Implementation of An Intravitreal Therapy Online Audit Tool For Neovascular Age-related Macular Degeneration Clinical Outcomes Reporting: Fight Retinal Blindness Spain (FRB-Spain) Project

Purpose: To implement an intravitreal therapy audit tool for neovascular age-related macular degeneration (nAMD) outcome reporting in a tertiary referral center. Methods: Implementation of an online audit tool (Fight Retinal Blindness platform) for nAMD eyes receiving anti-vascular endothelial growth factor (anti-VEGF) intravitreal therapy over a 24-month follow-up period. Data entry was compliant with the ICHOM dataset for nAMD. These data included visual acuity (VA), eye conditions, injection drug and number of injections. Subgroup analysis was performed for treatment-naive (TN) and previously treated (PT) eyes. Results: 234 eyes (191 patients) were included in the study. No signicant differences were observed in the subgroup analysis in mean baseline VA (logMAR letters: 58.8 TN vs 62.7 PT, p= 0.081) or nal VA at 12 months (TN 61.4 vs PT 62.4; p= 0.703). However, 12 months median VA change favoured TN cases (+4 TN vs +0 PT, p= 0.010) and median number of injections showed no differences (7 TN vs 7 PT, p= 0.644). No statistically signicant differences were found at two years on number of injections even though mean nal VA showed signicant differences (66 TN vs 59.5 PT, p = 0.032) and VA change favoured the TN group (+2.4 TN vs - 4.6 PT, p= 0.003). Conclusion: We have successfully implemented an online tool to evaluate nAMD anti-VEGF treatment delivered in our center. This feasibility study demonstrates that the online audit tool allows evaluating real world intravitreal therapy outcomes and benchmark these results with clinical practice guidelines and other real-life series.


Introduction
Pivotal clinical trials have demonstrated the e cacy of drugs against vascular endothelial growth factor (VEGF) in neovascular age-related macular degeneration (nAMD) using xed dosage treatment regimens [1][2] . These studies determine whether new treatments are effective in a highly selected group of patients that may not be representative of the general population (internal validity comes at cost of external validity of its results). Moreover, current trends in clinical practice tend to individualize treatment protocols, such as treat and extend approaches (TAE) 3 . Large-scale population-based observational studies therefore represent a different level of evidence that can complement ndings from clinical trials and determine treatment effects in routine clinical practice, with higher external validity. Such real-world studies track patients for longer periods of time and can provide enriching results on safety, quality of life, long-term effectiveness and can identify signi cant drug adverse events that were not detected in clinical trials 4 .
The Fight Retinal Blindness (FRB) online platform is a web-based data collection tool designed and developed at Save Sight Institute (Australia) to track clinical outcomes in routine clinical practice, with several modules including retinal diseases as nAMD or diabetic retinopathy, but also glaucoma and corneal diseases 5 . Data collection for the nAMD module includes visual acuity (VA), ocular characteristics, drug types, number of injections and lesion activity at baseline and at each follow-up visit. This structured data entry system allows the collection of accurate, systematic and high-quality clinical data that enable physicians to self-audit their clinical outcomes. This ability to evaluate individual centre results allows clinicians to benchmark their real-life results with pivotal clinical trials, to compare different drugs or treatment protocols, as well as to determine long-term safety of anti-VEGF therapies in unselected eyes outside a clinical trial scenario 6-9 . Deployed in a multicentre approach, this tool collates a large series of eyes in order to identify predictive factors of treatment response, as per recent publications 10 . Australia, New Zealand and Switzerland have been tracking their outcomes with this system since 2007, and as January 2019 the FRB registry included data from approximately 184.562 treatments from about 12.949 eyes of 10.100 patients with nAMD.
The opportunity to implement audit tools such as the FRB online platform appears specially interesting in countries like Spain, where few data about real-life outcomes of anti-VEGF drugs for nAMD have been previously reported. With this aim, we established a pilot study to evaluate the feasibility of performing a real-world clinical care audit using the FRB platform, to ascertain the outcomes of a single intravitreal therapy unit in a tertiary referral public centre of Barcelona.

Design and Setting
Observational study of data collected from eyes that commenced intravitreal therapy for nAMD and had been tracked in the FRB platform. The details of the FRB database have been already published 5

Patient Selection and Variables
Patients included in the analysis consisted of eyes initiating intravitreal therapy from March 2015 to January 2018 with either ranibizumab or a ibercept for nAMD treatment that had completed 12 or 24 months of follow-up. Treatment-naive eyes were included as well as previously treated ones. Given treatment approaches mainly included Treat and Extend (TAE) strategies as well as xed bimonthly injections regimes 8, 11 . Lesions were graded as "active" if presenting with sub-and intra-retinal uid at the optical coherence tomography (OCT) examination or new haemorrhage at dilated fundus exploration, "active-subretinal uid" if only subretinal uid was present or "inactive" if no changes of activity were observed by physician examination.

Outcomes
The main outcomes were nal VA, change in VA and number of injections and visits during follow-up. In addition, eyes were considered in two groups depending on prior treatment received [treatment-naive (TN) vs previously treated (PT)]. For each group of eyes, the mean baseline VA and mean nal VA, as well as the proportions of eyes with VA of > 70 letters (20/40 Snellen) and < 35 letters (20/200 Snellen) were obtained at 12 and 24 months. Median number of injections and visits were also reported at the same time points.

Statistical Analysis
Absolute frequencies and percentages (%) were used to describe categorical variables as well as overall number of eyes in each group. Description of quantitative variables was performed with mean and standard deviation (SD), median and interquartile intervals (Q1;Q3), and 95% con dence interval (CI) where appropriate. Student T, Wilcoxon rank-sum, ANOVA, Kruskal-Wallis, McNemars, Fishers and chisquare tests were used as appropriate to compare baseline characteristics between site groups for 12and 24-month completers. A bilateral type I error of 5% was established. All analyses were performed using R version 3.3.1.

