Evaluation of care with intravitreal aflibercept treatment for UK patients with diabetic macular oedema: DRAKO study 24-month real-world outcomes

The study addresses the effectiveness of intravitreal aflibercept (IVT-AFL) for diabetic macular oedema (DMO) in routine UK clinical practice over 24 months. Diabetes prevalence is rising globally, and DMO is the most common cause of vision loss among people with diabetes. Anti-VEGF therapies are first-line for DMO, with IVT-AFL enabling proactive dosing that can reduce monitoring burden. Randomised controlled trials (RCTs) may not reflect real-world outcomes due to adherence and resource constraints; therefore, DRAKO, a prospective observational study, evaluates outcomes under local standard-of-care conditions. The key questions are whether visual acuity (VA) and anatomical gains achieved with IVT-AFL are maintained or improved over two years, and how baseline characteristics and treatment patterns influence outcomes.

Prior pivotal RCTs (e.g., VIVID/VISTA) and observational studies (e.g., APOLLO) demonstrated the efficacy of anti-VEGF, including aflibercept, largely in populations with lower baseline VA (<73 letters). Registry-based studies indicate fewer injections and smaller functional gains in real-world practice compared with RCTs. Baseline VA and CST strongly influence potential for improvement with anti-VEGF therapy. The DRAKO study contributes UK-specific, prospective, real-world evidence across a broad range of centres and standard-of-care protocols.

Design: DRAKO (NCT02850263) was a 24-month, prospective, observational, multicentre, non-randomized UK study in routine clinical practice across 35 NHS hospitals (enrolment July 2016–April 2018). Patients had central-involving DMO; the decision to treat with IVT-AFL was made independently of the study. Two cohorts were enrolled: anti-VEGF treatment–naïve (N=507) and non-treatment–naïve (N=241; no anti-VEGF for DMO in prior 26 months). All participants provided informed consent; ethics approval was obtained (16/NW/0014). Treatment: IVT-AFL administered per local standard of care. Visits at baseline and follow-up; Month 12 and Month 24 windows were ±1 month from baseline anniversary. QoL measured with NEI VFQ-25 at baseline, Month 12, and Month 24. Outcome measures: Primary outcomes were mean change from baseline in BCVA (ETDRS letters) and CST at Month 12 for both cohorts (treatment-naïve Year 1 results previously reported). This analysis reports non-treatment–naïve primary outcomes at 12 months and secondary/exploratory outcomes for both cohorts through Month 24. Secondary outcomes included proportions gaining/losing ≥5, ≥10, or ≥15 letters; BCVA/CST changes stratified by baseline BCVA and CST subgroups; NEI VFQ-25 scores; diabetic retinopathy (DR) grading (English or Scottish schemes); and HbA1c at Months 12 and 24. Exploratory analyses examined number of injections relative to letter changes and baseline subgroups, and the impact of Year 2 injection numbers on outcomes. Statistical analysis: Interim (Month 12) and final (Month 24) analyses were performed separately by cohort. Two analysis populations per timepoint included a pre-protocol window (PPW) population with available baseline and Month 12/24 BCVA or CST, using last observation carried forward. Descriptive statistics summarized quantitative variables; categorical variables as frequencies/percentages. Outcomes were stratified by baseline BCVA (≥49.0–69.9, ≥70 letters) and CST (<300, 300–399, 400–499, ≥500 µm). No hypothesis testing for primary objectives; correlations between injection numbers and letter changes were assessed with correlation coefficients and p-values. Safety was assessed for all consented patients.

