Responsiveness of the National Eye Institute Visual Function Questionnaire to Progression to Advanced Age-Related Macular Degeneration, Vision Loss, and Lens Opacity: AREDS Report No. 14

Objective  To describe the ability of the National Eye Institute Visual Function Questionnaire (NEI-VFQ) to detect meaningful change over time (responsiveness) to the primary Age-Related Eye Disease Study outcomes.

Methods  The 25-item NEI-VFQ plus appendix was administered at 2 visits at 1- to 4-year intervals to 4119 participants in the Age-Related Eye Disease Study. Events evaluated were progression to advanced age-related macular degeneration (AMD), visual acuity (VA) loss of at least 15 letters, and lens opacity progression. Responsiveness was measured by the t statistic, effect size (ES), responsiveness statistic, and area under the receiver operating characteristic curve. Variance components were used to estimate the contributions of events to variability of the NEI-VFQ score.

Results  Overall NEI-VFQ score was responsive to AMD progression (t = 14.0; P<.001; ES = 0.81) and VA (t = 16.2; P<.001. ES = 0.74). Mean changes ranged from 11 to 25 points for the subscales of general vision, near and distance activities, social functioning, mental health, role difficulties, dependency, and driving. The NEI-VFQ was unresponsive to lens opacity progression, although when the event occurred in the eye with the best vision at the first administration, the lens opacity ES was moderate for the color vision (ES = 0.62) and driving subscales (ES = 0.66). Progression to advanced AMD and VA loss contributed significantly to the variation in the mean difference in overall VFQ score.

Conclusions  Changes in the NEI-VFQ overall and subscale scores of 10 points or more are associated with clinically significant changes in vision and AMD. This finding may assist the design of interventional studies of AMD and VA loss that include the NEI-VFQ as an outcome measure.

Age-Related Eye Disease Study (AREDS) is an ongoing multicenter study of the clinical course of lens opacity and age-related macular degeneration (AMD). The study incorporated a randomized clinical trial designed to evaluate the effect of high-dose micronutrient supplements on the incidence and progression of these 2 conditions. The main results of this trial have been published.^1,2 Secondary objectives of AREDS include further understanding the etiology of lens opacity and AMD and the effects that these conditions have on an individual’s health-related quality of life.

The National Eye Institute Visual Function Questionnaire (NEI-VFQ) was designed to measure areas of vision-targeted, health-related functioning and well-being that were identified as important by persons with eye disease. The NEI-VFQ was constructed to evaluate the impact of visual disability on health-related quality of life across several common eye conditions. Field tests of the instrument reported it to be a reliable and valid measure in patients with age-related cataracts, AMD, diabetic retinopathy, primary open-angle glaucoma, cytomegalovirus retinitis, and low vision from any cause.³

Responsiveness refers to the ability of a measurement tool to detect meaningful change in populations over time and is also called “sensitivity to change.”⁴ Many methods have been proposed to explore the responsiveness of questionnaires,⁵ and all involve the administration of the questionnaire before and after the time during which it is expected that changes in the population will occur. The Submacular Surgery Trials Research Group recently reported the overall NEI-VFQ score and those of 9 subscales (all subscales except general health, ocular pain, and color vision) to be responsive to a decrease in visual acuity (VA) of 3 lines or more during a 1-year period in the eye with better vision at first administration.⁶ There are several other such studies ongoing, but results have yet to be published.

In this report, we explore the responsiveness of the NEI-VFQ to documented progression in some of the primary outcomes used in AREDS, including advanced AMD, VA loss, and increased lens opacity. Our population provides the opportunity to assess the simultaneous effect of these outcomes on vision-targeted health-related quality of life.

Details of the AREDS study design have been published previously⁷ and are briefly described herein. A total of 4757 persons, aged 55 to 80 years at the time of enrollment, were entered into the study at 11 clinical centers from November 13, 1992, through January 15, 1998. Institutional review board approval for center participation was obtained at each site, and participants were required to sign the informed consent statement approved by the institutional review board before enrollment. Participants were required to have good VA in at least 1 eye (Snellen equivalent of ≥20/32. and media clear enough to obtain good-quality fundus photographs. This photographic eligibility requirement limited the severity of lens opacities in patients at baseline, but lens opacity status was not itself an eligibility requirement. The extent of AMD was classified in the following 4 categories: no macular abnormality in either eye to a few small drusen (category 1); many small or a few intermediate drusen, or pigment abnormalities (category 2); at least 1 large druse, extensive intermediate drusen, or noncentral geographic atrophy (category 3); and advanced AMD or lesions of AMD with VA of less than 20/32 in 1 eye (category 4).

