Will synthetic control arms revolutionize clinical trials?

SCIENCE CALLING

Clinical trials are an essential component in the research and development of therapeutic solutions.

They are used to demonstrate a drug’s efficacy and tolerability, as well as provide regulatory agencies with the data they need to authorize its release on the market. However, the design of such trials faces a number of ongoing challenges, particularly in the case of rare diseases, where patient recruitment can be especially complex.

In this respect, and with the rise of artificial intelligence, a new innovation has gradually emerged in recent years: the use of synthetic trial arms in clinical trials. This technique involves using external data as a basis for comparison to assess the results of the experimental group.¹ So, could synthetic data arms revolutionize the way clinical trials are designed? To answer this question, the Insights’ teams attended the Synthetic Data & The Power of Twins for Trials round table hosted by Dassault Systèmes at Science Week 2024, as part of the Science in The Age of Experience conference series. Over the course of the round table, experts from Institut Gustave Roussy, the National Institute for Health and Care Excellence, Medidata AI, and Servier discussed the prospects for using synthetic arms and the challenges and issues associated with their integration.

Randomization: the current gold standard for clinical trials

Randomized clinical trials are based on two patient cohorts:

Intervention arm: Patient group that will receive the new medicine or treatment being tested.
Control arm: Patient group that will be given the best current therapy for their condition, known as the “standard treatment”. In other cases, these patients may be given a placebo.

Clinical trial randomization is the practice of randomly assigning treatment to each patient taking part in a clinical trial. Randomization ensures a balanced distribution between patients with the kinds of variables that may interfere with trial results (age, comorbidities, etc.). Although randomization is crucial for assessing the efficacy of medicines and patient safety, it comes with certain challenges:

Creating targeted cohorts that reflect the diversity — especially geographical diversity — of patients afflicted with the same disease is a major challenge, particularly in the case of rare diseases. It also makes conducting clinical trials exceptionally time-consuming and costly. This can sometimes slow down the development of new treatments that could change patients’ lives.
Using placebos in some cases also makes it harder to recruit and enroll patients. On the one hand, they may be reluctant to take an inactive treatment; on the other, they may be averse to following strict protocols (follow-up visits, regular medication intake, etc.) if they suspect they are being treated with a placebo. Although these trials are designed to advance research rather than treat disease, they also raise ethical questions.

Precision medicine: an additional challenge for clinical trials

Artificial intelligence has paved the way for the emergence of increasingly individualized medicine. By identifying genetic mutations or anomalies, AI can better capture the specific biological characteristics of each patient. This evolution has led to the definition of disease sub-categories, each calling for differentiated treatments. Clinical trials aimed at developing new, increasingly targeted treatments must therefore reflect these sub-categories and incorporate new inclusion criteria. This makes recruiting patients for the trial even more challenging.

The rise of precision medicine provides an opportunity to better target a disease and develop more effective treatments for a defined population. This involves more complex inclusion criteria and patient recruitment, which prevents sub-categories of patients in the clinical trial from undergoing standard treatment when their response rate is low.

Cancer is a good example. Until a few years ago, cancer was seen as a single disease with a single standard treatment: chemotherapy. Oncologists now see cancer as a collection of multiple diseases, each with its own distinct causes. Lung cancer, for example, has no fewer than twenty different treatments, each corresponding to one or more genetic mutations or anomalies. For some patients, chemotherapy is no longer routinely recommended due to its limited efficacy. The same logic applies to many other cancer types. Consequently, subjecting all patients in the control arm to chemotherapy also raises ethical questions.

Synthetic control arm: revolutionizing clinical trials with new technologies

Over the past few years, synthetic arms have been introduced in clinical trials for the treatment of rare diseases. This new approach involves replacing data from patients in the control arm of a clinical trial with data from external sources.

