Relative Contributions of Reduced Vision and General Health to NEI-VFQScores in Patients With Neovascular Age-Related Macular Degeneration

Objective  To estimate the relative contributions of central vision loss and generalhealth to vision-targeted quality of life as measured by the National EyeInstitute Visual Function Questionnaire (NEI-VFQ).

Methods  Data on quality of life (NEI-VFQ and the 36-Item Short-Form Health Survey[SF-36]) and visual acuity were collected as part of the Submacular SurgeryTrials Pilot Study. Information on medical conditions was collected by patientchart review. Twenty-four–month data for 120 patients were analyzedusing linear regression methods.

Results  Median patient age at the 24-month examination was 77 years; 60% werewomen, and 98% were non-Hispanic whites. A 3-line decrement in visual acuityin the better-seeing eye was associated with a 5.1- to 17.1-point decrementin NEI-VFQ scores after adjustment for general health (SF-36 physical componentsummary [PCS] and mental component summary [MCS] scores). A 10-point decrementin the PCS score was associated with a 4- to 9-point decrement in NEI-VFQscores after adjustment for visual acuity in the better-seeing eye and MCSscore. A 10-point decrement in the MCS score was associated with a 4- to 8-pointdecrement in NEI-VFQ scores after adjustment for visual acuity in the better-seeingeye and PCS score. Diabetes, arthritis/rheumatism, and hypertension also hadlarge effects on NEI-VFQ scores in the adjusted analysis.

Conclusions  The NEI-VFQ is sensitive to differences in visual acuity in the better-seeingeye, as expected, and to differences in general health. Adjustment for generalhealth should be considered when comparing NEI-VFQ scores between patientgroups.

Because most individuals with vision loss caused by age-related maculardegeneration (AMD) are older than 65 years, they often have coexisting medicalconditions. The National Health Interview Survey¹ reportedthe following prevalences in individuals 65 years and older: high blood pressure,36%; arthritis, 48%; heart disease, 27%; diabetes, 10%; and asthma, 5%. Psychologicaldistress² and depression³ alsoare common in patients with AMD. In one study,³ nearlyone third of patients with advanced AMD screened positive for depression.Other sensory impairments, such as hearing loss, also may compound the effectsof vision impairment and make day-to-day functioning more difficult. For example,in the white population of Beaver Dam, Wis, AMD and hearing loss were bothpresent in 15% of individuals aged 48 to 92 years and in 40% of individuals75 years or older.⁴

Since vision and other health problems disproportionately affect olderpeople, it is important to understand how these conditions contribute to qualityof life (QOL). The knowledge of possible effects that nonocular medical conditionsmay have on QOL is particularly important when trying to assess the effectof vision loss on daily activities in people older than 65 years. One measureof vision-related QOL is the National Eye Institute Visual Function Questionnaire(NEI-VFQ).^5-8 Adjustmentof NEI-VFQ scores for general health, among other factors, has been reportedin the literature^6,8-10;however, no discussion has addressed the best way to measure general healthor the magnitude of the effect that general health may have on NEI-VFQ scores.The purpose of this study is to estimate the relative contributions of reducedvision and general health to vision-targeted QOL as measured by the NEI-VFQ.This information may be valuable for investigators who are measuring and interpretingvision-targeted QOL outcomes using the NEI-VFQ and for physicians who carefor people with AMD.

The Submacular Surgery Trials (SST) Pilot Study consists of 4 randomizedpilot trials undertaken to assess the feasibility of larger clinical trialsto evaluate submacular surgery for subfoveal choroidal neovascularizationassociated with AMD and other conditions. Patients who enrolled in 2 SST pilottrials, Group N and Group B, were eligible for the present study.¹¹ Group N pilot trial included persons with AMD whowere 50 years or older, had new untreated subfoveal choroidal neovascularization,and had visual acuity in the affected eye of 20/100 through 20/800 (approximateSnellen equivalent).¹² The subfoveal lesioncould be either well-demarcated and large (3.5-9.0 Macular PhotocoagulationStudy¹³ [MPS] disc areas, whereby 1 MPS discarea¹² corresponds to a circle with an areaof 2.5 mm²) or poorly demarcated with no lower size limit (≤9.0MPS disc areas). Group B pilot trial patients also were 50 years or olderand had AMD but had a large, predominantly hemorrhagic lesion (≥3.5 MPSdisc areas), with visual acuity in the study eye of 20/100 (approximate Snellenequivalent) or worse but at least light perception.¹² Inthese pilot trials, eligible consenting patients were randomized equally tosubmacular surgery or observation of the study eye.

