Purchase this article with an account.
M. Monhart, F. Dannheim; Evaluation of a Non-Linear Trend Analysis. Invest. Ophthalmol. Vis. Sci. 2010;51(13):5499.
Download citation file:
© ARVO (1962-2015); The Authors (2016-present)
The international Glaucoma Societies suggest to calculate the rate of progression based on a minimum of 6 examinations. Longer series though improve the accuracy of the calculation for the rate of progression. The goal of this retrospective study is to evaluate objective criteria for identifying more relevant current trends.
105 series of the initial 10 VF (Octopus G-pattern, TOP strategy) from 105 patients of the co-authors private practice were randomly selected. The average time span from the first to the last examination was 5,6 (±1,2) years. The first examination was excluded from analysis due to learning. VF tests 2..10 were analyzed with linear regression analysis and with a 2-segment (non-linear) regression analysis (see graph). The approximation consists of two connected linear segments, each comprising at least three data points, and satisfying minimum sum of squares of residuals. The non-linear trend analysis was considered to differ significantly from the linear trend analysis if a) the rate of progression of segment B differed more than 0,5dB/year from the linear regression rate and b) as compared to linear regression, the two-segment fit reduced the residuals (rms) to 70% or less. The calculations were performed using the OFA (Octopus Field Analysis) software version 2.55.
According to the above criteria 26 of the 105 series (25%) differed from the linear trend analysis. In 11 of the 26 series the linear regression analysis underestimated the rate of progression. In 4 series a trend was overestimated or recovery missed. 7 series showed a prolonged initial phase of learning or recovery. This might later lead to an underestimation in linear trend analysis. The remaining 4 series can be considered false positives as they only showed increased long term fluctuation without significant trends in either analysis.
The 2-segment approximation may be of help to identify non-linear trends. A recalculation of serial visual fields allows a clinically more relevant estimation of current trends.
This PDF is available to Subscribers Only