Summary of findings
The CLARITY study showed that intravitreal aflibercept monotherapy was superior to standard PRP treatment for PDR through 52 weeks; the effect was achieved with a median of one aflibercept injection only in the 40 week post-loading phase, indicating that aflibercept is a feasible new approach for compliant patients. The superior BCVA findings were supported by significantly better binocular visual acuity and binocular Esterman scores in the aflibercept arm. These observations have significant impact on eligibility to retain a driving licence. In the UK, the Driver and Vehicle Licensing Agency has designated both a minimum visual acuity and Esterman visual field standard to maintain a valid driving licence.83 With advances in laser technology and techniques, there are reports with short follow-up suggesting that modern-day laser techniques and technology, such as multispot laser, have reduced the prevalence of visual field loss with PRP.84,85 However, our study shows that, despite 69% of the study cohort being treated with multispot laser, aflibercept is associated with a lower risk of visual field loss than modern-day laser at 52 weeks, in keeping with findings noted in the recent ranibizumab trial in PDR at 2 years.53 Other visual outcomes that measured adverse effects of PRP, such as contrast sensitivity and low-luminance visual acuity, were not significantly different between arms, although removing outliers suggested greater preservation of low-luminance visual acuity letter score by 52 weeks in the aflibercept arm.
Panretinal photocoagulation remains an effective treatment in preventing severe visual loss.52 On average, the visual acuity remained very stable over 52 weeks but twice as many people lost 10 or more letters when treated with PRP. In contrast, the proportion of patients who lost visual acuity was minimal in the aflibercept arm. However, the study reiterates the Protocol S study in that PRP is not a one-off procedure: 65% of patients require supplemental therapy despite advances in laser technology.
Aflibercept also improved the level of diabetic retinopathy severity, which resulted in a higher proportion of eyes with total regression of new vessels than PRP. This anatomical effect should also be considered when choosing between anti-VEGF and PRP as a first-line option in PDR. As aflibercept is licensed for DMO, the findings of this study indicate that aflibercept is also effective in the management of PDR in the first year, allowing the use of a single agent to address both of these sight-threatening complications of diabetes mellitus.
These data are likely to promote a paradigm shift in the treatment of PDR.
Economic evaluation
Our objective was to evaluate the cost-effectiveness of intravitreal aflibercept therapy for PDR versus standard care, PRP.
Economic evaluation was undertaken on 202 participants (101 per arm) with complete cost and outcome data. This represents 96.7% of the clinical sample included in primary outcome ITT analysis. From a public sector multiagency perspective that covers health and social care services, treatment with aflibercept costs more in terms of total resource use (mean adjusted total additional cost per patient = £5475, bootstrapped 95% CI £5210.79 to £5749.82) than PRP laser treatment over the 52-week follow-up period. Sensitivity analysis in which we varied the costs of aflibercept from the list price to reflect possible NHS PAS showed this to be the case at any price because of the additional cost of administering aflibercept. Participants who received aflibercept gained some benefit in BCVA (mean = 3.93, bootstrapped 95% CI 3.84 to 4.02) but at an increased cost. No statistically significant difference was found in self-reported generic HRQoL (EQ-5D-3L) or in terms of capability (ICECAP-A). It may be that these measures were not sufficiently sensitive to pick up any changes over the 52-week follow-up period between arms. We have undertaken a secondary exploratory cost–utility analysis (i.e. cost-per-QALY analysis). Evidence is mixed, but points to the EQ-5D-3L not being sufficiently sensitive to be useful in studies of visual impairment.44,45 The results of our study speak for themselves: a mean adjusted cost difference of £5475 is divided by an extremely small and not statistically significant mean adjusted QALY difference of –0.022. This yields a cost per QALY of –£252,827, in which we do not have much confidence. Given that a positive significant difference was observed in the BCVA for the intervention group, we interpreted this as the EQ-5D-3L not being sufficiently sensitive in this context. Interestingly, the vision-specific self-reported health-related QoL measures (RetDQoL and NEI-VFQ-25, non-preference-based) also showed no statistically significant difference over the study period between arms. If society is willing to pay £1400 for an additional 1-point improvement in BCVA, then aflibercept has a 56.60% probability of being cost-effective at the list price of £816.
Mechanistic evaluation
This section is adapted from Nicholson et al.82 ©The Authors, 2018. This is an Open Access article distributed in accordance with the terms of the Creative Commons Attribution (CC BY 4.0) license, which permits others to distribute, remix, adapt and build upon this work, for commercial use, provided the original work is properly cited. See: http://creativecommons.org/licenses/by/4.0/.
