Glaucoma Progression Should Be Assessed in Terms of Both Structure and Function

Lexa W Lee

November 14, 2007

November 14, 2007 (New Orleans) -- Glaucoma progression should be assessed both structurally and functionally because there is poor agreement among current measurement methods, according to a study presented here at the American Academy of Ophthalmology 2007 Annual Meeting. This finding supports previous studies showing that eyes develop different structural and functional progression endpoints, according to lead author Gadi Wollstein, MD, assistant professor of ophthalmology, University of Pittsburgh School of Medicine, Pennsylvania.

"We wanted to compare conventional methods for glaucoma progression detection, specifically visual field [VF], optical coherence tomography [OCT], and Heidelberg retina tomography [HRT]," said Dr. Wollstein during the presentation. Comparison methods were based on gold standard definition (using separate measurements as the gold standard, as there are no widely accepted gold standard progression criteria) and latent class modeling (extracting shared properties among different parameters used to measure the same feature).

Fifty-seven eyes from 30 subjects with at least 5 reliable VF scans and 5 good-quality OCT and HRT scans were enrolled from a glaucoma service. Tests were performed within a 6-month window. The tests involved were Humphrey field analyzer full-threshold and Swedish Interactive Threshold Algorithm standard 24-2 perimetries(Carl Zeiss Meditec) for VF; fast retinal nerve fiber layer (RNFL) circumpapillary scan of Stratus OCT (Carl Zeiss Meditec); and HRT 3 (Heidelberg Engineering) optic nerve head scans.

The progression criterion for VF subjective analysis (SA) was the majority opinion of 3 glaucoma experts independently assessing VF printouts without clinical information; for VF glaucoma progression analysis (GPA), the progression criterion was the Early Manifest Glaucoma Treatment study criterion (3 progressive points in 3 consecutive visits); for OCT event analysis (OCT-e), the progression criterion was the change in RNFL thickness that exceeded reproducibility error; for OCT trend analysis (OCT-t), the progression criterion was that the RNFL thickness slope significantly exceeded the expected rate of progression in the normal population; for HRT topographic change analysis (TCA), the progression criterion was a cluster of 20 adjacent points with more than 95% probability of progression; and for HRT glaucoma probability score (GPS), the progression criterion was the GPS value (continuous parameter).

Using each parameter as the gold standard definition of progression, the specificity and sensitivity of all the other parameters were determined. For HRT-GPS, the area under the receiver characteristics curve was calculated. Latent class regression models were used to test the multiway frequency table of the binary progression indicators while adjusting for age, sex, and race. There were 37 women and 20 men enrolled into the study; 45 were white and 12 black. The mean age at baseline was 59.5 ± 10.2 years, and the mean follow-up period was 2.6 ± 0.8 years. None of the covariates was statistically significant.

Five eyes progressed as measured by VF-GPA, 18 by VF-SA, 19 by OCT-e, 19 by OCT-t, and 32 by HRT-TCA.

The specificity and sensitivity of the various parameters using the latent class analysis were as follows: VF-SA criteria: specificity was 83.3%, sensitivity was 38.5%; VF-GPA: specificity, 99.9%, sensitivity, 12.8%; OCT-e: specificity, 61.1%, sensitivity, 30.8%; OCT-t: specificity, 44.4%, sensitivity, 23.1%; and HRT-TCA: specificity, 44.4%, sensitivity, 56.4%. The intercept and slope for HRT-GPS were 0.19 and 0.00001 per day, respectively, for the stable eyes and 0.75 and 0.00002 per day for the progressing eyes.

The data showed that although specificity was moderate to high, using each of the parameters as the gold standard, the sensitivity was low. HRT-TCA showed slightly higher sensitivity with markedly lower specificity. This poor agreement was evident both between structure and function methods and within structural assessment methods. Latent class analysis showed poor agreement with all tested parameters

"This paper attempts to shed some light on a dilemma in glaucoma analysis: the definition of progression of disease and how to manage it," said Richard Bensinger, MD, an ophthalmologist from Seattle, Washington. "The manufacturers of these devices have accumulated statistics to analyze whether changes have occurred in a given patient with sequential scans over time. This study attempts to correlate changes in the visual fields to determine whether they are similar to the anatomical studies in degree and position.

"It is already agreed that very significant changes do cross over in these diagnostic studies. This paper shows that there is insignificant correlation when more subtle changes are sought," Dr. Bensinger pointed out. "The conclusion is that these tests have a way to go, as very modest changes cannot be detected in a reliable manner at this point in time. This is important, in that therapeutic decisions are made in part based on the assessment of changes that have been detected, and there is obviously some randomness in the data that prevents the cross correlation of these studies in subtly detected changes."

The study was supported in part by National Institutes of Health, the Eye and Ear Foundation (Pittsburgh, Pennsylvania), and an unrestricted grant from Research to Prevent Blindness (New York City). Dr. Wollstein received research funding from Carl Zeiss Meditec. Dr. Bensinger has disclosed no relevant financial relationships.

American Academy of Ophthalmology 2007 Annual Meeting: Scientific Poster 074. Presented November 11, 2007.