Abstract
Purpose :
To assess whether there is a bias towards certain visual acuity (VA) values when using an EDTRS logMAR chart and intraocular pressure (IOP) measurement using Goldmann Applanation Tonometry (GAT).
Methods :
This study involved VA and IOP data from 27 secondary care providers in the UK using the medisoft electronic medical record (EMR) platform and from the UK BioBank Eye and Vision Consortium (UKBB). Trial data was used from the ABC and LIGHT trial. LogMAR VA values were compared per line, from letter position 1-5 using the remainder modulus (RM) and IOP devices including GAT, iCARE, Tonopen and Airpuff were analysed in the EMR data, as well GAT IOP data from the LIGHT trial and ocular response analyser (ORA) data from the UKBB. Chi-squared tests comparing the frequency of RM 1-5 for LogMAR VA, as well for IOP for all devices was performed for all the data.
Results :
There was significantly unequal distribution between the letter positions, which favoured RM 5 in our EMR data (1=209295, 2=10368, 3=113341, 4=122606, 5=425664,X2(4,N=974,592)= 377889.8, p< 0.001). The UKBB also confirmed a significantly unequal distribution of values per RM on the LogMAR chart (1=8759, 2=14000, 3=18014, 4=11641, 5=13472,X2(1,N=65,886)=14292,p<0.001). Comparatively the ABC trial did not show any significant difference in the frequency of RM positions (1=255, 2=277,3=314, 4=269, 5=285, X2(4,N=1400)= 6.941,p=0.141). EMR data showed a significantly unequal distribution in even vs. odd IOP values for GAT compared to other IOP measuring devices (even/odd, GAT 83129/489324, iCARE 307929/288118, Tonopen 262890/255362, Airpuff 8776/66882,X2(3,N=2,510,935)=96863.010,p<0.001). Similarly, the LIGHT trial showed there was significantly unequal distribution for IOP values when using GAT (even/odd, 6433/4642,X2(1,N=718)=292.61,p<0.001). UKBB showed a small but statistically significant unequal distribution between values using ORA (even/odd, 32288/31731, X2(1,N=64,109)=4.865,p=0.028).
Conclusions :
LogMAR VA had a significantly unequal distribution and GAT IOP values were significantly more likely to be even in our EMR data set. Theses findings were backed up by UKBB and trial data. This rounding of LogMAR VA and GAT IOP could be due to examiner bias. Outcomes of this study may have consequences in clinical practice and therapeutic trials may require greater scrutiny.
This is a 2021 ARVO Annual Meeting abstract.