Skip to content

NSCLC Indication Guide: FDA Regulatory Endpoints & Trial Design Patterns

Purpose: Comprehensive synthesis of FDA NSCLC Endpoints 2015 (Final) guidance with contemporary Phase 3 trial design patterns (N=300 CTG trials). Provides biostatisticians with setting-specific regulatory requirements, endpoint selection frameworks, sample sizing guidance, and intercurrent event management strategies for NSCLC clinical trials.

1. Regulatory Foundation: FDA NSCLC Endpoints 2015 (Final)

Core Regulatory Position

"We consider OS to be the standard clinical benefit endpoint that should be used to establish efficacy of a treatment in patients with locally advanced or metastatic NSCLC. However, other endpoints can be considered for regulatory decision-making based on the population and risk-benefit profile of a drug." — FDA NSCLC Endpoints 2015, §III (Final guidance, April 2015)

NSCLC is heterogeneous: Disease presentation, molecular alterations (EGFR, ALK, ROS1, KRAS G12C, MET exon 14, RET, NTRK, BRAF V600E, PD-L1 status), histology (squamous vs. nonsquamous), and stage determine endpoint appropriateness and trial design.

Endpoint Acceptability Summary

Endpoint FDA Position Regulatory Use Conditions
OS Gold standard Regular approval Any setting; preferred for unselected/IO
PFS Conditional acceptance Regular approval HR ≤0.60 + ≥3–6 mo absolute diff; large effect mandatory
DFS (adjuvant) Acceptable Regular approval ODAC 17:2 vote (2015); first agents approved 2023+
ORR Limited Accelerated approval only Large response rate (≥20–30% above control); durable responses
pCR (neoadjuvant) Acceptable Accelerated approval Confirmatory EFS/OS trial required for regular approval

2. NSCLC Standard of Care & Molecular Context

Approved Regimens by Setting & Molecular Subtype

1L Metastatic

Subtype SOC Comparator Key Evidence HR/Outcome
EGFR del19/L858R Osimertinib (3rd-gen TKI) FLAURA (vs. std 1st-gen TKI) PFS HR 0.46
EGFR T790M Osimertinib AURA3 (2L post-TKI failure) PFS HR 0.40
EGFR exon 20 insertion TAK-788 or Platinum-doublet NCT04129502 (TAK-788 vs. chemo) PFS HR ~0.60–0.70
ALK rearranged Alectinib (preferred 1L) ALEX (vs. crizotinib) PFS HR 0.34
ROS1 rearranged Crizotinib PROFILE 1001 ~68% response rate
BRAF V600E Dabrafenib + Trametinib BRF113928 Dabrafenib/MEK inhibitor combo
MET exon 14 Capmatinib or Tepotinib GEOMETRY mono-1, VISION ~68% response rate
NTRK fusion Larotrectinib or Entrectinib LOXO-101, ALKA-372 Tissue-agnostic; high RR
Unselected/PD-L1 mixed Pembro/Nivo + chemo or chemo alone KEYNOTE-407, IMpower150 OS HR 0.71–0.78

2L+ and Maintenance

Setting SOC Comparator Key Evidence
Post 1L platinum (unselected) Docetaxel or Pemetrexed ASTER era (similar efficacy)
Post EGFR TKI (T790M−) Nivolumab + chemo or chemo NCT02864251 (Nivo + chemo)
Post ALK TKI Ceritinib or Alectinib ASCEND-5: HR 0.49 PFS
Any, IO-naive Pembrolizumab or Nivolumab KEYNOTE-010, KEYNOTE-024

Adjuvant (Stage II–IIIA Post-Resection)

Subtype SOC Comparator Key Evidence
EGFR-mutant Stage II–IIIA Osimertinib (3 years) ADAURA: DFS HR 0.17
ALK+ Stage II–IIIA Alectinib ALINA: DFS HR 0.36
Stage III unresectable (consolidation) Durvalumab (post-chemoradiation) PACIFIC: OS HR 0.68, PFS HR 0.55
General Stage II–IIIA Adjuvant chemotherapy (cisplatin-doublet) LACE meta-analysis: 4–5% OS benefit

Neoadjuvant (Pre-Surgery)

Setting SOC Comparator Primary EP Key Trials
Stage II–IIIB resectable Chemo + IO (nivo/pembro) pCR KEYNOTE-671, CheckMate 816

3. Endpoint Frequency: CTG Phase 3 NSCLC Database (N=300 trials)

Endpoint Count % of Trials Primary Settings Regulatory Status
PFS 109 36.3% 1L targeted (EGFR/ALK), 2L IO, Stage III Conditional acceptance (HR ≤0.60)
OS 109 36.3% 1L unselected/IO, 2L+, adjuvant co-primary Gold standard
ORR 26 8.7% EGFR/ALK/BRAF TKI (1L), response-enriched Limited; accelerated approval only
DFS/EFS 16 5.3% Adjuvant/neoadjuvant FDA-acceptable adjuvant; EFS confirmatory
Other 40 13.4% Specialized (DOR, TTNT, maintenance PFS) Context-dependent

Key Observations:

  • PFS and OS equiprevalent (109 each = 36.3%). PFS acceptable as primary with large HR (≤0.60) and absolute difference (≥3–6 mo).
  • ORR used only 8.7%, reflecting FDA's restrictive guidance (ORR does not reliably predict OS in NSCLC).
  • Adjuvant trials comprise <5.3%, despite regulatory interest in DFS.

