Treatment-Regimen Estimands in FDA Labels: Answering the Real Coverage Question

When a pivotal Phase 3 readout breaks, headline press releases routinely lead with the largest possible number: a 15% or 20% mean reduction in a primary outcome under an "efficacy," "trial-product," or "hypothetical" estimand. Months later, when the U.S. Food and Drug Administration (FDA) approves the drug and posts the prescribing information, the primary efficacy table in Section 14 (Clinical Studies) often tells a different story. The labeled number is smaller—a 9% or 11% mean reduction under an intention-to-treat (ITT) or "treatment-regimen" estimand.

For Pharmacy and Therapeutics (P&T) committees, health plan medical directors, and Health Economics and Outcomes Research (HEOR) teams reviewing Academy of Managed Care Pharmacy (AMCP) dossiers, this discrepancy is not an accounting quirk. It represents two entirely different clinical and economic questions:

The Treatment-Regimen Question: What is the effect of assigning this regimen, knowing that a substantial share of patients will discontinue, miss doses, or start rescue therapy after randomization?
The Hypothetical Question: What would the effect have been in a scenario where those intercurrent events did not occur?

For commercial payers and self-funded plan sponsors, the treatment-regimen (treatment-policy) number is the coverage-relevant question because it includes discontinuation and rescue. The hypothetical number answers a different clinical question and belongs in a supplementary appendix, not as the primary effect in a cost-effectiveness model.

This guide examines how the International Council for Harmonisation (ICH) E9(R1) framework operates across regulatory submissions, prescribing information, and formulary evaluations—and what clinical development and market access teams must coordinate before trial databases and electronic clinical outcome assessment (eCOA) systems are locked.

What Did FDA Actually Adopt in May 2021, and Is ICH E9(R1) a Binding Requirement or Nonbinding Guidance?

The structural gap between trial objectives, statistical analysis plans (SAPs), and regulatory claims led international regulators to overhaul clinical trial statistical principles. On November 20, 2019, the ICH Assembly endorsed the Step 4 final guideline for ICH E9(R1): Addendum on Estimands and Sensitivity Analysis in Clinical Trials to the Guideline on Statistical Principles for Clinical Trials.

FDA formally adopted the addendum into U.S. regulatory practice through a notice of availability published in the Federal Register on May 12, 2021 (86 FR 26047; Docket No. FDA-2017-D-6113; FDA guidance page content current May 11, 2021). The Federal Register notice confirmed that the document finalized the draft guidance issued on October 31, 2017.

┌────────────────────────────────────────────────────────────────────────────┐
│                    ICH E9(R1) REGULATORY MILESTONES                        │
├────────────────────┬───────────────────────────────────────────────────────┤
│ October 31, 2017   │ FDA issues draft guidance E9(R1) for public comment   │
├────────────────────┼───────────────────────────────────────────────────────┤
│ November 20, 2019  │ ICH Assembly endorses Step 4 Final Harmonised Guideline│
├────────────────────┼───────────────────────────────────────────────────────┤
│ May 12, 2021       │ FDA publishes Final Guidance notice (86 FR 26047)      │
├────────────────────┼───────────────────────────────────────────────────────┤
│ April 7, 2022      │ FDA finalizes companion ICH E8(R1) General Design     │
└────────────────────┴───────────────────────────────────────────────────────┘

Like all FDA guidance documents, ICH E9(R1) represents the agency's current thinking under 21 CFR 10.115 and does not establish legally enforceable responsibilities unless specific statutory or regulatory requirements are cited.

FDA's subsequent finalization of ICH E8(R1) (General Considerations for Clinical Studies, April 2022; content current April 7, 2022) states that study objectives are further refined through specification of estimands and that the protocol should define the estimand(s) following the ICH E9(R1) framework. That is guidance language, not a statute, but it is the design specification reviewers now look for before visit schedules and outcome collection are locked.

Why Is a Protocol Objective Plus a Named Endpoint Still Not an Estimand?

In legacy clinical trial protocols, teams often treated a statement such as "To evaluate the efficacy of Drug X compared to placebo in reducing body weight at Week 72" as a complete specification.

