Emerging Endpoints in Oncology Trials

Definition

Emerging endpoints are clinical trial endpoints that are under active regulatory evaluation but have not yet achieved the same established acceptance as OS, PFS, or DFS. The key emerging endpoints in oncology as of 2026 are:

MRD-negative CR rate — minimal residual disease negativity plus complete response; furthest advanced in multiple myeloma (FDA draft January 2026)
pCR (pathological complete response) — established surrogate for neoadjuvant breast cancer; emerging in other neoadjuvant settings
ctDNA (circulating tumor DNA) — liquid biopsy biomarker for early response and MRD detection; regulatory status exploratory across all indications
EFS-12/EFS-24 (event-free survival landmarks) — validated surrogates in aggressive lymphoma per Patterson et al. (2019); emerging in other hematologic malignancies
PFS2 (time to second progression) — capture entire treatment sequence value; supportive/descriptive endpoint, no formal regulatory guidance
MFS (metastasis-free survival) — established specifically for non-metastatic CRPC; not applicable to other tumor types

All emerging endpoints share the common characteristic: surrogacy for OS has been demonstrated or is plausible in specific contexts, but broad regulatory acceptance requires additional validation or formal FDA guidance.

Status: MRD/MM FDA Draft January 2026; pCR Final 2014; ctDNA = No FDA guidance; EFS-12/24 = Research only (not regulatory primary); PFS2 = Supportive only; MFS = Accepted in nmCRPC only

Regulatory Position

FDA general framework (2018): Emerging endpoints are typically accepted for accelerated approval as surrogate endpoints "reasonably likely to predict clinical benefit." Regular approval requires either validated OS correlation or demonstration that the endpoint itself represents direct clinical benefit.

ICH E9(R1) relevance: The estimand framework requires explicit IE strategy specification for all endpoints, including emerging ones. The fact that an endpoint is "emerging" does not reduce the need for pre-specified IE handling.

FDA 2026 Draft on MRD/CR: First indication-specific emerging endpoint guidance. Proposes MRD-negative CR for accelerated approval in MM with post-marketing confirmatory OS/PFS trial requirement.

1. MRD-Negative CR Rate (Multiple Myeloma)

Regulatory Status

FDA draft guidance (January 2026, Draft): First indication-specific MRD regulatory framework. Proposes MRD-negative CR rate as primary endpoint for accelerated approval in multiple myeloma when effect size is substantial (response rate difference ≥10–15%).

ODAC basis: April 2024 ODAC meeting — unanimous agreement (14/14 votes) that MRD is acceptable to support accelerated approval in MM, contingent on:

Assay standardization (10⁻⁵ sensitivity or better)
Central assessment
Confirmatory trial verifying PFS or OS

Prior regulatory milestone: FDA accepted MRD-negative CR as co-primary with PFS in GMMG-HD7 (isatuximab + RVd vs. RVd alone) — MRD negativity after induction as primary measurement.

Definition

MRD-negative CR = achievement of BOTH:

CR per IMWG 2016 criteria:
- Serum M-protein: negative on immunofixation AND
- Urine M-protein: <200 mg/24 hours AND
- Bone marrow plasma cells: <5%
MRD negativity in bone marrow at minimum sensitivity ≥ 10⁻⁵ (1 myeloma cell in 100,000 normal bone marrow cells)

Assay requirements (FDA Jan 2026 draft):

Validated next-generation flow (NGF): EuroFlow 8-color protocol; ≥5 × 10⁶ total cells assessed to reach 10⁻⁵ sensitivity
Validated next-generation sequencing (NGS): clonoSEQ, LymphoSeq, or equivalent; ≥200 million reads to achieve 10⁻⁵
Higher bar than 10⁻⁴ used in some published trials (GRIFFIN, MM3 studies)
Bone marrow aspirate bilateral sampling recommended (increases detection if disease heterogeneous)
Central assessment required for all primary endpoint claims

Distinction from MRD-positive CR:

CR + MRD positivity (detectable at 10⁻⁴ or 10⁻⁵) counts as MRD-positive CR, not MRD-negative. This distinction affects approval claims: MRD-negative CR → accelerated approval; MRD-positive CR → not sufficient for emerging endpoint claim.

