To the author’s knowledge this is the first cross-sectional review of one centre’s acute proliferative diabetic retinopathy related complications during the first COVID-19 wave. As routine eye appointments in NHS Grampian were suspended, all patients under HES care were risk stratified in order to identify those at highest risk of sight loss. These exceptional circumstances presented a unique opportunity for a complete dataset as all referrals were received at one point of entry. However, this cohort is likely to underrepresent the true number of patients who suffered a PDR complication as many may have not presented to acute services, particularly those who were shielding or patients who have experienced vitreous haemorrhages in the past.
Although many patients were not attending for their urgent appointments across the UK 3, we observed only three patients who did not attend for their review appointment following initial acute presentation. The reason for non-attendances is open to conjecture and may be due to COVID-related illness, fear of exposure to COVID or visual acuity may have improved following initial assessment. It is recognised that patients with diabetes are often overburdened with medical appointments for multiple systemic issues and may miss appointments due to work and family commitments. COVID is unlikely to have had an impact on the four patients who regularly did not attend.
63% of patients who presented were under HES care and two thirds of these had their scheduled review appointment affected by COVID cancellations. We are unable to say whether the complication could have been prevented if their follow-up had not been postponed.
2006 marked the rollout of Scotland’s national screening programme and has been shown to be an effective means of identifying diabetics at high risk of developing sight threatening retinopathy 4. With the pause of DRS many patients were vulnerable to the effects of PDR. We observed two patients who had their DRS appointment cancelled and subsequently presented with PDR related vitreous haemorrhages. In a report by Forster et al it was observed that missing one year of screening did not increase the risk detecting referrable retinopathy. However, not attending on two consecutive years increased the odds by 10.84 times of detecting referable retinopathy 5.
Development of an acute complication was seen in patients despite stable appearances at most recent hospital review and screening. Complications were observed in 9 patients who did not have their HES appointment delayed and 9 patients who were identified as having non-referrable retinopathy at screening. The mean time from screening to complication was 8.2 months. Reasons for progression may be the result of more erratic glycaemic control during lockdown, but may also be due to the unpredictability of DR.
11 (24%) patients had been discharged from HES back to DRS, 7 of whom had been shown to have stable retinopathy after treatment with PRP. Life-long follow-up is needed for patients with PDR as complications may still occur months to many years after a period of quiescence. The risk of recurrence can be anticipated by previous duration to achieve new vessel regression, duration of diabetes and metabolic control 6. Despite efforts to optimise modifiable risk factors retinopathy can still progress. Although this poses the question, how safe is it to discharge treated patients with PDR, a good liaison with community optometry, DRS and patient education may allow follow-up of stable treated PDR out with HES, thereby relieving stress on the already stretched HES.
HES have access to wide field fundus imaging, whereas, DRS use two 45-degree field images. Negretti et al observed 108 eyes and found that 17% of NVE were outside of the DRS imaging fields and 11% of patients with active PDR would have been missed 7. However, despite DRS not having 100% sensitivity it is considered a safe and effective screening tool to detect PDR. As the referral rate of PDR to ophthalmology in Scotland is reported to be 2.1–2.8% per year for type 1 diabetics and 0.4–0.7% for type 2 4, the discharge of stable patients avoids over burdening HES resources.
It is well recognised that the progression of diabetic retinopathy is directly related to poor glycaemic control. Our cohort had a mean HbA1c of 78 mmol/l. Lockdown related lifestyle disruption appears to have had an impact on glycaemic control. Khare et al demonstrated a 0.51% HbA1c rise in an Indian population over a 68 day lockdown 8. The compliance with diabetic medication and healthy living habits was significantly reduced after lockdown 9. Fernandez et al observed different results where type 1 diabetics in Spain using flash glucose monitoring system recorded improved glycaemic control over the period of lockdown 10.
One tertiary centre in Greece described the negative effect on visual acuity and progression to active PDR as a result of deferring appointments during lockdown 11. Ghosal et al created a predictive model with regard to diabetic glycaemic control during lockdown using multivariate regression analysis 12. They predicted an HbA1c increase of 3.68% over 45 days and a 2.9% increase in ocular PDR complication rates.
Although it is documented in the literature that deprivation is a major determinant of health and in the case of diabetic retinopathy, significant loss of vision, we found no correlation between SES and severity of vision loss at presentation or final visual outcome at last follow-up. Denniston et al reports an association between deprivation and late disease presentation with significant vision loss 13. In our cohort, the five non-attenders to screening were from more privileged areas. This may be attributed to the general demographic of Grampian’s population being more affluent and all patients out with Aberdeen City are equally disadvantaged by geography as great distances must be travelled in order to seek specialist ophthalmic support. It may also suggest that all patients were affected by COVID19 regardless of SES.
23 eyes (44%) of 18 patients received anti-VEGF injection as part of initial management of their PDR complication. We observed no statistical difference in visual improvement nor speed of visual improvement between the eyes that were injected and those that did not. Although all patients were offered follow-up, some patients received more regular follow-up than others due to complexity. In order to reduce omissions in data, visual acuities were taken at presentation, at 1 month and at 6 months.
The evidence for the use of intravitreal anti-VEGF injection in the context of VH secondary to PDR remains unclear and although there appear to be visual benefits, they are not long lasting. DRCR net conducted a randomised controlled trial where participants with VH secondary to PDR received either ranibizumab or saline 14. Little significance in vitrectomy rates were observed between the groups, although, patients who received anti-VEGF had greater improvement in visual acuity and reduced rate of VH recurrence in the short term. Huang et al suggested vitreous haemorrhage resolution hastened with bevacizumab to a mean of 12 weeks compared to controls at 18 weeks 15.
