Skip to content

DFS and EFS Endpoints

Definition

Disease-Free Survival (DFS):

"Generally, DFS is defined as the time from randomization until disease recurrence or death from any cause. The most frequent use of this endpoint is in the adjuvant setting after definitive surgery or radiotherapy." — FDA Cancer Endpoints 2018, §III.B.1 (Final guidance)

DFS is measured from randomization (post-surgery in adjuvant trials) until the earliest of: local/regional recurrence, distant metastasis, contralateral cancer, or death from any cause.

Event-Free Survival (EFS):

"An endpoint that is similar to DFS but is differentiated from it in that randomization takes place before definitive surgery or radiotherapy in the adjuvant setting is EFS. For the purpose of this guidance, EFS is defined as time from randomization to any of the following events: progression of disease that precludes surgery, local or distant recurrence, or death due to any cause." — FDA Cancer Endpoints 2018, §III.B.1 (Final guidance)

Key distinction: DFS is measured from a post-surgery/radiotherapy timepoint; EFS is measured from randomization before definitive local therapy. EFS includes the event "progression precluding surgery," which is absent in DFS.

Invasive DFS (iDFS): DFS component restricted to invasive disease events — excludes in situ recurrences. Standard definition in adjuvant breast cancer trials per STEEP (Standardized Definitions for Efficacy End Points) criteria: ipsilateral invasive breast tumor recurrence, regional node recurrence, distant recurrence, contralateral invasive breast cancer, or death from any cause.

Relapse-Free Survival (RFS): Functionally similar to DFS; may exclude non-cancer deaths in some definitions. Used interchangeably with DFS in melanoma and some solid tumor contexts but less preferred by FDA when all-cause mortality approach is available.

EFS with AML transformation (MDS-specific): Per FDA July 2025 MDS draft, AML transformation is explicitly endorsed as an EFS component event for higher-risk MDS trials. EFS incorporating AML transformation is preferred over OS for higher-risk MDS in disease-modifying therapy trials.

Regulatory Position

Traditional (regular) approval:

DFS has been "the primary basis of approval for adjuvant breast cancer hormonal therapy, adjuvant colon cancer, and adjuvant cytotoxic breast cancer therapy." (FDA 2018, Final)

DFS has supported traditional approval for: breast cancer (multiple subtypes), colorectal cancer (adjuvant), gastrointestinal stromal tumors (GIST), melanoma, renal cell carcinoma. (FDA 2018)

"In December 2003, the ODAC consensus was DFS prolongation represented clinical benefit if the magnitude of this benefit outweighed the toxicity of the adjuvant treatment. In May 2004, the ODAC recommended that DFS be considered a clinical endpoint for colon cancer drugs in the surgical adjuvant setting." (FDA 2018)

Accelerated approval:

DFS/EFS can serve as a surrogate for accelerated approval in adjuvant/neoadjuvant settings when the treatment effect is large and surrogacy for OS is plausible but OS is impractical to measure (long post-recurrence survival, effective salvage therapy).

EFS (lymphoma, hematologic):

EFS is the dominant primary endpoint in Phase 3 lymphoma trials (42/300, 14% of lymphoma Phase 3 trials). EFS-12 and EFS-24 (landmark rates at 12 and 24 months) are used as surrogates for OS in aggressive lymphoma.

EFS (AML/MDS):

EFS incorporating AML transformation endorsed by FDA July 2025 MDS draft as preferred endpoint for higher-risk MDS with disease-modifying therapies. AML transformation is a clinically meaningful event distinct from general progression.

Status: FDA Cancer Endpoints 2018 = Final; FDA Myelodysplastic Syndromes Endpoints July 2025 = Draft

When to Use

DFS as primary:

  • Adjuvant breast cancer (most common): 49/300 trials (16%). Examples include:
  • HR+/HER2- (CDK4/6 inhibitors: palbociclib PALLAS, ribociclib NATALEE)
  • HER2+ adjuvant (pertuzumab APHINITY, neratinib ExteNET, T-DM1 KATHERINE)

  • TNBC adjuvant (olaparib OlympiA, capecitabine CREATE-X)

  • Adjuvant colorectal cancer: 38/300 CRC trials (13%). FOLFOX, capecitabine adjuvant Stage III

  • Adjuvant melanoma: RFS/DFS in BRAF-mutated melanoma (vemurafenib, dabrafenib + trametinib), IO adjuvant (nivolumab, pembrolizumab)
  • Adjuvant RCC: 10/112 RCC trials (9%). Sunitinib adjuvant (S-TRAC), adjuvant pembrolizumab (KEYNOTE-564)
  • GIST adjuvant: Imatinib RFS (primary basis of adjuvant GIST approval)

