p16 immunohistochemistry in oropharyngeal squamous cell carcinoma: a comparison of antibody clones using patient outcomes and high-risk human papillomavirus RNA status.

High-risk human papillomavirus (HPV)-related oropharyngeal squamous cell carcinomas have a more favorable prognosis than HPV-negative ones. p16 immunohistochemistry has been recommended as a prognostic test in clinical practice. Several p16 antibodies are available, and their performance has not been directly compared. We evaluated three commercially available p16 antibody clones (E6H4, JC8 and G175-405) utilizing 199 cases of oropharyngeal squamous cell carcinoma from a tissue microarray, read by three pathologists with three different cutoffs for positivity: any staining, >50% and >75%. Positive predictive values for high-risk HPV status by RNA in situ hybridization for the E6H4, JC8 and G175-405 clones were 98%, 100% and 99% at the 75% cutoff, but negative predictive values were much more variable at 86%, 69% and 56%, respectively. These improved using the 50% cutoff, becoming similar for all three antibodies. Intensity varied substantially, with 85% of E6H4, 72% of JC8 and 67% of G175-405 showing strong (3+) intensity. With Kaplan-Meier survival plots at the 75% cutoff, the E6H4 clone showed the largest differential in disease specific and overall survival between p16-positive and -negative results. Decreasing the cutoff to 50% increased correlation with HPV in situ hybridization and improved the survival differential for the JC8 and G175-405 clones without worsening of performance for the E6H4 clone. Interobserver agreement was also assessed by kappa scores and was highest for the E6H4 clone. Overall, these study results show modest but important performance differences between the three different p16 antibody clones, suggesting that the E6H4 clone performs best because of strongest staining intensity, greatest differential in outcomes between positive and negative results, lowest interobserver variability, and lowest background, nonspecific staining. The results also suggest that a 75% cutoff is very functional but that, in this patient population with high HPV incidence, 50% and any staining cutoffs may be more effective, particularly for the non-E6H4 clones.