Cohorts and baseline: 750 patients enrolled (507 treatment-naïve; 241 non-treatment–naïve). PPW populations: treatment-naïve N=388 (Month 12) and N=326 (Month 24); non-treatment–naïve N=135 (Month 24). The non-treatment–naïve cohort was older (mean 64.5 vs 62.8 years). Most patients were male (~66%) and White (~64–68%). Type 2 diabetes in ~90% of both cohorts. Treatment-naïve patients had higher baseline BCVA (≥70 letters: 64.4% vs 59.3%); non-treatment–naïve had lower baseline CST (<300 µm: 37.8% vs 15.4%). Nearly 90% of non-treatment–naïve had prior ranibizumab. Visual and anatomical outcomes: Treatment-naïve (C1): Month 12 mean change BCVA +2.5 (SD 12.2) letters; Month 24 mean change BCVA +0.7 (12.7) letters. CST decreased further in Year 2: Month 12 mean change −119.1 (116.4) µm; Month 24 −123.3 (104.3) µm. Non-treatment–naïve (C2): Month 12 mean change BCVA +0.2 (10.2) letters; Month 24 −0.3 (13.0) letters. CST mean change Month 12 −79.1 (137.6) µm; Month 24 −91.6 (132.9) µm. Baseline subgroup effects: Greatest improvements occurred in those with worse baseline status. At Month 24, BCVA gains were ~17.6 letters (treatment-naïve) and 17.3 letters (non-treatment–naïve) for patients with baseline BCVA <50 letters; CST reductions were −214.8 µm (treatment-naïve) and −146.7 µm (non-treatment–naïve) for baseline CST ≥500 µm. Patients with baseline BCVA <70 letters gained above cohort means at Month 24. Injection patterns: Year 2 mean injections: 3.7 (treatment-naïve) and 4.3 (non-treatment–naïve). Correlations between injection numbers and letter gains/losses were generally low; injection numbers did not vary systematically by baseline status, suggesting limited personalization of treatment intensity. Letter gain/loss proportions (selected): At Month 24, among treatment-naïve, gains ≥5, ≥10, ≥15 letters occurred in 38.0%, 16.6%, 8.3%; losses ≥5 and ≥15 letters in 7.7% and 4.0%. Among non-treatment–naïve at Month 24, gains ≥5, ≥10, ≥15 letters in 39.6%, 9.4%, 3.8%; losses ≥5 and ≥15 letters in 26.4% and 0.9%. QoL and DR: NEI VFQ-25 overall baseline scores were high (approximately 804 treatment-naïve; 773 non-treatment–naïve) and were largely maintained over 24 months, mirroring VA outcomes. DR grading was predominantly stable over time; greater improvements were observed under the Scottish grading scheme (movement from R2 to R1). Glycaemic control: Mean HbA1c changes at 24 months were modest: +0.4 mmol/mol (treatment-naïve) and −0.7 mmol/mol (non-treatment–naïve).

In real-world UK practice, IVT-AFL maintained high baseline visual acuity and further improved anatomical outcomes over two years in DMO. Patients with poorer baseline BCVA or higher CST experienced the largest gains, reinforcing that baseline status predicts potential for improvement. Despite relatively low injection frequencies—particularly in Year 2—outcomes were generally favorable, though the modest correlation between injection numbers and vision gains and the lack of tailored dosing by baseline status suggest under-treatment and limited personalization. The maintenance of high vision-related QoL and stable DR grading align with stable VA outcomes. Small improvements in glycaemic control were observed; tighter glycaemic management could potentially enhance long-term outcomes. Overall, the findings support the effectiveness of IVT-AFL in routine care, while highlighting opportunities to optimize injection frequency and systemic control to potentially realize greater functional gains.

Year 2 DRAKO results confirm that IVT-AFL is an effective real-world treatment for DMO in the UK. High baseline VA and QoL were maintained over 24 months, and CST continued to improve in the second year despite relatively few injections. Outcomes were strongly influenced by baseline BCVA and CST, with the greatest improvements in those with worse baseline metrics. Future work should explore optimization of treatment intensity (especially in Year 1), personalization based on baseline characteristics, and improved glycaemic control to further enhance visual outcomes. The study provides UK-wide, prospective evidence to guide standard-of-care practices.

As an observational study, treatment administration and eligibility criteria were not standardized across sites, potentially introducing variability and confounding. Missing data (e.g., in DR grading) may have influenced some findings. Injection numbers did not appear to be tailored to baseline status, suggesting possible under-treatment. Safety data are not detailed in the provided results. Generalizability is to UK routine practice settings included in the study.