EFS as primary:

  • Neoadjuvant breast cancer: pCR → EFS/DFS as confirmatory endpoint (I-SPY2 platform, KEYNOTE-522)
  • Lymphoma Phase 3: EFS-12, EFS-24 as surrogate endpoints in DLBCL, follicular lymphoma; EFS incorporating relapse, progression, death
  • AML Phase 3: EFS (CR + duration) as co-primary with OS in some trials (15/489 AML trials, 3%)
  • Higher-risk MDS (2025 draft): EFS incorporating AML transformation; preferred over ORR surrogates for disease-modifying therapies

OS preferred over DFS when:

  • Post-recurrence survival is short (< 12–18 months)
  • Standard effective salvage therapy is unavailable
  • Colorectal cancer metastatic setting (OS dominates)
  • Pancreatic cancer (OS always preferred)

DFS/EFS insufficient alone when:

  • Post-recurrence survival is very long (> 10 years) and OS event rate is too low
  • DFS improvement does not translate to OS (proven failure: antiangiogenic adjuvant in some tumors)
  • Adjuvant setting with high cure rates (e.g., Stage I NSCLC) where absolute DFS benefit is tiny

Design Considerations

Assessment schedule and monitoring

  • Adjuvant solid tumor trials: Typically every 3–6 months for years 1–2, every 6 months for years 3–5, annually thereafter (varies by cancer type and recurrence patterns)
  • Assessment modality: Per disease-specific guidelines (CT for colorectal/lung, mammography for breast, PET-CT for lymphoma per Lugano)
  • Critical FDA requirement: "Differences between study arms in the frequency, timing, or reason for unscheduled assessments can introduce bias." (FDA 2018)
  • Symmetric visit schedules between arms are mandatory — any difference in imaging frequency must be pre-specified and justified

Death handling (critical design decision)

FDA 2018 guidance is explicit: "Considering deaths from all causes as recurrences can minimize bias." (§III.B.1)

Three options with tradeoffs:

  1. Count all deaths as DFS events (FDA-preferred):

    • Minimizes bias; non-informative censoring assumption likely holds
    • May inflate event count (deaths after long follow-up without recurrence)
    • Standard approach in modern adjuvant trials

  2. Count only cancer-related deaths:

    • Requires cause-of-death attribution — introduces attribution bias
    • Less favored by FDA unless competing risk analysis supports sensitivity
    • Less commonly used in contemporary trials

  3. Censor non-cancer deaths:

    • Violates non-informative censoring assumption if competing risk structure differs between arms
    • May bias results if one arm has higher non-cancer mortality
    • Not recommended as primary approach

FDA generally recommends option 1 for primary analysis; competing risk analysis (Fine-Gray) as sensitivity.

IRC requirements for recurrence verification

DFS/EFS in unblinded trials should have IRC verification for recurrence events:

  • Radiologic recurrence: IRC reviews imaging to confirm per protocol criteria (≥20% increase in sum of lesion diameters per RECIST 1.1, or new lesion)
  • Pathologic recurrence: Central pathology review when tissue confirmation is obtained
  • Blinded IRC may be waived in double-blind trials with robust effect sizes (HR ≤0.70)
  • Training: IRC members must be trained on protocol recurrence criteria; written charter required

Pre-specified DFS/EFS event lists

Protocol must explicitly enumerate all events counted:

Endpoint Standard Events Comments
DFS (solid tumors) Local recurrence, distant recurrence, contralateral cancer, death from any cause All-cause mortality approach preferred
iDFS (breast - STEEP) Ipsilateral invasive breast recurrence, regional node recurrence, distant recurrence, contralateral invasive breast cancer, death from any cause Does NOT include in situ recurrences or contralateral DCIS
EFS (neoadjuvant) Progression precluding surgery, local recurrence, distant recurrence, death Measured from pre-surgery randomization
EFS (lymphoma) Progression (≥50% increase per Lugano), relapse after CR, initiation of new therapy, death New therapy initiation is key event for EFS
EFS (AML/MDS) Failure to achieve CR at 3 months, relapse from CR, AML transformation (MDS only), death AML transformation: BM blasts ≥20% (MDS draft 2025)

Statistical analysis for DFS/EFS

Log-rank test and HR estimation:

For DFS with log-rank test:

- Null hypothesis: H0: HR = 1 (no treatment difference)
- Test statistic: Z ~ N(0,1) under H0
- Reject if |Z| > 1.96 (two-sided α=0.05)
- HR estimation via Cox regression with 95% CI

Sample size calculation (Schoenfeld):

