Background To evaluate strategies for cervical cancer screening among HIV-positive African women, against histological CIN endpoints.

MethodsThe HARP study enrolled HIV-positive women aged 25–50 in Burkina Faso (BF) and South Africa (SA). A stratified sampling strategy was used, with 2/3 of women on ART. Three cervical cancer screening methods were evaluated; (1) visual inspection (VIA/VILI); (2) high-risk HPV DNA (HC-2); (3) conventional cytology. Four-quadrant cervical biopsies were obtained among women with abnormalities detected by at least one test or by colposcopy.

Results 1252 women were enrolled (628 in BF; 624 in SA). The distribution of CD4 count (cells/µL) was similar in both sites: 68% with CD4+ ≥ 350 and 10% with CD4 < 200. Prevalence of high risk (HR)-HPV was 43% in BF and 61% in SA, and decreased with increasing CD4+ count (P-trend < 0.001). VIA/VILI positivity was 24% in BF and 41% in SA (P < 0.001) but did not vary by CD4 count (P-trend = 0.30). Prevalence of abnormal cytology (≥LSIL & ≥HSIL) was higher in SA (89% & 30%) than in BF (24% & 5%). 62% and 97% women were biopsied in BF and SA respectively, with CIN2+ prevalence of 6% and 29%, respectively (155 of 949 evaluated women thus far). CIN2+ prevalence decreased with increasing CD4+ (Table). Sensitivity of the single screening methods to detect CIN-2 decreased with increasing CD4+ count, whilst specificity tended to increase with increased CD4+ count (Table). Overall, HR-HPV DNA was the most sensitive test (94%) and HSIL+ cytology the most specific (90%). Further analyses with combination of tests did not show much improvement on performance.

Conclusions Cervical cancer screening tests among HIV-positive women are most sensitive among women with CD4+ count below 200 cells/µL. Screening strategies may vary according to CD4+ count but this will need to be evaluated prospectively.