Hormonal contraception and HIV: the methods have confused the message

OBJECTIVE:: To examine different scenarios through which confounding by condom use may lead to inaccurate conclusions about the effect of hormonal contraception on HIV acquisition in women. DESIGN AND METHODS:: Scenario analyses were conducted to evaluate the impact of coarse adjustment for condom use and condom misreporting on adjusted relative risk estimates for HIV acquisition in injectable hormonal contraception (IHC) users vs. nonusers. RESULTS:: Analyses crudely accounting for condom use through a binary variable result in biased hormonal contraception-related risk estimates if condoms are used during follow-up periods in which any unprotected sex is reported and condom use differs by hormonal contraception use. We found that over-reporting of condom use is plausible in at least one recent study, as demonstrated by high pregnancy rates given, reported IHC and condom use. Over-reporting of condom use also biases estimates, typically leading to underestimation of IHC-related risk if over-reporting is the same among IHC and non-hormonal contraception users, and overestimation of IHC-related risk if condom misreporting is differential by IHC use. The impact of misreported condom use is most pronounced in study populations with high condom uptake. CONCLUSIONS:: Discrepant findings in hormonal contraception-HIV-related research may result from inadequate measurement or adjustment for confounding by condom use. Future studies should precisely account for condom use in statistical analyses. Studies should aim to quantify the degree of condom use misreporting, by comparing reported condom use to pregnancy, HIV or other sexually transmitted infection rates, and if possible, testing stored genital swabs for prostate-specific antigen or Y chromosome.