Treatment Failure Endpoint Checklist

What it is

- This guideline is the checklist layer for treatment failure, non-response, loss of response, and proxy strategy-failure endpoints in RWE. The companion concept explains the endpoint family; this guideline defines what must be specified when the endpoint is used in a protocol, SAP, manuscript, HTA appendix, or regulatory package. The core rule is separation: clinical non-response, treatment-pattern proxies, intolerance, access barriers, non-adherence, death, and loss of observability are different mechanisms and should not be collapsed silently.

When to use

- Use it when a study endpoint includes discontinuation, switch, add-on, augmentation, rescue therapy, dose escalation, hospitalization, biomarker/imaging progression, line-of-therapy transition, or clinician-documented non-response as evidence that an initial strategy failed. It is especially important in claims-only studies where the endpoint is a proxy, and in linked EHR/registry studies where clinical assessments can validate or refute the proxy. Use it at design time because adequate-trial windows, assessment dates, and intercurrent-event strategies determine who is eligible to fail.

What it requires / checklist domains

- Define the estimand first: whether the endpoint is clinical non-response, treatment-pattern failure, loss of response, composite strategy failure, or an on-treatment/per-protocol failure event. State the adequate trial window, minimum exposure or persistence requirement, first eligible assessment date, and data sources for each component. List every component that can trigger failure and retain the triggering component in the analytic dataset. Specify how discontinuation, switching, add-on, rescue therapy, dose escalation, hospitalization, death, and loss to follow-up are handled as intercurrent events. Report component-specific counts and sensitivity analyses that remove weak or ambiguous components.

When NOT to use - limitations and common misapplications

- Do not label a claims-only treatment change as clinical non-response unless it has been validated against chart, lab, imaging, registry, or clinician-assessment evidence. Do not mix non-response with intolerance, formulary access, affordability, patient preference, or administrative censoring unless those are explicitly part of the estimand. Do not allow people to fail before an adequate trial could have occurred. Do not use next-line therapy as a universal failure definition in therapeutic areas where planned sequencing or finite therapy is standard. A composite treatment-failure endpoint is only interpretable if the component distribution is visible.

How it maps to this catalog

- This guideline cross-references `treatment-failure-non-response-rwe` for the endpoint concept, `persistence-time-to-discontinuation` and `switch-add-on-augmentation-rwe` for treatment-pattern proxies, `estimands-ate-att-intercurrent-events-rwe` for intercurrent-event handling, `target-trial-emulation` for time-zero and strategy definition, `ecog-performance-status-score-rwe` and `real-world-progression-rwpfs-rwe` for oncology response context, and `attrition-and-loss-to-follow-up-rwe` for observability and censoring. Use this checklist only for the reporting/implementation gate; the concept supplies the operational content.