The prevalence rates for binocular summation, inhibition, and equivalency were calculated from the difference in logMAR binocular visual acuity and logMAR better eye visual acuity using the following criteria (logMAR lines): binocular inhibition (difference ≤ −0.10), binocular equivalent (−0.10 < difference < +0.10), binocular summation (difference ≥ +0.10). A difference of 0.10 in the logMar scores is equivalent to a five-letter (one line) difference in ETDRS. Because previous research has shown that best corrected binocular visual acuity can be predicted with reasonable accuracy by best corrected visual acuity in the better eye alone,
11 we evaluated the predictive accuracy of two regression models, by using monocular visual acuities for predicting whether a participant exhibited equivalent binocular visual acuity or exhibited either binocular summation or inhibition. For the first model (linear regression model), binocular visual acuity was estimated from the multiple linear regression equation relating the logMAR binocular visual acuity (dependent variable) to the logMAR better eye visual acuity and the interocular difference (better eye minus worse eye) in visual acuities (independent variables). The resultant regression equation was then used to predict binocular visual acuity for each participant by using better eye and (better eye minus worse eye) visual acuities. The difference in the predicted binocular visual acuity and the actual measured better eye visual acuity was calculated. Participants were then classified, by using the logMAR line criterion as follows: inhibition (predicted difference ≤ −0.10), equivalent (−0.10 < predicted difference < +0.10), or summation (predicted difference ≥ +0.10). The percentage of participants correctly classified was then calculated, overall and for participants who exhibited equivalency and nonequivalency (i.e., summation or inhibition).
For the second model (logistic model), the likelihood of equivalency versus nonequivalency was estimated from the logistic regression equation relating the likelihood of equivalency versus nonequivalency (dependent variable) to the logMAR better eye visual acuity and the interocular difference in visual acuities (independent variables). As a measure of the predictive accuracy of the logistic model, the probabilities of correct classification were calculated.
In addition, the prevalence of visual impairment (defined to be worse than 20/40) was calculated using binocular visual acuity, better eye visual acuity, and the AMA algorithm for visual impairment. This definition of visual impairment is based on the U.S. Federal Motor Carrier Safety Regulations that requires a corrected or uncorrected visual acuity of 20/40 or better.
12 The AMA algorithm
13 for impairment of the visual system is (3 × better eye value + worse eye value)/4. Rates of concordance and discordance between the three criteria were compared by χ
2 procedures.
Analyses of covariance (ANCOVA) were conducted to compare the NEI-VFQ-25 scores for distance activities, near vision activities, and driving difficulties in participants with summation, inhibition, and equivalent visual acuity. These analyses were adjusted for age, gender, and comorbidity (e.g., cancer, cardiovascular disease, diabetes). When significant differences were found, pair-wise comparisons used the Tukey multiple comparison procedure. Finally, separate multiple linear regression analyses were used to determine the association of visual acuity (binocular, better eye, and the AMA measure) with difficulties in distance activities, near vision activities, and driving, by using scores from the NEI-VFQ-25. For these analyses, both the NEI-VFQ-25 subscales and visual acuity were log transformed to impose normality. All analyses were conducted at the 0.05 significance level, on computer (SAS, Cary, NC).