Examples of nonstudy fellow eyes of bilateral cases judged by Submacular Surgery Trials Photograph Reading Center personnel to have choroidal neovascularization (CNV) present at baseline or in the past. A, A fellow eye with subfoveal CNV. The photographs show an area of laser treatment inferonasal to the fovea with recurrent subfoveal CNV surrounded by a subretinal hemorrhage. The visual acuity of this eye was 20/250 (approximate Snellen equivalent). B, A fellow eye with a scar and CNV. The photographs show a subfoveal scar surrounded by depigmentation of the retinal pigment epithelium. The visual acuity of this eye was 20/63 (approximate Snellen equivalent). C, A fellow eye with a subfoveal scar. The photographs show a subfoveal scar surrounded by “histo spots.” The visual acuity of this eye at baseline was 20/25 (approximate Snellen equivalent). D, A fellow eye with a small laser scar. The photographs show a scar from earlier laser photocoagulation to a small area just temporal to the fovea. The visual acuity of this eye was 20/32 (approximate Snellen equivalent). E, A fellow eye with a large laser scar. The photographs show a scar from earlier laser photocoagulation to a large area just superotemporal to the fovea. The visual acuity of this eye was 20/250 (approximate Snellen equivalent).
Box-and-whisker summaries30 of distributions of National Eye Institute Visual Function Questionnaire (NEI-VFQ) scores by subscale for all patients enrolled in the Submacular Surgery Trials Group H Trial (n = 225). 0 Indicates the worst score; and 100, the best score. Each box (yellow) is bounded by the 25th and 75th percentile values of the distribution of scores for each subscale. The median score is indicated by the vertical line through the box; and the mean score, by the dot (red). The “whiskers” extend horizontally from the ends of the box to a maximum of 1.5 times the interquartile range (width of the box). Outliers are indicated by open circles.
Box-and-whisker summaries30 of distributions of National Eye Institute Visual Function Questionnaire (NEI-VFQ) scores by subscale separately for unilateral (n = 167) and bilateral (n = 58) cases. The upper (green) bar of each pair indicates distribution of scores for unilateral cases (n = 167); lower (blue) bar, distribution of scores for bilateral cases (n = 58). The legend to Figure 2 provides an explanation of the data in the figure.
Box-and-whisker summaries30 of distributions of physical component summary (PCS) and mental component summary (MCS) scale scores from the 36-Item Short-Form Health Survey. The total group of patients (N = 225, yellow bars) included unilateral (n = 167, green upper bars) and bilateral (n = 58) cases. The legend to Figure 2 provides an explanation of the data in the figure.
Box-and-whisker summaries30 of distributions of Hospital Anxiety and Depression Scale scores by subscale for Submacular Surgery Trials Group H Trial patients. The total group of patients (N = 222, yellow bars) included unilateral (n = 166, green upper bars) and bilateral (n = 56, blue lower bars) cases. The legend to Figure 2 provides an explanation of the data in the figure. However, in this figure, 0 indicates the best score; and 21, the worst score.
Submacular Surgery Trials Research Group. Health- and Vision-Related Quality of Life Among Patients With Ocular Histoplasmosis or Idiopathic Choroidal Neovascularization at Enrollment in a Randomized Trial of Submacular SurgerySubmacular Surgery Trials Report No. 5. Arch Ophthalmol. 2005;123(1):78–88. doi:10.1001/archopht.123.1.78
To (1) summarize vision-targeted and general health-related quality-of-life scores at baseline and quantify the effect of the ophthalmic problem, (2) evaluate the strength of relations between visual acuity and interview scores, and (3) compare scores for patients who also had choroidal neovascular lesions in the fellow eye (bilateral cases) with those of patients who had choroidal neovascularization in only the study eye (unilateral cases) at time of enrollment in a randomized trial of surgical removal of subfoveal choroidal neovascularization, either associated with the ocular histoplasmosis syndrome or of idiopathic origin.
Eligible patients had subfoveal choroidal neovascularization (including some classic choroidal neovascularization) and a visual acuity of 20/50 to 20/800 (Snellen equivalent), inclusive, in the eye to be assigned randomly to surgery or observation. Interviews that incorporated the 39-item version of the National Eye Institute Visual Function Questionnaire (NEI-VFQ) and 2 other instruments were conducted by telephone by trained interviewers before patients enrolled and were assigned randomly to surgery or observation. Information from baseline clinical examinations and fluorescein angiograms interpreted centrally by masked readers was used to classify patients as unilateral or bilateral cases and to provide potential explanations for variability of interview responses using linear regression models.
