Ideason building a better future
DavidTsaiIdeas
  1. Intro
  2. 01Single enrollment
  3. 02Protocol sharing
  4. 03Digital twin
  5. 04Proactive
  6. Closing

What does the clinical trial of tomorrow look like?

The future of clinical trials

What will define a remarkable trial in the next decade?

David Tsai

By David Tsai

I lead clinical operations in biotech and write about how AI is changing clinical trials and drug development.

8 min read

Introduction

These are ideas for where clinical trials could go. Some could sit inside one trial, some across many. I’ve spent the last 15 years in clinical research and drug development, with the first several years working at a large academic hospital enrolling patients onto clinical trials.

01

Single enrollment

You enroll once. But screen against all studies.

A sponsor is the company that pays for and runs a study.

Today, you screen for one study. You qualify and enroll for just that one study. But what about the multiple other trials you may qualify for?

They sit right underneath it, from other sponsors in the same disease. You were never checked against them, so you never knew. You qualified for one and missed the rest.

A person labeled You at the start of one study’s screening stepsThe same steps with 11 other sponsors’ studies beneath, each marked Never checked
1 / 2
Figure 1One study screened, and the studies from other sponsors that were never checked.

Screening and enrollment are locked to one sponsor’s study. We shouldn’t settle for that.

For the same disease and the same kind of therapy, those studies typically use the same procedures.

Now take one indication with many sponsors, each testing a different drug. You consent once, and the same screening and enrollment procedures go to all of them.

Over time your results count against each study’s eligibility criteria. You can take part in one drug trial at a time, but you can be screened against all of them. No waste. No redundancy.

Results count only inside each study’s window. In one published protocol, scans count for 28 days and labs for 72 hours, so only what has expired gets repeated.

The 12 studies with the same steps marked on every rowOne consent, one screening and one record fanning out to 12 studies, each with its own eligibility resultThe same fan, with data and samples traveling from the joined study to the others
1 / 3
Figure 2The same procedures across every study, then one consent, one screening and one record reaching all of them.

From the one study you join, your data and samples branch out to other studies and other sponsors. Today that is unheard of.

This is the future, a shared collective knowledge that benefits all patients.

How close is it? What one consent and shared data enable

2.7M

One consent that covers later invitations: people in the UK’s Our Future Health, whose consent includes being invited to other studies and trials.

Source: Our Future Health, programme design and recruitment

500K

One collection of samples and scans that many teams reuse: volunteers in UK Biobank, with more than 21,000 researchers in over 60 countries using the data.

Source: UK Biobank, news release via EurekAlert, July 14, 2025

880K

One record that many teams read: people in the NIH’s All of Us program, with more than 22,580 registered researchers working from their data.

Source: NIH All of Us Research Program, update to the Council of Councils (J. Denny), May 2026

Each treatment trial would still ask for your consent to its own risks; US rules require it (21 CFR 50.25). Screening windows and routine scans accepted for screening: protocol MK-6598-001, NCT05594043; every protocol sets its own.

02

Adaptive protocol sharing

What one trial learns, the next one already knows

Here are 3 diseases. Each line is one study from a different sponsor, Phase 1 to Phase 3. The gray circles are time points inside each study.

Right now: Sponsor A runs a trial. Learns something critical about a biomarker. Publishes 2 years later. Sponsor B uses it.

Across more than 600 trials, the median time from finishing a study to releasing the data was almost 2 years, and a third were still unpublished after 4. This is too slow. We need to move faster.

A timeline of studies from different sponsors in cancer, heart disease and blood disordersToday: one study’s learning reaches the others about 2 years after it ends
1 / 2
Figure 3Studies in 3 diseases, and how long one study’s learning takes to reach the others today.

Imagine instead: Sponsor A learns the biomarker predicts response. That learning goes into the shared knowledge base. Immediately.

Sponsor B, running a trial in the same therapeutic area, updates their eligibility criteria in real time. Sponsor C builds it into an upcoming trial. Collective learning.

In heart disease, one study learns something, and every heart study running or about to start can use it.

Then blood disorders. Inside each disease, every study now learns from the others.

Last, the diseases share with each other too, and every study becomes part of one shared collective learning trial.

The new way: one cancer study turns green from the point where it learned somethingEvery running cancer study picks up the learning, and it reaches studies that have not begunAll three diseases share inside themselvesEvery study merged into one band labeled Shared collective learning trial
1 / 4
Figure 4The new way: a learning shared the moment it is made, first inside one disease, then across all three.

It isn’t only the studies that finish. About 9 in 10 drugs that enter clinical trials never reach approval. When a program stops before its readout, what it learned rarely leaves the company.

Dot chart: 72 of 96 trials of approved drugs were fully published (75%), against 30 of 81 trials of drugs that were stopped (37%)
Figure 5US trials run from 2005 to 2009 in cancer, cardiovascular disease and neurological disorders, searched for full publication years later.Sources: Hakala et al., The BMJ, 2015; about 9 in 10: Sun et al., Acta Pharmaceutica Sinica B, 2022.

Those shelved programs hold answers the next team needs: a mechanism that didn’t work, a dose that was too high, a group of patients who didn’t respond. Buried, that is a loss for every patient who took part. Shared, it informs the trials still running and the ones about to start.

Why should one trial suffer from a learning about a mechanism of action that another trial already has?

Illustration: what the field knows rises slowly in steps when results are shelved, and compounds when they are shared; the gap between them is what shelving costs
Figure 6Shelved results leave each team to start over; shared results build on each other, and what the field knows compounds.