ICH E9(R1) established that an endpoint (or variable) is merely one component of a clinical question. A protocol that names a 72-week measurement without defining how post-baseline disruptions are handled does not define what is being estimated.

Under ICH E9(R1), an estimand is a precise description of the treatment effect reflecting the clinical question posed by the trial objective. It consists of five interrelated attributes that must be locked in the protocol and SAP:

                      ┌────────────────────────────────────────┐
                      │        THE 5 ESTIMAND ATTRIBUTES       │
                      └───────────────────┬────────────────────┘
                                          │
        ┌─────────────────┬───────────────┴───────────────┬─────────────────┐
        ▼                 ▼                               ▼                 ▼
 ┌──────────────┐  ┌──────────────┐                ┌──────────────┐  ┌──────────────┐
 │  Treatment   │  │  Population  │                │ Intercurrent │  │  Population  │
 │  Condition   │  │   of Target  │                │    Events    │  │   Summary    │
 │ (Investig-   │  │   Subjects   │                │  (Handling   │  │  (Difference,│
 │  ational vs  │  │ (Inclusion / │                │  Strategy    │  │  Odds Ratio, │
 │  Comparator) │  │  Exclusion)  │                │  Pre-Locked) │  │ Hazard Ratio)│
 └──────────────┘  └──────────────┘                └──────────────┘  └──────────────┘
                                          ▲
                                          │
                                   ┌──────────────┐
                                   │   Variable   │
                                   │  (Endpoint   │
                                   │ Measurement) │
                                   └──────────────┘

Treatment Condition: The investigational intervention (including dose, regimen, and background therapy) and the control/comparator condition.
Target Population: The broader patient population to which the clinical question applies, defined by inclusion and exclusion criteria and baseline disease characteristics.
Variable (Endpoint): The specific measurement obtained for each subject (e.g., percent change in body weight from baseline to Week 72, absolute change in HbA1c, or change in a patient-reported outcome score).
Handling of Intercurrent Events (ICEs): The explicit, pre-specified strategy used to account for events occurring after treatment initiation that either preclude observation of the variable or alter its clinical interpretation.
Population-Level Summary: The mathematical measure used to compare treatment conditions across the population (e.g., difference in mean change from baseline, odds ratio, relative risk, or hazard ratio).

Intercurrent Events Are Not Missing Data

A critical conceptual advance in ICH E9(R1) is the absolute distinction between intercurrent events and missing data:

An Intercurrent Event (ICE) is a clinical reality that occurred during the trial—such as treatment discontinuation due to an adverse event, patient non-compliance, initiation of prohibited rescue medication, dose reduction, or death. These events change the meaning or possibility of future measurements.
Missing Data is an operational failure where a measurement was intended to be collected according to the estimand definition but was not obtained (e.g., a missed clinic visit, a lost laboratory sample, or early withdrawal from trial follow-up).

When sponsors conflate ICEs with missing data, statistical analysis plans default to mathematical imputations (such as Last Observation Carried Forward or standard mixed models for repeated measures) that inadvertently alter the clinical question being answered.

How the Five ICH Strategies Change the Labeled and Payer Claim

ICH E9(R1) outlines five distinct strategies for addressing intercurrent events. Choosing among these strategies is not a mathematical preference; it fundamentally alters the clinical claim made to regulators and payers.

Strategy	Definition	How Data After ICE Are Treated	What Question Does It Answer?	P&T / Coverage Implication
Treatment Policy	The variable is evaluated regardless of whether the ICE occurs.	All data collected after ICE (e.g., post-discontinuation or on rescue therapy) are included in the primary analysis.	What is the effect of initiating the treatment regimen, inclusive of real-world tolerability and rescue needs?	Primary Coverage Metric. Reflects net clinical effectiveness and total drug cost in plan populations.
Hypothetical	A scenario is envisaged in which the ICE would not occur.	Post-ICE data are replaced by modeled values estimating what would have occurred had the ICE been prevented.	What would the biological efficacy be if patients never discontinued therapy and never required rescue?	Mechanistic / Sensitivity Only. Overestimates real-world benefit by ignoring tolerability failure.
Composite Variable	The ICE is integrated directly into the endpoint definition.	The occurrence of the ICE is classified as treatment failure or scored as an unfavorable outcome.	What proportion of patients achieve clinical benefit without experiencing treatment failure or death?	Defensible for Severe ICEs. Common in oncology (PFS, OS) and infectious disease composite cures.
While on Treatment	The variable is evaluated only up to the occurrence of the ICE.	Data collected after the ICE are excluded; response is measured during active exposure.	What is the average rate or effect per unit of time while patients are actively adhering to therapy?	Useful for Safety / Tolerability. Misleading as a primary efficacy endpoint for chronic maintenance drugs.
Principal Stratum	Analysis is restricted to the subpopulation in whom the ICE would (or would not) occur under both arms.	Evaluates the effect only in the subset of patients destined to tolerate both treatments.	What is the treatment effect in biological responders who can tolerate therapy without the ICE?	Exploratory / High Risk. Confounding post-randomization selection bias makes payer adoption risky.

