SWOG/NRG S1914
In high-risk, early-stage medically inoperable or surgery-declining NSCLC, does induction, concurrent and adjuvant atezolizumab added to SBRT improve overall survival compared with SBRT alone?
Absolute Effect
NNT: Not appropriately calculable.
Approximate NNH: 12 for one additional grade ≥3 adverse event.
What the trial showed
At median follow-up of 24.8 months, overall survival was not improved with atezolizumab plus SBRT: HR 1.04 (95% CI 0.69–1.58; one-sided P=.58). The trial closed early for futility. Two grade 5 respiratory events occurred with atezolizumab.
The initial interim analysis also showed no PFS advantage; 2-year PFS was 71% with SBRT alone versus 60% with atezolizumab plus SBRT. Regional and distant failure rates were similar, while local failures were numerically higher with the combination.
Critical review: why did atezolizumab fail after SBRT?
1. The residual systemic-risk problem is different. SBRT is exceptionally effective local therapy, with local control commonly exceeding 90%. After successful ablation of a small N0 tumour, the incremental benefit available to systemic immunotherapy depends on how much clinically meaningful micrometastatic disease remains. In many early-stage patients that burden may be too small to generate a measurable OS advantage. Stage III NSCLC and limited-stage SCLC have substantially greater systemic relapse risk, creating a much larger target for consolidation therapy.
2. Chemoradiotherapy is immunologically different from SBRT. Concurrent chemotherapy can contribute immunogenic cell death, antigen release, inflammatory signalling and modulation of the tumour microenvironment. Definitive CRT also treats a larger thoracic/regional disease field. SBRT is highly focal and deliberately ablative. The attractive laboratory concept of “radiation releases antigen, therefore checkpoint blockade should amplify it” is therefore incomplete: the magnitude, distribution and timing of the immune stimulus matter.
3. S1914 cannot be dismissed as a sequencing failure. Atezolizumab was deliberately started before SBRT and continued during and after it, based on preclinical reasoning that priming before radiation might be superior. Thus, the negative result is not simply evidence that post-SBRT immunotherapy was started too late.
4. Patient selection is fundamentally different. S1914 enrolled a predominantly elderly, medically inoperable population. Competing mortality and treatment vulnerability reduce the achievable OS benefit of an additional systemic therapy. PACIFIC and ADRIATIC enrolled fitter patients who had already tolerated definitive chemoradiotherapy and had no progression, enriching the population for patients able to receive prolonged consolidation therapy.
5. The biology of early-stage NSCLC is heterogeneous. Not all high-risk clinical features identify an immunotherapy-responsive tumour. Driver alterations such as EGFR can be associated with limited benefit from checkpoint blockade, while PD-L1, tumour immune phenotype and T-cell functionality may matter. S1914 was not designed as a biomarker-selected study, so a responsive subgroup could theoretically exist even though the unselected population did not benefit.
6. PACIFIC itself teaches that “more immunotherapy around radiation” is not automatically better. In PACIFIC-2, adding durvalumab concurrently with definitive CRT and then continuing consolidation durvalumab failed to improve PFS or OS compared with CRT followed by consolidation placebo. This reinforces the broader lesson that immunotherapy efficacy depends on treatment context and sequencing, not simply on adding an ICI to radiation.
Bottom line: S1914 does not show that PD-L1 blockade cannot cooperate with radiation. It shows that the PACIFIC/ADRIATIC paradigm cannot be extrapolated to SBRT simply because both strategies contain radiation and a checkpoint inhibitor. The most plausible explanation is a combination of lower residual systemic disease burden after SBRT, different immune priming from SBRT versus chemoradiotherapy, an older/frailer patient population, competing mortality, and lack of biomarker selection. The appropriate next step is not routine atezolizumab after SBRT, but better selection of patients whose biological and systemic-risk profile creates enough absolute room for benefit to justify added toxicity.
Why PACIFIC and ADRIATIC are different
PACIFIC: After definitive concurrent chemoradiotherapy for unresectable stage III NSCLC, durvalumab improved median PFS from 5.6 to 16.8 months and median OS from 29.1 to 47.5 months; 5-year OS was 42.9% versus 33.4%.
ADRIATIC: After concurrent chemoradiotherapy for limited-stage SCLC, durvalumab improved median OS from 33.4 to 55.9 months (HR 0.73) and median PFS from 9.2 to 16.6 months (HR 0.76). Grade 3–4 adverse-event rates were similar, 24.4% versus 24.2%.
The contrast is therefore not “durvalumab works after CRT but atezolizumab does not after SBRT.” It is more accurately: consolidation ICI has proven value when the preceding definitive therapy leaves a substantial systemic relapse problem in a selected population; S1914 shows that this premise cannot be assumed after SBRT for early-stage N0 disease.