Median enrollment: 450 patients | Double-blind: 38 trials (12.7%) | Open-label: 177 trials (59%)

4. Regulatory Requirements by Endpoint

Overall Survival (OS) — Gold Standard

FDA Position:

"OS is the preferable primary efficacy endpoint… as it is definitive and easy to determine. An observed OS benefit in a well-conducted, randomized trial can be directly attributed to the experimental therapy." (FDA 2015)

When required:

  • 1L unselected/IO combinations (primary or co-primary)
  • 2L+ immunotherapy (primary)
  • Adjuvant (co-primary with DFS)
  • Consolidation (co-primary with PFS)

Advantages:

  • Definitive, unambiguous (death is objective)
  • Not confounded by subsequent therapies in ITT analysis (treatment policy strategy)

Challenges:

  • Long follow-up (24–48+ months)
  • Multiple post-progression therapies dilute signal (>70% of patients receive subsequent lines)

Mitigation strategies:

  • RPSFT (Rank-Preserving Structural Failure Time) or IPCW (Inverse Probability of Censoring Weighting) adjustment for subsequent therapy if >30% crossover
  • Treatment policy as primary estimand (ITT, no adjustment); hypothetical estimand as sensitivity

Progression-Free Survival (PFS) — Conditional Acceptance

FDA Position (2015):

"PFS may be appropriate as the primary endpoint to establish efficacy for drug approval if the trial is designed to demonstrate a large magnitude for the treatment effect as measured by both the hazard ratio and absolute difference in median PFS and an acceptable risk-benefit profile of the drug is demonstrated." (FDA 2015)

Conditions for PFS-based approval:

  1. Large hazard ratio: HR ≤0.60 historically accepted; typical range 0.40–0.60 for targeted agents

    • Example: ARCHER 1009 (dacomitinib vs. erlotinib): HR 0.59 → approved on PFS
    • Example: ALEX (alectinib vs. crizotinib): HR 0.34 → approved on PFS

  2. Large absolute median PFS difference: ≥3–6 months improvement

    • Example: TAK-788 vs. platinum-doublet (EGFR exon 20): 12 mo vs. 6 mo → 6-month difference (acceptable)
    • Example: 2L IO vs. docetaxel: 3.5 mo vs. 2.8 mo → 0.7-month difference (likely insufficient)

  3. OS trend supportive: Immature OS data should trend toward benefit (not required to be significant at PFS approval)

  4. ODAC validation: ODAC voted 11:8 that PFS may support regular approval in metastatic NSCLC (2015)

NSCLC-Specific PFS Requirements (beyond general cancer guidance):

No early interim PFS stopping:

"Interim efficacy analyses of PFS before completion of patient accrual are discouraged. Early interim efficacy analyses of PFS that cross a stopping boundary often overstate the magnitude of the effect." (FDA 2015)

  • Implication: Cannot stop trial early for PFS efficacy; futility and safety monitoring acceptable
  • Rationale: Systematic overestimation at interim analyses due to regression to the mean
  • Practice: Typical interim analysis at 50% events is futility-only; efficacy decision made at final analysis (80–100% events)

IRC required for open-label trials:

"If investigator-assessed PFS is considered the primary endpoint for establishing efficacy, then evidence of lack of bias should be provided, for example, by verification of investigator assessment in a random sample audit conducted by an independent review committee." (FDA 2015)

  • Requirement: Minimum 10–20% random audit of investigator-assessed progressions by IRC
  • Timing: Audit must be completed before database lock
  • Discordance analysis: Concordance ≥85% expected; document reasons for discordance
  • Sensitivity analysis: IRC-reviewed PFS as sensitivity analysis

Justification narrative required:

  • Sponsors must prospectively justify PFS over OS in protocol
  • Discuss NSCLC heterogeneity (histology, stage, molecular subgroups)
  • Explain why PFS difference clinically meaningful given toxicity/tolerability
  • Document historical PFS-to-OS correlation in similar population

Assessment schedule symmetry:

  • Same imaging modality (CT preferred; MRI acceptable if consistent), timing (±3-day window), frequency in both arms
  • Unscheduled imaging documented and analyzed as sensitivity
  • Open-label trials: assess unscheduled imaging bias separately

