Highlights
– A pooled individual‑patient analysis of 11 GBG and AGO‑B randomized neoadjuvant trials (15 comparisons, 12,247 patients) shows strong trial‑level surrogacy of distant disease‑free survival (DDFS) for overall survival (OS) (R2trial = 0.91; 95% CI 0.82–1.00).
– Surrogacy was consistent across most clinical subgroups but varied by molecular subtype: strong in HR‑/HER2‑ and HR‑/HER2+ tumours, weak in HR+/HER2‑ and HR+/HER2+ tumours.
– Implications: DDFS can be a primary efficacy endpoint to reliably predict OS in many neoadjuvant RCTs, but careful subtype‑specific planning and further validation with contemporary therapies are needed.
Background: why surrogate endpoints matter in neoadjuvant breast cancer trials
Neoadjuvant systemic therapy is standard for many patients with operable early breast cancer and offers the opportunity to assess response in vivo and to tailor postoperative therapy. Randomized controlled trials (RCTs) in the neoadjuvant setting increasingly use intermediate endpoints to accelerate readout and reduce trial size. Pathologic complete response (pCR) is an early, patient‑level biomarker correlated with favourable outcomes in some subgroups, but it has limited trial‑level surrogacy for long‑term outcomes in many settings.
Distant disease‑free survival (DDFS) — time to distant recurrence or death — is a clinically relevant endpoint that reflects metastatic relapse and is closely tied to mortality. Validating DDFS as a surrogate for overall survival (OS) is important for trial design, regulatory decision‑making, and clinical interpretation of neoadjuvant RCTs.
Study design and methods
This pooled analysis used individual patient data from randomized neoadjuvant trials conducted by the German Breast Group (GBG) and the AGO‑B Study Group. Included were 11 RCTs providing 15 randomized treatment comparisons and 12,247 patients with available DDFS and OS data. The investigators applied two‑stage meta‑analytic copula methods to derive copula model‑based hazard ratios (HRs) for both endpoints and to evaluate the trial‑level association between treatment effects on DDFS and OS.
Trial‑level surrogacy was quantified using R2trial from the ReSEEM guideline thresholds: R2trial ≥ 0.70 = strong surrogacy; 0.50–0.69 = moderate; < 0.50 = weak. Subgroup analyses were prespecified for clinical and pathological features, including molecular subtype (combinations of hormone receptor [HR] and HER2 status). Heterogeneity across subgroups was formally tested.
Key findings
Overall surrogacy
– There was a very strong trial‑level association between treatment effects on DDFS and OS: R2trial = 0.91 (95% CI 0.82–1.00). This indicates that treatment‑induced changes in DDFS explained most of the variability in treatment effects on OS across included comparisons.
Subgroup findings and heterogeneity
– No significant heterogeneity of surrogacy was observed across most clinical and pathological subgroups (pheterogeneity > 0.05 for the majority of comparisons).
– Exceptions were molecular subtype–defined subgroups, where surrogacy varied significantly (pheterogeneity = 0.021):
– HR‑/HER2‑ (triple‑negative by HR/HER2 definition) tumours: R2trial = 0.89 (95% CI 0.75–1.00) — strong surrogacy.
– HR‑/HER2+ tumours: R2trial = 0.73 (95% CI 0.36–1.00) — strong surrogacy, though confidence bounds are wider.
– HR+/HER2‑ tumours: R2trial = 0.33 (95% CI 0.00–0.83) — weak surrogacy.
– HR+/HER2+ tumours: R2trial = 0.11 (95% CI 0.00–0.55) — weak surrogacy.
Interpretation of subgroup results
– In more proliferative, higher‑risk subtypes (HR‑ and/or HER2+), DDFS trackingly reflects OS benefit at the trial level — likely because distant relapse in these subtypes occurs earlier and contributes substantially to mortality within typical trial follow‑up intervals.
– In HR+ disease, particularly HR+/HER2‑ tumours, long natural histories, effective salvage and subsequent therapies, and late recurrences may decouple earlier gains in DDFS from OS within available follow‑up, reducing trial‑level surrogacy.
Other notable points
– The pooled analysis used copula models to account for correlation structure between endpoints and incorporated individual‑patient time‑to‑event data, strengthening the robustness of estimates compared with aggregate data approaches.
Clinical and research implications
For trialists and regulators
– For many contemporary neoadjuvant RCTs, particularly those enrolling patients with HR‑ or HER2+ disease, using DDFS as a primary endpoint appears justified to predict OS‑level treatment benefit, potentially allowing shorter trials and earlier decisions on efficacy.