Nuclear opacities were graded on a decimal scale using cut points set by a series of standard slitlamp photographs with increasingly severe nuclear sclerosis. Nuclear grades ranged from 0.9 (severity less than the lowest standard) to 6.1 (severity exceeding the highest standard). The extent of cortical and posterior subcapsular opacities was graded by estimating the area of lens involvement as seen on retroillumination photographs. Participants were followed up by means of clinic visits every 6 months, with ocular photographs taken at enrollment and annually thereafter beginning 2 years after enrollment. Visual acuity was tested by trained and certified examiners using a modified Early Treatment of Diabetic Retinopathy Study protocol, and photographs of the lens and fundus were sent to a reading center for grading.

The 1996 25-item NEI-VFQ plus appendix (NEI-VFQ-25; RAND Corporation, Santa Monica, Calif) was administered by trained and certified AREDS clinic coordinators. Administration took place during a clinic visit before an eye examination or by telephone when a participant was unable to come to the clinic for the interview (<5%). The first administration took place in 1997, with follow-up administration every 3 years thereafter and at each participant’s 5-year visit. For the purposes of this analysis, we included participants with at least 2 interviews that took place at least 1 year apart. The instrument measures 12 domains (subscales) of health-related quality of life, including overall health; overall vision; difficulty with near-vision and distance activities. ocular pain; driving difficulties; limitations with peripheral vision and color vision; social functioning; role limitations; dependency and mental health symptoms related to vision. Each subscale was scored using the average of all items within that subscale for an individual. Scores for each subscale ranged from 0 to 100. The average of all subscale scores is used to calculate an overall score.

Progression to advanced AMD in a study eye between NEI-VFQ administrations was defined as occurrence of any of the following in an eye without advanced AMD or VA loss due to AMD before the first administration: photocoagulation or other treatment for choroidal neovascularization (based on clinical center reports), or photographic documentation (based on reading center reports) of geographic atrophy involving the center of the macula, nondrusenoid retinal pigment epithelial detachment, serous or hemorrhagic retinal detachment, hemorrhage under the retina or the retinal pigment epithelium, and/or subretinal fibrosis. Progression to advanced AMD was evaluated on the basis of photographs or clinic reports after the first administration and until the second administration.

Progression to VA loss was defined as a decrease of 15 or more letters in Early Treatment of Diabetic Retinopathy Study VA between the measurements obtained at the 2 NEI-VFQ administrations in at least 1 eye for participants with VA of at least 34 letters (≥20/200) in at least 1 eye at the first NEI-VFQ administration. Visual acuity of 34 letters was chosen as a lower limit to allow enough residual vision from which to measure change. A decrease of 15 letters represents a doubling of the visual angle. Persons with VA of fewer than 34 letters in the better eye were omitted from VA analyses. A separate analysis evaluated the effect of vision loss in the eye with better vision at the first administration in persons with at least a 5-letter difference in VA between eyes.

Progression to a lens opacity event in a participant was defined as prespecified in the AREDS clinical trial. This definition was the occurrence in at least 1 eye (with a natural lens) of any of the following changes in photographic grade at the second administration compared with the photographic grade at the first administration: nuclear opacity (a 1.5-U increase), cortical opacity (10% absolute increase in the area of opacity within a standard central 5-mm circle), and posterior subcapsular opacity (5% absolute increase in the area of opacity within a standard central 5-mm circle). To be eligible for a nuclear event, participants must have had a nuclear score of 4.5 or less at the first administration to allow for a 1.5-U change. Similarly, progression for a posterior subcapsular event or cortical event required grades of less than 95% and 90%, respectively, at the time of first administration. Participants with cataract surgery between administrations were excluded from the lens opacity progression analyses. A separate analysis evaluated the effect of lens opacity progression when it occurred in the eye with better VA at the first administration in persons with at least a 5-letter difference in VA between eyes.