Objective Response Rate (ORR) — Limited Acceptability

FDA Position:

"Treatment effects on ORR have not been demonstrated to reliably predict corresponding effects on survival in NSCLC. We consider demonstration of ORR alone to be a surrogate endpoint reasonably likely to predict clinical benefit only when the treatment effect size is large and the responses are durable. In these circumstances, ORR has been used as the basis only for accelerated approval for NSCLC." (FDA 2015)

When ORR may support accelerated approval:

  • Large response rate difference (≥20–30% above control or historical rate)
  • Responses associated with documented symptom improvement
  • Durable responses (>6 months median DOR)
  • Example: Single-arm Phase 2 data in rare molecular subgroups (NTRK fusions, RET fusions, BRAF V600E)

ORR DOES NOT support regular approval in NSCLC (unlike hematologic malignancies)

Mitigation: If ORR-based labeling claim intended, require:

  • Large response rate differences (≥20–30%)
  • Confirmatory OS or PFS trial data
  • Documented symptom/QoL benefit

Disease-Free Survival (DFS) & Event-Free Survival (EFS) — Adjuvant/Neoadjuvant

FDA Position (2015):

"Since no drug was approved for the adjuvant treatment of NSCLC, hypothetically disease-free survival can be a reasonable endpoint to evaluate new therapy in an adjuvant setting." (ODAC voted 17:2, FDA 2015)

Regulatory developments post-2015:

  • Osimertinib (ADAURA, 2020): DFS HR 0.17 (EGFR-mutant adjuvant) → traditional approval based on DFS
  • Alectinib (ALINA): DFS (ALK+ adjuvant) → supported approval
  • Nivolumab (CheckMate 816): pCR (neoadjuvant) → accelerated approval; EFS confirmatory for regular approval
  • Pembrolizumab (KEYNOTE-671): pCR (neoadjuvant) → accelerated approval; EFS confirmatory

DFS definition in adjuvant NSCLC:

  • Time from randomization to first recurrence (locoregional or distant) OR death from any cause
  • All-cause mortality preferred (FDA 2015): non-cancer deaths included, not censored
  • Competing risk analysis optional: cancer-specific DFS (censoring non-cancer deaths) as sensitivity; Fine-Gray model for CI estimates

EFS definition in neoadjuvant NSCLC:

  • Time from randomization to progression, recurrence post-surgery, second malignancy, or death
  • Often used as confirmatory endpoint after pCR accelerated approval

5. When to Use: Endpoint Selection by Setting

1L Metastatic — Unselected or IO Combinations

Standard approach: OS primary or OS + PFS co-primary

Rationale:

  • Heterogeneous population (mixed EGFR/ALK/KRAS/wild-type); no reliable biomarker for response prediction
  • IO benefit modest (HR 0.70–0.75 vs. chemo); durability uncertain at approval time
  • OS maturation reasonable (24–36 months)

Real examples:

  • KEYNOTE-407 (pembrolizumab + chemo vs. chemo, squamous NSCLC): OS primary → OS HR 0.71 (p<0.001)
  • CheckMate 227 (nivolumab + ipilimumab vs. chemo): OS + PFS co-primary in unselected and PD-L1+ populations
  • IMpower130 (atezolizumab + bevacizumab + chemo): OS primary → OS HR 0.78

Sample size: 300–600 randomized; 180–260 PFS events + 100–140 OS deaths for maturity

HR assumptions: PFS HR 0.65–0.75; OS HR 0.75–0.85

Follow-up: 18–24 months to PFS; 36–48 months to OS maturity

1L Metastatic — Targeted Therapy (EGFR, ALK, ROS1, KRAS, MET, RET, NTRK)

Standard approach: PFS by IRC (RECIST 1.1) as primary; OS as secondary

Rationale:

  • Large treatment effect sizes (HR 0.20–0.55) in biomarker-selected populations
  • Rapid response kinetics; PFS reaches maturity faster than OS
  • OS confounded by post-progression therapies (multiple salvage options available)
  • OS secondary analysis still required; must show OS trend supportive of PFS benefit

Assessment requirements:

  • RECIST 1.1 with central (IRC) review mandatory for open-label trials (≥10–20% audit)
  • Assessment every 6 weeks (2 cycles) until progression
  • Unscheduled imaging documented and analyzed as sensitivity

Real examples:

  • FLAURA (osimertinib vs. standard EGFR TKI, EGFR+ 1L): PFS HR 0.46 → regular approval on PFS
  • ALEX (alectinib vs. crizotinib, ALK+ 1L): PFS HR 0.34 → regular approval on PFS
  • ARCHER 1009 (dacomitinib vs. erlotinib, EGFR+ 1L): PFS HR 0.59 → regular approval on PFS
  • ASCEND-5 (ceritinib vs. chemo, ALK+ 2L post-crizotinib): PFS HR 0.49 → accelerated approval; full approval with OS maturation

