KEYNOTE-689

In resectable locally advanced HNSCC, does perioperative pembrolizumab added to surgery and risk-adapted postoperative therapy improve event-free survival?

Evidence snapshot updated 23 October 2026. Source links are provided below; this is an evidence summary, not a substitute for current local guidelines or individual clinical judgement.

Population and treatment

714 participants with resectable locally advanced head and neck squamous-cell carcinoma; prespecified efficacy populations included PD-L1 CPS ≥10, CPS ≥1 and the overall population.

Intervention and comparator: Pembrolizumab 200 mg IV every 3 weeks for 2 neoadjuvant cycles and 15 adjuvant cycles, added to surgery and postoperative radiotherapy with or without concurrent cisplatin, versus standard care alone.

ClinicalTrials.gov: NCT03765918 ↗

Key results

At 36 months, EFS was 57.6% vs 46.4% in the overall population; HR 0.73 (95% CI 0.58–0.92; P=.0041). The EFS benefit was also observed in the prespecified CPS ≥1 and CPS ≥10 populations.

3-year event-free survival · 36 months

Absolute-effect interpretation: Absolute 36-month EFS difference 11.2 percentage points; approximate NNT 9 over 36 months based on reported fixed-time EFS estimates. Interpret in the context of the trial's sequential testing plan and PD-L1-defined populations.

Toxicity and treatment burden

Immune-related adverse events must be weighed against surgery, radiotherapy and possible cisplatin toxicity. The regimen adds a prolonged adjuvant treatment course; check discontinuation and postoperative complication details in the full report.

Interpretation and limitations

The trial tests adding perioperative immunotherapy to surgery-based standard care, not replacing indicated postoperative radiotherapy or cisplatin. Use the actual CPS eligibility and treatment sequence; avoid extrapolating to unresectable disease or definitive chemoradiation without surgery.

Use endpoint definitions, follow-up, analysis population and comparator exactly as reported. Do not infer NNT from a hazard ratio or median survival. Subgroup findings should not be treated as predictive unless the trial design and interaction testing support that interpretation.

Primary source and further reading

Primary publication / authoritative source ↗

Trial registry (NCT03765918) ↗

Automated structural checks can confirm the presence of links and outcome-visual markup; browser-rendered QA and complete source-by-source clinical freshness review remain separate quality-control steps.