Response-Based Endpoints (ORR, CR, DOR)

Definition

Objective Response Rate (ORR):

"ORR is defined as the proportion of patients with tumor size reduction of a predefined amount and for a minimum time period. Response duration usually is measured from the time of initial response until documented tumor progression. Generally, the FDA has defined ORR as the sum of partial responses plus CRs." — FDA Cancer Endpoints 2018, §III.B.2 (Final guidance)

Stable disease must NOT be included in ORR: "Stable disease can reflect the natural history of disease, whereas tumor reduction is a direct therapeutic effect." (FDA 2018)

Complete Response (CR):

"CR is defined as no detectable evidence of tumor. CR is generally measured through imaging studies (e.g., CT scans) or through histopathologic assessment (e.g., bone marrow biopsy or breast cancer resection specimens)." — FDA Cancer Endpoints 2018, §III.B.3 (Final guidance)

Duration of Response (DOR):

Time from first documented objective response (CR or PR) to the first documented disease progression or death. DOR is a conditional endpoint — reported in responders only, not the ITT population. DOR quantifies the durability of the ORR signal.

Partial Response (PR):

Per RECIST 1.1: ≥30% decrease in sum of longest diameters of target lesions (from baseline). PR is counted in ORR but not in CR rate.

Pathologic Complete Response (pCR):

Absence of residual invasive cancer in the breast and axillary lymph nodes at surgery following neoadjuvant therapy. Used exclusively in neoadjuvant breast cancer (see DFS and EFS Endpoints and Breast Cancer Trial Design Patterns: Indication-Specific Statistical Framework).

MRD-negative CR (myeloma):

CR per IMWG criteria (including sCR) AND MRD negativity ≥10⁻⁵ by NGF or NGS in bone marrow. Proposed as primary endpoint for accelerated approval in multiple myeloma per FDA January 2026 draft. See Emerging Endpoints in Oncology Trials.

Regulatory Position

ORR supports:

Accelerated approval (most common use): ORR is the single most widely used surrogate for accelerated approval when ORR is "reasonably likely to predict" OS or clinical benefit
Traditional (regular) approval: When ORR is "highly durable… substantiated by both response rate magnitude and duration" and particularly when high CR rate accompanies the ORR (FDA 2018)
Single-arm trials: ORR is uniquely evaluable in single-arm studies — the only major efficacy endpoint that does not require a control arm

CR rate supports:

Traditional approval in hematologic malignancies: "For hematologic malignancies such as acute leukemia, CR has been used as a clinical endpoint for traditional approval." (FDA 2018)
Accelerated approval as a surrogate in settings where CR rate predicts PFS/OS (lymphoma, myeloma, AML)

pCR supports:

Accelerated approval in high-risk early breast cancer (neoadjuvant setting): "For early high-risk breast cancer, pathologic CR (pCR) has been used as a surrogate endpoint for accelerated approval." (FDA 2018)
Confirmatory trial showing EFS/OS benefit required post-approval

DOR does not independently support approval but is required as an accompanying measure when ORR is the primary endpoint. "The significance of ORR is assessed by its magnitude and duration, and the percentage of CRs." (FDA 2018)

Status: FDA Cancer Endpoints 2018 = Final; FDA MRD/MM January 2026 = Draft

When to Use

ORR as primary endpoint — accelerated approval:

Any solid tumor or hematologic malignancy with high unmet need, single-arm Phase 2/3 design
AML (ORR: 48/489 trials, 10%), lymphoma (ORR: 27/300, 9%), NSCLC (26/300, 9%)
≥2L settings where randomized OS/PFS trial is infeasible or unethical
Rare tumors with limited patient population

ORR as primary endpoint — traditional approval:

When ORR is very high (>50–60%), durable (DOR >12 months), and includes high CR fraction
Examples: BRAF-mutated melanoma (vemurafumab, dabrafenib), ROS1+ NSCLC

