6
Original Articles The flashlight test and van Herick's test are poor predictors for occludable angles Ravi Thomas, MD* Thomas George, DOMS* Andrew Braganza, MS* Jayaprakash Muliyil, DrPHt Abstract Objective: To determine the reliabilityand validity of the flashlight test and van Herick's test in detecting occludable anterior chamber angles. Methods:The flashlight test, van Herick's test and gonioscopy were performedindependently by two observers on 96 consecutive new patients in our outpatient clinic. lnterobserver agreement was determined using the weighted Kappa statistic. Using the glaucoma specialist's assessment of occludability of the angle (assessed by gonioscopy) as a gold standard, the sensitivities and specificities of the two tests were calculated. Results: All three tests showed good agreement (Kappa more than 0.75). The sensitivity and specificity of the flashlight test were 45.5% and 82.7% respectively. For the van Herick's test they were 61.9% and 89.3%. Conclusion: The flashlight test and van Herick's test are of limited use as screening tests for occludable angles. Key words: Flashlight test, gonioscopy, occludability, predictive values, reliability, validity, van Herick's test. Angle-closure glaucoma can be prevented by prophylactic peripheral iridectomy. I Biometric parameters such as lens thickness, axial length and anterior chamber d e~th~-~ may be of poten- tial use in predicting occludable angles. Most of the field-based data on the prevalence of angle closure glaucoma are derived from studies that have used the flashlight and/or van Herick's tests (at least as the first ~ t a g e ) . ~ - ~ The prevalence of angle closure varies in different populations. A screening program which uses a multi-stage test- ing strategy has an overall sensitivity which is the product of the sensitivities of the individual tests used. There is little information on how these tests compare with gonioscopy. Gonios- copy is the usual clinical test used to assess the angle. We determined the reliability ~~ ~ * Schell Eye Hospital, Department of Ophthalmology, Christian Medical College, Vellore, India. Department of Community Health & Epidemiology, Christian Medical College, Vellore, India. Reprints: Dr Ravi Thomas, Schell Eye Hospital, Arni Road, Vellore 632 001, India. The flashlight test and van Herick's test are poor predictors for occludable angles 251

The flashlight test and van Herick's test are poor predictors for occludable angles

Embed Size (px)

Citation preview

Page 1: The flashlight test and van Herick's test are poor predictors for occludable angles

Original Articles

The flashlight test and van Herick's test are poor predictors for occludable angles

Ravi Thomas, MD* Thomas George, DOMS* Andrew Braganza, MS* Jayaprakash Muliyil, DrPHt

Abstract

Objective: To determine the reliability and validity of the flashlight test and van Herick's test in detecting occludable anterior chamber angles.

Methods:The flashlight test, van Herick's test and gonioscopy were performed independently by two observers on 96 consecutive new patients in our outpatient clinic. lnterobserver agreement was determined using the weighted Kappa statistic. Using the glaucoma specialist's assessment of occludability of the angle (assessed by gonioscopy) as a gold standard, the sensitivities and specificities of the two tests were calculated.

Results: All three tests showed good agreement (Kappa more than 0.75). The sensitivity and specificity of the flashlight test were 45.5% and 82.7% respectively. For the van Herick's test they were 61.9% and 89.3%.

Conclusion: The flashlight test and van Herick's test are of limited use as screening tests for occludable angles.

Key words: Flashlight test, gonioscopy, occludability, predictive values, reliability, validity, van Herick's test.

Angle-closure glaucoma can be prevented by prophylactic peripheral iridectomy. I Biometric parameters such as lens thickness, axial length and anterior chamber d e ~ t h ~ - ~ may be of poten- tial use in predicting occludable angles. Most of the field-based data on the prevalence of angle closure glaucoma are derived from studies that have used the flashlight and/or van Herick's tests (at least as the first ~ t a g e ) . ~ - ~ The prevalence of angle closure varies in different populations. A screening program which uses a multi-stage test- ing strategy has an overall sensitivity which is the product of the sensitivities of the individual tests used. There is little information on how these tests compare with gonioscopy. Gonios- copy is the usual clinical test used to assess the angle. We determined the reliability

~~ ~

* Schell Eye Hospital, Department of Ophthalmology, Christian Medical College, Vellore, India. Department of Community Health & Epidemiology, Christian Medical College, Vellore, India.

