Introduction
Accelerated and conditional approval pathways have fundamentally transformed oncology drug development, enabling earlier patient access to promising therapies for serious or life-threatening conditions. The U.S. FDA Accelerated Approval (AA) program, European Medicines Agency (EMA) Conditional Marketing Authorisation (CMA), and China National Medical Products Administration (NMPA) Conditional Approval pathway all permit marketing authorization based on surrogate endpoints "reasonably likely to predict clinical benefit" rather than requiring definitive evidence of overall survival (OS) or quality of life improvements12. However, recent evidence reveals substantial uncertainty around these surrogates, with approximately 19% of FDA accelerated approvals ultimately withdrawn and critical gaps in confirmatory trial execution, particularly in China where 57% of confirmatory trials missed required completion deadlines217.
As of January 2026, Project Orbis—the FDA Oncology Center of Excellence initiative established in May 2019—has expanded to eight international regulatory partners (Australia TGA, Brazil ANVISA, Health Canada, Israel Ministry of Health, Singapore HSA, Switzerland Swissmedic, and UK MHRA), facilitating concurrent submission and review of oncology products across jurisdictions1. While this international coordination accelerates global access, it also amplifies the consequences when surrogate endpoints fail to predict genuine clinical benefit.
Magnitude of the Problem: Withdrawal Rates and Clinical Benefit Assessment
A retrospective cohort analysis of 167 FDA accelerated approval indications for 113 anticancer drugs (1992–2022, tracked through August 2024) revealed that 31 indications (19%) were ultimately withdrawn, while 102 (61%) converted to regular approval and 34 (20%) remained ongoing17. Critically, among the 133 indications that were either converted or withdrawn, 112 (84%) used objective response rate (ORR) as the primary endpoint, while only 13 (10%) used progression-free survival (PFS) and 5 (4%) used disease-free survival (DFS)17. This heavy reliance on ORR—despite being the weakest predictor of clinical benefit—represents a fundamental practice gap.
Using the European Society for Medical Oncology Magnitude of Clinical Benefit Scale (ESMO-MCBS) framework, 66% (86/130) of pivotal trials supporting accelerated approvals demonstrated low clinical benefit, 24% (31/130) showed intermediate benefit, and only 10% (13/130) demonstrated substantial benefit at the time of approval17. Low ESMO-MCBS scores were strongly associated with subsequent withdrawal (odds ratio 4.63; 95% CI 1.50–14.33; p = 0.008), indicating that many drugs approved via accelerated pathways offer marginal clinical value17.
China's NMPA experience further illuminates global challenges. Between January 2018 and December 2022, NMPA granted conditional approval to 67 indications (88% oncology), with 79 pivotal trials using predominantly ORR (59%) as the primary endpoint2. However, among 66 confirmatory trials with reported completion timelines, median time to completion was 56 months (IQR 40–73), and 12 of 21 trials (57%) with defined deadlines failed to meet required timeframes by December 31, 20232. This systematic delay in confirmatory evidence generation leaves patients and physicians in prolonged uncertainty about true clinical benefit.
Surrogate Endpoint Validity: Tumor-Type Specific Evidence
Hematologic Malignancies: MRD as the Emerging Gold Standard
Minimal residual disease (MRD) negativity has emerged as the most rigorously validated surrogate endpoint in hematologic malignancies, particularly multiple myeloma (MM). A meta-analysis of 15,304 MM patients from 29 randomized controlled trials (published through June 2024) demonstrated significant negative associations between MRD negativity odds ratios and survival hazard ratios across the entire cohort (β_PFS = −0.20, p < 0.001; β_OS = −0.12, p = 0.023)16. However, critical disease-setting dependency emerged: associations were substantially stronger in newly diagnosed MM (NDMM) patients (β_PFS = −0.35, p < 0.001, R² = 0.79) compared to relapsed/refractory MM (RRMM) patients (β_PFS = −0.06, p = 0.635, R² = 0.0)16.
