The thirty-year flat line. And the room where I understood why.

This edition goes deep on two arguments I have been developing for a long time. They are connected, though they might not look it at first. One is about what happens when consent is sought for clinical trial participation. The other is about what happens outside that room, in the decisions that determine who ever gets that far.

Both are things I have not been able to stop thinking about. And both deserve more space than most formats allow.

Let me start with a number that I find genuinely shocking, even after years of working in this space.

A systematic review spanning three decades of research found that between a quarter and a half of clinical trial participants did not fully understand the key elements of what they had agreed to. What randomisation means. What the real risks are. Their right to withdraw without consequence. Comprehension across studies sat between 52% and 76%. And across those thirty years, that figure did not move once.

Not improved. Not worsened. Flat. While clinical trial design advanced in almost every measurable dimension:

However, on the single question of whether the participant actually understood what they were agreeing to, nothing changed.

I have spent years working directly with underserved communities. And I want to tell you what that flat line looks like from the inside, because I do not think the published literature captures it properly.

The standard response when comprehension fails in minority ethnic or migrant communities is translation. Produce the consent form in another language. Bring in an interpreter. Tick the box. The assumption is that language is the barrier, and that removing it removes the problem.

It does not.

I have sat in rooms with South Asian communities, with East African communities, with communities whose first language is not English and whose relationship with formal medical systems is shaped by decades of distrust and distance. And what I have seen, consistently, is that a translated consent form is not understood any better than the English original. Sometimes it is understood worse. Because the problem was never primarily the language. It was the register, the structure, the logic, and the implicit assumptions embedded in every line of the document.

A consent form translated into Urdu or Somali still uses the conceptual vocabulary of Western biomedical research. It still assumes familiarity with the concept of randomisation, with the distinction between research and treatment, with the idea that you might be allocated to a placebo arm of a study for reasons that have nothing to do with your individual clinical need. Those concepts do not become accessible because the words around them changed language. They are still foreign. And in communities that already carry a background anxiety about what healthcare systems do with people like them, they are not just confusing. They are frightening.

I have watched people read a translated consent form and become more reluctant to participate, not less, because the document raised questions they had not previously considered and gave them no reassuring framework to hold those questions in. The form told them about risks they had not imagined. It introduced uncertainty where there had been cautious willingness. And then it asked them to sign.

That is not informed consent. That is informed anxiety followed by a signature.

The reason this matters structurally, beyond the individual experience, is that the consent form is the first formal moment a trial communicates to a participant what kind of relationship this is going to be. And for communities that have historically experienced healthcare as something that happens to them rather than with them, that first moment carries enormous weight. A document that is dense, legal, and incomprehensible does not just fail to inform. It confirms a fear that was already present: that this system is not designed for people like me.

So why has the comprehension figure not moved in thirty years? I do not think it is because nobody cares. I think it is because the document was never primarily written for the participant. It is written for the institutions that need to demonstrate, in the event of any subsequent challenge, that the participant was informed. The legal team reviewing it needs it to be comprehensive. The ethics committee approving it needs it to cover every foreseeable risk. The sponsor defending it in a dispute needs every possible disclosure to have been made.

None of those stakeholders have a direct interest in whether the person signing it understood it. And that person, in almost every case, was not in the room when it was written.

Co-designed consent, where patients help shape what goes in, in what order, in whose words, consistently improves comprehension. The evidence on this is not new or contested. It works. And it remains the exception because it moves the document's primary loyalty back to the person signing it, which means it becomes shorter, clearer, and harder to hide behind. That is a shift worth making. It is also a shift that requires those with institutional power over the process to accept that they are giving something up.

For the communities I work with, this is not an abstract governance question. It is the difference between feeling like a partner in something and feeling like a subject of something. And that difference shows up in participation rates, in dropout rates, in whether the people who most need access to clinical research ever get it.