Reprints: Dr Ravi Thomas, Schell Eye Hospital, Arni Road, Vellore 632 001, India.

The flashlight test and van Herick's test are poor predictors for occludable angles 251

Page 2: The flashlight test and van Herick's test are poor predictors for occludable angles

(interobserver variation) and validity of the flash- light and van Herick’s tests using gonioscopy as a gold standard to determine occludable ante- rior chamber angles. Occludability, however, is not synonymous with angle closure glaucoma. There is limited information regarding the number of gonioscopically occludable angles that will develop angle closure, but it seems to be a rare event.y In the study by Willensky et al., 17 out of 139 eyes classified as occludable (using a Zeiss 4-mirror gonioscope) developed closure over a period of five years. The numbers quoted in the same study for the Koeppe lens were 15 out of 98. More than half of these patients developed ‘non-acute’ closure. The authors also noted that all angle closures occurred in eyes that were initially graded as occludable.2

Patients and methods

One hundred new patients consecutively attend- ing our outpatient clinic were studied. The eye to be examined was randomly selected on the toss of a coin by the examining ophthalmologist (who subsequently had nothing further to do with the study). Four patients were excluded as they had acute conditions -- phacolytic glaucoma in one, phacomorphic glaucoma in two and a cor- neal ulcer in one. The remaining 96 patients were first seen by a second-year resident in ophthal- mology and the following parameters were recorded in the order described below. . Flashlight test

van Herick’s test . Gonioscopy grade (Goldmann single

mirror) . Occludability (based on gonioscopy) . Presence of peripheral anterior synechiae

Immediately after this, the patients were sent to the glaucoma specialist with no information other than the eye to be examined. In addition to the tests mentioned above, he performed an indentation gonioscopy with a Sussmann four-mirror lens.

For the flashlight test, a flashlight beam was directed parallel to the iris from the temporal side as originally reported by Vargas et aLS The crescentic iris shadow thus formed was graded according to the area between the limbus and the pupillary edge that it occupied. For the purpose of this study four grades were thus defined as: more than half: half to one-third; minimal; and no shadow.

The van Herick’s test was performed as described by William van Herick.’ If peripheral anterior chamber depth was equal to or greater than the corneal thickness it was recorded as grade 4; half corneal thickness was grade 3; quarter thickness of cornea was noted as grade 2 and less than a quarter as grade 1.

The preferred grading used for gonioscopy in our institution is based on structures actually visualised (Table 1). Grades 3 and less are con- sidered narrow. Both the observers in this study were familiar with the grading system. Gonios- copy was performed on the Haag Streit slit lamp with the minimum intensity of illumination com- patible with good visualisation: the slit beam was shortened so that it did not fall on the pupil. In each case the observers were required to make a ‘forced’ decision as to whether the angle was occludable or not. This was based (in addition to the grading) on the type of entry into the angle (narrow/wide). Examination for the pres- ence of peripheral anterior synechiae (PAS) sug- gestive of angle closure was performed by manipulating (applying slight pressure on the lens as the patient looked toward the mirror) the Goldmann single mirror goniolens. For this pur- pose the glaucoma specialist first used the

Table 1. Gonioscopy grading used

0 1 Dipping of the beam 2

3 4 5 Scleral spur visible 6 Ciliary body band visible

No dipping of the beam

Schwalbe’s line and anterior one-third of the trabecular meshwork visible Middle third of the trabecular meshwork visible Posterior third of the trabecular meshwork visible

252 Australian and New Zealand Journal of Ophthalmology 1996; 24(3)

Page 3: The flashlight test and van Herick's test are poor predictors for occludable angles

Table 2. Interobserver agreement Test Weighted kappa

Flashlight test (4x4) 0.74 van Hemck’s test (4x4) 0.73

Gonioscopic occludability ( 1 x 7 ) 0.845* Gonioscopy grading (7x7) 0.8 1

Over0.75 i p usually considered good agreement. Figures in parentheses indicate the number of cells in the tables used for calculation of kappa values. *Weighted kappa is the same as the unweighted value for a 2x2 table.

Goldmann single mirror as described, followed by the Sussmann four-mirror lens in all cases. As further management in any particular case depends on the clinician’s impression of whether the angle is ‘gonioscopically occludable,’ this was used as the gold standard in statistical assessments.

