From the Eye to the Brain
Are stratification studies the key to identifying patients at risk of dementia?
Pearse Keane, Siegfried Wagner | | Quick Read
Estimates suggest that 50 million people were living with dementia in 2017. With the progressive aging of the population, the number is predicted to reach 75 million by 2030. Yet it has been noted that 50 to 80 percent of cases remain undiagnosed in high income-countries. Why? Part of the issue lies with the logistics of making a diagnosis. The gold standard for the most common form of dementia, Alzheimer’s disease, has classically been neuropathological confirmation, post-mortem. Research into newer techniques, such as amyloid positron emission tomography (PET) scanning and cerebrospinal fluid analysis, has supported their utility as potential biomarkers; however, these tests are invasive, expensive, and not pragmatic on a large scale. Could assessment of the neurosensory retina – derived embryologically from the same tissue – be the answer?
The impact of Alzheimer’s disease on ocular anatomy was first convincingly demonstrated in 1986 when widespread axonal degeneration was found in the optic nerve of eight recently deceased patients with the disease. Though subsequent work showed some evidence of an association between retinal venous diameter and Alzheimer’s disease, true relationships only began to emerge when cross sectional measurement of the retinal nerve fiber layer became possible. In particular, the introduction of OCT and the establishment of large prospective cohort studies that incorporate ocular imaging have demonstrated that people with dementia show thinning of the retinal nerve fiber (RNFL) and ganglion cell-inner plexiform layers. However, thinning of the inner retina is not just a feature of prevalent dementia; rather, it may be predictive of its development. Last September, two large prospective studies – UK Biobank and the Rotterdam Study – revealed that participants with thinner RNFL were significantly more likely to develop cognitive decline and dementia.
However, as noted by the Rotterdam Study team, prediction modeling to identify those individuals at risk of developing dementia has not yet been feasible because of the small number of cases in prospective cohorts, which generally recruit healthy middle-aged volunteers. Moreover, it remains unclear whether these relationships are generalizable to the non-Caucasian population. To address these observations and more, we designed AlzEye: a large-scale record linkage dataset combining all forms of retinal imaging captured over the last ten years at Moorfields Eye Hospital – the largest ophthalmic center in Europe and North America – with the national Hospital Episode Statistics (HES) database. HES is a centralized data warehouse, overseen by the UK’s National Health Service (NHS) Digital arm, which contains details of all hospital admissions, emergency attendances and outpatient appointments in England. In the AlzEye Study, we have linked approximately 2.3 million images of more than 250,000 patients across a diverse population of varying ethnicity and socioeconomic status with diagnostic codes – including dementia. The approach will provide an estimated 5,000 cases of incident dementia. Not only will AlzEye allow the development and validation of traditional statistical models, it will also provide an opportunity to employ cutting-edge artificial intelligence techniques for the potential for prediction. Leveraging the expertise at the Centre for Medical Image Computing of University College London, AlzEye aims to provide a much-needed risk stratification tool to identify people at risk of dementia.