A consent form should be the first moment a trial treats someone as a partner. Right now, for most people, it is the first moment it treats them as a liability.

Sources: Pietrzykowski T, Smilowska K. The reality of informed consent: empirical studies on patient comprehension - systematic review. Trials. 2021;22(1):57. European Health Literacy Survey (HLS-EU). European Journal of Public Health, 2015.

Years ago, on a site visit, a principal investigator told me, calmly and without lowering his voice, that he would not recruit Hispanic patients. His reasoning was too offensive to repeat. What stayed with me was not only what he believed. It was how comfortable he was saying it out loud, to a stranger, as though it were a scheduling preference rather than a decision that would shape the evidence base for a medicine.

I have thought about that moment many times since. And what I keep coming back to is not the man himself. It is what he revealed about the system around him.

Nothing in that trial's data would have caught him.

His numbers would have come in. His site would have reported its results. The people he refused to enrol would simply never appear in the dataset, and their absence would read as normal. No flag. No outlier. No question asked. The published results would carry no trace of the decision made in that room. The medicine would go on to be prescribed to the people whose exclusion never registered as a loss.

This is what I mean when I say the quiet version of exclusion does far more damage than the visible kind. That PI, at least, showed me where the problem was. He made the bias legible. The system around him was designed, not through any single malicious choice but through decades of accumulated indifference, to make it invisible.

I want to be precise about what I mean by the quiet version, because I think it is important not to let the PI in that room become a convenient villain who absorbs all the blame and lets everyone else off the hook.

For every PI who says something like that out loud, there are many others who make the same functional decision without ever articulating it. They open a site in a location their target community cannot easily reach. They set an eligibility criterion that is drawn more tightly than the science requires, and that happens to exclude the populations carrying the highest disease burden. They design a visit schedule around the assumption that participants have flexible working arrangements and reliable transport. They do not approach certain community organisations because they have learned, through experience or assumption, that those communities are hard to reach.

None of those decisions require malice. They require only the absence of a data layer that would make their consequences visible. And that absence is the norm. At most clinical trial sites, what gets recorded with precision is whether the patient met the eligibility criteria, whether they signed the consent form, whether they completed the required visits. What does not get recorded is who was approached and declined. Who was never approached. Which communities were never given the chance to say yes or no. Which sites made quiet decisions about which patients were worth the effort.

That absence does not look like a problem in the data. It looks like nothing at all. And that is exactly how structural bias travels from the room where a decision is made into the published evidence base, and from there into clinical practice, and from there into the treatment outcomes of people who were never represented in the trials that proved the medicine worked.

My father knew what it felt like to be on the wrong side of a system not built for him. He pushed for answers for over a year. He was not listened to. By the time they did listen, it was terminal. The human cost of a data gap is not a missing row in a spreadsheet. It is a person who was not seen in time.

The PI in that room taught me where the invisibility hides. Not in the published results. In the space between who was there and who was never asked. In the eligibility criterion that looked neutral. In the site that was never opened. In the community that was never approached because nobody wrote down that they had not been.

What this means practically is that changing individual attitudes, while worthwhile, is not sufficient. The PI who says it out loud can be removed from a trial. The PI who never says it, whose decisions produce the same outcome without the words, cannot be identified by any system that only measures what is present in the data. You cannot interrogate a bias that leaves no trace.

Unwritten Health exists to create the trace. To ask, before the protocol is locked, before the sites are contracted, before the first patient is approached: who is this trial designed to reach and who is it quietly designed to miss? To make the absence visible before it becomes a data gap that nobody ever has to explain.

The first job is not to change attitudes. It is to build systems where the consequences of those attitudes cannot hide.

Something I have been building for the last several months is nearly ready. And if you work in market access, regulatory affairs, or commercial strategy, this is directly relevant to what you are navigating right now.

The EU's Health Technology Assessment Regulation came into force in January 2025. Every member state is now reworking its national process to align with it. That alone would represent a significant shift to absorb. It is not happening in isolation.

