Glaucoma, Suspect

Analysis of the association between physical activity (mean daily step count) and rate of macular GCIPL thinning in the PROGRESSA cohort was conducted using a stepwise approach to regression. An initial univariable mixed-effects regression analysis assessed the correlation between mean daily step count and rate of macular GCIPL thinning. A second model was then constructed accounting for age and gender, and then a third final model was constructed accounting for covariates selected a priori with biological plausibility or previously documented correlations with macular thinning. A random intercept per patient was included in all analyses to account for the inclusion of more than one eye per patient (lme4 package in R; R Foundation for Statistical Computing, Vienna, Austria). Covariates for the final analysis included demographic (age and gender), ocular (maximum recorded IOP, baseline macular GCIPL thickness), and cardiovascular traits (maximum recorded systolic blood pressure, presence of diabetes mellitus). A secondary analysis using the final model subanalyzed participants characterized as glaucoma suspects. For descriptive purposes, participants were stratified into tertiles of physical activity based on mean daily step counts during the 7-day study period.
Having demonstrated associations between daily step count and macular GCIPL thinning, we then assessed the correlation between daily step counts and rates of peripapillary RNFL thinning, and we assessed the correlation between alternative measures of physical activity with rates of macular GCIPL thinning (mean time spent doing moderate/vigorous activity per day, mean daily active calorie expenditure). All secondary analyses utilized the same covariates as in the primary analysis and were performed using mixed-effects linear regression analysis. The P value for statistical significance in the secondary analysis of alternative measures of physical activity was set at 0.025 (Bonferroni correction). Physical activity was also correlated with each covariate using univariable linear regression to enable discussion of potential mediating pathways of effect.
Similarly, regression analysis in the UK Biobank was conducted using a stepwise approach. Initial univariable analysis was followed by a second analysis adjusting for age and gender, before a final multivariable regression analysis adjusting for traits with biological or known associations with macular thickness. The following traits were selected for inclusion in the model: age, gender, eye (right vs. left), IOP, cardiovascular traits (hypertension and diabetes), and neurological traits (history of stroke, Alzheimer's disease, or Parkinson's disease). Prior to construction of the final multivariable model, the association of these variables with total macular thickness was tested using multivariable regression. Variable variables demonstrating P > 0.1 were excluded from the final analysis. The measure of physical activity (internally centered and scaled milligravity) was then included in the multivariable modeling. Secondary analysis excluded participants with known ophthalmic conditions.
The summary tabulated data present the mean ± SD for continuous variables and the number and prevalence (percent) for discrete variables in the PROGRESSA study cohort. All continuous variables were internally scaled to z-scores prior to model fitting. Beta coefficients with 95% confidence intervals (CIs) were used to represent the 1-SD increase for continuous variables on the macular thickness.

Health factors included cardiovascular (CVD) disease, non-CVD disease, general health, eye disease, eyesight, and hearing. The variable of cardiovascular disease measured self-reported doctor-diagnosed heart condition that was divided into 9 groups (i.e., high blood pressure or hypertension, Angina, high cholesterol). Self-reported doctor-diagnosed chronic diseases were defined as non-CVD diseases and were classified into 7 groups (i.e., Parkinson’s disease, cancer or a malignant tumor, Asthma). The variable of general health indicated the participant’s self-response to the question ‘‘How is your health in general?”. Their response was divided into five-point scales (ranging from ‘‘poor” to ‘‘excellent”). Self-reported hearing measured the participant’s level of hearing and similar to the general health it was divided into five-point scales (ranging from ‘‘poor” to ‘‘excellent”). Self-rated eyesight was assessed by asking participants the three types of questions “Is your eyesight (using glasses or corrective lenses; if you use them) excellent/very good/good/fair/ or poor”, “How good is your eyesight for seeing things at a distance, like recognizing a friend across the street”, and “How good is your eyesight for seeing things up close, like reading ordinary newspaper print’. Their Responses were divided into five-point scales (ranging from ‘‘poor” to ‘‘excellent”). Finally, self-reported doctor-diagnosed eye condition was defined by the variable of eyesight which was divided into 4 groups (i.e., Glaucoma or suspected glaucoma, diabetic eye disease, macular degeneration)