Why Ophthalmology Trials Are Clinical Research's Toughest Tech Test

By John Oncea, Chief Editor, Clinical Tech Leader

Maggie Kelehan didn’t set out to specialize in eyes; she simply took a job.

“I didn’t know that I would be working in eyes, honestly,” she said. “I didn’t really care at the time what field of research I went into. I just really wanted to get into clinical research in the first place, get my foot in the door.” Three years later, she’s a Clinical Research Coordinator II working across multiple ophthalmology-focused sites, and she’s become genuinely attached to the specialty precisely because of what makes it technologically distinct.

“I love eyes,” she said. “I think they’re so interesting.”

That affection is useful context, because ophthalmology research runs on a technology stack most general-medicine trial platforms weren’t built for.

A Stack Built Around Imaging, Not Forms

Ask Kelehan what makes her work different from a typical coordinator, and the answer is immediate: imaging. “There are extensive imaging systems that are ophthalmology specific that we use,” she said, and integrating them into the EMR is routine. CTMS integration for imaging data is a different story; she isn’t sure it currently exists at her organization, though she’d welcome it.

Every image doesn’t just get filed. It gets submitted to a reading center, an independent company whose job is to confirm the images, meet study qualification standards, and screen for any major changes to a patient’s eyes. Some reading-center workflows are beginning to incorporate AI-supported review, according to Kelehan.

Kelehan’s assessment of that AI is practical rather than skeptical: it works, but it’s still slow. “Sometimes it takes a while for the AI to read things, but I think as time goes on, it’ll get faster.”

That’s a meaningfully different AI story than the one dominating most clinical tech conversations right now. This isn’t AI as a chatbot or a scheduling assistant; it’s AI performing a specific, high-stakes visual QC function inside a regulated submission pipeline, and the open question isn’t whether it works, but whether it’s fast enough yet.

The One Problem Software Can’t Touch

Here’s where ophthalmology trials expose something almost every clinical technology conversation skips: not every bottleneck is a software problem.

Kelehan works within an embedded-site model – her organization partners with multiple independent ophthalmology practices across the city, each running its own EMR. A given study might require a visual acuity examiner to be physically present at one site while she’s also needed at another. “If I’m the only person that can do something on a study and I have to be at site A, that limits what I can do at site B, because I can’t clone myself, unfortunately,” she said. “It’s more of a staffing problem, I would say, than necessarily something that could be resolved via technology, unless I could, I don’t know, have a robot I could control at a different site.”

No integration, no interface redesign, fixes that. It’s a headcount problem wearing a technology-shaped complaint. But it’s not entirely outside technology’s reach – Kelehan sees real value in AI-assisted scheduling that understands site-specific constraints automatically, flagging a conflict before a newer team member accidentally double-books her across town. “It would decrease headaches for me personally when I look at the calendar, and someone has tried to put me in two places at once, but they don’t know that they’ve tried to put me in two places at once.” It won’t solve the underlying staffing shortage. It would stop the calendar from making it worse.

That distinction – knowing which problems are staffing problems and which are software problems – is exactly the kind of judgment sponsors and vendors need from the coordinators they’re building for.

What Generalist Trial Platforms Miss

Kelehan hasn’t worked outside ophthalmology, so she’s careful not to generalize about other specialties. But her experience surfaces a pattern worth naming for anyone building or buying trial technology for a specialty indication: device-generated and image-heavy data doesn’t fit neatly into platforms designed around generic case report forms.

Her imaging systems integrate reasonably well with EMR, largely because ophthalmology has “been using advanced imaging for years” and the integrations have matured. But CTMS integration remains a gap, and the reading-center submission workflow sits somewhat outside the standard EDC-CTMS relationship entirely, a specialized pipeline bolted onto a general-purpose system rather than designed alongside it.

The lesson for vendors: a platform built around oncology or cardiology case report forms doesn’t automatically serve a specialty where the primary data isn’t a form field, it’s an image, and where a third-party reading center may be a required stop before image-based data can be used for study decisions or endpoints.

An Accidental Specialist, Now A Genuine One

There’s something worth noting in how Kelehan got here that’s relevant to anyone hiring or building for this workforce: she didn’t arrive with domain passion. She arrived wanting a career with flexibility, room to grow, and the chance to “advance science and help people.” A friend suggested clinical research; she spent six months deciding, took a certificate program, and got a job – one that happened to be in eyes.

The expertise came after the job, not before it. That’s a useful reminder for an industry that sometimes assumes deep specialty knowledge has to precede the role rather than grow inside it, and it’s a big part of why her read on ophthalmology’s technology gaps carries weight. She’s not reciting industry talking points. She’s describing three years of daily friction in a specialty she’s come to genuinely care about.

The Bigger Point

Ophthalmology is a small corner of clinical research, but it’s a useful stress test for trial technology. When a specialty depends on image-based data, third-party QC pipelines, and physically constrained staffing, generic platforms and generic scheduling logic reveal their limits fast. Vendors serving specialty indications would do well to talk to the coordinators living inside those constraints – not after launch, but before.