Evaluation of diabetic retinopathy severity on ultrawide field colour images compared with ultrawide fluorescein angiograms

Abstract

Purpose

To compare Early Treatment Diabetic Retinopathy Study (ETDRS) diabetic retinopathy (DR) severity on ultrawide field (UWF) colour imaging (CI) and UWF fluorescein angiography (FA).

Design

Cross-sectional retrospective review.

Subjects

Patients with diabetes mellitus and at least mild non-proliferative DR on UWF-CI.

Methods

UWF-CI and UWF-FA images acquired within 1 month of each other were evaluated independently using ETDRS DR Severity Scale (DRSS) for colour photography adapted for UWF-CI and UWF-FA. Extent of non-perfusion (NP, mm²) was determined from UWF-FA images.

Main outcome measures

Agreement rate between DRSS on UWF-CI and UWF-FA.

Results

Images from 218 eyes of 137 patients with diabetes were evaluated. Agreement rate for DRSS between UWF-CI and UWF-FA was moderate to substantial (K = 0.46, Kw = 0.65). Over-all, DRSS was worse in 73 (33.5%) eyes on UWF-FA and in 16 (7.3%) on UWF-CI. Compared to UWF-CI, UWF-FA identified more severe DRSS in 26.5% (1 step) and 7.34% (≥2 steps) of eyes. DRSS was worse than UWF-FA in 56 (51.4%) in early DR (ETDRS levels 20–47, N = 109) and 17 (15.6%) in eyes with severe DR (53 and higher, N = 109). In this cohort, the extent of NP significantly increased as eyes approach moderate non-proliferative DR (levels 43–47, p = 0.0065).

Conclusion

When evaluating UWF-FA images using the ETDRS colour severity scale, DRSS is graded as more severe in a substantial number of eyes than when evaluating UWF-CI. It is uncertain how the DRSS levels using UWF-FA translate to clinical outcomes, but the additional lesions detected might provide added prognostic value. These and other recent data emphasise the need of obtaining outcome data based on UWF-FA and the potential need to develop DRSS specifically tailored for UWF-FA images.

INTRODUCTION

Early identification of diabetic retinopathy (DR) lesions and accurate assessment of DR severity is critical for the care and treatment of patients with diabetes. The modified Airlie House classification of DR for grading colour fundus photographs as adapted for use in the Early Treatment Diabetic Retinopathy Study (ETDRS) has been the standard for the classification of DR severity and has guided the care of patients in both clinical and research settings.¹ With advances in retinal imaging and improved systemic control of diabetes, the identification of retinal markers for DR progression at early stages of DR has become more important. Landmark multicentre clinical trials such as the Diabetes Control and Complications Trial and ETDRS compared fluorescein angiography 9(FA), ETDRS FA fields 1 and 2, online supplemental figure 1) to ETDRS seven-field stereoscopic fundus photography and identified additional lesions and risk factors on FA.^{2 3} However, at the time, the studies were conducted, only a limited area of the posterior retina was evaluated and the studies concluded that the additional lesions and risk factors present on FA did not provide enough additional benefit in the management of the patients to be recommended routinely. Other subsequent studies have compared between colour imaging (CI) and FA, however, they focused on individual DR lesions and not DR severity.^{4 5}

Although the potential risks of FA are low, severe adverse events including death may occur in approximately 1 of every 200 000 patients.⁶ The current 2019 American Academy of Ophthalmology preferred practice patterns recommends FA mainly to guide focal laser treatment for diabetic macular oedema (DME) and to identify suspected but clinically undetectable neovascularisation (NV). Current preferred practice patterns do not recommend FA to screen for DR or to assess its severity level.

Furthermore, there is no widely adopted FA grading scale for DR severity that has been validated in multicentre prospective trials. Nevertheless, there is significant use of FA in real-world practice, even beyond current practice patterns and recommendations. McCarty et al reported the results of a survey examining practice patterns in the treatment of DR in Australia which noted that 10% almost always, 16% often and 10% would sometimes order FAs in patients with mild non-proliferative DR (NPDR), intact vision and no DME.⁷ Over 50% would always order an FA in patients with severe NPDR determined by exam. A recent survey regarding practice patterns in Egypt demonstrated that over 70% of respondents used FA to grade and base treatment decisions for DR either alone or in conjunction with CI.⁸ In these real-world settings, DR severity levels specifically designed for colour fundus photography such as the ETDRS or the more clinical International Classification of DR (ICDR) scale have been adapted when evaluating FA images.