CR rate as primary — hematologic malignancies:

AML induction therapy: CR/CRi rate is the standard primary endpoint for traditional approval
Lymphoma: CR rate at end of treatment (Lugano criteria) for accelerated approval
Myeloma: CR rate or MRD-negative CR for accelerated approval (2026 draft)

ORR as secondary (PFS/OS primary):

All Phase 3 randomized trials should include ORR as secondary
Provides supportive evidence of anti-tumor activity
ORR HR not computed — compared as proportion (CMH test, stratified by randomization factors)

Design Considerations

Response criteria — must be pre-specified

"The response criteria should be predefined in the protocol before the start of the study." (FDA 2018)

RECIST 1.1 (solid tumors): target lesion sum ≥30% decrease (PR); no detectable tumor (CR); ≥20% increase or new lesion (PD). Stable disease (SD): neither PR nor PD
Lugano 2014 (lymphoma): PET-CT based for FDG-avid histologies; CT-based for non-FDG. CR = Deauville score 1–3; PR = Deauville 4–5 with reduction
IMWG criteria (myeloma): CR = negative serum/urine electrophoresis + <5% plasma cells in marrow; sCR = CR + normal FLC ratio + no clonal plasma cells; VGPR, PR defined separately
IWG 2018 (MDS/AML): CR, CRi (complete remission with incomplete hematologic recovery), marrow CR, PR
WHO Histology (lymphoma): Histologically confirmed diagnosis required; pathology must support study eligibility

Assessment timing and response confirmation

Critical FDA requirement: Response must be confirmed by a second imaging assessment within 4–8 weeks. "Responses should be confirmed at least 4 weeks after initial response determination." (FDA 2018)

Initial response assessment: Typically 6–8 weeks after first dose (2 cycles chemotherapy) or per protocol schedule
Confirmatory response assessment: Required for CR and PR — typically 4 weeks after initial response (mandatory for single-arm studies)
DOR start date: First response assessment showing PR or CR; end date: first progression or death
Minimum response duration pre-specified in protocol: Typically 4 or 6 weeks for confirmatory scan

Unconfirmed ORR: If confirmatory scan is not available/missing, that patient cannot be counted in the primary ORR analysis. FDA strongly discourages unconfirmed response:

"To ensure that responses are not due to measurement error alone, responses should be confirmed at least 4 weeks after initial response determination." (FDA 2018)

IRC requirements and blinding

ORR as primary endpoint in single-arm trials: IRC is strongly recommended; provides blinded independent verification
ORR in randomized trials: IRC may be required when ORR is primary; optional when key secondary
IRC must adjudicate: All CR and PR determinations; discordant reads resolved by majority or third reader
Blinding standards: IRC must remain blinded to treatment arm, baseline characteristics, investigator assessments, and all clinical outcomes except imaging
Response criteria training: All IRC members must complete formal training and pass competency assessment on protocol response criteria before first assessment

Statistical analysis for ORR

Single-arm trials:

Primary analysis: Proportion responding with exact binomial 95% confidence interval (Clopper-Pearson)
Hypothesis test: One-proportion Z-test against historical control rate (e.g., H0: ORR ≤ 10%) Z = (p_observed - p_null) / sqrt[p_null(1-p_null)/n] Under H0: Z ~ N(0,1), reject if Z > 1.645 (one-sided α=0.05)
Sample size: Typically 30–100 patients to achieve 90% CI excluding the null ORR n = (z_α + z_β)² × p(1-p) / (p - p_0)² Example: p_0 = 0.10 (null), p = 0.30 (target), α=0.05, β=0.20 n ≈ 36 patients for 80% power

Randomized trials:

ORR comparison: Cochran-Mantel-Haenszel test stratified by randomization factors
Effect measure: Odds ratio (OR) with 95% CI; reported separately for each arm
Secondary: Risk difference (RD) with 95% CI
Statistical testing: Two-sided α=0.05 (unless hierarchy specifies otherwise)