Assessment Timing

At CR confirmation: First documentation of CR per above criteria (serology + BM assessment)
Post-consolidation: Pre-specified timepoints (12, 18, 24 months from randomization)
Sustained MRD negativity: Two consecutive negative assessments ≥12 months apart shows higher surrogacy for long-term OS than single-timepoint negativity (stratify by sustained vs. transient in SAP)

Landmark analysis: MRD-negative CR rate at 12 months post-randomization is a common reporting timepoint (though time-to-response preferred by FDA).

Estimand for MRD-Negative CR Rate

Primary analysis (treatment policy strategy):

Patients without post-baseline bone marrow assessment → count as non-responders (missing data = non-response imputation)
Patients with CR documented but no subsequent MRD assessment → count as MRD-positive CR (not MRD-negative)
Death before assessment → composite event (death = non-response)

IE 1: Treatment discontinuation before scheduled bone marrow biopsy

Primary strategy: Treatment policy — patient counted as MRD-positive if no assessment obtained regardless of reason
SAP language: "Patients who discontinue study treatment prior to obtaining a post-treatment bone marrow assessment will be counted as MRD-positive/non-responders in the primary analysis."
Sensitivity: Hypothetical — what rate if all patients completed protocol therapy? (Intent-to-treat subset with assessment)

IE 2: Death before assessment

Death before obtaining a bone marrow MRD assessment is a physical barrier to measuring the endpoint. Why is this "composite"?

Composite strategy definition: The intercurrent event is incorporated into the endpoint definition itself.
Strategy: Composite — The endpoint is redefined as: "Achieve MRD-negative CR OR experience death before assessment (unfavorable component)." Death is incorporated as part of the composite outcome, not ignored or censored.
Why composite (not treatment policy): Treatment policy would say "ignore death and continue follow-up" — but a deceased patient cannot undergo bone marrow assessment. We cannot pretend the patient is still alive to measure MRD. Therefore, we must incorporate death into the endpoint definition as an unfavorable outcome.
Why composite (not censoring): Censoring would exclude the death and treat it as "no event observed" — this biases results because treatment arms may differ in mortality, causing informative censoring.
SAP language: "Patients who die before any post-baseline bone marrow assessment will be counted as non-responders. The endpoint is composite: MRD-negative CR (favorable response) versus death before assessment or failure to achieve MRD-negative CR (unfavorable response). Death prior to assessment is incorporated as an unfavorable component of the composite outcome."

IE 3: Patient refuses bone marrow biopsy post-CR

A patient achieves CR clinically but refuses the MRD biopsy to confirm MRD-negative status. How is this patient classified?

Primary analysis (Composite strategy with non-responder imputation): Patients who achieve CR but do not undergo bone marrow MRD assessment (regardless of reason: patient preference, medical contraindication, logistical issues) will be classified as MRD-positive CR, not MRD-negative CR. This is the composite strategy — we're treating missing MRD assessment as equivalent to "MRD-positive" (the negative outcome). This rule is applied uniformly to all randomized patients analyzed as intent-to-treat (treatment policy principle: uniform application regardless of reason for non-compliance).
Sensitivity analysis (Hypothetical estimand): Among the subset of randomized patients with documented MRD assessment (i.e., those who complied with protocol), report MRD-negative CR rate separately. This hypothetical analysis estimates what the MRD-negative CR rate would be if all patients had complied with protocol assessments.

SAP language: "Patients who achieve CR but do not obtain a bone marrow MRD assessment will be counted as MRD-positive CR in the primary intent-to-treat analysis (composite strategy, non-responder imputation). Sensitivity analysis: among patients with documented MRD assessment, MRD-negative CR rate is reported as a per-protocol subset estimate (hypothetical estimand). The primary claim for accelerated approval is based on the intent-to-treat MRD-negative CR rate; the per-protocol rate is supportive."