A pooled analysis of individual patient data from 20 randomized multicenter MM trials (11 studies, 4,773 patients) evaluated MRD-negative complete response (MRD-CR) at 10⁻⁵ threshold as an intermediate endpoint, finding strong patient-level correlations across NDTE, NDTInE, and RRMM populations (global odds ratios 3.06–16.24)22. Trial-level correlations were encouraging (R² = 0.61–0.70 when pooling all populations; R² = 0.67–0.78 in newly diagnosed populations)22. Based on these analyses, the FDA Oncologic Drugs Advisory Committee voted in favor of using MRD negativity as an early endpoint in MM clinical trials for accelerated drug approvals16. This represents a landmark regulatory shift documented in Blood Cancer Discovery (2024) as "a roadmap for other areas of oncology to develop intermediate endpoints"7.
Real-world Evidence
Real-world evidence supports MRD utility: in a retrospective study of 331 newly diagnosed MM patients, MRD negativity (assessed by flow cytometry at Day+100 post-transplantation) was achieved in 47% and associated with significantly prolonged median PFS (49.2 vs. 18.4 months; HR 0.37; p < 0.001) and OS (not reached vs. 74.9 months; HR 0.50; p = 0.007)9. Critically, MRD-positive patients benefited substantially from lenalidomide maintenance (18-month PFS: 81% vs. 46%; HR 0.24; p = 0.002), while MRD-negative patients showed no benefit (p = 0.747)9, demonstrating MRD's utility in treatment personalization—an application not yet systematically integrated into clinical practice.
For acute myeloid leukemia (AML), the MRD Partnership and Alliance in AML Clinical Treatment (MPAACT) consortium documented that MRD-negative status independently predicts improved survival across diverse methodologies, timepoints, and patient subgroups8. However, formal integration as a primary surrogate endpoint in regulatory frameworks requires ongoing validation efforts involving regulatory agencies (FDA, EMA), health technology assessment bodies, and standardization of assessment methodologies8.
In chronic lymphocytic leukemia (CLL), patient-level correlation for PFS/OS was strong (Spearman Rho > 0.9), but treatment-dependent surrogacy emerged: weak correlation for chemotherapy/chemoimmunotherapy (tau = 0.52; 95% CI 0.49–0.55) versus strong correlation for targeted therapies (tau = 0.91; 95% CI 0.89–0.93)21. Critically, end-of-treatment MRD showed strong correlation with PFS (R = 0.88, R² = 0.78) but substantially weaker correlation with OS (R = 0.71, R² = 0.50), revealing a practice gap: MRD's correlation with OS is insufficient to serve as a direct OS surrogate despite regulatory acceptance21.
Solid Tumors: The PFS-ORR Disconnect
For antibody-drug conjugates (ADCs) in advanced solid tumors, a meta-analysis of 25 randomized controlled trials (26 treatment comparisons, 11,729 patients) demonstrated that PFS serves as a robust surrogate for OS (R² = 0.79; 95% CI 0.66–0.92), while ORR showed only moderate association (R² = 0.47; 95% CI 0.11–0.83)5. Importantly, tumor-type heterogeneity was evident: in breast cancer ADC trials specifically, ORR demonstrated strong association with OS (R² = 0.77; 95% CI 0.51–0.90)5. This finding supports PFS as the preferred endpoint for ADC accelerated approvals, but highlights that ORR's surrogate value varies substantially by cancer type despite uniform regulatory treatment.
In advanced biliary tract cancer, a comprehensive trial- and patient-level analysis of 41 studies (88 treatment arms, 7,817 patients) revealed that PFS showed moderate correlation with OS (R² = 0.71; 95% CI 0.56–0.86), while ORR demonstrated very low correlation (R² = 0.01; 95% CI 0–0.08)19. Critically, a responder analysis found no association between response and survival, indicating that achieving an objective response does not guarantee survival benefit—a fundamental disconnect between regulatory endpoints and patient outcomes19. The authors concluded that while PFS is currently the best candidate surrogate, "the need for novel endpoints in this field" is urgent19.
For immunotherapy trials, an analysis of 68 clinical trials submitted to the FDA (55 included) found moderate correlation between ORR and PFS across most immunotherapy regimens, but weaker correlation between ORR and OS, especially in combination therapies20. ORR predicted PFS better in first-line treatment but declined in later lines and remained a weak OS predictor overall20. This weak ORR-OS correlation in combination therapies suggests that regulatory reliance on ORR for immunotherapy combination drugs may systematically overestimate clinical benefit.
