Вы находитесь на странице: 1из 7

Self-Reported Assessment of Dry Eye in a PopulationBased Setting

Karen Bandeen-Roche,* Beatriz Munoz,^ James M. Tielsch^X Sheila K. West,\


and Oliver D. Scheinf

Purpose. To report on a subjective dry eye assessment method for use in large-scale research,
to evaluate its application in a population-based study of dry eye among elderly persons
in the United States, and to apply novel techniques to improve simple questionnaire item
summaries.
Methods. A dry eye questionnaire was administered to a population-based sample of 2520
volunteers ages 65 to 84 years in Salisbury, Maryland. Individual symptoms and signs, counts
of symptoms and signs, and latent class model summary of item responses were evaluated for
validity and internal consistency.
Results. Approximately 15% of participants reported experiencing one or more of six dry eye
symptoms often or all the time; 20% reported experiencing three or more symptoms sometimes, often or all the time. Four groups were derived on the basis of symptomatology, using
latent class analysis. The groups exhibited face validity, revealed symptom patterns that added
specificity to simple symptom counts, and were qualitatively similar when derived separately
within population subgroups. Internal consistency was moderate (Cronbach's alpha = 0.61),
indicating some variability in reporting.
Conclusions. Dry eye symptoms are commonly reported in a representative elderly population.
Symptom data were moderately consistent, suggesting their usefulness for dry eye assessment
if properly summarized. A latent class summary revealed biologically meaningful summary
patterns of symptoms reported in this population and holds promise for use in risk factor
investigations and in clinical trials. Invest Ophthalmol Vis Sci. 1997; 38:2469-2475.

U r y eye conditions comprise a syndrome the pathogenesis of which is increasingly well understood, but
for which substantial problems of measurement and
classification remain.1 Difficulties include nonspecific
diagnostic profiles, uncertain biologic significance
and interpretation of standard objective test results,2"4
and the frequent lack of concordance between objective test results and self-reported symptoms and of
symptomatology among patients who appear clinically
similar.5'6 Most of the assessment tools that exist have

From the * Department of Biostatislics, the -\Dann Center for Preventive


Ophthalmology, The Wilmer Eye Institute, Department of Ophthalmology, and the
XDRpartme.nt of international Health, Johns Hopkins University, Baltimore,
Maryland.
Supported in part, by National Institude on Aging, Bethesda, Maryland, grant
PO1AGI0IS4. Dr. West is a Research to Prevent Blindness Senior Scientific
investigator.
Submitted for publication September 20, 1996; revised April 30, 1997; accepted
June 16, 1997.
Proprietary interest category: N.
Reprint Requests: Dr. Karen J. Bandeen-Roche, E-3006 Hygiene, 615 N. Wolfe
Street, Baltimore, MD 21205.

been developed using clinic-based populations, leaving uncertain their utility in population-based settings.
Symptom assessment is a key component of clinical dry eye diagnosis. However, the literature reports
on few standardized symptom questionnaires designed for formal research purposes.7'8 Supporting the
potential usefulness of this approach, one clinic-based
study found that a screen based on symptoms alone
was better able to discriminate cases from noncases
than one based on symptoms plus diagnosed signs.9
For screening and in large-scale research, methods for
assessing dry eye must be brief and noninvasive. Here,
where symptom questionnaires may be particularly appropriate, the development of effective self-report assessment methods is especially needed.
The two principal challenges in developing a
questionnaire for dry eye are that symptom profiles
vary substantially among clinically similar people and
that dry eye diagnosis is typically individualized and
dependent on clinical judgment that is not systematic.5 These challenges raise two methodologic issues

Investigative Ophthalmology & Visual Science, November 1997, Vol. 38, No. 12
Copyright Association for Research in Vision and Ophthalmology

Downloaded From: http://iovs.arvojournals.org/pdfaccess.ashx?url=/data/Journals/IOVS/933420/ on 04/24/2016