Why both strategies matter: The primary claim (accelerated approval) uses composite/ITT (conservative, accounts for non-compliance, applicable to real-world use). The sensitivity (per-protocol/hypothetical) shows what the true treatment effect would be if compliance were perfect — useful for understanding mechanism but not for regulatory approval claims.

Confirmatory Trial Requirement

Per FDA Jan 2026 draft: Products approved via MRD-negative CR surrogate endpoint must verify clinical benefit with one of the following in post-marketing confirmatory trials:

PFS benefit (HR < 0.80 for recurrence/progression)
OS benefit (HR < 0.85 for death)
Timeframe: typically 5-year OS follow-up (median 3–5 years post-approval for accelerated approval)

Risk: If confirmatory trial fails (no PFS/OS difference), FDA may withdraw accelerated approval.

Real Trial Example: GMMG-HD7 (Isatuximab + RVd)

Feature	Detail
Population	Newly diagnosed MM, fit, age 18–75
Design	2:1 randomization; isatuximab + RVd vs. RVd
Primary endpoint	MRD negativity (10⁻⁵) after induction
Secondary	PFS, ORR, CR rate, OS (follow-up)
MRD assay	Central NGF assessment (EuroFlow)
Status	Positive for MRD: isatuximab arm 50% vs. control 35% MRD-negative CR at 12 months post-induction

SAP Language Template

"The primary endpoint is MRD-negative CR rate, defined as the proportion of randomized patients achieving both CR per IMWG 2016 criteria and MRD negativity (defined as <10⁻⁵ residual myeloma cells by validated next-generation flow [EuroFlow] or next-generation sequencing [clonoSEQ] in bone marrow aspirate) at the first post-treatment assessment or at any time during the treatment period.

Primary analysis: Patients without a documented post-treatment bone marrow MRD assessment will be classified as MRD-positive/non-responders (treatment policy strategy with non-responder imputation). Patients who achieve CR but have MRD positivity will be counted as MRD-positive CR (not in the numerator for MRD-negative CR rate).

Sensitivity analysis: Among patients with documented MRD assessments (per-protocol subset), report MRD-negative CR rate separately.

Sustained MRD negativity (secondary): Report proportion with sustained MRD negativity (two assessments ≥12 months apart, both negative) as a supporting measure."

2. pCR (Pathological Complete Response)

Regulatory Status

FDA guidance (2014, Final): pCR supports accelerated approval in neoadjuvant high-risk breast cancer as a surrogate endpoint "reasonably likely to predict EFS/iDFS benefit." Confirmatory data (EFS/iDFS) required for regular/traditional approval within 5 years post-accelerated approval.

Indication-specific surrogacy:

Validated (strong OS correlation): TNBC, HER2+ breast cancer
Not validated: HR+/HER2- (luminal) breast cancer — pCR does NOT predict OS
Emerging: Neoadjuvant NSCLC (EFS confirmatory, not yet final FDA guidance); gastric, bladder cancers (exploratory only)

Definition

ypT0ypN0 (FDA-preferred standard for breast):

No residual invasive cancer in breast tissue
No residual invasive cancer in axillary lymph nodes
DCIS (ductal carcinoma in situ) allowed
Non-invasive tumor burden OK

ypT0/Tis ypN0 (alternative, slightly less stringent):

ypT0 = no invasive cancer in breast, DCIS allowed
Tis = carcinoma in situ only (includes DCIS, lobular carcinoma in situ)
ypN0 = nodes clear
Used in some trials (KEYNOTE-671 NSCLC)

pCR + EFS co-primary (KEYNOTE-522 design):

pCR (ypT0ypN0) at surgery → primary endpoint supporting accelerated approval
EFS (event-free survival, time-to-event) → co-primary for regular approval
Trial declared positive when both endpoints significant (or each independently at pre-specified alpha allocation)
This design allows accelerated approval on pCR while confirmatory EFS data collected during post-approval follow-up

Estimand for pCR

Population: Intent-to-treat (all randomized patients, including those who do not complete neoadjuvant therapy)

Primary analysis:

Denominator: All randomized patients
Numerator: Patients with ypT0ypN0 documented on surgical specimen (must have surgery and pathology assessment)
Patients without surgery: count as non-pCR regardless of reason

IE 1: Treatment discontinuation before surgery

Strategy: Composite — no surgery = pCR failure; count as non-pCR
SAP language: "Patients who do not undergo surgical resection, regardless of reason, will be classified as non-pCR."