Sample size: 200–400 randomized; 120–180 PFS events (targeted); 200–260 PFS events (2nd-gen vs 1st-gen comparison)

HR assumptions: 1L vs chemo: HR 0.40–0.50; 2nd-gen vs 1st-gen: HR 0.55–0.65

Follow-up: 18–24 months PFS; OS follow-up extended (24–36 months for subset maturity)

2L+ Targeted Therapy (Post-1L TKI Failure)

Standard approach: PFS by IRC; OS secondary

Rationale:

  • T790M or other acquired resistance mutations drive treatment selection
  • Rapid progression expected in control arm
  • OS confounded by subsequent lines of therapy (3–5+ lines common in 2L+ setting)

Example:

  • AURA3 (osimertinib vs. platinum/pemetrexed, T790M+ NSCLC): PFS HR 0.40 → regular approval on PFS

Sample size: 150–250 randomized; 120–180 PFS events

HR assumptions: HR 0.45–0.60

Follow-up: ~18–24 months PFS; OS follow-up extended

2L+ Immunotherapy (Post-Chemotherapy)

Standard approach: OS primary; PFS secondary

Rationale:

  • Modest benefit in unselected population (HR 0.70–0.75)
  • Long post-progression survival common (even with chemo progression); OS becomes measurable
  • Historical single-arm data limited; randomization with OS essential

Real examples:

  • KEYNOTE-010 (pembrolizumab vs. docetaxel, 2L+ NSCLC): OS primary → OS HR 0.71 (p=0.0008)
  • OAK (atezolizumab vs. docetaxel, 2L+ NSCLC): OS primary → OS HR 0.73

Sample size: 200–400 randomized; 100–150 PFS events + 100–120 OS deaths

HR assumptions: PFS HR 0.70–0.75; OS HR 0.80–0.90

Follow-up: 18–24 months PFS; 30–48 months OS maturity

Unresectable Stage III — Consolidation or Maintenance

Standard approach: PFS + OS co-primary (consolidation); PFS primary (maintenance)

Rationale:

  • Distinct from metastatic disease; longer follow-up feasible for OS
  • Consolidation therapy post-chemoradiation: OS gains possible with durable benefit
  • Maintenance setting: PFS endpoint suitable as bridge therapy

Real example:

  • PACIFIC (durvalumab vs. placebo post-chemoradiation, unresectable Stage III):
  • PFS + OS co-primary
  • Median PFS not reached (durva) vs. 5.6 months (placebo); HR 0.55 (p<0.001)
  • OS: HR 0.68 (median not reached at 5-year follow-up)
  • Became standard of care for Stage III consolidation

Sample size: 600–900 randomized; ~350 PFS events + ~330 OS deaths at maturity

HR assumptions: PFS HR 0.65; OS HR 0.75

Follow-up: 18–24 months PFS; 48+ months OS maturity

Adjuvant (Post-Surgery, Stage IB–IIIA)

Standard approach: DFS primary; OS secondary

Rationale:

  • Rapidly evolving field; adjuvant OS follow-up requires 5–10 years
  • DFS acceptable per FDA 2015 ODAC statement (17:2 vote)
  • Molecular subgroup enrichment standard (EGFR+ in TKI trials; ALK+ in ALK inhibitor trials)

DFS event definition:

  • First recurrence (locoregional, distant, or brain) or death from any cause
  • All-cause mortality preferred (cancer + non-cancer included)
  • Competing risk sensitivity: cancer-specific DFS (censoring non-cancer) via Fine-Gray

Real examples:

  • ADAURA (osimertinib vs. placebo post-surgery, EGFR+ Stage IB–IIIA):
  • DFS primary → DFS HR 0.17 (2-year DFS: osimertinib 89% vs. placebo 78%; p<0.001)

  • OS secondary (immature); approval based on DFS

  • ALINA (alectinib vs. chemotherapy post-surgery, ALK+ Stage IB–IIIA):

  • DFS primary → DFS HR 0.36
  • Approval based on DFS

Sample size: 400–700 randomized; 300–420 DFS events at primary (median 24–36 months follow-up)

HR assumptions: EGFR TKI adjuvant: HR 0.20–0.35 (osimertinib: HR 0.17); other adjuvant: HR 0.70–0.85

Follow-up: Median 24–36 months DFS analysis; 5+ years OS follow-up

Neoadjuvant (Pre-Surgery, Stage II–IIIB)

Standard approach: pCR (pathologic complete response) primary; EFS confirmatory for regular approval

Rationale:

  • Emerging setting; accelerated approval on pCR + clinical benefit; confirmatory OS/EFS trial required
  • Rapid pathway to approval (pCR measurable at surgery); tumor regression definitive

pCR definition:

  • Absence of residual invasive disease (no cancer cells in surgical specimen)
  • May include in situ lesions (ypT0 or ypT0/is)