Additional data collection for the present study was limited to 10 clinicalcenters that participated in Group N and Group B pilot studies and that alsowere participating currently in the larger randomized trials of submacularsurgery sponsored by the NEI of the National Institutes of Health (the SST).This analysis is limited to cross-sectional data from 120 patients who underwentinterviews using the NEI-VFQ and visual acuity measurements 24 months afterenrollment in the SST Group N and Group B pilot trials (Figure 1). The institutional review board at each clinical centerreviewed the SST Pilot Study protocol before patient enrollment began. Approvalwas obtained from the SST Operations Committee and the local institutionalreview boards at the 10 centers for the medical chart reviews conducted forthe present study.

Information on baseline characteristics and best-corrected visual acuitywas collected as part of the SST Pilot Study. Visual acuity was measured usingmodified Bailey-Lovie (Early Treatment Diabetic Retinopathy Study) charts¹⁴ and a standard protocol after careful refractionto obtain the best correction.^12,15 Visualacuity was measured monocularly at a 2-m test distance from the vision chart.Whenever a patient could read at least 15 letters correctly at this distancewith the eye being tested, the final visual acuity score for that eye wasthe number of letters read correctly plus 30. However, when the patient couldnot read at least 15 letters, he or she was moved to 0.5 m from the chartand the measurement was repeated; the final visual acuity score for that eyewas the number of letters read correctly at the 0.5-m distance. When the patientcould not read any letters on the vision chart at the 0.5-m distance (visualacuity <20/1600), vision was coded as light perception or no light perception,as assessed by the vision examiner and confirmed by the ophthalmologist. Forthe purpose of SST Pilot Study data analysis, light perception was given ascore of –10 (ie, 3 lines worse than 20/1600) and no light perceptiona score of –30 (ie, ≥7 lines worse than 20/1600). Scores were convertedto lines of visual acuity by dividing visual acuity scores by 5 (5 lettersper chart line).

During the 24-month follow-up examination, a traveling vision examiner,who was masked to treatment assignment and to study eye, measured the visualacuity of each patient at each clinical center whenever possible. Ninety-eight(82%) of 120 patients had masked visual acuity measurements 24 months afterenrollment. For this analysis, those visual acuity measurements were usedwhen available.

Interviews were conducted at baseline and 6, 12, and 24 months afterenrollment in the SST Pilot Study. Initially, the 36-Item Short-Form HealthSurvey (SF-36)^16,17 was administeredin-person by a member of the clinical staff as part of baseline data collectionin the SST Pilot Study for most of the patients enrolled. The QOL data collectionwas modified in January 1997 when the interview procedure changed from in-personlocal interviewer administration to masked central telephone administration.Also at that time, the NEI-VFQ was added to the interview. Because the NEI-VFQwas introduced in the course of the study, the number of patients with NEI-VFQinterviews was largest 24 months after enrollment. Thus, 24-month NEI-VFQdata were analyzed for this study to include the maximum number of patients.

Three versions of the NEI-VFQ have been published, containing 25 items,39 items, and 51 items.^5-7,18 Forthis analysis, a 37-item NEI-VFQ was created by excluding 2 general healthitems from the 39-item version. The 37 items are divided into 11 subscales:general vision, ocular pain, near activities, distance activities, socialfunctioning, mental health, role difficulties, dependency, driving, colorvision, and peripheral vision. Individual subscales consist of 1 to 6 questions,and each question is given a score of 0 to 100. One of the driving subscalequestions (difficulty driving in difficult conditions) was added late to theNEI-VFQ interview, and, therefore, for 55% of individuals included in thisanalysis, the driving subscale was calculated from a maximum of 2 items insteadof 3. A subscale score is an unweighted average of all questions for whicha response was recorded. Subscale scores range from 0, which is the worstscore (poorest function), to 100, which is the best possible score (best function).The overall score for the NEI-VFQ is an unweighted average of the 11 subscalescores.

