advertisement
Editors Selection IGR 14-1
Progression: Combination of imaging and perimetric data to better assess visual field change
Comment by William Swanson on:
49288 Improved Estimates of Visual Field Progression Using Bayesian Linear Regression to Integrate Structural Information in Patients with Ocular Hypertension, Russell RA; Malik R; Chauhan BC et al., Investigative Ophthalmology and Visual Science, 2012; 53: 2760-2769
One of the key needs in clinical care and clinical trials is improved ability to assess rate of progression of visual field loss in patients with glaucoma. This study by Russell et al. applied insights from the past decade of research on relations between ganglion cell damage and perimetric defects, and developed a Bayesian method for combining imaging and perimetric data to assess change in visual field during a prospective nine-year longitudinal study of patients with ocular hypertension. The rate of progression was characterized as the slope of a regression line over time, for both ordinary and Bayesian methods. The authors derived a relation between perimetric sensitivity and area of neuroretinal rim in cross-sectional data, and used this to compare estimates of rate of change for perimetry and rim area in longitudinal data. The two rate estimates were combined to give a Bayesian estimate of rate, where relative weight of the imaging data depends on the relative variability in the two datasets. For instance, when perimetry alone shows a learning effect, the Bayesian rate estimate does not. The Bayesian rate and the rate for perimetry alone were each used to predict perimetric sensitivity at the next clinic visit. When the total number of visits was less than seven, the Bayesian rate was on average a better predictor than the rate from perimetry alone. For more than seven visits the Bayesian rate became a worse predictor than perimetry alone. The authors carefully discuss this and the other limitations of their method, and the need for future research. Their study establishes the validity of their basic approach, and shows that more work is needed before clinicians will be able to use a Bayesian method to assess stability of the visual field in routine patient care. This is a valuable and careful step forward.