Ultrawide field (UWF) imaging visualises nearly threefold more retinal area (~80%) than obtained with ETDRS photography. The UWF images include peripheral retina which may experience changes early in diabetes. Retinal diabetic lesions beyond the ETDRS seven fields have been shown on to be associated with increased non-perfusion (NP) and risk of retinopathy progression.^{9 10} Evaluation of these additional regions on UWF-FA might provide substantial disease prognosis and clinical benefit to better identify eyes at high risk for DR progression.^{9 11 12} Our prior work has shown that UWF-FA allows precise quantification of retinal NP¹² and identifies 3–5 fold more haemorrhages and/or microaneurysms (H/Ma) as compared with UWF CI (UWF-CI).⁴ In addition to H/Ma, intraretinal microvascular abnormalities (IRMA), venous beading (VB) and NV are all used in the ETDRS colour grading scale to assess severity and the risk for DR progression, development of PDR and risk of visual loss. Greater retinal NP measured on UWF-FA is associated with increasing DR severity.¹²

Currently, it is unknown how the grading of DR severity on CI applies to DR severity on FA, or whether FA grading might provide the same or more accurate disease stratification.^{13 14} Thus, this study compared DR evaluation of UWF-CI and UWF-FA images using ETDRS colour grading scales and determined agreement rates between both imaging modalities across a broad range of DR severity levels. Additional analysis was performed to determine which modality correlated more closely with retinal NP, an objective reproducible metric for quantifying disease severity.

METHODS

This was a cross-sectional retrospective chart review. The study included patients over 18 years old, with type 1 or type 2 diabetes imaged by both 200° UWF-CI and UWF-FA within 1 month of each other as part of routine care at the Beetham Eye Institute of the Joslin Diabetes Center from 21 March 2012 to 1 December 2018. Of the 218 eyes included in the current study, 210/218 eyes (96%) had both CI and FA on the same day while the remaining 8 eyes had both imaging modalities within 1 week of each other.

Eyes included for analysis had adequate image quality for grading ETDRS DR Severity Scale (DRSS) and no evidence of panretinal photocoagulation (PRP) or other significant non-diabetic retinal diseases.

Retinal image acquisition and grading

On-axis, non-steered 200° UWF-FA and UWF-CI images were acquired using Optos retinal imaging devices (California or 200Tx, Optos plc, Dunfermline, UK). All images were stereographically projected using proprietary prototype software available from the manufacturer (Optos, plc, Dunfermline, Scotland, UK). This stereographic projection was based on ray tracing of each pixel through a combined optical model of the Optos California or Optos 200TX and a Navarro UWF model eye. This software corrects for peripheral image warping that occurs when a spherical three-dimensional object (fundus) is flattened onto a two-dimensional image.¹⁵ Images were then graded by two independent graders (OA and MA) using the ETDRS DRSS (online supplemental table 1). Each grader evaluated only one image modality (either UWF-CI or UWF-FA) and was masked to the DRSS grade obtained from other imaging modality. All UWF-CI were graded by OA while all UWF-FA were graded by MA. Both graders were certified by the Joslin Vision Network to assess DR severity and were experienced in the grading of images for clinical trials. The graders used the ETDRS DRSS for UWF-CI and directly adapted the colour severity scale to the grading of UWF-FA.¹³ In addition to the DRSS, the extent and severity of individual DR lesions (HMAs, VB, IRMAs and NV) was evaluated. A retina specialist (PSS) experienced in grading UWF images performed side-by-side adjudication between UWF-FA and UWF-CI images with any discrepancies in the DRSS and the cause of the difference was recorded for each case. To determine intergrader agreement, a subset of 100 UWF-CI were regraded by grader (MA) 1 month after he completed grading the UWF-FA images. The agreement rate between both graders was 93% within one step of DRSS with a 53% exact agreement rate comparable to that of established reading centres and current multicentre trials.¹⁶

Measurement of retinal NP

Total NP area (NPA) and total gradable area (TA) were measured on each UWF-FA image using a previously validated technique at the Joslin Diabetes Centre.^{12 17} The NPA and TA were calculated in square millimetres (mm²) by summing the size of all pixels within the appropriate masked area using a proprietary tool (Optos plc, Dunfermline, Scotland, UK) that implements DICOM Supplement 173.¹⁸ Using this tool, the size of an individual pixel was individually defined by its location in the image and was calculated using spherical trigonometry after projecting it back onto a spherical surface, thus allowing an accurate estimation of retinal area (mm²) independent of peripheral image distortion.¹⁹ The NP index (NPI) for each eye was calculated by dividing NPA by gradable area.

