Screening for Narrow Angles in the Singapore Population: Evaluation of New Noncontact Screening Methods

Screening for Narrow Angles in the Singapore Population: Evaluation of New Noncontact Screening Methods Raghavan Lavanya, MD,1 Paul J. Foster, FRCS,2 Lisandro M. Sakata, MD, PhD,1 David S. Friedman, MD, PhD,3 Kenji Kashiwagi, MD, PhD,4 Tien-Yin Wong, MD, PhD,1 Han T. Aung, MD,1 Tamuno Alfred, MSc,5 Hong Gao, MD,5 Adrian G. Ee, MD,6 Steve K. Seah, FRCS,1 Tin Aung, PhD, FRCS(Ed)1,7 Purpose: To assess the screening effectiveness of 3 new noncontact devices, the scanning peripheral anterior chamber depth analyzer (SPAC) (Takagi, Nagano, Japan), which measures peripheral anterior chamber depth (ACD); IOLMaster (Carl Zeiss Meditec, Jena, Germany), which measures central ACD; and Visante anterior segment optical coherence tomography (AS-OCT) (Visante, Carl Zeiss Meditec, Dublin, CA), which images the angles, and to compare these instruments with gonioscopy in identifying people with narrow angles (NAs). Design: Cross-sectional, observational, community-based study. Participants: Phakic subjects aged ⱖ50 years without ophthalmic symptoms who were recruited from a community polyclinic in Singapore. Methods: All subjects underwent examination with SPAC, IOLMaster, and AS-OCT in the dark by a single operator. Gonioscopy was performed by an ophthalmologist masked to the instruments’ findings. The area under the curve (AUC) receiver operating characteristic (ROC) was generated to assess the performance of these tests in detecting people with a NA in either eye. Main Outcome Measures: Eyes were classified as having NAs by gonioscopy if the posterior pigmented trabecular meshwork could be seen for ⱕ2 quadrants of the angle circumference with or without peripheral anterior synechiae. Results: A total of 2052 subjects were examined and underwent all 3 tests. The prevalence of a NA in at least 1 eye diagnosed by gonioscopy was 20.4% (422 subjects). The AUC for the SPAC using a numeric grade ⬍5 as a cutoff was 0.83 (95% confidence interval [CI], 0.82– 0.85), with a sensitivity of 90.0% (95% CI, 86.8 –92.7) and a specificity of 76.6% (95% CI, 74.4 –78.6). The AUC for the IOLMaster at an ACD cutoff of ⬍2.87 mm was 0.83 (95% CI, 0.81– 0.85), with a sensitivity of 87.7% (95% CI, 84.2–90.7) and a specificity of 77.7% (95% CI, 75.6 –79.7). The AUC for the AS-OCT was 0.76 (95% CI, 0.74 – 0.78), with a sensitivity of 88.4% (95% CI, 84.9 –91.3) and a specificity of 62.9% (95% CI, 60.5– 65.2). Conclusions: The low specificity found with the SPAC, IOLMaster, and AS-OCT may limit the usefulness of these devices in screening for NAs. Financial Disclosure(s): Proprietary or commercial disclosure may be found after the references. Ophthalmology 2008;115:1720 –1727 © 2008 by the American Academy of Ophthalmology.

Glaucoma is the most common cause of irreversible blindness worldwide, affecting approximately 67 million people.1–3 Primary angle closure glaucoma (PACG) is a major form of glaucoma in Asia. Because of its high prevalence in the populous nations of China and India, and because it is an aggressive and visually destructive disease, PACG is a significant cause of global visual morbidity.4 – 8 The high costs of health care and the economic burden of visual loss caused by PACG have prompted many to consider whether there is a need for public health initiatives to combat blindness from this condition. Eyes with PACG are characterized by irido-trabecular contact, a pathologic state in which aqueous outflow is

1720

© 2008 by the American Academy of Ophthalmology Published by Elsevier Inc.

being partially impaired as a consequence of physical obstruction of the trabecular meshwork (TM) by appositional or synechial closure. This leads to glaucomatous damage to the optic nerve caused by an elevation of intraocular pressure. In the early stages of the disease, eyes are considered at risk of angle closure if they possess particular anatomic characteristics that place the iris and the TM in sufficiently close proximity for contact to be considered possible. These eyes are said to have narrow angles (NAs). To effectively prevent PACG by the use of prophylactic laser iridotomy, it is necessary to identify people with the early stages of disease. Once manifest angle closure has occurred, and particularly after an acute attack or when ISSN 0161-6420/08/$–see front matter doi:10.1016/j.ophtha.2008.03.015

Lavanya et al 䡠 Screening for Angle Closure Using IOLMaster, AS-OCT, and SPAC glaucomatous optic neuropathy has developed, laser iridotomy is less effective in preventing an increase in intraocular pressure.9,10 Detection of anatomically NAs (latent stage) is thus a key component of any screening program to prevent PACG. Gonioscopy is the clinical reference standard for assessing the angle and diagnosing angle closure. However, it is a semisubjective technique, requires considerable skill and experience, and involves contact with the cornea, making it an inappropriate initial screening test. Intraocular pressure measurement and optic disc evaluation have been evaluated as screening techniques for open angle glaucoma and were reported to be limited in their usefulness as screening tools.11 Anatomic biometric characteristics, such as a shallower anterior chamber depth (ACD), thick anteriorly positioned lens, shorter axial length, and steeply curved cornea, have been linked with angle closure.12–15 Among these, shallow ACD, both centrally and peripherally, is regarded as a cardinal risk factor for angle closure in most racial groups. Thus, the role of central and peripheral ACD measurements in screening for PACG has been explored.16 –19 We previously reported that Singapore has a high prevalence of PACG5 and the world’s highest incidence of symptomatic angle closure.20 The aim of this study was to evaluate 3 new noncontact screening methods for NAs in an older Singaporean population: the IOLMaster (Carl Zeiss Meditec, Jena, Germany), which measures central ACD; the scanning peripheral anterior chamber depth analyzer (SPAC) (Takagi, Nagano, Japan), which assesses peripheral ACD; and the anterior segment optical coherence tomography (AS-OCT) (Visante, Carl Zeiss Meditec, Dublin, CA), which images the angles directly. We compared these instruments with gonioscopic examination of the angle. Although these new devices have been evaluated in hospitalbased studies, no data have been published assessing the screening potential of these noncontact instruments in a community setting.

