High attrition and irregular testing for human immunodeficiency virus (HIV) in cohort studies for high-risk populations can bias incidence estimates. We compare incidence trends for high-risk women attending a dedicated HIV prevention and treatment clinic, using common methods for assigning when seroconversion occurs and whether seroconversion occurs among those with attrition.
Between April 2008 and May 2009, women were enrolled into cohort 1 and from January 2013 into cohort 2, then scheduled for follow-up once every 3 months. Incidence trends based on assuming a midpoint in the seroconversion interval were compared with those of assigning a random-point. We also compared estimates based on the random-point with and without multiple imputation (MI) of serostatuses for participants with attrition.
By May 2017, 3084 HIV-negative women had been enrolled with 18,364 clinic visits. Before attrition, 27.6% (6990 of 25,354) were missed visits. By August 2017, 65.8% (426 of 647) of those enrolled in cohort 1 and 49.0% (1194 of 2437) in cohort 2 were defined with attrition. Among women with 1 or more follow-up visit, 93 of 605 in cohort 1 and 77 of 1601 in cohort 2 seroconverted. Periods with longer seroconversion intervals appeared to have noticeable differences in incidences when comparing the midpoint and random-point values. The MI for attrition is likely to have overestimated incidence after escalated attrition of participants. Based on random-point without MI for attrition, incidence at end of observation was 3.8/100 person-years in cohort 1 and 1.8/100 in cohort 2.
The random-point approach attenuated variation in incidence observed using midpoint. The high incidence after years of ongoing prevention efforts in this vulnerable population should be investigated to further reduce incidence.
In a cohort of human immunodeficiency virus high-risk women, assigning random-point values as the seroconversion timing attenuated variation in incidence observed with the midpoint assignment during observation periods with longer seroconversion intervals.
From the *Medical Research Council/Uganda Virus Research Institute and London School of Hygiene & Tropical Medicine (MRC/UVRI and LSHTM) Uganda Research Unit, Entebbe, Uganda;
†London School of Hygiene and Tropical Medicine (LSHTM), and
‡MRC Tropical Epidemiology Group, London School of Hygiene and Tropical Medicine (LSHTM), London, United Kingdom
Acknowledgements: The authors thank the cohort participants and the many staff who have contributed to the success of the Good Health for Women Project over the last 10 years. Particularly, Yunia Mayanja for her comments to the final draft, Daniel Bagiire and Onesmus Kamacooko for their endless response to the many data requests. The authors acknowledge Jim Todd, Kathy Baisley and Maryam Shahmanesh for the initial input regarding the overall direction of the work of which this article is part.
Ethical issues: This study was approved by the Research Ethics Committee of the Ugandan Virus Research Institute, Uganda National Committee for Science and Technology, and Ethics Committee of the London School of Hygiene and Tropical Medicine. Each participant gave written informed consent and confidentiality was kept by using numerical identifiers only.
Conflicts of Interest: none declared.
Sources of Funding: This work was jointly funded by the UK Medical Research Council (MRC) and the UK Department for International Development (DFID) under the MRC/DFID Concordat agreement (K012126/1). The clinic receives support from PEPFAR through the Centrex for Disease Control and Prevention for the provision of clinical services. I.K. is jointly funded for a PhD by MRC Tropical Epidemiology Group, London School of Hygiene and Tropical Medicine (LSHTM), UK and UK Medical Research Council (MRC).
Correspondence: Ivan Kasamba, MSc, MRC/UVRI and LSHTM Uganda Research Unit, Plot 51-59 Nakiwogo Rd, Entebbe, Uganda. E-mail: Ivan.Kasamba@lshtm.ac.uk.
Received for publication September 23, 2018, and accepted January 14, 2019.
Supplemental digital content is available for this article. Direct URL citations appear in the printed text, and links to the digital files are provided in the HTML text of this article on the journal’s Web site (http://www.stdjournal.com).