The Mathematical Divergence: Treatment Policy vs. Hypothetical

Consider a Phase 3 trial in a chronic metabolic or immunological condition where 25% of active-arm patients discontinue study medication due to gastrointestinal adverse events by Week 24:

Under a Hypothetical Estimand, the statistical model imputes Week 72 outcomes for those discontinuing patients based on the trajectory of patients who remained on active drug. The resulting treatment difference might show a robust -15.0% effect.
Under a Treatment-Policy (Treatment-Regimen) Estimand, the trial collects actual Week 72 measurements from all randomized patients, including those who stopped drug and regained weight or switched to alternative therapies. The resulting treatment difference shows a -9.0% effect.

If a manufacturer submits a pharmacoeconomic model to a health plan using the -15.0% hypothetical figure while plan members experience real-world discontinuation rates matching the 25% trial drop-out, the plan's realized cost-per-responder will be substantially higher than modeled.

For broader context on how clinical outcome assessments translate into payer policies, see our analysis of PRO trial-to-payer evidence gaps.

Which Labeled Foundayo Table 6 Number Is the Treatment-Regimen Analysis?

The real-world implementation of ICH E9(R1) in U.S. labeling is exemplified by the FDA prescribing information for Foundayo (orforglipron) oral tablets, approved in 2026 for chronic weight management (SPL effective 20260729; SetID: 8ac446c5-feba-474f-a103-23facb9b5c62).

In Section 14 (Clinical Studies) of the Foundayo label, Table 6 details the primary efficacy results at Week 72 across Trial 1 (adults with obesity or overweight without type 2 diabetes) and Trial 2 (adults with type 2 diabetes and overweight/obesity).

┌────────────────────────────────────────────────────────────────────────────┐
│         FOUNDAYO (ORFORGLIPRON) TABLE 6 — WEEK 72 PRIMARY EFFICACY         │
├───────────────────────────────┬────────────┬───────────┬───────────────────┤
│ Trial / Arm                   │ N (ITT)    │ % Change  │ Placebo-Diff (95% CI)│
├───────────────────────────────┼────────────┼───────────┼───────────────────┤
│ TRIAL 1 (without diabetes)    │            │           │                   │
│ • Placebo                     │    949     │  -2.1     │        —          │
│ • Foundayo 5.5 mg             │    723     │  -7.4     │ -5.3 (-6.1, -4.4) │
│ • Foundayo 9 mg               │    725     │  -8.3     │ -6.2 (-7.1, -5.3) │
│ • Foundayo 17.2 mg            │    730     │ -11.1     │ -9 (-10, -8.1)    │
├───────────────────────────────┼────────────┼───────────┼───────────────────┤
│ TRIAL 2 (with type 2 diabetes)│            │           │                   │
│ • Placebo                     │    630     │  -2.5     │        —          │
│ • Foundayo 5.5 mg             │    329     │  -5.1     │ -2.6 (-3.5, -1.6) │
│ • Foundayo 9 mg               │    332     │  -7       │ -4.5 (-5.4, -3.6) │
│ • Foundayo 17.2 mg            │    322     │  -9.6     │ -7.1 (-8.1, -6.1) │
└───────────────────────────────┴────────────┴───────────┴───────────────────┘
Label Table 6 reports ITT percent change from baseline (ANCOVA). The percent-change
differences shown here carry multiplicity-controlled p<0.001; some responder rows
in the full table are not multiplicity-controlled.