Results
Data from a total number of 234 eyes (191 patients) with nAMD was collected in the FRB online platform, corresponding to more than 100.000 data elds, VA measurements from 2.815 outpatient visits and more than 2.114 treatment episodes.

Discussion
This study reports the successful implementation of the FRB platform in the intravitreal therapy unit of a single tertiary referral centre, an online audit tool that allowed structured data collection and provided long-term outcomes of a real-world cohort of patients receiving VEGF inhibitors for nAMD. This system enabled the e cient capture of high-quality data, allowing us to benchmark our real-life clinical results and compare them with those reported in randomized clinical trials 5 .
Real world studies on large-scale population-based data and phase IV studies, which are carried out after regulatory approval is given, are becoming increasingly important nowadays. They track patients for longer and ensure that results obtained in clinical trials extend to the general population, evaluating endpoints as long-term e cacy, safety, quality of life and cost-effectiveness. This is especially important in anti-VEGF therapies because they are usually long-term treatments and nAMD causes a strong negative impact on quality of life [11][12] .
Anti-VEGF drugs represent a large proportion of the health system expenditure all around Europe 13 . Previous reports analysing nAMD patients management with intravitreal anti-VEGF in routine clinical practice in our country concluded that this was highly variable between different centres and distant from recommended European guidelines 14 . New technologies give us the opportunity to enhance self-auditing practice in Spain and re-evaluate outcomes, to broaden our knowledge and to optimize the management of our patients. Nonetheless, this could be considered of great importance given the paucity of such information in southern Europe countries, which may socially and economically differ from its northern neighbours.
The results observed in our series present similarities and differences with other real world study cohorts. As expected, we found a higher prevalence of nAMD in women, with a 1.6:1 female-to-male ratio, in line with previously published studies 15 . However, the baseline VA was substantially better in our cohort than other real-world studies (59.7 vs 55-56 logMAR letters) 12,16 Usually, patients with better baseline VA have a better prognosis and tend to maintain VA in time 11 , but can also present with less room for improvement (ceiling effect) as per VA change. No patient was excluded according to pre-treatment status (as done in most reports 8-10 ) but a sub-analysis was made according to these criteria. Mean nal VA was 61.9 logMAR letters at 12 months and 62.4 at 24 months, with signi cant differences between TN and PT eyes at 24 months (66 TN, 59.5 PT, p = 0.032). VA change favoured TN eyes at 12 and 24 months, with gain of + 4.0 and + 2.4 letters respectively (p = 0.01; 0.003). These results are comparable to other international cohorts described using FRB platform, such as the Australian 12 , including 1198 eyes under TAE regimen only in which mean VA increased from 56.5 logMAR letters at baseline to 61.8 at 24 months.
In the UK, a real-wold multicentre study was performed to study the outcomes of ranibizumab under a PRN approach, obtaining poor results: a total of 12951 eyes with baseline VA 55 logMAR letters and VA gain of -2.0 letters at 24 months 17 . More recently, another UK group reported results using a xed bimonthly a ibercept treatment regime that achieved better outcomes: 1180 eyes with baseline VA 56.3 logMAR letters and VA gain of + 2.3 letters at 24 months 18 . The differences observed between TN and PT eyes in our study cohort, with greater VA gains at 12 and 24 months in the TN group, could be explained by the natural history of the disease after anti-VEGF treatment: initial VA improvement that usually peaks after the loading dose, then stabilizes for a variable period, and gradually decreases over time. 19 Regarding treatment frequency, no signi cant differences were observed in the number of injections at 12 months (7 vs 7) or 24 months (11 vs 12) between both study subgroups. These results are similar to the outcomes reported in randomised clinical trials such as CATT, with a total of 7 injections of ranibizumab in the rst year 20 , and comparable to other real-world outcomes such us UK cohort with median number This study has several strengths and limitations. Using the FRB database allowed a detailed analysis of nAMD patients treated in routine clinical practice, re ecting real-world utilisation of both anti-VEGF drugs for the treatment of nAMD. Furthermore, different treatment regimens such as xed bimonthly dosing and TAE were included, being selected at the discretion of the treating physician 21 . Like all observational studies, it has low internal validity, but it is still meaningful, because this is how clinical practice is actually delivered and re ect routine clinical care. Also, results from long-term observational studies may be affected by external factors, for example, loss of follow-up.

Conclusions
In conclusion, online platforms such as FRB enhance self-audit practices and facilitate rapid analysis of intravitreal therapy units results, allowing comparisons with other national and international cohorts. The results reported in this study support the use of these systems for auditing purposes in centers transitioning from paper notes to computer systems or unstructured EMR softwares. Further research is warranted with larger cohorts and longer follow-up to determine whether the positive results herein reported are sustained over time. Availability of data and materials: The datasets analysed during the current study are available from the corresponding author on reasonable request.
Competing interests: Gillies and Barthelmes are inventors of the software used to collect the data for this analysis. Zarranz-Ventura is a consultant for Novartis and Bayer, and a grant holder from Novartis.
Funding: The Fight Retinal Blindness! Project is supported by a grant from the Macular Disease Foundation Australia and unrestricted educational grants from Bayer and Novartis.
Authors' contributions: APC collected and interpreted data and wrote the manuscript. MFR and RPCM collected data and edited the manuscript. VN performed the statistical analysis. DB and MCG are inventors of the software and edited the manuscript. JZV wrote and edited the manuscript, is the main director of this study and the national coordinator of the FRB project Spain. All authors read and approved the nal manuscript.