Data Collection

By design, LSOCA participants without ocular opportunistic infections were seen every 6 months. At baseline (study enrollment) and at each follow-up visit, participants were interviewed about medical and treatment histories, and each underwent an ophthalmic examination that included determination of best-corrected visual acuity, determination of CS, Goldmann visual field testing, slit-lamp biomicroscopy, and dilated indirect ophthalmoscopy. Laboratory tests obtained at each visit included complete blood cell count, including leukocyte differential; platelet count; lymphocyte subset analyses; serum creatinine; serum lipid determinations; plasma HIV RNA level; and urine protein level. The method by which proteinuria was determined was not requested from investigators at the time of data collection. Beginning in 2005, blood pressure was measured and smoking status was determined at each visit. A Framingham mortality risk score was calculated for each participant; this score takes into account the following factors: gender, age, systolic blood pressure, treatment for hypertension, total cholesterol, high-density lipoprotein, and smoking status. Continuous covariates were categorized using clinically meaningful cut points, where applicable.

CS was determined using the method of Pelli-Robson. Briefly, participants sat or stood 1 m from the Pelli-Robson contrast sensitivity chart, using a +0.75-diopter lens in addition to their refractive corrections. Participants read the chart until at least 2 letters in a triple were missed. The logCS score was calculated as the total number of letters read correctly minus 3, then multiplied by 0.05. Because CS is linear on a log ₁₀ scale, analyses used log ₁₀ CS. The protocol used in LSOCA does not permit confusion between “C” and “O,” which is consistent with the technique described by Myers and associates but is unlike the protocol used in some other reported studies; this convention will potentially lower the distribution of CS values in LSOCA when compared to established norms in those other studies. This issue has been addressed previously by the SOCA Research Group.

We categorized study participants who died during follow-up on the basis of immediate and contributing causes of death, using information from available death certificates; specifically, we identified whether death was associated with diseases characterized by microvasculopathy (diabetes, cardiovascular disease, renal disease, stroke), liver disease (a major cause of morbidity and mortality among people with AIDS), AIDS-related opportunistic infections and malignancies, other AIDS-related disorders, or trauma.

Definitions

Abnormal CS was defined as logCS <1.5, which corresponds to the lower 2.5th percentile of a normal control population described by Myers and associates. The rationale for use of this threshold in studies of the LSOCA population has been discussed elsewhere. For purposes of this study, HAART was defined as the concurrent use of 3 or more antiretroviral drugs.

A participant was considered to have diabetes mellitus if the condition was self-reported at baseline or during follow-up, or if he or she was on any therapy for diabetes mellitus.

A participant was considered to have hypertension if the condition was self-reported at baseline; if he or she received treatment with antihypertensive medications; or if he or she had a systolic blood pressure >140 mm Hg or a diastolic blood pressure >85 mm Hg on at least 1 LSOCA visit.

A participant was considered to have cardiovascular disease if either coronary heart disease or peripheral vascular disease was self-reported or mentioned in medical records at baseline or if he or she was reported to have a cardiovascular event during follow-up.

A participant was considered to have a history of stroke if the condition was self-reported at baseline or if he or she was reported to have had a cerebrovascular event during follow-up.

A participant was considered to have renal disease if the condition was self-reported at baseline; if he or she had proteinurea of 2+ or more; or if his or her calculated glomerular filtration rate (GFR) was <60 mL/min, using the Modification of Diet in Renal Disease (MDRD) Study equation: 186 × (serum creatinine [mg/dL]) ^−1.154 × (age [y]) ^−0.203 × (0.742, if female) × (1.212, if black).

Data Analysis

Unless otherwise noted, the unit of analysis was the eye. Demographic, medical, and laboratory data were compared between study participants with abnormal CS in at least 1 eye versus those with normal CS in both eyes. CS at baseline was compared to the occurrence of death, both as a categorical variable (abnormal CS vs normal CS) and as a continuous variable; relationships were investigated for all participants and for subgroups based on best-corrected visual acuity (20/20 or better vs worse than 20/20), with and without adjustment for the following factors: CD4+ T-lymphocyte count, plasma HIV RNA level, HAART status (current use vs no current use), and date of CS measurement. We adjusted for date of examination in all multivariate analyses because we observed a secular trend in which there was a small, but significant, increase in mean logCS that may have been related to replacement of old contrast sensitivity charts in some clinics during the latter years of the study.

In a secondary, person-specific (as opposed to eye-specific) analysis, we compared abnormal CS to death after categorizing study participants into the following groups: abnormal CS in both eyes at baseline; abnormal CS in 1 eye at baseline; and abnormal CS in neither eye at baseline.

CS (abnormal vs normal) at each visit was compared to the following factors at the same visit: history of stroke; the presence or absence of diabetes, hypertension, cardiovascular disease, or renal disease; laboratory markers of renal disease (proteinuria, GFR); and Framingham 10-year mortality risk score.

Abnormal CS at baseline and abnormal CS as a time-dependent variable were assessed as risk factors for death in crude comparisons, and time-dependent abnormal CS was assessed as a risk factor for death in several models that adjusted for various combinations of the following potential confounders: HAART status; current and nadir CD4+ T-lymphocyte counts; plasma HIV RNA level; visual acuity; history of stroke; the presence or absence of diabetes, hypertension, cardiovascular disease, or renal disease; Framingham 10-year mortality risk score; and date of CS measurement.

For study participants who died and for whom death certificates were available, causes of, or factors contributing to, death were compared between those with and those without abnormal CS.

Statistical Techniques

Differences were assessed using the Wilcoxon rank sum test for continuous data and the χ ² test for categorical data. Assessment of time-independent and time-varying risk factors for death used Cox regression with staggered entry; time since diagnosis of AIDS was the time metric, and analysis accounted for the correlation between eyes of a given participant. P values were 2-sided and were not adjusted for multiple outcomes or multiple looks. Analyses were performed with both SAS (SAS/STAT User’s Guide, Version 9.2, 2008; SAS Institute, Cary, North Carolina, USA) and Stata (Stata Statistical Software: Release 10, 2007; StataCorp LP, College Station, Texas, USA) statistical packages.