Our study on 40 patients (20 in each arm) did not reveal any significant change in retinal oximetry between arms. This may be because of the small sample size. The study was not powered to detect important clinically significant differences between arm. The study was exploratory and hypothesis driven to decide on sample size of future studies in this area if changes observed were worth pursuing. The SaO2 and SvO2 at baseline were both high compared with the normative data on retinal oximetry, in keeping with reports on retinal oximetry in PDR. In addition, the standard deviations of the values for both SaO2 and SvO2 are similar to reports on PDR, suggesting that the non-significant changes are unlikely to be errors in measurement. After accounting for test–retest variability, we would require a mean of 5% change in SO2 in the arteries and veins to confidently report a treatment effect.78 This was not achieved in either the treatment-naive group or the post-PRP-treated non-naive group in either arm. Therefore, although there may be an overall change in the distribution of oxygen supply and consumption with either treatment, our study results show that these treatments do not affect the overall oxygen exchange in the retina. Guduru et al.86 reported a correlation between retinal capillary non-perfusion and retinal arterial SaO2. Our study participants in the mechanistic study did show significant areas of capillary non-perfusion but both aflibercept and PRP failed to show any difference in change in capillary non-perfusion at 52 weeks, with patients in both arms showing a continuing increase in area of total non-perfusion. This increase in non-perfusion was not associated with a corresponding change in retinal oximetry values, emphasising that retinal oximetry represents only global change in oxygen and may not be accurate enough to measure local changes in redistribution of oxygen demand and supply.87
Strengths and limitations
This study included patients presenting prospectively to the medical retina clinics where patients are managed in a real-life setting so the study allowed accurate testing of the hypothesis. The study also included both treatment-naive eyes and eyes with persistent NV post-initial PRP and was therefore representative of the UK population who would benefit from an alternative treatment to PRP. Patients who have been previously treated are often excluded from ophthalmology trials and, therefore, this is the first study that demonstrated the impact of anti-VEGF in this indication. We excluded eyes with macular oedema to avoid any confounding effect on the primary outcome analysis. Therefore, the study provided the true benefit of anti-VEGF agent in PDR. Participants underwent thorough phenotyping including clinical, qualitative and objective assessment and an independent Reading Centre also further graded the retinal images to ensure a robust validation of the study anatomical outcomes.
The study was slow to recruit initially but we successfully met the target by increasing recruitment of five clinical trial sites in time. The main reason for the slow recruitment in the initial phase was the additional burden on busy clinics, the provision of PRP being available on the same day as the retinal consultation, and the perception that anti-VEGF therapy is short-acting. The increase in performance across the sites in the later half of the study may be because of the availability of new data on the effectiveness of anti-VEGF on diabetic retinopathy and so the research team became more reassured that anti-VEGF agents may indeed be a useful treatment for PDR too.
The robust randomised controlled trial (RCT) design, high statistical power and excellent retention rates are particular strengths of this study. The study patients are representative of the PDR population and, therefore, these findings can be generalised to clinical practice for the first year of therapy. Retreatment criteria used in CLARITY were very similar to those followed in the ranibizumab trial17 and were determined by treating investigators at each study visit. Compliance with treatment (94% aflibercept arm and 97% PRP arm) was very good in CLARITY, indicating that these retreatment criteria can be easily applied to routine clinical practice.
The safety evaluation of aflibercept in CLARITY revealed no new concerns on this drug. There were no statistically significant differences in APTC events or other systemic adverse events between arms at 52 weeks, substantiating the reports of previous intravitreal aflibercept studies in diabetic patients. Long-term studies on anti-VEGF in PDR patients are required for meta-analysis of the safety of these agents in PDR.
The limitation of this study is that it was a Phase IIb study with follow-up for only 52 weeks. To date, the only other well-designed study on anti-VEGF for PDR included patients with DMO and so the treatment regimen was preplanned to be more intense than this study.11 However, as a 5-year study, it will provide long-term outcomes of ranibizumab in PDR, information on the disease-modifying effect of anti-VEGF and the long-term compliance of patients. The study also excluded patients with significant fibrovascular proliferation or tractional retinal detachment in the posterior pole and eyes with prior vitrectomy. We have limited experience of treating severe retinopathy at the posterior pole with anti-VEGF therapy and so we decided that this group is best avoided as risks may outweigh benefits.
Approximately 75% of the PDR patients included in this study had low-risk PDR. This mirrors the study population in Protocol S suggesting that PRP is now initiated in patients with low-risk PDR.
We did not collect information on ADL that would have told us something about how treatment of this condition with aflibercept versus PRP affects home and work life.