The median overall NEI-VFQ score was 75 (interquartile range, 60-84). The median scores on individual subscales ranged from 55 (general vision) to 100 (color vision). The visual acuity of the better-seeing eye accounted for much of the variability in scores on most NEI-VFQ subscales; a 3-line difference in visual acuity was associated with a 10-point or greater difference in scores on 5 subscales after adjustment for other characteristics of patients and eyes. Scores on most scales of all 3 instruments differed between unilateral cases (n = 167) and bilateral cases (n = 58). Even after adjustment for visual acuity and reading speed of the better-seeing eye, age, gender, and scores on the other instruments, scores on the NEI-VFQ near and distance activities subscales differed by almost 13 and 10 points, respectively, between unilateral and bilateral cases. Neither age nor gender was an important independent explanatory variable for NEI-VFQ scores.
Unilateral and bilateral cases had vision-targeted health-related quality-of-life scores worse than those published for a reference population without eye disease. Furthermore, despite younger age, better visual acuity, and better short-term visual prognosis, bilateral cases had NEI-VFQ scores at baseline similar to those published for 2 groups of patients with age-related macular degeneration. Unidentified factors, in addition to the visual acuity of the better-seeing eye, affected patients’ perceptions of visual function.
The ocular histoplasmosis syndrome, first described by Woods and Wahlen in 1959,1 is diagnosed frequently by ophthalmologists in the region of the United States in which infection with Histoplasma capsulatum is endemic. The endemic region (“histo belt”) includes most of the Ohio and Mississippi River valleys.2 Because choroidal neovascularization (CNV) typically is diagnosed in patients with the ocular histoplasmosis syndrome when they are in their 30s, 40s, and 50s,3- 9 these patients are not only at risk of loss of vision but also at risk of loss of employment in their usual occupations. Idiopathic CNV commonly is diagnosed in patients at ages similar to those who have the ocular histoplasmosis syndrome.10 Thus, vision-related and general health-related quality of life (HR-QOL) may be affected by the vision problems of these patients.
In 1997, the Submacular Surgery Trials (SST) Research Group began to enroll patients with subfoveal CNV in randomized trials to evaluate surgical removal of the subfoveal lesion (submacular surgery). The first trial initiated was for patients who had subfoveal CNV that was idiopathic or associated with the ocular histoplasmosis syndrome (SST Group H Trial). When the SST was planned, measurement of vision-targeted and general HR-QOL was included in the design. Measurement of HR-QOL has been advocated as an adjunct to clinical measurements in randomized trials of treatment and other study designs in which patient outcomes are assessed.11- 13 When the condition for which treatment is being evaluated is not life threatening and when differences between treatments with respect to clinical outcomes are small, assessment of patients’ perceptions of treatment benefit or harm may be particularly important.14 One of the challenges confronting designers of randomized trials and other prospective studies in which HR-QOL is to be measured, either as a primary or a secondary outcome of interest, is to ascertain the likely distribution of scores at enrollment and, from that distribution, to assess the ability of individual instruments to measure change in the expected direction. Such information is being assembled for ophthalmology patients as studies initiated during the past decade, including the SST, publish data collected for patients with specific diagnoses.
This article (1) reports findings from vision-targeted and general HR-QOL interviews conducted with patients before enrollment in the SST Group H Trial and quantifies the effect of their ophthalmic condition, (2) evaluates the strength of the relation between visual acuity of the better-seeing eye to scores on a multidimensional vision-targeted interview in this group of patients, and (3) compares vision-targeted and general HR-QOL scores between patients who had CNV in only the study eye with those of patients who also had a choroidal neovascular lesion in the (nonstudy) fellow eye.
Members of an independent Data and Safety Monitoring Committee were appointed late in 1996 by the director of the National Eye Institute (NEI) to review findings from a multicenter pilot study for patients similar to those enrolled in the SST Group H Trial. They approved the design and methods proposed for the SST on January 29, 1997, before enrollment of patients was initiated on April 1, 1997. In addition, institutional review boards at all participating institutions reviewed and approved the study design and the consent forms to be used locally. All patients gave signed consent before enrollment and random treatment assignment.
