 
Blind Weaver - photograph by
Eudora Welty, 1930s
índice
-
Objective: To determine the association between performance on selected tasks
of everyday life and impairment in visual acuity and contrast sensitivity.
-
Methods: Visual acuity and contrast sensitivity were obtained on a
population-based sample of 2520 older African American and white subjects.
Performance was assessed on mobility, daily activities with a strong visual
component, and visually intensive tasks. Disability was defined as performance
less than 1 SD below the mean. Receiver operating characteristic curve analyses
were used to evaluate the sensitivity and specificity of thresholds in acuity
and contrast loss for determining disability.
-
Results: Both visual acuity and contrast sensitivity loss were associated with
decrements in function. The relationship of function to the vision measures was
mostly linear, therefore, receiver operating characteristic curves were not
helpful in identifying cutoff points for predicting disabilities. For mobility
tasks, most persons were not disabled until they had significant acuity loss
(logMAR visual acuity >1.0 or <20/200) or contrast sensitivity loss (0.9 log
units contrast sensitivity). For heavily visually intensive tasks, like reading,
visual acuity worse than 0.2 logMAR (20/30) or contrast sensitivity worse than
1.4 log units was disabling.
-
Conclusions: Both contrast sensitivity and visual acuity loss contribute
independently to deficits in performance on everyday tasks. Defining disability
as deficits in performance relative to a population, it is possible to identify
visual acuity and contrast loss where most are disabled. However, the cutoff
points depend on the task, suggesting that defining disability using a single
threshold for visual acuity or contrast sensitivity loss is arbitrary.
-
The link between visual impairment and physical disability, psychosocial
distress, and general quality of life has been a growing area of research, and a
number of studies including our own have documented the decrements in function
associated with loss of various components of visual function.1- 9 In all but a
few of these, the measure of visual impairment was confined to either
self-report of visual loss, or a measure of visual acuity. Yet, there are data that suggest other measures of visual function are also
important for predicting functional loss. Rubin et al3 have shown that a 2-fold
reduction in contrast sensitivity is associated with 3- to 5-fold odds of
self-report of difficulty on everyday tasks, and the effect is independent of
the contribution of visual acuity loss. Others have linked loss of contrast
sensitivity to mobility and balance decrements.10,11
-
In the Salisbury Eye Evaluation (SEE) Project, our goal is to determine, in a
population-based sample of older adults, the relationship of visual impairment
with functional disability. Measures of functional disability include both
self-report of difficulty and actual performance on a variety of tasks.12 The
tasks chosen are meaningful for comparisons with tasks of everyday life, as we
have previously shown that performance on these tasks mirrors the performance on
the same tasks conducted in the home.13 In this article, we show the
relationship of decrements in visual acuity and contrast sensitivity to
decrements in the performance on a variety of tasks, and try to determine
meaningful cutoff points in loss of visual acuity or contrast that are rooted in
functional loss.
Population
The SEE Project is a population-based, longitudinal study of the effect of
visual impairment, and age-related eye diseases, on functional status in older,
community-dwelling adults.12 To achieve the aims of this project, a random
sample of residents of Salisbury, Md, aged from 65 to 84 years, was recruited
for a home interview and an examination at the SEE clinic. The sample was
selected from the Health Care Financing Administration Medicare database.
Eligibility criteria excluded those who were institutionalized or completely
housebound, and those who scored less than 18 on the Mini-Mental State
Examination (MMSE).14 Written, informed consent was obtained at the home
interview in accord with the tenets of the Declaration of Helsinki. Details on
the population and recruitment are described elsewhere.15 In summary, of the
original sample, 73% participated in the home interview and 65% participated in
both the interview and the clinical examination.
Permission was also sought to administer a 12-question screener questionnaire to
both the refusals and the participants, to investigate the comparability between
those for whom data were available and those who refused. Of the 1301 refusals
to the clinic examination or the home questionnaire, 65% agreed to answer the
questions on the screener. There were no differences by age, race, or sex
between the refusals who answered the screener and those who did not answer the
screener.15 There was no difference in participation rates by race;
participation rates declined with age from 68% in the age group 65 through 69
years to 55% in age group 80 through 84 years. There was no difference in the
sex- and age-adjusted proportion rating their vision as 6 or better, on a scale
of 1 to10 where 10 was excellent. Among participants, 82.9% were 6 or better
compared with 84.0% among refusals. These data provide some assurances that the
sample of 2520 participants did not seem to be biased on self-reported vision
status.