IE 2: Death before surgery

Strategy: Composite — death = non-pCR

IE 3: Residual disease not fully excised (R1/R2 resection)

Strategy: Composite — count based on final specimen; if residual invasive cancer remains, not pCR regardless of margins
Sensitivity: R0 (complete resection) subset analysis

IE 4: Missing pathology report or insufficient specimen assessment

Strategy: Composite — missing pathology = non-pCR

Limitations and Surrogacy Concerns

pCR predicts EFS in TNBC and HER2+; NOT OS surrogate in HR+/HER2- breast cancer
HR+/HER2- pCR rates: 10–20% with chemotherapy; limited clinical benefit at population level
Luminal breast cancers: pCR is not a validated surrogate — FDA does not accept pCR for approval in HR+ disease
pCR at surgery is a single-timepoint measure — residual micrometastatic disease in blood/organs not captured
Tumor heterogeneity: pCR in primary does not exclude residual disease in distant nodes or occult metastases
ypT0ypN0 vs. ypT0/Tis: More stringent ypT0ypN0 correlates better with OS than ypT0/Tis (includes DCIS-only)

Regulatory Precedent: Neoadjuvant Breast Trials

Trial	Regimen	Population	pCR Rate (Exp vs. Control)	Accelerated Approval	EFS/iDFS Result
KEYNOTE-522	Pembrolizumab + chemo	TNBC	55% vs. 33%	Yes (pCR)	Positive EFS (HR=0.60)
KEYNOTE-671	Pembrolizumab + chemo	NSCLC neoadjuvant	pCR secondary	No (still under review)	EFS confirmatory pending
NeoTRIPaPDL1 (European)	Atezolizumab + chemo	TNBC	58% vs. 44%	Not FDA (European data)	EFS pending

SAP Language Template

"The primary endpoint is pathological complete response (pCR), defined as ypT0ypN0 (no residual invasive cancer in breast or axillary nodes) on final surgical pathology specimen. Patients who do not complete neoadjuvant therapy and do not undergo surgical resection, or who die before surgery, or for whom final pathology is not available, will be classified as non-pCR. pCR rate will be reported as the proportion of randomized patients achieving ypT0ypN0, with exact binomial 95% confidence interval (Clopper-Pearson).

Subgroup analyses (pre-specified): pCR rates by hormone receptor status (TNBC vs. HR+/HER2-) and HER2 status (HER2+ vs. HER2-). Surrogacy of pCR for EFS is expected to differ by subtype."

3. ctDNA (Circulating Tumor DNA)

Regulatory Status

FDA position (as of April 2026): ctDNA remains an exploratory biomarker endpoint in all regulatory submissions. It cannot support approval claims (primary or secondary inferential) without:

Pre-specified analysis with clinical outcome correlation
Fully validated assay (analytical validation per FDA bioanalytical guidance)
Demonstrated correlation with validated endpoint (OS, PFS, DFS, or pCR)

Informal FDA communications (from review letters, ODAC discussions):

ctDNA as pharmacodynamic biomarker: Acceptable as secondary/exploratory (informative but not confirmatory)
ctDNA clearance as pCR surrogate: Plausible mechanism but not yet validated prospectively
ctDNA-MRD for adjuvant solid tumors: Active area of investigation (CIRCULATE-Japan, COBRA, DYNAMIC III trials); no regulatory acceptance yet but strong FDA interest

Expected timeline to acceptance: 2027–2030, pending outcomes of major adjuvant trials (CIRCULATE-Japan with OS readout 2027–2028).