The SST Manual of Procedures15 and the SST Forms Book16 provide detailed information regarding study design, methods, and policies. Only methods pertinent to the goals of this analysis are summarized.
Patients eligible for the SST Group H Trial were identified from referrals to the clinical centers participating in the SST. One or more patients were enrolled at 21 of the 27 SST centers. The principal eligibility criteria for the SST Group H Trial are summarized in Table 1. Eligible patients had subfoveal CNV in at least 1 eye (study eye) either of unknown cause or attributable to the ocular histoplasmosis syndrome and were 18 years or older. The details of eligibility requirements for patients and eyes, clinical evaluation procedures, and clinical data collection methods and schedules are given elsewhere.15,16
As part of the baseline examination, the best-corrected visual acuity of each eye of each patient was measured by an SST-trained and SST-certified vision examiner according to a standard protocol.15,17 The vision examiner also measured the contrast threshold of each eye using Pelli-Robson charts18 and the reading speed of each eye using cards with enlarged text developed for the Macular Photocoagulation Study.19 The visual acuity of the study eye of eligible patients had to be between 20/50 and 20/800 (Snellen equivalent), inclusive. There was no restriction on the neovascular status of the fellow eye; however, the visual acuity of the fellow eye had to be light perception or better. To be eligible for the randomized trial, patients had to agree to complete centrally administered telephone interviews before enrollment and at scheduled times after enrollment. Enrollment and random treatment assignment did not take place until the baseline interview was completed by telephone with an interviewer located at the SST Coordinating Center, Baltimore, Md.
Patients judged eligible by the examining ophthalmologist were invited to participate. After the patient completed the study education process and signed the consent form for the trial, a study identification number and an alphabetic code assigned by the clinical center staff were recorded on the baseline data forms. Following completion of the telephone interview, the data forms were telecopied to the Coordinating Center for preliminary review of eligibility and completeness of baseline data recording. After this review and satisfactory resolution of any eligibility or data-reporting issues, the next treatment assignment was selected automatically from the file of random assignments that had been prepared for that clinical center specifically for this trial. Thus, the patient and the clinical center staff were unaware of the treatment that would be assigned to the study eye, surgery or observation, until an automated message was sent to them by telecopier after the telephone interview had been completed and all baseline clinical data had been received at the Coordinating Center. Baseline photographs of both eyes of the patient were sent to the Photograph Reading Center, Baltimore, as soon as possible after enrollment of the patient.
The interview instruments (standardized questionnaires) selected for the SST were the 39-item version of the NEI Visual Function Questionnaire (NEI-VFQ),20,21 the 36-Item Short-Form Health Survey (SF-36),22 the Hospital Anxiety and Depression Scale (HADS) (14 items),23 and the 3-item SST Vision Preference Value Scale.24 The NEI-VFQ was designed to be applicable to patients with several different vision-limiting or vision-threatening conditions. Initially, the 25-item version of the NEI-VFQ was used with a few additional items; ultimately, the entire 39-item version was used. The 11 individual subscales and the overall NEI-VFQ have possible scores from 0 (worst) to 100 (best).20,21,25 The SF-36 measures general health status and health-related functioning; scores for the 2 summary scales are calibrated to the general population to have mean scores of 50 points and SDs of 10 points.26 The HADS instrument was included in the interview because of anecdotal and published evidence of depression among patients with poor vision27 and anxieties regarding future vision among patients with early vision loss. Possible scores on the 2 HADS subscales range from 0 (best) to 21 (worst).23 The SST Vision Preference Value Scale was incorporated into the interview to measure patients’ perceptions of the value of their current health and vision. Baseline data from the latter instrument are reported elsewhere.24
All interviews were conducted via telephone by trained interviewers using a computer-assisted system. The interview system interacted with the primary study database to confirm study identifiers assigned to the patient and recorded on the baseline data forms so that the interview data could be integrated with the clinical data for each patient. In addition, the system prompted the interviewer regarding items for which a response had not been entered. Because interviews were conducted before enrollment and random treatment assignment, interviewers were masked to the treatment assignment and to clinical data.
In the absence of published data that indicated whether a general instrument or a condition-targeted instrument should be administered first when multiple instruments are used, we decided to randomize the order of administration. A random selection of the first instrument (NEI-VFQ or SF-36) was made at initiation of each interview using an automated procedure. The HADS instrument always was administered immediately after the SF-36 and the SST Vision Preference Value Scale after the NEI-VFQ. The effect of order of administration is reported elsewhere for patients in all SST clinical trials.28 A small, but statistically significant, order effect was observed, with worse scores on mental health scales obtained when the NEI-VFQ was administered first.