Why Discontinuation and Missingness Must Be Inspected Alongside Table 6

In the footnotes to Figures 1 and 2, the prescribing information defines "W72 TRE" as the treatment effect under the treatment regimen estimand at Week 72 and states that the Week 72 model-based estimates use primary modified multiple imputation.

The label reports that in Trial 1:

Study drug discontinuation occurred in 22% (5.5 mg), 22% (9 mg), and 24% (17.2 mg) of Foundayo-treated patients, compared to 30% of placebo patients.
Missing Week 72 body weight measurements occurred in 16% (5.5 mg), 14% (9 mg), and 15% (17.2 mg) of Foundayo patients, compared to 24% of placebo patients.

Because nearly one in four patients on the 17.2 mg arm discontinued trial drug before Week 72, the treatment-regimen estimand incorporates post-discontinuation trajectories rather than asking what would have happened if everyone remained on assigned therapy. An "efficacy" or hypothetical number that conditions on remaining on treatment would look larger, but it would answer a different question than the labeled ITT percent-change table.

This article uses Foundayo only as a current labeled example of that distinction. For access, pricing, and prior-authorization workflow, see the Foundayo coverage guide. For earlier pipeline readouts and approval milestones, compare this with our Q1-Q2 2026 Foundayo approval recap and our analysis of the CagriSema REDEFINE estimand split.

Registry and Label Census: How Common Are Explicit Estimand Citations?

To assess how widely the terminology of ICH E9(R1) has penetrated regulatory filings and public trial registries, we conducted an empirical census across official U.S. datasets.

┌────────────────────────────────────────────────────────────────────────────┐
│                    REGISTRY & LABEL CENSUS SUMMARY                         │
├─────────────────────────────────────┬──────────────────────────────────────┤
│ Metric                              │ Census Count (As of Aug 29, 2026)     │
├─────────────────────────────────────┼──────────────────────────────────────┤
│ openFDA Labels Containing "Estimand"│ 1 SPL (Foundayo)                     │
├─────────────────────────────────────┼──────────────────────────────────────┤
│ ClinicalTrials.gov Total "Estimand" │ 68 Studies                           │
│ • Industry-Sponsored                │ 44 Studies                           │
│ • Industry Phase 3                  │ 25 Studies                           │
├─────────────────────────────────────┼──────────────────────────────────────┤
│ ClinicalTrials.gov Quoted Phrases   │                                      │
│ • "treatment policy estimand"       │ 9 Studies                            │
│ • "efficacy estimand"               │ 6 Studies                            │
│ • "hypothetical estimand"           │ 5 Studies                            │
│ • "treatment-regimen estimand"      │ 2 Studies                            │
│ • "trial product estimand"          │ 1 Study                              │
├─────────────────────────────────────┼──────────────────────────────────────┤
│ CDISC NCI-EVS Terminology (66,529)  │ 16 DDF Rows (C188813, C188815, etc.) │
│ • Legacy ADaM / Protocol Terminology│ 0 Matches                            │
├─────────────────────────────────────┼──────────────────────────────────────┤
│ NIH Common Data Elements (22,308)   │ 5,772 endpoint-relevant (not estimands)│
└─────────────────────────────────────┴──────────────────────────────────────┘

1. openFDA Drug Label Search

A live query of the openFDA Drug Product Labeling API on August 29, 2026, for the keyword estimand returned exactly 1 Structured Product Label (SPL): Foundayo (orforglipron).

Searches for quoted phrases such as "trial product estimand", "treatment policy estimand", and "efficacy estimand" yielded 0 indexed SPLs.

Methodological Caution: The absence of the literal word "estimand" in major labels (such as Wegovy or Zepbound) does not mean those products lack treatment-policy analyses. Most FDA labels present ITT and treatment-regimen data under traditional terminology (e.g., "Full Analysis Set," "regardless of treatment discontinuation") without explicitly incorporating the technical term "estimand" into Section 14 narrative prose.

