Cataract Classification Using Serial Examinations in the Age-Related Eye Disease Study: Age-Related Eye Disease Study Report No. 24

Cataract Classification Using Serial Examinations in the Age-Related Eye Disease Study: Age-Related Eye Disease Study Report No. 24

Cataract Classification Using Serial Examinations in the Age-Related Eye Disease Study: Age-Related Eye Disease Study Report No. 24 ROBERT D. SPERDUTO...

170KB Sizes 0 Downloads 33 Views

Cataract Classification Using Serial Examinations in the Age-Related Eye Disease Study: Age-Related Eye Disease Study Report No. 24 ROBERT D. SPERDUTO, TRACI E. CLEMONS, ANNE S. LINDBLAD, AND FREDERICK L. FERRIS III, ON BEHALF OF THE AGE-RELATED EYE DISEASE STUDY RESEARCH GROUP ● PURPOSE:

To describe use of serial lens examinations to assign cataract phenotype in the Age-Related Eye Disease Study (AREDS). ● DESIGN: Cohort study. ● METHODS: Lens photographs were graded annually using the AREDS system for classifying cataracts. Nuclear grades (0.9 to 6.1) were assigned using standard photographs. Percentage of pupillary involvement was used to assign cortical and posterior subcapsular grades. Cutpoints were established for the presence or absence of each type of opacity (absent <4.0 for nuclear, <10% for cortical, and <5% of central 5 mm for posterior subcapsular). An algorithm weighted for grades at the last three examinations was used to assign cataract phenotype. Separately, cataract phenotype was assigned as grade predicted at final visit by linear regression of serial grades. Results from the two approaches were compared and final person phenotypes were established. These person cataract phenotype assignments were compared with phenotype assignments based on the last serial grade alone. ● RESULTS: Four thousand six hundred and twenty-eight AREDS participants aged 55 to 80 years at baseline had median follow-up of 10.6 years. Person phenotype assignments agreed for the two approaches in 4,557 (98.5%) participants after some algorithmic adjudication. Phenotypes were no cataract (n ⴝ 1,418), nuclear (n ⴝ 1,287), cortical (n ⴝ 1,396), posterior subcapsular (n ⴝ 541), cataract surgery and no specific opacity type (n ⴝ 335), and questionable (n ⴝ 426). Phenotype assignments based on serial grades and on last examination alone were in good agreement. ● CONCLUSIONS: Serial lens photographs obtained over a 10-year period were used to provide a robust assignment of cataract phenotype. Well-characterized cataract phenotypes are of importance as genetic studies of the AREDS cohort are considered. (Am J Ophthalmol

Accepted for publication Oct 24, 2007. From the National Eye Institute, National Institutes of Health, Bethesda, Maryland (R.D.S., F.L.F.); and The EMMES Corporation, Rockville, Maryland (T.E.C., A.S.L.). Inquiries to Traci E. Clemons, The EMMES Corporation, 401 North Washington Street, Suite 700, Rockville, MD 20850; e-mail: tclemons@ emmes.com

504

©

2008 BY

2008;145:504 –508. © 2008 by Elsevier Inc. All rights reserved.)

M

ANY CLINICAL STUDIES, INCLUDING GENETIC

studies of age-related cataract,1– 4 establish cataract phenotype based on the results of a single examination. Even when established clinical or photographic grading systems5–9 are used, misclassification can be of concern because of the many sources or error that can affect the largely subjective cataract grading process. Differences in phenotype assignment can occur because of intraobserver or interobserver grading variability, edge effects when cutpoints are used to establish the presence or absence of cataract, inadequate examiner training, and inherent differences between grading systems. Many clinical studies can tolerate small amounts of misclassification, but in some studies, such as genetic studies and studies with few participants, even relatively little misclassification of phenotype can lead to substantial misinterpretation of results. The Age-Related Eye Disease Study (AREDS),10 an 11-center prospective study of the clinical course of agerelated cataract and macular degeneration, is unique in that lens and macular photographic examinations were conducted at baseline and annually starting in year 2. Examinations of the large AREDS cohort continued for more than 10 years. Participant deoxyribonucleic acid (DNA) and data on environmental exposures were collected. It is important that cataract phenotypes be well characterized as genetic studies are considered. The serial examinations in AREDS provide an opportunity for robust assignment of cataract phenotype. In this study, we describe how the serial lens examinations were used to make phenotype assignments and how the assignments based on the serial grades compare with assignments based on the grade at the last visit alone.