Measures of health
The General Health Questionnaire was administered in the clinic; the subscale
on depressive symptoms was used as a measure of depression. Cognitive status was
assessed at home visit using the MMSE; possible scores for this study range from
18 to 30.
Data on diabetes mellitus were based on self-report, validated by use of insulin
or oral hypoglycemics, or by hospital or physician records. For those who did
not report having diabetes mellitus, we included as having diabetes mellitus
those with a hemoglobin A1c value greater than 7%.16 Hypertension was based on
self-report plus evidence of taking an antihypertension medication. In addition,
blood pressure was measured in a sitting position, 3 times, according to a
standard protocol.17 Persons who did not report hypertension but had an average
diastolic pressure of 90 mm Hg or higher or a systolic pressure of 160 mm Hg or
higher were classified as subjects possibly having hypertension for our study.
Other conditions were based on self-report such as stroke, cancer, parkinsonism,
arthritis, myocardial infarction, congestive heart failure, and pulmonary
problems. We calculated an index of number of comorbid conditions(excluding
cataract).
Measures of vision
Details on vision testing have been described previously.18 All tests were
administered by a trained technician using forced-choice procedures. In essence,
distance acuity was measured using Early Treatment Diabetic Retinopathy Study
charts, monocularly and binocularly, with habitual correction followed by best
correction after subjective refraction. Visual acuity was scored as the number
of letters read correctly, and converted to logMAR scores (logMAR is the
logarithm of the minimum angle of resolution). Contrast sensitivity was measured
with the Pelli-Robson letter sensitivity test, scored as number of letters read
correctly. These data are reported as log contrast sensitivity(with the number
of letters read of 48 letters indicated). Our previous article demonstrated that
models using vision in the better seeing eye were as good as models using
measures of binocular visual acuity in predicting self-reported function.1 Other
analyses also indicated negligible contribution from the worse seeing eye in
models of contrast sensitivity loss and function. Therefore, for these analyses,
we used presenting acuity and contrast sensitivity in the better seeing eye.
Outcome measures: performance on tasks
Details on the tasks and scoring procedures are described elsewhere.12,13 For
the purposes of this study, we chose performance-based tests in 3 categories:
mobility, tasks of daily living that require a visual component, and visually
intensive tasks. Four measures of mobility were included: a timed 4-m walk, a
timed stair ascent, stair descent, and a timed get-up-and-go test (the latter
requires the participant to get up from a chair with arms, and step away from
the chair). The tasks of daily living included 3 timed measures: inserting a key
in a lock, inserting a plug in a socket, and dialing a rotary telephone. The
times to complete the task for mobility items and items of daily living were
converted to speed. All tasks were performed in the clinic under standard
conditions, including lighting between 400 and 600 lux. The visually intensive
tasks included face recognition and reading speed. Face recognition involved
presenting participants with a panel of 4 faces (3 of the same individual) in
different poses, and having them select the one that differed from the other 3.1
The faces subtended 2.5° (horizontal) by 3.2° (vertical). Fifteen sets of panels
were presented on a monochrome monitor. Data were recorded as number of faces
identified correctly. Reading speed involved reading aloud short passages of
text presented on a computer screen for 15 seconds. Four print sizes were
tested; data for newsprint-sized text of 0.26° were used in these analyses.
Reading speed was calculated as the number of words read correctly per minute.
Data analyses
The association between contrast sensitivity and visual acuity was examined
using a plot of logMAR vs contrast for right eyes, with correlation coefficient
and R2 calculated to determine the degree of independence of the 2 measures. The
distributions of the speeds for the functional tests, and the score for face
recognition, were examined for normalcy. All speeds of task performance were
converted to z scores by subtracting the mean and dividing by the SE to permit
meaningful comparisons. Simple smoothed graphs, followed by linear spline
regression analyses that adjusted for age, sex, race, education, cognition, and
number of comorbid conditions, were done. Linear regression analyses were
carried out to determine the joint and unique contribution of visual acuity and
contrast sensitivity loss, adjusted for other predictors, to disability. The
unit of change was 0.1 logMAR for visual acuity, equivalent to 1 line change,
and 0.05 log contrast sensitivity, equivalent to 1 letter for contrast
sensitivity.
