Digital Twins: Advancing Clinical Evidence Strategy (Sponsored)

Nadea Leavitt

Imagine you have a real patient. You know their medical history how their disease responded to a specific treatment. However, what you want to know is how this same patient would respond to a different treatment. This is where digital twins come in.

Nadea Leavitt

A digital twin is created or we often say initialized from data about a real patient, and depending on the use case, can actually be updated or re-estimated as the real patient's clinical state evolves and new data becomes available.

Nadea Leavitt

Digital twins help researchers design better studies, evaluate trial strategies, generate individualized predictions, and in some scenarios, even create comparator cohorts, especially when traditional control groups are difficult to obtain.

Nadea Leavitt

The key difference is that external controls are based on observed outcomes from a cohort of real patients, whereas digital twins generate patient-level model-derived simulations of those outcomes.

Nadea Leavitt

Digital twins actually aren't a specific analytic method.

Nadea Leavitt

As the naming suggests, these combine biological knowledge with data-driven learning, often using things like knowledge graphs, ontologies, or mechanistic constraints to keep predictions clinically plausible or to help model scenarios that are not represented in real-world data, while still benefiting from the scale and flexibility of modern AI.

Nadea Leavitt

IQVIA Life Science models are a family of foundation models that are powered by such multimodal data that create a richer digital twin and make patient-level simulation scalable across diseases and use cases in a way that was previously very difficult.

Nadea Leavitt

Specifically, Life Science models learn from longitudinal real-world data on more than 200 million patients to understand how diseases progress, how patients respond to treatment, how clinical events unfold over time.

Nadea Leavitt

The richer and higher quality the information we have about the real patient, the more precisely we can initialize the digital twin.

Nadea Leavitt

For example, in the US, the 21st Century Cures Act and subsequent rules have driven adoption of standardized APIs, including fire-based APIs that make healthcare data increasingly accessible and exchangeable across health systems in a common format.

Nadea Leavitt

If a digital twin is going to influence a trial design decision, a regulatory submission, or ultimately patient access to therapy, then the question isn't whether the model is sophisticated, the question is whether it's fit for purpose we're asking it to serve.

Nadea Leavitt

The real question is whether it has demonstrated that it can reliably reproduce clinically meaningful trajectories for the specific patient population, intervention and outcome being studied.

Nadea Leavitt

Second, transparency is essential. People need to understand where the underlying data came from, how the model was trained, what assumptions are built into it, and how uncertainty is quantified.

Nadea Leavitt

As the model moves closer to generating evidence that could affect treatment effect estimates or regulatory conclusions, expectations around validation, documentation, robustness testing, and auditability become much higher.

Nadea Leavitt

It's earned by demonstrating that the model is transparent, scientifically defensible, robust under scrutiny, and reliable in the specific context where it's being applied.