Interobserver reproducibility of cytologic p16INK4a/Ki‐67 dual immunostaining in human papillomavirus‐positive women

• Published in: Cancer cytopathology 2017-03, Vol.125 (3), p.212-220
• Main Authors: Benevolo, Maria, Allia, Elena, Gustinucci, Daniela, Rollo, Francesca, Bulletti, Simonetta, Cesarini, Elena, Passamonti, Basilio, Giovagnoli, Maria Rosaria, Carico, Elisabetta, Carozzi, Francesca M., Mongia, Alessandra, Fantacci, Giulia, Confortini, Massimo, Rubino, Teresa, Fodero, Cristina, Prandi, Sonia, Marchi, Natalina, Farruggio, Angelo, Coccia, Anna, Macrì, Luigia, Ghiringhello, Bruno, Ronco, Guglielmo, Bragantini, Emma, Polla, Enzo, Maccallini, Vincenzo, Negri, Giovanni, Giorgi Rossi, Paolo
• Format: Article
• Language: English
• Subjects: cervical cancer; cyclin‐dependent kinase inhibitor 2A (p16ink4a)/Ki‐67; dual immunostaining; human papillomavirus; inter‐rater agreement
• ISSN: 1934-662X; 1934-6638
• DOI: 10.1002/cncy.21800

BACKGROUND The accumulation of cyclin‐dependent kinase inhibitor 2A (p16ink4a) protein in a cell is associated with neoplastic progression in precancerous cervical lesions. Dual staining for p16ink4a and Ki‐67 has been proposed as a triage test in cervical cancer screening for women who test positive for human papillomavirus DNA. In this study, interobserver reproducibility of the interpretation of this test was assessed. METHODS Forty‐two immunostained, liquid‐based cytology slides were divided into 2 sets and were interpreted by 17 to 21 readers from 9 different laboratories, yielding a total of 816 reports. Immunostaining results were classified as positive, negative, inconclusive, or inadequate. After evaluation of the first set of slides and before circulation of the second set, the results were discussed in a plenary meeting. The 10 slides with the most discordant results were evaluated again by selected expert cytopathologists. RESULTS The overall κ value was 0.612 (95% confidence interval [CI], 0.523‐0.701), it was higher for the positive and negative categories (κ = 0.692 and κ = 0.641, respectively), and it was almost null for the inconclusive category (κ = 0.058). Considering only readers from laboratories with documented experience, the κ value was higher (κ = 0.747; 95% CI, 0.643‐0.839) compared with nonexperienced centers (κ = 0.498; 95% CI, 0.388‐0.616). The results were similar in both sets of slides (κ = 0.505 [95% CI, 0.358‐0.642] and κ = 0.521 [95% CI, 0.240‐0.698] for the first and second sets, respectively). Reinterpretation of the slides with the most discordant results did not provide any improvement (first evaluation, κ = 0.616 [95% CI, 0.384‐0.866]; second evaluation, κ = 0.403 [95% CI, 0.182‐0.643]). CONCLUSIONS Dual staining for p16 ink4a and Ki‐67 demonstrated good reproducibility, confirming its robustness, which is a necessary prerequisite for its adoption as a triage test in cervical cancer screening programs that use human papillomavirus DNA as a primary test. Cancer Cytopathol 2017;125:212–220. © 2016 American Cancer Society. In human papillomavirus (HPV)‐based cervical cancer screening, dual‐staining for cyclin‐dependent kinase inhibitor 2A (p16ink4a) and Ki‐67 has been proposed as a biomarker for triaging women who are positive for HPV DNA. Staining for p16ink4a/Ki‐67 has good reproducibility between readers from 9 different laboratories, confirming its robustness, which is a necessary requisite for its introduction into