Table 3:

The AUC from prediction of 5-year incidence of advanced AMD using baseline AREDS severity scale without or with inclusion of demographics and treatment among patients with baseline AREDS severity scale 5 or above (N=135 patients, 198 eyes)

Analysis Approach	AREDS severity scale only*			AREDS severity scale + demographics + treatment + fellow eye group*			Difference in AUC (95% CI)			P-value
Two Eyes Analysis	AUC	95% CI	Width	AUC	95% CI	Width	AUC	95% CI	Width
Naïve approach	0.719	0.645, 0.794	0.149	0.783	0.709, 0.857	0.148	0.064	0.009, 0.119	0.110	0.02
Nonparametric clustered ROC analysis	0.719	0.641, 0.797	0.156	0.783	0.708, 0.859	0.151	0.064	0.008, 0.120	0.112	0.03
Cluster Bootstrap: percentile-based in SAS	0.722	0.641, 0.793	0.152	0.785	0.706, 0.855	0.149	0.063	0.009, 0.120	0.111	0.03
Cluster bootstrap: BCa in R	0.720	0.640, 0.798	0.158	0.785	0.702, 0.856	0.154	0.065	0.004, 0.121	0.117	NA
Left Eye Analysis (N=102)
Standard logistic regression	0.691	0.583, 0.801	0.218	0.786	0.692, 0.890	0.198	0.094	−0.004, 0.192	0.196	0.06
Right Eye Analysis (N=96)
Standard logistic regression	0.745	0.643, 0.848	0.205	0.823	0.722, 0.924	0.202	0.077	0.006, 0.149	0.143	0.03
Worse eye analysis (N=135)
Standard logistic regression	0.727	0.647, 0.808	0.161	0.772	0.692, 0.853	0.161	0.045	−0.013, 0.103	0.116	0.13

^*Include age, gender, current smoking status, fellow eye status, and randomized treatment group. AREDS severity scale was modelled as categorical measure; BCa=Bias-corrected/accelerated; NA=not available.