The SF-36 interview scores were evaluated as a measure of general healthstatus. The SF-36 consists of 8 subscales that can be combined to form 2 summaryscales.^16,17 Data analysis forthis study concentrated on the 2 summary scales: the physical component summary(PCS) and the mental component summary (MCS) scores 24 months after enrollment.¹⁷ All 8 subscales contribute to each of the 2 summaryscales, but they are weighted differently for each summary scale. A nonmissingvalue is required for all 8 subscale scores to create the summary scales.The summary scale scores are standardized to have a mean of 50 and an SD of10 in the general US population.¹⁷ Theoretically,the worst possible value for the summary scale scores is 0, and the best possiblevalue is 100, but these are highly improbable scores for individuals participatingin an outpatient study.

A list of medical conditions was taken from the Medical Outcomes Study¹⁹ by the developers of the NEI-VFQ⁷ andwas used as a model for collecting information on medical conditions. Forthe present study, the 16-item checklist of medical conditions was retained;items were added regarding clinic note dates, source of information (ie, medicalhistory or examination), and information on prescription medications.¹¹ Psychiatric disease was added to the list of conditionsafter data collection based on responses to an open-ended question on thequestionnaire.

Information on medical conditions was extracted from the medical chartsthat were available to the ophthalmologist when study patients were examined.These medical charts typically were the total ophthalmology charts (6 centers)or the ophthalmology subspecialty charts (retina service charts) (2 centers);at 2 centers, the full outpatient medical record typically was available tothe ophthalmologist. Differences reflected the variety of clinical centersettings, that is, community-based centers and university-based centers. Chartsat all 10 clinical centers were reviewed by one of us (P.H.M.); a physician(N.M.B.) provided advice regarding categorization of medical conditions, surgeries,procedures, and therapies. Information on prescription medications was usedas a secondary source when the clinical notes were unclear as to whether amedical condition was present.

Descriptive statistics were calculated for each variable of interest.The χ² test or the Fisher exact test was used to compare patientcharacteristics that were categorical. The Wilcoxon 2-sample rank sum testwas used to compare patient characteristics that were continuous. An extensionof the Wilcoxon rank sum test²⁰ was used toevaluate trends in QOL scores across visual acuity categories.

Linear regression models were used to explore relations among visualacuity, general health, and QOL. The dependent variable in all regressionmodels was the NEI-VFQ overall score or 1 of the 11 subscale scores 24 monthsafter enrollment. The independent variables were visual acuity in the better-seeingeye and general health status measured by the presence of medical conditionsor SF-36 PCS and MCS scores 24 months after enrollment. The NEI-VFQ and SF-36scores were treated as continuous variables in the analysis. Visual acuityin the better-seeing eye, also a continuous variable, was defined to be thatof the eye with the higher visual acuity score 24 months after enrollment.Medical conditions were treated as indicator variables and were coded as 0when the condition was absent or when the disease status could not be determinedand as 1 when the condition was definitely or probably present. Data analysisconcentrated on the 5 most common medical conditions. Hearing problems werestudied in detail to investigate how another sensory impairment contributesto NEI-VFQ scores. The number of medical conditions was categorized for somelinear regression models: coding was 0 when the patient had less than 2 medicalconditions other than AMD and 1 when the patient had 2 or more additionalmedical conditions. This categorization was based on our observation thatthere seemed to be a difference in the distribution of NEI-VFQ scores betweenindividuals with less than 2 conditions and those with 2 or more conditions.Gender and age were evaluated as potential confounders, using linear regressionmethods, for the association between vision loss and QOL.

Power calculations (post hoc) were carried out. With a 5% α level(P≤.05) and a sample size of 120 for most scales,there was 11% to 33% power to identify 5-point differences in NEI-VFQ overalland subscale scores between individuals who had less than 2 medical conditionsvs individuals who had 2 or more medical conditions and 30% to 86% power toidentify 10-point differences. Unadjusted P≤.05was considered statistically significant. Biological plausibility and clinicalsignificance were considered when determining whether findings were meaningful.Data were analyzed primarily using SAS software (SAS Institute Inc, Cary,NC) on a UNIX platform.