An initial training session for interviewers was conducted before the trial began; additional training was provided during interviewer workshops held at least once each year. Training emphasized standard administration of the interview and acceptable interactions with trial patients. In addition, the interviewers met periodically during the first years of the trial to reach agreement on issues of interview administration and interactions with patients and to recommend improvements in the computer-assisted system. The SST Patient-Centered Outcomes Subcommittee met at least twice yearly to review data regarding interview administration and completion.
As required by the eligibility criteria, all patients had active CNV in the study eye, documented by leakage of dye during fluorescein angiography. Baseline stereoscopic photographs of the macula and disc of the nonstudy fellow eye of each patient, together with frames of the film-based stereoscopic fluorescein angiogram taken of the nonstudy eye at 2 and 5 minutes after dye injection, were evaluated promptly on receipt at the SST Photograph Reading Center for presence of a choroidal neovascular lesion. Patients whose nonstudy eye was judged to have active CNV in any macular or peripapillary location, identified by leakage of fluorescein dye on frames of the angiogram, a disciform scar, or a macular lesion resulting from laser photocoagulation or submacular surgery, were defined as bilateral cases. Additional patients for whom no such lesion was identified during central review of photographs but who had been reported by the SST ophthalmologist to have undergone laser photocoagulation or other treatment for CNV in the macula of the nonstudy eye before enrollment in the SST Group H Trial also were classified as bilateral cases. All other patients were defined as unilateral cases.
Scores were calculated using the recommendations of the developers of the questionnaires.23,25,26 However, for the NEI-VFQ interviews, a subscale score was calculated only when at least half the items that made up the subscale had been answered. In addition to classification of patients as “noncases” (score, 0-7), “doubtful [possible but unlikely] cases” (score, 8-10), and “definite cases” (score, ≥11) of anxiety and depression, as proposed by the developers of the HADS, distributions of raw scores also were examined.
Data from all baseline interviews completed by patients who had enrolled in this clinical trial by the time accrual ended (September 30, 2001) were analyzed. Characteristics of unilateral and bilateral cases were compared using χ2 tests for trend when they could be described in ordered categories and χ2 tests for homogeneity when they were dichotomous.29 Distributions of scores from interviews have been summarized for display using box-and-whisker plots.30 Distributions of vision measurements and interview scores were compared between subgroups of patients using the Wilcoxon rank sum test.31 No adjustment has been made for multiple comparisons; however, only statistical tests that yielded P ≤ .01 have been deemed to indicate statistical significance in this report.
Relations between scores on NEI-VFQ subscales (dependent variable) and visual acuity and other baseline covariates (explanatory or independent variables) were explored using correlation analysis32 and linear regression models.33 Stepwise linear regression with forward selection (“step-up” approach) was used to identify covariates within a predefined subset to include in the most parsimonious model (the model with the fewest covariates) that explained the most variability in scores of the overall NEI-VFQ scale and each NEI-VFQ subscale. The baseline covariates evaluated in stepwise regression were the age and gender of the patient, systemic hypertension (present or absent), measurements of best-corrected visual acuity, contrast threshold, reading speed of the better and worse eyes, and either the 2 SF-36 component summary scales or the SF-36 physical component summary (PCS) scale and the HADS subscales. Covariates retained in linear regression models in the stepwise selection procedure for each NEI-VFQ subscale were those that yielded P ≤ .05. Covariates consistently selected for parsimonious models over most NEI-VFQ scales and subscales were included in the multiple linear regression models (final models). The same set of explanatory clinical variables was included in the final models in which the score on each of the SF-36 summary scales and HADS subscales was the dependent variable. The SAS/STAT software (SAS Institute Inc, Cary, NC) was used for all data analyses.
When accrual ended on September 30, 2001, 225 patients had enrolled; 58 patients subsequently were classified as bilateral cases. Examples of neovascular lesions observed in the nonstudy fellow eyes of bilateral cases are shown in Figure 1. Of the bilateral cases, 2 were judged at the Photograph Reading Center to have idiopathic CNV and 56 were judged to have CNV as part of the ocular histoplasmosis syndrome.