Agreement between a resident’s observations and that of the glaucoma specialist was deter- mined using the weighted Kappa statistic.’”,’’ The weighted kappa is a good indicator of intraclass correlation for ordinate data. A kappa value of 0 to 0.2 suggests poor agreement; 0.2 to 0.6 fair, 0.6 to 0.8 substantial and 0.8 to 1.0 represents almost perfect agreement; over 0.75 is usually considered good agreement.

Validity was assessed using occludability of the angle on gonioscopy as the gold standard.*

Results The mean age of our patients was 45.45 years (range 14 to 74 and standard deviation 14.90). There were 50 males and 46 females in the group.

The kappa values for the four tests are shown in Table 2. All tests had good interobserver agreement.

The prevalence of occludable angles accord- ing to gonioscopy was 21.9% (21 patients).

Peripheral anterior synechiae were detected in four cases.

The sensitivities and specificities of the vari- ous tests (compared to gonioscopic occludability) are shown in Table 3. Using half or more iris shadow as the cut off, the sensitiv- ity and specificity of the flashlight test were 45.5% and 82.7% respectively. For one-third iris shadow the flashlight test yielded a sensitivity of 85.71% and a specificity of 70.67%. For the van Herick’s test (Grade 1 taken as narrow), the sensitivity and specificity were 61.9% and 89.3%. Not surprisingly, the validity of gonio- scopic grading was high. Since occludability and grading are related (and assessed by the same instrument), this is not discussed further.

The usefulness of these tests in detecting nar- row angles can be judged by estimating the posi- tive and negative predictive values. Predictive values in turn primarily depend on the preva- lence of narrow angles in the various test situa- tions. Table 4 shows the estimated predictive values for the prevalence in our clinic setting (21.9%).

Discussion

Angle-closure glaucoma is a sight-threatening condition that can result in profound visual loss.’ In the early stages and in predisposed eyes the disease is curable by a non-invasive laser iridotomy.’ The prevalence of angle closure var- ies in different populations. The very body of knowledge on the prevalence of primary angle closure in communities is mostly based on stud-

Table 3. Sensitivity and specificity values Test Sensitivity Specificity

Flashlight test (half shadow) Flashlight test (one-third shadow) van Herrick’s test Gonioscopy grading Flashlight test (one-third shadow) or van Henick‘s test Flashlight test (one-third shadow) and van Herrick’s test

45.45% 85.71% 61.9% 98.67% 95.24% 52.38%

82.67% 70.67% 89.3% 95.23% 66.67% 90.67%

The flashlight test and van Hericks test are poor predictors for occludable angles 253

Page 4: The flashlight test and van Herick's test are poor predictors for occludable angles

Table 4. Predictive values of screening tests for ACG (at prevalence of occludable angles of 21.9%) Tests Positive predictive value Negative predictive value

Flashlight test (half shadow) 43.48% Flashlight test (one-third shadow)

Flashlight test (one-third shadow) or van Hemck’s test

45 % van Herrick’s test 61.9%

44.2% Flashlight test (one-third shadow) and van Herrick’s test 66.7%

84.93% 94.64% 89.33% 98.04% 87.16%

ies using the flashlight test (at least as the first stage). A screening program which uses a multi- stage testing strategy has an overall sensitivity which is the product of the sensitivities of the tests used. The flashlight test is fast, simple and cheap and has been used in prevalence studies of angle closure to identify eyes requiring goni- oscopy.* The van Herick’s test is also used to identify eyes predisposed to angle closure. How- ever, despite their widespread use there is little information available regarding their reliability or suitability as screening tests. Also, to the best of our knowledge these tests have not been di- rectly compared with gonioscopy.

All three tests had good interobserver agree- ment (Table 2). However, the sensitivity and specificity of the flashlight test were found to be low: at half iris shadow, sensitivity was 45.5% and specificity was 82.7%; the values were 85.7% and 70.7% respectively at one-third iris shadow. As van Herick’s test takes into account the peripheral anterior chamber depth, a good correlation with gonioscopic findings is expected. But even van Herick’s test had low sensitivity and specificity values (61.9% and 89.3% respec- tively).

