Koilocytotic atypia in Papanicolaou smears. Reproducibility and biopsy correlations

It would be useful if cervical smears containing koilocytes alone could be reliably separated from those containing other forms of nuclear atypia within the spectrum of low grade squamous intraepithelial lesions (LSIL) and that this separation was predictive of differences in biopsy follow-up. In this article, the authors sought to test this possibility.

Consecutive smears diagnosed as LSIL from 140 patients who had follow-up colposcopic biopsies were reviewed independently by three observers. The inter- and intraobserver reproducibility in diagnosing smears with koilocytes alone was assessed by the kappa statistic. Comparison of each reviewer's cytologic diagnosis with a reviewed biopsy diagnosis was assessed using chi-square analysis. The quantity of abnormal cells was compared with the presence or absence of a lesion on biopsy.

Kappa values for any 2 observers agreeing on koilocytosis as a separate category ranged from 0.22-0.47 (poor to good). Intraobserver reproducibility across all cytologic categories ranged from 0.35-0.62 (poor to good). A cytologic diagnosis of koilocytosis predicted a lower rate of high grade SIL (HSIL) on initial biopsy for one of the observers, but not for the other two. Koilocytosis did not predict a lower rate of LSIL on initial biopsy or HSIL in follow-up biopsies for any observer. The quantity of cells on the initial smear did not correlate with a lesion on biopsy.

Separation of LSIL into two diagnostic categories is not feasible because of its poor cytologic reproducibility and inability to predict differences in biopsy diagnosis. The quantity of abnormal cells in LSIL is not predictive of the detection of a lesion on biopsy.