Sociodemographic characteristics and visual status of patients are summarized in Table 2 and Table 3, respectively, for all patients who enrolled in the trial and separately for unilateral and bilateral cases. Compared with unilateral cases, bilateral cases were older (mean age, 57 vs 46 years), fewer were employed outside the home, and more had hypertension (P ≤ .01, χ2 tests). Even after adjustment for age, 2.3 times as many bilateral cases as unilateral cases had hypertension at baseline.
Information regarding major comorbidities, other than hypertension and diabetes mellitus which were recorded for all enrollees, was obtained at baseline for only 124 (55%) of the 225 patients who enrolled in the SST Group H Trial. Of these, 78 patients, 57 unilateral cases and 21 bilateral cases, reported having some other major health condition. Altogether, 111 (49%) patients had 1 or more known health problems in addition to CNV and reduced visual acuity in at least the study eye.
In Table 3, the better eye has been defined as the eye with better visual acuity. As expected, the distributions of the visual acuity, contrast threshold, and reading speed of the better eye differed between unilateral and bilateral cases (P < .001, Wilcoxon rank sum tests). Similarly, the distributions of visual acuity and reading speed of the worse eye differed between unilateral and bilateral cases (P < .001, Wilcoxon rank sum tests). Although the distributions of contrast threshold of the worse eye differed somewhat between the 2 groups, the difference was not statistically significant by our criterion (P = .05, Wilcoxon rank sum test).
Distributions of baseline scores from the NEI-VFQ overall scale and subscales are summarized in Figure 2 for the total group of 225 patients and in Figure 3 separately for unilateral and bilateral cases. Many patients had scores at or near the ceiling (best possible score) of most subscales; a few patients, typically bilateral cases, had scores near the floor (worst possible score) of some subscales. The median overall NEI-VFQ score for all 225 patients in the SST Group H Trial was 75 (interquartile range, 60-84); median scores on individual subscales ranged from 55 (general vision) to 100 (color vision). A wide range of scores was observed on all scales (Figure 2); the width of the interquartile range was 25 points or larger for each subscale.
As shown in Figure 3, scores for bilateral cases were worse (lower) than scores for unilateral cases on most NEI-VFQ subscales (P ≤ .001, Wilcoxon rank sum tests, for all subscales except peripheral vision and ocular pain). Differences between median scores of the 2 patient subgroups on subscales for which the distributions differed to a statistically significant degree ranged from 12 points (color vision) to 50 points (driving). For the overall NEI-VFQ, the difference between median scores was 26 points.
Pearson product moment and Spearman rank correlation coefficients for visual acuity of the better eye and scores on NEI-VFQ, SF-36, and HADS instruments are given in Table 4. Correlations with visual acuity of the better eye were statistically significant (P < .001) for all NEI-VFQ subscales except for peripheral vision and ocular pain. Pearson product moment correlation coefficients were 0.4 or larger and Spearman rank correlation coefficients were 0.3 or larger for all other NEI-VFQ subscales.
The baseline covariates retained in the final multiple linear regression models were those selected from stepwise linear regression (visual acuity and reading speed of the better eye, bilateral or unilateral case, and SF-36 component summary scale scores) plus age and gender of the patient. Inclusion of these covariates provided estimates of the independent association of each variable with NEI-VFQ scores after adjustment for the effects of all others in the final model. The estimated coefficient (model parameter or slope estimate) of each explanatory variable in the final linear regression model (except baseline age and gender) is shown for each subscale in Table 5. The coefficient of each covariate measured on a continuous scale can be interpreted as the number of points the score on the specified NEI-VFQ subscale differs, on average, between patients who differ by 1 unit on the measurement scale. The percentage of total variability in scores that was explained by the covariates in the final multiple linear regression model (model R2) also is shown for each NEI-VFQ subscale. In addition, the model R2 is given for the linear regression model with visual acuity of the better eye as the only independent variable (Table 5). Typically, the visual acuity of the better eye was the first independent covariate selected during stepwise linear regression analysis and accounted for 75% or more of the variability in NEI-VFQ scores explained by the multiple linear regression models. Except for peripheral vision and ocular pain, a difference of 1 line in the visual acuity of the better eye was associated with differences of 1.9 to 6.9 points in scores on NEI-VFQ subscales (Table 5). The largest estimated differences in NEI-VFQ scores between unilateral and bilateral cases (Table 5), independent of visual acuity and other explanatory variables, were 9.8 points on the distance activities subscale and 12.9 points on the near activities subscale (P ≤ .01).The reading speed of the better eye was related independently to scores on 5 NEI-VFQ subscales; however, a difference of 20 or more words per minute was required to observe a difference of 2 or more points in NEI-VFQ scores.