2. ClinicalTrials.gov Protocol Registry

Querying the ClinicalTrials.gov API v2 on August 29, 2026, identified 68 registered clinical studies referencing estimand. When filtered by sponsor and phase:

44 studies had LeadSponsorClass categorized as INDUSTRY.
25 studies were industry-sponsored PHASE3 trials.

Specific quoted phrase occurrences across all registry records remained highly concentrated:

"treatment policy estimand": 9 studies
"efficacy estimand": 6 studies
"hypothetical estimand": 5 studies
"treatment-regimen estimand": 2 studies
"trial product estimand": 1 study

While registry mentions reflect increasing awareness, full-text registry fields do not guarantee that the trial's SAP or operational data collection adheres to the five ICH E9(R1) attributes.

3. CDISC Standards and NIH Common Data Elements

Inspection of the CDISC NCI-EVS Controlled Terminology snapshot (August 26, 2026; 66,529 terms) shows that explicit estimand entities—including Estimand (C188813), Intercurrent Event (C188815), and Intercurrent Event Strategy (C188857)—are coded in the Digital Data Flow (DDF) collection (16 matching terms). The corresponding ADaM and Protocol terminology files in that snapshot contain no estimand or intercurrent-event rows. Naming a CDISC term still does not specify an estimand.

Similarly, an audit of the NIH Common Data Elements (CDE) repository (22,308 total records; 5,772 endpoint-relevant records; 1,264 patient-reported instruments) confirms that while CDEs standardize instrument variables, naming a CDE does not specify an estimand. An instrument defines the measurement; the estimand defines the clinical effect.

When Can a Secondary COA Endpoint Enter Labeling Under FDA's 2022 Multiplicity Guidance?

In many Phase 3 programs, sponsors attempt to supplement primary treatment-regimen outcomes with patient-reported outcomes (PROs) or clinical outcome assessments (COAs) evaluating fatigue, physical function, or symptom burden. However, presenting a statistically significant secondary PRO in an AMCP dossier does not mean the claim is eligible for FDA labeling.

Under FDA's final guidance, Multiple Endpoints in Clinical Trials (October 2022; Docket No. FDA-2016-D-4460), any secondary endpoint intended to support a labeling claim must be included in a pre-specified statistical testing strategy that controls the study's overall Type I error probability. The guidance notes that the most widely used two-sided alpha is 0.05.

                              ┌────────────────────────┐
                              │  PRIMARY ENDPOINT(S)   │
                              │ (Multiplicity Locked)  │
                              └───────────┬────────────┘
                                          │
                                          ▼ If p < 0.05
                              ┌────────────────────────┐
                              │ KEY SECONDARY ENDPOINT │
                              │ (Formal Gatekeeping /  │
                              │  Hierarchical Testing) │
                              └───────────┬────────────┘
                                          │
                        ┌─────────────────┴─────────────────┐
                        ▼ If p < 0.05                       ▼ If p >= 0.05
             ┌─────────────────────┐             ┌─────────────────────┐
             │ Labeled Secondary   │             │ Testing Stops /     │
             │ Claim Permitted     │             │ Exploratory Only    │
             │ (If Hierarchy Holds)│             │ (No Label Claim)    │
             └─────────────────────┘             └─────────────────────┘

If a trial tests secondary endpoints without controlling overall Type I error—or if testing fails at a higher hierarchical node—subsequent PRO findings are classified as exploratory. FDA's multiplicity guidance treats exploratory endpoints as research or hypothesis-generation analyses; they should not appear in Section 14 in a way that implies a statistically rigorous conclusion.

PFDD Guidance 4 Remains Draft

Sponsors frequently cite FDA's Patient-Focused Drug Development (PFDD) series as authority for incorporating COAs into trial endpoints. However, regulatory and market access teams must recognize that:

PFDD Guidance 3 (Selecting What to Measure and Selecting or Developing Fit-for-Purpose Clinical Outcome Assessments) was finalized in October 2025 (Docket No. FDA-2022-D-1385; content current October 23, 2025).
PFDD Guidance 4 (Incorporating Clinical Outcome Assessments into Endpoints for Regulatory Decision-Making), which addresses responder definitions, COA estimands, and score interpretation, remains a draft guidance (issued April 2023; Docket No. FDA-2023-D-0026; content current April 5, 2023) and is explicitly marked "Not for implementation."

