Update on Retinal Research With the IRIS Registry

Key insights into a number of retinal conditions.


The IRIS (Intelligent Research in Sight) Registry was established in March 2014 by the American Academy of Ophthalmology (AAO) to provide a clinical data repository from ophthalmology practices in the United States. The goals were to extract clinical measurements for quality improvement, scientific, clinical, and policy purposes. Initially, the most important of these goals was to help members improve their quality of care with actionable feedback, and then to help them meet government mandates to assess quality in a pay-for-performance system, called at the time the Medicare Physician Quality Reporting Initiative (PQRI) and later the Merit-Based Incentive Payment System (MIPS).

Critical to the formation and maintenance of the IRIS Registry is the protection of patient identifiers according to the Health Insurance Portability and Accountability Act (HIPAA), retention of ownership of individual practice data by the participating physicians without an accessible link of the data to individual physicians or their practices without their permission, and electronic capture from electronic health record (EHR) systems to ease data entry. The aggregated de-identified data is owned by the AAO and the effort has been almost entirely funded until recently by the organization. The AAO leadership had decided that no industry or government funds would be used in the construction of the IRIS Registry to avoid any question of conflict of interest or claim to data access. The only monetization of the data would be in keeping with the goals stated above and with the AAO’s mission.1

The IRIS Registry software maps to the EHR systems used by almost all ophthalmic practices in the United States using a systems integrator installed on a practice’s server containing the practice’s EHR database. The patient and doctor de-identified data are collected by the IRIS Registry automatically in almost real time and about 70% of ophthalmic practices and 15% of the US population are now included.

Retinal research with the IRIS Registry can look at disease prevalence, different treatment approaches, and clinical outcomes. Certain factors currently can limit data acquisition. Diagnosis coding and/or data storage in EHR systems may not have sufficient granularity to identify less common disorders, such as some uveitic or genetic conditions. Physicians may also not document or IRIS cannot adequately map certain EHR systems to access key data points. For example, we found this problem in our study described below2 when we sometimes could not identify the eye the physician injected, thus we had to exclude several thousand patients from our data analysis. Topical or systemic medication data may not be adequately documented in some EHR systems, and presently the IRIS Registry does not collect diagnostic images, such as fundus photographs or fluorescein angiograms. For optical coherence tomography (OCT), central macular thickness is recorded but not more detailed findings.

So far, 4 large retinal research studies have been reported evaluating the real-world treatment of major retinal conditions, such as age-related macular degeneration (AMD), diabetic retinopathy (DR), macular hole (MH), and epiretinal membrane (ERM) using the IRIS Registry. These are highlighted below.


Prethy Rao, MD

Anti-VEGF agents are the cornerstone of AMD therapy in the United States and worldwide. Numerous landmark trials have compared anti-VEGF agents in neovascular AMD (nAMD). However, there are minimal data that evaluate the relationship of all 3 current anti-VEGF agents (bevacizumab, ranibizumab, aflibercept) head-to-head in real-world US clinical practice on a large epidemiological scale. The purpose of this study was to evaluate the real-world visual acuity (VA) of nAMD patients treated with a single anti-VEGF agent only (bevacizumab only, ranibizumab only, or aflibercept only) for 1 year without switching drug therapies (Figure 1).3

Figure 1. Visual acuity by drug type and visual gain over 1 year for AMD patients. Figure reprinted with permission from Rao et al. Real-world vision in age-related macular degeneration patients treated with single anti-VEGF drug type for 1 year in the IRIS Registry. Ophthalmology. 2018;125(4):522-528.3

In this study of 13,859 patients with 84,828 injections who were followed for 1 year, there was no statistical difference in the final or mean change in logMAR VA among all 3 groups after multivariate adjustment (age, baseline logMAR VA, diabetes mellitus, posterior vitreous detachment, number of injections, race, insurance type, primary open angle glaucoma, number of injections, mean follow-up). The mean change in logMAR VA was -0.048 (0.44 standard deviation [SD]) for bevacizumab, -0.053 (0.46 SD) for ranibizumab, and -0.040 (0.39 SD) for aflibercept (P=.46).

In a subset analysis, patients were divided into those with a VA loss of ≥3 lines and those with a VA gain of ≥3 lines. Interestingly, aflibercept exhibited a 1.25 increase in log odds of achieving a ≥3 line VA loss compared to bevacizumab in the multivariate model. However, there were no differences in risk between ranibizumab and bevacizumab.