A Bristol Myers Squibb integrated dataset analysis demonstrated consistently stronger correlation between OS and PFS than between OS and best overall response (BOR)18. First-line therapy patients showed stronger correlations between BOR, PFS, and OS compared with second- or third-line patients, indicating that surrogate validity is compromised in heavily pretreated populations—a critical gap for patients with relapsed/refractory disease18.
Early-Stage Disease: Disease-Free Survival Complexity
In early-stage HER2-negative breast cancer, a SEER-Medicare analysis of 28,655 elderly patients (≥66 years) observed significant positive correlation between DFS and OS, with strongest correlation between distant DFS (DDFS) and OS in HR+/HER2-negative BC (correlation coefficient 0.60; 95% CI 0.57–0.62; p < 0.001) and triple-negative BC (0.69; 95% CI 0.65–0.71; p < 0.001)23. However, the authors emphasized that "additional validation in trial-level meta-analyses" is required before DFS can be broadly accepted as an OS surrogate in regulatory contexts23.
Postmarketing Requirement Gaps and Confirmatory Trial Failures
A cross-sectional analysis of 181 postmarketing requirement (PMR) statements for 161 oncology indications granted FDA accelerated approval (January 2011–July 2023) revealed critical specification gaps4. While most PMR statements specified target population (98%) and endpoints (81%), critical design parameters were frequently omitted: follow-up duration (30%), enrollment targets (26%), and double-blinding (13%)4.
PMR statements for indications that converted to regular approval differed significantly from those withdrawn: converted indications more commonly specified follow-up duration <1 year (27% vs. 0%; p < 0.001), endpoints other than OS/PFS (27% vs. 4%; p = 0.01), and specific trial/protocol (71% vs. 36%; p = 0.003)4. On-time PMR submissions were associated with less rigorous trial designs (fewer sites, less blinding) but more flexible endpoint strategies, suggesting regulatory flexibility on design parameters may facilitate timely submissions at the cost of trial rigor4.
FDA data (1992–2024) show temporal evolution: median time to conversion decreased from 4.3 to 2.3 years (p < 0.001) and median time to withdrawal decreased from 9.5 to 3.2 years (p < 0.001) when comparing 1992–2013 to 2014–2024 periods3. The proportion of indications with confirmatory studies already underway at accelerated approval increased from 63% to 85% (p = 0.003)3, indicating improved planning but also heightened regulatory oversight.
Representative Case Examples Across Tumor Types and Modalities
Multiple Myeloma – Daratumumab (D-Rd)
The MAIA trial post hoc subgroup analysis (median follow-up 64.5 months) demonstrated that daratumumab plus lenalidomide/dexamethasone significantly improved PFS versus lenalidomide/dexamethasone alone across clinically important subgroups, including aged ≥75 years (HR 0.59; 95% CI 0.44–0.79), frail patients (HR 0.64; 95% CI 0.48–0.85), and high-risk cytogenetics (HR 0.59; 95% CI 0.44–0.80)14. D-Rd demonstrated superior MRD-negativity rates (10⁻⁵) and sustained (≥12 months) MRD-negativity across subgroups, supporting its use as a standard of care for transplant-ineligible NDMM14.
Ovarian Cancer – Mirvetuximab Soravtansine (MIRV)
The PICCOLO phase II single-arm study in 79 patients with FRα-positive, third-line or greater recurrent platinum-sensitive ovarian cancer demonstrated investigator-assessed ORR of 51.9% (95% CI 40.4%–63.3%), median duration of response 8.25 months (95% CI 5.55–10.78), and median PFS 6.93 months (95% CI 5.85–9.59)12. Efficacy was maintained in patients with prior PARPi resistance (ORR 45.8% in patients with PD on/within 30 days of prior PARPi)12. This exemplifies ORR-based single-arm accelerated approval in heavily pretreated populations with limited options.