Statistical analysis

Sensitivity (SN) and specificity (SP), and simple (K) and linear weighted (Kw) kappa statistics were evaluated. Kappa agreement of 0.21–0.40 was considered fair, 0.41–0.6; moderate, >0.61–0.80 substantial and >0.81 almost perfect agreement.²⁰ Non-parametric analyses (Wilcoxon rank-sum tests) were used to compare distributions of continuous variables between groups of eyes with different DR severity levels. The Fisher’s exact test or χ² test was used to compare frequencies of categorical variables as appropriate. All analyses were performed using SAS software V.9.4 (SAS). A p<0.05 was considered significant.

RESULTS

The study included 218 eyes of 137 patients with a mean age was 50.6±13.7 years, 45.4% were female, 52.7% type 1 DM with a mean duration 22.1±11.6 years and average A1c of 8.47±1.6. Distribution of DRSS using UWF-CI versus UWF-FA was: mild NPDR (level 35) 8.7% vs 9.6%, moderate NPDR (level 43) 19.3% vs 8.3%, moderate to severe NPDR (level 47) 22.0% vs 15.1%, severe NPDR (level 53) 35.8% vs 49.1 %, PDR <high risk (level 61, 65) 12.4% vs 12.4%, and high-risk PDR (level 71) 1.8% vs 5.5%. The cross tabulation of DRSS by UWF-CI and UWF-FA is presented in table 1.

Table 1.

Cross-tabulation of the number of eyes by level of Diabetic Retinopathy Severity Scales derived from ultrawide field colour images (CI) and ultrawide field fluorescein angiograms (FAs)

CI DR grade	FA DR grade
	Mild NPDR (level 20, 35) n (%)	Moderate NPDR (level 43) n (%)	Moderately Severe NPDR (level 47) n (%)	Severe NPDR (level 53) n (%)	PDR<HRC (level 61, 65) n (%)	PDR≥HRC (level 71, 75) n (%)	Total
Mild NPDR (level 20, 35) n (%)	15 (6.88)	1 (0.46)	3 (1.38)	0 (0.00)	0 (0.00)	0 (0.00)	19 (8.72)
Moderate NPDR (level 43) n (%)	6 (2.75)	9 (4.13)	15 (6.88)	12 (5.50)	0 (0.00)	0 (0.00)	42 (19.27)
Moderately Severe NPDR (level 47) n (%)	0 (0.00)	8 (3.67)	15 (6.88)	25 (11.47)	0 (0.00)	0 (0.00)	48 (22.02)
Severe NPDR n (%)	0 (0.00)	0 (0.00)	0 (0.00)	68 (31.19)	9 (4.13)	1 (0.46)	78 (35.78)
PDR<HRC (level 61,65) n (%)	0 (0.00)	0 (0.00)	0 (0.00)	2 (0.92)	18 (8.26)	7 (3.21)	27 (12.39)
PDR≥HRC (level 71,75) n (%)	0 (0.00)	0 (0.00)	0 (0.00)	0 (0.00)	0 (0.00)	4 (1.83)	4 (1.83)
Total n (%)	21 (9.63)	18 (8.26)	33 (15.14)	107 (49.08)	27 (12.39)	12 (5.50)	218 (100)

Exact agreement: 59.17%; within one-step agreement: 92.66%; simple kappa statistic 0.46 (95% CI 0.37 to 0.54); weighted kappa statistic (linear scale): 0.65 (95% CI 0.58 to 0.71).

HRC, high-risk characteristics; NPDR, non-proliferative diabetic retinopathy; PDR, proliferative diabetic retinopathy.