Patients with worse baseline VA had a 4.8-fold increase in log odds of achieving ≥3 line VA gain. Those that exhibited ≥3 line VA gain had a greater number of injections (6.6 [0.05 SD]) compared to those with no change or ≥3 line VA loss (6.1 [0.03 SD] and 5.9 [0.06], P<.0001, respectively). Similarly, there was a 1.1-fold increase in log odds of achieving a ≥3 line VA gain for every unit increase in number of injections, irrespective of drug subtype. Further subset analyses are warranted to evaluate treatment protocols and OCT characteristics.

Analysis of UK Electronic Medical Record to Improve Retina Therapies

Usha Chakravathy, MD, PhD, FRCS, FRCOphth

The UK EMR is a robust repository of data. We benefit hugely from being able to look at real-world outcomes from all the different interventions that we now perform in retina. We also have the ability to review information from past years and see patterns of outcomes over time.

The Relationship Between Dry and Wet AMD

We’ve always had the view that both geographic atrophy (GA) and neovascular AMD were unique phenotypes. But, in fact, we know that is not the case because the risk factors and risk estimates are the same for both conditions. So, using the EMR, we showed that when you looked at very large numbers of people referred for a variety of reasons, in the older age group, you see that GA is often the first manifestation and that neovascularization supervened on this picture of GA.1

The bottom line is that there are microscopic areas of atrophy. Sometimes these coalesce into big areas of atrophy, and other times, patches of neovascularization arise because the small areas of atrophy are trying to overcome the lack of nutrition and oxygen. We dissected some of these relationships, and we used the EMR system to show the temporal relationships between these late-stage manifestations to show that both conditions are basically the same disease but that there are different manifestations.

This study included 10 centers with some 90,000 patients. There was bilateral GA in about 2,000 and we had early AMD in both eyes in about 30,000 people. Also, we had people who had wet AMD in one eye and early AMD in the fellow eye, as well as people who had GA in one eye and early AMD in the fellow eye. We looked at these permutations and combinations and worked out the progression rates for new-onset GA and for wet AMD.

Analyzing Outcomes for Anti-VEGF

With more than 10 years of data and because of the very large numbers included in this dataset, we were also able to show that for those people who are receiving treatments from wet AMD despite benefitting considerably a proportion continue to lose sight. Unfortunately, this information shows that there is a huge unmet need.2

The outcomes from the treatment of diabetic retinopathy after introduction of anti-VEGF therapies is another area that we studied using EMR systems. By combining data from some 20 sites, we showed that using the anti-VEGF treatments with diabetic macular edema was beneficial, just as it was in the clinical trials, but the proportion showing improvements was not as high as was observed in the clinical trials. It could be that the timing was such that many patients in the study had chronic disease and had been waiting for treatment. So, there may have been a slightly worse outcome. We now have about 6 years’ worth of data and, of course, we’ll be looking at that in greater detail.

Also, because of the way in which diabetic retinopathy severity is automatically captured and recorded in the EMR systems, we’re able to see how baseline severity determines risk of progression.3 And to our delight and amazement, we found that the data are so robust that the findings almost replicate what was shown in the DRCR Network studies from a few years ago where you have survival curves showing progression. You could almost superimpose our curves collected from routine clinical examinations onto those data.

Injections and the Posterior Capsule

We also showed that intravitreal injection increases the risk of posterior capsule rupture when a patient has a cataract removed. What we observed was that with more than 10 injections, the risk of suffering a posterior capsule rupture was approximately 2.5 times greater than with no injections.4 This is interesting because it hadn’t been reported before, and because the number of patients receiving anti-VEGF injections who also need cataract surgery is increasing. We also found that the probability of developing diabetic macular edema increases at the point of cataract surgery. These are the types of things we can spin off from interrogating the EMR systems.

Features of a Useful EMR

We think the UK EMR system is extremely good for several reasons. One, it is intuitively structured so it’s easy to use. There are certain key fields that one can now get quickly without entering the data. For example, the visual acuity absolutely has to be entered, and it’s all by drop-down fields and by eye. And in the diabetic retinopathy module, images pop up as part of the EMR system, so that when they’re looking at our patients, we can then say what are the features that are present in the diabetic eye, and that then allows us to give, to get an automated calculation of how severe the retinopathy surface is.

Diagnostic drop-down fields are key. That makes the data easy to code when it’s exported into an Excel file, whereas if the system is searching by free text, it could be an absolute nightmare. There were instances for the GA study, however, where free text was preferred. We wrote algorithms to extract those free-text entries. But we did do a data validation on those free-text entries and found the negative predictive values were in excess of 90%, which suggests that the clinician, even if they were entering free text they were entering the data correctly.