A 2-point difference on each SF-36 summary scale was associated with approximately a 1-point difference in NEI-VFQ scores, independently of other covariates (Table 5). However, a 1-point difference on the SF-36 mental component summary (MCS) scale was associated with a 1-point difference on the NEI-VFQ mental health subscale. Although included in the final linear regression models as explanatory variables, patient age at baseline and gender had little or no independent relation to NEI-VFQ scores after adjusting for other covariates.
Distributions of PCS and MCS scale scores from the SF-36 are summarized in Figure 4 for all patients and separately for unilateral and bilateral cases. Mean SF-36 scores for all SST Group H Trial patients were 49.7 on the PCS scale and 49.4 on the MCS scale. Based on published scores for the general population,26 expected mean scores for the population with the same age and gender distribution as SST Group H Trial patients were calculated to be 48.8 for the SF-36 PCS scale and 50.4 for the SF-36 MCS scale. Scores in the central portion (box) of the distributions of SF-36 PCS and MCS scores were slightly lower for bilateral cases than for unilateral cases (Figure 4). A larger shift in distributions was observed for the MCS scale.
Correlation coefficients between visual acuity of the better eye and scores on SF-36 summary scales were smaller than for scores on most NEI-VFQ subscales (Table 4). In the final linear regression models used to estimate independent associations between covariates and scores, visual acuity was related independently to scores on the SF-36 MCS scale but not to those on the PCS scale (Table 6). Based on the estimated regression coefficient, a 3-line difference in visual acuity of the better eye corresponded to a difference of about one third of an SD for the SF-36 MCS score. Baseline covariates included in the model (visual acuity and reading speed of the better eye, unilateral or bilateral case status, age, and gender) explained only 8% of the variability in SF-36 PCS scores and 13% of the variability in SF-36 MCS scores (Table 6).
The distributions of scores from the HADS subscales are summarized in Figure 5 for all patients and separately for unilateral and bilateral cases. Scores on the anxiety subscale were not available for 3 patients, 1 unilateral case and 2 bilateral cases, because of refusal to answer 1 or more items that contribute to that subscale. Unilateral and bilateral cases had more similar scores on the anxiety subscale than on the depression subscale. Depression scores of bilateral cases were worse (higher) than those of unilateral cases (P < .001, Wilcoxon rank sum test). Bilateral cases were classified more often as definite and doubtful cases of anxiety and depression, using the criteria adopted by the HADS developers23; however, as when distributions of raw scores were compared, the difference was statistically significant only for depression (P = .001, χ2 test for trend).
The same set of explanatory variables as used for the SF-36 summary scales was included in the multiple linear regression models in which the HADS subscale scores were the dependent variables. The estimated regression coefficients also are shown in Table 6. As with the SF-36 summary scales, this set of covariates explained only a small portion of the total variability in scores: 10% of the variability in scores on the HADS anxiety subscale and 18% of the variability in scores on the HADS depression subscale.
To our knowledge, baseline interviews of SST Group H Trial patients provide the first data regarding vision-targeted and general HR-QOL collected using standard instruments for patients who have subfoveal CNV either associated with the ocular histoplasmosis syndrome or idiopathic in origin. A wide range of scores was observed on the NEI-VFQ, overall and by subscale. Scores were lowest (worst) on the general vision subscale (median, 55), followed by the mental health (median, 65), driving (median, 67), and near activities (median, 67) subscales. Scores were highest (best) on the color vision (median, 100) and social functioning (median, 92) subscales.
As expected, based on the purpose and design of the NEI-VFQ, scores on this vision-targeted instrument among patients who enrolled in the SST Group H Trial correlated with the visual acuity of the better-seeing eye (Table 4). In multiple linear regression models, the visual acuity of the better eye accounted for much of the difference in NEI-VFQ scores observed between unilateral and bilateral cases (Figure 3 and Table 5). A 3-line difference in baseline visual acuity of the better eye accounted, on average, for a difference of 9 or more points in the overall NEI-VFQ score and scores on 7 of 11 subscales, including a 20-point difference in driving subscale scores (Table 5). The visual acuity of the better eye also was an independent explanatory variable in multiple linear regression models for scores on the SF-36 MCS and HADS depression scales after adjustment for other covariates (Table 6).
