Pembrolizumab Adds Toxicity Without Extending Event-Free Survival in Early-Stage NSCLC
Pembrolizumab did not improve event-free survival when added to stereotactic body radiotherapy for unresected stage I or II non-small cell lung cancer, and the combination produced substantially more treatment-related harm than radiation alone. The phase III KEYNOTE-867 study was halted early after the benefit-risk profile turned unfavorable, ending what was meant to be a definitive test of immunotherapy layered onto curative-intent radiation in a setting where SBRT already performs well as a local treatment backbone.
The trial enrolled adults with untreated stage I or II disease, including T1 to limited T3, N0, M0 tumors that were inoperable or in patients who refused surgery. Participants were randomized 1:1 to thoracic SBRT over 2 weeks or less with pembrolizumab 200 mg every 3 weeks for up to 17 cycles, or to SBRT with placebo on the same schedule. The primary endpoint was event-free survival assessed by blinded independent central review, a design intended to capture not just local control but the broader consequence of adding systemic therapy to a disease state that is often managed definitively with local treatment alone.
The result was numerically close, but not clinically persuasive. Median EFS was 31.2 months with pembrolizumab and 28.3 months with placebo, yielding a hazard ratio of 0.92 and a P value of .2932. In practical terms, the immune checkpoint inhibitor did not shift the survival curve in a meaningful way. For a population already receiving a highly effective local therapy, that matters because any additional systemic intervention must earn its place by improving outcomes enough to justify the added risk, cost, and treatment burden.
The safety findings moved in the opposite direction. Among participants who received at least one dose of study treatment, grade 3 to 5 treatment-related adverse events occurred in 20.4% of those assigned to pembrolizumab and 3.7% of those assigned to placebo. Five treatment-related deaths occurred, all in the pembrolizumab arm. That pattern is difficult to ignore in early-stage lung cancer, where many patients may otherwise remain free of major treatment toxicity after SBRT alone. Immunotherapy can amplify pulmonary and systemic inflammatory injury, and when paired with thoracic radiation the concern is not abstract: the lung is the target organ, and immune activation in irradiated tissue can cross from theoretical synergy into clinically meaningful harm.
The median time from random assignment to the data cutoff for the first interim analysis was 20.6 months, with follow-up ranging from 1.3 to 56.8 months. Even with that follow-up window, no efficacy signal emerged strong enough to offset the toxicity burden. The trial was stopped prematurely, and all study treatment was discontinued. In a field that has been searching for ways to intensify cure-oriented therapy without compromising tolerability, this result redraws the boundary between biologic rationale and bedside utility.
Why the Biology Did Not Translate Into Better Outcomes
The appeal of combining pembrolizumab with SBRT was straightforward. Radiation can release tumor antigens and alter the local immune microenvironment, creating a plausible platform for checkpoint blockade to extend control beyond the irradiated field. In early-stage NSCLC, where microscopic disease may still seed recurrence, the concept of systemic immune priming seems especially attractive. Yet biologic plausibility does not guarantee clinical benefit, and this study shows how quickly a promising mechanistic model can fail once it is tested in patients who already have a strong standard treatment.
Several factors likely work against easy synergy in this setting. SBRT delivers ablative local control, so the incremental room for improvement is limited from the start. Patients with stage I or II unresected disease also represent a heterogeneous population, including individuals who are medically inoperable and others who decline surgery, so the competing risks and baseline prognosis vary widely. Against that backdrop, a checkpoint inhibitor must produce a clear reduction in recurrence or death to justify its toxicity profile. Here, the benefit was not large enough to separate from chance, while the safety signal was immediate and clinically consequential.
The early-stop decision reflects that imbalance. When a trial in potentially curable disease exposes patients to more serious toxicity without improving the primary endpoint, continued treatment becomes difficult to defend. Five treatment-related deaths in the pembrolizumab arm alter the interpretation from a simple negative study to a cautionary one. Thoracic immunotherapy combinations remain attractive in advanced disease, where the potential upside is larger and the disease burden is greater. In unresected stage I or II NSCLC, the threshold for acceptable harm is much lower because SBRT alone is already a standard of care and many patients can be treated effectively without systemic therapy.
The clinical implication is narrow but important: adding pembrolizumab to SBRT should not be assumed to improve outcomes in early-stage unresected NSCLC, and the combination introduced measurable excess risk in this randomized phase III trial. The absence of an EFS advantage was paired with a clear increase in severe treatment-related adverse events, including deaths, which is the wrong direction for a regimen intended for curative-intent management.
The most relevant unanswered issue is not whether immunotherapy can ever complement radiation in lung cancer, but which disease states, schedules, and biomarkers might identify a group that can benefit without inheriting this level of toxicity. KEYNOTE-867 provides a direct answer for the regimen tested here: concurrent pembrolizumab with SBRT did not improve event-free survival and carried a less favorable safety profile than SBRT alone.