Integrating multimodal real-world data at trial design: a framework for improving oncology trials

Executive Summary

The study-concept and protocol design phase is where the largest share of a trial’s probability of technical and regulatory success (PTRS) is determined. Multimodal real-world data (RWD) can be a high level input at that phase, however it is systematically under-used because sponsors typically engage data partners after the protocol has been drafted, at which point the highest-value design decisions are already locked in.

This guide describes a three-step framework for integrating multimodal RWD during trial design in order to improve the way oncology trials are constructed:

1) Characterize the target population and appropriate comparator
2) Simulate trial performance and feasibility
3) Inform enrollment forecasts and site selection.

For each step, we walk through specific protocol decisions RWD can inform and the methods that support them. The intent is a practical reference for sponsor teams designing oncology trials.

Why trial design is the right entry point for RWD

Approximately 90% of drug candidates that enter clinical trials fail to reach the market. The two largest drivers, lack of efficacy (roughly 40–50% of failures) and unmanageable toxicity (roughly 30%), are heavily influenced by decisions made before enrollment opens including who the trial enrolls, what endpoints it measures, and what comparator population is chosen.¹

We know that sponsors often engage technology and data partners after they have developed the full protocol, creating redesign risk, protocol amendments, and execution delays.²
RWD-informed decisions at trial design phase are far cheaper to make than amendments after protocol lock, and they are compounding: better cohort characterization improves feasibility assumptions which lead to better enrollment forecasts.

Which decisions can RWD inform?

Real-World Data (RWD) can play a pivotal role during the study-concept phase by informing key protocol-design decisions, including the refinement of inclusion and exclusion criteria, selection of endpoints, and calibration of sample size and event-rate assumptions. Furthermore, RWD can help establish valid comparator arms, guide site and geographic enrollment strategies, and address FDA priorities regarding equity and population representativeness. Because all of these trial parameters stem from a clear understanding of the target real-world patient population, utilizing patient-level, multimodal RWD—integrating clinical, molecular, imaging, and longitudinal outcome data—can help establish a rigorous and defensible foundation for the study design.

A three-step framework

Step 1: Characterize the target population and appropriate comparator

When drafting the trial protocol, use real-world data to characterize the target patient population and establish a realistic comparator arm. Analyzing baseline patient characteristics—including molecular profiles, treatment history, and outcomes—helps avoid efficacy failures that stem from misaligned patient selection or poorly characterized control arms. This foundational step allows researchers to define key trial parameters, such as inclusion/exclusion criteria and stratification factors, based on evidence from real world populations rather than assumptions.

Methods:

• Analyze linked clinical, molecular, and imaging data to build a comprehensive view of the patient population.
• Simulate your draft inclusion and exclusion criteria to assess the size and characteristics of the resulting cohort.
• Evaluate the distribution of key biomarkers and their correlation with outcomes to refine enrichment strategies and define cut-points.

Step 2: Simulate trial performance and feasibility

Use real-world data to simulate trial performance and feasibility including modeling how selected endpoints may behave and stress-testing the design against realistic enrollment scenarios and comparator arm outcomes. This proactive analysis helps identify potential issues—from slow-maturing endpoints to optimistic enrollment assumptions—when adjustments are still possible, reducing the risk of costly protocol amendments.

Methods:

• Benchmark potential endpoints, such as real-world overall survival and progression, to evaluate their relevance and behavior in the current standard-of-care landscape.
• Simulate the impact of different inclusion and exclusion criteria on endpoint outcomes to refine patient selection.
• Stress-test comparator-arm assumptions by analyzing real-world outcomes or constructing a formal external control arm where appropriate following FDA guidance.^3,4

Step 3: Inform enrollment forecasts and site selection

Before protocol lock, use real-world data to inform forecasts for trial enrollment and inform site selection. By applying draft protocol criteria to a large, de-identified dataset, researchers can replace guesswork with evidence-based models of patient recruitment. This enables mapping patient density, predicting potential screen-failure rates, and building more realistic enrollment timelines, helping ensure the trial is set up for success from the start.

Methods:

• Model enrollment velocity based on real-world disease incidence and treatment patterns to create more realistic timelines.
• Identify and rank potential trial sites based on actual patient density to optimize recruitment efforts.
• Stress-test the protocol by modeling potential screen-failure rates and attrition to identify overly restrictive criteria before the trial begins.

How this framework can improve PTRS

This framework can improve the probability of technical and regulatory success (PTRS) by informing key decisions throughout the trial design process:

• Improve patient selection: By characterizing the target population with multimodal data, you can enrich for responder subpopulations, supporting the primary endpoint evaluation.
• Accelerate decision-making: Simulating trial feasibility and performance allows you to select criteria that could provide earlier readouts on efficacy, enabling faster go/no-go decisions.
• Optimize trial execution: By forecasting enrollment and informing site selection with real-world data, you can create more realistic timelines and design a trial that is operationally sound, potentially reducing the risk of costly delays.

Because expected portfolio value scales with PTRS (eNPV = PTRS × value − cost), even modest compounding gains at each of these decision points may help redirect capital toward the assets most likely to reach patients.⁵

Practical considerations for sponsors

To effectively leverage real-world data in trial design, sponsors must adopt a proactive and integrated approach from the earliest stages of study development. The most valuable change is to engage RWD partners during the concept phase, not after the protocol is locked. This timing allows sponsors to evaluate platforms on their ability to support rapid, iterative analyses—a critical factor for making evidence-based decisions. As researchers begin these concept-phase discussions, they should also confirm that the data is fit for your specific purpose and consider the procurement process lead time needed to prevent delays. Internally, this approach requires earlier cross-functional coordination. Appointing a dedicated RWD lead can help align clinical development, biostatistics, health economics and outcomes research, and regulatory affairs, ensuring all functions work from a unified, data-driven strategy.

About Tempus AI

Tempus AI is a technology company advancing precision medicine through the application of AI in healthcare. Tempus operates one of the largest multimodal oncology databases, linking de-identified clinical, genomic, imaging, and outcome data at the patient level. While this guide reflects patterns from Tempus’s collaborations with biopharma partners on protocol design, endpoint validation, and external control arm construction, neither this guide nor the framework guarantees future study performance or outcomes. Tempus contributes this framework to the CTTI Clinical Trials Roadmap as a resource for sponsor teams designing oncology trials.

1. Sun D, Gao W, Hu H, Zhou S. Why 90% of clinical drug development fails and how to improve it? Acta Pharmaceutica Sinica B. 2022;12(7):3049–3062.
2. Clinical Trials Transformation Initiative. (2025, April 29). Tech, device, and data company: Final protocol. https://ctti-clinicaltrials.org/roadmap/tech-and-device-and-data-final-protocol/
3. External Control Arm with Synthetic Real-world Data for Comparative Oncology using Single Trial Arm Evidence (ECLIPSE): a case study using Lung-MAP S1400I. medRxiv preprint, 2024.
4. Considerations for the Design and Conduct of Externally Controlled Trials for Drug and Biological Products” (Feb 2023)
5. Paul et al., Nature Reviews Drug Discovery (2010), “How to improve R&D productivity.”

* Kapilivsky J, Islam F, Roth EK, et al. Validation of a composite mortality end point in a large clinicogenomic real-world database of patients with advanced cancer. JCO Clinical Cancer Informatics. 2026.

** Duke-Margolis Center for Health Policy. Determining Real-World Data’s Fitness for Use and the Role of Reliability. September 2019.