Interobserver reproducibility and accuracy of p16/ K i‐67 dual‐stain cytology in cervical cancer screening

BACKGROUND Dual‐stain cytology for p16 and Ki‐67 has been proposed as a biomarker in cervical cancer screening. The authors evaluated the reproducibility and accuracy of dual‐stain cytology among 10 newly trained evaluators. METHODS In total, 480 p16/Ki‐67–stained slides from human papillomavirus‐positive women were evaluated in masked fashion by 10 evaluators. None of the evaluators had previous experience with p16 or p16/Ki‐67 cytology. All participants underwent p16/Ki‐67 training and subsequent proficiency testing. Reproducibility of dual‐stain cytology was measured using the percentage agreement, individual and aggregate κ values, as well as McNemar statistics. Clinical performance for the detection of cervical intraepithelial neoplasia grade 2 or greater (CIN2+) was evaluated for each individual evaluator and for all evaluators combined compared with the reference evaluation by a cytotechnologist who had extensive experience with dual‐stain cytology. RESULTS The percentage agreement of individual evaluators with the reference evaluation ranged from 83% to 91%, and the κ values ranged from 0.65 to 0.81. The combined κ value was 0.71 for all evaluators and 0.73 for cytotechnologists. The average sensitivity and specificity for the detection of CIN2+ among novice evaluators was 82% and 64%, respectively; whereas the reference evaluation had 84% sensitivity and 63% specificity, respectively. Agreement on dual‐stain positivity increased with greater numbers of p16/Ki‐67–positive cells on the slides. CONCLUSIONS Good to excellent reproducibility of p16/Ki‐67 dual‐stain cytology was observed with almost identical clinical performance of novice evaluators compared with reference evaluations. The current findings suggest that p16/Ki‐67 dual‐stain evaluation can be implemented in routine cytology practice with limited training. Cancer (Cancer Cytopathol) 2014;122:914–920. © 2014 American Cancer Society .