Dossier reviewers should reject manufacturer assertions that a custom PRO responder threshold is "validated under finalized FDA Guidance 4."

The Operational Rule: Locking Intercurrent Event Strategies Before eCOA Calendars Are Built

The most frequent point of failure in modern Phase 3 programs is not biostatistical theory; it is operational disconnect. Clinical operations, data management, and digital health technology teams often build eCOA platforms and clinical visit calendars before the biostatistics team finalizes the ICH E9(R1) estimand strategy.

If a trial specifies a Treatment-Policy Estimand, the operational protocol must require that subjects who discontinue study medication continue to attend clinic visits, undergo laboratory evaluations, and complete eCOA questionnaires through the final trial milestone.

┌────────────────────────────────────────────────────────────────────────────┐
│                    OPERATIONAL INTERLOCKED WORKFLOW                        │
├────────────────────────────────────────────────────────────────────────────┤
│ 1. Lock Primary Estimand in Protocol (Treatment Policy vs. Other)          │
│                                    │                                       │
│ 2. Define Site Retention Procedures for Discontinued Subjects              │
│                                    │                                       │
│ 3. Program eCOA Devices: Never Auto-Deactivate Accounts on Drug Stop       │
│                                    │                                       │
│ 4. Track Specific ICE Types (Adverse Event vs. Lack of Efficacy vs. Other) │
│                                    │                                       │
│ 5. Execute Primary Analysis (Modified Multiple Imputation / Direct Post-ICE│
└────────────────────────────────────────────────────────────────────────────┘

When clinical operations teams fail to align protocol operations with the statistical estimand, severe implementation errors occur:

Automated eCOA Lockout: Provisioned eCOA devices or BYOD applications automatically deactivate subject profiles when a site records "study drug discontinued," destroying the ability to collect post-ICE patient-reported outcomes.
Missing ICE Reason Coding: Electronic Case Report Forms (eCRFs) capture that a patient stopped treatment but fail to record why (e.g., intolerable toxicity vs. elective withdrawal vs. disease progression), preventing biostatisticians from applying differential imputation strategies.
Unplanned Informative Censoring: Sites assume that "treatment stopped" equals "trial completed," converting an intended treatment-policy analysis into an unintended while-on-treatment or complete-case analysis.

For trial operations teams, specifying intercurrent-event strategies before visit windows and outcome collection are locked is an essential safeguard against building data collection architectures that inadvertently bias regulatory endpoints.

For related digital measurement and sensor qualification standards, see our analysis of the digital health endpoint qualification gap. For comparative trial architectures outside randomized designs, review our guide on the external control arm evidence standard.

A 2025 Statistics in Medicine Perspective, Including FDA Biostatistics Co-Authors

A May 2025 Statistics in Medicine perspective (Fleming TR, Carroll KJ, Wittes J, Emerson SS, Rothmann MD, Collins S, Levin G; Stat Med 2025; 44(10-12): e70104; doi: 10.1002/sim.70104; PMID 40394856; PMC12092964) includes co-authors from the FDA CDER Office of Biostatistics. It is a peer-reviewed perspective, not a substitute for ICH E9(R1) and not binding FDA policy.

The authors argue that, under the influence of the Addendum, many trials have proposed while-on-treatment, hypothetical, or principal-stratum primaries for intercurrent events such as treatment discontinuation or rescue medication:

Randomization Preservation: While-on-treatment, hypothetical, and principal-stratum strategies often use post-randomization information in ways that do not preserve the protection of baseline randomization.
Adoption Relevance: The authors prefer treatment-policy strategies for discontinuation and rescue when the question is the net effect of adopting an intervention, not the effect that would have been observed if those events had not occurred.
Handling Terminal Events: Treatment policy cannot be applied to death because post-death measurements do not exist. In those settings, the authors discuss composite strategies as a defensible alternative.

Important Caveat: Co-authorship by FDA Office of Biostatistics staff does not convert this article into FDA guidance or into a description of how every review division will treat an End-of-Phase 2 package.