To characterize disability, we used a cutoff of 1 SD below the population mean
for performance on a test. Receiver operating characteristic (ROC) curves were
used to evaluate the sensitivity and specificity of acuity and/or contrast
sensitivity loss for predicting disability at this level of dichotomization.
Graphs were also created of the proportion disabled by level of visual acuity
and contrast sensitivity decrement.
The population in the SEE Project study was 58% female, and 26% were African
American (Table 1).
The distributions of each of the performance-based outcome
measures were close to normal, with the exception of face recognition where the
distribution was skewed to high performance. The mean and SD for each of the
outcome measures are given in Table 2.
Table 2
As expected, those with very good contrast sensitivity generally also had good
visual acuity as well. However, there was a substantial spread of visual acuity
scores in those with contrast sensitivity below 1.65 log units(36 letters)
(Figure 1). Overall, the correlation was 0.81 with an R2 of 0.66. (Data shown
for right eye.)
Figure 1
For all the functional outcomes, the relationship with either visual acuity or
contrast sensitivity was close to linear, or linear with changes in the slope.
There was no evidence for a threshold of either acuity or contrast sensitivity
below which a sharp decrement in performance was observed; rather, gradual
declines in performance were observed concomitant with modest decrements in
visual acuity or contrast sensitivity. An example of a linear relationship with
a change in slope (where a spline regression was appropriate) is shown in Figure
2A, where the slope of the relationship with speed of plug insertion changes
with loss of contrast sensitivity.
Figure 2A and Figure 2B
A straight linear relationship is typified by
Figure 2B, the relationship between logMAR visual acuity and number of faces
read correctly. We created linear regression models, including spline terms
where significant, for visual acuity and contrast sensitivity separately for
each of the outcome measures (Table 3).
Table 3
For all the outcome measures
(transformed into z scores so change is per SD unit), significant associations
were observed with age, race, sex, cognitive score, and number of comorbid
conditions, as well as the vision variables, and these confounders were also
included.
To investigate the independent contributions of visual acuity loss and contrast
loss, both variables were included in final models for each outcome in the set
of mobility items, activities of daily living items, and visually intensive
tasks (Table 4).
Table 4
All outcomes were significantly related to both measures of
vision, except visual acuity was unrelated to decreased speed of stair ascent or
descend once contrast sensitivity was included in the model. These data suggest
that for daily living tasks and reading speed, the sharpest declines were
associated with a contrast sensitivity between 0.85 and 1.60 log units (20-35 of
48 letters). For daily living tasks, decrements were observed until visual
acuity was about 0.7 logMAR(20/100); for reading speed, the sharpest declines
were observed until visual acuity reached about 0.5 logMAR (20/60), then leveled
off.
Except possibly for reading speed, there are no commonly accepted performance
standards for "disability" that could be applied to our data. Therefore, we
arbitrarily defined disability for each outcome as the value at 1 SD below the
mean value (the values are shown in Table 1). The essentially linear
relationship between visual loss and functional loss, and the effect of other
comorbid conditions besides visual loss on these functional decrements,
suggested that there is no optimal cutoff for either visual acuity or contrast
that was both highly sensitive and highly specific for disability in these
domains. The ROC curves for each outcome confirm this (Figure 3). Except for
reading, where for example at logMAR visual acuity value 0.176 there was 50%
sensitivity and over 95% specificity, the other outcomes have true- and
false-positive rates that are almost equal. Thus, for any arbitrarily defined
cutoff for disabling visual acuity loss, there would be persons with less visual
loss with only slightly less risk of disability. Similar ROC curves were
observed for contrast sensitivity (data not shown).
Figure 3
In light of these findings, we took another approach to evaluating the effect of
visual acuity and contrast sensitivity loss on disability: the determination of
the level of visual acuity loss or loss of contrast sensitivity at which more
than half of the affected population would be expected to be disabled(ie,
performing at levels <1 SD below that of the population mean). For each of the
domains, that level was different, as might be expected by the visual demands of
the tasks (Table 5).