Real examples:

  • KEYNOTE-671 (pembrolizumab + chemo vs. chemo, neoadjuvant NSCLC):
  • pCR primary for accelerated approval: 55% vs. 31% (p<0.001)

  • EFS confirmatory for traditional approval

  • CheckMate 816 (nivolumab + chemo vs. chemo, neoadjuvant NSCLC):

  • pCR primary: ~45% vs. 25% (p<0.001)

Sample size: 300–500 randomized; 150–250 pCR responses (binomial, not TTE-based)

Follow-up: pCR assessed at surgery (3–6 months post-randomization); EFS follow-up 2–3 years

6. Design Considerations: NSCLC-Specific Requirements

PFS as Primary — Mandatory Requirements (Conditional on Setting)

Context: PFS is an acceptable primary endpoint for specific NSCLC settings where treatment effects are large and OS is confounded by multiple post-progression therapies (see Section 5: 1L targeted therapy, select 2L targeted settings). When PFS is chosen as primary for these appropriate settings, FDA imposes the following mandatory design requirements:

  1. Large HR AND large absolute median PFS difference

    • HR alone insufficient per FDA 2015
    • Example acceptable: HR 0.50 + 6-month improvement in median PFS (12 mo vs. 6 mo)
    • Example questionable: HR 0.70 + 1-month improvement (3 mo vs. 2 mo)

  2. Justification narrative required in protocol

    • Discuss NSCLC heterogeneity (histology, stage, molecular subgroup)
    • Explain why PFS difference clinically meaningful given drug toxicity
    • Document historical PFS-to-OS correlation in similar population

  3. Interim analysis restrictions

    • No interim efficacy stopping on PFS (FDA explicitly discourages)
    • Futility or safety interim acceptable
    • Final analysis timing pre-specified (e.g., "at 80% of planned PFS events")

  4. IRC verification for open-label trials

    • Mandatory random audit: ≥10–20% of investigator-assessed progressions
    • Audit must be conducted before database lock
    • Discordance analysis and sensitivity included in Statistical Analysis Plan (SAP)
    • Target concordance ≥85%

  5. Assessment schedule symmetry

    • Same imaging modality (CT preferred; MRI if consistent), timing (±3-day window), frequency in both arms
    • Unscheduled imaging documented and analyzed as sensitivity
    • Open-label trials: assess unscheduled imaging bias

Tumor Measurement Requirements (Appendix A, FDA 2015)

Critical documentation in CRF:

  • Target lesions selected BEFORE treatment (no retrospective selection)
  • Lesion-specific ID assigned and tracked throughout trial
  • Same imaging modality at baseline and all follow-ups
  • Zero must be recorded when lesion disappears — cannot be confused with missing assessment
  • New lesions recorded at both scheduled and unscheduled visits
  • Measurable non-target lesions assessed
  • "Missing" vs. "assessed but undetectable" clearly distinguished

"Substantial numbers of missing tumor assessments can potentially overestimate or underestimate treatment differences." (FDA 2015)

Molecular Subgroup Design (NSCLC-Specific)

FDA 2015 explicit statement:

"NSCLC is a heterogeneous disease with varying response to treatment across different molecular and histopathologic subgroups (e.g., pemetrexed, erlotinib). We recommend that clinical trials be prospectively designed to evaluate such differences in treatment effect." (FDA 2015)

Recommended approaches:

  1. Enrichment design — biomarker-selected population

    • Companion diagnostic co-developed and validated
    • All-comers excluded or enrolled as secondary cohort
    • Example: EGFR-mutation testing mandatory for enrollment in EGFR TKI 1L trials

  2. Biomarker-stratified design (common in IO trials)

    • Primary population: Biomarker+ (e.g., PD-L1 ≥50%)
    • Secondary: All-comers
    • Hierarchical testing: biomarker+ at α=0.025; all-comers at α=0.025 if biomarker+ significant

  3. Histology-specific cohorts (squamous vs. non-squamous)

    • Pemetrexed contraindicated in squamous; separate efficacy assumptions
    • Bevacizumab avoided in squamous (bleeding risk); separate arms

Assessment Schedule (Phase 3 Practice)

Setting Assessment Frequency Rationale
Targeted therapy (1L, 3-wk cycles) Every 6 weeks (2 cycles) Aligns with cycle frequency
IO combinations (3-wk chemo) Every 6 weeks; extend to 8 wk after response Flexible after initial response
2L chemotherapy Every 6–8 weeks Standard practice
Maintenance Every 9–12 weeks Less frequent post-stabilization
Adjuvant Every 12 weeks (2 yrs); annually thereafter Lower event rate; longer intervals feasible

Eligibility Criteria Patterns (NSCLC Phase 3, N=300)