AMCP Dossier and P&T Review Checklist: What Access Teams Must Demand

When a manufacturer submits an AMCP Format Version 5.0 formulary dossier to a health plan or PBM, the market-access and clinical review team should run a systematic estimand audit.

┌────────────────────────────────────────────────────────────────────────────┐
│                    AMCP DOSSIER ESTIMAND AUDIT CHECKLIST                   │
├────────────────────────────────────────────────────────────────────────────┤
│ [ ] 1. Identify the Primary Estimand Strategy                              │
│     • Does the primary efficacy claim use a Treatment-Policy strategy?     │
│     • If a Hypothetical estimand is presented, is it labeled supplementary?│
├────────────────────────────────────────────────────────────────────────────┤
│ [ ] 2. Quantify Discontinuation and Rescue Rates                           │
│     • What percentage of active-arm patients discontinued study drug?      │
│     • What percentage initiated non-protocol rescue medication?            │
├────────────────────────────────────────────────────────────────────────────┤
│ [ ] 3. Audit Missing Data Imputation Methods                               │
│     • How were post-discontinuation data obtained or imputed?              │
│     • Are jump-to-reference or copy-reference sensitivity analyses shown?  │
├────────────────────────────────────────────────────────────────────────────┤
│ [ ] 4. Verify Multiplicity Control on Secondary PRO/COA Endpoints          │
│     • Was the secondary PRO included in the formal gatekeeping hierarchy?  │
│     • Did all preceding primary and secondary endpoints achieve p < 0.05?  │
├────────────────────────────────────────────────────────────────────────────┤
│ [ ] 5. Align Budget Impact Models to the Labeled Effect                    │
│     • Does the cost-effectiveness model use the labeled treatment-regimen  │
│       number rather than the press-release efficacy number?                │
└────────────────────────────────────────────────────────────────────────────┘

Key Decision Rules for P&T Committees

Rule 1: Base Tier Placement on the Labeled Treatment-Regimen Number. If a drug demonstrates an 18% reduction under a hypothetical estimand but only an 8% reduction under the labeled treatment-regimen analysis, evaluate cost-effectiveness and comparative value using the 8% figure.
Rule 2: Disallow Unadjusted Headline Comparisons. If Drug A quotes an efficacy estimand (-15%) from a Phase 2 trial and Drug B quotes a labeled treatment-regimen estimand (-10%) from a pivotal Phase 3 trial, reject the cross-trial comparison as structurally confounded.
Rule 3: Require Same-Estimand Sensitivity Analyses. Under ICH E9(R1), a valid sensitivity analysis must target the same estimand under alternative statistical assumptions (e.g., testing different imputation models for missing post-ICE data). Analyses that alter the handling of intercurrent events (e.g., converting a treatment-policy analysis to an on-treatment analysis) are supplementary analyses, not sensitivity checks.

Frequently Asked Questions

Is a treatment-regimen estimand the same thing as intention-to-treat (ITT)?

Not entirely. While a treatment-regimen estimand relies on the treatment-policy strategy and analyzes subjects according to their randomized assignment regardless of post-baseline adherence (the core ITT philosophy), ICH E9(R1) provides a far more rigorous framework. Legacy ITT principles did not explicitly define how to distinguish between missing data mechanisms, rescue therapy initiation, or composite terminal events. The treatment-regimen estimand formalizes the exact handling of each specific intercurrent event alongside the population summary measure.

Can a hypothetical or efficacy estimand support a coverage decision if it is statistically significant?

A hypothetical estimand can provide supportive evidence regarding biological mechanism of action, but it should rarely serve as the primary basis for a formulary coverage or pricing decision. Hypothetical estimands answer what would happen if patients never stopped drug or suffered side effects. Because commercial health plans pay for real-world prescription fills—including patients who discontinue due to adverse events—basing coverage on hypothetical numbers substantially overestimates cost-effectiveness and real-world return on investment.

Did FDA finalize PFDD Guidance 4 on turning COA scores into endpoints?

No. As of August 2026, PFDD Guidance 4 (Incorporating Clinical Outcome Assessments into Endpoints for Regulatory Decision-Making) remains an official draft guidance (issued April 2023; Docket No. FDA-2023-D-0026; content current April 5, 2023) and is marked "Not for implementation." In contrast, PFDD Guidance 3 was finalized by FDA in October 2025 (Docket No. FDA-2022-D-1385; content current October 23, 2025).