A total of 197 of 206 patients were included in the study (Figure 1); 120 of these 197 patients underwentNEI-VFQ interviews and visual acuity measurements 24 months after enrollmentand were included in this analysis. The median age of these patients was 77years (Table 1). Most patientswere non-Hispanic whites (98%), reflecting the underlying distribution ofneovascular AMD in the US population. Median visual acuities in the better-seeingand worse-seeing eyes were 20/100 and 20/500, respectively. Twenty-eight percent(33/120) of the individuals included in the cross-sectional analysis of 24-monthNEI-VFQ data were legally blind at that time (visual acuity ≤20/200 inthe better-seeing eye). The most common medical conditions at 24-month follow-upwere hypertension, arthritis or rheumatism, heart attack or angina, cancer,and diabetes (Table 2). Eighty-fivepercent of the patients had 1 or more medical conditions in addition to neovascularAMD (Figure 2). At 24 months, themedian SF-36 PCS score was 46 (mean, 44; range, 18-60) and the median MCSscore was 57 (mean, 55; range, 17-68) in the 120 patients included in theanalysis (data not shown).

Thirteen of 77 patients who were included in the study but excludedfrom the cross-sectional analysis of 24-month data died before this point.No large differences were found on demographic characteristics (age, gender,and race) between the 120 patients who were included in the analysis and theremaining 64 patients who had reached 24 months of follow-up but were excludedfrom the analysis because of missing data (data not shown). However, patientswho were included in the analysis less often had a major paralysis or neurologicproblems (5.0% vs 21.9%; χ²P = .001)and hearing problems (5.8% vs 18.8%; χ²P = .006) at 24 months of follow-up than patients who were excludedfrom the analysis.

The NEI-VFQ overall score and 9 of the 11 subscale scores were sensitiveto differences in visual acuity in the better-seeing eye (Table 3). That is, the NEI-VFQ overall and subscale scores werehighest in the subgroup with best visual acuity and lowest in the subgroupwith poorest visual acuity in the better-seeing eye. The exceptions were theperipheral vision and ocular pain subscales.

Linear regression models that included 1 of 6 medical conditions (hypertension,arthritis or rheumatism, heart attack or angina, cancer, diabetes, or hearingproblems) indicated that visual acuity in the better-seeing eye was the primaryinfluence on the NEI-VFQ overall score and on 9 of the 11 subscale scores(Table 4). In the unadjusted analysis(data not shown) and the analysis adjusted for individual medical conditions,the ocular pain and peripheral vision subscale scores were not associatedwith visual acuity in the better-seeing eye. The effect of visual acuity inthe better-seeing eye was consistent within each of the subscales and wasindependent of single medical conditions included in the linear regressionmodel (ie, the slope for visual acuity in the better-seeing eye was stable).Visual acuity in the better-seeing eye had the greatest impact on the driving,near activities, dependency, and distance activities subscales, where 3- to6-point differences were observed per 1-line difference in visual acuity inthe better-seeing eye after considering each of the 6 medical conditions.

Despite the dominant effect of visual acuity, some medical conditionshad a large effect on some of the NEI-VFQ subscales: the dependency subscalescores were, on average, 10.7 points lower for patients who had arthritisor rheumatism, the distance activities subscale scores were 12.2 points lowerfor patients with diabetes, and the ocular pain subscale scores were 7.6 pointshigher for patients with hypertension after adjustment for visual acuity inthe better-seeing eye (Table 4).The presence of less than 2 vs 2 or more medical conditions was consideredin alternate regression models; findings are summarized in Table 5. Slopes for visual acuity in the better-seeing eye werenearly identical to those in Table 4.The presence of multiple medical conditions was related to scores on the roledifficulties subscale. After taking into account visual acuity in the better-seeingeye, patients with 2 or more medical conditions averaged 7.6 points loweron the role difficulties subscale than patients with less than 2 medical conditions.