Current ctDNA Applications in Phase 3 Trials

1. Surveillance/detection of molecular residual disease (MRD) in adjuvant settings:

Colorectal cancer adjuvant: ctDNA positivity post-surgery predicts recurrence risk (CIRCULATE-Japan, DYNAMIC III); exploratory but FDA-encouraged monitoring
NSCLC adjuvant: ctDNA clearance after resection as potential risk stratification factor; monitoring post-op (ADRIATIC, KEYNOTE-091 substudies)
Breast cancer adjuvant: ctDNA-guided de-intensification of adjuvant therapy under investigation (PHARE trial, DESTINY-Breast05 substudies); no approval guidance

2. ctDNA as predictive biomarker (acceptable use):

Pre-specify subgroup by ctDNA status at baseline (e.g., ctDNA-positive vs. negative) to assess differential treatment benefit
ctDNA as stratification factor for randomization: acceptable per FDA (ensures balance of MRD status across arms)
Post-hoc ctDNA subgroups without pre-specification: exploratory only, no formal claims

3. ctDNA as secondary endpoint (acceptable reporting):

ctDNA response rate: Reduction in variant allele frequency (VAF) after treatment — descriptive, no formal statistical test
ctDNA clearance at landmark timepoint: Proportion achieving undetectable ctDNA at end-of-treatment — descriptive
ctDNA kinetics: Time to ctDNA negativity — supportive measure of early response
ctDNA-PFS/OS correlation: Post-hoc correlation analysis — hypothesis-generating

Design Considerations for ctDNA

1. Assay platform selection and validation:

Tumor-informed assay (e.g., Signatera, Guardant LUNAR, Natera Signatera):
Personalized mutation panel based on tumor sequencing
High specificity; lower sensitivity early in disease
Requires tumor tissue availability before treatment start
Tumor-agnostic assay (e.g., Foundation Medicine, Guardant 360):
Fixed panel of common cancer mutations
Broader applicability; variable sensitivity by tumor type
No tumor sequencing required
Analytical validation requirement: Assay must be analytically validated per FDA guidance before regulatory submission (sensitivity/specificity at specified LOD)

2. Timepoint selection (critical for validity):

Must be pre-specified in protocol and SAP (Cycle 1 Day 1, end of Cycle 2, end of treatment, 6-month surveillance, etc.)
Ad hoc or data-driven timepoint selection introduces selection bias — unacceptable for regulatory claims
Multiple timepoints acceptable if hierarchically ordered (primary timepoint at end-of-treatment; secondary at 3, 6, 12 months)

3. Statistical analysis pre-specification (even for exploratory endpoints):

ctDNA response definition: ≥50% VAF reduction, ≥1000-copy reduction, conversion to undetectable (specify in SAP)
Handling of missing ctDNA data: patients lost to follow-up or with missing samples treated as missing (not imputed)
Correlation analysis (if planned): Spearman or Kendall correlation between ctDNA clearance and PFS/OS — specify method in SAP

4. Assay cost and logistics:

Tumor-informed: $3,000–5,000 per patient (tumor sequencing + multiple plasma tests)
Tumor-agnostic: $1,500–2,500 per patient (direct plasma sequencing)
Plasma collection, shipping, storage must maintain DNA integrity (protocols vary by vendor)
Budget and logistics must be specified in protocol to ensure compliance

Real Trial Example: CIRCULATE-Japan (Adjuvant CRC)

Feature	Detail
Population	Stage III CRC post-resection + adjuvant chemo
Design	3-arm: ctDNA-positive post-chemo → intensive surveillance (enhanced imaging) vs. standard; ctDNA-negative → standard surveillance
ctDNA assay	Tumor-informed (tumor sequencing + targeted plasma ddPCR)
Primary endpoint	RFS (recurrence-free survival)
ctDNA endpoints	Exploratory: ctDNA as predictor of RFS; ctDNA status post-chemo as risk stratifier
Status	Enrollment complete; RFS and OS readout anticipated 2027–2028

