The Future Of Clinical Trials: Artificial Patients, Synthetic Data And Real-Time Analysis
Bertalan Meskó, MD, PhD
Director of The Medical Futurist Institute (Keynote Speaker, Researcher, Author & Futurist)
In this article, we summarise three concepts that are already/ will soon become familiar for everyone interested in the future of clinical trials. These are 1. the concept of the artificial patient, 2. using synthetic data and 3. real-time analyses.
What connects the three is that all are potential tools to make clinical trials faster, cheaper and safer. That is so, in an ideal world. Because, as almost always, there is a catch, not even a small one. But more on it later.?
The artificial patient
What is it? As of today, there is no final, widely accepted definition of what an artificial/virtual/synthetic patient is. A very detailed explanation about some (but not all) of the different definitions can be found in this study .?
So let me phrase it simply what we mean by it in this analysis:
an artificial patient is a set of data representing the desired human characteristics the best possible way that is based on large amounts of real patient data, without actually including any backtracable real-patient data.?
Why is it? Artificial patients can be the answer to more than one problems of modern medicine. One of them is patient privacy. With ever more machine learning and deep learning models being used, A.I. needs huge amounts of data to learn from. But providing a lot of real patient data is against their privacy rights, and we have seen ample examples of how bad an idea it is to allow random companies to access heaps of sensitive health data. On the other hand, taking a real-life dataset of existing humans, and generating a synthetic dataset that resembles the original in all important aspects (*more on it later) without actually including anything personal can be a solution.
What is it good for? Artificial patients can be used for a number of things, from medical education to clinical trials, this time we are only focusing on the latter. One day, virtual patients might become the go-to tools for
As many hope, one day artificial patients may be able to completely substitute humans and animals in clinical trials, most likely with animals being the first.?
While using artificial patients for drug development or medical device development is a promising field, there is a long way to go until the models can reach the required complexity while being truly representative of the human population.
On the other hand, artificial patients as the placebo control group have arrived. AppliedVR recently conducted a trial for VR treatment for chronic back pain patients. And instead of recruiting patients to sign up for the trials to not receive the treatment (being the control group), they decided to turn to an existing database of chronic pain patients, provided by healthcare data company Komodo Health.?
Using real-world data as a patient group in a trial, often known as a synthetic control arm, can make research trials more efficient — companies don’t have to enrol as many people in clinical trials and can guarantee that those who apply will indeed receive the treatment.
Synthetic control groups can also improve equity in clinical research. “That allows us to go look at all those different subpopulations and underrepresented patient populations to see if they have different outcomes,” Web Sun, president and co-founder of Komodo Health says.?
Another interesting example was this virtual trial (or in-silico trial) carried out to predict the efficiency of using flow diverters for brain aneurysms. Researchers created 82 virtual patients based on data of real patients from previous flow diverter trials. The experiment projected an 82.9% success rate for the use of diverters, pretty close to the results of three real-world flow diverter trials that had 86.8%, 74.8% and 76.8% success rates respectively.
Synthetic data
If you felt like your head started spinning from dealing with the concept of the artificial patient, behold, synthetic data is probably an even wilder ride.?
What is it? The definition is simpler this time: synthetic data is the use of A.I. to create datasets that mimic the real world.
Why is it? Because we 1. don’t have enough real-world data or 2. don’t want to use real-world (sensitive) data.
What is it good for? Feeding any algorithm that needs massive amounts of data to learn and either develop new prediction capabilities or recognise patterns. Synthetic data is widely used in a number of industries and segments, not just in medicine, but also in self-driving vehicles, security, robotics, fraud protection, insurance models, military and so on.
领英推荐
Artificial intelligence has earned its place in multiple fields of medicine , from recognising patterns, supporting diagnoses and setting up treatment pathways to optimising healthcare logistics. Smart algorithms can sift through large volumes of data no man can, deriving clear-cut trends from such analyses.?