Agreement rates for DRSS

Overall agreement between UWF-CI and UWF-FA for DRSS was moderate to substantial (K = 0.46 and Kw = 0.65). DRSS agreement between UWF-CI and UWF-FA images was within one step in 92.7% of eyes and exact in 59.2% (table 1). Only 7.3% (n = 16) eyes were graded as having more severe DRSS on UWF-CI images than on UWF-FA. Compared with UWF-CI, 4.6 times more eyes were graded as having more severe DR on UWF-FA (73 (33.5%) vs 16 (7.3%)). Compared with UWF-CI, UWF-FA identified one step more severe DRSS in 26.2% and two steps more severe DRSS in 7.3% of eyes (table 1). In eyes graded as mild NPDR on UWF-CI and moderate on UWF-FA, the discrepancy was driven only by H/Mas. However, in eyes graded as moderate NPDR on UWF-CI but severe NPDR on UWF-FA, the difference was driven by H/Mas in 81.0% of eyes, IRMA in 16.2% of eyes and both in 2.7% of eyes (online supplemental table 2 and figure 1).

Figure 1.

Retinal images and angiograms of an eye with moderate NPDR on colour composite images and severe NPDR on the fluorescein angiogram. Inset shows the magnified area outlined in white and red in the colour composite image and angiogram, respectively. Colour Picture (CP-left), green channel/red-free (RF-middle), angiogram (FA-right). Magnified view shows an area with intraretinal microvascular abnormalities greater than 8A that is visible on the angiogram but not clearly visible on the colour composite images. NPDR, non-proliferative diabetic retinopathy.

Among eyes with PDR on UWF-FA (N = 39), 29 (74.4%) were identified on UWF-CI, yet 10 were missed (25.6%) (SN = 0.74 (95% CI 0.58 to 0.86), SP = 0.98 (95% CI 0.94 to 0.99))(figure 2). There were two eyes that were graded as having PDR on UWF-CI but did not have signs of NV on FA figure 2. Among eyes with vision threatening DR (severe NPDR or worse) on UWF-FA (N = 146), 37 eyes (25.3%) were graded as less severe by UWF-CI (SN = 0.74 (95% CI 0.67 to 0.81), SP = 0.97 (95% CI 0.90 to 0.99)). The SN, SP and predictive values for UWF-CI as compared with UWF-FA are presented in table 2.

Figure 2.

Retinal images and angiograms of an eye with moderate NPDR on colour composite images and early PDR on the fluorescein angiogram. Inset shows the magnified area outlined in white and red in the colour composite image and angiogram, respectively. Colour Picture (CP-left), green channel/red-free (RF-middle), angiogram (FA-right). Magnified view shows new vessels larger than half disc area that is visible on the angiogram but not clearly visible on the colour composite images. NPDR, non-PDR; PDR, proliferative diabetic retinopathy.

Table 2.

Sensitivity, specificity, positive and negative predictive values for advanced DR grading on ultrawide field colour imaging compared with ultrawide field fluorescein angiography, using fluorescein angiography as the gold standard

	Sensitivity (95% CI)	Specificity (95% CI)	PPV (95% CI)	NPV (95% CI)
Vision threatening DR	0.74 (0.67 to 0.81)	0.97 (0.90 to 0.99)	0.98 (0.93 to 1.00)	0.66 (0.56 to 0.75)
PDR	0.74 (0.58 to 0.86)	0.98 (0.94 to 0.99)	0.87 (0.71 to 0.96)	0.95 (0.90 to 0.97)

DR, diabetic retinopathy; NPV, negative predictive value; PDR, proliferative DR; PPV, positive predictive value;

An exploratory analysis was conducted looking at type 1 and type 2 DM separately. There were no substantial differences in the agreement rates between both types of DM. Results are summarised in online supplemental table 3.

Association of retinal NP and DR severity on colour and fluorescein angiographic grading

Based on ETDRS DRSS grading of both UWF-CI and UWF-FA images, there was a significant trend of increasing total NPI with increasing DR severity across all DR severity levels (CI: p = 0.003, FA: p = 0.0015) (figure 3). The extent of NPI begins to increase substantially at level 53 in both colour and FA ETDRS grading. In eyes with early DR by colour grading, the extent of NPI increased by 4.7-fold between ETDRS levels 35 and 43/47 (NPI level 35: 0.0044871, 43/47: 0.0211597, p = 0.0065). In contrast, eyes graded on UWF-FA as having ETDRS level 35 approached but did not reach a significantly increased NPI by level 43/47 (NPI level 35: 0.0038177, 43/47: 0.0088718, 2.3-fold increase, p = 0.0598).

Figure 3.

Box and whisker plot demonstrating increasing non-perfusion with worsening DR severity on both colour images (CI) (left panel) and on fluorescein angiography (FA) (right panel). DR, diabetic retinopathy; ETDRS, Early Treatment Diabetic Retinopathy Study; NPI, Non-Perfusion Index.