We evaluated the performance of the NEI-VFQ using a number of measures of responsiveness. The statistics do not depend on the measurement scale, and larger values indicate more responsive NEI-VFQ measures. Paired t tests were used to compare the difference in NEI-VFQ scores between the first and second administrations for participants with progression to an event between administrations. Significant changes in score would be evidence of responsiveness.⁸ The effect size (ES) and responsiveness statistic (RS) are calculated as the mean change in NEI-VFQ score between the first and second administrations, divided by the standard deviation of the NEI-VFQ at the first administration (ES) and the standard deviation of change for participants without progression to an event between administrations (RS). A larger ES indicates greater likelihood that the instrument as a whole or the various subscales will detect progression. Cohen^9,10 suggests that an ES of 0.2 to 0.49 represents a small change; 0.5 to 0.79, a medium change; and 0.8 or more, a large change. An ES of more than 0.8 indicates that the scale, on average, changed by 0.8 SDs, suggesting that the scale or domain is responsive. An RS of more than 1.0 indicates that the measure is highly responsive to change.¹¹

The sensitivity and specificity of the NEI-VFQ as a diagnostic test for progression to an event was computed for each possible threshold value to obtain a receiver operating characteristic curve.¹² The area under the curve (AUC) was used as a quantitative method for assessing the scale’s ability to distinguish participants with true progression from those without it. The AUC can be interpreted as the probability of correctly identifying participants with true progression from randomly selected pairs of changed and unchanged participants¹³ and ranges from 0.5 (no diagnostic accuracy beyond chance) to 1.0 (perfect diagnostic accuracy).

An additional analysis consisted of comparing the difference in mean scores between participants with and without progression with each of the outcome measures (advanced AMD, VA loss, and lens opacity) using a generalized linear model adjusted for age, sex, race, AMD category, and time between administrations. Significance was tested at P<.01 to partially adjust for multiple comparisons. Separate models were generated for each outcome. We used variance components to estimate how much each ocular factor contributed to the overall variability in the NEI-VFQ score. A larger variance indicated greater contribution of the factor to the overall variability of the NEI-VFQ score. All analyses were performed using SAS version 8.0 software (SAS Institute, Inc, Cary, NC).

Beginning in December 1997, the NEI-VFQ plus appendix was administered to 4119 AREDS participants on at least 2 occasions at least 1 year apart. Characteristics of the 4119 participants are given in table 1. The median age of the population at the first administration was 72 years (age range, 57-84 years). The time between administrations ranged from 1 to 4 years, and more than half (59%) of participants completed the questionnaires 3 to 4 years apart. Women constituted 57% of the participants, and 96% were white. At enrollment in the AREDS, 24% of participants had no AMD (category 1), and 19% had advanced AMD in 1 eye (category 4).

Responsiveness of the NEI-VFQ to progression to advanced AMD was assessed in 3885 AREDS participants with fundus photographs available at each administration. Of these 3885 participants, 364 demonstrated progression to advanced AMD in at least 1 eye between NEI-VFQ administrations. Table 2 shows the mean scores at the first and second administrations, the mean difference in scores for participants with progression, a standard paired t test comparing the means, the 2 measures of responsiveness (ES and RS), and a measure of the sensitivity and specificity of the scales (AUC) for the overall score and the 12 subscale scores. When using the paired t statistic to determine which subscales are more responsive, we found that the overall score of the NEI-VFQ is responsive (t = 14.0; P<.001) to progression to advanced AMD. The subscales of the NEI-VFQ that were more responsive to progression to advanced AMD using the t statistic were general vision, near activities, distance activities, social functioning, mental health, role difficulties, dependency, and driving.

Using the ES measure, the difference detected using the overall score showed a large decrease in score for progression to advanced AMD (ES = 0.81). The subscales showing a moderate to large decrease in scores for progression to AMD included general vision (ES = 0.67), near activities (ES = 0.76), distance activities (ES = 0.76), social functioning (ES = 0.84), mental health (ES = 0.54), role difficulties (ES = 0.67), dependency (ES = 0.83), and driving (ES = 0.85). The RS was equal to or exceeded 1 for the overall NEI-VFQ score (RS = 1.56) and for each of these subscales, with the exception of general vision (RS = 0.87. and mental health (RS = 0.92).