Materials and Methods Study Population and Recruitment The study had the approval of the ethics review board of the Singapore Eye Research Institute and was performed in accordance with the tenets of the Declaration of Helsinki. In this prospective cross-sectional study, subjects aged ⬎50 years who did not have any ophthalmic symptoms were recruited from a government-run community polyclinic providing primary health care services. This polyclinic serves ⬎10,000 people per month, mainly of lower to middle socioeconomic status with a high proportion requiring chronic disease management. Subjects were identified by systematic sampling (every fifth patient registered at the polyclinic) and asked to participate in the study after obtaining written informed consent. A detailed questionnaire, including demographic, socioeconomic details, and medical and ocular history, was administered. The exclusion criteria were a history of glaucoma, previous intraocular surgery or penetrating eye injury, and corneal disorders, such as corneal endothelial dystrophy, corneal opacity, or pterygium, preventing ACD measurement.

Imaging The IOLMaster, AS-OCT, and SPAC (all of which are noncontact) measurements were performed on all subjects in the same order under standardized dark conditions (20 lux) by an experienced operator who was masked to the results of the clinical ophthalmic examination. The IOLMaster is an ocular biometric device that measures axial length, ACD, and corneal curvature.21–22 The IOLMaster takes 5 simultaneous ACD measurements, and the mean of these 5 readings is displayed with an accuracy of 10 ␮m.23 The AS-OCT uses 1.3 ␮m infrared light to obtain highresolution, cross-sectional tomographic images of the anterior segment24 –26 Scans were centered on the pupil and taken along the horizontal (nasal-temporal angles at 0°–180°) and vertical meridians (superior–inferior angles 90°–270°). Images were then processed using inbuilt software that dewarps the images (adjusting for distortions arising from corneal optical properties) and assessed by 2 individuals (LS and TA working together) who were masked to the gonioscopy findings. A closed angle on AS-OCT was defined by contact between the iris and any part of angle wall anterior to the scleral spur. An eye was defined as having closed angles on AS-OCT if ⱖ2 quadrants (nasal, temporal, superior, and inferior) were closed. An individual was classified as having closed angles if ⱖ2 quadrants of the angle were closed in either eye. The SPAC uses optical methods to assess the peripheral ACD and indirectly evaluates the iris profile.27–29 The SPAC scans the ACD from the optical axis to the temporal limbus in approximately 0.66 seconds, taking 21 consecutive slit-lamp images at 0.4-mm intervals from the visual axis, recording cross-sectional slit images from the cornea to the anterior iris. Three measurements are performed automatically, and the mean of the ACD values at each point is measured. The SPAC converts these measurements into numeric and categoric grades by comparison with a normative database derived from a sample of Japanese subjects. Each eye is classified by the device on a numeric scale from 1 to 12, with 12 representing the deepest ACD. The SPAC also reports categoric grades for risk of angle closure: S (for “suspect angle closure”), P (for “potential angle closure”), and no suffix (for “normal”).

Gonioscopy Gonioscopy was performed by a single trained ophthalmologist (RL) with extensive experience in performing gonioscopy in epidemiologic studies, who was standardized against another ophthalmologist with subspecialty glaucoma training (TA). A weighted kappa of 0.82 was achieved for the assessment of angle grading. The gonioscopist was masked to the findings of the imaging devices. Dynamic gonioscopy was performed using a Sussman 4-mirror lens (Ocular Instruments Inc, Bellevue, WA) at high magnification (⫻16) with the eye in the primary position of gaze. A 1-mm light beam was reduced to a very narrow slit, and the vertical beam was offset horizontally for assessing superior and inferior angles, and vertically for nasal and temporal angles. Care was taken to avoid light falling on the pupil during gonioscopy. The angle in each quadrant was graded using the Scheie grading system according to the anatomic structures observed during gonioscopy.30 Indentation gonioscopy was used to establish presence or absence of peripheral anterior synechiae (PAS). We used the International Society for Geographical and Epidemiological Ophthalmology glaucoma classification.31 Because the primary aim of the study was to screen for NAs in the community using the 3 devices, we combined primary angle closure suspect, primary angle closure (PAC), and PACG into a single category, namely, NAs. An eye was defined as having NAs by gonioscopy if the

1721

Ophthalmology Volume 115, Number 10, October 2008 posterior pigmented TM was not visible on non-indentation gonioscopy for at least 180 degrees, with or without PAS. An individual was classified as having NAs if at least 1 eye had NAs. In addition to gonioscopy, all participants underwent a comprehensive eye examination by the same ophthalmologist at the same visit. This included visual acuity measurement using a logarithm of minimum angle of resolution (logMAR chart, The Lighthouse, Long Island, NY), slit-lamp (model BQ 900; Haag-Streit, Bern, Switzerland) examination of the anterior segment, and intraocular pressure measurement using Goldmann applanation tonometry and stereoscopic optic disc examination with a 78 diopter lens (Volk Optical Inc, Mentor, OH) through an undilated pupil.

Sample Size Calculation A population-based study in Chinese Singaporeans (The Tanjong Pagar study)5 had estimated the prevalence of NAs in subjects aged ⬎50 years to be 10.6%. By using data from this study, the estimated area under the curve (AUC) for peripheral ACD receiver operating characteristic (ROC) curve was 0.9. The AUC for the shallowest optical central ACD ROC was 0.86. With a type I error rate of 5% and a power of 80%, the sample size would be 417 for an unpaired comparison. By assuming that the numbers of those with normal findings and abnormal findings are equal, approximately 210 people with NAs would be required. By assuming a rate of NAs of 10% in those aged ⬎50 years, with a 95% confidence interval (CI) of 87.8% to 92.2% and 83.5% to 88.5%, respectively, there was a sample size of 4200 eyes, that is, 2100 people was calculated.

