Abstract

The aim of this study was to assess the value of Ki-67 immunoquantitation with a computerized image analysis system for grading support in cervical intraepithelial neoplasia (CIN). Sixty-five 'blind' consensus biopsies (23 CIN 1, 22 CIN 2, and 20 CIN 3) were used as a learning set. Measurements were done in the carefully selected most severely dysplastic part of the epithelium of each CIN case. The resulting discriminating combination of quantitative features was then prospectively applied on 121 new biopsies (test set) and compared with the classical CIN grade assessed routinely by six different pathologists and with the blind review grades assessed by two experienced pathologists. In the learning set of 65 cases, a jack-knifed stepwise discriminant analysis showed that the 90th percentile of the stratification index and the number of positive nuclei per 100 microm basal membrane are the best discriminating set of features to distinguish the three CIN grades at the same time. With these features, two CIN 1 cases were 'misclassified' as CIN 2 and nine CIN 2 cases as CIN 3. Overall agreement, therefore, was only 83%. However, recut of the paraffin blocks in the two 'misclassified' CIN 1 cases revealed CIN 2 in the first and CIN 3 in the other, while the other CIN 1 cases that were correctly classified with Ki-67 quantitation remained CIN 1. Likewise, nine CIN 2 cases were misclassified as CIN 3, but in two of these nine cases histological follow-up clearly indicated CIN 3. Agreement may thus be higher than the 83% in the learning set suggests. In the subsequent prospective evaluation on 121 routine CIN cases (test set), agreement between routine CIN grades (by six independent different pathologists) and quantitative Ki-67 classification was 78%. However, when compared with the blind review CIN grades of two expert pathologists, agreement was 97% and sensitivity, specificity, and positive and negative predictive value were very high. It is concluded that Ki-67 immunoquantitation is a useful diagnostic adjunct to distinguish different CIN grades and may also be a sensitive biological indicator of progression of seemingly low-grade CIN.

Full Text
Published version (Free)

Talk to us

Join us for a 30 min session where you can share your feedback and ask us any queries you have

Schedule a call