Most Common Inclusion Criteria

Criterion Prevalence Details
Histologically/cytologically confirmed NSCLC 100% Biopsy or cytology required
ECOG Performance Status 0–1 ~95% Rarely includes ECOG 2
Stage (setting-dependent) 100% 1L: III/IV/recurrent; Adjuvant: II–IIIA resected
Measurable disease (RECIST 1.1) ~90% ≥10 mm long axis by CT/MRI
Molecular marker status 50–80% Required if targeted; optional for unselected
Prior treatment lines Setting-dependent 1L: none; 2L+: 1–2 prior
Organ function (renal, hepatic) ~95% CrCl ≥30–50 mL/min; AST/ALT ≤3–5× ULN
Bone marrow reserve ~90% Plt ≥100 K/μL; ANC ≥1.5 K/μL; Hgb ≥9 g/dL

Most Common Exclusion Criteria

Criterion Prevalence Rationale
Small cell component 100% Non-small cell histology only
Active secondary malignancy ~95% Exception: non-melanoma skin, in situ cervix, cured >5 yr
Untreated brain metastases ~90% Some trials allow stable, treated CNS
Pregnancy/lactation 100% All TKI/IO trials; contraception required
Significant cardiac disease ~80% LVEF <50%, MI <6 mo excluded
Active infection/severe comorbidity ~85% Uncontrolled diabetes, severe COPD/asthma
Prior same drug class Setting-dependent 1L targeted: no prior TKI; IO: varies
Liver cirrhosis ~80% Decompensated liver disease excluded
HIV/Hep B/Hep C ~70% Exclusionary unless on appropriate treatment

Notably Restrictive Criteria

  • Molecular Testing: EGFR/ALK/PD-L1 status must be known; "unknown" often excluded in biomarker-driven trials
  • Prior Treatment Washout: IO trials typically require 4-week washout; some extend to 12 weeks
  • CNS Disease: Untreated brain metastases near-universal exclusion; recent IO trials allow asymptomatic lesions

7. Sample Size Patterns: NSCLC Phase 3 Trials (CTG Dataset, N=300)

Enrollment Distribution

Metric Value Notes
Median Enrollment 450 Modern Phase 3 standard
25th Percentile ~250 Smaller trials often enriched (biomarker, single-arm)
75th Percentile ~600 Larger unselected or global registration studies
Largest (ARCHER 1009) 878 Large EGFR TKI Phase 3 trial
Smallest Viable ~60–100 Rare; typically single-arm, enriched, or supportive

Typical HR Assumptions by Setting

Setting Endpoint Typical HR Justification Events (80% power)
1L Targeted (TKI vs Chemo) PFS 0.40–0.50 Historical superiority (EGFR del19: HR 0.32–0.46) 120–150
1L Targeted (2nd-gen vs 1st-gen) PFS 0.55–0.65 Incremental improvement (dacomitinib: HR 0.59) 180–220
1L IO ± Chemo vs Chemo PFS 0.65–0.75 Checkpoint inhibitor class effect 180–240
1L IO ± Chemo vs Chemo OS 0.75–0.85 Modest OS benefit; longer follow-up 100–140
2L Targeted (Post-TKI) PFS 0.45–0.60 ALK post-crizotinib: HR 0.50 120–180
2L+ IO PFS 0.65–0.75 Class effect 150–220
Adjuvant DFS (EGFR TKI) DFS 0.25–0.35 Osimertinib: HR 0.17 300–420
Adjuvant OS OS 0.80–0.90 Smaller HR; longer follow-up 200–300

Follow-Up Duration Patterns

Setting Median Follow-Up Rationale
1L Metastatic (PFS primary) 18–24 months Targeted trials faster; TKI trials extend to 24+ for subset maturity
1L Metastatic (OS co-primary) 30–48 months OS slower; requires 60–70% death events
2L+ (PFS primary) 12–18 months Shorter survival context
2L+ (OS required) 24–36 months OS maturation faster than 1L
Adjuvant DFS 24–36 months ~30% DFS events at 2 years Stage IIIA
Adjuvant OS 48–60+ months Extended follow-up for long-term benefit
Consolidation Stage III 18–24 months (PFS); 48+ (OS) PFS faster; OS requires extended maturity

8. Intercurrent Events (NSCLC-Specific): Strategies & SAP Language

IE 1: Subsequent Anti-Cancer Therapies (Multiple Lines Available)

Unique NSCLC Challenge: Patients frequently receive 4–6+ lines of therapy. This substantially dilutes OS signals. Nearly 90% of trials address this IE.

