Structural equation modeling (SEM) of kidney function markers and longitudinal CVD risk assessment

Area of Science:

• Nephrology
• Cardiology
• Biostatistics

Background:

• Reduced kidney function is a known risk factor for cardiovascular disease (CVD).
• The optimal method for predicting CVD risk using kidney function markers remains unclear.
• Integration of multiple kidney markers may enhance CVD risk prediction.

Purpose of the Study:

• To compare the predictive performance of different estimated glomerular filtration rate (eGFR) equations and a novel pooled kidney function index derived from structural equation modeling (SEM).
• To determine if integrating multiple kidney function markers improves cardiovascular disease (CVD) risk prediction compared to established eGFR equations.

Main Methods:

• Utilized structural equation modeling (SEM) to create a pooled index of latent kidney function using serum creatinine (eGFRcre), cystatin C (eGFRcys), uric acid (UA), and blood urea nitrogen (BUN).
• Compared the predictive performance of the SEM-based index against established eGFR formulas for 10-year incident CVD risk using the C-statistic and DeLong test in a longitudinal population-based cohort.
• Defined 10-year incident CVD risk using Framingham risk score (FRS) >5% and Pooled Cohort Equation (PCE) >5%.

Main Results:

• The SEM-based estimate integrating eGFRcre, eGFRcys, UA, and BUN demonstrated superior prediction performance for both FRS>5% (C-statistic: 0.70) and PCE>5% (C-statistic: 0.75) compared to other SEM models and eGFR formulas.
• Despite improved prediction, the novel SEM-derived marker did not significantly outperform cystatin C-based eGFR (eGFRcys) in predicting incident CVD risk (DeLong test p-values = 0.88 for FRS>5% and 0.20 for PCE>5%).

Conclusions:

• Structural equation modeling (SEM) is a valuable approach for identifying latent kidney function signatures.
• For predicting incident cardiovascular disease (CVD) risk, cystatin C-based estimated glomerular filtration rate (eGFRcys) may be preferable due to its simpler derivation and comparable performance to more complex integrated markers.