advertisement

Editors Selection IGR 20-3

Screening and Detection: Two birds with one stone? Piggy-backing on Diabetic Retinopathy Screening

Comment by Benton Chuter & Linda Zangwill on:

112698 Evaluating the outcome of screening for glaucoma using colour fundus photography-based referral criteria in a teleophthalmology screening programme for diabetic retinopathy, Tan R; Teo KYC; Husain R et al., British Journal of Ophthalmology, 2023; 0:

Find related abstracts

Establishing optimal threshold values for a glaucoma screening test is crucial to balance the costs of potential missed diagnoses and associated increased disease burden with that of overdiagnosis and overutilization of limited clinical resources. A lower threshold misses fewer cases of glaucoma and may help prevent irreversible vision loss, but leads to a higher number of false positives that in clinical practice can impose high resource utilization costs.¹

This prospective study illustrates this tradeoff in the context of a real-world opportunistic detection of glaucoma using color fundus photographs (CFPs) acquired through Singapore's Integrated Diabetic Retinopathy Programme (SiDRP) to ascribe glaucoma suspect status. Specifically, vertical cup-to-disk ratio (VCDR) values (determined by the Singapore Optic Disc Assessment (SODA) tool) with and without other glaucoma features identified in the CFPs were used to identify glaucoma suspects.

The study included 5023 diabetic patients participating in SiDRP between 2017 and 2018, with 2625 with glaucoma suspects (GS) identified by the CFPs. Sensitivity, specificity, and positive predictive value (PPV) of the screening criteria were evaluated using electronic medical record (EMR) diagnosis codes from the follow-up visit used to establish ground truth (GT). Using a VCDR of ≥ 0.65 alone, the sensitivity was 81.6%, specificity 50.6%, and PPV 14.0%. Increasing the VCDR threshold to ≥ 0.80 alone improved specificity to 93.9% and PPV to 15.4%, but dramatically reduced sensitivity to 11.3%. Adding glaucoma features to the VCDR cut-off of ≥ 0.65 and ≥ 0.80 alone resulted in a modest increase in sensitivity and reduction in specificity.

The results highlight the challenges of using VCDR in glaucoma screening

This study has several strengths. It is one of the largest community-based opportunistic telemedicine glaucoma screening in diabetic patients in the literature. In addition, the study includes over three years of follow-up of the referred glaucoma patients, providing sufficient time to determine whether the patient has glaucoma. Most importantly, the results highlight the challenges of using VCDR in glaucoma screening and provide quantitative information on the resulting trade-offs in sensitivity, specificity and PPV when different VCDR cut-offs are used.

Several limitations were highlighted by the authors including questions regarding the generalizability of the results as the study diabetic screening population was predominantly Asian, with less than 4% of the study population is not Chinese, Indian, or Malay. The authors also mention that VCDR may not be the most appropriate metric for glaucoma detection. The reported performance in glaucoma detection is somewhat lower than that from other recent studies using alternate CDR and estimation and other glaucoma detection techniques, potentially also due to different study populations.^2-7 Other possible metrics such as cupto- disc area ratios (aCDR) have been demonstrated to be less prone to image orientation or localized defects.²

Overall, this study provides important insight regarding appropriate VCDR threshold values for use in glaucoma screening, which are crucial to implementation of effective screening practices.

• ← Previous
• Next →