METHODS THE AREDS IS A STUDY OF THE NATURAL HISTORY OF

age-related cataract and macular degeneration.10 It included a randomized clinical trial that evaluated the effect of high doses of selected nutritional supplements on the

ELSEVIER INC. ALL

RIGHTS RESERVED.

0002-9394/08/$34.00 doi:10.1016/j.ajo.2007.10.024

FIGURE 1. Algorithm demonstrating the assignment of cataract phenotypes in eyes using serial eye grades in the AgeRelated Eye Disease Study (AREDS).

incidence and progression of the two conditions.11,12 Details of the study design have been published elsewhere10 and are described briefly herein. A total of 4,757 persons, 55 to 80 years of age at the time of enrollment, were entered into the study at 11 eye care centers from 1992 through 1998. Ocular eligibility requirements were determined largely by the age-related macular degeneration component of the study. Except for a requirement that the media be sufficiently clear to allow photographic documentation of macular status, lens opacity status was not considered specifically in selecting participants. The large sample size and the expected high prevalence of lens opacities in the targeted age group made it likely that a diverse array of age-related lens opacities of mild to moderate severity would be present in the cohort at the conclusion of the study. The present analysis of 4,628 participants excludes 129 persons with bilateral aphakia or pseudophakia. Specially modified Topcon slit-lamp cameras (Topcon Corp, Tokyo, Japan) and Neitz retroillumination cameras (Neitz Instruments Co, Ltd, Tokyo, Japan) were used by certified photographers to obtain color photographs of the lenses of participants at baseline and annually thereafter, starting in year 2. The lens photographs were evaluated for quality and were graded for cataract at a reading center by trained and certified examiners. A quality control program included masked replicate gradings of samples of photographs to assess contemporaneous and temporal grading reliability. The Age-Related Eye Disease System for Classifying Cataracts,9 an extension of the Wisconsin System for Classifying Cataracts from Photographs,7 was used to assess the presence and severity of nuclear (N), cortical (C), and posterior subcapsular (PSC) lens opacities. The extent of cortical and PSC opacities was graded by estimating the area of lens involvement in sectors of a grid overlay on the retroillumination photographs. The individual subfield percentages were combined to estimate an overall percentage of involvement within the central 5 mm of the lens and, for cortical opacities, within the full visible lens. Nuclear opacity grades ranged from 0.9 to 6.1 VOL. 145, NO. 3

CATARACT PHENOTYPE

FIGURE 2. Algorithm demonstrating the assignment of cataract phenotype in eyes using linear regression of serial grades in the AREDS.

TABLE 1. Adjudication Rules for Resolving Some Differences between Algorithm and Regression Cataract Phenotype Assignments of Eyes in the Age-Related Eye Disease Study One Method

Other Method

Questionable Present

Absent Cataract surgery, no typespecific opacity Present *Nuclear, ⱖ 3.5 *Cortical, ⱖ 5% *PSC, ⱖ 2.5%

Missing data Present Present Present

Resolution

Questionable Present Present Present Present Present

PSC ⫽ posterior subcapsular. *Two of three last photographs using algorithm or the predicted value from linear regression, for the same type as present.

using cutpoints set by a series of standard photographs with increasingly severe nuclear sclerosis. ● ASSIGNMENT OF PHENOTYPE USING THE SERIAL GRADES: First, cutpoints for the presence or absence of

each specific type of cataract were established by AREDS ophthalmologists who came to a consensus about the amount of change that was likely to be of clinical significance after examining a large number of lens photographs. The cutpoints used to define cataract are: absent, ⬍4.0 for nuclear, ⬍10% of the entire visible lens for cortical, and ⬍5% of the central 5 mm for PSC opacities. Eye-specific phenotype then was defined using two approaches. The first applied an algorithm to the serial reading center grades separately for each opacity type (Figure 1). Special weight in diagnosing the presence of cataract by the algorithm was given to consistency of present grades at the last three visits and to cataract

FROM

SERIAL EXAMINATIONS

505

RESULTS

TABLE 2. Cataract Phenotype Assignments Using the Serial Grades (Algorithm/Regression Assignment) and the Last Visit Grade in the Age-Related Eye Disease Study Algorithm/ Regression Method

Last Photograph

Person Phenotype

No.