  1. Chakravarthy U, Bailey CC, Johnston RL, et al. Characterizing disease burden and progression of geographic atrophy secondary to age-related macular degeneration. Ophthalmology. 2018;125(6):842-849.
  2. Lee AY, Lee CS, Butt T, et al; UK AMD EMR Users Group. UK AMD EMR USERS GROUP REPORT V: benefits of initiating ranibizumab therapy for neovascular AMD in eyes with vision better than 6/12. Br J Ophthalmol. 2015;99(8):1045-1050.
  3. Lee CS, Lee AY, Baughman D, et al; UK DR EMR Users Group. The United Kingdom Diabetic Retinopathy Electronic Medical Record Users Group: Report 3: Baseline retinopathy and clinical features predict progression of diabetic retinopathy. Am J Ophthalmol. 2017;180:64-71.
  4. Lee AY, Day AC, Egan C; United Kingdom Age-related Macular Degeneration and Diabetic Retinopathy Electronic Medical Records Users Group. Previous intravitreal therapy is associated with increased risk of posterior capsule rupture during cataract surgery. Ophthalmology. 2016;123(6):1252-1256.


Jeffrey Willis, MD

Diabetic macular edema (DME) is one of the leading causes of visual impairment among working-age adults. Treatment options include anti-VEGF therapy, laser, and steroids. When DME is left untreated, individuals are at risk for long-term vision loss. At present, there are few national-level studies that describe how ophthalmologists in the United States manage treatment-naïve DME patients. In particular, there are few data on the anti-VEGF treatment burden and the associated visual acuity outcomes with anti-VEGF treatment. To address this knowledge gap, we conducted a study using data from the IRIS registry to characterize the real-world treatment patterns surrounding incident DME in the United States.

Based on data from the IRIS Registry, we identified 13,410 individuals with treatment-naïve newly onset DME in the United States diagnosed between July 2013 and March 2016.4 Analysis of treatment patterns showed that during the initial 28 days after diagnosis, most individuals (74.5%) were observed.

When treatment was given, anti-VEGF (15.6%) was the preferred modality. Among these individuals treated with anti-VEGF, 71.3%, 17.1%, and 11.6% received bevacizumab, aflibercept, and ranibizumab, respectively. Furthermore, among these that were started on anti-VEGF within 28 days of diagnosis, the mean number of injections administered over the 1-year follow-up period was 4.3. More specifically, 50% received 3 injections or fewer over the 1-year follow-up period. Finally, analysis of visual acuity data showed a trend where visual acuity gains 1-year after the initial diagnosis was greater among those that received a greater number of injections. The logMAR gain was 0.11 and 0.06 among those that were treated with 1-5 injections and ≥6 injections, respectively (note: 0.02 logMAR is approximately equal to 1 ETDRS letter). In order to understand the impact of anti-VEGF injections on visual acuity in the real-world setting, further research is warranted.


Elizabeth Atchison, MD

Previous studies have found a high rate of sustained intraocular pressure (IOP) rise after anti-VEGF injections, as high as 23% at 2 years in the MARINA and ANCHOR studies.5 The rate of such a rise outside of a clinical trial setting was unknown. Using the IRIS Registry, we looked at patients who received a single type of anti-VEGF injection for nAMD, DME, or retinal vein occlusion in their right eyes who did not have any anti-VEGF treatment in their left eyes. Thus, the left eyes served as a natural control. At least 1 year of follow-up was required. Patients who received steroids any time in the year before or during the study period were excluded. The main outcome measures were change in IOP from baseline and rate of clinically significant IOP increase (an increase of at least 6 mmHg that leads to an IOP greater than 21 mmHg). Results were subdivided by type of anti-VEGF agent, treatment diagnosis, and number of anti-VEGF injections received (Figure 2).

Figure 2. Rate of clinically significant intraocular pressure increase in treated eyes (solid) vs fellow eyes (hollow) by group and drug for any number of injections. Asterisks indicate a statistically significant difference between treated eyes and untreated fellow eyes. Figure reprinted with permission from Atchison et al. The real-world effect of intravitreous anti-vascular endothelial growth factor drugs on intraocular pressure an analysis using the IRIS Registry. Ophthalmology. 2018;125(5):676-682.