Estimated Timeline to Regulatory Acceptance

ctDNA Application	Evidence Status	Expected FDA Acceptance
Adjuvant CRC post-chemo risk stratification	Strong (CIRCULATE-Japan 2027)	2027–2028 if OS benefit confirmed
Adjuvant NSCLC MRD detection	Moderate (ongoing trials)	2028–2030
Neoadjuvant response (pCR surrogate)	Weak (requires prospective validation)	2029+
Metastatic cancer response monitoring	Exploratory only	2030+ or never (OS benefit unclear)
Standalone primary endpoint	Unvalidated	2030+ or never (requires paradigm shift)

SAP Language Template

"ctDNA is measured at baseline, end-of-treatment, and at 6-month surveillance intervals using [assay platform and method]. ctDNA analysis is exploratory and intended to assess correlation with clinical outcomes (PFS, RFS, OS) and to evaluate ctDNA as a potential predictor of recurrence risk. ctDNA endpoints are not subject to formal Type I error control; results are descriptive and hypothesis-generating only. ctDNA will not be used to modify treatment decisions or to support efficacy claims unless pre-specified in a formal statistical analysis plan amendment."

4. EFS-12 / EFS-24 (Lymphoma Landmark Surrogates)

Regulatory Status

Not formally validated by FDA as a primary regulatory endpoint. EFS is used as a time-to-event endpoint in Phase 3 lymphoma trials (ZUMA-7, TRANSFORM, etc.), but EFS-12 and EFS-24 as landmark rate surrogates for OS are research measures only.

Validation basis (Patterson et al., 2019, Journal of Clinical Oncology):

Analyzed 6,000+ DLBCL patients treated with R-CHOP (rituximab + CHOP chemotherapy)
EFS-24 (proportion event-free at 24 months) strongly correlates with long-term OS
Patients achieving EFS-24 have OS approaching age-matched general population
Caveat: Validation was for R-CHOP era; surrogacy may not hold for CAR-T, BTK inhibitors, bispecific antibodies

Current use: EFS-12/24 serve as secondary exploratory endpoints in some lymphoma trials to assess durability of response in early relapses.

Design Considerations

EFS vs. EFS landmark: Regulatory primary = EFS as time-to-event (HR-based); EFS-12/24 as supplementary landmark rates
Multiple timepoints acceptable: EFS-3, EFS-6, EFS-12, EFS-24 as supportive measures (no multiplicity adjustment needed if labeled exploratory)
Censoring at landmark: Patients not experiencing event by 12 or 24 months censored; use Kaplan-Meier to estimate EFS at landmark

Example: ZUMA-7 Design

Feature	Detail
Population	Relapsed/refractory DLBCL
Design	Axicabtagene (CAR-T) vs. standard salvage therapy
Primary endpoint	EFS (time-to-event)
Secondary	EFS-12, EFS-24 (landmark rates); ORR; OS
Results	EFS HR=0.27 (p<0.001); EFS-12: ~60% vs. 35% control

5. PFS2 (Time to Second Progression)

Regulatory Status

No FDA formal guidance. PFS2 is an emerging secondary/supportive endpoint capturing time from randomization to second progression (under second-line therapy) or death. Designed to assess whether a first-line regimen's benefit extends through the subsequent treatment sequence.

Estimand complexity: PFS2 involves multiple IEs:

Second-line treatment choice (patient/physician discretion)
Definition of second progression (may differ from first progression criteria)
Death from any cause
Treatment discontinuation for toxicity between lines

No standard SAP template established. Design decisions case-by-case.

