Objective
This population‐based study was conducted to evaluate the performance of colposcopy to assess women with positive visual inspection with acetic acid (VIA) and/or human papillomavirus (HPV) tests.

Materials and Methods
A total of 30 773 women were screened by VIA and oncogenic HPV test. Hybrid capture 2 was used for oncogenic HPV detection. All VIA‐ and/or HPV‐positive women and 8.7% test‐negative women had the colposcopy. International Federation of Cervical Pathology & Colposcopy (IFCPC) 2011 nomenclature was used for colposcopic classification of abnormalities. All women with grade 1 or worse lesions had punch biopsies. Biopsies were also obtained from HPV‐positive women with normal colposcopy.

Results
Colposcopy and satisfactory biopsy reports were available for total 2466 women. The overall strength of agreement between colposcopy and histologic classification of cervical neoplasias was poor (kappa = 0.17). Agreement was better when colposcopy was performed on HPV‐positive women compared to VIA‐positive women. Sensitivity of colposcopy to detect high‐grade squamous intraepithelial lesions (HSIL) at referral threshold of grade 1 abnormality was 84.8% after correction of verification bias. Colposcopy was most inaccurate in identifying non‐neoplastic conditions often encountered in VIA‐ and/or HPV‐positive women. In 68.8% women with normal histology, colposcopic impression was grade 1 and above. Overestimation of disease severity on colposcopy was more common in VIA‐positive women. Colposcopy also underestimated severity of disease in 52.6% of women with HSIL diagnosis on biopsy.

Conclusions
Colposcopy performed well in the overall detection of cervical neoplasias, though its capability for accurate categorisation of degree of abnormality was poor.