Linear regression models also were constructed using the SF-36 PCS andMCS scores as measures of general health (Table 6). Although the relation between NEI-VFQ scores and visualacuity in the better-seeing eye remained unchanged before and after inclusionof SF-36 summary scale scores, the SF-36 PCS and MCS scores contributed substantiallyto most of the NEI-VFQ subscale scores. A 10-point decrement in the PCS scoretranslated to a 4- to> 9-point decrement in the NEI-VFQ overall score andin 10 of 11 subscale scores after adjustment for visual acuity in the better-seeingeye and MCS score, and a 10-point decrement in the MCS score translated toa 4- to 8-point decrement in the NEI-VFQ overall score and in 7 of 11 subscalescores after adjustment for visual acuity in the better-seeing eye and PCSscore.

The SF-36 PCS and MCS scores were unrelated to visual acuity in thisstudy (data not shown). Visual acuity in better-seeing and worse-seeing eyesexplained 0.02% and 2.0% of variability in PCS scores, respectively, and 0.01%and 1.0% of variability in MCS scores, respectively.

In the unadjusted analysis and after adjustment for visual acuity inthe better-seeing eye, distribution of the ocular pain subscale scores onlywere statistically significantly different between men and women (data notshown). Men had, on average, 7.6-point higher ocular pain subscale scores(ie, less ocular pain) than women (P = .02) in theunadjusted analysis and after adjustment for visual acuity in the better-seeingeye. Of the 16 medical conditions, women had higher prevalences of arthritisand rheumatism (35% vs 15%; χ²P =.02) and back problems (11% vs 0%; Fisher exact test P =.02) than men. However, adjustment for either of these conditions and visualacuity in the better-seeing eye did not diminish the association of ocularpain with gender. Adjustment of ocular pain subscale scores for visual acuityin the better-seeing eye and general health status measured by the SF-36 PCSand MCS resulted in a gender difference of 5.7 points (P = .06).

We observed apparent gender differences for other NEI-VFQ subscales;however, none of these differences reached statistical significance. Aftertaking into account visual acuity in the better-seeing eye and PCS and MCSscores, men's scores, on average, were 7.7 points higher than women's scoreson the peripheral vision subscale (P = .13), 4.6points lower on the general vision subscale (P =.12), 4.5 points lower on the distance activities subscale (P = .19), and 3.8 points lower on the role difficulties subscale (P = .31). The distribution of visual acuity in the better-seeingeye was not different between men and women.

Age was independently associated with 3 of 11 NEI-VFQ subscale scores:role difficulties, ocular pain, and driving (data not shown). A 1-year incrementin age resulted, on average, in a 0.7-point decrement in role difficultiessubscale score (P = .02), a 0.5-point increment inocular pain subscale score (P = .03), and a 1.1-pointdecrement in driving subscale score (P = .02). However,age was associated only with the ocular pain subscale scores when the relationshipbetween age and NEI-VFQ scores was adjusted for visual acuity in the better-seeingeye or for visual acuity in the better-seeing eye and SF-36 PCS and MCS scores.A 1-year increment in age was associated with a 0.5-point increment in ocularpain subscale scores in both of these adjusted analyses (P = .03). Age distributions of men and women were similar.

The results of this study show that a 37-item version of the NEI-VFQis sensitive to differences in visual acuity in the better-seeing eye andto differences in general health. Visual acuity in the better-seeing eye wasconsistently associated with NEI-VFQ scores before and after adjustment forgeneral health, whether measured by individual medical conditions or by SF-36PCS and MCS scores. A 3-line decrement in visual acuity in the better-seeingeye was associated with a 5.1- to 17.1-point decrement in the NEI-VFQ overallscore and in 9 of the subscale scores. Ocular pain and peripheral vision subscalescores were unrelated to visual acuity in the better-seeing eye in this groupof patients.