Table 5
Mobility was unaffected by visual acuity loss until levels
worse than 1.0 logMAR (20/200) and even then, a sizable number were not
disabled. However, for daily living tasks and visually intensive tasks, the
cutoff point for 50% disabled occurred at more modest levels of visual acuity
and contrast sensitivity loss. Disability in reading, that is, reading fewer
than 90 words per minute, was observed in the 50% of the population with visual
acuity worse than 0.2 logMAR (20/30) and in 90% of those with acuity worse than
0.3 logMAR(20/40). Disability in reading speed was also sensitive to contrast
sensitivity loss, with 50% disabled among those with contrast sensitivity of 1.4
log units or worse (seeing ≤31 letters), and 90% disabled among those with
contrast sensitivity of 0.55 log units (14 letters) or fewer.
The loss of vision has been associated with numerous deleterious outcomes,
from decrements in self-report of function,3- 7,19,20 use of social services,21
nursing home admissions,22 and even mortality.23- 25 However, few
population-based studies have attempted to determine the effect of vision on
actual performance in tasks of everyday life. Performance-based tests of
function provide data that is different from self-reported function in several
important ways.26 First, standardized testing provides a continuous measure of
performance across the population that is not subject to vagaries of
self-report. Self-report of function combines an individual's assessment of his
or her ability, his or her expectation of that ability, and his or her
understanding of the degree of difficulty involved in carrying out the task in
the presence of any limitations. Performance-based testing requires the
individual to simply do the task. Second, it has been argued that cognitively
impaired persons can perform a task, whereas self-report of difficulty
performing the task may be inaccurate. In our study, severe cognitive impairment
was not an issue. Finally, performance-based tests usually provide a continuous
measure of function that may be more sensitive to change over time compared with
categorical data. Performance-based tests as carried out in research settings
are criticized as unrepresentative as they are conducted under ideal conditions
(lighting among others). However, we have found that performance at the clinic
and the performance at home on the same tasks were highly correlated.13
Moreover, normative data in representative populations enable the determination
of performance levels that are distinctly poorer than the average performance.
If curves indicating a sharp dropoff in performance were observed, it would
enable ROC curve analyses to determine possible cutoff point values for visual
loss that had good sensitivity and specificity for poor performance.
In our population-based study of 2520 persons aged from 65 to 84 years, we found
reasonably normal distributions of performance on tasks in domains of mobility,
activities of daily life, and visually intensive tasks (with the exception of
the face recognition test, which was skewed to high performance). Moreover, both
visual acuity and contrast sensitivity were significantly related to performance
on all the tasks, except stair ascent and descent where contrast sensitivity
alone predicted performance. Our data suggest that contrast sensitivity loss and
visual acuity loss each contribute independently to decrements in performance.
Together with other cofactors, the vision variables predicted anywhere from 21%
to 57% of the variance associated with the outcome of interest.
Linear relationships between the functional tasks and visual acuity and
contrast, with some spline regressions where the slope changed, were the best
characterization of the association. Thus, decrements in visual acuity from 0.0
logMAR (20/20) or better, or in contrast from about 1.85 log units(40 letters),
were likely to have a negative effect on performance, and any cutoff point to
determine visual disability would be arbitrary. This finding was further
supported by the use of ROC curve analyses.
To explore the issue of vision loss associated with functional disability using
ROC curve analyses required us to define disability for these tests. We chose to
define disability as performance less than 1 SD below the population mean. Our
selection allowed us to have reasonable numbers in the disabled group, and yet
have disability cutoff points that were marked. For example, the cutoff point of
less than 1 SD defined as disabled those who read standard newsprint-size
letters at fewer than 90 words per minute, or recognized about half the faces in
the test, or required close to 19 seconds to insert a key into a lock.
Regardless of the definition of disability, though, no level of vision (either
visual acuity or contrast sensitivity) provided reasonable values for
sensitivity and specificity in determining disability. This result was not
surprising, given the nearly linear relationship of the vision variables with
performance on the tasks, and the contribution of other factors to performance.
An alternative approach is proposed: determining at what level of visual acuity
or contrast sensitivity loss more than 50% of the population is disabled(in our
study, disabled was defined as performance on tasks 1 SD below the population
mean, but it could be defined in other ways). Using this approach, the variation
in level of visual acuity or contrast sensitivity loss is easily observed,
according to the degree of visual demand required to do the task. For mobility
tasks, less than 50% of persons are disabled until visual acuity becomes worse
than 1.0 logMAR (20/200). We were unable to determine at what point beyond 1.0
logMAR the cutoff point would be reached because of small numbers in our
population with vision worse than 1.0 logMAR.