2469

2470

Investigative Ophthalmology & Visual Science, November 1997, Vol. 38, No. 12

in applying a standardized instrument in populationbased research: the need for effective scoring or other
method for summarizing symptoms to filter out incidental fluctuations in reporting and appropriate instrument validation. Internal validation in populationbased research is not feasible in the absence of a
broadly applicable gold standard diagnositic approach, and the opportunity for external validation is
also limited because of the acknowledged disparity
between objective and subjective dry eye tests. This
leaves, as feasible alternatives, the determination of
whether symptom patterns are consistent with biologic
mechanisms that give rise to dry eye (face validity)
and whether comparable populations report symptoms similarly (cross-validation).
The criteria for determining an effective method
for summarizing symptoms will vary according to the
scientific purpose. For the purpose of detecting clinically significant dry eye in a person, one reasonable
approach might be to identify whether a symptom
is experienced persistently (a yes/no summary). In
contrast, in a population-based study in which the object is to identify risk factors for dry eye, the investigator must define dry eye status and severity in participants as precisely as possible. To assess study participants' status, the most common strategy is to create a
symptom scorefor example, the number of symptoms reported with at least a given severity or frequency. To assess disease status, this works well when
items reflect severity for a single cause. Dry eye does
not fit this paradigm but rather includes a heterogeneous collection of disorders with diverse pathogenesis involving at least two separate causes: insufficient
tear flow and abnormal tear content.1'10 This suggests
the potential importance of accounting not only for
severity (persistence or number) of symptoms, but also
for patterns of symptoms. We propose that grouping
persons who report symptoms in similar patterns may
serve as a more useful construct for summarizing dry
eye symptoms than does simply calculating severity
scores.
In this report, we will use the statistical method
of latent class analysis11"13 to formalize the idea of
grouping people with similar symptom patterns. This
and related methods have served to clarify a symptombased definition of health status in other substantive
areasfor example, gerontology.14 Latent class analysis views an overall study population as comprised of
several "diagnostic" groups, possibly including "no
disease," each of whose members report symptoms
similarly. Key quantities described by the analysis are
the prevalence of each group and, within groups, the
proportions of people who report each symptom. These
latter quantities identify the qualitative composition
of the derived groups. For instance, only a small proportion of disease-free people should report dry eye
complaints, whereas most people with conditions

caused by insufficient tear flow would be expected


to report symptoms associated with dryness. Allowing
such anomalous reporting as occasional symptoms in
disease-free people acknowledges that symptoms may
arise from conditions other than dry eye and that no
instrument perfectly discriminates diagnostic groups.
Synthesizing, latent class analysis identifies the primary patterns in which symptoms are reported, which
can then be checked for biologic significance, and
summarizes symptoms by associating subjects with reporting profiles. Thus, it is well suited to two of our
objectives in this report: to validate a dry eye questionnaire that we have applied in a population-based setting and to derive usable summaries of reported symptoms for syndrome description and for use as outcomes in risk factor analyses.
Considering the variability in individual symptom
reporting by apparently similar patients, self-report assessment methodology for broad-based populations
must include strategies for summarizing the responses
to specific symptom questions. Evaluation of such
methods cannot be undertaken appropriately in a
clinical setting, where the effectiveness of the questionnaire may vary widely, depending on the severity
distribution or other characteristics within a specific
case series. Here, we report on a subjective dry eye
assessment method for large-scale use, and we evaluate
its application in a population-based study of dry eye
among the elderly in the United States.
METHODS
The Questionnaire
Our questionnaire was designed for use in a large,
population-based study in conjunction with objective
dry eye tests and an examination protocol that included many assessments unrelated to dry eye. For
this type of use, a streamlined instrument was vital.
Pilot testing of multiple questions about dry eye symptoms in a clinical setting led to a final instrument
containing questions about six symptoms that effectively discriminated dry eye cases from control subjects
(see Appendix 1). For each symptom endorsed, the
subject was asked to indicate the frequency of the
symptom (rarely, sometimes, often or all the time).
For subsequent discussion, we abbreviate the six symptoms: eyes feel dry, eyes feel gritty or sandy, eyes burn,
eyes appear red, crust forms on lashes, and eyelids stick
shut in the morning. Informed consent was obtained
from all study participants. The tenets of the Declaration of Helsinki were followed, and all procedures
were approved by the Joint Committee on Clinical
Investigation of The Johns Hopkins University School
of Medicine.
Study Population
We administered our questionnaire as part of the baseline examination of the Salisbury Eye Evaluation