A previous report‘i compared the flashlight and van Herick’s tests with central anterior cham- ber depth in patients with angle-closure glau- coma and controls. Using pachymetry (central anterior chamber depth of less than 2mm) as the gold standard, the flashlight test yielded a sensitivity of 89% and a specificity of 88% (the presence of any iris shadow was taken as the test positive). For van Herick’s test (consider-

ing Grades 1 and 2 as narrow) the sensitivity was 82%, but specificity was not reported. On comparing gonioscopy with pachymetry, 7% of eyes with anterior chamber depth less than 2mm had open angles (93% narrow). With depths between 2 to 2.49mm and 2.5 mm or more, nar- row angles were present in 56% and 4.5% re- spectively. In other words going by central ante- rior chamber depth alone, a significant number of narrow angles would be missed.

Our results based on occludability on goni- oscopy show that the flashlight test and van Herick’s test are of little use in screening for narrow angles. The original reference’ reported that van Herick’s test is ‘consistent’ with goni- oscopy, but the data to calculate reliability and validity were not presented.

The prevalence of occludable angles in our clinic was 21.9%. The prevalence reported from an optometry practice in the west was 6%.” There are no published community-based data from our part of the world on the prevalence of occludable angles. However, there is reason to believe that our hospital-based prevalence of occludable angles (of which very few actually develop angle closure) may be representative of our population. Narrow angles are commoner in A s i a n ~ . ~ . ’ ~ Our own population-based study (to be published) found occludable angles in 10.33% of the population between the ages of 30 to 60 years (972 individuals selected by strati- fied cluster sampling from census data of a local urban area were examined). Our hospital records show that primary angle-closure glaucoma (acute plus chronic) is as common as

254 Australian and New Zealand Journal of Ophthalmology 1996; 24(3)

Page 5: The flashlight test and van Herick's test are poor predictors for occludable angles

primary open-angle glaucoma. We have also ana- lysed consecutive axial length measurements on patients undergoing routine cataract surgery. The mean axial length was 22.69 k 0.98mm (range, 19.21 to 29.52mm) in 774 eyes compared to a Western mean of 24.2 f 0.85mm (data from Sorsby et al. 1 4 ) . This difference was significant (P<O.OOl). Our patients may have an anatomic predisposition to angle closure.

The significance of our findings is better appreciated when we consider the expected predictive values (Table 4).

Considering the prevalence of occludable angles in our clinic (2 1.9%) to be representative of a general ophthalmic practice in our region, the positive predictive values are: flashlight test half shadow 43.48%; one-third shadow 45%; and van Herick’s test 61.9%. Considering either a positive van Herick’s or a positive flashlight test as a positive result, the positive predictive value is 44.2% and for both tests positive it is 66.7%. The value of a positive result is minimal and gonioscopy is still required. Note that the positive predictive value increases with an increase in prevalence. It will be higher at higher prevalences and lower at lower prevalences as would be the case during screening.

Ideally, every patient attending an ophthal- mology clinic should undergo a gonioscopic evaluation. In a busy outpatient clinic this ideal is impractical; it would be useful if we could identify those unlikely to need a gonioscopy. The negative predictive values of the flashlight and van Herick’s tests are high and they are of value in ruling out narrow angles. At a prevalence of 21.9% the negative predictive values are: flash- light test half shadow 84.93%; one-third shadow 94.64%; and van Herick’s test 89.33%. On combining the two tests (considering both a nega- tive van Herick’s and the flashlight test one-third iris shadow being negative as a negative test), the negative predictive value is 98.04%. Thus, if both the flashlight (one-third shadow) and the van Herick’s test are negative (suggesting an

open angle) then gonioscopy need not be man- datory to confirm this (of 100 negative results 98 would have open angles). In this scenario a negative flashlight test (one-third shadow) alone has a similar implication (of 100 negative re- sults 95 would have open angles); the test is thus useful in ruling out narrow angles. (Note that the negative predictive values increase as the prevalence decreases. Hence if the prevalence is markedly higher than 20% this guideline may not be applicable.) In the Baltimore eye survey (at an estimated prevalence of between 0.72% and 2.1 %) a combination of the flashlight test and a history of glaucoma as screening criteria had a negative predictive value of 99.7%.15

It must be emphasised that if a patient is a glaucoma suspect, has raised intraocular pres- sure or a family history of glaucoma, gonios- copy should be performed irrespective of the results of the screening tests. Gonioscopy may also.be necessary to evaluate other conditions such as uveitis or a suspected intraocular for- eign body; our guidelines do not cover those situ- ations where gonioscopy is required for other reasons.

