Influence of missed catch trials on the visual field in normal subjects

Improving predictive models of glaucoma severity by incorporating quality indicators

OBJECTIVE

In this paper we present an evaluation of the role of reliability indicators in glaucoma severity prediction. In particular, we investigate whether it is possible to extract useful information from tests that would be normally discarded because they are considered unreliable.

METHODS

We set up a predictive modelling framework to predict glaucoma severity from visual field (VF) tests sensitivities in different reliability scenarios. Three quality indicators were considered in this study: false positives rate, false negatives rate and fixation losses. Glaucoma severity was evaluated by considering a 3-levels version of the Advanced Glaucoma Intervention Study scoring metric. A bootstrapping and class balancing technique was designed to overcome problems related to small sample size and unbalanced classes. As a classification model we selected Naïve Bayes. We also evaluated Bayesian networks to understand the relationships between the different anatomical sectors on the VF map.

RESULTS

The methods were tested on a data set of 28,778 VF tests collected at Moorfields Eye Hospital between 1986 and 2010. Applying Friedman test followed by the post hoc Tukey's honestly significant difference test, we observed that the classifiers trained on any kind of test, regardless of its reliability, showed comparable performance with respect to the classifier trained only considering totally reliable tests (p-value>0.01). Moreover, we showed that different quality indicators gave different effects on prediction results. Training classifiers using tests that exceeded the fixation losses threshold did not have a deteriorating impact on classification results (p-value>0.01). On the contrary, using only tests that fail to comply with the constraint on false negatives significantly decreased the accuracy of the results (p-value<0.01). Meaningful patterns related to glaucoma evolution were also extracted.

CONCLUSIONS

Results showed that classification modelling is not negatively affected by the inclusion of less reliable tests in the training process. This means that less reliable tests do not subtract useful information from a model trained using only completely reliable data. Future work will be devoted to exploring new quantitative thresholds to ensure high quality testing and low re-test rates. This could assist doctors in tuning patient follow-up and therapeutic plans, possibly slowing down disease progression.

Introduction of the Low Vision Evaluator (LoVE) for children

PURPOSE

To determine whether the Low Vision Evaluator (LoVE) can grade the visual acuity of young children with light perception and hand movement acuity into finer acuity steps and at what age reliable measurements can be obtained.

METHODS

Two hundred twenty children were tested with the LoVE. Each eye was tested separately, and each stimulus magnitude (intensity x duration) was presented three times. Three catch trials per eye also were presented.

RESULTS

Scores ranged from -8 to -1 on variable-duration tests and from 22.5 to 37.5 dB on fixed-duration tests for four children with hand movement vision. Scores ranged from -12 to 0 on variable-duration tests and from 12.5 to 40 dB on fixed-duration tests for five children with light perception vision. Reliable measurements were obtained at different times on different days. Mean scores for children with counting finger vision or better were significantly better than scores for eyes with light perception and hand movement (P < .001 and P < .01, respectively). Reliability was less for children younger than age 4 years.

CONCLUSIONS

The LoVE is capable of grading the visual function of children with light perception and hand movement vision into finer steps. Reliable measurements can be obtained for children age 4 years and older.

The influence of patient reliability on visual field outcome

The reliability of subjects to perform to perimetry correctly should be carefully evaluated to interpret visual field examinations adequately. Clinicians generally agree that numerous false-positive responses to catch trials cause measured thresholds to be falsely high and numerous false-negative responses cause measured thresholds to be falsely low. We studied the effect of false-positive and false-negative responses on the outcome of visual field measurements. Of 47 eyes, the results of 106 stable glaucomatous visual field tests (Program G1, Octopus 201, Interzeag, Schlieren, Switzerland) with false-positive responses and no more than one false-negative response to catch trials were compared to the results of reliable visual field tests (no false-positive and no more than one false-negative response) performed on the same eye. Similarly, 60 stable visual fields with false-negative responses and no more than one false-positive response were used to study the effect of false-negative responses on visual field sensitivities. Linear regression analysis disclosed a mean sensitivity increase of 1.5 dB for every 10% of false-positive responses (r = .34, P = .000) and a mean sensitivity decrease of 1.2 dB for every 10% of false-negative responses (r = .26, P = .04). These results may be used to help reduce the magnitude of unexplained long-term fluctuation in visual field interpretation.

Threshold variation in automated perimetry

A physiologic fluctuation in threshold levels exists in automated perimetry, which is greater in glaucoma and ocular hypertensive patients than in normal subjects. Fluctuation increases with greater eccentricity from fixation and in areas of reduced retinal sensitivity. Other factors related to automated perimetry testing potentially may influence threshold fluctuation including: a learning effect, reliability, pupil size, age, and the mode of stimulus presentation. Statistical software may aid in analyzing both the single field and the changes between successive fields. However, statistical programs cannot replace physician judgment, and all factors that may influence threshold levels should be considered when interpreting automated visual fields.

Association of reliability with reproducibility of the glaucomatous visual field

We studied 207 glaucoma patients (207 eyes) who had two successive visual tests on the Humphrey Field Analyzer to determine the association of unreliability (> 10% incidence of a missed catch trial test) with threshold variation. We found that in patients where one visual field showed > 30% false negative or positive errors, and the other visual field also was unreliable, a significant increase in the variance of the mean defect existed from reliable patients (p < 0.05). Of patients who had only one of two visual fields which was unreliable (from false negative errors), only those with marked glaucomatous visual field loss (< -15 dB depth of defect) showed a greater variance in mean defect between examinations over reliable patients (p < 0.05). In total, 41 patients (19.8%) had unreliable visual fields associated with statistically increased threshold variation. This study suggests that patient reliability should be considered when interpreting changes in threshold between automated visual field examinations.