An analysis of the AUC confirmed these findings. The overall NEI-VFQ moderately discriminated between participants with or without progression to advanced AMD (AUC = 0.74). The general vision, near activities, distance activities, social functioning, mental health, role difficulties, dependency, and driving subscales suggested some ability to detect progression, with the AUC ranging from 0.65 for the mental health and general vision subscales to 0.73 for the near activities subscale.

After adjustment for age, sex, race, AMD category, and time between NEI-VFQ administrations, the mean difference in scores for participants with vs those without progression to advanced AMD between the 2 administrations differed significantly for the overall NEI-VFQ and for all of the subscales (P<.001) with the exception of general health and ocular pain (data not shown). Significant differences ranged from 8 points (general vision) to slightly more than 19 points (driving).

We assessed the responsiveness of the NEI-VFQ to progression to VA loss (at least a 15-letter drop in VA in either eye) in 3624 AREDS participants with VA of at least 34 letters (≥20/200) in at least 1 eye at the first. administration. Of these 3624 participants, 485 demonstrated progression to VA loss in at least 1 eye between NEI-VFQ administrations. Table 3 shows that the overall score of the NEI-VFQ has a moderate ES and is highly responsive to VA loss with an ES of 0.74 and an RS of 1.76 (t = 16.2; p<.001). The subscales of the NEI-VFQ that had moderate ESs and were also highly responsive to VA loss were general vision (ES = 0.73; RS = 1.07), near activities (ES = 0.72; RS = 1.38), distance activities (ES = 0.73; RS = 1.54), social functioning (ES = 0.70. RS = 1.46), role difficulties (ES = 0.61; RS = 1.23), dependency (ES = 0.69; RS = 1.50), and driving (ES = 0.75. RS = 1.42).

The overall NEI-VFQ moderately discriminated between participants with or without progression to VA loss (AUC = 0.76). The general vision, near activities, distance activities, social functioning, mental health, role difficulties, dependency, and driving subscales suggested some ability to detect progression, with an AUC ranging from 0.66 for the mental health subscale to 0.74 for the distance activities subscale.

After adjustment for age, sex, race, AMD category, and time between administrations, the mean difference in scores for participants with vs those without VA loss between the 2 administrations differed significantly for the overall NEI-VFQ and for all subscales (data not shown). The driving subscale had the largest difference of nearly 19 points.

An analysis of the subset of 1313 persons who had a difference of 5 or more letters in VA between eyes at the first NEI-VFQ administration shows that when VA loss occurs in the eye with better vision, the NEI-VFQ is more responsive to the loss than when the event occurs in the eye with worse vision. The overall scale and subscales are most responsive when VA loss occurs in both eyes. table 4 lists the mean differences in scores adjusted for age, sex, race, AMD category, and time between administrations and the ES by event type. When a loss of 15 or more letters occurred in the eye with better VA at the first administration, the ES was in the moderate range for the overall score (ES = 0.73) and for 8 of the subscale scores. When the event occurred in the eye with poorer VA at the first administration, the ES was in the small range for the overall score and 8 of the subscale scores. The largest ES was seen for persons who experienced an event in both eyes (ES = 1.38). The ESs for the overall score and all subscales were in the large range except for the general health, ocular pain, color vision, and peripheral vision subscales. The mean differences in scores between administrations when VA loss occurred in both eyes were at least twice the difference observed when VA loss occurred in the eye with better vision. When differences in scores were compared in eyes with a difference in VA within 5 letters at the first VFQ administration, results for VA loss in 1 eye were similar to results for VA loss in the worse eye (table 4).