Statistical Analysis Although the performance of the devices was assessed individually for both eyes and subjects, the unit of analysis for presentation in this article was individuals, not eyes. Statistical analysis was done using the statistical package Stata 9.1 (StataCorp LP, College Station, TX). Data were entered into CLINTRIAL, and reporting of the study followed the STARD guidelines. ROC curves, with calculations of AUC and 95% CIs, were used as an index of global test performance. Values for specificity, sensitivity, negative predictive value (NPV), positive predictive value (PPV), and the ROC curve were calculated for each of the screening methods with gonioscopy as the reference standard. PPV and NPV were calculated both by using a population prevalence of NAs of 10% from the Tanjong Pagar population-based study5 and 20% from the current study. We also assessed the performance of the 3 devices for detecting PAC and PACG using population prevalences of 2.2% and 0.8%, respectively,5 for calculating PPV and NPV. A nonparametric approach32 was used to compare the areas under the ROC curves from the different instruments using Stata’s roccomp command.

Results A total of 2114 subjects were screened from December of 2005 to June of 2006. Twelve subjects were ineligible because they were pseudophakic in both eyes or were known to have glaucoma and were excluded. Approximately 90% of subjects were Chinese, with a mean age of 63.3⫾8.0 years, and 1085 (52.9%) were women (Table 1). Of the 2102 eligible subjects, 50 subjects could not complete the tests for various reasons: alignment errors (12); inability to follow instructions (16) or focus on the fixation light (4); refused gonioscopy (4); or other reasons (14). Therefore, data

1722

from the 2052 subjects who completed all the tests were included in the analysis. There were no significant differences in age (P ⫽ 0.13), race (P ⫽ 0.17), refractive spherical error (P ⫽ 0.16), axial length (P ⫽ 0.36), or central ACD (P ⫽ 0.55) between those who completed the tests and those who did not. NAs in at least 1 eye on gonioscopy were present in 422 individuals (20.6%). Subjects with NAs were significantly older and more hyperopic, and had higher intraocular pressures, shallower ACDs, and shorter axial lengths than subjects without NAs (Table 1). The AUC for detecting subjects with NAs compared with gonioscopy was lowest for AS-OCT: 0.76 (95% CI, 0.74 – 0.78). The SPAC numeric grade of ⱕ5 had an AUC of 0.83 (95% CI, 0.82– 0.85), whereas the categoric grade S or P had an AUC of 0.80 (95% CI, 0.79 – 0.82) (Table 2). Central ACD measurement with the IOLMaster (using a cutoff of 2.87 mm) had an AUC of 0.83 (95% CI, 0.81– 0.85). Nonparametric tests32 showed that the AUC of the 3 devices differed significantly from each other (P⬍0.001). By using an alternative cutoff with SPAC grades of <6, S or P, sensitivity using the SPAC increased to 97.6% (95% CI, 95.7–98.9) and specificity decreased to 55.6% (95% CI, 53.2– 58.1); the PPV was 19.7% (95% CI: 18.8 –20.6), and the NPV was 99.5 (95% CI, 99.1–99.7) (Table 3). For the IOLMaster, the specificity was higher at a cutoff of 2.53 mm for central ACD (97.2%; 95% CI, 96.3–98.0), but sensitivity decreased to 35.8% (95% CI, 31.2– 40.6) (Table 4). The performance characteristics of the 3 devices seem to be similar regardless of age or gender (Table 5; available at http://aaojournal.org). For the highest risk population in this study (women aged ⬎60 years), the sensitivity of SPAC using a combined grade of <6 and categoric grades S and P was approximately 100%, with a specificity of 41%. By using the IOLMaster (using a cutoff of 2.87 mm), the PPV was 5.6% for detecting subjects with PAC (95% CI, 5.1– 6.1) and 1.9% for detecting subjects with PACG (95% CI: 1.6 –2.2), whereas the corresponding NPVs were 99.6% (95% CI, 99.3–99.8) and 99.8% (95% CI, 99.5–99.9), respectively. The PPV of SPAC (using a combined grade <5 and categoric grades S and P) was 4.5% for detecting PAC (95% CI, 4.2– 4.9) and 1.4% for detecting PACG (95% CI, 1.2–1.7), whereas the corresponding NPVs were 99.6% (95% CI, 99.3–99.8) and 99.7% (95% CI: 99.4 –99.9), respectively. For the AS-OCT, the PPV was 4.1% for PAC (95% CI, 3.8 – 4.5) and 1.5% for PACG (95% CI, 1.3–1.7), whereas the NPV was 99.5% for PAC (95% CI, 99.1–99.7) and 99.8% for PACG (95% CI, 99.6 –99.9).

Discussion The ideal community-based screening test should be clinicianindependent, quick, and noninvasive, and have a very high specificity. Although a high sensitivity is important, the choice of the specific thresholds used to trigger a referral for examination is as much a question of public health and health economics issue as a clinical one. The severity of the disease and its impact on the national economy and personal finances need to be balanced against individual and national costs of testing, definitive examination, and preventive treatment. Furthermore, for screening to be a viable proposition, additional resources (infrastructure, instrumentation, and staff) are needed to confirm the diagnosis, treat, and provide long-term care for those in whom the diagnosis is confirmed.33,34 All 3 instruments assessed in this community-based screening study were safely operated by technicians, pro-

Lavanya et al 䡠 Screening for Angle Closure Using IOLMaster, AS-OCT, and SPAC Table 1. Demographic Data of the Study Population With and Without Narrow Angles* Total Subjects (n ⴝ 2052)

Measures 1.

2.

3.

4.

5.

6.

7.

