Purchase this article with an account.
elizabeth tsianta, Anthony G Robson, Angela Rees, Graham E Holder, Carlos Pavesio, Mark Westcott; Linear regression analysis of electrodiagnostic parameters against time in patients with Birdshot Retinochoroidopathy (BRC). Invest. Ophthalmol. Vis. Sci. 2016;57(12):4503.
Download citation file:
© ARVO (1962-2015); The Authors (2016-present)
To evaluate the usefulness of univariate linear regression analysis of serial electrodiagnostic parameters against time in patients with Birdshot Retinochoroidopathy (BRC)
Twenty three patients with BRC were ascertained (11 Males and 12 females) with a mean age of 48 years (range 41 -78 years). Patients underwent pattern ERG and full-field ERG testing incorporating the International standard. The mean number of ERG test sessions per patient was 14 (range 7–24) recorded over a mean follow-up period of 90 months (range 41-171 months). The PERG P50 and multiple ERG parameters were assessed for each patient by linear regression with least squares analysis, using time as the independent variable. The level of statistical significance for the gradient was set at p < 0.05. PERG and ERG parameters that showed significant fit were reviewed and predicted values at baseline and final visit compared, to gauge if the change was clinically relevant in that patient. The predicted PERG P50 or ERG amplitude change was considered significant if greater than 30% of the predicted baseline amplitude (DA10.0 ERG a- & b-waves; LA3.0 30Hz ERG) or if peak time change was 3ms or greater (LA3.0 30Hz ERG).
In conclusion 1 to 5 electrophysiological parameters in each of nineteen patients showed a significant gradient (p<0.05) when compared against time (table 1). Pattern ERG normalised in 1 and became abnormal in 1 other case with worsening macular function. DA10.0 a-waves became severely abnormal in 1 eye but were stable in most; b-waves showed significant decline and were subnormal in 3 eyes at final follow up; 1 eye showed marked improvement. Photopic 30Hz ERG peak times were abnormal in all cases at baseline and showed improvement in 15 eyes (3ms – 8ms), normalising in all but 1 case.
Successful linear fit varied according to parameter, and was obtained in a minority of patients. Both improvement and worsening were detected. Linear regression analysis can be a useful method of identifying clinically relevant change in EDD parameters in some patients. The technique could also be used for forecasting.However, LRA failed as a fit in a significant proportion of Birdshot patients. These series are either stationary, or if changing might be better analysed with event based analysis, or with models other than simple linear regression.
This is an abstract that was submitted for the 2016 ARVO Annual Meeting, held in Seattle, Wash., May 1-5, 2016.
This PDF is available to Subscribers Only