Association with progression

A total of 109 eyes were graded as having either mild or moderate NPDR on UWF-CI. Of those, 65 eyes had a follow-up of 4 years or more (6.37±1.24 years) and were included in the longitudinal analysis. In the group with similar DRSS on both CI and FA, progression rates to PDR were 7.6%. In eyes with more severe DRSS grades on FA compared with CI, progression rates were higher (FA: 18.8% vs CI: 7.6%) but not statistically significant (χ² p = 0.41). When excluding eyes with mild NPDR (n = 11), eyes with moderate on both CI and FA (n = 25) had a progression rate to PDR of 8.7% while those graded as moderate on CI and severe on FA (n = 29) had a higher but non statistically significant progression rate to PDR (20.8%, p = 0.55).

DISCUSSION

Using the standard ETDRS colour DR severity grading scale on stereoscopically projected UWF-FA images, substantially more DR was observed as compared with using UWF-CI. In this cohort, DR severity was worse in 73 (33.5%) eyes on UWF-FA and in only 16 (7.3%) eyes on UWF-CI. In eyes with early DR (ETDRS levels 20–47, N = 109 eyes) DR severity was worse by UWF-FA in 56 (51.4%). In eyes with more advanced DR (ETDRS levels 53 and higher, N = 109), only 17 (15.6%) eyes were graded more severe by UWF-FA. As observed previously, there was a significant association between increasing NP and increasing DR severity.^{21 22} However, in eyes with ETDRS levels 35–47 when graded by ETDRS colour grading scales on UWF-FA, the significant trend of greater NPI in eyes with more severe DR only approached statistical significance. These data suggest that there are distinct differences between UWF-CI and UWF-FA grading of DR severity. Which correlates better with DR progression remains unknown and will require large longitudinal studies to determine. The added lesion detection may improve ability to identify eyes at greater risk, or detecting more lesions at baseline may make detection of interim change harder.

Data from Protocol AA of the DRCR retina network has shown that increasing NP and predominantly peripheral lesions graded on UWF-FA were associated with increased risk for disease worsening over time.²³ The measurement of retinal NP is a reproducible and objective method of quantifying disease severity that has been used in large-scale research settings.¹² By quantifying the underlying disease pathology, NP measurements may serve as an objective indicator of the applicability of a subjective categorical grading scale. The current study looked at whether NP measurement trends differed in the same group of eyes based on the imaging modality used. As seen in prior reports, in both UWF-CI and UWF-FA, there is a clear trend for increase NP with increasing DR severity. However, in eyes with early DR by colour grading (ETDRS levels 20–47), there was significant rise in the extent of NP in eyes with level 43–47 that continues to increase as severity worsens. This trend was not observed in the same group of eyes when graded using the ETDRS colour scale directly adapted to UWF-FA images. In eyes with early DR on UWF-CI, the increased identification of retinal lesions such as microaneurysms and IRMA on UWF-FA was not necessarily accompanied by a significant increase in retinal NP. The prognostic implications of the substantially more visible lesions of UWF-FA compared with UWF-CI is still unclear. While the current study was not powered to conduct a longitudinal analysis, an analysis for 65 eyes that completed 4 years follow-up showed no significant difference in the progression rates to PDR between eyes with similar and different DRSS on UWF-CI and UWF-FA. However, given the small sample size, more studies with a larger sample size will be necessary to confirm these findings. These findings demonstrate that depending on the imaging modality used to grade DR there can be significant differences, which could clearly effect outcomes of clinical studies and clinical trials. Our present findings suggest that colour-grading scales cannot be readily applied to FA images at this time. Pending longitudinal study results, investigations that will employ FA based imaging will need to identify a validated method to determine DR severity that should be correlated with an objective measure of disease severity and predictive of DR progression.

