The Performance of a New Scoring System for Colposcopy in Detecting High-Grade Dysplasia in the Uterine Cervix

Abstract

This article describes the development of a simple scoring system for colposcopic grading of high-grade cervical dysplasia. Two different clinic populations with a 48% and a 35% rate of high-risk lesions were used to develop the model, which was based on data accumulated from March 2001 to June 2003 during colposcopic examination of women referred for abnormal cytology. Colposcopically directed biopsies were taken of all suspicious areas using punch biopsy, laser cone, or large loop excision of the transformations zone (LLETZ). Only patients with histopathology of the cervix were included in the study (n = 297). All colposcopic examinations were graded by experienced colposcopists for acetowhiteness of the epithelium, lesion margin and surface contour, vessel pattern, lesion size, and iodine staining. A three-point scoring system was used for each colposcopic characteristic (see Table 1).TABLE 1: Scoring modelBenign conditions were seen in 30% of the histopathologic samples, and koilocytosis or cervical intraepithelial neoplasia (CIN 1) was diagnosed in 25%. Forty-four percent of the samples contained high-grade lesions, including cervical intraepithelial neoplasia (CIN) 2 (13%), CIN 3 (28%), and invasive cancer (3%). Using the biopsy diagnosis of high-grade CIN (defined as CIN 2 or greater) as a standard, an “ideal” grading system was constructed that assigned each colposcopic characteristic a weighted value that would produce the best sensitivity and specificity performance for the model. These values were then rounded off to simplify the final grading system shown in Table 1. When the colposcopic lesion characteristics were analyzed by multiple logistic regression, acetowhiteness had the highest correlation to high-grade dysplasia, followed by lesion margin and surface contour, lesion size, iodine staining, and vessel pattern. The distribution of total colposcopic scores for each lesion was then analyzed for the sensitivity and specificity to predict a CIN 2 lesion or greater. No patient with CIN 2 had a colposcopic score of four or less. Only two patients had a score of five. Ninety-five percent of the patients with CIN 2 or greater had a colposcopic score of six or higher. Only 10% of women with CIN 1 or less had a colposcopic score of eight or more. Four women in the series had glandular dysplasia and received scores of eight to 10. All had adenocarcinoma in situ on final histologic evaluation. There were 10 invasive cancers diagnosed. Five women had squamous cell carcinoma, four had adenocarcinoma, and one had a mixed adenosquamous carcinoma. The colposcopic scoring index graded one of these patients with seven points; the other nine received scores of eight to 10. As a control, another 199 women patients were diagnosed with benign, low-grade, or high-grade cervical dysplasia based on the general colposcopic impression of the experienced examiner without using any formal grading system. Eighteen of 85 biopsy-proven high-grade lesions were underdiagnosed by colposcopic impression and six of 114 low-grade lesions were misidentified as high-grade (sensitivity = 79% and specificity = 95%).