Table 4.

Summary of reported multiple regression analyses in included studies

Study	Type of regression model	Results of analysis			Notes
		Dependent variable	β-coefficient	p-value
			(SE)
predictors of tto values
Brown et al. [30]	OLS regression	VA (Snellen) in BSE	0.37	<0.0001	The following equation was developed from the model:
					Utility value = 0.37 (VA) + 0.514,
Brown et al. [31]	OLS regression	VA (Snellen), 1 'good’ eye	-0.0902	0.001	Significant differences in reported utility values were noted when patients with two 'good’ eyes (bilateral good vision) were compared with those with one 'good’ eye (unilateral good vision).
Brown et al. [32]	OLS stepwise model	VA (Snellen), BSE	NR	<0.0001	A significant relationship was demonstrated between decreasing vision in the BSE and decrements in utility values. This relationship was absent for VA in the WSE.
		VA (Snellen), WSE	NR	0.43
Espallargues et al. [22]	OLS Stepwise model	Distant VA (logMAR), BSE	-0.04	0.686	An association was observed between distant VA in the BSE and TTO scores. Selection criteria for significant predictors were p < 0.1. Age and time since diagnosis were important for TTO values.
			(0.05)
Sharma et al. [34]	OLS model	VA (logMAR), BSE	0.176	<0. 01	VA levels in both the affected eye (p < 0.01) and unaffected eye (p < 0.01) were independently associated with reported utilities. Better vision was associated with higher scores.
predictors of sg values
Lloyd et al. [24]	Mixed model analysis	VA (Snellen), BSE	NR	NR	The VA levels were based on the levels of vision used in the health state cards developed for the study. The authors reported that described states were significant in predicting utility. Further analysis showed that SG values were not associated with a patient’s visual acuity level.
Sharma et al. [34]	Bivariate analysis	VA (logMAR), BSE	0.193	<0. 01	VA levels in both the affected eye (p < 0.01) and unaffected eye (p < 0.01) were independently associated with reported utilities. Better vision was associated with higher scores.
	Multivariate analyses
predictors of hui-3 values (global)
Espallargues et al. [22]	Multiple linear regression	Distant VA (logMAR), BSE	-0.12	0.226	A selection criterion of p < 0.1 was adopted for a backward stepwise regression model of relevant variables. Significant variables were contrast sensitivity, illness (es) of long duration and age.
			(0.43)
	Univariate regression	VA (logMAR), BSE	-0.14	<0.01
			(0.03)
Sahel et al. [26]	Multiple regression	BSE: WSE	NR	NR	The adjusted R-squared showed that 21% of the variance in the global score was due to the VA levels [p < 0.01 (BSE); p = 0.31(WSE)].
		≥20/40: ≥ 20/200	0.6	NR
		≥20/40: < 20/200	0.57	NR
		<20/40: ≥ 20/200	0.41	NR
		< 20/40: <20/200	0.42	NR
predictors of hui-3 values (vision dimension)
Espallargues et al. [22]	Univariate regression	VA (logMAR), BSE	-0.25	<0.01
			(0.26)
	Multivariate analyses	VA (logMAR), BSE	-0.21	<0.01
			(0.04)
Sahel et al. [26]	Multiple regression	BSE: WSE			Authors reported that 36% of the variance in the visual dimension of the HUI-3 score was expressed by the adjusted R-squared value [p < 0.01 (BSE); p = 0.7(WSE)].
		≥20/40: ≥ 20/200	0.75	NR
		≥20/40: < 20/200	0.74	NR
		<20/40: ≥ 20/200	0.42	NR
		< 20/40: <20/200	0.37	NR

Abbreviations:BSE better-seeing eye, EQ-5D euroQol values, HUI-3 health utilities index mark3, logMAR logarithm of minimum angle of resolution, NR not reported, OLS ordinary least square, SE standard error, SG standard gamble values, TTO time-trade-off values, VA visual acuity, WSE worse-seeing eye.