Estimands have become a central concept in modern clinical trials, particularly in oncology, where treatment pathways are complex and patient outcomes can be influenced by many post-randomization events. At their core, estimands force us to be explicit about the exact question a trial is trying to answer which has not always been clearly defined in the past. This blog explores the role of estimands in oncology clinical trials and explains how clearly defining the treatment effect of interest can improve trial design, analysis and interpretation in increasingly complex treatment settings.
In oncology, this clarity is especially important. Trials often involve long follow-up periods, multiple lines of therapy, and outcomes such as overall survival (OS) that can be affected by events occurring after treatment initiation. The estimand framework provides a structured way to align the trial objective, design, analysis, and interpretation.
Defining Estimands
An estimand is defined through five key components:
- the treatment condition,
- target population,
- variable (endpoint),
- intercurrent events, and
- summary measure.
While these elements may sound straightforward, their careful specification can significantly influence how results are interpreted.
Consider a typical oncology trial comparing an immunotherapy with chemotherapy in patients with advanced melanoma. The treatment condition might specify not just the drugs being compared, but also their dosing and administration schedules. The target population would define the specific group of patients, such as those who have progressed after prior therapy, ensuring that conclusions are appropriately scoped.
The choice of endpoint is often where complexity begins. Overall survival is widely regarded as the gold standard in oncology because it is objective and clinically meaningful. However, surrogate endpoints like objective response rate (ORR) are frequently used to provide earlier signals of efficacy. Discrepancies between these endpoints are not uncommon. For example, a treatment may show a strong improvement in tumour response without translating into a survival benefit. This raises an important question: what treatment effect are we truly trying to estimate?
This is where intercurrent events (ICEs) come into play. These are events that occur after treatment initiation and can affect either the interpretation or the measurement of the endpoint. In oncology trials, common ICEs include treatment discontinuation, initiation of subsequent therapies, or dose modifications. Each of these can blur the relationship between the assigned treatment and the observed outcome.
Different strategies exist to handle intercurrent events, and the choice among them defines the estimand. A treatment policy strategy, closely aligned with the intention-to-treat principle, includes all events as part of the outcome, reflecting real-world practice. A hypothetical strategy, on the other hand, asks what would have happened if the intercurrent event had not occurred. Other approaches, such as while-on-treatment or composite strategies, offer alternative perspectives depending on the clinical question.
Regulators View of Estimands
Draft regulatory guidance from the FDA supports a treatment policy approach for the primary analysis of overall survival. This is because survival is an unambiguous endpoint, and incorporating subsequent therapies reflects actual clinical pathways. Excluding or censoring patients could introduce bias and reduce the relevance of the findings to real-world settings.
Finally, the summary measure, such as a hazard ratio for survival or a difference in response rates, quantifies the treatment effect across the population. But even this seemingly technical choice should align with the broader clinical question defined by the estimand.
The growing emphasis on estimands reflects a broader shift in clinical research: from simply analyzing data to carefully defining the question before the data are even collected. In oncology, where treatment landscapes are rapidly evolving and patient journeys are increasingly complex, this shift is not just methodological, it is essential for generating results that are both scientifically robust and clinically meaningful.
Conclusion
As oncology treatment pathways become more complex, it is increasingly important to clearly define the question a clinical trial is intended to answer. The estimand framework provides a structured approach to addressing intercurrent events, aligning trial objectives with analysis methods, and ensuring that treatment effects are interpreted consistently. By carefully specifying estimands during trial planning, sponsors can generate evidence that is not only statistically robust but also clinically meaningful and relevant to regulators, clinicians and patients.
Looking to incorporate estimands effectively into your oncology development program? Phastar’s oncology and estimands experts support sponsors with trial design, endpoint strategy, estimand definition, statistical analysis and regulatory interactions.
