Macuject, software developed in Australia, uses machine learning artificial intelligence for analysing macular optical coherence tomography images to find and measure intraretinal and subretinal fluid volumes. In doing so, the aim is to guide treatment decisions which ultimately will help reduce the emerging issue of preventable vision loss due to the under-treatment of neovascular age-related macular degeneration (nAMD)
Macular Month coincides with my migration from London to Melbourne to work as a retinal specialist in May 2006. My arrival happily coincided with the greatest leap in medical retinal practice since the introduction of retinal photocoagulation in 1956 – intravitreal anti-VEGF injections for neovascular agerelated macular degeneration (nAMD).
…examination of the OCT scans of patients from the CATT study also demonstrated that intraretinal fluid was often missed by retinal specialist clinical investigators, even though they were arguably looking harder…”
It is no exaggeration to state that medical retina has been at the leading edge of medicine, let alone ophthalmology, since then – it has been both the first specialty for which so many new technologies have been developed and in which they have been applied. Examples include gene therapy, stem cell therapy, high resolution in vivo imaging, minimally invasive and robotic surgery, and bionics.
Furthermore, many of the earliest and most significant clinical applications of artificial intelligence (AI) have addressed medical retinal conditions. Indeed, the first FDA approved autonomous AI software in medicine (IDx) is a screening tool for diabetic retinopathy in the community. Within medicine, ophthalmologists appear to be the specialists most likely to embrace and use AI in their daily practice; more so than radiologists, oncologists or dermatologists.
While all of these technological advances have enormous potential to improve the health and lives of millions of people in the future, the prosaic reality today is that intravitreal anti-VEGF injections for nAMD have generated the greatest transformation of all. And yet there is a barrier to delivering this therapy that is stopping us from achieving even more.
Global under-treatment is the most significant problem in nAMD therapy today. Just a few years ago, I founded Macuject to address this emerging issue, which is resulting in easily preventable blindness, even in the United States, Europe and Australia.
THE BURDEN OF UNDER-TREATMENT
Lucentis (ranibizumab) was approved by the Therapeutic Goods Administration (TGA) in 2006 and became available on the Pharmaceutical Benefits Scheme (PBS). Since then, we have seen the expectations of clinicians and patients alike shift from nAMD being a sentence to permanent blindness to the expectation of almost normal vision. Eylea (aflibercept) became available in 2012 and bolstered our ability to meet these expectations.
In Australia, early detection and prompt referral of patients by optometrists to ophthalmologists has meant that treatment usually starts so early in the disease process that it is initiated before permanent blinding structural changes in the macula have occurred.
Furthermore, studies have shown that more injections are administered in Australia, on average, in the first two years than almost any other country. This is likely to relate to the early and highly commendable PBS decision to pay for anti-VEGF drugs, even though they are priced at over AU$1300 per injection and the Medicare fee for injections is approximately $260. That decision was likely to have been informed by a health economic analysis demonstrating unequivocal benefits, even with monthly injections, which was the protocol used in the registration trials.
With continued treatment we have seen longterm vision improvement and maintenance of good functional vision. Indeed, I have patients who have had over 150 injections in each eye over nearly 15 years and maintain a visual acuity of 6/7.5 in both eyes.
Additionally, unlike even a decade ago, there is no longer any need to dissuade nAMD patients from cataract surgery – many achieve 6/6 vision postoperatively.
However, there are still many patients who lose vision over time, even when treated strictly according to proven protocols. This is usually related to the development of geographic atrophy (dry macular degeneration), for which there is currently no known treatment nor preventative measure available. In other words, a loss of vision is no longer expected and is generally caused by a condition other than nAMD.
Unfortunately, the positive nAMD patient outcomes achieved in Australia, which closely match those in pivotal anti-VEGF trials and their follow-up studies, are not the global norm. Indeed, they are probably the exception; several studies have shown that visual outcomes in real-world practice at two years or so after the beginning of treatment are poor, with patients returning to their initial presenting vision or worse. There are differences between countries, with Italian patients faring worse than the United Kingdom’s, for example, but the common factor seems to be most obviously related to the number of injections administered in the first two years of treatment and thereafter.
