The treatment policy strategy describes a treatment effect regardless of whether a specific intercurrent event has occurred. Endpoint values are therefore attributed to the originally assigned treatment strategy even if a person discontinues the study treatment, receives an add-on therapy, or switches to another therapy. The strategy thus answers a question about the treatment decision originally initiated under the events that actually occur over the course of the study.
Which clinical question it defines
With this strategy, the event is neither removed from the endpoint nor reinterpreted as a hypothetical course without the event. Instead, the comparison asks about the difference between the randomized strategies when their subsequent consequences are also taken into account. A value after initiation of rescue medication therefore remains informative for the defined endpoint in principle, provided it was collected.
The estimate may therefore include consequences of the actual care pathway, including different add-on therapies. This is not a weakness of the strategy, but its specific meaning.
The term does not mean that events would be irrelevant for treatment and patient safety. “Irrelevant” refers exclusively to how the event in question is handled in the estimand attribute. Whether a treatment discontinuation is clinically meaningful continues to be assessed via its reasons, frequency, safety data, and additional endpoints.
Follow-up requirements
A treatment policy estimate depends on values after the event. The protocol must therefore specify that endpoints should, where possible, also be collected after discontinuation of the randomized treatment or after a therapy switch. If such visits are stopped because sites only follow the treatment phase, the very data range that this strategy needs for the primary question is missing.
The evaluation cannot be guaranteed by merely assigning all randomized individuals to their arm. Missing measurements after the event remain a separate analysis issue. The assumptions and sensitivity analyses must explain how unobserved courses are handled without turning the treatment policy question into a question about an event-free course.
For a terminal event such as death, the strategy cannot be applied to a symptom or functional endpoint that is only measured later, because after the event no value arises that could be attributed to the original treatment strategy. For such constellations, ICH E9(R1) refers to a composite variable or an endpoint that captures the event itself as a component. The event catalogue in the protocol must therefore specify, for each event type, whether further endpoint collection is possible at all and until what time point it is planned to continue.
The strategy is particularly appropriate when care after treatment initiation is part of the clinical reality that shapes the decision for a treatment strategy. It does not answer the narrower question of the pharmacological effect under full adherence. For interpretation, the reasons for discontinuation, the type and timing of concomitant therapy, and the distribution of values collected after the event should therefore be presented transparently. Otherwise, a large difference in endpoint collection between arms can undermine the intended strategy.
Distinction from intention-to-treat analysis
The treatment policy strategy is not a synonym for intention-to-treat analysis. It specifies which treatment effect the estimand refers to for a given intercurrent event. Intention-to-treat analysis and the full analysis set, by contrast, relate to the analysis population and the assignment of participants to the randomized group.
Both concepts can align, because an analysis according to randomization often supports the treatment policy question. Nevertheless, an analysis population alone cannot decide whether values after rescue medication, after a switch, or after a discontinuity count toward the endpoint. This decision precedes the choice of analysis set and must be specified in the estimand for each relevant event.
Relevance for clinical trials
The operational challenge lies in separating end of treatment from end of study. Study teams must capture the reasons for and timing of the event without prematurely stopping the scheduled endpoint collection. In long-running studies, contacts with individuals who are no longer receiving study medication are also important; otherwise, the treatment policy would only be described for those with complete follow-up.
Full-service CROs such as Mediconomics support the development of follow-up rules after discontinuation, coordinate the capture of concomitant and rescue therapies in the eCRF, monitor the completeness of subsequent endpoints, and, together with biostatistics and medical writing, prepare the estimand-appropriate description for the protocol and analysis plan.
Frequently Asked Questions (FAQ)
Are values after a therapy switch discarded under this strategy?
No. The inclusion of such values is precisely what characterizes it, provided the therapy switch is the relevant intercurrent event and the endpoint can subsequently be measured.
Does the treatment policy strategy require that all participants be treated until the formal end of the study?
No. Treatment may end; what is required is carefully planned collection of the results specified for the endpoint beyond the end of treatment.
Is the strategy suitable for every intercurrent event?
No. If the clinical question explicitly concerns a course without an add-on therapy or after a clearly delineated safety event, another strategy may better describe the intended treatment effect.
Regulatory References
- ICH E9(R1), Addendum on Estimands and Sensitivity Analysis in Clinical Trials – describes the treatment policy strategy as an approach to handling intercurrent events.
- ICH E9, Statistical Principles for Clinical Trials – explains analysis according to randomization and the intention-to-treat principle.
- ICH E6(R3), Guideline for Good Clinical Practice – anchors prospective documentation of study procedures and data.
