Pathological response and exploratory assessment of pretreatment PD-L1 combined positive score after four cycles of neoadjuvant pembrolizumab plus chemotherapy in locally advanced head and neck squamous cell carcinoma: a retrospective cohort study

Technical Summary

Problem Statement
In locally advanced head and neck squamous cell carcinoma (LA-HNSCC), neoadjuvant immunochemotherapy has shown potential to induce pathological regression prior to surgery. However, it remains uncertain whether the pretreatment Programmed Death-Ligand 1 (PD-L1) Combined Positive Score (CPS) can reliably identify patients who will respond to this regimen. While CPS is a validated biomarker for recurrent or metastatic disease, its utility in the neoadjuvant setting is inconsistent, and small cohorts often suffer from unstable cutoffs and overfitting. Furthermore, distinguishing between primary tumor response and nodal response is critical, as a complete response in the primary site does not guarantee the eradication of regional disease.

Methodology
This study is a single-center, retrospective cohort analysis of 31 consecutive, treatment-naive patients with resectable stage III–IVA LA-HNSCC (all p16-negative) treated between September 2022 and June 2024 at Tianjin Cancer Hospital Airport Hospital.

• Intervention: All patients received four 3-week cycles of neoadjuvant therapy consisting of pembrolizumab (200 mg), nab-paclitaxel (260 mg/m²), and cisplatin (75 mg/m²), followed by definitive surgery 3–6 weeks later.
• Biomarker Assessment: Pretreatment biopsies were analyzed using PD-L1 IHC 22C3 pharmDx. CPS was calculated as a continuous variable, with a prespecified cutoff of 10 (CPS < 10 vs. CPS ≥ 10) used for group comparisons to avoid data-derived bias.
• Pathological Evaluation: Two blinded pathologists independently assessed resection specimens for Residual Viable Tumor (RVT).
  • Primary Endpoint: Primary-tumor Major Pathological Response (MPR), defined as RVT ≤ 10% (including Pathological Complete Response, pCR).
  • Secondary Endpoints: Primary-tumor pCR, nodal MPR/pCR (in clinically node-positive patients), and patient-level overall MPR/pCR (requiring response in both primary and nodal compartments).
• Statistical Approach: The study utilized Fisher's exact test for categorical comparisons and Mann–Whitney U tests for age. Exploratory Receiver Operating Characteristic (ROC) analysis with 30,000 bootstrap resamples was performed to assess the discriminatory power of continuous CPS. No multivariable modeling was attempted due to the small sample size.

Key Results

• Pathological Response:
  • Primary Tumor: The primary endpoint of primary-tumor MPR was achieved in 22/31 patients (71.0%). Primary-tumor pCR was observed in 15/31 patients (48.4%).
  • Patient-Level: When considering both primary and nodal compartments, patient-level overall MPR was 18/31 (58.1%), and overall pCR was 9/31 (29.0%).
  • Nodal Response: Among the 14 patients with baseline clinical nodal involvement, nodal MPR was achieved in only 5/14 (35.7%), and nodal pCR in 2/14 (14.3%).
• Biomarker Association (CPS):
  • There was no statistically significant association between pretreatment CPS and pathological response.
  • Primary-tumor MPR rates were 66.7% for CPS < 10 and 72.7% for CPS ≥ 10 (Odds Ratio [OR] 1.33; 95% CI 0.25–7.11; P = 1.000).
  • Continuous CPS demonstrated limited discrimination for primary-tumor MPR, with an Area Under the Curve (AUC) of 0.525 (95% CI 0.277–0.767), performing little better than chance.
  • At the CPS 10 cutoff, specificity was low (33.3%), and the Youden index for overall MPR was negative (−0.103).
• Safety:
  • Any-grade adverse events occurred in 71.0% of patients; grade ≥ 3 events occurred in 12.9%.
  • Common events included nausea/vomiting and myelosuppression. One patient experienced grade 3 hypophysitis, and one experienced grade 2 pulmonary embolism. There were no treatment-related deaths.

Significance and Claims
The paper concludes that a four-cycle neoadjuvant regimen of pembrolizumab plus chemotherapy demonstrates substantial pathological activity in a p16-negative LA-HNSCC cohort, achieving high rates of primary-tumor regression. However, the study explicitly states that pretreatment CPS is not associated with response and should not be used as a standalone tool for patient selection without further validation.

The authors emphasize that the divergence between primary-tumor response and patient-level (nodal) response highlights the necessity of evaluating both compartments separately. The study positions its findings as hypothesis-generating rather than definitive, noting that the small sample size and retrospective nature limit the precision of the biomarker analysis. The authors argue that while CPS remains a candidate component of a broader biomarker strategy, its use as a clinical cutoff requires validation in larger, prospective, and independently validated cohorts. The paper does not propose new applications or future experimental protocols but calls for rigorous external validation before CPS can be integrated into neoadjuvant selection criteria.