Design Considerations for PFS2

1. Defining "second progression":

Same criteria as first PFS? (e.g., RECIST 1.1 if solid tumor)
Or more lenient criteria? (e.g., symptomatic progression if imaging inconclusive)
Must be pre-specified in SAP; discordance in criteria introduces bias

2. Second-line treatment assumptions:

Ideal: protocol mandates second-line regimen (allows comparison)
Real-world: second-line left to physician choice (unmeasured confounding)
Sensitivity: analyze by actual second-line treatment received (post-hoc, descriptive)

3. Death handling in PFS2:

Composite: second progression or death from any cause
Censoring: patients not progressed on second-line therapy censored at last assessment
Deaths after second progression not included (event already occurred)

Where PFS2 is Used

Myeloma Phase 3 trials: Daratumumab combinations, CAR-T designs (CARTIFAN, MAIA)
Lymphoma trials: Emerging for CAR-T post-approval trials
NSCLC targeted therapy: PFS2 as exploratory (limited use; OS more important)

FDA Position (Informal)

PFS2 is descriptive and supportive — it does not substitute for OS data. PFS2 showing no attenuation of benefit (i.e., experimental arm remains better at second progression) supports the durability of the primary PFS claim but does not establish clinical benefit on its own. FDA would not approve a drug on PFS2 alone without OS data.

SAP Language Template

"PFS2 is defined as time from randomization to progression on second-line therapy or death from any cause, whichever occurs first. Patients receiving different second-line therapies are analyzed together (ITT); a post-hoc subgroup analysis by second-line regimen will be provided as supporting information. PFS2 is a supportive endpoint intended to assess the durability of the first-line treatment benefit through the treatment sequence."

6. MFS (Metastasis-Free Survival) — Established but Narrowly Applicable

Regulatory Status

Accepted by FDA specifically and exclusively for non-metastatic castration-resistant prostate cancer (nmCRPC). MFS has NOT been accepted for other tumor types.

Key approvals:

Apalutamide (SPARTAN): MFS HR 0.28 (95% CI 0.21–0.37); supported regular approval
Enzalutamide (PROSPER): MFS HR 0.29 (95% CI 0.22–0.39); supported regular approval

Why MFS is Specific to nmCRPC

Well-defined M0 population: Non-metastatic defined by conventional imaging (bone scan, CT); excludes patients with visceral/nodal disease
Measurable event rate: Median time to metastasis ~16 months in control arm (SPARTAN, PROSPER) — sufficient for timely trial readout
High OS correlation in nmCRPC: Historical data show strong MFS-OS correlation specific to this population
Clinical relevance: Metastasis is the major prognostic milestone in nmCRPC

Why MFS is Not Applicable to Other Cancers

Breast cancer: Distant metastasis inevitable over time; endpoint not sensitive enough; DFS/iDFS preferred
NSCLC: Metastatic progression indistinguishable from local recurrence + distant; not a distinct milestone
Colorectal: Peritoneal/liver metastasis not reliably detected; endpoint underestimates true disease progression
Pancreatic: Metastasis universal at diagnosis/early recurrence; endpoint not feasible

Cross-Endpoint Comparison Table

Endpoint	Regulatory Status	Primary Setting	Assay/Assessment	OS Surrogacy	Evidence Level
MRD-negative CR	Draft Jan 2026 — accelerated approval MM	Multiple myeloma	NGF/NGS at 10⁻⁵	Plausible; confirmation required	Phase 3 (GMMG-HD7)
pCR	Final 2014 — accelerated approval	Neoadjuvant breast (TNBC, HER2+)	Central pathology (ypT0ypN0)	Strong (TNBC, HER2+); None (HR+)	Phase 3 (KEYNOTE-522)
ctDNA	Exploratory only	All solid tumors, adjuvant	Platform-dependent (tumor-informed or agnostic)	Unvalidated	Ongoing Phase 3 (CIRCULATE-Japan, DYNAMIC III)
EFS-12/24	Research only (not regulatory primary)	DLBCL	PET-CT per Lugano	Moderate (R-CHOP era only)	Retrospective validation (Patterson 2019)
PFS2	Supportive/descriptive	Myeloma, lymphoma	IMWG/Lugano criteria	Not validated	Phase 3 secondary analyses
MFS	Regular approval (nmCRPC only)	Non-metastatic CRPC	Imaging (bone scan + CT)	Strong (nmCRPC only)	Phase 3 (SPARTAN, PROSPER)

Intercurrent Events in Emerging Endpoint Trials

MRD-Negative CR: IE Handling

IE: Death before bone marrow assessment for MRD

Strategy: Composite (death = non-response)
SAP: "Deaths before any post-treatment assessment will be included in the denominator and counted as non-responders."

