.700 to a maximum of 0.815 with a median of 0.748. We compare our risk equation to the 12-year, 8-variable Framingham heart risk equation as a gold standard risk score. As a result, with a mean C statistic of 0.751, our risk equation obtained a significant improvement over the Framingham heart risk equation that had only yielded an average C statistic of 0.681 in the GCKD cohort. The proposed risk equations facilitate the timely identification of CKD patients at risk of experiencing a major adverse CV event, only relying on readily available patient parameters.
The authors declare that they have no competing interests.
The authors declare that a positive ethics committee vote has been obtained.