Downloaded From: http://iovs.arvojournals.org/pdfaccess.ashx?url=/data/Journals/IOVS/933420/ on 04/24/2016

Subjective Assessment of Dry Eye

2471

l. Demographic Characteristics of
Sample (n = 2520)

TABLE

Age (years) at Lime of clinic examination

65-69
70-74
75-79
80-86

Sex
Male
Female
Race
White
Black
Education
<7 years
7-12 years
>12 years
Mini-Mental State Examination Score
<20

20-24
25-29
>29

31.0
33.1
22.0
13.9
42.1
57.9

lation, to validate our questionnaire, to explore


whether summarizing symptoms by die patterns in
which they were observed to occur was more informative than counting symptoms, and to assess the variation in reporting symptoms. We calculated the frequency distributions of diy eye symptoms in our population separately for each complaint we assessed. On
the basis of these distributions, we calculated two positive-versus-negative symptom definitions for use in
subsequent analyses. For definition 1, a response was
defined as positive when the subject reported a symptom to occur sometimes, often or all the time and as nega-

73.6
26.4
8.4

63.7
28.1
1.2

15.0
65.4
18.4

(SEE), a population-based study of the risk factors for


eye diseases and the impact of eye disease and visual
impairment on physical disability in older people. Details on sample selection and recruitment are given
elsewhere.15 Briefly, 2520 volunteers, ages 65 to 84
years and living in the Salisbury, Maryland, metropolitan area, were recruited for the study from a random
sample drawn from the Medicare database. Fifty-six
percent of the 65- to 74-year-old and 62% of the 75to 84-year-old white population in the database were
selected, with slight oversampling of people in the
higher age bracket because of a presumed lower response rate. An attempt was made to contact and recruit all 65- to 84-year-old blacks to ensure an adequate
sample for comparison of rates with those in the white
population. A 2-hour in-home interview was administered to each eligible participant. A 4- to 5-hour clinical examination followed, which included all of the
dry eye assessments: clinical testing, questionnaire administration, and lid-margin grading for blepharitis.
The plugging of Meibomian glands and collarettes
were graded from 0 to 3 in each subject without knowledge of the subject's dry eye questionnaire responses
or test (Schirmer's, rose bengal) results. Seventy-six
percent of the initial sample agreed to participate in
the study. A screening questionnaire administered to
half of those declining to participate revealed that they
were somewhat older and were less likely to report a
target symptom of dry eye (eyes feel gritty or sandy)
than were those who participated.
Statistical Analysis
Statistical analysis was designed to describe the severity
and frequency of dry eye symptoms in our study popu-

tive when reported to occur rarely or never. For definition 2, a response was defined as positive when the
subject reported a symptom to occur often or all the
time and as negative when reported to occur sometimes,
rarely, or never. Next, we derived two symptom summaries: counts (number of symptoms), and diagnostic
group profiles (symptom reporting patterns) resulting
from latent class analysis. Derived diagnostic groups
were examined for face validity and cross-tabulated
with clinical dry eye outcomes as a measure of external
validity. To assess internal consistency of symptom reporting, we used odds ratios to measure associations
between individual symptoms, Cronbach's alpha10 to
assess variability of counts, and latent class reporting
probabilities to assess variability in symptom patterns.
To assess stability of latent class model findings, symptom pattern summaries were derived separately and
examined for similarity (cross-validated) across randomly selected halves of the sample as well as age-,
race-, and sex-specific subgroups.
RESULTS
Demographic characteristics of the sample are listed
in Table 1, including age at the time of clinical examination, sex, race, Mini-Mental Status Examination
(MMSE)17 score, and years of education.

gritty burning

red

crust

shut

FIGURE 1. Distributions of symptoms by the frequency of responses (n = 2482).