In spite of encouraging Australian data, it is also known that, even here, up to 50% of patients cease treatment within three years of starting, which is likely to result in severe loss of vision. In the US, the societal cost of under-treatment has been estimated at US$35 billion over five years.
Although there is some controversy regarding whether poor outcomes relate solely and directly to fewer injections, it is widely held that under-treatment is strongly associated with poor visual outcomes. Under-treatment therefore warrants intense scrutiny globally, as understanding and addressing the causes could lead to the prevention of extensive avoidable blindness.
FACTORS CONTRIBUTING TO UNDER-TREATMENT
The huge adoption of optical coherence tomography (OCT) scanners in optometry practices has been revolutionary for patients, both in helping optometrists pick up early disease but also in assisting ophthalmologists to triage referrals. The effect has been to ensure that those who need to be seen quickly by an ophthalmologist do so in a timely fashion.
However, there are many stages in the process of patients undergoing their longterm treatment at which under-treatment can manifest.
Many patients need assistance when visiting their treating specialist, whether it is in terms of transportation or other forms of support. While many nAMD patients are not of working age, their carers often are; for someone in a full-time job, taking one day off every month is very difficult, to say the least. Patients can see themselves as a burden to their carer, sometimes either not informing them about the next appointment or simply missing it to reduce the need for the carer to take time off or make alternative arrangements.
The tyranny of distance can play a significant role. There are several schemes in Australia, some of which are public and others organised by private clinics, which aim to address this issue by facilitating transport for patients to their retinal specialists. Personally, I used to fund a coach that would transport patients from rural Gippsland for ongoing regular injections, sometimes for an eight-hour round trip. This approach has been replaced more recently by a more community-based approach with Red Cross and local council volunteers assisting patients with their travel arrangements.
The need to travel significant distances is, in part, a reflection of the lack of retinal specialist expertise outside major cities. In Australia, about 10% of ophthalmologists are retinal specialists and they perform approximately 80% of injections. The remaining injections are performed by general ophthalmologists, particularly in suburban or regional practices. Retinal specialists tend to be based near CBDs, in Melbourne and Sydney at least, and visit regional areas less frequently. In the United States, a study has shown that adherence and persistence with treatment is significantly lower when patients live more than 50 miles away from the clinic in which the injections are delivered.
It is difficult to determine whether the local needs of patients are being fully met in regional areas. MBS statistics for the rates of injections per 100,000 people are only readily available at a state-by-state level, not by postcode. Even so, in 2019 (quoted here to avoid confounding COVID-19 effects) rates of injection per 100,000 varied significantly between the states: 298 (NT), 1,536 (Vic), 1,548 (WA), 1,570 (Qld), 1,904 (ACT), 2,082 (SA), 2,785 (NSW) and 3,254 (Tas), with a national mean of 2,013/100,000. The precise reasons for these differences are not known, but may include population demographics, geographical distribution and readiness of clinicians to provide an injection service. In some countries, such as the United Kingdom and New Zealand, nurses are delivering intravitreal injection services and there has been discussion about the role of optometrists too. It is clear that local access to high quality injection services is needed throughout.
Interestingly, on the potentially controversial subject of doctors’ skillsets, there are no studies that have looked at the difference in the OCT reading skills of retinal specialists vs. general ophthalmologists, nor at the level of their knowledge and application of treatment protocols. It is accepted wisdom, however, that there is likely to be a difference, but this may well either narrow or disappear as general ophthalmologists become increasingly experienced.
Regardless of who performs the injections, however, the treatment burden for patients and clinicians alike is a significant problem. The American Society of Retinal Specialists’ Global Trends in Retina survey (2018) determined that 73.2% of US retinal specialists prioritised solving treatment burden over longer-acting drugs (56.3%).
