Relationship between the presence of oncogenic HPV DNA assessed by polymerase chain reaction and Ki-67 immunoquantitative features in cervical intraepithelial neoplasia

The aims of this study were firstly to determine which Ki-67 immunoquantitative parameters correlate with the presence of oncogenic human papillomavirus (HPV) in cervical intraepithelial neoplasia (CIN) lesions; and secondly to compare prospectively the routinely assessed CIN grades with the Ki-67 quantitative pathological CIN grade, the expert revised grade, and the presence of oncogenic HPV DNA. HPV polymerase chain reaction (PCR) and Ki-67 immunoquantitation were performed on 90 consecutive biopsies (16 CIN 1, 35 CIN 2, and 39 CIN 3). CIN grade was assessed routinely by six different pathologists. The presence of the lesion was confirmed in a histological section following the material used for PCR and Ki-67 analysis. In a second prospective routine test set analysis, 66 more CIN lesions (14 CIN 1, 15 CIN 2, and 37 CIN 3) were routinely graded (also by six different pathologists, routine CIN grade=CIN(ROUT)), studied for oncogenic HPV DNA, and graded by quantitative Ki-67 features (quantitative pathological CIN grade=CIN(QP)). These latter cases were blindly revised by one of the authors (reference CIN grade=CIN(REF)). Eight of the nine Ki-67 immunoquantitative features showed a significant difference between the oncogenic HPV-positive and -negative cases. The best single discriminator was the 90th percentile of the stratification index (SI90). All 61 cases with Ki-67 SI90>0.60 were HPV-positive (68% of the total group studied). Of the 29 cases with SI90< or =0.60, 16 were negative and 13 positive for oncogenic HPV and none of the Ki-67 features (either single or combined) could distinguish them. Using stepwise multivariate analysis, the best discriminating combination of features was SI90 and the percentage of Ki-67-positive nuclei in the deep third layer of the epithelium (PERC DL). The combination of SI90 and the percentage of Ki-67-positive nuclei per 100 microm basal membrane was nearly as strong as that of SI90 and PERC DL. With these two features, 86% of the cases were correctly classified. The subjective estimate of SI90 (>0.60 or < or =0.60) by two independent observers was not accurate and not reproducible. In the prospective routine test set analysis of 66 cases, the 37 CIN(ROUT)=3 all had CIN(QP) and CIN(REF)=3 and all these cases were oncogenic HPV-positive. Eight of the 14 original CIN(ROUT)=1 grades were oncogenic HPV (=HPV)-positive and five of these eight were upgraded by CIN(QP) to CIN 2 and CIN 3. These upgrades were in agreement with the blind reference revisions. The six HPV-negative CIN(ROUT)=1 cases were CIN 1 both by CIN(QP) and by CIN(REF). Thirteen of the 15 original CIN(ROUT)=2 grades were HPV-positive and seven of these were CIN(QP)=3. All six HPV-positive CIN(ROUT)=2 cases that were CIN(QP)=2 were also CIN(REF)=2 at blind revision. In conclusion, this study has shown firstly, that in CIN lesions, Ki-67 immunoquantitative features and the presence of oncogenic HPV are highly correlated, and also within one subjective CIN grade; secondly, that subjective impressions of SI90 are not as accurate or reproducible as quantitative image analysis results; and thirdly, that the routine application of QP CIN-grading gives results that are in very good agreement with CIN grades assessed by an expert. Thus, routine QP-grading may be used to correct the subjective grade assessed by non-expert pathologists.