Age (y) Mean (SD) Median (range) Gender, No. (% ) Male Female Race, No. (% ) Chinese Malay Indian Others Autorefraction spherical error§ Mean (SD) Median (range) Intraocular pressure (mmHg)§ Mean (SD) Median (range) Axial length (mm)§ Mean (SD) Median (range) Central ACD (mm)§ Mean (SD) Median (range)

Narrow Angles (n ⴝ 422)

Open Angles (n ⴝ 1630)

P Value

63.3 (8.0) 63 (50–93)

65.5 (8.2) 65 (50–89)

62.8 (7.9) 62 (50–93)

⬍0.001†

967 (47.1) 1085 (52.9)

152 (36.0) 270 (64.0)

815 (50.0) 815 (50.0)

⬍0.001‡

1840 (89.7) 43 (2.1) 146 (7.1) 23 (1.1) (n ⫽ 1996) 0.21 (2.64) 0.75 (⫺15.70 to 8.87)

399 (94.5) 7 (1.7) 13 (3.1) 3 (0.7) (n ⫽ 411) 1.47 (1.65) 1.37 (⫺5.62 to 8.37)

14.8 (2.4) 14 (8–26)

15.3 (2.5) 16 (10–26)

1441 (88.4) 36 (2.2) 133 (8.2) 20 (1.2) (n ⫽ 1585) ⫺0.12 (2.75) 0.50 (⫺15.70 to 8.87) 14.6 (2.4) 14 (8–26)

0.002‡

⬍0.001† ⬍0.001†

23.91 (1.33) 23.67 (20.47–31.02)

23.08 (0.86) 23.02 (20.78–26.94)

24.13 (1.35) 23.87 (20.47–31.02)

⬍0.001†

3.09 (0.37) 3.09 (1.72–4.30)

2.68 (0.22) 2.67 (1.72–3.32)

3.20 (0.33) 3.19 (2.29–4.30)

⬍0.001†

SD ⫽ standard deviation; ACD ⫽ anterior chamber depth. Missing data: Autorefraction 108; intraocular pressure 3; axial length 41; central ACD 2. *NAs includes all subjects with primary angle closure suspect (PACS), primary angle closure (PAC), and primary angle closure glaucoma (PACG). † Wilcoxon rank-sum test. ‡ Chi-square test. § Measures 4 to 7 include measurements from right eye only.

vided immediate results, and performed reasonably well, but none gave very high combinations (⬎95%) of sensitivity and specificity when trying to identify subjects with NAs. This is not consistent with previous findings from Mongolia16 and Singapore,18 which suggested better performance for central and limbal ACD measurements in screening for NAs (summarized in Table 6). Axial ACD measurement gave an AUC of 0.9316 and 0.86,18 respectively, compared with 0.83 for the IOLMaster ACD, at a cutoff of 2.87 mm in the current study. A higher cutoff of ⬍3.0 mm

(AUC 0.81) was highly sensitive (96.7%) in identifying NAs but resulted in substantially lower specificity, whereas a slightly lower ACD cutoff (⬍2.53 mm) resulted in a low sensitivity (35.8%), a high specificity (97.2%), and an AUC of 0.67. In a previous article, we discussed the difference between figures from Mongolia and Singapore, which showed a much tighter relationship between ACD and PAS (used as an index of angle closure) in Mongolians than for Singaporeans.35 We believe this may reflect a difference in the mechanism of disease, with pure pupil-block pre-

Table 2. Sensitivity, Specificity, Positive Predictive Value, Negative Predictive Value, and Area Under the Curve (95% Confidence Interval) for Each Device Compared with Gonioscopy, by Subjects (n ⫽ 2052)

Measure

IOLMaster (Carl Zeiss Meditec, Jena, Germany) (<2.87 mm)

SPAC (Takagi, Nagano, Japan) Numeric Grade <5

SPAC Categoric Grade P or S

SPAC Combined Grade <5 and/or S or P

AS-OCT (Visante, Carl Zeiss Meditec, Dublin, CA) 2 Quadrants Closed

Sensitivity, % (95% CI) Specificity, % (95% CI) PPV,* % (95% CI) NPV,* % (95% CI) PPV,† % (95% CI) NPV,† % (95% CI) AUC (95% CI)

87.7 (84.2–90.7) 77.7 (75.6–79.7) 30.4 (28.4–32.5) 98.3 (97.8–98.7) 49.5 (47.1–52.0) 96.2 (95.1–97.0) 0.83 (0.81–0.85)

90.0 (86.8–92.7) 76.6 (74.4–78.6) 29.9 (28.0–31.9) 98.6 (98.1–98.9) 49.0 (46.7–51.3) 96.9 (95.8–97.6) 0.83 (0.82–0.85)

92.4 (89.5–94.8) 68.3 (66.0–70.6) 24.5 (23.1–25.9) 98.8 (98.3–99.1) 42.2 (40.3–44.1) 97.3 (96.3–98.1) 0.80 (0.79–0.82)

92.9 (90.0–95.2) 66.7 (64.4–69.0) 23.7 (22.4–25.0) 98.8 (98.4–99.2) 41.1 (39.3–42.9) 97.4 (96.4–98.2) 0.80 (0.78–0.82)

88.4 (84.9–91.3) 62.9 (60.5–65.2) 20.9 (19.8–22.1) 98.0 (97.4–98.5) 37.3 (35.6–39.0) 95.6 (94.3–96.6) 0.76 (0.74–0.78)

SPAC ⫽ scanning peripheral anterior chamber depth analyzer; AS-OCT ⫽ anterior segment optical coherence tomography; CI ⫽ confidence interval; PPV ⫽ positive predictive value; NPV ⫽ negative predictive value; AUC ⫽ area under the curve. *Assuming 10% prevalence. † Assuming 20% prevalence.

