New research conducted by investigators from the University College London (UCL) and Moorfields Eye Hospital—and published in Neurology—is the first of its kind to potentially identify retinal biomarkers of Parkinson’s disease (PD) using artificial intelligence (AI) several years before an actual diagnosis.
Let’s start with some background.
Eyecare practitioners (ECPs) routinely use optical coherence tomography (OCT) to conduct retinal scans. To note, OCT is the only non-intrusive way to quickly scan neuroretinal tissue.
As of late, researchers have considered OCT for depicting diagnostic and prognostic data in neurological disorders (Alzheimer’s disease, multiple sclerosis (MS), and schizophrenia); in fact, prior studies have used OCT to unveil potential (but inconsistent) findings associated with PD.
These retinal signatures of systemic disease—called “oculomics”—are increasingly being identified via a combo of high-resolution ophthalmic imaging (OCT) and modeling strategies such as machine learning AI.
Now this research.
In this cross-sectional analysis, investigators from the University College used OCT to determine the inner retinal anatomy of PD patient data from the AlzEye data cohort set, which was then repeated using the wider UK Biobank (UKBB) database from the UKBB study.
Refresh me on these two studies.
The AlzEye, a retrospective cohort study, linked systemic disease data from hospital admissions with the world's largest single institution retinal imaging database (the National Health Service [NHS]) of patients ages 40+ to create a diverse, large-scale cohort.
The UKBB study is a prospective, population-based, multicenter cohort of an estimated 500,000 healthy individuals (ages 40-69 years).
To note, a subset of 67,000 UKBB participants underwent an ophthalmic exam that included retinal imaging (OCT) at their initial assessment visit.
Gotcha. And the hypothesis?
Investigators hypothesized that PD patients would exhibit three characteristics where retinal thinning was noted in the following:
- Macular ganglion cell inner plexiform layer (GCIPL)
- Macular retinal nerve fiber layer (mRNFL)
- Inner nuclear layer (INL)
The finding’s difference would be associated with the incident disease (the number of new diagnosed cases).
So how was this OCT data studied?
The researchers used linear mixed effects models fitted to the data to study the connection between retinal thickness and prevalent PD; hazard ratios were used to determine the association between time to PD diagnosis, while retinal thickness was determined via frailty models.
Findings?
Looking at the AlzEye data of 154,830 participants, 700 had existing PD (and 105,700 did not).
PD patients had thinner GCIPL (-2.12 μm, 95% confidence interval (CI)l: -3.17, -1.07, p = 8.2 × 105) and INL (-0.99 μm, 95% CI: -1.52, -0.47, p = 2.1 × 104).
And the UKBB patients?
For the UKBB data of 50,405 participants, 53 patients developed PD. The data also associated with newly diagnosed PD cases included:
- Thinner GCIPL (hazard ratio: 0.62 per standard deviation increase, 95% CI: 0.46, 0.84, p=0.002)
- Thinner INL (hazard ratio: 0.70, 95% CI: 0.51, 0.96, p=0.026)
So what does that mean?
The findings of thinning GCIPL confirmed prior research on PD patients while also identifying—for the first time—an association to a thinner INL several years before a clinical diagnosis.
What else?
A reduced thickness of these layers was associated with an increased risk of developing PD, far beyond that resulting from other factors and comorbidities such as age, sex, ethnicity, hypertension, and diabetes mellitus.
Any limitations?
Yup … with the first being a lack of detailed clinical information regarding PD status (disease state, treatment patterns, current therapy)—thus they were unable to connect retinal morphology to disease duration or severity.
Others included differing definitions of PD (case-based vs disease-specific reference standard) as well as no correlative OCT and retinal histology data for the INL hypothesis.
Is further research needed?
According to the study authors, yes.
They stated that additional research is needed to identify if GCIPL atrophy is driven by brain changes in PD, or if INL thinning develops prior to GCIPL atrophy.
Why is this so important to determine?
Answering these questions could reveal whether retinal imaging like OCT could play a key role in diagnosing, prognosis, and managing potential PD patients years before the disease progression.