While it was expected that UWF-FA detects significantly more NV compared with UWF-CI, the fact that CI missed approximately 25% of eyes with PDR was surprising. The sensitivity of UWF-CI at detecting eyes with PDR in the current study was 75% which is lower than that previously reported by Haridas et al (92.5%).²⁴ There are some differences in the methodology between both studies including the use of the ICDR scale (vs ETDRS) and the use of WF-FA on the Heidelberg (vs UWF-FA). The majority of lesions missed were NV elsewhere (NVEs), many of which were subtle and located in the retinal periphery. Data from landmark studies such as the Diabetic Retinopathy Study (DRS) and the ETDRS study addressed the risk of severe visual loss (SVL) was based on the assessment of NV within the ETDRS fields and relied on colour photographs not FA.^{13 14} The risk of bleeding and SVL from NVs in the peripheral retina or subtle NVs anywhere in the retina remains to be fully determined. The DRS study reported that in eyes with NVE but without NVD the cumulative event rates for SVL was 1.4% at 1 year and 9.6% at 2 years.²⁵ The ETDRS study reported that the 3-year rate for developing SVL in eyes with severe NPDR or PDR less than high-risk criteria was 6.5% compared with 3.8% in eyes receiving full scatter PRP.²⁶

Understanding the differences in DR severity brought about by using FA is important given the substantial use of FA in various regions of the world. In 2000, prior to the implementation of Australia’s national DR guidelines, only 39% of ophthalmologists never ordered Fas for patients with mild NPDR, no vision changes and no oedema, while 50% sometimes/often ordered them and 10% almost always did.⁷ The National Health and Medical Research Council (NHMRC) guidelines recommended the use of FA in eyes with diffuse DME to identify sources of leakage to guide therapy, to diagnose eyes with diabetic macular ischaemia or in eyes with PDR to assess response. One year later and contrary to the NHMRC guidelines, only 50% of ophthalmologists never ordered Fas for early DR and that number remained stable 10 years later.^{27 28} A recent survey in Egypt which included approximately 6% of all Egyptian ophthalmologists suggested that FA was being widely used in the management of patients with DR.⁸ In those with DR and no DME, 32.5% of respondents said that they would use FA in patients with mild DR, 44.3% in patients with moderate DR and 51.2% in patients with severe non-proliferative diabetic retinopathy (NPDR). Approximately 70% of respondents used colour grading scale to evaluate FA to determine disease severity and make treatment decisions. Given that over 50% of respondents said they would treat eyes with severe NPDR with PRP, the potential for over treatment is substantial.

Limitations of the current study include its retrospective cross-sectional nature and thus its inability to evaluate the prognostic value of using either imaging modality in predicting DR progression. Strengths of the current study include that both the UWF-FA and UWF-CI were acquired within 1 month of each other and all images were graded by certified trained graders in a masked manner with adjudication performed by senior trained retina specialists (PSS and MA).

In conclusion, grading of UWF-CI and UWF-FA are not readily interchangeable. The ETDRS colour severity scale adapted to UWF-FA images identifies substantially more retinal lesions and as a result suggests more severe DR than UWF-CI. In early DR, only colour grading was associated with the extent of NP. These findings need to be considered among those using FA for DR severity determination. How well the two modalities relate to each other with regard to the risk of DR progression will require future prospective evaluation.

Supplementary Material

Supplemental material This content has been supplied by the author(s). It has not been vetted by BMJ Publishing Group Limited (BMJ) and may not have been peer-reviewed. Any opinions or recommendations discussed are solely those of the author(s) and are not endorsed by BMJ. BMJ disclaims all liability and responsibility arising from any reliance placed on the content. Where the content includes any translated material, BMJ does not warrant the accuracy and reliability of the translations (including but not limited to local regulations, clinical guidelines, terminology, drug names and drug dosages), and is not responsible for any error and/or omissions arising from translation and adaptation or otherwise.

► Additional supplemental material is published online only. To view, please visit the journal online (http://dx.doi.org/10.1136/bjo-2022-322163).

WHAT IS ALREADY KNOWN ON THIS TOPIC

• Ultrawide field (UWF) colour imaging (CI) and fluorescein angiography (FA) are widely used in clinical practice to manage diabetic retinopathy (DR). UWF-FA detects significantly more DR lesions as compared with UWF-CI.

WHAT THIS STUDY ADDS

• When grading DR severity on UWF-FA images using the classic Early Treatment Diabetic Retinopathy Study (ETDRS) colour severity scale, a substantial number of eyes are graded as more severe.

HOW THIS STUDY MIGHT AFFECT RESEARCH, PRACTICE OR POLICY

• ETDRS colour severity scale should not be employed to grade the severity of DR on UWF-FA images as it may lead to overtreatment. Studies looking to develop separate DR severity scales for FA and assess its ability to detect eyes at higher risk of progression are needed.

Data availability statement

Data are available on reasonable request.