1723

Ophthalmology Volume 115, Number 10, October 2008 Table 3. Performance of Scanning Peripheral Anterior Chamber Depth Analyzer Using Alternate Cutoffs in Identifying Narrow Angles* by Subjects (n ⫽ 2052) Measure

SPAC Combined Grade <5 and/or S or P

SPAC Combined Grade <6 and/or S or P

SPAC Numeric Grade <6

Sensitivity, % (95% CI) Specificity, % (95% CI) PPV,† % (95% CI) NPV,† % (95% CI) PPV,‡ % (95% CI) NPV,‡ % (95% CI) AUC (95% CI)

92.9 (90.0–95.2) 66.7 (64.4–69.0) 23.7 (22.4–25.0) 98.8 (98.4–99.2) 41.1 (39.3–42.9) 97.4 (96.4–98.2) 0.80 (0.78–0.82)

97.6 (95.7–98.9) 55.6 (53.2–58.1) 19.7 (18.8–20.6) 99.5 (99.1–99.7) 35.5 (34.2–36.8) 98.9 (98.1–99.4) 0.77 (0.75–0.78)

97.6 (95.7–98.9) 57.4 (55.0–59.8) 20.3 (19.4–21.3) 99.5 (99.2–99.8) 36.4 (35.1–37.8) 99.0 (98.1–99.4) 0.78 (0.76–0.79)

SPAC ⫽ scanning peripheral anterior chamber depth analyzer; CI ⫽ confidence interval; PPV ⫽ positive predictive value; NPV ⫽ negative predictive value; AUC ⫽ area under the curve. *NAs include all subjects with PACS, PAC, and PACG. † Assuming 10% prevalence. ‡ Assuming 20% prevalence.

dominating in Mongolia, whereas in Singapore anatomic factors affecting the peripheral anterior chamber are more important. In our previous report on possible screening tests in Singapore,18 limbal ACD estimation gave an AUC of 0.90. In this study, the performance of the SPAC was lower (AUC 0.83). We previously reported the performance of SPAC relative to modified limbal ACD grading, citing AUCs of 0.79 versus 0.87, respectively, in a group of hospital patients.36 Our current data suggest that this previous AUC for SPAC screening was at the lower end of the range of expected values. The SPAC combined grade (P or S and/or ⱕ5) gave a sensitivity and specificity of 92.9% and 66.7%, respectively. By assuming a population prevalence of NAs of 10% (from the Tanjong Pagar population-based study), the PPV of the test would be 23.7% in people aged ⱖ50 years. This means that approximately only 1 in 4 of those with abnormal findings on screening would actually have anatomically NAs. However, in our study, in which participants were drawn from patients attending a general polyclinic, the prevalence of NAs was 20.6%. In this case, approximately 1 of 2 people who were positive on screening would actually have anatomically NAs. Even with the rel-

atively weak performance of the SPAC, using this device to trigger a referral to an ophthalmologist may be a feasible design for a strategy of prevention of angle closure. Given the high sensitivity of SPAC, an alternative or additional approach to screening could rely on SPAC as a universal first-line screening test for all patients attending an ophthalmologist in Singapore. Those identified by SPAC as having a risk of angle closure could undergo a definitive gonioscopic examination at the same visit and managed accordingly. An alternate low-cost strategy could be assessing the limbal chamber depth by the van Herick test for screening for NAs. In light of the good performance of the van Herick test18 and by assuming all ophthalmologic examinations will incorporate a slit-lamp examination, a limbal chamber depth (van Herick) examination could be considered at no additional cost and only a few extra seconds of time. AS-OCT tended to detect more closed angles than gonioscopy. This finding agrees with data from a previous hospital-based study.25 The differences between gonioscopy and AS-OCT may be due to the different anatomic landmarks and levels of irido-TM contact used to define a closed angle. One possible explanation for the high “false-positive” rates with AS-OCT could be the presence of low irido-TM con-

Table 4. Performance of IOLMaster Using Different Central Anterior Chamber Depth Measurements in Identifying Narrow Angles* by Subjects (n ⫽ 2052) Measures

IOLMaster ACD Measurements <2.40 mm

<2.53 mm

<2.60 mm

<2.87 mm

<3.00 mm

Sensitivity, % (95% CI) Specificity, % (95% CI) PPV,† % (95% CI) NPV,† % (95% CI) PPV,‡ % (95% CI) NPV,‡ % (95% CI) AUC (95% CI)

15.9 (12.5–19.7) 99.1 (98.5–99.5) 65.7 (52.5–76.9) 91.4 (91.0–91.7) 81.2 (71.3–88.2) 82.5 (81.9–83.1) 0.58 (0.56–0.59)

35.8 (31.2–40.6) 97.2 (96.3–98.0) 59.0 (51.2–66.4) 93.2 (92.7–93.6) 76.4 (70.3–81.6) 85.8 (84.9–86.7) 0.67 (0.64–0.69)

48.3 (43.5–53.2) 94.8 (93.6–95.8) 50.7 (45.0–56.4) 94.3 (93.8–94.8) 69.9 (64.8–74.5) 88.0 (87.0–89.0) 0.72 (0.69–0.74)

87.7 (84.2–90.7) 77.7 (75.6–79.7) 30.4 (28.4–32.5) 98.3 (97.8–98.7) 49.5 (47.1–52.0) 96.2 (95.1–97.0) 0.83 (0.81–0.85)

96.7 (94.5–98.2) 65.0 (62.7–67.3) 23.5 (22.3–24.8) 99.4 (99.1–99.7) 40.9 (39.2–42.5) 98.7 (97.9–99.2) 0.81 (0.79–0.82)

ACD ⫽ anterior chamber depth; CI ⫽ confidence interval; PPV ⫽ positive predictive value; NPV ⫽ negative predictive value; AUC ⫽ area under the curve. *NAs include all subjects with PACS, PAC, and PACG. † Assuming 10% prevalence. ‡ Assuming 20% prevalence.

