Systematic identification of correlates of HIV infection: an X-wide association study

Objective: Better identification of at-risk groups could benefit HIV-1 care programmes. We systematically identified HIV-1 risk factors in two nationally representative cohorts of women in the Demographic and Health Surveys. Methods: We identified and replicated the association of 1415 social, economic, environmental, and behavioral factors with HIV-1 status. We used the 2007 and 2013–2014 surveys conducted among 5715 and 15 433 Zambian women, respectively (688 shared factors). We used false discovery rate criteria to identify factors that are strongly associated with HIV-1 in univariate and multivariate models of the entire population, as well as in subgroups stratified by wealth, residence, age, and past HIV-1 testing. Results: In the univariate analysis, we identified 102 and 182 variables that are associated with HIV-1 in the two surveys, respectively (79 factors were associated in both). Factors that were associated with HIV-1 status in full-sample analyses and in subgroups include being formerly married (adjusted OR 2007, 2.8, P < 10−16; 2013–2014 2.8, P < 10−29), widowhood (aOR 3.7, P < 10−12; and 4.2, P < 10−30), genital ulcers within 12 months (aOR 2.4, P < 10−5; and 2.2, P < 10−6), and having a woman head of the household (aOR 1.7, P < 10−7; and 2.1, P < 10−26), while owning a bicycle (aOR 0.6, P < 10−6; and 0.6, P < 10−8) and currently breastfeeding (aOR 0.5, P < 10−9; and 0.4, P < 10−26) were associated with decreased risk. Area under the curve for HIV-1 positivity was 0.76–0.82. Conclusion: Our X-wide association study identifies under-recognized factors related to HIV-1 infection, including widowhood, breastfeeding, and gender of head of the household. These features could be used to improve HIV-1 identification programs.