Fig 1.
All study participants are divided into two parts: Model building set (n = 647, left panel) and Longitudinal set (n = 670, right panel).
Fig 2.
SEM conceptual framework under different assumptions (model 1–5).
We developed five models based on different combinations of kidney biomarkers and demographic variables (sex and age). Circles represent latent variables and squares represent observed variable. BUN: blood urea nitrogen; Cys: cystatin C; eGFRcre: creatinine-based eGFR (eGFRCKD-EPI Cre 2021); eGFRcys: cystatin c-based eGFR (eGFRCKD-EPI Cys); SCr: serum creatinine; UA: uric acid.
Table 1.
Baseline characteristics of participants included in the model-building and the longitudinal setsa.
Table 2.
Factor loadings and goodness-of-fit statistics for the five SEM models.
Fig 3.
The distribution of latent kidney trait from model 2 (SEM 2).
Panel A: Histogram of SEM 2. Panel B: Scatter plot for SEM 2 and eGFRCKD-EPI Cre 2021. Panel C: Scatter plot for SEM 2 and eGFRCKD-EPI Cys. Panel D: Scatter plot for SEM 2 and eGFRCKD-EPI CreCys.
Fig 4.
Distribution of Framingham risk score (FRS) and pooled cohort equation (PCE) in longitudinal set.
Pink- and blue-colored density plots corresponds to FRS and PCE, respectively. The grey dotted line indicates the cut-off value for dichotomization (5%).
Fig 5.
Comparisons of receiver operating characteristics (ROC) curves for Framingham risk score (FRS)>5 for general CVD risk estimation and pooled cohort equation (PCE)>5 for risk estimation of hard atherosclerotic cardiovascular disease in 10 years.
Panel A: Comparisons of the C-statistics for FRS>5 within SEM-based latent kidney traits. Panel B: Comparisons of the C-statistics for FRS>5 with the 2nd SEM model (SEM 2) and eGFR equations. Panel C: Comparisons of the C-statistics for PCE>5 within SEM-based latent kidney traits. Panel D: Comparisons of the C-statistics for PCE>5 with the 2nd SEM model (SEM 2) and eGFR equations. The x-axis shows specificity ranging from 1 to 0, while the y-axis shows sensitivity ranging from 0 to 1. The right-hand table summarizes the C-statistics and its 95% confidence intervals in descending order. CKD-EPI: the Chronic Kidney Disease Epidemiology Collaboration; Cre: creatinine; Cys: cystatin C; eGFR: estimated glomerular filtration rate; MDRD: the Modification of Diet in Renal Disease study; SEM: structural equation modeling.