There were 23,776 patients in the study who were followed an average of 1.8 years during which time they received a mean of 8 injections. By diagnosis: 76% were AMD, 11% were RVO, and 12% were DME. By drug: 56% received bevacizumab, 25% ranibizumab, and 9% aflibercept. The average change in IOP for treated eyes was a decrease of 0.9 mmHg and 0.2 mmHg for untreated eyes. This decrease was statistically significantly larger for aflibercept and bevacizumab but there was no significant difference in IOP for treated and untreated eyes for ranibizumab. The rate of clinically significant IOP increase was 2.6% for treated eyes and 1.5% for untreated eyes. The rate of clinically significant IOP rise was statistically significantly higher in eyes treated with ranibizumab and bevacizumab but not aflibercept. Less than 1% of treated and untreated eyes required a glaucoma procedure. Patients who had a clinically significant IOP increase had a history of glaucoma diagnosis 3 times more often than those without a clinically significant IOP increase (2.1% vs 0.7%).


D. Wilkin Parke III, MD

The reported complication rate after vitrectomy for epiretinal membrane or macular hole has varied significantly, even in the last decade of small gauge surgery. The anatomic success rate with macular hole closure has ranged from 85 to 100%. The rate of postoperative retinal detachment has been reported as 1% to 14% for macular hole surgery and 1% to 7% for epiretinal membrane surgery.

We used the IRIS Registry to investigate additional ophthalmic surgery within 1 year after vitrectomy for macular hole or epiretinal membrane.6 All vitrectomies for macular hole or epiretinal membrane between January 1, 2013 and June 30, 2017 were analyzed. Eyes with additional surgical indications like vitreous hemorrhage, retinal detachment, or proliferative diabetic retinopathy or concurrent surgical procedural codes like retinal detachment repair were excluded. After applying the exclusionary criteria, 41,475 eyes underwent vitrectomy for macular hole and 73,219 eyes underwent vitrectomy for epiretinal membrane.

In the macular hole vitrectomy group, 6.8% of eyes underwent a second surgery that was not cataract-related. The most common second procedures were macular hole repair (4.6%), retinal detachment repair (1%), and complex retinal detachment repair (1%). In the epiretinal membrane vitrectomy group, 5.5% required a second noncataract procedure, the most frequent of which were vitrectomy with membrane stripping (1.4%), complex retinal detachment repair (1.4%), macular hole repair (1.3%), and retinal detachment repair (1.1%). The final best corrected visual acuity was worse in those eyes that required a second, noncataract surgery within 1 year (logMAR 0.62) vs those eyes that did not undergo additional surgery (logMAR 0.51, P<.001) (Figure 3).

Figure 3. Protocol for identification of macular hole and epiretinal membrane cases. Figure reprinted with permission from Parke and Lum. Return to the operating room after macular surgery IRIS Registry analysis. Ophthalmology. 2018;125(8):1273-1278.6


This is just the beginning of the real-world research that can be performed for vitreoretinal diseases with the IRIS Registry. In the future, studies will have expanded opportunities with improvements in data entry and extraction and the addition of digital downloads from ophthalmic cameras and other devices such as OCT scanners and visual field devices. In time, the information in the IRIS Registry can be combined with data from other medical registries as well as outputs from the operating room, laboratories, and genetic repositories. The IRIS Registry is ideal for performing phase 4 studies of new drugs and treatments as required by the FDA. Ultimately, we may be able to access in real time the outcomes of patients who have similar vitreoretinal conditions to those in our practice, and thus allow us to optimally personalize our patients’ treatment approach. RP


  1. Parke DW, Rich III WL, Sommer A, Lum F. The American Academy of Ophthalmology’s IRIS Registry (Intelligent Research in Sight clinical data): a look back and a look to the future. Ophthalmology. 2017;124(11):1572-1574.
  2. Atchison EA, Wood KM, Mattox CG, Barry CN, Lum F, MacCumber MW. The real-world effect of intravitreous anti-vascular endothelial growth factor drugs on intraocular pressure: an analysis using the IRIS Registry. Ophthalmology. 2018;125(5):676-682.
  3. Rao, P, Lum F, Wood, K, et al. Real-world vision in age-related macular degeneration patients treated with single anti-VEGF drug type for 1 year in the IRIS Registry. Ophthalmology. 2018;125(4):522-528.
  4. Willis JR, Chia Y, Morse LS, et al. Treatment patterns for diabetic macular edema in the united states: analysis of the IRIS Registry (Intelligent Research In Sight). Abstract presented at: The American Academy of Ophthalmology 2017 annual meeting; November 11-14, 2017; New Orleans, LA.
  5. Bakri SJ, Moshfeghi DM, Francom S, et al. Intraocular pressure in eyes receiving monthly ranibizumab in 2 pivotal age-related macular degeneration clinical trials. Ophthalmology. 2014;121(5):1102-1108.
  6. Parke DW 3rd, Lum F. Return to the operating room after macular surgery IRIS registry analysis. Ophthalmology. 2018;125(8):1273-1278.