Privacy concerns limit the amount of available data in medicine. Working with sensitive patient data is a tricky issue. It seems we cannot keep our privacy intact AND also benefit from A.I.’s advantages in our care. We saw in many cases how sensitive information can get leaked even unintentionally – and we are not even talking about hacking or privacy, just a poorly protected database. New methods like federated learning might make it possible to do this without breaching patients’ privacy, but its scope is limited.
That’s when synthetic data comes in. It can fill in the missing data, making it possible to produce entirely fabricated patient datasets that are just as useful for training A.I. as the real thing, while keeping patient data protected.
Using synthetic data could help overcome this challenge as the training could focus on such variables, making use of real-world environments. Using the above-mentioned example, how to diagnose melanoma on dark skin toned patients – as often previous algorithms have failed to be able to do so.
But not everything is just sunshine and sandy beaches
This insightful analysis not only explains how and why synthetic data is used, but also why we should be scared of it. In short: any dataset we create will be imperfect to some extent. It will contain biases we are not aware of. It will not include important variables we either overlooked or are not aware of their importance. Even, in the best of cases, it will be like a snapshot of a given moment of a given situation.?
If we let machine learning and deep learning algorithms develop on these synthetic, imperfect datasets, chances are they will come to conclusions that are more or less false in the real world.?
And the reason why this is worrying is the speed at which the use of synthetic data is spreading. Gartner predicts that by 2024 some 60% of all data used for AI will be synthetic. And not just in medicine.?
Synthetic data is cheap and easy to come by, much easier and cheaper than collecting huge amounts of messy real-world data. What happens if decisions affecting large groups of people or whole societies will be made based on it??
It is especially worrying as the world already faces challenges regarding “truth”. Introducing alternative truths based on ‘data’ to back decisions affecting societies - like healthcare funding, insurance models and so on - can have devastating consequences.
Real-time / decentralised clinical trials
What is it? The use of electronic health data/records/devices to carry out clinical trials in near real-time with patients not needed to be present on site.
Why is it? Real-time trials offer faster results and the possibility of participants to directly connect to other patients, share their experiences and get access to results.?
What is it good for? More committed participants and faster results.?
We have seen a few good examples in recent months.
One is that of Royal-Philips rolling out?a new at-home ECG system for decentralised clinical trial use. The company is pitching this new technology as a way for clinical trial participants to record ECG data without travelling to a clinical site or requiring an in-home clinician. Data of trial participants can be transmitted near real-time to the cloud servers for analysis.?
A number of digital health companies are designing tools to enable decentralised trials. In September, digital wound-care company Swift Medical launched a new digital-imaging platform designed to support decentralised clinical trials. The technology was designed to aid in large-scale image collection and management in order for researchers to monitor the impact of medical interventions at various sites or at home.
Deploying such advanced technologies in clinical trials will require pharmaceutical and biotech companies to commit not only financially but also to the idea that technologies can significantly contribute to clinical trials, making drugs cheaper, making the process faster and much more importantly, making the lives of patients participating in them more comfortable.
Change Management & Business Excellence (Digital Transformation, Community Building, PROSCI Training, Coaching & Consulting)
2 年We are currently working on these synthetic data and digital twins. I would be also highly interested to know the difference between artificial patients and digital twins. Many thanks !
I empower Non-Native English Leaders in global Clinical Studies to express themselves clearly and confidently, so they can grow their careers and make a bigger professional impact.
2 年This sounds very interesting and encouraging Bertalan Meskó, MD, PhD
Clinical Research Nurse, RN, MS
2 年Thanks for sharing. It's interesting. Artificial patients.. is the concept similar with digital twins? I am working on my grad thesis and have been reading few articles about digital twins. After reading your article, I would like to know what's the difference between artificial patients and digital twins?
Entrepreneur physician scientist. Health tech and digital health innovation leader | Strategic planning and business design | Building win-win partnerships between tech start ups and healthcare institutions and KOLs.
2 年Interesting insights Bertalan Meskó, MD, PhD I looked at it from a slightly different angle here: https://www.dhirubhai.net/posts/saber-ghadakzadeh_awareness-recruitment-failure-activity-6943017253814632448-PtPX?utm_source=linkedin_share&utm_medium=android_app