Prediction of high-risk cervical papillomavirus infection by biopsy morphology

Certain human papillomavirus (HPV) types (such as type 16) have been linked to high-grade precancers and invasive carcinomas of the cervix. However, the accuracy with which morphologic characteristics will predict the presence and type of HPV infection is controversial. Three pathologists independently classified 102 consecutive cervical biopsies with the use of specific criteria and correlated their findings with the presence of HPV 11, 16, and 18 RNA sequences by in situ hybridization. Based on the presence and distribution of nuclear atypia, abnormal mitotic figures, and koilocytosis, biopsies were classified into borderline condyloma, condyloma, borderline cervical intraepithelial neoplasia (CIN) with koilocytotic atypia (CINK), and CIN. Two or more observers agreed on the diagnosis in 96% of cases. HPV 16-related sequences alone were detected in 0% of borderline condylomata, 17% of flat condylomata, 43% of borderline CINK, 67% of CINK, and 77% of CIN lesions. Other HPVs, including those producing signals with more than one probe, were present in 0, 50, 14, 9, and 0% of these lesions, respectively. The authors data suggest that consistent identification of HPV-related cervical disease requires the presence of specific cytologic changes. In the authors' series, when HPV-related disease is present, CIN is the most common lesion and most (71%) contain HPV 16-related nucleic acids. Thus, a high proportion (88%) of histologic abnormalities associated with HPV-16 could be distinguished as CIN by morphologic characteristics alone, and this distinction could be made by most observers.