Abstract: FR-PO940
CKD Progression Model (CKD-PM): Development and Validation
Session Information
- CKD Epidemiology, Risk Factors, Prevention - II
November 03, 2023 | Location: Exhibit Hall, Pennsylvania Convention Center
Abstract Time: 10:00 AM - 12:00 PM
Category: CKD (Non-Dialysis)
- 2301 CKD (Non-Dialysis): Epidemiology, Risk Factors, and Prevention
Authors
- Uster, Anastasia, Boehringer Ingelheim International GmbH, Ingelheim, Rheinland-Pfalz, Germany
- Ramos, Mafalda, IQVIA Ltd, Brussels, Belgium
- Gerlier, Laetitia, IQVIA Ltd, Brussels, Belgium
- Muttram, Louise Emma, Boehringer Ingelheim Ltd, Bracknell, Bracknell Forest, United Kingdom
- Steubl, Dominik, Boehringer Ingelheim International GmbH, Ingelheim, Rheinland-Pfalz, Germany
- Frankel, Andrew H., Imperial College Healthcare NHS Trust, London, United Kingdom
- Lamotte, Mark, IQVIA Ltd, Brussels, Belgium
Background
Patients with CKD are at risk of disease progression and related complications. Existing CKD models are limited to the risk of end stage kidney disease (ESKD) and selected cardiovascular disease (CVD) outcomes. Our aim was to develop a CKD-PM, that uses evolving risk factors to project risk of a broad range of CKD related complications.
Methods
Development of the CKD-PM was informed by systematic and targeted literature reviews of relevant models and risk factors for CKD progression and risk of complications.
The result was a patient-level state transition model with KDIGO categories as health states. Evolving risk factors project CKD progression and occurrence of complications, with submodules for: CVD, infections, anemia, diabetes, hypertension, acidosis, hyperkalemia, mineral and bone disorders, hospitalisations, acute kidney injury, cancer and ESKD.
Internal and external model validation was performed by comparing predicted outcomes with those observed in cohort studies, utilizing source study patient characteristics and follow up time. An ordinary least squares (OLS) linear regression line (LRL) was fitted to the data, and the slope was used to categorize the quality of the prediction. Deviation <25% from the perfect prediction (LRL slope =1) was considered accurate (mild deviation), 25-50% deviation was considered moderate, and severe beyond 50%.
Results
Core inputs (mortality, CVD and evolution of risk factors including eGFR and uACR) were sourced primarily from two research groups: Chronic Renal Insufficiency Cohort [CRIC] and Chronic Kidney Disease Prognosis Consortium [CKD-PC]).
Seven large CKD cohorts studies were used to validate the CKD-PM predictions. Both internal and external validation results demonstrated robust modeling properties. All-cause mortality (ACM) was accurately projected, with mild under-prediction, either through direct prediction (hazard ratio by eGFR and uACR levels) or as a composite of CVD mortality, renal and other deaths. CVD mortality was mildly under-predicted in internal validation, and mildly over-predicted in external validation. Projected mean change of eGFR or uACR values and ESKD rates were within an acceptable range compared to external values.
Conclusion
CKD-PM is a comprehensive tool with robust modeling properties demonstrated through internal and external validation.
Funding
- Commercial Support – Boehringer Ingelheim International GmbH