Downloaded From: http://iovs.arvojournals.org/pdfaccess.ashx?url=/data/Journals/IOVS/933420/ on 04/24/2016

Investigative Ophthalmology & Visual Science, November 1997, Vol. 38, No. 12

2472

TABLE

2. Symptoms: Subgroup Severity Percentages


Gritty

Dry

Age (years)
65-69
70-74
75-79
80+

N/R

774
824
540
344

80.0
76.2
78.2
77.6

15.6
18.3
17.2
18.6

O/A N/R
4.4
5.5
4.6
3.8

68.2
67.7
69.3
65.7

Red

Burn
O/A N/R

26.4 5.4
26.3 6.0
25.6 5.2
23.8 10.5

73.9
75.4
79.3
76.2

S
23.8
22.1
17.4
19.5

O/A N/R
2.3
2.6
3.3
4.4

76.4
76.6
79.4
82.9

S
18.7
19.6
17.2
14.5

Crust
O/A N/R
4.9
3.9
3.3
2.6

84.9
82.8
85.2
81.1

S
12.3
13.1
10.9
14.2

Shut
O/A N/R
2.8
4.1
3.9
4.7

1,052 81.9 14.6 3.4 72.7 21.7


1,430 75.1 19.2 5.7 64.4 28.9
2

(2> = 17.5;
P< 0.001

Race
White
Balck

0/A

93.9 5.6 0.5


94.1 5.2 0.7
93.3 5.6 1.1
88.9 10.2 0.9
X2(6) = 13.4;
P = 0.04

X (6) = 12.8;
P = 0.05

Sex
Male
Female

5.6 77.0 20.8 2.2 76.8 18.2 4.4 87.1 10.3 2.7 92.2
6.7 75.0 21.5 3.4 78.9 17.6 3.6 81.3 14.2 4.6 93.8

6.8
5.6

1.0
0.6

X2(2) = 15.7;
P< 0.001

X2(2) = 19.5;

P < 0.001

1,832 77.7 16.9 5.4 72.0 22.3 5.7 77.2 19.4 3.2
650 78.8 18.3 2.9 56.3 35.9 7.8 72.0 26.0 2.0
X2(2) = 55.4;
v = 6.6;
(2> = 13.6;
P< 0.001
P = 0.001
P = 0.04

79.8 16.6 3.6 82.8 13.1 4.1 94.4


72.9 22.3 4.8 86.5 10.9 2.6 89.7
X2(2) = 13.4;
P= 0.001

4.9 0.7
9.4 0.9

r<2> = 17.2;
P = 0.001

N/R = never or rarely experiences; S = sometimes experienced; O/A = experienced often or all the time.

