The relationship of HIV prevalence in pregnant women to that in women of reproductive age: a validated method for adjustment

Abstract

Objective: To devise and validate a method for adjusting HIV seroprevalences in pregnant women to estimate population prevalences among all women in their child-bearing years.

Design: Birth and termination rates from women with known HIV infection in the United Kingdom were calculated according to the likely route of HIV infection and whether HIV infection was diagnosed.

Methods: Birth and termination rates were weighted and combined to produce summary statistics. Comparisons were then made with population birth and termination rates to derive summary relative inclusion ratios (RIRs), the relative probabilities of including HIV-infected and uninfected women in seroprevalence surveys of pregnant women.

Results: The derived RIRs for women having live births were close to unity: 1.03 [95% confidence intervals (CI) 0.90-1.17] for London and 0.80 (Cl, 0.71-0.89) for elsewhere in England and Wales. This indicates that currently observed overall seroprevalences among pregnant women having live births in London would be similar to those among all women of the same age, while elsewhere it would be slightly underestimated. Sensitivity analysis indicated that RIRs could, however, vary three-fold (0.47-1.56) according to the proportion of diagnosed maternal infections and the mix of maternal HIV-exposure categories. The method was validated by using it to predict the ratio of unlinked seroprevalences between women having terminations and live births in London. It predicted a ratio of 1.74: 1, which is close to the observed ratio of 2.07 : 1.

Conclusions: Application of HIV seroprevalences from pregnant women to whole populations may need adjustment for fertility rates among HIV-infected women. A general method for this has been derived and validated. Gathering fertility data for HIV-infected women is a useful adjunct to serosurveillance.