advertisement

Editors Selection IGR 15-2

Progression: Combination of event- and trend-analysis of SAP data

Comment by Gustavo de Moraes on:

48936 Integrating event- and trend-based analyses to improve detection of glaucomatous visual field progression, Medeiros FA; Weinreb RN; Moore G et al., Ophthalmology, 2012; 119: 458-467

Find related abstracts

Automated achromatic perimetry (AAP) is widely used in glaucoma to diagnose and monitor disease progression. In addition, it was the main outcome measure in clinical trials investigating treatment efficacy and risk factors for progression. However, there is little consensus among investigators regarding how to define progression using AAP. Although, progression criteria can vary widely they can be classified into two main modalities: event- and trend-based analyses. These two approaches have been traditionally employed separately and each has inherent strengths and limitations. Medeiros et al. described a prospective, longitudinal study combining event and trend analysis to detect glaucoma progression. The goal was to overcome each method's limitations and lack of gold standard to define progression by integrating a-priori information from one method to increase the level of certainty over the results provided by the other test. This statistical method is a practical application of the Bayes' theorem, named after the English mathematician and priest Thomas Bayes. For the present study, Medeiros et al. used a-priori information based on event analysis provided by commercially available software (Guided Progression Analysis, GPA, Humphrey Field Analyzer, Carl-Zeiss Meditec, Inc., Dublin, CA) to improve the accuracy of estimates of rates of progression using trend analysis. The Bayes' estimates combining the two approaches detected a greater number of progressing eyes than each method individually. In addition, eyes progressing only by the Bayesian method had faster rates of visual field decline than those progressing only by ordinary least square (OLS) linear regression ‐ the method provided in the GPA printout using the visual field index (VFI). In practical terms, the Bayesian method described by the authors uses apparently conflicting information from the GPA event- and trend- based methods to improve the likelihood of detecting true glaucoma progression as opposed to test-retest variability. Although the determination of the specificity of a given method to detect glaucoma progression is limited by the lack of a gold-standard, Medeiros et al. analyzed the data of a subset of glaucomatous eyes tested with AAP at short intervals ‐ so that progression is unlikely to have occurred ‐ to compare the false-positive results between the Bayesian and OLS linear regression methods. Both methods had a high specificity of 96%.

Nevertheless, one should be reminded that, by definition, Bayes statistics employs the information of both event and trend analysis from GPA and hence it is expected that this approach has a better agreement with each of these methods taken individually. On the other hand, OLS missed a significant number of eyes progressing at fast rates based on the Bayesian method. This can be explained at least in part by the limitations of OLS due measurement variability which is currently minimized by repeated testing. Since the rates of visual field change using the Bayes approach were faster than OLS, the former may be more accurate in detecting clinically significant progression ‐ as opposed to statistically significant progression - than methods clinicians have traditionally used in practice.

The next step is to test this method in other population samples by different centers. Once cross-validated, it is important to test how the application of integrated event- and trend-based methods can modify the current treatment paradigm, which based on each method individually.

• ← Previous
• Next →