However, for reading speed or face recognition, visual acuity loss as modest as
0.3 logMAR (20/40) and contrast sensitivity loss of 1.35 log units(30 letters)
was associated with disability, values of impairment that are otherwise
considered rather minor. Leat et al27 suggest that a log contrast sensitivity
score of less than 1.5 log units (33 letters) is impaired, while a score of 1.05
would result in disability. Our data suggest that the level of disability
associated with vision loss is highly dependent on the task and the visual
demands of the task. Thus, the use of an arbitrary cutoff point of visual
impairment to define disability may well ignore disabilities experienced at
lesser levels of visual acuity or contrast loss.
There are limitations to our study that must be discussed. First, this is a
cross-sectional analysis of associations. Because we do not know the time of
visual loss or the onset of difficulty in performing tasks, these associations
should not be interpreted as causal. Moreover, it is likely that those with
visual loss of longer duration have arrived at compensatory strategies for
performing tasks that persons with more recent onset have not yet learned. It
will be critical to add a longitudinal component to this study.
In addition, the level of visual acuity or contrast sensitivity loss associated
with disability will be sensitive to the definition of disability. The question
of whether a significant slowing of speed in performing a task is a true problem
in everyday life of our participants was not addressed. For a few, a reading
speed of fewer than 90 words per minute was reported as not a significant
problem, while others reported extreme difficulty or inability to read at such a
speed.28 For this analysis, we chose 1 SD below the mean, and that choice
resulted in performances that were clearly abnormal. However, defining
disability as 2 SDs below the mean would result in great loss of visual acuity
and contrast sensitivity necessary for disability. Regardless of the cutoff
point for disability, though, the same linear function will be operative.
Finally, many factors, apart from vision, have an effect on the ability of
persons to perform everyday activities. In the SEE Project, we found that age,
sex, race, education, and comorbid conditions also affected performance on tasks
in several domains. Our study was carried out in an aging population, so it is
not surprising that these other factors also are involved in the performance on
tasks. We have tried to control for these factors by models that adjust for
confounding. However, it is clear that using a visual acuity or contrast
sensitivity cutoff point to predict performance on tests of function, even
highly visually demanding tests, will never be able to simultaneously achieve
reasonable sensitivity and specificity.
This study has documented the independent contributions of visual acuity and
contrast sensitivity loss to decrease in performance in a variety of tasks of
daily living. The relationships between visual loss and performance are
essentially linear, with no obvious thresholds for sharp declines in
performance. Using a population-based approach to defining disability, it is
apparent that varying levels of visual acuity or contrast sensitivity loss that
are associated with most of the population at that level of loss being disabled
can vary, depending on the task. Choosing any cutoff point for disability is
arbitrary. A more scientific approach to disability would consider the level of
function for a given task that is required, and the likelihood that persons with
visual acuity or contrast sensitivity impairments would be able to perform at
that level.
-
Rubin GSMunoz BBandeen-Roche KWest SK Monocular versus binocular
visual acuity as measures of vision impairment and predictors of visual
disability. Invest Ophthalmol Vis Sci. 2000;413327- 3334
-
Rubin GSRoche KBPrasad-Rao PFried LP Visual impairment and
disability in older adults. Optom Vis Sci. 1994;71750- 760
-
Rubin GSBandeen-Roche KBHuang GH et al. The association of multiple
visual impairments with self-reported visual disability: SEE project.
Invest Ophthalmol Vis Sci. 2001;4264- 72
-
Hakkinen L Vision in the elderly and its use in the social
environment. Scand J Soc Med Suppl. 1984;355- 60
-
Havlik RJ Aging in the eighties, impaired senses for sound and light
in persons age 65 years and older: preliminary data form the Supplement
on Aging to the National Health Interview Survey: United States,
January-June 1984. Adv Data Vital Health Stat. 1986;1251- 7
-
Carabellese CAppollonio IRozzini R et al. Sensory impairment and
quality of life in a community elderly population. J Am Geriatr Soc.
