Last week I was fortunate enough to give a talk at the fantastic Hardian Health Summit. I opened with a story about working in ALS/MND early in my career. When you've made a promise to participants with just a short time to live that the data they’re sharing with researchers will be used for good, "we'll get to the paper next year" is not an acceptable answer, and I felt the pain of waiting for peer review comments as patients died from the terrible illness. 

That promise is what taught me to move fast. 

The most useful surprise from the day: how many founders came up afterwards to say they'd never thought about adding an evidence production line to their regulatory journey. They had a regulatory plan. They had a product roadmap. The evidence was an afterthought. Now, they will be using some ProofStack frameworks to start building the necessary foundation.

I met MHRA regulators, former FDA digital health experts, health economists, frontline clinicians, and a fair few startups. 

I also called out a few bad actors who'd been bending the rules on publication best practice (i.e. Microsoft, OpenAI, Quadrivia) which is how I came to be officially dubbed "the LinkedIn police" by Hugh. I'm not mad, though I might invest in a bulletproof vest.

This issue, I'm sharing something a little different. Eighteen months ago I was on the other side of the clinical trial table… as a participant. The deep dive is what I learned.

Enjoy!

– Paul

In this episode, Dr Tauseef Mehrali, VP of Regulatory at femtech juggernaut Flo Health shares his experience of designing safe products at a scale that touches hundreds of millions of users, and he spills the beans on the job offer he refused because a certain book was on every employee’s desk.

We also covered how to introduce medical device regulation into a 400-million-user consumer app without breaking it and much more.

For most of my career, I've been on one side of the clinical trial table: designing protocols, reviewing them, criticising them, occasionally celebrating them. Back in 2024, I spent three months on the other side. As a participant.

But first, a bit of context about why this matters.

Clinical trials are quietly getting harder to run.

Last week, I spoke online at the National Institute on Aging, talking about how to improve trials for older people with dementia. That led me to a recent finding.

Return on R&D investment for new drugs has fallen from roughly 12–15% at the turn of the millennium to between 3–5% today. The proportion of new drugs that succeed has dropped from 6–8% over the same window. Translation: it's much harder for pharma to turn investment into approved therapies than it was when I was starting out.

Why? In part because we now ask far more of patients.

The average trial participant went through about 187 procedures in the period 2013–2015. In the past two years, that figure is 301.

By procedures, we mean things like questionnaires, blood draws, pulse checks, weigh-ins, scans. Phase 3 pivotal trials now span 13 countries on average instead of 9. Total data points per trial have tripled — from roughly 2 million a decade ago to 6 million today. The reason is because, simply, we have far more data points to track than ever before. 

So as a participant you're no longer filling out a paper questionnaire on a clipboard. You're wearing a Fitbit, taking home a device with an app, doing scans, and sometimes letting AI watch you do tests in your kitchen.

That additional burden has consequences. 

About one in 10 trials never enrols a single patient. The average study visit takes more than two hours. That's two hours of your day, plus the childcare, the pet care, the time off work. And it disproportionately shuts out the people whose data we most need: shift workers, hourly workers, people without paid leave, people who don't drive, people who don't live near the research hospital. The result is what we keep seeing in the data — under-representation of non-white participants. 

This is why I developed a service called “Patient-Proof Your Protocol,” but maybe we’ll save that for another newsletter. 

I'd been on a rooftop at a client party, drinking a beer with colleagues when I noticed the tickle. By the time I'd walked back to my hotel, I had a full cough coming on. I flew home, isolated in the spare bedroom for 12 days, and missed my own birthday (my loving daughters left little gift parcels at the door for me to retrieve).

I came out the other side with reduced lung capacity, asthma I haven't been able to shake (still on inhaled steroids), tinnitus, and the thing that hung around longest — no sense of smell or taste.

Breathing is, as I joke, one of my favourite activities, so the lung issues got most of my attention initially. But over the next eighteen months, the missing smell and taste started to wear on me. I tried the standard online protocol of smelling essential oils every morning to retrain the nose. However, all I could detect was a faint, alcohol-like smell from these potent oils.

What pushed me over the edge was an Apple TV show called Drops of God, about elite sommeliers. Watching characters describe wine in terms of peach, earthiness, the terroir of soil… and I realised I couldn't smell my own daughters' hair after a bath. Naturally, the scientist I am, I started to worry about which memories I'd be unable to form properly. Smell is one of the strongest memory cues we have. I'd taken to adding cashews to every meal just for some sensory texture. I went out for a Sri Lankan feast and didn't flinch at heat that should have melted my face. The waiter was waiting for me to keel over.

So when a charity called SmellTaste posted a call for participants for a trial at the University of Nottingham, I signed up. (I wrote up the experience on LinkedIn here.)

The trial was unusual because it wasn't run by the medical school. It was run by the food science department, where Professor Nicole Yang and her PhD student Jing Feng (recently graduated!) worked on the kinds of flavour and sensation problems that the food and confectionery industries pay close attention to. They had developed flavour-changing chewing gum.

Yes, really.

Four packets, four flavours per packet, each gum changing across multiple sensations within a two-minute chew. Some had menthol for cooling. Some had capsicum for heat. The premise: if you stimulate a dulled sense with strong, varying inputs, and pay close attention while you do it, the brain can be coaxed into rebuilding the connections that COVID severed.

The protocol was twice a day, two minutes a time, for three months: twelve weeks. No coffee or toothpaste before chewing. Once a week, I would scan a QR code and answer adherence and outcome questions. I even had to pack the gum in my carry-on for a business trip to New York with a letter from the university explaining that I was not, in fact, smuggling narcotics in small clear baggies.

Around week four, I had a blueberry 🫐

For the first time in roughly two and a half years, it was a blueberry. Not just texture and acidity. The most blueberry-ish blueberry I have ever eaten that I felt like I was in the Blue Man Group. The world had been painted in blueberry! 

A week or so later, I was mowing the lawn and ran the mower over a small ‘offering’ from a neighbour's dog. The smell of cut grass and dog faeces in equal measure hit me like a freight train. I remember celebrating it. I can smell dog poo. This is brilliant.

By the end-of-study visit, Jing was sitting across the room and uncapping smell sticks. Garlic, leather, licorice, orange, cloves, peppermint, cheddar. I called them all from across the room. My score had improved by 7 points. The threshold for statistical significance was 5–5.5. So it worked, and I'm grateful every time I walk past a bakery or my daughters come out of the bath with a headful of strawberries.

Even though all I had to do was chew gum, it was still hard to stick to. I had to skip my morning coffee. I couldn't brush my teeth first. Two minutes of concentration is a long time when you have small children getting ready for school. 

I stuck with it because I could feel the benefit. That's an enormous luxury most trial participants never get. 

If you're designing a trial right now, the question is whether a real human being can actually complete the study. And whether the people you most need in the data can do it without quitting their job, their childcare, or both.

If you don't know the answer, that's the gap to close before the trial starts, not after enrolment stalls. 

Closing that gap is the difference between a trial that proves a point and one that doesn't. 

I want more people to have a good clinical trial experience. That's how we build proof that works in the real world. This is how we create extraordinary results!

— Paul

Today I am delivering a talk on the exact topic of today's deep dive: “A taste of my own medicine.” Funny story but I received an invitation to speak at this talk due to my LinkedIn post here!

Next week I am running a leadership reflections session called "Hat GPT" for the Academy of Medical Sciences FLIER program in London on May 13th.