This study used a 12-month recall period for contacts with health and social care services, which is considered reasonable but towards the boundary of recall for patients.88
For the economic evaluation, we were primarily interested in our health economic analysis on how the cost per change in BCVA compared with the cost per QALY and hence decided to use complete-case analysis where full data were available for these outcome measures. We do not impute costs and, therefore, this left us with 202 participants (this represents 96.7% of the clinical sample included in primary outcome ITT analysis).
Comparison with existing literature
The DRCR.net performed a multicentre (55 sites), randomised clinical trial comparing panretinal photocoagulation with 0.5 mg of intravitreal ranibizumab in 305 patients with PDR.52,53 PRP was performed at baseline and ranibizumab was given at baseline and week 4 pro re nata. Eyes with DMO in both groups were eligible to receive ranibizumab. The primary outcome was change in BCVA and the secondary outcomes included area under the visual acuity curve, peripheral visual field loss (as measured on Humphrey automated visual field testing), incidence of vitrectomy, development of DMO, and persistent or new neovascularisation. Improvements in BCVA for the ranibizumab and PRP groups were +2.2 and +0.2 letters, respectively (95% CI –0.5 to + 5.0). The group receiving ranibizumab experienced less peripheral visual field sensitivity loss (–23 dB vs. –422 dB, 95% CI 213 dB to 531 dB; p < 0.001), fewer vitrectomies (4% vs. 15%, 95% CI 4% to 15%; p < 0.001), and a lower incidence of DMO (9% vs. 28%). Ranibizumab-treated eyes required a median of seven injections through year 1 and 10 injections through year 2. Forty-five per cent of eyes in the PRP group required additional laser and 53% of eyes required ranibizumab for DMO. The authors concluded that ranibizumab may be a reasonable alternative to PRP through 2 years. The decreasing number of injections in year 2 suggests that some disease modulation occurs after 1 year of ranibizumab therapy.
Comparison with other economic evaluation studies
A US study exploring the cost-effectiveness of intravitreal ranibizumab compared with PRP for PDR found that over 2 years, compared with PRP, 0.5 mg of ranibizumab was within the US$50,000 to US$150,000 per QALY range frequently cited as cost-effective in the USA for eyes presenting with PDR and vision-impairing DMO, but not for those with PDR without vision-impairing DMO.88 The US study measured costs associated with ocular treatments and potential complications only, and based quality-of-life adjustments in QALY calculations on utility weights by mapping from Brown et al.89 for a 2-year period. Sensitivity analysis explored the effect on the ICER of time trade-off responses from trial participants. Brown et al.89 did not collect preference-based utility data directly from patients.
Implications for practice and research recommendations
In addition to improving the visual acuity, intravitreal aflibercept also reduced the level of diabetic retinopathy severity and resulted in a higher proportion of eyes with total regression of new vessels than PRP, fewer visually disabling vitreous haemorrhages and fewer cases of incident macular oedema. These anatomical effects should be considered when choosing between anti-VEGF and PRP as a first-line option in PDR. As aflibercept is licensed for DMO, the findings of this study indicate that aflibercept is also effective in the management of PDR in the first year, allowing the use of a single agent to address both of these sight-threatening complications of diabetes mellitus in patients presenting with both PDR and DMO.
We recommend a longer study investigating the efficacy and effects of this treatment, including safety events over a duration of up to 5 years to inform future treatment guidelines for this indication.
Future health economics studies in this area should ensure the inclusion of a range of outcome measures (vision specific and generic, as we did in this study) and should model costs and benefits over a longer period. Although the CLARITY study was limited to 1 year, the rate of complications was significantly higher in the PRP arm in terms of new-onset macular oedema and vitreous haemorrhage by 52 weeks. A similar study, Protocol S,53 showed that, over 2 years, significantly more vitrectomies were performed in the PRP arm. Cataract surgery is also frequently needed after vitrectomy. Similarly, more patients in the PRP arm need to be treated with anti-VEGF therapy for macular oedema. The costs of these complications have to be taken into account when modelling costs over a 5-year period or longer.
Some physicians who prefer anti-VEGF therapy over laser will probably use bevacizumab although data for these drugs are not yet available. Confirmatory studies have to be done before translating the effect of ranibizumab and aflibercept studies to bevacizumab as the number of injections varied between Protocol S and CLARITY and the study designs of these two studies were not identical. Therefore, the CLARITY study results should be used cautiously if other anti-VEGF agents are used to treat PDR.
If there is an intention in a future economic evaluation study to use QALYs as an outcome measure to allow comparison with, for example, the NICE payer threshold of £20,000 per QALY, then feasibility work is necessary to identify ways of measuring utility relating to visual impairment that make the QALY a meaningful outcome measure in this context.