Table 4.

Comparison of performance of a prognostic model that included only clinical covariates versus a prognostic model that included clinical covariates and AGP proteins

Selected Covariates and Prognostic Performance Tests	Models
	Only Clinical		Clinical and AGP Proteins
	Statistic	P Value	Statistic	P Value
Effect estimates, HR (95% CI)a
HbA1c (%)	1.25 (1.16 to 1.34)	6.8 × 10−10	1.31 (1.22 to 1.41)	1.2 × 10−13
eGFR per 10 ml/min per 1.73 m2	0.85 (0.79 to 0.92)	2.4 × 10−5	0.93 (0.86 to 1.00)	4.0 × 10−2
log2ACR	1.37 (1.28 to 1.46)	1.6 × 10−22	1.28 (1.20 to 1.36)	1.1 × 10−13
EFNA4 per quartile			1.27 (1.03 to 1.56)	2.4 × 10−2
EPHA2 per quartile			1.42 (1.16 to 1.75)	7.9 × 10−4
C-statistic±SEMb	0.7971±0.0122	—	0.8162±0.0112	4.0 × 10−3
−2 log-likelihood ratio	2708		2650
AIC	2720		2666
NRI (versus clinical model 1) (95% CI)c			0.19 (0.13 to 0.25)	6.5 × 10−10d

Cox regression models were used to evaluate 10-year ESKD risk in the four combined cohorts (n=745). The covariates selected by backward elimination for clinical model included HbA1c, baseline eGFR, ACR, and cohort indicator out of seven examined covariates (see Supplemental Table 2). The covariates selected by backward elimination for combined clinical and AGP proteins included HbA1c, baseline eGFR, ACR, cohort indicator, and six AGP proteins.

^aThe effects are shown as HRs (95% CIs) per one-quartile change of EFNA4 or EPHA2 (continuous variables).

^bUno concordance statistics with two-sided P values. Null values for C-statistics are 0.5 and 3013 for AIC, respectively. P value versus model 1.

^cRisk categories for ESKD are 0%–4.9%, 5.0%–9.9%, 10.0%–19.9%, and ≥20% over 10 years.

^dP value versus model 1.