%

No.

%

Absent Nuclear alone Cortical alone PSC alone Nuclear/cortical Nuclear/PSC Cortical/PSC Nuclear/cortical/PSC Cataract surgery/no specific type Questionable

1418 773 870 139 265 141 153 108 335 426

30.6 16.7 18.8 3.0 5.7 3.1 3.3 2.3 7.2 9.2

1646 836 864 183 262 171 166 107 343 50

35.6 18.1 18.7 4.0 5.7 3.7 3.6 2.3 7.4 1.1

AGREEMENT BETWEEN SYSTEMS THAT USED THE SERIAL

grades was exact in 3,530 (76.3%) participants. Of those with exact agreement, cataract was present in 2,221 (62.9%) and no cataract was present in 1,309 (37.1%). An additional 1,098 (23.7%) required adjudication. After adjudication, phenotype assignments remained discrepant in 71 participants (1.5%). Final person-specific cataract phenotypes after adjudication were as follows: no cataract ⫽ 1,418 (30.6%), N cataract ⫽ 1,287 (27.8%), C cataract ⫽ 1,396 (30.2%), PSC cataract ⫽ 541 (11.7%), cataract surgery but no specific cataract type ⫽ 335 (7.2%), and questionable ⫽ 426 (9.2%). Pure nuclear cataracts were found in 773 (16.7%), pure cortical cataracts were found in 870 (18.8%) and pure PSC cataracts were found in 139 (3.0%) participants (Table 2). We also computed a person phenotype using the last available photograph and the predefined cutpoints for type-specific cataract. Table 2 shows the person phenotypes by cataract type using the method defined above compared with using the last available photograph. Using the last photograph to classify phenotype resulted in only 50 cases of questionable cataract. There was a five percentage–point increase in the number of participants classified as cataract absent in the use-the-last-photograph approach compared with the approach using serial photographs. Table 3 shows the sensitivity and specificity of using the results of the last examination alone to assign cataract phenotype when the phenotype assignments based on the serial grades were used as the gold standard. Both sensitivity and specificity were high, indicating relatively few false positives and false negatives, for any N, any C, any PSC, absence of cataract, and cataract surgery but no specific type assignment.

PSC ⫽ posterior subcapsular.

surgery. Type-specific assignments were combined to assign eye grades (cataract absent; cataract present, by type; cataract surgery, but no specific cataract type; or questionable). The second approach applied a linear regression model to the serial grades to predict a grade for an eye at the last visit. A linear regression model was developed for each eye for each opacity type, estimating the slope of a linear fit to the serial grades. The slope was used to predict the opacity grade at the visit of the last available photograph for each eye. The predicted grade then was used to define presence or absence of opacity type using the cutpoints defined above. A summary of the regression approach is provided in Figure 2. Eye-specific phenotype assignments for the two approaches were compared and some, but not all, differences were adjudicated systematically (Table 1). Eye-specific phenotype assignments were made for each approach that took into account changes in grade allowed by the adjudication rules. For each approach, each participant then was assigned a cataract phenotype using the eye-specific phenotype assignments of both eyes. Person assignments were made in priority order as follows: 1) cataract absent, no cataract in either eye; 2) cataract present, cataract in at least one eye (participants then classified as C alone; N alone; PSC alone; C and N; C and PSC; N and PSC; or C, N, and PSC); 3) history of cataract surgery but no specific type; and 4) questionable, everyone else. Person assignments for the two approaches were compared and final person phenotype assignments were established. Persons with discrepant assignments for the two approaches were classified as questionable. Using the final phenotype assignment based on the serial grades as the gold standard, we determined the sensitivity and specificity of the cataract grade assignments based on the last visit alone. 506

AMERICAN JOURNAL

DISCUSSION FEW CLINICAL STUDIES HAVE THE OPPORTUNITY TO USE

serial lens examinations over a prolonged period to establish cataract phenotype. Lens photographic examinations were conducted in AREDS at baseline and then annually starting in year 2 for a median follow-up of 10.6 years (range, zero to 12.9 years). Lens photographs were obtained in a standardized fashion by certified photographers and were graded at a reading center by specially trained and certified observers. A quality control program included masked replicate gradings of samples of the photographs to assess contemporaneous and temporal grading reliability. An algorithm that gave special weight to the cataract gradings at the last three visits and a linear regression model that predicted the cataract grades at the last visit were in exact agreement in assigning cataract phenotype in 76% of participants. After adjudication, phenotype assignOF