Of 850 patients diagnosed with PDR in a Finnish population, Wirkkala et al observed vitreous haemorrhage occur in 16% of type 1 diabetics and 9% type 2 diabetics 16. Two thirds received anti-VEGF injection and VH clearance occurred within 3 months in 92% of these eyes. They concluded that timely injection accelerated VH resolution, improved visual outcome, reduced recurrent VHs and reduced the need for vitrectomy by 72% over the 5 year study period.
The previously mentioned studies included patients with variable levels of PRP. Park et al studied the effect of bevacizumab on diabetic VH in patients with complete PRP 17. Injection increased the likelihood of VH clearance and reduced the need for vitrectomy. Although a greater visual improvement was seen, this effect was also short lived.
Urgent vitrectomy offers a surgical option for the rapid clearance of an acute diabetic VH. Although due to the invasive nature, associated risks and resource dependence other treatment modalities for the initial management of diabetic VH have been investigated. Protocol AB identified no significant difference in visual outcome at 6 months for patients initially treated with aflibercept compared to urgent vitrectomy with panretinal photocoagulation 18. Surgery was subsequently avoided in two thirds of patients that received aflibercept.
A concern when administering anti-VEGF to patients with advanced PDR is the risk of tractional detachment. Although there is strong evidence for anti-VEGF’s role in the management of PDR, it can facilitate fibrosis and subsequent traction through upregulation of the fibrin-fibronectin complex 19. Of 608 eyes that received anti-VEGF prior to vitrectomy for active PDR and non-macular involving traction, the incidence of tractional macula detachment was 10% at the time of surgery. The risk of detachment increased significantly if the surgery was performed greater than 6 days after injection.
With the concern of COVID-19 virus transmission via direct contact, droplet and airborne routes, efforts were made to ensure the time in close proximity with patients was kept to a minimum. Despite these risk mitigation measures, the delivery of urgent PRP was not delayed in this cohort of patients. In those previously treated with PRP, anti-VEGF was a practical option to manage active PDR due to the short procedural time.
Anti-VEGF therapy is a useful adjunct to PRP in the management of PDR 20, particularly if PRP is not practical. Protocol S suggested that ranibizumab is non-inferior and may be more effective than PRP for visual acuity at 2 years, although half of the participants in the PRP group also received anti-VEGF treatment for DMO 21. Indefinite intraocular injections is invasive, disruptive and costly when the alternative can be completed in a few sessions and without the risk of endophthalmitis or retinal detachment.
The CLARITY study demonstrated the advantageous effect on visual outcome at one year with a lower incidence of vitreous haemorrhage following loading treatment with aflibercept for PDR over standard PRP treatment 22. A meta-analysis by Gao et al demonstrated fewer PDR complications with the use of anti-VEGF when compared to PRP alone 23. Through the use of OCTA (optical coherence tomography angiography), He et al suggested that combination PRP plus anti-VEGF treatment was more effective at regressing NVE than PRP alone 24. If combination treatment were to be offered in routine practise one must consider the practicality in an era of COVID risk mitigation measures for example, cost, availability of resources and multiple hospital visits. PRP is thought to remain the mainstay in management in reducing the risk of sight threatening diabetic retinopathy.
The need for vitrectomy and development of neovascular glaucoma are advanced complications of PDR. We observed 11 eyes of 12 patients progress onto vitrectomy and 3 eyes developed neovascular glaucoma: one responded well to both IVT and PRP, and two required further IOP lowering procedures. The ETDRS study reported the 5-year incidence rate of vitrectomy to be 5.3% in diabetics with established DR 25. The progression to vitrectomy has reduced over the past three decades. In 2010, Ostri et al reports a 10-year incidence vitrectomy rate of 2.9% in type 1 diabetics 26.
In a Northern English population, Vaideanu et al report that although the prevalence of diabetes had increased from 2.8–5.5% from 2000 to 2010, respectively, the rate of PDR within the diabetic population had reduced from 2.4–1.8% and the vitrectomy rate in these patients had reduced from 7.7–5.7% over the same time period 27. This can be partly attributable to the introduction of screening, allowing patients to receive timely treatment prior to the development of sight threatening DR.
In our cohort, with 6 months follow-up, we observed a 32% re-bleed rate (n = 15) and a 25% (n = 3) rate of post vitrectomy vitreous haemorrhage. At 6 months post vitrectomy, 50% (n = 6) had a BCVA better than 6/60 and 50% had CF vision or worse. Yorston et al evaluated the post-operative outcomes of patients who underwent vitrectomy for PDR in a Scottish population 28. 22% re-bled within 6 months, 3% re-detached and 3% progressed onto neovascular glaucoma. The visual outcomes were described as unpredictable with 72% achieving a vision of 6/60 or better and 16% had CF vision or worse.
Limitations of the study include a lack of control group to compare how presentations to HES have changed. Comparing our data to pre-COVID conditions would have been preferred to determine if the differences in how patients had presented were significant. As a result of Grampian’s geography and the strong relationship between community and hospital eye care asynchronous telemedicine is commonplace, allowing many patients to be managed in the community. Comparison with prospective “post-COVID” data collection will be the focus of future research. Modifiable risk factors that affect diabetic retinopathy including hypertension and BMI were not included.
As patients with advanced complications of PDR are, by definition, complex there is no clear consensus on optimal management. Management styles are likely to have been influenced by social distancing measures, for example, intravitreal injection may have been preferentially offered over PRP. This decision-making process will have inter-clinician variation within the unit.
7 eyes of 6 patients had no follow-up data and efforts were made to contact patients and optometry practices to organise review. Closer review of the patients that were under DRS care would be of interest to determine if the retinal photograph and outcome of the most recent appointment demonstrated evidence of evolving PDR activity.