– For HR+ subtypes, however, trials that rely solely on DDFS to infer OS benefit should be cautious. Trial designs may require longer follow‑up for OS, larger event numbers, or the use of alternative strategies (co‑primary endpoints, adaptive designs) to ensure clinically meaningful conclusions.
For clinicians
– When interpreting neoadjuvant trials that report DDFS as the primary long‑term outcome, clinicians should consider tumour biology: strong DDFS→OS surrogacy in HR‑ and HER2+ disease supports translating DDFS benefit into a likely OS advantage, while in HR+ disease such extrapolation is less certain.
For researchers
– The findings underscore the need to validate surrogate relationships in modern trial series that incorporate current systemic agents (anti‑HER2 dual blockade, T‑DM1, CDK4/6 inhibitors, immune checkpoint inhibitors) and patient populations enriched for high‑risk features.
Strengths and limitations
Strengths
– Individual patient data meta‑analysis allows consistent endpoint definitions, time‑to‑event modelling, and flexible subgroup exploration.
– Use of copula methodologies and formal trial‑level surrogacy metrics (R2trial) adheres to contemporary statistical standards for surrogate validation.
Limitations
– The pooled trials span eras with shifting standard therapies and patient selection; older comparator regimens and fewer contemporary targeted agents may limit applicability to current practice.
– Some molecular subtype subgroup estimates have wide confidence intervals; statistical power is limited for less frequent subtype/treatment combinations, which affects precision.
– The copula approach assumes that the bivariate dependence structure is appropriately modelled; departures from model assumptions (non‑proportional hazards, changing salvage therapies) can influence results.
– Because event timing differs by subtype (late recurrences in HR+ disease), follow‑up duration needed for reliable OS capture may exceed that available in some trials, reducing observed surrogacy.
How this fits with prior evidence
Previous pooled analyses in the neoadjuvant and adjuvant settings have produced mixed results regarding the trial‑level validity of early endpoints. The CTNeoBC pooled analysis showed that pCR is prognostic at the patient level and predictive in certain subgroups but not a consistent trial‑level surrogate for OS across all subtypes (Cortazar et al., Lancet 2014). The current GBG/AGO‑B pooled analysis complements that literature by demonstrating strong trial‑level surrogacy for DDFS overall, while refining the picture by identifying subtype‑specific limitations.
Next steps and research priorities
– Validate DDFS→OS surrogacy in contemporary RCT datasets that include modern systemic therapies (dual HER2 blockade, antibody‑drug conjugates, CDK4/6 inhibitors, immunotherapy).
– Explore alternative or composite intermediate endpoints (e.g., invasive disease‑free survival, event‑free survival) in subtype‑specific contexts.
– Develop statistical methods and trial designs that explicitly account for subtype‑dependent recurrence timing and the influence of effective salvage therapies on OS.
– Consider coordinated meta‑analyses across international trial groups to increase power for rarer subtype/treatment combinations.
Conclusions
This pooled individual‑patient analysis from GBG and AGO‑B demonstrates that, with adequate follow‑up, distant disease‑free survival is a robust trial‑level surrogate for overall survival in neoadjuvant randomized trials of early breast cancer in most contexts. However, surrogacy is substantially weaker for hormone receptor‑positive subtypes — especially HR+/HER2‑ disease — where late recurrences and effective subsequent therapies may dilute a near‑term DDFS signal for OS. Trial designers, regulators, and clinicians should therefore apply DDFS as a primary outcome with attention to tumour biology, trial follow‑up, and evolving therapeutic landscapes.
Funding and trial registration
The pooled analysis reported no external funding. Individual trials included in the pooled dataset had varied sponsorship and registration; investigators seeking details should consult the primary trial publications and registries.
References
1. Conforti F, Holtschmidt J, Nekljudova V, et al. Distant disease‑free survival as a surrogate endpoint for overall survival in randomised trials of neoadjuvant therapy for early breast cancer: a pooled analysis of GBG and AGO‑B Study Group trials. Lancet Oncol. 2025;26(12):1584–1597. doi:10.1016/S1470‑2045(25)00546‑7.
2. Cortazar P, Zhang L, Untch M, et al.; CTNeoBC Investigators. Pathological complete response and long‑term clinical benefit in breast cancer: the CTNeoBC pooled analysis. Lancet. 2014;384(9938):164–172. doi:10.1016/S0140‑6736(13)62422‑8.
3. Burzykowski T, Molenberghs G, Buyse M. The Evaluation of Surrogate Endpoints. Springer; 2005. ISBN: 978‑0387248078.
4. Prentice RL. Surrogate endpoints in clinical trials: definition and operational criteria. Stat Med. 1989;8(4):431–440. doi:10.1002/sim.4780080407