Italy has replaced its pricing and reimbursement dossier with a full HTA-based framework, effective April 2026. The change is not cosmetic. It requires a fundamentally different approach to how evidence is structured and how value is argued. France's 2026 Social Security Financing Law has narrowed early access pathways and shifted financial risk onto manufacturers when reimbursement timelines slip, which they increasingly do. The Netherlands already runs one of the most demanding health-economic standards in Europe, and the new European layer has added complexity rather than simplification. Germany continues to operate under AMNOG while adapting to the new joint clinical assessment requirements. Spain is moving in its own direction, at its own pace, with its own interpretation of what the regulation requires at national level.

The cumulative picture is this: more assessment, more scrutiny, more national variation stacked on top of a new European layer. And every one of those layers represents a point at which a submission can stall, a launch can be delayed, or a commercial team can spend months correcting errors that the right preparation would have avoided entirely.

It is worth naming something that often goes unsaid in market access conversations. The countries where launches are delayed, where submissions are challenged, where reimbursement is denied or restricted, are often the same countries where underserved populations have the least alternative access to new treatments. Market access failure is not just a commercial problem. It is a health equity problem. The patients who most need access to innovation are the ones with the fewest alternatives when the system fails to deliver it.

Navigate the System is our response to both of those problems. Six country playbooks, for the UK, Germany, France, Italy, Spain and the Netherlands, written for the people who actually have to get a medicine reimbursed in each one. Not an overview. Not a framework document. Operational detail: what the process is now, what just changed, what it means for your next submission, and where the specific risks and opportunities lie in each market at this moment.

Keep an eye out for the launch email.

I published a short video this week, the clearest account we have given of why Unwritten Health exists and how our platform actually works. I want to go deeper here than the video allows.

The question we get most often from clinical development teams is a version of this: we care about inclusion, we have a diversity plan, we have a patient engagement strategy. What does Equity Engine add that we are not already doing?

The honest answer is: it adds the data layer that makes everything else credible.

Here is how it works on a programme. A sponsor or CRO comes to us with a specific problem. A protocol that has already had one amendment and is about to have another. A feasibility question about whether a particular underserved population can realistically be recruited at the planned sites. A regulatory submission that needs to demonstrate how patient experience evidence from underrepresented groups shaped the design.

We start by defining the specific questions that patient experience data needs to answer for that programme. Not generic questions about diversity or engagement, but precise questions: where will this visit schedule break in real life for someone with insecure work? What would make this digital diary tool unusable for older patients with limited digital access? What trade-offs around procedures and time commitment are this community realistically willing to make?

We then work through trusted community partners — people embedded in underserved neighbourhoods and networks who have existing relationships with the communities we need to reach. This is not a recruitment exercise. It is an access model built on existing trust, which is the only kind that produces honest answers. One-off surveys sent to community organisations produce one kind of data. Ongoing relationships with people who feel genuinely safe sharing what their lives are really like produce another.

The data we collect is longitudinal and structural. Not what people think about a trial in a moment, but how they live with their condition over time — what gets in the way, what would help, how their perception of research shifts, what their actual daily reality looks like in terms of work, family, transport, housing, language, and culture. We use structured interviews, digital diaries, and ethnographic methods that go significantly deeper than tick boxes.

Equity Engine then takes that lived experience and makes it usable for clinical development teams. It maps insights to specific elements of a protocol — not just flagging that there is a risk, but showing where the risk sits, which communities are most affected, and what the practical options are. A visit at a hospital site at 9am on a Wednesday looks different to a community nurse in a local clinic on a Saturday morning. The data shows you which of those two designs a broader range of patients can actually sustain across a six-month study.

The output clinical teams see in Equity Engine is structured and navigable: how different communities will experience the proposed protocol, where feasibility and burden are likely to be a challenge, what specific changes would reduce that risk, and where further evidence might be needed before design lock.

