Clinical trials traditionally rely on control groups — typically receiving placebos or standard-of-care treatments — to establish the safety and efficacy of new therapies. However, recruiting patients for these arms can be ethically fraught or practically unfeasible in rare, progressive, or life-threatening conditions, where withholding treatment may be unreasonable or recruitment pools are too small. In complex diseases, where achieving endpoints carries greater risk, control groups add significant time and cost to the R&D process, potentially discouraging investment.

Drug Discovery News spoke with Muhunthan Thillai, Chief Executive Officer (CEO) of Qureight Ltd, and Lyn Baranowski, CEO of Avalyn Pharma, to explore how synthetic control arms are emerging as a powerful, ethical, and efficient alternative in such challenging trial landscapes.

What are synthetic control arms and why are they becoming increasingly important in clinical research?

MT: Synthetic control arms are patient groups generated for comparison purposes in a clinical trial, but that aren’t directly enrolled into the study itself. Developed using statistical methods, these models enable researchers to evaluate the effectiveness of a treatment by comparing outcomes against a well-matched virtual control group derived from existing data.

Qureight’s approach advances this concept by digitally twinning placebo or other control arm patients with treated subjects by incorporating both clinical and image data. This allows comparisons between treated and control groups that are more tightly matched to accurately reflect disease progression and treatment impact.

Synthetic control arms are increasingly important in reducing clinical trial size, duration, cost and for accelerating new therapies to patients. - Muhunthan Thillai, CEO of Qureight Ltd

Synthetic control arms are increasingly important in reducing clinical trial size, duration, cost and for accelerating new therapies to patients. This approach is especially valuable when recruiting traditional control groups is difficult or unethical, such as in rare diseases or severe conditions with limited treatment options.

While already used in rare cancer research, this approach is now expanding into diseases where progression is better characterized, and data availability is improving. In idiopathic pulmonary fibrosis (IPF), for example, Qureight’s methodology is reducing the burden on drug developers in a field where new therapies are urgently needed, but traditional study designs require large patient cohorts and multiple years to evaluate efficacy.

How does the concept of a synthetic control arm differ from the traditional use of historical control data?

MT: Historical control data involves using results from previous studies or real-world patient records to benchmark. These datasets are static and unlikely to closely match specific characteristics of a current trial’s patient population. Differences in patient demographics, disease severity, enrollment criteria, or changes in standard of care can introduce bias and limit the accuracy and relevance of the comparison.

Our approach goes a significant step further. Rather than simply referencing historical data, we use advanced imaging AI and data analytics to build digitally-twinned cohorts that closely match treatment arm groups in terms of baseline disease state.

This multi-modal approach ensures that synthetic controls are likely to reflect comparable rates of disease progression, enabling a more accurate assessment of treatment effect. This approach, especially in rare diseases or where using a placebo or no background therapy arm is unethical or impractical, reduces the need for recruiting large numbers of control patients, without compromising results.

Can you briefly describe the AP01 study and why it was an appropriate setting for introducing a synthetic control arm? What specific challenges or gaps did this approach help to address?

LB: Avalyn is aiming to redefine the standard of care in pulmonary fibrosis through an inhaled approach. Avalyn's investigational therapy, AP01, is an optimized inhaled formulation of pirfenidone for the treatment of IPF and progressive pulmonary fibrosis (PPF) (1). Oral pirfenidone is one of two approved medicines that slow pulmonary fibrosis progression, but these oral medicines are associated with significant toxicities that restrict their use and dosing. Given the severity of IPF and the limited treatment options available, including patients on no background therapy in a clinical study can be challenging and raises ethical concerns.

In order to evaluate data from a completed historic study that lacked a placebo group, Avalyn partnered with Qureight to develop a treatment-naïve synthetic placebo arm for that study evaluating AP01 in IPF patients. Using Qureight’s platform, AP01-treated patients were matched to real-world individuals based on clinical and radiological markers linked to disease progression. This enabled the creation of a tightly matched synthetic control population. From this patient dataset, more than 10,000 randomly sampled control groups were generated, and the top 1,000 best-matched cohorts were used to assess comparative efficacy for AP01, providing a rigorous, ethical alternative to a traditional placebo arm.

What types of data are critical to building reliable synthetic control arms, and how do you ensure this data is representative and robust?