1724

Lavanya et al 䡠 Screening for Angle Closure Using IOLMaster, AS-OCT, and SPAC Table 6. Comparison of Performance of Screening Tests in Previous Studies Alsbirk13 Location Setting No. of subjects Test

Greenland Community 1067 Optical ACD

Cutoff ACD Sensitivity (%) Specificity (%) Reference standard

⬍2.00 mm 86 88 Symptoms or tonometry, gonioscopy

Congdon et al19

Devereux et al16

Devereux et al16

Devereux et al16

Taiwan Community 562 Ultrasound ACD* ⬍2.70 mm 76.9 87.0 Gonioscopy

Mongolia Community 1717 Optical ACD*

Mongolia Community 937 Slit-lamp ultrasound ⬍2.60 mm 83 81 Gonioscopy

Mongolia Community 461 Handheld ultrasound ⬍2.53 mm 86 73 Gonioscopy

⬍2.22 mm 85 84 Gonioscopy

Nolan et al18 Singapore Population 1092 Ultrasound ACD ⬍2.53 mm 75.6 73.7 Gonioscopy

Nolan et al18 Singapore Population 1092 Limbal ACD–van Herick grade (ⱕ15%) 83.0 88.1 Gonioscopy

ACD ⫽ anterior chamber depth.

tact just above the scleral spur in any one quadrant, which would have been labeled “open” by gonioscopy. Further subanalysis37 in a smaller cohort of 502 subjects revealed that a low irido-angle contact just above the scleral spur on AS-OCT was observed in 71% of angles graded as open by gonioscopy and closed on AS-OCT. On AS-OCT, this low irido-angle contact would be considered a closed angle because this was defined as the presence of any contact between the iris and the angle wall anterior to the scleral spur. However, because this apposition may not have occurred up to the level of the posterior TM, the angle may not have been graded as closed on gonioscopy. Another explanation for the discrepancy between the 2 tests could be the difference in illumination. Although both gonioscopy and AS-OCT examinations were undertaken under dim lighting conditions, complete darkness was possible for the AS-OCT examination because AS-OCT uses infrared light. Despite efforts to use as little light as possible for gonioscopy, the anterior segment and pupil are nevertheless exposed to light during gonioscopy. This small amount of light may artificially open up an angle that may have been closed in the dark. The scleral spur could not be visualized in approximately 30% of AS-OCT images.38 However, the presence or absence of closed angles could still be assessed in ⬎90% of the eyes. Some eyes had deep and open angles with the iris base clearly going into the ciliary body. In these cases, the examiners who graded the AS-OCT images understood that scleral spur location was not necessary for diagnosis of an open angle. Similarly, some angles seemed clearly closed, because they showed a large area of iris contact to the angle wall anterior to the insertion of the iris. Only 10% of the images could not be classified because of a lack of visualization of the scleral spur, and these were excluded from the analysis. We previously suggested that the ASOCT may actually be outperforming even an expert human gonioscopist,25 because it performs noncontact angle imaging in near-dark conditions. Only longitudinal studies will accurately determine whether eyes labeled as closed by only the AS-OCT are at risk of developing acute PAC attacks, PAC, or PACG.39 We used a definition of NAs of ⱖ180 degrees (ⱖ2 quadrants) of angle circumference in which the posterior

TM was hidden from view. Using a grading system based on visible angle structures is not the same as detecting irido-trabecular contact. The Liwan Eye Study,40 using the definition of ⱖ180 degrees, demonstrated a prevalence of NAs that was fairly similar to ours (17.5%). By using a more stringent definition of ⱖ270 degrees, the Tanjong Pagar Study5 and the Liwan Eye study40 reported a population-based prevalence of NAs of 10.6% and 11.0%, respectively (⬎50 years of age). Of note, using the 3 quadrant definition did not greatly alter the AUC, sensitivity, or specificity values of the 3 devices (data not shown). There are several explanations for the differences in the prevalence of NAs in different Singapore studies. First and most likely is the inherent variability of gonioscopy in determining the configuration of the angle. Second, it is possible that people attending a polyclinic truly differ from a randomly selected sample of the general population. If the polyclinic population does indeed have a higher prevalence of NAs, this would suggest they represent a “pathology-enriched” group in whom screening would be more productive. Indeed, a productive approach would be to target high-risk groups, such as the elderly, and in particular, women. Most certainly the devices we assessed could play an important role in clinics where they could be used to detect at-risk persons who can easily undergo definitive evaluation while at the same visit. Although the risks and benefits of prophylactic laser iridotomy for all people with NAs are a matter of ongoing research, early detection of anatomically closed angles will certainly be an important component of future blindnessprevention programs in Asia. Additional research is needed to investigate the cost-effectiveness of different strategies of prevention. Ultimately, the decision to embark on any form of screening is a matter for individuals, governments, and health care providers, weighing the relative benefits and costs. The exact role for these devices will vary between nations and between different health care sectors (general physicians, optometrists, and ophthalmologists). The viability of universal, proactive screening remains unproven, despite the fact that PACG causes substantial amounts of blindness in Singapore and throughout Asia. All ophthalmologists in Asia should be aware of the impact of PACG

1725

Ophthalmology Volume 115, Number 10, October 2008 and take all possible steps to identify the condition. Without the need for any additional imaging or biometric equipment, it should be emphasized that all patients with glaucoma and those with suspected glaucoma must undergo gonioscopy. Case detection among patients attending general ophthalmology clinics could be the first small yet significant step in reducing blindness caused by PACG in Asia.