For PFS:

  • Strategy: Treatment policy (does not affect PFS directly; censoring at new therapy start handles it)
  • SAP language: "Patients initiating subsequent anti-cancer therapy before documented progression will have their PFS censored at the date of the last adequate tumor assessment prior to initiation of new therapy."
  • Operational requirement: Must document date of new therapy initiation

For OS:

  • Strategy: Treatment policy (ITT, per randomized assignment)
  • SAP language: "Overall survival will be analyzed per intent-to-treat principle. All subsequent anti-cancer therapies are documented but do not affect the assignment for OS analysis."
  • Sensitivity: Hypothetical (RPSFT/Two-Stage Regression with Subsequent Treatment) if substantial crossover detected (>30%)

Adjustment methods:

  • RPSFT (Rank-Preserving Structural Failure Time): Assumes structural model of treatment effect; often requires untestable assumptions about counterfactual outcomes
  • IPCW (Inverse Probability of Censoring Weighting): Weights observations by probability of remaining uncensored; less assumption-heavy than RPSFT
  • Two-Stage Regression: Direct adjustment for post-progression treatment effect

IE 2: Treatment Discontinuation Without Progression

Common in IO and targeted therapy trials (toxicity-driven; discontinuation rates 10–30%). ~70–80% of trials address this.

Primary Estimand (Treatment Policy/ITT):

Efficacy evaluated in Intent-to-Treat population. Patients discontinuing prematurely 
(toxicity, preference, physician discretion) remain on-study; progression and death assessed 
regardless of active drug exposure. This strategy reflects real-world compliance and tolerability.

Sensitivity Estimand (Per-Protocol):

Per-protocol analysis includes only patients completing planned treatment duration without 
premature discontinuation. This estimates PFS/OS benefit in compliant subset, providing 
context for tolerability impact.

Adjustment methods:

  • Standard Kaplan-Meier (ITT and per-protocol, no adjustment needed)
  • If dropout >20% in one arm: IPCW can adjust for informative dropout

IE 3: Progression Assessment Delay or Missed Imaging (Bias in Open-Label)

Frequency: ~60% of trials manage via IRC/BICR mandates.

Primary Estimand (Investigator Assessment with IRC Sensitivity):

Primary analysis uses investigator assessment of progression per RECIST 1.1. Imaging 
scheduled every 6–8 weeks metastatic, 12 weeks adjuvant. If delayed >3 weeks, assessed 
based on most recent imaging or clinical progression. For open-label trials, Blinded 
Independent Review Committee (IRC) reviews and adjudicates progression independently 
(pre-specified sensitivity analysis).

Sensitivity Estimand (IRC Assessment):

IRC-reviewed PFS (blinded to treatment) serves as sensitivity analysis to account for 
investigator bias in open-label designs. IRC uses same RECIST 1.1 criteria with central 
adjudication.

IRC Requirement (FDA NSCLC 2015: "IRC is expected for open-label trials"):

  • Mandatory blinded assessment
  • 10–20% audit rate typical
  • Concordance Analysis: Target >80–85% agreement between investigator and IRC
  • Sensitivity analysis if concordance <80%

IE 4: Death from Non-Cancer Cause (Competing Risk in Adjuvant)

Frequency: ~40% metastatic trials; ~80% adjuvant trials explicitly define handling.

Primary Estimand (All-Cause Mortality in Adjuvant):

Disease-Free Survival (DFS) defined as time from randomization to locoregional recurrence, 
distant metastasis, second primary, or death from any cause—whichever first. All-cause 
mortality included; non-cancer deaths NOT censored. Reflects real-world benefit of surgery 
and adjuvant therapy on overall patient outcomes.

Sensitivity Estimand (Cancer-Specific, Competing Risk):

As sensitivity, non-cancer deaths censored (considered competing risk). Cancer-specific 
DFS isolates anti-cancer effect from life-expectancy factors (cardiac comorbidity, second 
malignancies).

Adjustment methods:

  • Fine-Gray Competing Risk Regression: For cancer-specific DFS; summarize cumulative incidence functions (CIF)
  • Kaplan-Meier (all-cause) for primary analysis

Rationale: FDA prefers all-cause mortality for adjuvant endpoints (conservative, reflects clinical reality)

Frequency: ~30–40% of trials; rare in modern trials but possible in long-duration studies.

Primary Estimand (Last Observed):

Patients withdrawing consent or lost to follow-up censored at date of last known contact 
for disease progression status. For patients with no progression prior to withdrawal, event 
time censored at withdrawal date. Assumes missing-at-random (MAR) mechanism.

Sensitivity Estimand (Worst-Case Imputation):

Conservative worst-case sensitivity: lost-to-follow-up patients treated as experiencing 
the event (progression or death) at censoring date. Provides upper bound on failure rates 
if non-random missing data.

IE 6: EGFR/ALK Resistance Mutation Emergence (Targeted Therapy)

Scenario: Acquired resistance mutation (T790M in EGFR, ALK resistance mutations) detected; new therapy initiated before RECIST progression.

Handling:

  • If new therapy before RECIST progression: Censor PFS at last assessment before new therapy
  • Sensitivity: Document whether RECIST progression preceded or followed therapy switch
  • SAP language: "Patients who initiate new therapy based on resistance mutation testing before documented RECIST progression will have PFS censored at the date of the last adequate tumor assessment prior to new therapy initiation."

