Glaucomatous visual field progression is a clinical research topic that has been evaluated by a number of groups over extended time periods with many different approaches, but no clear procedure has emerged that is universally accepted. A review of visual field progression techniques, ranging from simple to highly sophisticated approaches, has recently been published1 and several investigations have reported that various methods for determining glaucomatous visual field progression only agree with each other on a limited basis.2-5 There are dramatic differences in the sensitivity and specificity of various approaches. In the current paper by Montesano and colleagues, three types of hierarchical Bayesian analysis models have been applied to a large visual field data set and are compared to several methods that have been utilized in the past. The investigation is comprehensive and very thorough by a group of researchers with extensive backgrounds and expertise in clinical and analytic methodologies. The three hierarchical Bayesian models consisted of a linear regression of pointwise measures over time (Hi-linear), a similar model censored at 0 dB (Hi-censored), and a heteroskedastic censored model (Hi-HSK). All three models yielded excellent results and outperformed permutation analysis (PoPLR) and simple linear regression. A most noteworthy aspect of the study is the extensive assessment of data censoring approaches for sensitivity values of 0 dB or near 0 dB. Because there is an upper intensity limit that automated perimeters have for stimulus presentations, it is not clear how the data should be censored and earlier studies have mostly avoided this issue. It is encouraging that the authors have undertaken this challenge. Although alternative light sources may allow automated perimeters to achieve higher intensities, this may create undesirable effects due to scattered light and possible transient adaptation changes.
Although alternative light sources may allow automated perimeters to achieve higher intensities, this may create undesirable effects due to scattered light and possible transient adaptation changes
There appear to be two limitations associated with this investigation: (1) the majority of cases evaluated are representative of early to moderate visual field loss (or normal visual field results), and (2) implementation of these procedure on automated perimeters may be difficult because the computations are time-consuming. In spite of this, the investigation by Montesanto and colleagues represents a meaningful contribution to the glaucomatous visual field progression literature and should be an incentive to other investigators to continue to explore this topic.