References 1. Quigley HA. Number of people with glaucoma worldwide. Br J Ophthalmol 1996;80:389 –93. 2. Thylefors B, Negrel AD, Pararajasegaram R, Dadzie KY. Global data on blindness. Bull World Health Organ 1995;73: 115–21. 3. Wong TY, Loon SC, Saw SM. The epidemiology of age related eye diseases in Asia. Br J Ophthalmol 2006;90: 506 –11. 4. Foster PJ, Baasanhu J, Alsbirk PH, et al. Glaucoma in Mongolia: a population-based survey in Hovsgol Province, northern Mongolia. Arch Ophthalmol 1996;114:1235– 41. 5. Foster PJ, Oen FT, Machin D, et al. The prevalence of glaucoma in Chinese residents of Singapore: a cross-sectional population survey of the Tanjong Pagar district. Arch Ophthalmol 2000;118: 1105–11. 6. Dandona L, Dandona R, Mandal P, et al. Angle closure glaucoma in an urban population in southern India: the Andhra Pradesh Eye Disease Study. Ophthalmology 2000;107:1710 – 6. 7. Jacob A, Thomas R, Koshi SP, et al. Prevalence of primary glaucoma in an urban south Indian population. Indian J Ophthalmol 1998;46:81– 6. 8. He M, Foster PJ, Ge J, et al. Prevalence and clinical characteristics of glaucoma in adult Chinese: a population-based study in Liwan district, Guangzhou. Invest Ophthalmol Vis Sci 2006;47:2782– 8. 9. Nolan WP, Foster PJ, Devereux JG, et al. YAG laser iridotomy treatment for primary angle closure in East Asian eyes. Br J Ophthalmol 2000;84:1255–9. 10. Alsagoff Z, Aung T, Ang LP, Chew PT. Long-term clinical course of primary angle-closure glaucoma in an Asian population. Ophthalmology 2000;107:2300 – 4. 11. Tielsch JM, Katz J, Singh K, et al. A population-based evaluation of glaucoma screening: the Baltimore Eye Survey. Am J Epidemiol 1991;134:1102–10. 12. Alsbirk PH. Primary angle closure glaucoma: oculometry, epidemiology, and genetics in a high risk population. Acta Ophthalmol Suppl 1976;127:5–31. 13. Alsbirk PH. Anterior chamber depth and primary angle closure glaucoma I An epidemiological study in Greenland Eskimos. Acta Ophthalmol (Copenh) 1975;53:89 –104. 14. Lowe RF. Aetiology of the anatomical basis for primary angle closure glaucoma: biometrical comparisons between normal eyes and eyes with primary angle-closure glaucoma. Br J Ophthalmol 1970;54:161–9. 15. Sihota R, Lakshmaiah NC, Agarwal HC, et al. Ocular parameters in the subgroups of angle closure glaucoma. Clin Experiment Ophthalmol 2000;28:253– 8. 16. Devereux JG, Foster PJ, Baasanhu J, et al. Anterior chamber depth measurement as a screening tool for primary angleclosure glaucoma in an East Asian population. Arch Ophthalmol 2000;118:257– 63. 17. Kashiwagi K, Tokunaga T, Iwase A, et al. Usefulness of peripheral anterior chamber depth assessment in glaucoma screening. Eye 2005;19:990 – 4.

1726

18. Nolan W, Aung T, Machin D, et al. Detection of narrow angles and established angle closure in Chinese residents of Singapore: potential screening tests. Am J Ophthalmol 2006; 141:896 –901. 19. Congdon N, Quigley HA, Hung PT, et al. Screening techniques for angle closure glaucoma in rural Taiwan. Acta Ophthalmol Scand 1996;74:113–9. 20. Seah SK, Foster PJ, Chew PT, et al. Incidence of acute primary angle closure glaucoma in Singapore: an island wide survey. Arch Ophthalmol 1997;115:1436 – 40. 21. Drexler W, Findl O, Menapace R, et al. Partial coherence interferometry: a novel approach to biometry in cataract surgery. Am J Ophthalmol 1998;126:524 –34. 22. Santodomingo-Rubido J, Mallen EA, Gilmartin B, Wolffsohn JS. A new non-contact optical device for ocular biometry. Br J Ophthalmol 2002;86:458 – 62. 23. Vogel A, Dick HB, Krummenauer F. Reproducibility of optical biometry using partial coherence interferometry: intraobserver and interobserver reliability. J Cataract Refract Surg 2001;27:1961– 8. 24. Radhakrishnan S, Rollins AM, Roth JE, et al. Real time optical coherence tomography of the anterior segment at 1310 nm. Arch Ophthalmol 2001;119:1179 – 85. 25. Nolan WP, See JL, Chew PT, et al. Detection of primary angle closure using anterior segment optical coherence tomography in Asian eyes. Ophthalmology 2007;114:33–9. 26. Radhakrishnan S, Goldsmith J, Huang D, et al. Comparison of optical coherence tomography and ultrasound biomicroscopy for detection of narrow anterior chamber angles. Arch Ophthalmol 2005;123:1053–9. 27. Kashiwagi K, Kashiwagi F, Toda Y, et al. A newly developed peripheral anterior chamber depth analysis system: principle, accuracy, and reproducibility. Br J Ophthalmol 2004;88: 1030 –5. 28. Kashiwagi K, Kashiwagi F, Hiejima Y, Tsukahara S. Finding cases of angle closure glaucoma in clinic setting using a newly developed instrument. Eye 2006;20:319 –24. 29. Kashiwagi K, Abe K, Tsukahara S. Quantitative evaluation of changes in anterior segment biometry by peripheral laser iridotomy using newly developed scanning peripheral anterior chamber depth analyser. Br J Ophthalmol 2004;88:1036 – 41. 30. Scheie HG. Width and pigmentation of the angle of the anterior chamber: a system of grading by gonioscopy. Arch Ophthalmol 1957;58:510 –2. 31. Foster PJ, Buhrmann RR, Quigley HA, Johnson GJ. The definition and classification of glaucoma in prevalence surveys. Br J Ophthalmol 2002;86:238 – 42. 32. DeLong ER, DeLong DM, Clarke-Pearson DL. Comparing the areas under two or more correlated receiver characteristic operating curves: a nonparametric approach. Biometrics 1988; 44:837– 45. 33. Wilson JM, Jungner YG. Principles and practice of mass screening for disease (in Spanish). Bol Oficina Sanit Panam 1968;65:281–393. 34. Stamper RL. Glaucoma screening. J Glaucoma 1998;7: 149 –50. 35. Aung T, Nolan WP, Machin D, et al. Anterior chamber depth and the risk of primary angle closure in 2 East Asian populations. Arch Ophthalmol 2005;123:527–32. 36. Baskaran M, Oen FT, Chan YH, et al. Comparison of the scanning peripheral anterior chamber depth analyzer and the modified van Herick grading system in the assessment of angle closure. Ophthalmology 2007;114:501– 6. 37. Sakata LM, Lavanya R, Friedman DS, et al. Comparison of gonioscopy and anterior segment ocular coherence tomography