Distributions of the symptom frequencies among


study participants appear in Figure 1. The proportion
of participants who reported ever experiencing each
symptom ranged from more than 40% who reported
eyes that felt gritty or sandy to 12% who reported
eyelids stuck shut in the morning. The majority of
people who reported a symptom said that they experienced it sometimes, as seen by the preponderance of
each lightest bar in the overall bar heights. Thus, three
symptom frequency categories of (1) never or rarely,
(2) sometimes, and (3) often or all the time lose little
severity information relative to the original five-point
measure. Table 2 reports symptom distributions using
these categories and stratifying the respondents separately by age, sex, and race. Five percent to 8% more
women than men reported dry, gritty, and crusting
symptoms (chi-squared statistics = 17.5, 19.5, and
15.7, respectively; 2 df;P< 0.001), and four symptoms
were more prevalent among blacks than among
whites: eyes feel gritty, burn, appear red, stick shut
(range of differences = 4% to 16%, all chi-squared
statistics more than 13.4; 2 df; P < 0.001). Consistent
variation of response by age was not observed.
Figure 2 displays the distribution of the number
of symptoms reported. By definition 1, more than 40%
of participants reported no symptoms at all, 20% reported at least two, and less than 5% reported five or
more. Using the stricter definition 2, approximately
15% reported symptoms. Latent class modeling for
definition 2 symptoms (not shown) did not reveal
clear reporting patterns but rather grouped people

according to the number of symptoms (zero, one, and


two or more). In contrast, latent class modeling for
definition 1 symptoms distinguished people by number of symptoms and by varying reporting patterns at
higher numbers. This analysis is summarized in Table
3, with entries representing the estimated proportion
of people reporting each symptom in each dry eye
symptom pattern group. It suggests that four such
groups are sufficient to describe the data (likelihood
goodness of fit chi-squared = 45.83; df= 37; P= 0.15,
so that the null hypothesis that the data are consistent
with the latent class description is not rejected). Essentially, the first group reports no or isolated symptoms,
the second reports one or two symptoms, predominately eyes feel gritty, the third reports multiple symptoms including eyes burn and largely excludes crusting
on lashes, and the fourth reports multiple symptoms
including crusting. The prevalences of the more symptomatic groups 3 and 4 were estimated at 13% and
9%, respectively. Assessing face validity of the latent
class symptom patterns is complicated by clinical overlap in the symptoms of tear-deficient versus other
forms of dry eye and by the existence of people with
nonspecific or transitory symptoms. Keeping these
cautions in mind, the most symptomatic groups appear broadly consistent with the recognized causes of
dry eye of insufficient tear flow (class 3) and abnormal
tear content related to blepharitis (class 4).1 More
precise support of this observation is provided in Table 4, in which symptom pattern groups- are crosstabulated against indicators of artificial tear use and

Downloaded From: http://iovs.arvojournals.org/pdfaccess.ashx?url=/data/Journals/IOVS/933420/ on 04/24/2016

2473

Subjective Assessment of Dry Eye

OVERALL

g
LJJ
C

NUMBER OF SYMPTOMS > SOMETIMES

OVERALL

NUMBER OF SYMPTOMS FREQUENTLY OR ALWAYS


FIGURE 2. Distribution of symptom counts (n = 2482). The
top panel displays distributions for definition 1 (positive
symptom = sometimes or more often); the bottom panel
displays distributions for definition 2 (positive symptom =
often or all the time).

blepharitis evidence (a score of grade 2 or 3 on lid


margin), measured by a masked, independent clinician. Whereas artificial tear use is similarly elevated in
the two severely symptomatic groups 3 and 4, anatomic evidence of blepharitis is pronounced only in
group 4.
The associations among individual symptoms are
summarized in Table 5 as odds ratios. Here, higher
values represent stronger association, so that people
who reported a gritty sensation had more than four
times the odds of also reporting eyes that felt dry than
did people who did not report a gritty sensation
(definition 1). Odds ratios in the observed range of
2 to 8 indicate that item responses are moderately
consistent but also that redundancy has been acceptably minimized. At the stricter level of recognizing a
symptom as positive (definition 2), odds ratios are
much higher. This result indicates either that definition 2 gives a more reliable measure of dry eye than
definition 1 or that patterns in which severe symptoms