Although NEI-VFQ scores were mostly driven by the level of visual acuityin the better-seeing eye, physical and mental aspects of general health alsocontributed to the NEI-VFQ scores: a 10-point decrement in the PCS score wasassociated with a 4- to 9-point decrement in NEI-VFQ scores after adjustmentfor visual acuity in the better-seeing eye and MCS score, and a 10-point decrementin the MCS score was associated with a 4- to 8-point decrement in NEI-VFQscores after adjustment for visual acuity in the better-seeing eye and PCSscore. All 6 of the medical conditions studied in detail had an impact on1 or more subscales of the NEI-VFQ after adjustment for visual acuity in thebetter-seeing eye, although many of the associations did not reach statisticalsignificance, perhaps because of the small sample size. However, 3 large associationswere observed: the dependency subscale scores were affected by the presenceof arthritis or rheumatism, the distance activities subscale scores by diabetes,and the ocular pain subscale scores by hypertension after accounting for visualacuity in the better-seeing eye. Hip and knee replacements were included inthe arthritis or rheumatism category and may explain some of the effect onthe dependency subscale. Patients with more severe diabetic retinopathy wereexcluded from the SST Pilot Study at baseline, but some patients with diabetesand no diabetic retinopathy or less severe diabetic retinopathy were included.It is possible that some of these patients developed visually impairing diabeticretinopathy by 24 months after enrollment, when this cross-sectional analysiswas performed, which could explain the association between diabetes and distanceactivities subscale scores. Other potential explanations could be the presenceof retinal ischemia, reduced peripheral vision, diminished capacity of visualpathways due to some disease process associated with diabetes or AMD, or chancealone. We cannot formulate a biologically plausible hypothesis for the associationof hypertension with higher ocular pain subscale scores (ie, less ocular pain),which may represent a chance finding.

The vision-specific role difficulties subscale was found to be associatedwith the presence of multiple nonocular medical conditions. Patients who had2 or more medical conditions had, on average, 7.6-point lower role difficultiessubscale scores than individuals with less than 2 medical conditions. A possibleexplanation for this could be that many patients who had only 1 nonocularmedical condition had hypertension only; however, patients who had more than1 medical condition often had potentially debilitating conditions in additionto hypertension, such as heart problems, arthritis, cancer, diabetes, or seriouslung problems. Several other NEI-VFQ subscales also seemed to be affectedby the presence of multiple medical conditions, although these values didnot reach statistical significance.

With respect to determining whether another sensory impairment in additionto vision loss would have an impact on NEI-VFQ scores, we noted that individualswho had hearing problems had 11-point lower social functioning subscale scores,12-point lower mental health subscale scores, and 14-point lower color visionsubscale scores, on average, than individuals without hearing problems aftertaking into account visual acuity in the better-seeing eye. Although thesevalues were not statistically significant, they indicate that hearing problemsmay lower several of the NEI-VFQ subscales scores more than would be expectedbased on the level of visual acuity alone. Second, these data suggest thatphysicians may need to pay particular attention to the needs of vision- andhearing-impaired patients with AMD, for whom several domains of QOL seem tobe affected.

It is possible that some of the effect of vision impairment on NEI-VFQscores may be eliminated when adjusting NEI-VFQ scores for SF-36 PCS and MCSscores. The SF-36 has been reported in some studies^21-23 tobe associated with aspects of vision, whereas other studies^9,24-26 haveshown small or insignificant associations. Visual acuity of better- or worse-seeingeyes was not associated with SF-36 PCS or MCS scores in our study of patientswith central vision loss due to AMD. Therefore, we would not expect to eliminatevision-related disability from the NEI-VFQ scores by adjusting for SF-36 PCSand MCS scores. In other studies, we recommend careful examination of associationbetween clinical vision of interest and SF-36 PCS and MCS scores before adjustmentis applied.

A potentially important finding from this study was that gender wasassociated with the ocular pain subscale of the NEI-VFQ, with women havingworse scores than men, on average. This association could not be explainedby any gender difference in random treatment assignment in the SST Pilot Studyor in the presence of medical conditions. This finding prompted an analysisof the bodily pain subscale scores of the SF-36. A similar finding was observed,with women having bodily pain scores that were 9 points lower (ie, worse),on average, than men. This finding also agrees with published findings forindividuals 65 years and older.¹⁶ Consistencyof the result in 2 different, but related, scales and the similarity of theresults to what has been reported elsewhere argues against a chance association.There also was some evidence that gender differences may exist in other domainsof the NEI-VFQ; however, this area needs further investigation.