IE: Patient refuses post-CR bone marrow biopsy

Strategy: Composite (non-response imputation)
SAP: "Patients who decline post-CR assessment will be counted as MRD-positive (treatment policy approach)."

Sensitivity analysis: Per-protocol subset (patients with documented assessments) — may show higher MRD-negative CR rate but less representative of intent-to-treat

pCR: IE Handling

IE: Treatment discontinuation before surgery

Strategy: Composite (no surgery = non-pCR)
SAP: "Patients who do not undergo surgery will be classified as non-pCR regardless of pathological assessment."

Sensitivity analysis: Among-surgery subset (patients who underwent resection) — shows pCR rate in surgical population only

ctDNA: IE Handling

IE: Lost to follow-up before ctDNA assessment

Strategy: Missing data (not imputed)
SAP: "Patients without ctDNA assessment will be excluded from ctDNA-specific analyses. Intent-to-treat population includes all randomized patients; ctDNA-evaluable subset is pre-specified for analysis."

Limitations and Pitfalls

1. MRD-negative CR with immature OS follow-up:

Accelerated approval granted on MRD without long-term OS data introduces risk of false surrogate validation. If confirmatory trial fails (no OS difference), accelerated approval may be withdrawn.

Mitigation: Plan 5-year OS follow-up; commit to post-approval trial in SAP

2. pCR surrogacy is indication-dependent:

pCR is not validated in HR+/HER2- breast cancer. Using pCR as approval basis in luminal disease would be inappropriate.

Mitigation: Pre-specify surrogacy assumption by receptor subtype; if trial enrolls mixed subtypes, report pCR separately by subtype

3. ctDNA assay variability across platforms:

Different ctDNA assays (tumor-informed vs. tumor-agnostic, ddPCR vs. NGS) have different LOD, specificity. Results not comparable across assays.

Mitigation: Pre-specify assay platform; analytic validation before trial initiation; sensitivity analysis by assay type if multiple platforms used

4. EFS-12/24 surrogacy decay in novel therapies:

Validation of EFS-24 was for R-CHOP (conventional chemotherapy). CAR-T, BTK inhibitors, bispecific antibodies have different disease kinetics — EFS-24 surrogacy may not hold.

Mitigation: For novel therapy trials, treat EFS-12/24 as exploratory; do NOT assume OS surrogacy without prospective validation

5. PFS2 confounding by second-line treatment:

If second-line therapy differs between arms (different access, patient choice), PFS2 reflects both first-line benefit and second-line differences — uninterpretable as first-line efficacy measure.

Mitigation: If second-line uncontrolled, treat PFS2 as supportive/exploratory only; sensitivity analysis by actual second-line regimen

6. MFS only applicable to nmCRPC:

Inappropriate use of MFS as approval endpoint in other cancers (attempted in early breast cancer, early NSCLC) has been rejected by FDA.

Mitigation: Only consider MFS for nmCRPC; use standard endpoints (DFS, EFS) for other indications

Backlinks

Source: FDA Draft Guidance on MRD and CR in Multiple Myeloma (January 2026, Draft); FDA Pathological Complete Response in Neoadjuvant Breast Cancer 2014 (Final); FDA Cancer Endpoints 2018 (Final); ODAC meeting minutes April 2024; Patterson et al. (2019) on EFS surrogacy; CIRCULATE-Japan protocol Status: MRD/MM January 2026 = Draft (comment period closed March 2026); pCR 2014 = Final; ctDNA = No FDA guidance; EFS-12/24 = Research measures only; PFS2 = Supportive/descriptive; MFS = Accepted in nmCRPC only Compiled from FDA draft chunks + CTG myeloma trials + published clinical trial literature