are reported are less differentiated than are patterns


in which less severe symptoms are reported. Regarding
internal consistency of counts, Cronbach's alpha was
estimated at 0.61 for definition 1; results for definition
2 were similar. This value predicts moderate variability
in test-re test reporting for both definitions. Together
with the latent class analysis reported earlier, this finding indicates that definition 2 differentiates symptom
patterns less than definition 1, not that it gives a more
reliable measure of dry eye. For latent class summary
of symptom reporting, the within-group proportions
endorsing symptoms measure reporting variability. Extremely low proportions endorsing symptoms in group
1 suggest that our symptoms assess dry eye widi specificity. However, in the other three groups, multiple
symptom proportions fall in a middle range (between
0.2 and 0.8). This is consistent with substantial variation in reporting among people with similar underlying disease.
Regarding stability of latent classes derived in different population subsamples, consistency of key defining features was observed. Specifically, four-class
models, applied separately to two random data halves
and across demographic subsamples, agreed in identifying an essentially asymptomatic group, a one- to twosymptom group predominately reporting gritty sensation, a multiple symptom group with low "crusting"
prevalence, and a multiple symptom group widi high
"crusting" prevalence. However, derived summaries
were not statistically equivalent across subsamples
(likelihood ratio goodness-of-fit criterion violated), reflecting ambiguity in assigning certain patterns to
classes. (For instance, the model predicts people who
report dry eyes in the absence of any other symptom
to be members of classes 1 and 2, with approximately
equal probability.) Moreover, diere was some evidence of additional symptom groups in women and
in 65- to 69-year-olds (best fitting four-class model inadequate by goodness-of-fit criterion). The observed
fluctuations in symptom patterns across demographic
TABLE 3.

Latent Class Analysis of Symptom


Data: Probability of Positive* Symptom
Report Within Classes
Class
1

0.34
0.70
0.08
0.21
0.15
0.05

0.46
0.68
0.98
0.50
0.01
0.10

0.55
0.67
0.63
0.57

Crust
Shut

0.07
0.04
0.08
0.11
0.07
0.02

Prevalence

0.57

0.21

0.13

0.09

Symptom

Dry
Grit
Burn
Red

1.0

0.37

* Positive report defined as frequency = at least "sometimes."

Downloaded From: http://iovs.arvojournals.org/pdfaccess.ashx?url=/data/Journals/IOVS/933420/ on 04/24/2016

2474

Investigative Ophthalmology & Visual Science, November 1997, Vol. 38, No. 12

TABLE 4. Association Between Derived Dry Eye Groups and Clinical Findings
Dry Eye Group

1 (n = 1499)

Clinical Finding

Artificial tear use?

88
5.9
48
0.032

Number

Blepharitis evidence?

2(n = 428)

Number

55

12.9
11

0.026

3 (n = 343)

4 (n = 209)

70
20.4
11
0.032

46
22.0
18
0.087

df

Value

102.4

<0.001

17.7

<0.001

df= degrees of freedom.

subsamples are consistent with our finding of systematic differences in reporting individual symptoms (Table 2). This should be considered in age-, sex-, or racespecific analyses.
In summary, dry eye symptoms were common and
were reported throughout a range of severity in our
population-based setting. Some heterogeneity in reporting among clinically similar people and across demographic subgroups was indicated. However, two distinct and qualitatively stable symptom patterns were
observed at substantial severity. Thus, our methodology added more information than does simple counts
and provided symptom summaries that are promising
for use in further research. Derived patterns had good
biologic and clinical validity, suggesting that our instrument is useful for assessing dry eye.
DISCUSSION
For screening and in large population-based studies,
simple approaches for assessing disease are often mandatory. Because treatment of dry eye is primarily designed to relieve patient symptoms, it is essential to
incorporate self-reported symptoms as outcomes in
dry eye epidemiologic research and randomized trials
in a clinically usable way. We report on results of a
self-reported assessment of dry eye and describe the
development and application of methods to summarize dry eye symptoms using latent class analysis. The
questionnaire used detected substantial dry eye symptomatology in a population-based setting. Moreover,
dry eye profiles derived from latent class analysis were

