The diagnostic accuracy of colposcopy in previously treated cervical intraepithelial neoplasia

To determine whether colposcopy is reliable in diagnosing cervical intraepithelial neoplasia in women who have undergone a previous cervical excision biopsy.

A prospective study of women attending the colposcopy clinic at the University Hospital of North Staffordshire was performed between January 1998 and December 1999.

A clear histological diagnosis of the grade of cervical intraepithelial neoplasia was available for 469 in the treatment-naive group and 58 in the treatment group. kappa coefficients comparing the colposcopic impression (negative, low-grade, high-grade, or invasion) with histological diagnosis showed that there was no difference between the treatment-naive group, weighted kappa=0.46, and the previous treatment group, weighted kappa=0.47. The sensitivity, specificity, positive predictive value, and negative predictive value of colposcopy for any cervical disease in the treatment-naive women were 93.9%, 51.9%, 96.7%, and 34.1%, respectively, compared with 77.6%, 66.7%, 86.4%, and 35.3% in previously treated women. The sensitivity of colposcopy fell when it was used to differentiate normal and low-grade disease from high-grade disease and invasion: 82.4%, 55.9%, 82.6%, and 49.6% for treatment-naive women, compared with 61.5%, 84.2%, 60.0% and 51.6%, respectively, for the previous treatment group.

Previous treatment to the cervix does not seem to impair the ability of colposcopy to differentiate normal cervix from all grades of cervical abnormality in women where the squamocolumnar junction is visible. However, there is a suggestion that the sensitivity of colposcopy to differentiate negative/low-grade disease from high-grade disease/invasion is lower in previously treated women.