Lavanya et al 䡠 Screening for Angle Closure Using IOLMaster, AS-OCT, and SPAC in detecting angle closure in different quadrants of the anterior chamber angle. Ophthalmology 2007 Oct 2 [Epub ahead of print]. 38. Sakata LM, Lavanya R, Friedman DS, et al. Assessment of the scleral spur in anterior segment optical coherence tomography images. Arch Ophthalmol 2008;126:181–5.

39. Thomas R. Anterior segment optical coherence tomography. Ophthalmology 2007;114:2362–3. 40. He M, Foster PJ, Ge J, et al. Gonioscopy in adult Chinese: the Liwan Eye Study. Invest Ophthalmol Vis Sci 2006;47: 4772–9.

Footnotes and Financial Disclosures Originally received: November 13, 2007. Final revision: February 22, 2008. Accepted: March 14, 2008. Available online: May 16, 2008.

7

Yong Loo Lin School of Medicine, National University of Singapore, Singapore.

Manuscript no. 2007–1464.

1

Singapore National Eye Centre and Singapore Eye Research Institute, Singapore.

2

UCL Institute of Ophthalmology, London, United Kingdom.

3

Wilmer Eye Institute and Johns Hopkins Bloomberg School of Public Health, Baltimore, Maryland.

4

University of Yamanashi, Yamanashi, Japan.

5

Clinical Trials and Epidemiology Research Unit, Singapore.

6

Singhealth Polyclinics, Singapore.

Financial Disclosure(s): Dr Kashiwagi has a Japanese patent on the SPAC (Japanese patent No. 3878164). Dr Friedman has been a paid consultant to Carl Zeiss-Meditec. Dr Foster has received honoraria and travel support from Carl Zeiss Meditec. Dr Aung has received research funding and travel support from Carl Zeiss Meditec. Supported by a grant from Singhealth Foundation, Singapore. The authors’ work was independent of the funders (the funding source had no involvement). These authors (R.L. and P.J.F.) contributed equally to this work. Correspondence: Tin Aung, PhD, FRCS(Ed), Glaucoma Department, Singapore National Eye Centre, 11 Third Hospital Avenue, Singapore 168751. E-mail: tin11@ pacific.net.sg.

1727

Ophthalmology Volume 115, Number 10, October 2008 Table 5. Performance of the Three Tests IOLMaster (ACD <2.87 mm) Patient Group

Prevalence (% of NAs*)

Sensitivity (95% CI)

Specificity (95% CI)

AUC (95% CI)

50–59 y ⱖ60 y ⱖ70 y Male Female Female, ⱖ60 y

14.5 (121/832) 24.7 (301/1220) 27.8 (123/443) 15.7 (152/967) 24.9 (270/1085) 31.1 (183/589)

81.8 (73.8–88.2) 90.0 (86.1–93.2) 91.9 (85.6–96.0) 86.8 (80.4–91.8) 88.1 (83.7–91.8) 90.7 (85.5–94.5)

85.8 (83.0–88.3) 71.4 (68.3–74.3) 63.7 (58.2–69.0) 79.6 (76.7–82.3) 75.7 (72.6–78.6) 68.0 (63.2–72.5)

0.84 (0.80–0.88) 0.81 (0.79–0.83) 0.78 (0.74–0.81) 0.83 (0.80–0.86) 0.82 (0.80–0.84) 0.79 (0.76-0.82)

ACD ⫽ anterior chamber depth; SPAC ⫽ scanning peripheral anterior chamber depth analyzer; AS-OCT ⫽ anterior segment optical coherence *NAs include all subjects with PACS, PAC, and PACG.

1727.e1

Lavanya et al 䡠 Screening for Angle Closure Using IOLMaster, AS-OCT, and SPAC in Different Subgroups of Patients SPAC (Combined Grade P or S and/or <5)

AS-OCT (2 Quadrants Closed)

Sensitivity (95% CI)

Specificity (95% CI)

AUC (95% CI)

Sensitivity (95% CI)

Specificity (95% CI)

AUC (95% CI)

84.3 (76.6–90.3) 96.3 (93.6–98.2) 96.7 (91.9–99.1) 90.1 (84.2–94.4) 94.4 (91.0–96.9) 97.3 (93.7–99.1)

76.8 (73.5–79.9) 59.0 (55.7–62.2) 53.4 (47.8–59.0) 68.0 (64.7–71.2) 65.5 (62.1–68.8) 55.2 (50.2–60.1)

0.81 (0.77–0.84) 0.78 (0.76–0.80) 0.75 (0.72–0.78) 0.79 (0.76–0.82) 0.80 (0.78–0.82) 0.76 (0.74–0.79)

90.9 (84.3–95.4) 87.4 (83.1–90.9) 80.5 (72.4–87.1) 89.5 (83.5–93.9) 87.8 (83.3–91.4) 87.4 (81.7–91.9)

65.5 (61.9–69.0) 60.8 (57.6–64.0) 62.5 (56.9–67.8) 61.7 (58.3–65.1) 64.0 (60.6–67.3) 62.8 (57.9–67.5)

0.78 (0.75–0.81) 0.74 (0.72–0.77) 0.72 (0.67–0.76) 0.76 (0.73–0.79) 0.76 (0.73–0.79) 0.75 (0.72–0.79)

tomography; CI ⫽ confidence interval; AUC ⫽ area under the curve.

1727.e2