Article Text

Download PDFPDF

173 Sensitivity and concordance of CD274 expression by RNA sequencing (RNA-seq) in comparison with three PD-L1 immunohistochemistry methods in head and neck squamous cell carcinoma (HNSCC)
  1. Mary Nesline1,
  2. Sarabjot Pabla1,
  3. Jeffrey Conroy1,
  4. Paul DePietro1,
  5. Shengle Zhang1,
  6. Roger Klein1,
  7. Achyut Bhagelu2,
  8. Rebecca Previs2,
  9. Prasanth Reddy2,
  10. Shakti Ramkissoon2 and
  11. Eric Severson2
  1. 1OmniSeq, Inc., Buffalo, NY, USA
  2. 2Labcorp, Durham, NC, USA


Background PD-L1 expression by immunohistochemistry (IHC) is associated with HNSCC immunotherapy response.1 The performance of different PD-L1 IHC clones has shown variability and poor concordance for immune vs. tumor cell scoring in HNSCC. Crucially, this leads to poor reproducibility in the combined positive score (CPS) method by the PD-L1 IHC 22C3 companion diagnostic.2 We explored the clinical sensitivity and concordance of CD274 (PD-L1) expression by RNA-sequencing compared to three PD-L1 IHC methods.

Methods A retrospective cohort of HNSCC patients (n=258) with FFPE tissue was tested by comprehensive immune profiling3(2017–2022), including CD274 by RNA-seq (normalized percentile rank 0–100). IHC was performed with either the 28–8 or 22C3 PD-L1 clones. 28–8 was scored with% tumor cells stained (TC, n=34), while 22C3 was scored with either tumor proportion score (TPS, n=61) or combined positive score (CPS, n=163). For 22C3, CPS≥1 is low positive, and ≥20 is high positive. For 28–8 TC and 22C3 TPS, ≥1 is low positive, and ≥50 is high positive. ROC models for each IHC method were constructed for 5 sets of patients with different pairwise interpretation groups and used to determine RNA-seq cutoffs based on individual PD-L1 IHC scoring methods and accuracy at those cutoffs. Concordance between standard IHC scoring methods and CD274 by RNA-seq was also assessed.

Results PD-L1 IHC results varied depending on the clone and scoring method used. Not surprisingly, CPS had the fewest negative cases (2.7%) and most high cases (47.2%). Tumor cell scoring by TPS (29.5% negative, 24.6% high) and TC (17.6% negative, 26.5% high) was similar for high vs. not high. For all three IHC approaches, PD-L1 RNA-seq classified IHC high v negative, high v low, and high v not high status with at least fair range of AUC (0.758–0.981), sensitivity (0.636–1.00), and specificity (0.785–1.00). RNA-seq could not discern between IHC low v negative status for any method. Pairwise comparisons showed significant concordance between median RNA-seq percentile ranks for IHC high v low and high v negative status for all IHC methods. By RNA-seq, the frequency of PD-L1 high cases for each scoring method increased from 47% to 55% (CPS), 24.6% to 42.6% (TPS) and 26.5% to 47.1% (TC) based on modeled RNA-seq cutoffs.

Conclusions RNA-seq accurately discerns PD-L1 high vs. not high HNSCC tumors based on IHC scoring methods and may more reliably identify patients for frontline immunotherapy.


  1. NCCN, ‘NCCN Clinical Practice Guidelines in Oncology (NCCN Guidelines®) Head and Neck Cancers, Version 2.2022,’ National Comprehensive Cancer Network, Fort Washington, Pennsylvania, Version 2.2022, Mar.

  2. J. Ribbat-Idel et al., Performance of Different Diagnostic PD-L1 Clones in Head and Neck Squamous Cell Carcinoma, Front. Med., vol. 8, no. April, pp. 1–8, 2021.

  3. J. M. Conroy et al. Analytical Validation of a Next-Generation Sequencing Assay to Monitor Immune Responses in Solid Tumors, J. Mol. Diagnostics, vol. 20, no. 1, pp. 95–109, Jan. 2018.

Statistics from

Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.