Bilateral cases demonstrated poorer HR-QOL than unilateral cases, as measured by a vision-targeted instrument (NEI-VFQ), a generic health status instrument (SF-36), and a mental health–targeted instrument (HADS). Mean scores were worse for bilateral cases than for unilateral cases on the overall NEI-VFQ and all 11 subscales (Figure 3), on the SF-36 summary scales (Figure 4), and on the HADS subscales (Figure 5). Even after adjustment for visual acuity of the better eye and other explanatory variables, bilateral cases had scores on the NEI-VFQ near activities and distance activities subscales that were almost 13 and 10 points worse, respectively, than scores of unilateral cases (Table 5). Physical and mental health, as measured by the SF-36 PCS and MCS scale scores, respectively, accounted for some of the variability in scores on most NEI-VFQ subscales (Table 5). Substitution of scores from the 2 HADS subscales for the SF-36 MCS scale scores in the exploratory linear regression models yielded parameter estimates for other explanatory variables nearly identical to those in Table 5.
Bilateral cases differed from unilateral cases with respect to several baseline measurements and medical history (Table 2 and Table 3). Each of these factors was considered in our exploration of explanatory variables for HR-QOL scores. Apart from hypertension and diabetes mellitus, major comorbidity data were available at baseline for only 55% of patients because collection of these data began in February 1999, nearly 2 years after initiation of the trial. We repeated the linear regression analyses, adding an indicator variable for any known comorbidity to the final models. Any known comorbidity at baseline included hypertension, diabetes mellitus, or any other major health condition reported whenever the additional question was asked. Although this analysis yielded minor changes to the estimated coefficients shown in Table 5, comorbidity was not related to NEI-VFQ scores, after adjustment for other covariates, on any subscale for patients in the SST Group H Trial. This finding agrees with that of Miskala et al,34 who reported that the SF-36 PCS score was at least as important an explanatory variable for NEI-VFQ scores as detailed data on major comorbidity. The SF-36 PCS scale scores may have captured most of the effects of age, gender, and coexisting health problems of SST Group H Trial patients. The contrast threshold of the better eye showed no relation to NEI-VFQ scores, except for the color vision subscale, and, thus, was not included as an explanatory variable in final multiple linear regression models. The influence of visual acuity and reading speed of the worse eye, occasionally selected as covariates in exploratory regression models, may have been captured by the relations with bilateral disease in the final models. The contrast threshold of the worse-seeing eye never was selected in any stepwise linear regression analysis for any NEI-VFQ scale.
Because this investigation was exploratory, several variables were evaluated for their relation to scores on 3 different HR-QOL instruments; regression models were explored for a total of 16 different scales and subscales. Although we have provided parameters estimated in the final multiple linear regression models, these models and estimates are not intended to be interpreted as recommendations for adjustment of scores but merely to provide estimates of the relative strength and direction of the associations in this group of patients. To minimize overstatement of the strength of relations, we elected P ≤ .01 as the criterion for judging statistical significance. Based on sample size guidelines provided by the developers of the NEI-VFQ,25 the number of patients enrolled in this trial provided 80% power to detect 10-point differences between equally sized patient subgroups on most scales. In our analysis, a 3-line difference in visual acuity, considered a clinically significant difference by most ophthalmologists, was estimated to account for a 10-point or larger difference in scores on 5 of 11 NEI-VFQ subscales and the overall score, independently of other explanatory variables in the final model (Table 5). Similarly, the differences in adjusted scores on the NEI-VFQ near and distance activities subscales between bilateral and unilateral cases were almost 13 and 10 points, respectively (Table 5). These estimates met our conservative criterion for statistical significance.
Instrument order was randomly assigned at initiation of the interview for 205 SST Group H Trial patients. For 97 patients, 74 unilateral cases and 23 bilateral cases, the SF-36 was administered first. The effect of instrument order on baseline scores was evaluated using linear regression models. For no scale, including mental health scales, was instrument order associated with scores on any vision-targeted or HR-QOL instrument in this study (P > .05). Thus, we ignored order of administration in our analysis.
