Innovating Clinical Trials

In the final part of this three-part conversation, Liam and Ted ask Paul Ivsin, Principal at First Patient In, a question that made everyone laugh: when should you fire a recruitment vendor?

Paul's answer is that the decision belongs before the first patient is enrolled. Once a study is running, nobody agrees to change anything, so the only window is early. Agree a checkpoint, such as a hard look at 15% or 20% of target, and write down what "off track" means. Then treat it the way a trial treats an interim analysis: a planned moment to look at the data and make a go or no-go call on the operations. He admits the weakness too. Interim analyses often come back ambiguous, and so will this one.

That leads to the theme of the episode: we run careful experiments on the drug and almost none on the enrollment. Paul uses pre-screeners as the example. Some have three questions, and he recently saw one with 21 that he couldn't finish himself. Which is better? After 25 years, he doesn't know, and says the honest answer is to put both versions on the same website, randomise, and count. When he has suggested this to larger pharma companies, nobody has liked it, because "we don't know" is hard to say out loud. Liam adds that no question count will ever be right for every patient, so segmenting by urgency and unmet need, even into eight rough buckets, beats pretending there is one answer.

Ted closes with the big question: with all the tools we have now, are we getting better at recruitment or worse? Paul says both. The best digital programs are smarter and work better with sites than anything from ten years ago. But outreach has concentrated on Facebook, Instagram and Google, and anyone with a laptop and some overconfidence can now call themselves a recruitment company. He expects a reshuffle, because sites that understand what a screened patient is worth to them can run those ads themselves.

Find Paul at firstpatientin.com and on the First Patient In on Substack.

In Part 2 of this three-part conversation, Liam and Ted pick up with Paul Ivsin, Principal at First Patient In, at the point where most campaigns leak: the handoff between pre-screening and the site.

It opens with an exercise the hosts ran on a number of central advertising campaigns. They took the pre-screening questions meant to qualify candidates and laid them against the protocol language. Every campaign had questions that either let the wrong patients through or turned the right ones away, and each had passed an ad agency and a sponsor review. Paul thinks this is getting worse. Turning a protocol criterion into a question a patient can answer always means choosing which errors you can live with, and some pre-screeners now look like they took under a minute to generate.

From there the conversation turns to ownership. Liam has been reading CRO budgets with biotechs and keeps finding the same thing: the most senior recruitment person on the study is an associate director at around two hours a month, while the project manager and data manager have full seats at the table. Paul sees it from the vendor side. Some CROs still make money on change orders and have little reason to shorten a timeline. Many sponsors hand recruitment to a vendor and stop looking, so ask how enrollment is going overall, and someone starts digging through a spreadsheet. When a study is three months in and behind, the usual answer is "it'll catch up once more sites activate." Paul's point is that this is an answerable question: you have activation dates and patients in, so you have a rate.

Ted raises no-shows, which he sees rising at sites. Paul's experience is that a persistent 20 to 30% of referrals never answer the phone once handed over, and the fixes that work are basic: send the text reminder every dentist sends, and hire someone for intake so study coordinators, who are often better with data than with people, aren't the ones phoning every referral.

In Part 1 of this three-part conversation, Liam and Ted sit down with Paul Ivsin, Principal at First Patient In, who has spent more than 20 years building enrollment programs with sponsors, sites and recruitment vendors.

Paul studied philosophy, drifted into database work, and landed in recruitment through a client running radio and newspaper ads for a trial. The phones were ringing and leads were being faxed to sites, but nobody could prove a single call had turned into a patient. He was hired to build the tracking. Twenty-five years later, he says, the industry still struggles to connect recruitment activity to enrollment.

That habit of asking "where did this number come from?" led him to the most quoted statistic in the industry: 80% of trials miss their enrollment timelines. The number has not moved in two decades, which made Paul wonder whether anyone was still measuring it. He followed the citations backwards and the trail ends at a source book from around 2000 that he has never been able to find. Worse, the question is broken anyway. Most studies have three timelines, the feasibility estimate, the CRO's bid and the date the CEO announced, so a trial can beat one and miss another. Paul's bigger worry is how the stat gets used: when everything is declared broken, any solution looks reasonable.

The conversation then turns to what Paul's data shows about digital advertising. Ads built for retail are made to maximise clicks, and a trial needs the opposite: only the people who will qualify. When he has compared the best-performing ads by clicks with the ones that produced enrolled patients, the correlation has often been negative.

Paul closes on where he would look first to improve any campaign: the handoff from pre-screening to the site. In a single campaign, with the same ads and the same pre-screener, he has seen one site consent 20% of its referrals while three others consented none.