IE 7: Unscheduled Imaging (Open-Label Bias)

Scenario: Patient on experimental arm experiences toxicity; physician orders additional scans, detecting earlier progression vs. control arm.

Handling:

  • Document all unscheduled imaging
  • Primary analysis uses all assessments (scheduled + unscheduled) per ITT
  • Sensitivity analysis excludes unscheduled assessments to assess potential bias

SAP language: 

Unscheduled imaging will be documented separately. Primary analysis includes all assessments 
(scheduled and unscheduled) per ITT. Sensitivity analysis excludes unscheduled imaging to 
assess potential bias from differential assessment frequency in open-label trials.

9. Regulatory Precedent: Real NSCLC Phase 3 Trials

NCT# Drug Setting Design Primary EP Key Result Approval
NCT02220894 Durvalumab (PACIFIC) Stage III unresectable consolidation Randomized, placebo PFS + OS co-primary PFS HR 0.52, OS HR 0.68 Regular (PFS+OS)
NCT02142127 Nivolumab ± ipi (CheckMate 227) 1L unselected Randomized 3-arm OS ± PFS OS/PFS significant Regular
NCT01360554 Dacomitinib (ARCHER 1009) 1L EGFR+ Randomized, DB PFS by IRC HR 0.59; 15.3 vs 9.6 mo Regular (PFS)
NCT02143856 Alectinib (ALEX) 1L ALK+ Randomized, open PFS by IRC HR 0.34; 35 vs 11 mo Regular (PFS)
NCT04129502 TAK-788 (Mobocertinib) 1L EGFR exon 20 Randomized, open PFS by IRC HR ~0.60–0.70 Regular (PFS, 2023)
NCT01828112 Ceritinib (ASCEND-5) 2L ALK+ post-crizo Randomized, open PFS by BIRC HR 0.49; 8.6 vs 5.5 mo Accelerated→Regular (PFS)
NCT02511129 Osimertinib (ADAURA) Adjuvant EGFR+ Randomized, DB DFS HR 0.17; 89% vs 78% 2-yr Regular (DFS)
NCT04743922 Pembrolizumab (KEYNOTE-671) Neoadjuvant Randomized pCR 55% vs 31%; p<0.001 Accelerated (pCR)

10. Limitations and Pitfalls (NSCLC-Specific)

1. PFS early stopping (major FDA concern):

FDA explicitly discourages early PFS stopping in NSCLC. Trials stopping at interim PFS analyses systematically overestimate treatment effect due to regression to the mean and random variation.

  • Mitigation: Pre-specify no early efficacy stopping for PFS; futility and safety monitoring acceptable

2. ORR does not predict OS:

Unlike hematologic malignancies, ORR has NOT been validated as OS surrogate in NSCLC. ORR-based regular approval is rare.

  • Mitigation: If ORR-based claim intended, require large response rate differences (≥20–30%) and documented symptom benefit

3. Non-inferiority PFS trials are problematic: NSCLC landscape evolving rapidly (new imaging, molecular subgroup definitions, new SOC). Constancy assumptions are difficult to verify.

  • Mitigation: OS-based NI trials less fraught; FDA prefers superiority designs in NSCLC

4. Unscheduled imaging bias (open-label trials): In open-label targeted therapy, toxicity-driven additional scans may detect progression earlier in experimental arm vs. control.

  • Mitigation: Document all imaging (scheduled/unscheduled); sensitivity analysis excluding unscheduled assessments; DSMC oversight

5. Molecular subgroup drift: Approvals in one molecular subgroup (e.g., EGFR exon 19/21 deletions) do not automatically generalize to other mutations (exon 20 insertions, uncommon mutations).

  • Mitigation: Separate development programs for distinct molecular subgroups; do not assume cross-subtype generalizability

6. Histology-specific effects (squamous vs. non-squamous): Pemetrexed less active in squamous NSCLC; bevacizumab contraindicated in squamous (pulmonary hemorrhage risk).

  • Mitigation: Separate efficacy assumptions and subgroup analyses by histology; pre-specify if primary populations differ

Primary Source: FDA Clinical Trial Endpoints for the Approval of Non-Small Cell Lung Cancer Drugs and Biologics (April 2015, Final) Status: Final guidance (April 2015) Data Sources: ClinicalTrials.gov Phase 3 NSCLC dataset (300 trials); endpoint frequency: PFS 109, OS 109, ORR 26, DFS/EFS 16, other 40 Design Patterns: ingest/study_design_patterns.json (NSCLC); ingest/endpoint_frequency_by_indication.json CTG Index: ingest/ctg_index/ctg_nsclc_phase3_index.json (sample trials, design metadata) Last Updated: 2026-04-10