OPHTHALMOLOGY

MARCH 2008

TABLE 3. Sensitivity and Specificity of Cataract Phenotype Assignment Using the Last Photograph When Cataract Phenotype by Serial Grades Is the Standard in the Age-Related Eye Disease Study Person Phenotype

Agreement

Absent Any nuclear Any cortical Any PSC Cataract surgery/no specific type

94% 93% 97% 98% 99%

Sensitivity (95% CI)

97.7 90.3 93.3 95.2 91.9

(96.9 (88.7 (91.9 (93.4 (89.0

to to to to to

Specificity (95% CI)

98.5) 91.9) 94.6) 97.0) 94.9)

91.9 93.6 97.0 97.3 99.2

(90.9 (92.8 (96.4 (96.8 (98.9

to to to to to

92.8) 94.4) 97.6) 97.8) 99.5)

CI ⫽ confidence interval; PSC ⫽ posterior subcapsular.

ment was discrepant in 1.5% of participants. Adjudication resolved the edge effect that can occur when a cutpoint is used to establish the presence or absence in 29% of participants who required adjudication. Typically, an edge effect occurred when the grade for one of the systems was at or above the cutpoint and the other was just below the cutpoint. In such cases, an adjudicated assignment of present was made. Adjudication also resolved in favor of the presence of the specific type(s) of opacity when one system had a present grade for a specific type(s) and the other had an assignment of cataract surgery of no specific type. In many clinical studies, assignment of cataract phenotype is based on the results of a single subjective lens examination. Both standardized grading systems that rely on a slit-lamp assessment of the lens and those that rely on the grading of lens photographs to establish cataract phenotype can be affected by many potential sources of error. These include inadequate examiner training, intraexaminer or interexaminer grading variability, temporal drift in grading, equipment differences, edge effects, and, in the case of photographic systems, poor quality photographs and changes in film and film processing. Serial lens examinations by a corps of trained and certified photographers and cataract graders and phenotype assignments that emphasize consistency in the gradings over time or trends in the grades over a long period may be expected to lessen the potential for misclassification of phenotype. In the AREDS, when we compared phenotype assignments based on only the last lens examination with those derived from the grades from the serial examinations, we found a high degree of concordance for the two approaches. Using the serial grade phenotypes as the gold standard, the sensitivity and specificity, respectively, of using the last examination grade to assign phenotype was high for any N (90%; 94%), any C (93%; 97%), any PSC (95%; 97%), no cataract (98%; 92%), and cataract surgery but no specific type (92%; 99%). The high levels of agreement are at least in part because the last visit grade is an important component of the algorithm and regression phenotype assignment. The high degree of concordance between phenotype assignments based on the serial grades and those made using the last examination alone may reflect the fact that both VOL. 145, NO. 3

CATARACT PHENOTYPE

TABLE 4. Quartiles of Opacity Grade in the Worse Eye at the Last Visit with a Photograph in the Age-Related Eye Disease Study Percentile Opacity Type

Nuclear Cortical PSC

25th

50th

75th

2.4 0.3% 0%

3.2 3.1% 0%

4.0 13.6% 0.5%

PSC ⫽ posterior subcapsular.

approaches used the same corps of certified graders and shared the same rigorous quality control procedures for photography and grading. Also, when a single cutpoint is used to divide subjects into present or absent groups, there is probably less potential for misclassifying subjects whose true grade is not close to a cutpoint. Table 4 shows 25th, 50th, and 75th percentiles for the three opacity types for the worse eye of participants at the last visit at which a grade was available. The table suggests that most grades were remote from cutpoints, making present-or-absent agreement between different grading approaches more likely. Nevertheless, the discrepancies seen here for absent and questionable phenotypes suggest caution in relying on a single (last) grade. This observation is particularly important for analyses where misclassification error must be minimized, as is the case for genetic studies. The AREDS was a large natural history study of agerelated cataract and macular degeneration designed, in part, to search for clues about the causes of the two diseases and possible strategies for intervention. The cohort includes a large pool of participants with a broad range of well-characterized cataract phenotypes (no cataract ⫽ 1,418, N cataract ⫽ 1,287, C cataract ⫽ 1,396, PSC ⫽ 541, cataract surgery and no specific opacity type ⫽ 335, and questionable ⫽ 426). Data were collected on a wide range of possible risk factors for cataract, including educational status, smoking, diabetes, sunlight exposure, body mass index, drug use, estrogen replacement therapy, and dietary intake of various nutrients. Participant

