［摘要］目的:比较Frisby、Randot、Howard-Dolman三种立体视检查方法的可重复性，确定此三种临床用立体视检查方法的正常值变化的95%可信区间。方法:选择26位无斜视、弱视和眼手术史的健康个体，平均年龄为24.42±4.26岁（14～32岁），所有的受检者具有正常的双眼视且双眼视力为6/6。测量视远瞳距和习惯视力后，以随机的顺序用Frisby、Randot、Howard-Dolman三种检查方法测量立体视，测量时保持光照度为670lx。约1周后在相同的测量条件下同样用三种检查方法测量立体视，平均间隔时间为7.8±2.4天。 结果:三种临床立体视检查方法的可重复性系数分别为±2.7″(Frisby)、±8.0″（Randot）和±9.3″(Howard-Dolman)，再次试验的平均值低于首次试验的平均值，t检验提示这改变差异无显著性意义（P＞0.05）。重复试验的相关系数分别为0.91（Frisby）、0.56（Randot）、0.60（Howard-Dolman）。三种检查方法之间的相关系数分别为0.31、0.31、0.33。结论：三种临床立体视检查方法的相关性较差。Frisby检查法有最好的重复性和正常值有较小的范围。用该法测得成年人立体视改变的95%可信区间为±2.7″。如果用Frisby法测得年轻成年人的立体视的变化超过这个范围，可认为是有显著意义的临床改变。

［中图分类号］R778.3［文献标识码］A

［文章编号］1008-1801(2000)04-0216-03

The detection of change in stereoacuity

Ping Situ

(Centre for Contact Lens Research,School of Optometry, University of Waterloo,Waterloo, Ontario, Canada, N2L 3G1）

David Elliott

（Department of Optometry,University of Bradford,Bradford,BD7 1DP,UK.)

Abstract：Objective：To compare the repeatability of the Frisby, Randot and Howard-Dolman stereotests, and to determine the 95% confidence limits for the change of each test.Methods：Stereoacuity using the three clinical stereoacuity tests was measured in 26 healthy adult subjects with normal binocular vision. The subjects were retested approximately one week later.Results：The coefficients of repeatability for the three tests were ±2.7″ (Frisby) ±8.0″ (Randot) and ±9.3″ (Howard-Dolman) and test-retest correlation coefficients were 0.91 (Frisby), 0.56 (Randot) and 0.60 (Howard-Dolman).Conclusion：The Frisby test showed the best repeatability and smallest range of normal values. For young adults, the 95% confidence limits for change were calculated to be ± 2.7″. If a young adult's stereoacuity measured with the Frisby test changes by more than this amount, then this should be considered a significant clinical change.

Key words：stereoacuity; stereoacuity tests/methods; test-retest repeatability

［CLC number］R778［Document code］A

［Article ID］1008-1801(2000)04-0216-03

Introduction

Measurements of stereoacuity can be used clinically to determine the presence of binocular single vision and how well developed or established it is, and to monitor possible change with treatment or time. The detection of change in stereoacuity or deviation from normality is clinically important, as it is liable to influence the likely management of a particular condition. Changes in score on a stereoacuity test (or any other test) can occur due to chance and despite no real change having occurred. For a clinician to decide that a real change has occurred, confidence limits for change need to be determined for a particular test^［1,2］. These confidence limits can be determined as the outer limits within which the vast majority (95%) of ‘normal’ subjects will change from test to retest^［1,2］.These limits are influenced by the inherent variability of a test, in that tests which are inherently variable are likely to provide larger confidence limits for change^［3］. Such tests will be less capable to detect changes in score or deviations from normality^［3］. The purpose of this study was to determine the confidence limits for change for three commonly used clinical stereoacuity tests: the Howard-Dolman, Frisby, and Randot stereotests.

Although several papers have assessed stereoacuity tests in terms of their screening ability^［4～9］, very few have determined their ability to detect change in stereopsis. Frisby and colleagues determined correlation coefficients between test and re-test for the Frisby, TNO and Titmus circles stereotests^［10］. However, they did not provide information regarding the confidence limits for change for the three tests. In addition, correlation coefficients, although frequently used to assess repeatability, only give a measure of association but not necessarily of agreement^［11］. Moreover, they are dependent upon the range values in the sample, as well as their association. A more appropriate indicator of repeatability is gained from the coefficient of repeatability (COR)^［11］. The COR describes the 95% confidence limits for any discrepancy between test and retest data. These confidence limits also provide information for longitudinal assessment of function so the significance of any change in performance can be assessed^［1,2］.

1Materials and methods

1.1Frisby stereotest

This is a real depth test consisting of a series of three perspex plates of varying thickness (6mm,3mm,1.5mm). Each plate has four 6mm squares of random dot patterns printed on one side. One of the squares contains a small central circular area of dots on the back of the plate. Provided the patient's head and test plate are parallel and do not move during testing, so the disparity produced by the circular patch can only be detected using stereoacuity. At a 40cm testing distance, the disparities of the plates are 340″, 170″ and 85″ for a PD of 64 mm. The amount of stereopsis presented to a subject can be altered by changing the viewing distance. Stereoacuity was determined using a bracketing technique with the plate being presented in a number of random positions and different distances. Threshold was recorded as the largest distance at which correct responses were given in at least 3 out of four presentations.

1.2Randot test

The Randot test with random dot stereograms in vectograph form views using polaroid glasses. It consists of three parts, of which the first two parts assess gross stereopsis only and were not used. Contoured circles at ten levels of crossed disparity provide a finely graded sequence for critical testing. The circles can be presented in uncrossed disparity by turning the booklet over. The range of disparity provided is from 400″ to 20″ when used at 40cm. Threshold was recorded as the largest distance where the last level of stereopsis was chosen correctly.

1.3Howard-Dolman test

This is another real-depth test. The version used consisted of one stationary rod and one moveable rod, which were viewed through a rectangular aperture against an evenly illuminated white background at a testing distance of 3 meters. Subjects were required to align the moved rod with the stationary rod. The offset from alignment in seconds of arc of retinal disparity was calculated. The process was repeated six times and the average was taken.

1.4Subjects

26 healthy subjects who had no history of strabismus, amblyopia or ocular surgery participated in the study. The mean age (±1SD) was 24.42±4.26 years (range from 14-32 years). All subjects had binocular visual acuity 6/6.

1.5Procedure

After informed consent<