In Part 3 of this three-part conversation, Liam and Ted wrap up their discussion with Robert Goldman, Vice President and Global Head of Clinical Operations at Contraline, on a question that lands in every ClinOps inbox: is “AI-powered” innovation or marketing? Robert explains why a dashboard from eight years ago isn’t new just because it has an AI label, how to test a vendor’s pitch, and why AI should support clinical judgment, not replace it. He also asks the question he puts to every AI vendor: if the hours saved are real, why aren’t drugs reaching patients faster?

It’s the same demand for evidence Robert brings to feasibility estimates and site selection in Parts 1 and 2 of the series. Listen to all parts of the conversation to see how it runs from the first plan to the latest technology.

In Part 2 of this three-part conversation, Liam and Ted continue their discussion with Robert Goldman, Vice President and Global Head of Clinical Operations at Contraline, to talk about why feasibility so often fails to predict what happens next. 

Robert explains why historical benchmarks mean little for a study that has never been done, and what he would use instead: measurable assumptions, early performance indicators, and intervention points agreed in advance. Ted raises the feasibility question he has never seen on a questionnaire, and Robert explains why sites are afraid to answer it honestly. They discuss why sponsors budget for backup sites, why the fastest responders aren’t always the best, and how sponsors can build real relationships with sites before the selection process begins.

Join us in Part 3 of the conversation as Robert gives his take on AI in clinical operations.

In Part 1 of this three-part conversation, Liam and Ted sit down with Robert Goldman, Vice President and Global Head of Clinical Operations at Contraline, who brings nearly 20 years of clinical research experience spanning CRO roles and sponsor-side clinical operations, to talk about why the industry rewards optimism in planning and demands realism in execution.

Robert shares how a detour through Caribbean medical school led him into clinical research by accident, and what he has learned about how lean biotech teams succeed: knowing what to build, what to outsource, and what good looks like. Together they look at why feasibility estimates stay unrealistic when everyone is incentivized to present the best case, and why optimism is not an enrollment plan.

Robert leaves listeners with the question he would ask instead of “how many patients can you enroll?”: show me the evidence for your estimate, and tell me what happens when you’re wrong.

In Part 3 of this three-part conversation, Liam and Ted wrap up their discussion with Duncan, President at DTS Language Services about AI, risk mitigation, and the human factors that are hard to capture in clinical trial translation.

Duncan calls translation the “last mile” of clinical trials, where small details can decide whether a site activates or a patient stays in a study. He explains why AI has a place when millions of words are on tight timelines, but argues that low-volume, high-risk clinical content is a very different proposition. He compares AI translation to an autopilot where the human is only brought in after something goes wrong, and describes how translators are spending 80% to 90% of the usual effort correcting AI output while being asked to accept much lower prices. 

And just when it gets serious, Duncan closes with a story about what happened when he translated a stack of Chinese tattoos. Spoiler: most of it was gibberish!

In Part 2 of this three-part conversation, Liam and Ted continue their discussion with Duncan, President at DTS Language Services about patient trust, consistency, and the operational realities of translating clinical trial content. 

Duncan explains how small issues, from inconsistent terminology to differences in date formats, can undermine a patient’s confidence in a study and even lead to missed visits. Trust comes from consistency, fewer handoffs, and an approach that fits the purpose of the content, and describes the “quilt factor” that builds up when previous translations, multiple CROs, and competing terminology databases pile on top of each other. 

The conversation ends on a question every sponsor should ask before translation begins: what does success actually look like?

In Part 1 of this three-part conversation, Liam and Ted sit down with Duncan, President at DTS Language Services, to talk about the good, the bad, and the ugly of clinical trial translation. Drawing on experience in both pharmaceutical and language services, Duncan explains why translation is far more than converting words from one language to another, and why a mistranslation in a clinical study isn’t just a typo but a potential risk to a patient, a trial, or a regulatory submission.

Duncan explains why the real work of translation starts before the translation itself: understanding who is reading the content, their language, reading level, context, fears, and emotions, and making sure the final version is both linguistically trusted and faithful to the approved source. He argues that translators have to resist the temptation to “improve” content. Liam connects this to a problem he's seen: the constant addition of complexity, particularly in consent forms.

This episode is a candid look at the hidden complexity behind clinical trial translation and the role it plays in patient trust, compliance, and study execution.

In this episode, Ted introduces the “ENROLL Corporation” (Everyone Needed for Recruitment Optimization Leadership and Logistics) an imaginary joint venture owned by the sponsor, the CRO, every site, and every vendor. He explains why most recruitment providers solve one small piece of the problem, much like a wildlife photographer so focused through a 500mm lens that an animal wanders past unnoticed. 

Liam adds a river analogy about who “eats their own cooking,” and together they look at the 80-plus factors that affect recruitment, the holes in the enrollment funnel that nobody measures, and how adding complexity to a protocol can quietly cost months.