We assessed the responsiveness of the NEI-VFQ to progression to lens opacity (nuclear, cortical, or posterior subcapsular) in the 3055 AREDS participants with at least 1 natural lens and photographic follow-up at the first NEI-VFQ administration. Of these 3055 participants, 348 experienced a progression of lens opacity in at least 1 eye between NEI-VFQ administrations. Table 5 shows that the overall score of the NEI-VFQ was somewhat responsive to progression to lens opacity as measured by the paired t statistic (t = 5.24; P<.001), but other measures of responsiveness demonstrated that the amount of change was small (ES = 0.22; RS = 0.36. AUC = 0.54). When we evaluated the subscales, we found the general health (ES = 0.26), general vision (ES = 0.25), distance activities (ES = 0.21), social functioning (ES = 0.23), and driving (ES = 0.27) subscales to be the only subscales with an ES that can be classified as at least small. The overall NEI-VFQ did not adequately discriminate between participants with or without progression to lens opacity (AUC = 0.54). Similar results were found when the analysis was conducted for opacity type (data not shown). The mean difference in scores for participants with progression to lens opacity vs participants without progression between the 2 administrations did not differ for the overall NEI-VFQ or the subscales, after adjustment for covariates (p>.01. data not shown).

As shown in table 6, an analysis of lens opacity progression occurring in the eye with better visual acuity at baseline (by ≥5 letters, which occurred in 49 participants) found ESs in the moderate range for driving (ES = 0.66) and color vision (ES = 0.62). All other subscales except ocular pain and peripheral vision (ES<0.20 for both) had small ESs ranging from 0.22 to 0.47. Effect sizes remained in the small or the no-effect range when events occurred in both eyes or in the eye that had worse VA at first administration. Significant differences in mean change in score between those who experienced lens opacity progression in an eye with better VA at the first administration compared with those without progression between administrations were observed only for the role difficulties, driving, and color vision subscales.

The contribution to the variability in the NEI-VFQ overall and subscale scores by type of event was assessed using a variance component analysis. Progression to advanced AMD and VA loss of 15 or more letters each contributed to the overall variability of the NEI-VFQ score (variance component estimates of 25 for advanced AMD and 32 for VA loss). Lens opacity progression had little impact (variance component estimate of 0.4 for overall score). Visual acuity loss had the greatest impact on the subscales of general vision, near activities, distance activities, and driving, whereas role difficulties had a greater contribution from progression to advanced AMD. Very little contribution from these types of events was observed for the general health and ocular pain subscales (data not shown).

The figure illustrates that the most severe VA loss (≥15 letters) predominates in persons with progression to advanced AMD (68%), whereas only 19% of those with lens opacity progression experienced such a loss.

The AREDS population includes participants whose lens status, AMD status, and vision are well characterized at the time of VFQ administration. We found that the NEI-VFQ plus appendix overall score was highly responsive to progression to both advanced AMD and VA loss of at least 3 lines. The subscales of general vision, near and distance activities, driving, social functioning, role difficulties, mental health, and dependency were also responsive. Responsiveness was accentuated within the subgroup of patients experiencing a VA loss of 15 or more letters in both eyes between administrations. Moderate ESs were observed when VA loss occurred in the eye with better vision, and small ESs were observed when VA loss was limited to the eye with worse vision at the first administration.

We did not find the NEI-VFQ overall and subscale scores to be responsive to any lens opacity progression, except when that progression occurred in the eye with better vision at the baseline administration. For these participants, the driving and color vision subscales had moderate ESs. Other health-related quality-of-life instruments such as the 14-item visual function questionnaire (VF-14) have been reported to be highly responsive to cataract surgery (ES = 0.99).¹⁴ Our population includes participants with macular degeneration, and we have shown AMD to have a significant impact on health-related quality of life.¹⁵ In a study of 1073 cataract operations, researchers found that 99 cases had preoperative diagnoses of age-related maculopathy. Of those cases, 2% had exudative disease and the remainder had dry or unspecified AMD. Among the patients with AMD in this study of cataract surgery, 17% did not think the cataract surgery was worthwhile.¹⁶ The severity of AMD in this study’s population is likely much less than that in the AREDS population. Because of the small sample sizes, we were unable to evaluate the effect of cataract surgery on NEI-VFQ responses in a subset of persons without AMD. It is possible that the lack of responsiveness to lens opacity progression in our study may have been influenced by the exclusion of persons who had cataract surgery, which may have left in the study mainly those who had no or only small limitations in visual function due to opacity progression or who have lesser demands for good visual function. Also, our definition of opacity progression may be too mild to elicit an impact on visual function among persons with no or with mild opacity at the start of follow-up.

