Bridging the gap: How Target Trial Emulation transforms real-world data into credible evidence

Adrienne Brackey: In clinical trials, stricter eligibility criteria create a more homogeneous cohort. There are often requirements on age or performance status that enroll healthier patients, often without additional comorbidities. In contrast, real-world populations more accurately reflect the patients seen in routine care, who may have varying functional status, be older, or simply not qualify for trials.

Beyond the patient characteristics, there are also differences in how data elements are ascertained. In a clinical trial, outcomes are measured under protocol-driven conditions at regular intervals using standardized criteria like RECIST v1.1. In the real world, data is often event-driven. Images are not typically taken at such regular intervals, response criteria are not standardized, and patients have variable follow-up windows. Comparing RWD to clinical trial data without adjustment is risky because it overlooks these fundamental differences and can lead to flawed conclusions about a therapy’s effectiveness.

Adrienne Brackey: TTE is a structured methodology that attempts to create a randomized clinical trial structure using observational data. The goal is to mitigate common biases encountered in observational research, such as immortal time bias. Instead of comparing observational data directly to trial data, TTE requires you to explicitly define the protocol of a hypothetical trial and then emulate that protocol as closely as possible with your RWD.

This involves prespecifying several key components: the inclusion and exclusion criteria, the treatment strategies being compared, an appropriate index date (time zero), a relevant follow-up period, and clear definitions for the outcomes. By applying a more robust set of operational definitions, we ensure we are only including patients who meet the prespecified criteria, creating a more focused and reliable comparison.

Adrienne Brackey: We define real-world objective response rate as the proportion of patients who achieve a best overall response of a complete or partial response, based on physician assessment in the clinical record. When calculating this, we recommend an approach that aligns with the intention-to-treat principle used in oncology trials, as recommended by the FDA.¹ This involves using non-responder imputation (NRI), where patients with missing or undocumented responses are conservatively counted as non-responders.

While this clinician-assessed endpoint differs from the RECIST 1.1 criteria used in trials, its validity has been reinforced through collaborative research. We participated in a multi-stakeholder study with Friends of Cancer Research that demonstrated clinician-assessed responses were relatively consistent across various RWD sources. More importantly, the study showed that these real-world responses correlated with time-to-event endpoints like overall survival and time to next treatment, giving us confidence that we are measuring a clinically meaningful outcome.²

Adrienne Brackey: It’s important to emphasize that no single organization can solve the challenges of using real-world evidence alone. For instance, for a long time, there was limited industry-wide consensus on the reliability of clinician-dictated real-world responses. Rather than having each RWD client attempt to establish best practices in a silo, collaborative projects bring partners together. This allows us to test and validate different approaches across diverse data sources, including EHR data, disease registries, and claims databases. When you demonstrate consistency and validity across a number of different databases, it dramatically improves the credibility of the findings compared to a report from a single organization. Ultimately, this transparency and alignment across the industry help increase regulatory trust and accelerate the acceptance of real-world evidence as a whole.

Adrienne Brackey: I see two major advancements on the horizon. First, as our datasets grow, we have an increased opportunity to benchmark outcomes for niche, molecularly defined subgroups, like patients with EGFR mutations or high PD-L1 expression. These populations are often too small to study in traditional trials. Having a large RWD source enables precision medicine at scale and helps de-risk biomarker-driven trial designs before companies need to invest in a prospective study.

Second is the use of "digital twins"—computational models of individual patients built from rich, longitudinal multimodal data. In a research setting, these models aim to predict how an individual patient will respond to a therapy, allowing for more personalized insights beyond the cohort level, such as simulating dosage optimization.