1993;41401- 407
-
Rudberg MAFurner SEDunn JECassel CK The relationship of visual and
hearing impairments to disability: an analysis using the longitudinal
study of aging. J Gerontol. 1993;48M261- M265
-
Salive MEGuralnik JGlynn RJChristen WWallace RBOstfeld AM
Association of visual impairment with mobility and physical function. J
Am Geriatr Soc. 1994;42287- 292
-
Appollonio ICarabellese CMagni EFrattola LTrabucchi M Sensory
impairment and mortality in an elderly community population: a six-year
follow-up study. Age Ageing. 1995;2430- 36
-
Marron JABailey IL Visual factors and orientation-mobility
performance. Am J Optom Physiol Opt. 1982;59413- 426
-
Turano KADagnelie GHerdman SJ Visual stabilization of posture in
persons with central visual field loss. Invest Ophthalmol Vis Sci.
1996;371483- 1491
-
West SKMunoz BRubin GSSchein OD et al. Function and visual
impairment in a population based study of older adults; SEE project.
Invest Ophthalmol Vis Sci. 1997;3872- 82
-
West SKRubin GSMunoz BAbraham DFried LPfor the Salisbury Eye
Evaluation Project Team, Assessing functional status. J Gerontol A Biol
Sci Med Sci. 1997;52M209- M217
-
Folstein MFFolstein SEMcHugh PR "Mini-mental state". J Psychiatr
Res. 1975;12189- 198
-
Munoz BWest SRubin GSSchein ODFried LPBandeen-Roche K Who
participates in population based studies of visual impairment? the SEE
project experience. Ann Epidemiol. 1999;953- 59
-
Peters ALDavidson MBSchriger DLHasselblad V A clinical approach for
the diagnosis of diabetes mellitus. JAMA. 1996;2761246- 1252
-
The 1988 report of the Joint National Committee on Detection,
Evaluation, and Treatment of High Blood Pressure. Arch Intern Med.
1988;1481023- 1038
-
Rubin GSWest SKMunoz B et al. for the Salisbury Eye Evaluation
Project, A comprehensive assessment of visual impairment in a population
based study of older adults: SEE study. Invest Ophthalmol Vis Sci.
1997;38557- 568
-
Mangione CMGutierrez PRLowe GOrav EJSeddon JM Influence of
age-related maculopathy on visual functioning and health-related quality
of life. Am J Ophthalmol. 1999;12845- 53
-
Mangione CMPhillips RSLawrence MGSeddon JMOrav EJGoldman L Improved
visual function and attenuation of declines in health-related quality of
life after cataract extraction. Arch Ophthalmol. 1994;1121419- 1425
-
Wang JJMitchell PSmith WCumming RGAttebo K Impact of visual
impairment on use of community support services by elderly persons: the
Blue Mountains Eye Study. Invest Ophthalmol Vis Sci. 1999;4012- 19
-
Wang JJMitchell PSmith WCumming RGLeeder SR Incidence of nursing
home placement in a defined community. Med J Aust. 2001;174271- 275
-
Taylor HRMcCarty CANanjan MB Vision impairment predicts five-year
mortality. Trans Am Ophthalmol Soc. 2000;9891- 96
-
Klein RKlein BEMoss SE Age-related eye disease and survival: the
Beaver Dam Eye Study. Arch Ophthalmol. 1995;113333- 339
-
Thompson JRGibson JMJagger C The association between visual
impairment and mortality in elderly people. Age Ageing. 1989;1883- 88
-
Guralnick JMLa Croix AZ Assessing physical function in older
populations. Wallace RBWoolson RFeds The Epidemiologic Study of the
Elderly. Oxford, England Oxford University Press1992;159- 181
-
Leat SJLegge GEBullimore MA What is low vision? Optom Vis Sci.
1999;76198- 211
-
Friedman SMuñoz BRubin GSWest SKBandeen-Roche KFried LPfor the
Salisbury Eye Evaluation Project Team, Characteristics of discrepancies
between self-reported visual function and measured reading speed. Invest
Ophthalmol Vis Sci. 1999;40858- 864
ϟ
How Does Visual Impairment Affect Performance on Tasks of Everyday Life?
Sheila K. West, PhD; Gary S. Rubin, PhD; Aimee T. Broman, MS; et al Beatriz
Muñoz, MSc; Karen Bandeen-Roche, PhD; Kathleen Turano, PhD;
for the SEE Project Team
Arch Ophthalmol. 2002;120(6):774-780. doi:10.1001/archopht.120.6.774
(June 2002)
Δ
|