FROM

SERIAL EXAMINATIONS

507

DNA also was collected and stored. The availability of DNA and data on environmental exposures that may be associated with cataract development for a large cohort with well-characterized cataract phenotypes offers some

unique opportunities to study possible genetic influences on age-related cataract development and the possible interplay between genetic factors and environmental exposures.

THIS STUDY WAS SUPPORTED BY CONTRACTS FROM THE NATIONAL EYE INSTITUTE, NATIONAL INSTITUTES OF HEALTH, Department of Health and Human Services, Bethesda, Maryland. The authors indicate no financial conflict of interest. Involved in conception, design, and conduct of study (R.D.S., A.S.L., F.L.F.); analysis and interpretation of the data (T.E.C., R.D.S.); preparation and critical revision of the manuscript (R.D.S., T.E.C., A.S.L., F.L.F.); and review and approval of the manuscript (R.D.S., T.E.C., A.S.L., F.L.F.). Institutional review board approval was obtained by each clinic for the collection of AREDS data, and informed consent was obtained from all participants before enrollment. THE AGE-RELATED EYE DISEASE STUDY RESEARCH GROUP A complete list of the members of the Age-Related Eye Disease Study (AREDS) Research Group appears in Arch Ophthalmol 2004;122:716 –726.

8. Sparrow JM, Bron AJ, Brown MAP, Ayliffe WM, Hill AR. The Oxford clinical cataract classification and grading system. Int Ophthalmol 1986;9:207–225. 9. The Age-Related Eye Disease Study Research Group. The Age-Related Eye Disease Study (AREDS) system for classifying cataracts from photographs: AREDS Report No. 4. Am J Ophthalmol 2001;131:167–175. 10. The Age-Related Eye Disease Study Research Group. The Age-Related Eye Disease Study (AREDS): design implications. AREDS Report No. 1. Control Clin Trials 1999;20: 573– 600. 11. Age-Related Eye Disease Study Research Group. A randomized, placebo-controlled, clinical trial of high-dose supplementation with vitamins C and E, beta-carotene, and zinc for age-related macular degeneration and vision loss. AREDS Report No. 8. Arch Ophthalmol 2001;119:1417–1436. 12. Age-Related Eye Disease Study Research Group. A randomized, placebo-controlled, clinical trial of high-dose supplementation with vitamins C and E, beta-carotene, and zinc for age-related cataract and vision loss. AREDS Report No. 9. Arch Ophthalmol 2001;119:1439 –1452.

REFERENCES 1. The Framingham Offspring Eye Study Group. Familial aggregation of lens opacities: the Framingham Eye Study and the Framingham Offspring Eye Study. Am J Epidemiol 1994;140:555–564. 2. Alberti G, Oguni M, Podgor M, et al. Glutathione Stransferase M1 genotype and age-related cataracts. Invest Ophthalmol Vis Sci 1996;37:1167–1173. 3. Heiba IM, Elston RC, Klein BEK, Klein R. Evidence for a major gene for cortical cataract. Invest Ophthalmol Vis Sci 1995;36:227–235. 4. Heiba IM, Elston RC, Klein BEK, Klein R. Genetic etiology of nuclear cataract: evidence for a major gene. J Med Genet 1993;47:1208 –1214. 5. Chylack L, Wolfe J, Singer D, et al. The Lens Opacities Classification System III. Arch Ophthalmol 1993;111:831– 836. 6. Taylor HR, West SK. The clinical grading of lens opacities. Aust N Z J Ophthalmol 1989;17:81– 86. 7. Klein BEK, Klein R, Linton KLP, Magli YL, Neider M. Assessment of cataracts from photographs in the Beaver Dam Eye Study. Ophthalmology 1990;97:1428 –1433.

508

AMERICAN JOURNAL

OF

OPHTHALMOLOGY

MARCH 2008