A limitation of the present study was that information on medical conditionswas collected retrospectively. It is possible that the clinic charts infrequentlymay have included information for conditions that are not common or life threateningor that do not require surgical intervention. Any missing information on medicalconditions, and on severity and duration of conditions, may have caused residualconfounding when the association of interest, in this case QOL and vision,was adjusted imperfectly. Thus, incomplete information on coexisting medicalconditions could be another potential explanation, in addition to the smallsample size, for the few statistically significant associations of the NEI-VFQscores with individual medical conditions. However, general health measuredby the SF-36 PCS and MCS scores, which were collected prospectively, had asubstantial effect on most NEI-VFQ subscale scores. There are inherent problemsin collecting and summarizing information on individual medical conditions;perhaps the SF-36 is a more accurate and reliable way to assess the impactthat a particular condition or multiple medical conditions have on an individual.

It is likely that patients enrolled in the SST Pilot Study were healthierthan "typical" patients with AMD and choroidal neovascularization becauseof the outpatient setting and the eligibility and exclusion criteria imposed.It also is possible that patients with hearing problems may have less oftenparticipated in a study that included an interview. Thus, results regardingQOL cannot be generalized to all patients with neovascular AMD. Furthermore,patients who were not interviewed during follow-up had a worse health statusmore often than those who were interviewed.

The main strength of the study was that QOL and clinical informationwas collected prospectively using standard protocols and trained and certifiedpersonnel. The telephone interviewers, the traveling vision examiners, andthe individual who retrieved information on medical conditions did not haveany other relationship with the patients that could have biased data collection.Furthermore, patients with choroidal neovascularization secondary to AMD camefrom 10 clinical centers in different parts of the United States, yieldinga more heterogeneous study population than if all the patients had been identifiedand followed at a single clinical center.

In conclusion, NEI-VFQ scores were associated with the level of visualacuity in the better-seeing eye. Physical and mental aspects of general healthcontributed to NEI-VFQ scores. The extent to which general health contributedto NEI-VFQ scores was surprising since the questionnaire is supposed to measurevision-related disability. Thus, adjustment for general health should be consideredwhen comparing NEI-VFQ scores between groups of patients.

Corresponding author and reprints: Päivi H. Miskala, PhD, WilmerClinical Trials and Biometry, The Johns Hopkins University, 550 N Broadway,Ninth Floor, Baltimore, MD 21205 (e-mail: pmiskala@jhmi.edu).

Submitted for publication September 12, 2002; final revision receivedJuly 22, 2003; accepted August 14, 2003.

The SST Pilot Study was funded through numerous private and public sourcesthat have been published previously (Am J Ophthalmol. 2000;130:408).Funding for the additional data collection and analysis for this ancillarystudy was provided by the Michael B. Panitch Stop AMD Fund (Baltimore).

We thank Barbara S. Hawkins, PhD, Michael X. Repka, MD, Marie Diener-West,PhD, and Barbara Martin, PhD, for their advice on this project; Barbara S.Hawkins, PhD, for her review of this manuscript and editorial comments; andthe SST principal investigators and clinic coordinators at the following clinicalcenters for providing access to the medical records: Emory University EyeCenter, Atlanta, Ga; The Wilmer Ophthalmological Institute, Baltimore; BarnesRetina Institute, St Louis, Mo; Cole Eye Institute, Cleveland, Ohio; RetinaVitreous Associates of Kentucky, Lexington; the Department of Ophthalmology,Duke University Medical Center, Durham, NC; Associated Retinal Consultants,Royal Oak, Mich; Schatz, McDonald, Johnson & Ai, San Francisco, Calif;Retina Vitreous Consultants, Pittsburgh, Pa; and Jules Stein Eye Institute,Los Angeles, Calif.

A complete listing of the SST investigators has been published elsewhere(Am J Ophthalmol. 2000;130:405-406).

The authors of this article assume authorship responsibility and hadcomplete access to the data.