Duration of Response (DOR) analysis

Population: Only responders (CR + PR); not ITT
Method: Kaplan-Meier estimate of DOR among responders
Reporting: Median DOR with 95% CI; percentage with DOR >6, >12, >18 months
Right censoring: Patients not progressed at data cutoff censored at last adequate assessment
Sample size: DOR requires ≥20–30 responders minimum for stable KM estimates; FDA cautions against DOR estimates based on <10 responders

IRC vs. Investigator concordance analysis

Required pre-specification:

Primary analysis: ORR by IRC (blinded central review)
Sensitivity analysis: ORR by investigator assessment; compare magnitude and statistical significance
Discordance table: Proportion of cases where IRC and investigator agree/disagree on response
Breakdowns by: ≤30 days delay, 31–90 days delay, >90 days delay
Breakdowns by: IRC-CR vs. Investigator-PR (or vice versa)

When discordance is substantial (>10% of responses):

FDA views IRC-driven ORR as the more reliable estimate
Investigator ORR is supporting sensitivity analysis
Large discordance raises questions about assessment criteria clarity or investigator training

Alpha allocation when ORR is co-primary with PFS/OS

When ORR and PFS are co-primary endpoints:

Gatekeeping not typical — both tested at α=0.025 one-sided (or α=0.05 two-sided)
OR: ORR at α=0.025, then PFS at α=0.025 sequentially if ORR is significant
See Multiple Endpoints and Alpha Allocation for formal strategies

Single-arm vs. randomized design considerations

Single-arm advantages:

Smaller sample size (30–100 vs. 200–500 in randomized design)
Shorter trial duration
Ethical when no effective control exists

Single-arm requirements:

Historical control ORR must be well-established, prospectively defined, and documented in protocol
Patient population must match historical controls (same line of therapy, molecular subgroup, performance status)
Response criteria must be identical to historical data
IRC strongly recommended
Confirmatory randomized trial required for traditional approval

Intercurrent Events

1. Subsequent anti-cancer therapy before response assessment

Rarely affects ORR (response assessed before new therapy in most cases). Relevant when response assessment window extends beyond treatment discontinuation.

ICH E9(R1) strategy: Treatment Policy — assess per schedule regardless of treatment status

SAP language: "Response assessments will be conducted per protocol schedule regardless of treatment discontinuation. Responses occurring after initiation of non-protocol anti-cancer therapy will not be counted in the primary ORR analysis."

2. Treatment discontinuation before response (non-progression)

Patients who stop treatment for toxicity before achieving measurable response.

ICH E9(R1) strategy: Composite (count as non-responders in ITT denominator)

SAP language: "All randomized patients will be included in the ORR denominator (Intent-to-Treat population). Patients who discontinue treatment before any response assessment will be considered non-responders."

3. Death before first response assessment

ICH E9(R1) strategy: Composite (count as non-responder)

SAP language: "Patients who die before the first post-baseline tumor assessment will be classified as non-responders in the primary ORR analysis."

4. Response assessment unavailable due to loss to follow-up

ICH E9(R1) strategy: Composite (count as non-responder) or missing data strategy (sensitivity analysis)

SAP language: "Patients lost to follow-up before response assessment will be classified as non-responders. Sensitivity analyses excluding lost-to-follow-up patients will be reported."

