Use of an expanded gold standard to estimate the accuracy of colposcopy and visual inspection with acetic acid

Abstract

We estimate the accuracy of colposcopy and visual inspection with acetic acid (VIA) while minimizing the effects of misclassification bias, and maximizing ascertainment of disease. VIA was performed by experienced physicians on a population-based sample of women aged 30 to 49 years in rural Shanxi province, China. Each woman received VIA, liquid-based cytology (LBC) and hybrid capture 2 (hc2, QIAGEN, Gaithersburg, MD; formerly Digene Corporation). Any woman who tested positive on any test had colposcopy, endocervical curettage (ECC) with directed biopsies as necessary and 4-quadrant random biopsies from normal-appearing areas of the cervix. A standard diagnosis based on colposcopy and directed biopsy, and an expanded diagnosis including ECC and 4-quadrant random biopsy were generated for each woman. In 1,839 women, use of the expanded versus the standard diagnostic criteria increased the prevalence of histologically confirmed high-grade cervical intraepithelial neoplasia and cancer (CIN2+) from 3.2% (59/1,839) to 4.2% (77/1,839) and decreased the sensitivity of VIA for CIN2+ from 69.5% (95% CI: 56.8-79.8) to 58.4% (95% CI: 47.3-68.8%) with little change in specificity of approximately 89%. Compared with the expanded diagnostic criterion, the sensitivity of a visual diagnosis of high-grade CIN or cancer by a colposcopist was 49.4% (95% CI: 38.2-60.5). The use of an expanded diagnostic criterion in this study yielded more conservative estimates of the sensitivity of VIA and colposcopy.