And here is the thing that compounds over time. When you run multiple programmes through Equity Engine, you are not starting from zero each time. You are building a longitudinal evidence base on underserved communities that gets more valuable with every study. The insights from a Phase II oncology trial inform a Phase III trial in the same indication. The community relationships built for one programme can be activated for the next. The regulatory narrative that shows how patient experience shaped design becomes stronger each time it is used, because it is backed by an evidence base with real depth and continuity.

That is what we mean when we say this is infrastructure, not engagement. It is the difference between a diversity plan that gets written for a submission and a clinical development capability that genuinely de-risks how you design and run trials.

If this is the problem your team is wrestling with, we would like to talk. We have opened a small number of conversations about Equity Engine over the coming weeks. If you want one of them, the link to book is below.

BOOK A CONVERSATION

Three things from this week worth your attention.

PharmaTimes published a feature this month by Julie Massicotte, Senior Director of Regulatory Affairs at Indero, making the case that inclusive protocol design is an upstream development decision, not a downstream recruitment fix. The argument will be familiar to readers of this newsletter. But it is significant when it appears in mainstream pharma trade press, because it signals that the framing is moving from the margins of the conversation toward the centre. The practice has not caught up. The language is getting there.

PharmaTimes. A Design for Life. June 2026. pharmatimes.com/magazine/2026/june-2026/a-design-for-life

The MHRA published its landmark report on public views of AI in healthcare this week. This is one of the most comprehensive engagement programmes the agency has undertaken on any topic. 760 people and institutions contributed to the Call for Evidence. The National Commission worked with National Voices specifically to reach underrepresented groups. The Health Foundation ran public deliberative sessions to ensure the findings reflected communities that do not typically show up in institutional consultations.

The headline finding is that the public broadly supports AI in healthcare, provided it comes with ongoing monitoring, transparency, accountability, and genuine human oversight. The finding beneath the headline that matters more is this: public trust in AI is not unconditional, and it is not distributed evenly. The communities that already have the least trust in healthcare institutions are the ones most concerned about algorithmic decision-making in clinical settings. That is not a communications problem. It is a design and governance problem. And it will not be solved by publishing reassuring FAQs.

The line from Patient Safety Commissioner Henrietta Hughes is the one I want to keep: the AI Commission puts patients' voices at the centre, bringing too often unheard perspectives from the margins into the heart of decision-making. That is the standard. The question is whether the AI systems being deployed in clinical research will be held to it.

MHRA. National Commission into the Regulation of AI in Healthcare: Research, Engagement and Call for Evidence Findings. 11 June 2026. gov.uk/government/news/mhra-landmark-report-reveals-public-views-on-ai-in-healthcare

The FDA announced this week that two proof-of-concept real-time clinical trials are already underway. AstraZeneca's TRAVERSE trial in mantle cell lymphoma and Amgen's STREAM-SCLC trial in small cell lung cancer are both reporting endpoints and data signals to the FDA in real time rather than through traditional batch reporting. A broader pilot programme is planned for this summer with final selections in August.

FDA Commissioner Marty Makary described it directly: for sixty years, key data signals have taken years to reach the FDA. The lag delays decisions and slows timelines. That is true and the initiative is genuinely significant. The thing also worth naming is that real-time data sharing only improves the evidence base if the populations generating that data are representative of the patients who will eventually use the medicine. Moving faster with a skewed dataset does not close the evidence gap. It accelerates it. As the FDA builds this infrastructure, the question of who is in the trial — answered before the first endpoint is ever reported — matters more than it ever has.

FDA. FDA Announces Major Steps to Implement Real-Time Clinical Trials. April 2026. content.govdelivery.com/accounts/USFDA/bulletins/414ec5f

Thanks for reading. This newsletter exists because I believe the right framing, in the right hands, changes decisions. If it did that for you this week, even a little, that's enough.

Ashish.