Limited data exist describing the responsiveness of the NEI-VFQ and its subscales to changes in ophthalmic status.¹⁷ Studies of the VF-14 as an index of visual function in recipients of a corneal graft found significant changes in the overall score of 13 U 1 year after graft placement.¹⁸ Stelmack et al¹⁹ found the NEI-VFQ-25 plus appendix to be a useful measure of the effects of low-vision rehabilitation. In the AREDS population, a cross-sectional study suggested that differences between those with and without advanced AMD ranged from 1. to 20 points in many of the subscales found to be responsive in this analysis and about 15 points in the overall score.¹⁵ These differences are similar to those reported for the Activities of Daily Living Scale by Scilley et al²⁰ and Mangione et al.²¹ The Submacular Surgery Trials Research Group⁶ reported on the responsiveness of the NEI-VFQ in a population of individuals with subfoveal neovascularization in at least 1 eye. Eyes with better vision at baseline that lost 3 or more lines of VA during a 12-month period had an expected decrease in score ranging from 3.6 to 16.. points in the overall score and 9 of the subscales. We found in our longitudinal study that VA loss of 3 or more lines in an eye with better vision at baseline was associated with mean changes of 11 points for the overall score and ranging from 7 to 19 points for the same 9 subscales reported by the Submacular Surgery Trials Research Group.⁶ The higher scores (more change) in the AREDS may be due to longer follow-up (median follow-up, 3 years vs 12 months), better initial VA, or methodological differences.

A cross-sectional study of the VF-14 found that, after adjustment for VA, AMD severity was not an independently significant predictor of VF-14 score for the overall study population. In this same study, an analysis restricted to patients with VA of 20/20 in the better eye found that AMD was a significant predictor of score after adjustment for acuity in the fellow eye.²² Mangione et al²¹ found that AMD severity did not explain a significant portion of visual functioning after taking into account VA in an assessment of the Activities of Daily Vision Scale. We found that the NEI-VFQ was responsive among those with progression to advanced AMD. A variance components analysis suggests that a 15-letter VA loss, a standard ophthalmic outcome in clinical trials, and progression to advanced AMD contribute to the variability of the responses on the NEI-VFQ. Advanced AMD appeared to have slightly more impact in the subscale of role difficulties.

A recent report²³ of the potential impact of depression on responses to self-reported visual function by people with recent VA loss due to AMD suggested that an increase in depressive symptoms over time was predictive of a decline in self-reported visual functioning independent of changes in vision or medical status. Although some participants in AREDS were administered the Center for Epidemiological Studies Depression Scale, its administration was not concurrent with the administration of the NEI-VFQ, and so the effects of depression cannot be evaluated within this study of responsiveness. Analyses of the Center for Epidemiological Studies Depression Scale as it relates to AMD status and vision are ongoing.

We did not find an association of progression of age-related cataract by the AREDS definition with clinically relevant changes in vision-targeted, health-related quality of life, although the ability to find an association may be limited in this study. Progression to advanced AMD as defined by AREDS had a significant impact on vision-related quality of life, as did loss in VA of at least 3 lines. The largest impact on vision-targeted quality of life occurred in persons who lost VA in both eyes. Persons who lost VA in 1 eye were most affected when that eye began with better vision than the fellow eye. The NEI-VFQ was responsive to change in 2 clinically important outcome measures, VA loss and progression to advanced AMD, and these results suggest that further exploration of this tool as a surrogate for patient outcome when patients are unable to return to a clinical center for examination may be warranted.

Our results suggest that mean changes in the overall and subscale scores of 10 points or more are associated with clinically significant changes in vision status. This finding may be useful in the design of interventional studies of AMD and VA loss that include the NEI-VFQ as an outcome measure.

Correspondence: AREDS Coordinating Center, The EMMES Corporation, 401 N Washington St, Suite 700, Rockville, MD 20850-1707 (aredspub@emmes.com).

Submitted for Publication: October 22, 2003; final revision received October 14, 2004; accepted November 18, 2004.

Financial Disclosure: None.

Funding/Support: This study was supported by contracts from the National Eye Institute, National Institutes of Health, Department of Health and Human Services, Bethesda, Md.

Group Information: A complete list of members of the AREDS Research Group was published in Arch Ophthalmol. 2004;122:723-724.