Required DFS events: d = (z_α + z_β)² / [log(HR)]²

Example — Adjuvant breast HR+ (CDK4/6 inhibitor):

- Baseline 5-year DFS rate (control): 75% 
  → Median DFS ~18 months

- Target HR: 0.75 (25% DFS improvement)
- α = 0.025 one-sided (z_α = 1.96), power = 85% (z_β = 1.04)
- d = (1.96 + 1.04)² / [log(0.75)]² 
- d ≈ 9.0 / 0.0625 = 144 events

For 2-arm trial with 1:1 randomization, ~240 patients 
enrolled; ~180 DFS events expected at 5-year follow-up

Competing risk analysis:

When non-cancer mortality is substantial (>10% of events), Fine-Gray subdistribution hazard analysis should be reported as sensitivity:

# Fine-Gray model (competing risks)
library(cmprsk)
fit <- crr(ftime = dfs_time, fstatus = dfs_status, 
           cov1 = treatment_arm, failcode = 1, 
           cencode = 0)  # Treats non-cancer death as competing

Alpha allocation: DFS primary + OS secondary

When DFS is primary and OS is key secondary (standard adjuvant design):

  • Hierarchical testing: OS tested only if DFS is significant (gatekeeping)
  • OS interim analysis at DFS primary analysis: Very small alpha (0.001) to preserve overall Type I error
  • OS maturity at DFS primary: Often <30% — immature OS supports DFS but does not confirm OS benefit independently
  • Final OS analysis: Typically 3–5 years after DFS primary analysis when sufficient OS events accrue

Minimum follow-up and event count targets

  • Adjuvant breast (CDK4/6 inhibitors): Typical 150–300 DFS events for 80% power at HR 0.70–0.75
  • Adjuvant colorectal cancer: 250–400 DFS events (5-year DFS rate ~70–75% in control arm)
  • Adjuvant melanoma (stage III): 200–350 DFS events; longer follow-up due to late relapses
  • Neoadjuvant breast (EFS primary): 300–500 EFS events; longer follow-up post-pCR to capture delayed recurrences
  • Minimum median follow-up: At least 2× expected median DFS in control arm (e.g., if median DFS = 5 years, follow ≥10 years)

Intercurrent Events

1. Death without prior documented recurrence

Most impactful IE in DFS trials. Patients die (non-cancer cause) without documented progression.

ICH E9(R1) strategy: Composite — count all deaths as DFS events regardless of cause

SAP language: "Death from any cause without prior documented disease recurrence will be counted as a disease-free survival event. Non-cancer deaths are not censored and are included in the primary analysis."

Sensitivity strategy: Competing risk analysis treating non-cancer death as a competing event (Fine-Gray subdistribution hazard model)

2. New anti-cancer therapy before recurrence

Patient initiates additional adjuvant or systemic therapy outside protocol without recurrence documented.

ICH E9(R1) strategy: Treatment policy — continue on-study follow-up; censor only if clear loss to follow-up

SAP language: "Patients who initiate non-protocol anti-cancer therapy before a documented DFS event will remain in the analysis. Censoring will occur only at the last known alive date if the patient is subsequently lost to follow-up. New therapy initiation is not censored."

3. Second primary cancer (DFS-specific handling)

In adjuvant breast cancer (iDFS), contralateral invasive breast cancer counts as a DFS event. Second primary cancers of other types may or may not be counted depending on protocol.

STEEP-compliant iDFS strategy: Contralateral invasive breast cancer = DFS event; other second primaries per pre-specification

SAP language: "Contralateral invasive breast cancer will be counted as an invasive disease-free survival event. Other second primary malignancies will be included as DFS events per the pre-specified definition in this SAP [specify inclusion/exclusion]."

4. Treatment discontinuation without recurrence

Patients stop adjuvant treatment early (toxicity) but continue disease monitoring.

ICH E9(R1) strategy: Treatment policy — continue assessments; recurrence/death still counted

SAP language: "Early discontinuation of study treatment does not constitute censoring of the DFS endpoint. Patients who discontinue treatment will continue to be followed for disease recurrence and survival per protocol schedule."