MT: The foundation is having the right patient population. IPF patients have a unique disease trajectory and diagnostic criteria, so the starting point is a large dataset of the right patients. Key data types include demographic and clinical variables that align with study inclusion and exclusion criteria, as well as multiple longitudinal measures of disease progression, captured consistently at defined follow-up intervals. In this case, lung function and imaging data are essential. Both are strongly associated with progression risk, and our image-based biomarkers have been peer-reviewed and recognized as a breakthrough in the field.

Data must come from credible, validated sources and be benchmarked against known disease trajectories. We assess how patient cohorts perform against established clinical endpoints, such as lung function decline over 12 months, to confirm data quality and relevance.

These approaches ensure that the data we draw from is representative and robust. We’ve developed this approach over multiple years, with continuous improvements in data curation, statistical matching, and comparative efficacy modeling. This ensures that the virtual control arms we build are both scientifically rigorous and representative of real-world patient populations.

How do you address variability in real-world data — such as differences in imaging techniques or clinical record-keeping — when building your control models?

MT: Variability is inherent in clinical data for complex diseases like IPF. We manage this in the clinic by having large patient datasets and long follow-up periods. Our approach to real-world data is identical.

Image-wise, Qureight's platform uses proprietary artificial intelligence (AI)-based quantitative tools trained on diverse, real-world data. These tools are designed to handle variation in image acquisition methods while still reliably quantifying lung features linked to disease progression, ensuring consistency in how patients are matched.

We have the largest biorepository of IPF patients in the world. Curation involves removing low-quality or incomplete data — an intensive process, but one that only needs to be done once to ensure ongoing data integrity. We also benchmark datasets from different sources against each other, confirming consistency in clinical outcomes such as lung function decline.

From there, our data scientists have developed statistical techniques that cope with noise in the data, such as collecting imaging and lung function data at different timepoints, to ensure we can leverage as much data as possible. Collectively, after filtering by clinical trial inclusion criteria and disease-related measures, we can create tightly matched synthetic control pools from which to randomize into control arms.

How scalable is this model to other rare or chronic lung diseases with limited patient populations?

LB: This model is particularly impactful for companies like Avalyn, that are working to redefine the standard of care, especially for diseases where traditional placebo arms are challenging to recruit or unethical. This is true across many rare lung diseases, including those that Avalyn is currently focused on.

MT: The model is highly scalable, particularly in diseases with well-characterized progression and access to well-curated datasets. As long as the underlying data meets quality and volume thresholds, the same synthetic control framework can be applied effectively.

Ongoing advances in AI and statistical modeling allow us to extract meaningful insights from smaller datasets, making this approach increasingly applicable even in ultra-rare diseases. Scaling this model is central to Qureight’s future R&D strategy, enabling faster, more efficient clinical development across a wider range of conditions.

Have regulators engaged with this synthetic control model yet, and what feedback has been received from health authorities?

MT: Both the US Food and Drug Administration (FDA) and the European Medicines Agency (EMA) have issued guidance on the use of external control arms. Their use is becoming more accepted, particularly in clinical trials for rare cancers and severe diseases where traditional control arms are difficult or unethical to implement.

We hope that by building an evidence base, as we are in IPF, we can engage regulators in a rational discussion, given the development challenges and unmet patient needs. Synthetic control arms are not intended to fully replace enrolled control patients — these are still necessary to confirm that trial participants reflect a representative and unbiased population. However, synthetic arms can meaningfully reduce the number of control patients required. This not only eases the burden on drug developers but also makes clinical trial participation more appealing to patients, especially in high-need areas.

LB: Avalyn hasn’t formally proposed a synthetic control arm yet or submitted this approach for regulatory feedback but will continue to collaborate with Qureight to build evidence supporting a possible future discussion with health authorities.

What’s next for the synthetic control platform — are there plans to integrate additional biomarkers, longitudinal clinical data, or multi-omics inputs?

MT: We already incorporate longitudinal clinical and imaging data, and we’re beginning to build multi-omics datasets. While these are currently limited in size, they’re growing steadily. Our initial focus has been on generating synthetic placebo arms, and we have additional publications forthcoming that support this approach.

Now we’re confident in our statistical approach, the next step is to expand into synthetic control arms based on standard of care, enabling more precise benchmarking of investigational therapies. We also plan to extend the platform into other interstitial lung diseases, particularly where disease trajectories are well characterized.

LB: Avalyn will continue imaging analysis in the ongoing programs. If the opportunity becomes relevant to look at specific synthetic control groups, we’ll absolutely evaluate that.

References:

1. West, A. et al. Inhaled pirfenidone solution (AP01) for IPF: a randomised, open-label, dose-response trial. Thorax 78, 882–889 (2023).