Abstract: TH-PO385
Discovery of CKD Patient Subgroups by Consensus Clustering: The CRIC Study
Session Information
- CKD: Risk Scores and Translational Epidemiology
November 07, 2019 | Location: Exhibit Hall, Walter E. Washington Convention Center
Abstract Time: 10:00 AM - 12:00 PM
Category: CKD (Non-Dialysis)
- 2101 CKD (Non-Dialysis): Epidemiology, Risk Factors, and Prevention
Authors
- Zheng, Zihe, University of Pennsylvania, Philadelphia, Pennsylvania, United States
- Waikar, Sushrut S., Harvard Medical School, Boston, Massachusetts, United States
- Yang, Wei, University of Pennsylvania, Philadelphia, Pennsylvania, United States
- Shafi, Tariq, University of Mississippi Medical Center, Jackson, Mississippi, United States
- Hsu, Chi-yuan, University of California San Francisco, San Francisco, California, United States
- Wilson, Francis Perry, Yale School of Medicine, New Haven, Connecticut, United States
- Chen, Jing, Tulane School of Medicine, New Orleans, Louisiana, United States
- Anderson, Amanda Hyre, Tulane University, New Orleans, Louisiana, United States
- Ricardo, Ana C., University of Illinois at Chicago, Chicago, Illinois, United States
- Isakova, Tamara, Feinberg School of Medicine, Northwestern University, Chicago, Illinois, United States
- Rincon-Choles, Hernan, Cleveland Clinic, Cleveland, Ohio, United States
- Kallem, Radhakrishna Reddy, University of Pennsylvania, Philadelphia, Pennsylvania, United States
- Feldman, Harold I., University of Pennsylvania, Philadelphia, Pennsylvania, United States
- Landis, J. Richard, University of Pennsylvania, Philadelphia, Pennsylvania, United States
Background
CKD is a heterogeneous condition with multiple underlying causes, risk factors, and outcomes. Consensus clustering may reveal CKD subgroups with different risk profiles of adverse outcomes.
Methods
Among 2696 participants in the prospective CRIC Study, we performed unsupervised consensus clustering with K-means, without using outcome information, on 72 baseline characteristics of traditional and novel factors to discover patient subgroups. We calculated the standardized difference of all predictors across subgroups, and used the cut-off of ±0.3 to show key features. We examined the associations of each subgroup with ESRD, cardiovascular diseases (composite of heart failure, MI, stroke and PAD), and death.
Results
Three unique CKD subgroups, identified using only baseline factors, were associated with low (Cluster1, N=1203), medium (Cluster2, N=1098), and high (Cluster3, N=395) risks of the outcomes (Fig 1). Patients in cluster 1 (lowest risk) had lower bone & mineral, diabetes, cardiac, and obesity markers, greater eGFR, and used fewer medications. Patients in cluster 2 had higher diabetes and obesity markers, and used more medications. Patients in cluster 3 (highest risk) had higher bone & mineral (except for lower serum calcium), cardiac (except for lower serum CO2), inflammation (except for lower serum albumin), and kidney markers (except for lower eGFR) (Fig 2).
Conclusion
Consensus clustering discovered distinct subgroups of CKD patients with distinguishing patterns of baseline clinical and laboratory factors yielding markedly different risks of important clinical outcomes. Specific biomarkers featuring high-risk CKD subgroup could provide potential treatment targets.
Funding
- NIDDK Support