The inclusion of a standard vision-targeted instrument (NEI-VFQ) permitted comparison of scores of SST Group H Trial patients with those of other individuals who have different ocular conditions. Table 7 summarizes NEI-VFQ scores for SST Group H Trial patients for comparison with published scores for other ophthalmology patients. The mean age of patients, the percentage who were women, and the mean visual acuity of the better eye are given whenever the information could be found in the publications cited. At enrollment in this clinical trial, patients had poorer scores on all subscales than the reference group (no eye condition except possibly refractive error) who participated in the NEI-VFQ Field Test (column K, Table 7),21 regardless of whether the scores of all 225 patients are considered or scores of unilateral and bilateral cases are considered separately. For most subscales, bilateral cases who enrolled in the SST Group H Trial had scores similar to those of patients with age-related macular degeneration in the study by Brody et al35 (column A, Table 7) and worse than scores of patients with age-related macular degeneration who participated in the NEI-VFQ Field Test (column B, Table 7). Mean scores for unilateral cases were similar on most subscales to those of NEI-VFQ Field Test patients who had diabetic retinopathy, were somewhat worse than those of NEI-VFQ Field Test patients who had preoperative age-related cataract, and were much worse than those of patients interviewed 5 years after enrollment in the Age-Related Eye Disease Study39 (columns C, F, and J, respectively, Table 7).
When the SST was designed, some members of the planning group speculated that any effect of surgery on vision-targeted QOL would be observable only in the subgroup of patients whose study eye was the second eye to be affected by CNV (bilateral cases). They reasoned that the unaffected fellow eyes of other patients (unilateral cases) would have excellent visual acuity, that patients would use that eye for visual tasks, and that vision-targeted QOL was unlikely to be affected by the status or treatment of the study eye. However, findings from the baseline interviews of patients who enrolled in the SST Group H Trial demonstrate that unilateral and bilateral cases of CNV had reduced vision-targeted QOL. Despite younger age, better visual acuity, and a better short-term visual prognosis, bilateral cases had NEI-VFQ scores similar to those of patients with age-related macular degeneration in 2 different studies (Table 7). Ophthalmologists who care for patients similar to those in the SST Group H Trial may not have appreciated fully the magnitude of the adverse effect of neovascular ocular histoplasmosis and idiopathic CNV on vision-targeted QOL, particularly when only 1 eye is affected.
As documented by the Macular Photocoagulation Study Group,40 patients with neovascularization in 1 eye due to the ocular histoplasmosis syndrome have a nearly constant risk of CNV developing in the second eye at a rate of 2% per year for 5 years or longer. Thus, ophthalmologists not only should inform patients of the risk of CNV in unaffected fellow eyes but also should continue to monitor these eyes so that treatment, when appropriate, can be administered promptly. Early diagnosis of CNV in the second eye should be followed by referral for low vision or other professional counseling. In addition, ophthalmologists should be aware of the possibility of poor mental health in these patients and should consider prompt referral for psychiatric examination and management when depression is suspected.
Our findings suggest that patients’ perceptions of vision and ability to function are based on more factors than visual acuity of the better-seeing eye. Aspects of physical health and personal values, goals, and expectations undoubtedly contribute to perceptions of vision and function, and should be considered in the overall management of the patient’s ophthalmologic problem.
Correspondence: Barbara S. Hawkins, PhD, SST Coordinating Center, 550 N Broadway, Ninth Floor, Baltimore, MD 21205-2010 (firstname.lastname@example.org).
Submitted for Publication: November 13, 2003; final revision received July 29, 2004; accepted July 29, 2004.
Financial Disclosure: None.
Funding/Support: The SST is supported by cooperative agreements U10 EY11547, EY11557, and EY11558 between the NEI, National Institutes of Health, Department of Health and Human Services, Bethesda, Md, and The Johns Hopkins University, Baltimore, Md. Participating clinical centers were supported by contracts with The Johns Hopkins University.
*Group Information: A list of the members of the Submacular Surgery Trials (SST) Research Group who contributed data for the patients enrolled in this clinical trial was published in Arch Ophthalmol. 2004;122:1609-1610. The members of the SST Patient-Centered Outcomes Subcommittee who take responsibility for the content of this article on behalf of the SST Research Group are Barbara S. Hawkins, PhD; Paivi H. Miskala, PhD; Carol M. Mangione, MD, MSPH; Eric B. Bass, MD, MPH; Neil M. Bressler, MD; Ashley Childs Mann, MS; LiMing Dong, PhD; and Marta J. Marsh, MS.