775    

Background:  The operating characteristics of screening program components are important factors in evaluating screening policy. The study aims were:  to estimate the test-retest reproducibility of consecutive anal cytologic examinations; to estimate the agreement between simultaneous anal cytologic examinations and anal histopathology sampled at high-resolution anoscopy; to estimate agreement between colposcopic visual impression and anal histopathology; and to estimate the prevalence of severe dysplasia by simultaneous PAP category.

Methods:  Prospective study of HIV-infected males receiving anal dysplasia screening as part of routine care at dedicated HIV clinic. High-resolution anoscopy was performed by a single examiner. Agreement between consecutive PAP tests and between simultaneous PAP tests and biopsies was estimated using weighted kappa statistics. Biopsy results were coded as the most severe lesion if multiple biopsies were taken.

Results:  Between July 2000 and September 2003, 1864 patients underwent 2947 anal PAP smears. Excluding unsatisfactory smears (6%), 642 had repeat cytologic evaluation at median 387 (range 16 to 940) days and 154 had simultaneous PAP and biopsy. Using 4-category PAP grading (no atypical/malignant cells, ASCUS, LSIL, HSIL), overall weighted (1/0.67/0.33/0) kappa agreement between the first 2 PAP measurements was 0.36, but declined from 0.44 to 0.21 as the interval between PAPs increased from 0 to 3 months to >12 months. Comparing simultaneous PAP (£ASCUS, LSIL, HSIL) and biopsy (normal, AIN I, ³AIN II) overall weighted (1/0.5/0) kappa was 0.36. Comparing most severe colposcopic visual impression (³HSIL, <HSIL) and biopsy (³AIN II, <AIN II), kappa agreement was 0.32. Prevalence of AIN III or CIS at biopsy, by simultaneous PAP category, was 0 (PAP normal), 21% (PAP ASCUS), 27% (PAP LSIL), and 54% (PAP HSIL).

Conclusions:  Reproducibility of sequential PAP tests decreases as interval between tests increases. Observed agreement thus reflects sampling and interpretation variability as well as biological transitions. Kappa agreement between simultaneous PAP and biopsy, and between colposcopic impression and biopsy was only slight. Severe histology was common among those with ASCUS or LSIL cytology results. These results are relevant both for screening policy and for evaluation of response to therapies for dysplasia.

Keywords: anal dysplasia; screening; reliability