There are multiple reasons for using a synthetic arm in a clinical trial. First of all, it reduces the development time for new medicines, particularly for rare diseases where patient recruitment for clinical trials is especially time-consuming. Recently, synthetic arms have been used to evaluate the efficacy of innovative therapies for a rare form of lung cancer (LCNEC – Large Cell Neuroendocrine Carcinoma of the Lung) and for pediatric diseases. Furthermore, reducing the duration of clinical trials, and ultimately their cost, means that new treatments can be more effectively explored and made available to patients sooner.

Use of a synthetic arm also reduces certain biases that can arise in clinical trial design. In particular, it provides a means of dealing with any bias linked to the recruitment of patients for a clinical trial, which may result from the use of overly restrictive inclusion criteria (age limit, exclusion of certain comorbidities, place of residence close to the clinical trial site, etc.). By relying partly on real-life data, the bias correlated with the fact that patients in the control arm sometimes benefit from better management as a result of closer medical follow-up, which may increase the response rate of standard treatment in the trial, is also mitigated.

Testimonial

Randomization remains the gold standard for clinical trials. However, for rare diseases, the time required to gather a representative sample of patients significantly delays the market approval of life-saving drugs. More generally speaking, asking patients to forgo treatment on a control arm also raises serious ethical concerns. By integrating synthetic control arms – or at least hybrid control arms – we can supplement collected data, uphold scientific rigor, and offer patients a fairer choice. This approach might just be the ethical and legal breakthrough the industry needs.

Sammi Tang Former Vice-President Global Head of Biometrics at Servier

Synthetic and external control arms: what’s the difference?

When it comes to reducing the number of patients in the control arm of a clinical trial, there are two possible solutions: the use of an external control arm or a synthetic control arm.

These two alternatives differ in the type of data on which they are based. In the case of the external control arm, the data come from patients who have previously participated in clinical trials or who have already received treatment. These data are used as is.For the synthetic control arm, data are generated using advanced statistical models. They come from a combination of various sources, including previous clinical trials and real-life observational data. These data are used to create a control group as close as possible to the intervention group.

Synthetic control arms: What are the best conditions for optimal application?

The development of synthetic control arms is opening up promising new perspectives, however, several conditions must be met before they can be fully introduced in order to guarantee the integrity of trial findings and maximum patient safety.

The first requirement is the availability of high-quality, standardized health data that are representative of the populations impacted by the disease under investigation. In this respect, regulatory authorities have a key role to play in overseeing the use of personal medical data. This involves proposing a regulatory framework that both guarantees the complete security of patient data and is sufficiently flexible to allow access to a suitable volume of data for those involved in research and the pharmaceutical industry.

Second, the regulatory authorities must set out a clear regulatory framework for the use of synthetic arms. Health authorities will need to demonstrate a high degree of agility to address patients’ expectations regarding the timely availability of new treatments, while at the same time retaining the necessary objectivity to assess all the implications of the use of synthetic arms and define appropriate regulations.

What circumstances are currently conducive to the use of synthetic arms?

In the wake of the explosion of real-world data, the medical and pharmaceutical sectors have become increasingly interested in synthetic control arms in recent years. However, their use is subject to strict control by regulatory bodies, who are keen to protect patient safety by making sure that clinical trials are carried out in accordance with the very highest scientific standards. In Europe, the European Medicines Agency encourages the creation of synthetic arms in certain very specific cases, but discourages their use in clinical trials where randomized trials can be conducted ethically and within a reasonable timeframe.

^[1^] Utiliser des données externes pour l’évaluation clinique d’un DM – DeviceMed.fr – https://www.devicemed.fr/dossiers/sous-traitance-et-services/etudes-cliniques/utiliser-des-donnees-externes-pour-levaluation-clinique-dun-dm/36476#:~:text=Un%20bras%20de%20contrôle%20synthétique%20ou%20bras%20virtuel%20est%20une,les%20résultats%20du%20groupe%20expérimental

All the news from Servier Group