Background:  Cervical cancer is a leading cause of cancer death in sub-Saharan Africa. In HIV-infected women from developed settings, higher rates of premalignant cervical lesions with more rapid progression compared to uninfected women are reported. Life-prolonging ART may alter the risk of cervical cancer. We report progression and regression of cervical dysplasia in a prospective cohort of HIV-infected women in Soweto, South Africa.

Methods:  Women, 18 years and older, attending an HIV wellness clinic between August 2003 and July 2008 were offered cervical smears as part of comprehensive care. Smears are assessed using the Bethesda system. Women with high-grade squamous intraepithelial lesions (ASC-H or HSIL), or worse are referred for colposcopic treatment. Our prevalence analysis includes women with baseline CD4 counts within 6 months of their cervical smear and incidence rates are from the same women with >1 smear ≥6 months apart. Cox models were used to estimate hazard ratios (HR) for predictors of incident events as follows: HSIL in women with baseline normal, ASCUS or low-grade squamous intraepithelial lesions (LSIL) smears; regression to normal in women with baseline LSIL.

Results:  A total of 1837 women had a baseline smear; their median age and CD4 count was 32 years and 296 cells/μL; 27% were already on or initiated ART during follow-up. At baseline, 60%, 24%, and 12% had normal, LSIL, and HSIL smears, respectively. Of those with baseline normal, ASCUS or LSIL smears, 687 women had at least another smear, a median time of 466.5 days (IQR 384 to 703) apart. Of these, 8.6% progressed to HSIL (4.0/100 person-years 95%CI 3.1 to 5.1). Of 484 women with normal baseline smears, 23% progressed to LSIL (10.9/100 person-years 95%CI:9.4-13.5).  Of 171 women with LSIL at baseline and with another smear ≥1 yr later, 33.5% regressed to normal (17.3/100 person-years, 95%CI 13.2 to 22.6). In multivariate analysis, predictors for progression to HSIL were CD4 counts ≤200 vs >500 cells/μL (aHR 4.5 95%CI 1.7 to 12.1), but not HAART (aHR 0.6 95%CI 0.3 to 1.2). For regression to normal from LSIL, predictors were CD4 count/100 (aHR 1.35 95%CI 1.2 to 1.6), and age/5 (aHR 1.4 95%CI 1.2 to 1.8). 

Conclusions:  HIV-infected women have high rates of prevalent and incident HSIL and LSIL with low risk of regression of LSIL; suggesting short cervical screening intervals in women with CD4 counts <200 and investigating more proactive management of LSIL in HIV-infected women. Our results also add urgency to improving access to an affordable, effective human papillomavirus (HPV) vaccine.