Does naming an NIH common data element or a CDISC term establish an estimand?

No. Naming an NIH Common Data Element (CDE) or a CDISC instrument variable defines only the variable (measurement) attribute of an estimand. It does not define the treatment condition, the target population, the strategy for handling intercurrent events (such as drug discontinuation or rescue), or the population-level summary measure. All five attributes must be specified to construct an estimand.

Sources

U.S. Food and Drug Administration (FDA). E9(R1) Statistical Principles for Clinical Trials: Addendum: Estimands and Sensitivity Analysis in Clinical Trials. Final Guidance for Industry. Content current May 11, 2021. Docket No. FDA-2017-D-6113.
https://www.fda.gov/regulatory-information/search-fda-guidance-documents/e9r1-statistical-principles-clinical-trials-addendum-estimands-and-sensitivity-analysis-clinical
Federal Register / FDA. E9(R1) Statistical Principles for Clinical Trials: Addendum: Estimands and Sensitivity Analysis in Clinical Trials; International Council for Harmonisation; Guidance for Industry; Availability. 86 FR 26047; May 12, 2021 (Doc. No. 2021-10066).
https://www.federalregister.gov/documents/2021/05/12/2021-10066/e9r1-statistical-principles-for-clinical-trials-addendum-estimands-and-sensitivity-analysis-in
International Council for Harmonisation (ICH). ICH Harmonised Guideline: Addendum on Estimands and Sensitivity Analysis in Clinical Trials to the Guideline on Statistical Principles for Clinical Trials E9(R1). Step 4 version; November 20, 2019.
https://database.ich.org/sites/default/files/E9-R1_Step4_Guideline_2019_1203.pdf
U.S. Food and Drug Administration (FDA). E8(R1) General Considerations for Clinical Studies. Guidance for Industry. April 2022; content current April 7, 2022. Docket No. FDA-2019-D-3049.
https://www.fda.gov/regulatory-information/search-fda-guidance-documents/e8r1-general-considerations-clinical-studies
U.S. Food and Drug Administration (FDA). Multiple Endpoints in Clinical Trials: Guidance for Industry. October 2022. Docket No. FDA-2016-D-4460.
https://www.fda.gov/media/162416/download
National Library of Medicine (NLM) / DailyMed. FOUNDAYO (orforglipron) tablets prescribing information. SPL SetID: 8ac446c5-feba-474f-a103-23facb9b5c62; effective July 29, 2026.
https://dailymed.nlm.nih.gov/dailymed/drugInfo.cfm?setid=8ac446c5-feba-474f-a103-23facb9b5c62
Fleming TR, Carroll KJ, Wittes J, Emerson SS, Rothmann MD, Collins S, Levin G. A Perspective on the Appropriate Implementation of ICH E9(R1) Addendum Strategies for Handling Intercurrent Events. Statistics in Medicine. 2025; 44(10-12): e70104. doi: 10.1002/sim.70104. PMID: 40394856; PMCID: PMC12092964.
https://pmc.ncbi.nlm.nih.gov/articles/PMC12092964/
U.S. Food and Drug Administration (FDA). Patient-Focused Drug Development: Selecting, Developing, or Modifying Fit-for-Purpose Clinical Outcome Assessments (Guidance 3). Final Guidance. October 2025; content current October 23, 2025. Docket No. FDA-2022-D-1385.
https://www.fda.gov/regulatory-information/search-fda-guidance-documents/patient-focused-drug-development-selecting-developing-or-modifying-fit-purpose-clinical-outcome
U.S. Food and Drug Administration (FDA). Patient-Focused Drug Development: Incorporating Clinical Outcome Assessments into Endpoints for Regulatory Decision-Making (Guidance 4). Draft Guidance for Industry. Content current April 5, 2023. Docket No. FDA-2023-D-0026.
https://www.fda.gov/regulatory-information/search-fda-guidance-documents/patient-focused-drug-development-incorporating-clinical-outcome-assessments-endpoints-regulatory