Perioperative immunotherapy has reshaped resectable NSCLC. But as neoadjuvant regimens produce high rates of pathologic complete response, a harder question is emerging: does every patient who responds completely need the full year of adjuvant checkpoint inhibition that follows? The evidence driving label approvals was built on trial populations — not pCR subgroups. That gap is where current practice quietly breaks.
What’s changing
CheckMate-77T and similar perioperative designs have established neoadjuvant chemo-IO followed by adjuvant PD-1 inhibition as a near-standard framework in resectable NSCLC. Gains in pCR and event-free survival have been compelling enough to drive broad adoption. The perioperative year — neoadjuvant cycles, surgery, adjuvant IO — has become a monolithic pathway. What’s changing is the biology of patients moving through it. A meaningful subset are arriving at surgery with no residual viable tumor. For these patients, the clinical question shifts. The pathway doesn’t.
Where practice quietly breaks
The perioperative regimen was designed around the full trial population, and the regulatory label reflects that. When a patient achieves pCR, clinical teams face a structurally constrained choice: follow the label, or deviate in a way that guidelines and trial schemas don’t currently support. There are no validated de-escalation frameworks based on depth of pathologic response. That absence forces maximum systemic exposure by default — regardless of what pathology showed. Standard tumor boards have no decision node that asks: does this pCR change the adjuvant calculus?
Decision tension: the clinical scenario
A 58-year-old patient with stage IIIA NSCLC — no actionable driver mutation, PD-L1 50% — completes four cycles of neoadjuvant chemo-IO without significant toxicity and proceeds to lobectomy with R0 resection. Pathology returns with a pathologic complete response: no viable tumor in the primary specimen or any sampled lymph node. The patient returns to clinic fatigued but recovering. They have a family history of autoimmune thyroid disease and ask directly: “My doctor said there’s no cancer left. Why do I need another year of treatment?” The clinical team has no clean answer. The label supports adjuvant nivolumab. The protocol expects it. But the patient’s biology — as far as pathology can show — no longer contains measurable disease. The team proceeds with adjuvant IO not because the evidence in this specific patient is clear, but because deviating feels unsupported.
What’s not being said
The field is more comfortable discussing under-treatment in curative settings than overtreatment. The implicit assumption behind completing the full perioperative year is that more is safer — that continuing IO post-surgery is at worst neutral. That assumption hasn’t been tested. Endocrine toxicities — thyroid dysfunction, adrenal insufficiency, hypophysitis — can be permanent. Immune-related pneumonitis carries real morbidity in patients who just had a portion of their lung removed. The system is structurally biased toward completion. A clinician who abbreviates adjuvant IO in a pCR patient outside a trial is deviating from the label. A clinician who completes the full course is following the protocol — regardless of whether the biology still warrants it.
Where the data are unclear
Overall survival data from perioperative NSCLC trials remain immature. Event-free survival and pCR serve as primary or co-primary endpoints — meaningful surrogates, but not answers to whether adjuvant IO specifically, in a pCR patient, improves long-term survival. Crossover and salvage therapy rates in both trial and real-world settings further obscure the survival contribution of the adjuvant phase alone. There are no published randomized data asking, in a pCR-selected NSCLC population, whether completing full-course adjuvant IO improves outcomes versus observation or a shortened course. Until those data exist, the recommendation to complete a full year of adjuvant IO in pCR patients is extrapolated — not validated — in the subgroup for whom the question matters most.
Deeper implications
The adjuvant trap is a systems issue. Pathways without response-adapted decision points push teams toward maximum exposure by default. Tumor boards are not equipped to make de-escalation recommendations when no guideline supports them. As perioperative IO expands to earlier stages and broader populations, the proportion of patients achieving pCR will grow — and so will the aggregate toxicity burden of continuing adjuvant IO in patients who may not need it. The efficiency argument is real too: healthcare systems absorb the cost and visit burden of a full adjuvant year without confirmed benefit data in the patients most likely to be cured by surgery alone. The conversation about response-adapted de-escalation needs to happen before the next generation of perioperative trials reads out.
Decision Snapshot: The Adjuvant Trap
- Protocol adherence overriding dynamic, patient-specific pathologic response signals
- Adjuvant benefit assumed in pCR patients without subgroup-specific randomized data
- Lifelong toxicity risk — endocrine, pulmonary — accepted for unproven incremental survival
- Regulatory labels and trial schemas locking in fixed treatment durations regardless of pathologic outcome
- Absence of response-adapted de-escalation frameworks defaulting all patients to maximum exposure
- Systems bias toward completion — deviating from the label feels unsupported even when clinical logic is sound
A question worth carrying: when the pathology says complete response, are we still treating the patient — or finishing the protocol?