Regulatory Precedent

NCT#	Trial	Drug	Indication	Endpoint	Key Data
NCT01649856	Rituximab SC vs IV + CHOP	Rituximab	Untreated DLBCL	CR/CRu Rate	Primary endpoint for efficacy
NCT05171647	Mosunetuzumab + Polatuzumab vs. R-GemOx	Mosunetuzumab	R/R LBCL	ORR by IRF (Lugano 2014)	Accelerated approval pathway
NCT03976102	DRL_RI vs. MabThera in FL	Rituximab biosimilar	Low tumor burden FL	Best Overall Response Rate	Efficacy in follicular lymphoma
NCT01324947	Pomalidomide monotherapy	Pomalidomide	R/R MM	Objective response (IMWG)	Demonstrated activity in refractory myeloma
NCT00378222	Autologous transplant	Stem cell transplant	Multiple myeloma	CR/near-CR rate	Gold-standard consolidation approach

Note: Lymphoma CTG dataset = 300 Phase 3 trials (9% ORR primary); Myeloma CTG dataset = 256 Phase 3 trials (response rates vary 15–50% depending on drug class and line of therapy).

Limitations and Pitfalls

1. ORR does not equal clinical benefit:

A high ORR without durability does not establish clinical benefit. "The significance of ORR is assessed by its magnitude and duration, and the percentage of CRs." (FDA 2018). Short-lived responses (DOR < 3 months) are insufficient.

Mitigation: Pre-specify minimum DOR target (typically 6–12 months); FDA may require additional OS/PFS evidence if DOR is short

2. Stable disease exclusion is firm:

Including SD in the response denominator is a common protocol error. FDA will reject definitions that count SD as part of ORR.

Mitigation: Clear protocol language: "Stable disease is NOT included in ORR." SAP must define SD explicitly per response criteria

3. Single-arm ORR requires rigorous historical control:

If historical control ORR is not well-characterized, the hypothesis test against it is unreliable. FDA may reject single-arm data if the historical control population is not comparable.

Mitigation: Document historical control source in protocol (publication, trial database, registry); justify population comparability; pre-specify statistical method and null hypothesis

4. Confirmatory response assessment often missing: DOR requires a confirmed response (repeat scan showing PR or CR within 4–8 weeks). Protocols that do not mandate a confirmatory scan have unconfirmed ORR — FDA will require confirmation for single-arm accelerated approval.

Mitigation: Protocol must mandate confirmatory imaging; missing confirmatory scans render patient ineligible for ORR numerator; document timing and reasons for any missing confirmations

5. pCR as surrogate for OS in breast cancer is indication-specific: pCR predicts EFS in high-risk subtypes (TNBC, HER2+) but not in HR+/HER2- disease. Confirmatory EFS/OS trial required regardless.

Mitigation: Separate analysis by receptor subtype; pre-specify pCR-to-EFS correlation for indication; establish surrogate validation plan with FDA in type-C meeting

6. CR vs. CRi in AML: FDA distinguishes CR (complete hematologic recovery) from CRi (incomplete platelet/neutrophil recovery). For traditional approval in AML, the proportion achieving CR proper (not just CRi) matters significantly.

Mitigation: Report CR and CRi separately; present CR as primary efficacy endpoint; justify if CRi alone is used as primary

7. MRD-negative CR in myeloma (2026 draft): Requires NGF or NGS with ≥10⁻⁵ sensitivity. Earlier assays (10⁻⁴) are insufficient. Assay must be validated before trial initiation; FDA recommends EOP1 meeting for assay alignment.

Mitigation: Lock assay and methods before first patient enrolled; submit validation data in EOP1; include assay-qualified population in SAP; pre-specify assay failure rules

8. Unblinding through IRC discordance: If IRC assessments are heavily discordant from investigators (>15–20%), this may signal unblinding or investigator bias. FDA scrutinizes large discordances.

Mitigation: DSMC oversight of IRC-investigator concordance; implement investigator training and re-certification; pre-specify re-review criteria if initial discordance is high

Backlinks

Source: FDA Cancer Endpoints 2018 (Final); FDA MRD/CR in Multiple Myeloma Draft Guidance (January 2026, Draft) Status: Cancer Endpoints 2018 = Final; MRD/MM January 2026 = Draft Compiled from retrieved FDA chunks + ClinicalTrials.gov records (300 lymphoma Phase 3 trials, 256 myeloma Phase 3 trials, 489 AML Phase 2/3 trials)