OBJECTIVE: To determine which condom-use measures best predict biological outcomes in STI/HIV-prevention research. METHODS: In a prospective cohort study of 2296 HIV-negative Zimbabwean women aged 18-35 followed for up to 2 years, the authors compared four measures of condom use (use since last visit, use at last sex, frequency of use and count of unprotected acts). The authors evaluated the performance of each in predicting incidence of pregnancy, cervical STIs (chlamydia/gonorrhoea) and HIV. RESULTS: Over follow-up, 19.3% of women became pregnant, 10.3% acquired a cervical STI, and 6.9% acquired HIV infection. In multivariable analysis, all four condom-use measures were significantly associated with a reduced pregnancy incidence; statistical tests of fit suggest that the frequency of use measure was most predictive. The time to pregnancy was longer for women who, in a typical month during the previous 3 months, reported always using condoms as compared with those who never used a condom (HR 0.19; 95% CI 0.14 to 0.26). Among those women diagnosed as having prevalent chlamydia/gonorrhoea at study enrolment, three of the four reported condom-use measures were associated with a non-significant but decreased risk of incident cervical STI. Reported condom use was associated with an increased risk for cervical STI among women without chlamydia/gonorrhoea at enrolment. None of the condom-use measures were associated with HIV infection. CONCLUSIONS: The frequency of reported condom use measure best predicted pregnancy incidence; however, the authors found no evidence for a clear 'best' condom-use measure for use in STI/HIV prevention research in this population of Zimbabwean women