A shared collective knowledge that benefits all patients. Fewer people on a design someone else already learned from.

The gap from study end to paper uses the 23-month median in Ross et al., BMJ, 2012 (635 NIH-funded trials; 32% unpublished at a median 51 months). Changing eligibility in the middle of a study still means a protocol amendment and approval from an ethics board, and sharing from a study still running needs rules that protect its blinding. A piece of this already runs: TransCelerate member companies share de-identified historical control-arm data to improve trial design.

03

Your digital twin

Your living digital twin

At first, it is all you.

Then it becomes strands of digital DNA.

An X-shaped X-ray pattern of DNAA double helix with one strand in ink and one in blue
1 / 2
Figure 7After Photo 51, Rosalind Franklin and Raymond Gosling’s 1952 X-ray of DNA

Not “a patient digital twin for drug discovery research” but your personalized, ever-evolving, patient digital twin.

Based on your data. Your genetics. Your lifestyle. Your current health history. And it foreshadows disease.

Your record copies over: every visit, lab and scan. It is a copy of you that you own, and you decide who sees which part.

On the twin those lines stop being separate. Each links to the others, and what they add up to can be inferred.

Two dotted figures, You and Your twin, with a health record of 8 entriesThe twin with record entries pinned to parts of its body and a note, Inferred: coronary risk rising
1 / 2
Figure 8You and your twin, and your health record copied onto it.

Then we can run trials on your twin without running them on your biological self.

In your body you can take part in one trial at a time. Through your twin you can test many drugs and see the outcomes.

And your digital twin, built from your continuous health data, runs 10,000 simulations. And gives you treatments built on your data.

The twin with Drug C passing over it and a counter of simulationsA table of Drug A to Drug F with predicted benefit and side effects, Drug C picked as the best predicted match
1 / 2
Figure 9Model prediction. For you and your doctor to weigh.

And it says: “This treatment is built for you, from you, without ever having you participate in testing something out that may not work.”

Today, the only way to find out whether a drug works for you is to take it.

04

Proactive (not Reactive)

Proactive health care (not reactive trials)

Right now, you get sick, then you look for a solution. If there is none, you go on a clinical trial. You enroll, you complete, you leave.

10 years

Type 2 diabetes found at 40 takes about 10 years off a life.

Source: Lancet Diabetes & Endocrinology, 2023

11.1 years

Life expectancy lost, on average, after a first heart attack at 50, for a woman whose heart pumps weakly after it.

Source: SWEDEHEART, Circulation, 2024

12.8 years

Life expectancy lost, on average, by a man found to have colon cancer at 55.

Source: Sweden, diagnosed in 2005. Andersson, BMC Cancer, 2015

A chart of years of living: one red line climbs from diabetes at 40 and ends earlyThree red lines: diabetes at 40, heart attack at 50, colon cancer at 55, each ending early
1 / 2
Figure 10Three illnesses, each climbing in steps and ending a life early.

When you get sick, or before you get sick, the system matches you to an intervention.

The rise is caught before it becomes an illness.

58%

fewer new cases of type 2 diabetes when people with prediabetes followed a lifestyle program.

Source: Diabetes Prevention Program. Knowler, NEJM, 2002

54.5%

lower risk of coronary heart disease for each 1 mmol/L lower LDL cholesterol carried from early in life, about 3 times what a statin started later achieves.

Source: Ference, JACC, 2012

53%

fewer deaths from colorectal cancer than expected in people who had precancerous polyps removed.

Source: National Polyp Study. Zauber, NEJM, 2012

A blue line reaches age 40, where the diabetes line has come down to a small bumpThe blue line with three small bumps: caught at prediabetes, cholesterol lowered early, polyp removedOne long blue line running to the end of the scale
1 / 3
Figure 11The same three, each caught early. Numbers are from the sources named.

How long you live. And how well you live.

Life is enough of a roller coaster. Our health doesn’t need to resemble one.

In closing

The future of clinical trials

It’s not about “better” clinical trials. It’s about what healthcare should be.

Trials and care aren’t separate. They’re the same thing. Generating evidence while delivering care. Not after. Not separate. Integrated.

One line from one consent branching to several studiesThe studies run into a double helix, then red lines that climb in stepsOne blue line running through all four, nearly flat after one small bump
1 / 3
Figure 12The four ideas as one line: one consent, shared learning, your twin, and a rise caught early.
Single enrollment
One consent and one screening, checked against every study in your disease.
Adaptive protocol sharing
What one trial learns, the next one knows while it runs.
Your digital twin
Many drugs run through your twin, and you see the outcomes first.
Proactive (not Reactive)
The rise is caught before it becomes an illness.

What it means for us

When care and research are one, everyone gets better treatments, sooner.

This was the future of clinical trials and where they should go. Part 1 shows the trials we run today and challenges we face and how we overcome them together.

Read nextWhat makes a clinical trial resilient?Building resilient clinical trials

Start a conversation.

About the author

David Tsai

I lead clinical operations in biotech and write about how AI is changing clinical trials and drug development.

I’ve spent the last 15 years in clinical research and drug development, with the first several years working at a large academic hospital enrolling patients onto clinical trials.

Views and opinions expressed here are my own and do not necessarily reflect the position of current or past employers.

Full background on davidts.ai ↗Opens in a new tab, so this page stays open.

Before you go

The animated version

This is the ideal way to view this piece. The animations are there to drive the points home.

If you get motion sick, scroll slowly.

Continue to the animated version