TABLE 5. Internal Consistency, Individual


Symptoms: Pairwise Odds Ratios*

Dry
Grit
Burn
Red

Crust
Shut

Dry

Grit

Burn

Red

Crust

Shut

4.4

3.3
4.1

2.5
2.8
4.4

7.5
5.2

2.8
2.8
2.2
2.5

2.6
3.3
2.6
2.8
7.5

16.8
19.4

15.8

7.1
7.2
6.3

6.7
6.7
8.6

11.9
12.7
2.5

22.3

* Above diagonal: positive report = frequency at least sometimes;


below diagonal: positive report = frequency at least often.

consistent with recognized dry eye clinical categories,


as was evident descriptively and through logical patterns of association with clinical variables. This suggests that in our population-based setting, symptom
elicitation provides useful information about dry eye.
We proposed to investigate the coherence and
precision with which symptoms are reported, using
our instrument. Results of our analyses supported
broad reproducibility of the number of symptoms reported as well as the within-subject symptom report
pattern but also suggested variability in reporting
among clinically similar people. This is consistent with
the recognized variation in dry eye symptomatology
among people with similar clinical symptoms.5'6 Many
factors could contribute to such variability, including
individual differences in perceiving symptom severity
and external influences (weather or time of day). Such
factors might also lead to significant within-subject
variability in reporting. Indeed, ad hoc retest data collected in the homes of 97 Salisbury Eye Evaluation
participants18 yielded weighted kappa statistics19 of 0.6
to 0.65 for definition 1 symptom reporting. These reliability estimates are limited by 1- to 3-month time
intervals between assessments, systematic differences
between clinic and home data collection, and inclusion of few people with severe dry eye in the retest.
Taking these in combination with our reported findings, however, we recommend that strategies to increase precision in reporting be pursued. We further
recommend that item response summary that recognizes syndrome heterogeneity is critical for quantitating dry eye status in research settings. The latent class
analysis approach described in this report provides
one such method for summarizing responses.
In Table 2 and in validation analysis, we found
some variability in reporting individual symptoms
across demographic subgroups. It is unclear whether
this reflects differences in underlying disease or systematic differences in reporting equivalent disease.
Because of this, we do not recommend diat formal
scores be derived directly from our latent class model
(Table 3) for application to other data sets. Instead,
we propose that it will be useful to distinguish counts
of two or fewer symptoms by whether gritty sensation
is reported and counts of three or more symptoms by

Downloaded From: http://iovs.arvojournals.org/pdfaccess.ashx?url=/data/Journals/IOVS/933420/ on 04/24/2016

2475

Subjective Assessment of Dry Eye

whether crusting is reported. In analyses in which such


variables as age, race, or sex are important, we intend
as one analytic technique to derive separate pattern
summaries within demographic subgroups. Because
demographic differences in measuring dry eye have
substantial implications for research and patient care,
more investigation is needed.
We do not suggest that a questionnaire-based approach obviates the need for measurement of tear
function. However, symptom reporting clearly provides different information,6 and in our own analyses
(not shown) counts of symptoms as well as latent class
groups were only weakly associated with Schirmer's
test results and rose bengal scores. Neither do we recommend that our population-based findings be uniformly generalized to samples of severely diseased patients. However, several aspects of this work may prove
useful in a range of settings. The instrument validly
captured a range of symptoms while minimizing redundancy. The latent class method is particularly appropriate for research on dry eye when definitive diagnostic definitions do not exist. Isolating different
symptom subgroups, as we have done in our population-based sample, may be useful for application to
clinical trials of dry eye treatments. For example, a
potential treatment may offer benefit to one set of
patients with a specific class of symptoms but not to
another, a finding that might have been difficult to
detect with traditional symptom-scoring techniques.
Finally, we hope that our analytic approach, which has
yielded specific symptom subclasses, will help us and
others to determine risk factors for various forms of
dry eye.
Key Words
aging, diagnosis, dry eye, epidemiology, latent class analysis
References
1. Lemp MA, Chairman. Report of the National Eye Institute/Industry Workshop on Clinical Trials in Dry
Eyes. ClJiOJ. 1995;21:221-232.
2. Farris RL, GilbardJP, Stuchell RN, Mandel ID. Diagnostic tests in keratoconjunctivitis sicca. CLAO J.
1983;9:23-28.
3. Goren MD, Goren SB. Diagnostic tests in patients with
symptoms of keratoconjunctivitis sicca. Am J Ophthalmol. 1988; 106:570-574.
4. Lucca JA, Nunez JN, Farris RL. A comparison of diagnostic tests for keratoconjunctivitis sicca: Lactoplate,
Schirmer, and tear osmolality. CLAO J. 1990; 16:109112.
5. Nelson JD. Diagnosis of keratoconjunctivitis sicca. International Ophthalmology Clinics, Dry Eye. Vol 34. Boston: Little, Brown; 1994;34:37-56.