Regulatory Precedent

NCT# Drug Indication Endpoint Outcome
NCT01805271 Everolimus + endocrine HR+ adjuvant breast DFS at 2 years Negative; HR >1.0
NCT00038467 Exemestane vs. tamoxifen HR+ early breast DFS at Month 36 Exemestane superior; HR ~0.80
NCT02625441 Trastuzumab + pertuzumab + docetaxel HER2+ adjuvant iDFS Traditional approval; HR ~0.62
NCT04622319 T-DXd vs. T-DM1 HER2+ residual disease iDFS T-DXd superior in DESTINY-Breast05
NCT01772472 Trastuzumab emtansine (T-DM1) HER2+ residual post-neoadjuvant iDFS at 3 years ~88% iDFS (KATHERINE trial)
NCT00878709 Neratinib after trastuzumab HER2+ adjuvant iDFS at year 2 Neratinib superior; ExteNET trial
NCT00493870 TAC vs. TC chemotherapy Adjuvant breast iDFS at 3 years TAC superior; foundational adjuvant trial

Note: Adjuvant breast DFS trials comprise 49/300 (16%) of breast Phase 3 trials in CTG dataset. Median iDFS follow-up: 5–7 years for detection of late breast cancer recurrences.

Limitations and Pitfalls

1. DFS may not translate to OS: Several large adjuvant trials have shown DFS benefit without OS benefit (palbociclib PALLAS, bevacizumab adjuvant trials in multiple tumor types, sunitinib SSGXXX adjuvant GIST). Long post-recurrence survival means patients who recur can still achieve prolonged OS with effective salvage therapy.

  • Mitigation: Pre-specify OS follow-up regardless of DFS result; FDA mandates OS collection even when DFS is approval basis

2. OS follow-up is required regardless of DFS approval: Even when DFS is the approval endpoint, OS data collection is mandatory in randomized trials (FDA 2025 guidance). OS must be monitored as a safety endpoint to rule out potential harm and document long-term survival trends.

  • Mitigation: Plan 10-year OS follow-up; establish interim OS analyses schedule; publish OS data with delayed recurrence patterns

3. Unscheduled assessment bias (detection bias): If one arm has more frequent imaging (e.g., patients on experimental arm are monitored more closely for toxicity-related scans), progression/recurrence may be detected earlier in that arm → appears to show worse DFS in the experimental arm despite equivalent tumor biology.

  • Mitigation: Mandate symmetric assessment schedules in protocol; pre-specify handling of unscheduled assessments; sensitivity analysis excluding unscheduled assessments

4. Cause of death attribution bias: In older DFS definitions that censor non-cancer deaths, attribution errors (cancer vs. cardiovascular death) introduce bias and increase bias risk if treatment has differential non-cancer toxicity. FDA's all-cause approach avoids this.

  • Mitigation: Use all-cause mortality approach; pre-specify competing risk analysis; DSMC oversight of cause-of-death attribution

5. EFS-12/EFS-24 as lymphoma surrogate validity varies: EFS landmarks are widely used in aggressive lymphoma but their surrogacy for OS is population-dependent. In older chemotherapy regimens (RCHOP), EFS-24 correlates well with OS. In novel therapy settings or in slow-responding patient subsets, this surrogacy must be re-validated.

  • Mitigation: Establish surrogate validation in FDA type-C meeting; pre-specify subgroup analyses by histology/prognostic score; plan OS confirmatory trial for accelerated approval

6. AML transformation definition in MDS: Including AML transformation as an EFS component event (FDA 2025 MDS draft) is appropriate for higher-risk MDS but adds complexity to event adjudication. AML transformation definition must be pre-specified (bone marrow blast threshold ≥20% is standard).

  • Mitigation: Lock AML transformation definition before study start; implement central bone marrow review; train sites on blast counting methodology; DSMC oversight of transformation adjudication

7. Contralateral breast cancer handling in iDFS: STEEP criteria explicitly include contralateral invasive breast cancer as an iDFS event, but in situ lesions (DCIS) are excluded. Protocol must clearly distinguish invasive vs. in situ; misclassification affects event count and HRs.

  • Mitigation: Require pathology review for all contralateral lesions; SAP clearly defines invasive vs. in situ; conduct sensitivity analysis excluding/including contralateral DCIS

8. Very long follow-up creates loss-to-follow-up bias: In 10+ year DFS follow-ups, patient migration, death unrelated to cancer, or voluntary withdrawal increases censoring rates. If treatment arm differs in follow-up adherence, informative censoring can bias DFS estimates.

  • Mitigation: Implement robust patient tracking systems; annual contact attempts; analyze follow-up rates by arm; sensitivity analysis with alternative censoring rules

Source: FDA Cancer Endpoints 2018 (Final); FDA Myelodysplastic Syndromes Endpoints Draft Guidance (July 2025, Draft) Status: Cancer Endpoints 2018 = Final; MDS July 2025 = Draft Compiled from retrieved FDA chunks + ClinicalTrials.gov records (300 breast Phase 3 trials with iDFS/DFS focus; 300 CRC adjuvant trials; 112 RCC adjuvant trials)