The flashlight and van Herick’s tests are firmly established in ophthalmic practice. We have demonstrated that both these are poor pre- dictors for an occludable angle. On the other hand a negative test indicates that gonioscopy may not be necessary in the individual patient.

References 1. Ritch R, Liebmann J, Solomon IS. Laser iridectomy and iridoplasty. In: The glaucomas, Vol2. Ritch R, Sheilds MB, Krupin T, eds. St Louis: CV Mosby, 1989;582-3. 2. Willensky JT, Kaufman PL, Frohlichstein D, Gieser DK, Kass MA, Ritch R, Anderson R. Follow-up of angle closure glaucoma suspects. Am J Ophthalmol 1993;115:338-46. 3. Lowe R. Aetiology of the anatomical basis for primary angle-closure glaucoma. Br J Ophthalmol 1970;54: 161-9. 4. Tomlinson A, Leighton DA. Ocular dimensions in the heredity of angle-closure glaucoma. Br J Ophthalmol 1973;57:475-86.

The flashlight test and van Herick‘s test are poor predictors for occludable angles 255

Page 6: The flashlight test and van Herick's test are poor predictors for occludable angles

5. Vargas E, Drance SM. Anterior chamber depth in an- gle-closure glaucoma: clinical methods of depth determi- nation in people with and without the disease. Arch Ophthalmol 1973;90:438-9. 6. Vargas E, Schulzer M, Drance SM. Use of oblique illumination test to predict angle-closure. Can J Ophthalmol 1974;9: 104-5. 7. van Herick W, Shaffer RN, Schwartz A. Estimation of width of angle of anterior chamber: incidence and sig- nificance of the narrow angle. Am J Ophthalmol 1969;68:626-9. 8. Congdon N, Wang F, Tielsch JM. Issues in the epide- miology and population-based screening of angle-closure glaucoma. Surv Ophthalmol 1992;36:4 1 1-23. 9. Ritch R, Liebmann J, Solomon IS. Angle closure glau- coma: mechanisms and epidemiology. In: The glaucomas, Vol 2. Ritch R, Sheilds MB, Krupin T, eds. St Louis: Mosby, 1989;834.

10. Cohen J. A coefficient of agreement for nominal scales. Educ Psycho1 Meas 1960;20:37-46. 11. Cohen J. Weighted kappa: nominal scale agreement with provision for scaled disagreement or partial credit.

12. Cockburn DM. Prevalence and significance of narrow anterior chamber angles in optometric practice. Am J Optom Physiol Opt 1981;58(2):171-5. 13. Loh RCK. The problems of glaucoma in Singapore. Singapore Med J 1968;9:76-80. 14. Sorsby A, Leary GA, Richards MJ. Correlation ametropia and component ametropia. Vision Res 19622: 309. 15. Patel KH, Javit JC, Tielsch JM, Street DA, Katz J, Quigley HA, Sommer A. Incidence of acute angle clo- sure glaucoma after pharmacological mydriasis. Am J Ophthalmol 1995;120:709-17.

Psycho1 Bull 1968;70:213-20.

Appendix: calculation of kappa k = observed - expected

1 - expected In calculating the weighted kappa the overall disagreement was measured using the square of the distance from the diagonal as the weight measure. The table shows the results ofthe flashlight test for both observers. In assessing agreement due consideration was given to the fact that the degree of disagreement between no shadow and < one-third shadow was minimal in comparison to being classified as, say no shadow and > half shadow.'' We used the square of the distance from the diagonal as the weight measure in calculating disagreement and the weighted kappa was obtained by subtracting the overall disagreement from one. Weighted kappa = 1 -disagreement = 0.74 The kappa values for the four tests are shown in Table 2. All tests had good interobserver agreement.

Interobserver agreement for the flashlight test Observer1 + No Minimal 112- 1 13 >1/2 Total Observer2 ) shadow shadow No shadow 22 7 1 1 31 Minimal shadow 12 11 4 4 31 112-113 0 3 9 4 16 >I12 0 1 3 14 18 Total 34 22 17 23 96

256 Australian and New Zealand Journal of Ophthalmology 1996; 24(3)