Statistics from Altmetric.com
I refer to the interesting article1 by Wedner et al published recently in the BJO. The authors have concluded, without enough justification, that “an eye screening programme for significant refractive errors is indicated in this population.” I believe more data are required for such a conclusion.
Wedner et al could have reported the true positive predictive rate for the detection of myopia using the stated referral criterion of “worse than 6/12” on the Snellen visual acuity chart: 141/174 or 81% (95%CI: 74 to 86). However, they have referred to the prevalence of “significant” myopia being 141/2511 or 5.6%. In epidemiological convention it is incorrect to apply this denominator and report this prevalence rate because the reference standard for the determination of refractive errors—that is, refraction, has not been performed in the majority (2337) of the 2511 students. The positive predictive value of a screening test is influenced by the prevalence of the condition to be screened (that is, myopia). Nevertheless, without experimental data on the false negative subjects, the actual prevalence cannot be determined. Therefore, true prevalence of myopia cannot be determined given the study design. In a study conducted in Singapore,2 for example, each one of the 1003 children in the sample has been subjected to refraction as well as the screening test (visual acuity). With a cut-off criterion of logMAR 0.30, the positive predictive value was as high as 95.4% (95%CI: 92.6 to 97.2), but the actual prevalence of myopia was only 33.8% (95%CI: 30.9 to 36.8).2 At least one of the studies cited3 by Wedner et al has employed refraction in the entire study sample and this may very well be a reason for the previously reported relatively higher myopia prevalence rates (in the true sense of the word).
In the discussion section of the paper,1 it was assumed without justification, that only the “lower degrees of hyperopia” would be undetected by their visual acuity test. In fact, some proportion of all types of refractive errors could escape the best compromise threshold on visual acuity testing.2 If a subject could pass the visual acuity criterion on one occasion, one should not assume that he is not myopic or even, to have such low myopia that he is functionally normal.4