Abstract
AIMS To analyse the effect of the expert end-point committee (EPC) review on histological endpoint classification of cervical intraepithelial neoplasia (CIN). Methods A cohort of women living with HIV were recruited in Burkina Faso (BF) and South Africa (SA) and followed over 18 months. Four-quadrant cervical biopsies were obtained in women with abnormalities detected by at least one screening test. A central review by a panel of five pathologists was organised at baseline and at endline. Results At baseline the prevalence of high-grade CIN (CIN2+) was 5.1% (28/554) in BF and 23.3% (134/574) in SA by local diagnosis, and 5.8% (32/554) in BF and 22.5% (129/574) in SA by the EPC. At endline the prevalence of CIN2+ was 2.3% (11/483) in BF and 9.4% (47/501) in SA by local diagnosis, and 1.4% (7/483) in BF and 10.2% (51/501) in SA by EPC. The prevalence of borderline CIN1/2 cases was 2.8% (32/1128) and 0.8% (8/984) at baseline and endline. Overall agreement between local diagnosis and final diagnosis for distinguishing CIN2+ from ≤CIN1 was 91.2% (κ=0.82) and 88.9% (κ=0.71) for BF at baseline and endline, and 92.7% (κ=0.79) and 98.7% (κ=0.97) for SA at baseline and endline. Among the CIN1/2 cases, 12 (37.5%) were graded up to CIN2 and 20 (62.5%) were graded down to CIN1 at baseline, and 3 (37.5%) were graded up to CIN2 and 5 (62.5%) were graded down to CIN1 at endline. Conclusions This study highlights the importance of a centralised rigorous re-reading with exchange of experiences among pathologists from different settings.