The US National Institute of Health’s CATT study confirmed conclusively that a monthly regimen of treatment with anti- VEGF injections significantly improved and maintained vision gains over the first two years of treatment. In addition to showing that Avastin had a non-inferior outcome compared with Lucentis, it also showed that an as required, or pro re nata (prn), approach was inferior unless patients were reviewed on a monthly basis. However, monthly reviews negate the potential benefit that drove adoption of a prn approach in the first instance, i.e., the reduction of treatment burden for clinicians, patients and their carers.
Subsequent examination of the OCT scans of patients from the CATT study also demonstrated that intraretinal fluid was often missed by retinal specialist clinical investigators, even though they were arguably looking harder than in their usual clinical practice. Furthermore, fluctuation of intraretinal fluid volumes over a patient’s treatment was associated with significantly worse visual outcomes. Evidently, improved detection of fluid and maintaining a ‘dry’, (intraretinal) fluid-free macula should result in improved outcomes.
The treat and extend regimen (TER), in which injections are initially given monthly, followed by an iterative extension of the interval between injections, achieves these goals in the real world. Guided by qualitative inspection of OCT scans for the presence of exudative fluid, the TER aims to establish the longest fluid-free interval that maintains a ‘dry’ macula. Adopted by Australian ophthalmologists (retinal specialists and general ophthalmologists alike) over a decade ago, the TER is now globally accepted as the optimal approach to managing nAMD patients, by maximising visual outcomes with the fewest injections. An Australian study has also shown that vision loss and recurrence of choroidal neovascularisation occurs in 91% of patients if untreated for more than 12 weeks, so many clinicians impose an upper limit of 12 weeks or so on intervals.
WHAT CAN BE DONE?
Pragmatic measures are required for the problem at hand while we await longer acting drugs, devices that prolong the action of injections, and alternatives to intravitreal injections. In spite of proven treatment protocols such as TER, the issues of treatment burden, adherence by clinicians to protocol, adherence by patients to doctors’ instructions, and persistence with treatment, continue to be significant issues globally. Addressing these issues from both the perspective of the clinicians, and their patients and families, seems an appropriate approach. Recent publication of the results of a patient support program have demonstrated significant benefits through engaging patients directly.
It is with these issues in mind that I started on my journey to develop Macuject. Our AI-enabled clinical decision support software aims to provide workflow efficiencies and improve adherence and persistence. Supported by a Commonwealth BioMedtech Horizons 2 grant, the growing Macuject team has been refining the software to streamline and assist in the decision-making required at each patient visit. Macuject’s software consists of two components – machine learning AI for analysing macular OCT images to find and measure intraretinal and subretinal fluid volumes.
These values, in the context of patients’ clinical history (for up to 15 years and counting…), feed into a real-worldadapted TER that is configured by treating ophthalmologists for their treatment preferences. The output is a recommendation for treatment that is, in essence, that which the doctor would make, albeit much faster. The OCT findings and decision-making process are all clearly seen by the doctor and every aspect of the recommendation and other key findings by the AI can be overridden by the doctor, who maintains complete control throughout.
Macuject has been in beta testing with a select group of ophthalmologists for several months now and their feedback is being integrated into the next version, to be released in June. Preliminary feedback is positive, with workflow improvements in some cases and stricter adherence to TER in others. We are looking to expand the usage of Macuject over the course of 2021.
Analysis of my own patients’ adherence statistics going back to 2006, through the use of Macuject software, has yielded some interesting findings with respect to the effects of non-adherence by patients on their visual outcomes. Early data suggest that several novel metrics of patients’ adherence are highly statistically significantly associated with the degree of vision gained with treatment, the time taken to achieve this and subsequent loss of vision. These findings, some of which are counterintuitive, are currently being written up for submission for publication.
We envisage introducing a patient engagement arm to our software, which will have the distinct advantage of being personalised to patients’ specific stages of treatment and visual status. We, at Macuject, are excited by the huge international opportunities we see to help reduce preventable blindness by standing on the shoulders of the revolution that began 15 years ago with intravitreal anti- VEGF therapy.
Dr Devinder Chauhan is the Founder and Medical Director of Macuject Pty Ltd. He is a retinal ophthalmologist with over 20 years of experience, and specialises in the medical and surgical treatment of retinal, macular and vitreous conditions.