6. Schein OD, Tielsch JM, Muiioz B, Bandeen-Roche


K, West S. Relationship between signs and symptoms
of dry eye: A population-based perspective. Ophthalmology, (in press).
7. McMonnies CW, Ho A. Patient history in screening
for dry eye conditions. JAm Optom Assoc. 1987;58:296301.
8. Sargent C, Ludlow M, McCredie M, Brooks PM, Webb
J. Assessment of a patient-completed questionnaire for
keratoconjunctivitis sicca. MedJ Aust. 1988; 148:395398.
9. Anderson JA, Whaley K, Williamson J, Buchanan WW.
A statistical aid to the diagnosis of keratoconjunctivitis
sicca. QJMed, Neio Series. 1972;41:175-189.
10. Holly FJ, Lemp MA. Tear physiology and dry eyes.
Surv Ophthalmol. 1977; 22:69-87.
11. Goodman LA. Exploratory latent structure analysis using both identifiable and unidentifiable models. Biometrika. 1974;61:215-231.
12. Lazarsfeld PF, Henry NW. Latent Structure Analysis.
New York: Houghton-Mifflin; 1968.
13. McCutcheon, AL. Latent Class Analysis. Newbury Park,
CA: Sage Publications; 1987.
14. Manton KG, Stallard E, Liu K. Forecasts of active life
expectancy: Policy and fiscal implications. / Geiontol.
1993; 48:11-26.
15. West SK, Muiioz B, Rubin GS, et al. Function and
visual impairment in a population-based study of older
adults. The SEE Project. Invest Ophthalmol Vis Sci.
1997; 38:72-82.
16. Cronbach LJ. Coefficient alpha and the internal structure of tests. Psychometrika. 1951; 16:63-70.
17. Folstein MF, Folstein SE, McHugh PR. Mini-mental
state: A practical method for grading the cognitive
state of patients for the clinician. / Psych Res.
1975; 12:189-198.
18. West SK, Rubin GS, Munoz B, Abraham D, Fried LP,
and SEE Project Team. Assessing functional status:
Correlation between performance on tasks conducted
in a clinic setting and performance on the same task
conducted at home. / Geiontol 1997;52(A):M209M217.
19. Fleiss JL. Statistical Methods for Rates and Proportions.
2nd ed. New York: John Wiley & Sons; 1989:220.

APPENDIX 1
Dry Eye Questionnaire*
1. Do your eyes ever feel dry?
2. Do you ever feel a gritty or sandy sensation in
your eye?
3. Do your eyes ever have a burning sensation?
4. Are your eyes ever red?
5. Do you notice much crusting on your lashes?
6. Do your eyes ever get stuck shut in the morning?
* Allowable responses: never, rarely, sometimes,
often or all the time

Downloaded From: http://iovs.arvojournals.org/pdfaccess.ashx?url=/data/Journals/IOVS/933420/ on 04/24/2016

Вам также может понравиться