EGFR-Mutant NSCLC – Osimertinib + Ramucirumab
The RAMOSE randomized phase II trial (2:1 randomization, 147 treated, median follow-up 16.6 months) demonstrated that osimertinib plus ramucirumab significantly prolonged PFS versus osimertinib alone (median 24.8 vs. 15.6 months; HR 0.55; 95% CI 0.32–0.93; p = 0.023)13. The combination showed manageable safety (53% vs. 41% grade 3 TRAE)13, supporting dual-pathway targeting in first-line EGFR-mutant NSCLC.
Precision Medicine – Entrectinib (NTRK Fusions; ROS1-Positive NSCLC)
Entrectinib received tumor-agnostic approval for NTRK gene fusions across multiple jurisdictions (Spain, France, USA, Germany) and is approved for non-small cell lung cancer stage IV in France and Spain15. This landmark precision medicine case demonstrates how molecular alterations rather than tumor histology can drive regulatory decisions, with Phase II activity in RET gene fusion cancer, pancreatic carcinoma, and neuroblastoma15.
Strategies to Reduce Practice Gaps and Improve Patient-Centered Outcomes
Tumor-Type Specific Surrogate Validation
Establish mandatory pre-approval validation of surrogate-OS correlations using trial-level meta-analyses specific to tumor type and treatment modality. The ADC analysis demonstrating strong PFS surrogacy (R² = 0.79) but tumor-specific ORR variation exemplifies the need for context-dependent endpoint selection5.
Concurrent Confirmatory Trial Requirements
Require confirmatory trials to be randomized, ongoing, and adequately enrolled at the time of accelerated approval rather than permitting sequential initiation. The increase from 63% to 85% of trials underway at approval (1992–2013 vs. 2014–2024) shows this is feasible3, yet China's 57% confirmatory trial delay rate indicates enforcement gaps2.
MRD-Guided Treatment Personalization
Systematically integrate MRD testing into clinical practice guidelines for hematologic malignancies, with treatment intensification protocols for MRD-positive patients and de-escalation options for MRD-negative patients. Real-world evidence showing differential benefit of lenalidomide maintenance by MRD status (HR 0.24 in MRD-positive vs. no benefit in MRD-negative)9 supports personalized approaches.
Enhanced PMR Specification and Transparency
Mandate comprehensive PMR statements including specific trial identification, enrollment targets, follow-up duration, and rigorous design elements (randomization, blinding). The finding that only 45% of PMRs specified a specific trial and 13% mentioned double-blinding reveals substantial specification gaps4.
International Data-Sharing Consortia
Expand collaborative frameworks like MPAACT (for AML MRD validation)8 and i2TEAMM (for MM MRD validation)16 to other tumor types, facilitating pooled analyses of surrogate-OS correlations. Multi-indication meta-analysis methods can increase precision of target indication estimates by sharing information across indications25.
Real-World Evidence Integration
Leverage real-world data to monitor post-approval effectiveness and safety, identify surrogate-outcome disconnects, and trigger expedited regulatory review when confirmatory trials fail or are delayed. The real-world MM MRD study validating trial findings and identifying treatment-specific benefit patterns demonstrates feasibility9.
Conclusion
Accelerated and conditional approval pathways have revolutionized oncology drug development, enabling earlier patient access through surrogate endpoint acceptance. However, a 19% FDA accelerated approval withdrawal rate, systematic confirmatory trial delays (57% in China NMPA program), and heterogeneous surrogate validity across tumor types reveal substantial practice gaps. MRD demonstrates strong validation in newly diagnosed hematologic malignancies (R² = 0.79 for PFS in MM; R² = 0.78 in CLL for targeted therapies) but remains incompletely integrated into formal regulatory frameworks for AML. PFS shows moderate surrogacy in solid tumors (R² = 0.79 for ADCs; R² = 0.71 for biliary tract cancer), while ORR—used in 84% of accelerated approvals—demonstrates weak OS correlation (R² = 0.01 in biliary tract cancer; weak in immunotherapy combinations). Critical unmet needs include tumor-type–specific surrogate validation, mandatory concurrent randomized confirmatory trials, MRD-guided treatment personalization protocols, enhanced PMR specification with rigorous design elements, international data-sharing consortia expansion, and real-world evidence integration. Addressing these gaps through evidence-based strategies will strengthen the scientific foundation of accelerated pathways while maintaining their essential role in expediting access to breakthrough therapies.