The study was reviewed and approved by two Institutional Review Boards at FHI 360 in North Carolina, USA, and the Kenya Medical Research Institute, Nairobi, Kenya. Women voluntarily enrolled through a written informed consent process. All participants were receiving standard HIV-related care at the comprehensive care centre of Kenyatta National Hospital.
We recruited 60 women on ART who wanted to use a levonorgestrel (LNG) contraceptive implant (the product is marketed as Jadelle®). The following additional inclusion criteria were applied: aged between 18 and 44 years, last CD4 count of at least 200 cells, at least six months on a first line therapy of stavudine or zidovudine plus lamivudine+nevirapine, sexually active and willing to continue ART. (We chose first line therapies to maximize participation, though other ART regimens could be studied, since contraindications to implant use are non-existent.). We applied the following exclusion criteria to allow enough of a wash-out period for previous hormonal contraception: ≤ five months since last DMPA injection, ≤ two months since last dose of oral contraceptives, ≤ two months since removal of an LNG intrauterine device and ≤ two months since removal of contraceptive implant. We excluded women who were not appropriate candidates for an implant: currently pregnant, desire for pregnancy in the next 12 months, surgically sterilized (including partner vasectomy) and medical contraindications for implant use. To avoid unstable health histories that might require other interventions, we excluded women who were pregnant within the past six months, women currently taking rifampicin and women in unstable WHO stage 3 or 4 HIV disease. On the day of enrolment, a new blood draw was done and sent to the lab for analysis and the sub-dermal implant was inserted. Once the new CD4 count was available and confirmed to be at least 200 cells, we added the implant user to a list of women needing a matched control and recorded participant ID number, current age and CD4 count.
We recruited a cohort of matched controls who were not using hormonal contraception. We applied three matching criteria: using the same ART regimen as cases; current age (±5 years); and CD4 count (±50 cells). We accepted baseline CD4 counts up to six months old as long as the participant had not used hormonal contraception for the wash-out periods (as specified above) before the blood specimen was collected. Furthermore, the baseline CD4 count was only usable if the participant had been on ART for at least six months prior to that specimen being collected and if it was at least 200 cells. In addition to the criteria outlined above, we applied the following exclusion criteria for controls to better avoid possible contamination: any use of hormonal contraception between baseline CD4 and enrolment date; plans to use hormonal contraception in the next 12 months.
Regular hospital services included periodic check-ups, ART provision and periodic blood draws to measure CD4 counts. We did not interfere with clinical management of patients but did advocate for blood draws and lab tests to maintain the routine standard of care at that facility (every six months, if stable).
We aimed to recruit 60 implant users and an equal number of matched controls, based on variance of change in CD4 counts from previous research in Uganda and Zimbabwe [10
] and the following assumptions: the mean change from baseline CD4 will not differ by more than 100 cells per µL compared with the non-hormonal group, two-sided 0.05 level test, 80% power to detect a difference of 105 cells per µL in the change from baseline and 15% lost to follow-up or death.
We followed the participants prospectively and recorded CD4 counts as they became available through the regular services (approximately every six months). In addition, each time participants returned for regular services, they were asked to visit the study nurse and provide an update on contraceptive use, illnesses and pregnancy. At that time, the study nurse also reviewed medical records, if feasible. At the close of the study, final interviews and medical record reviews were conducted. The following key information was transcribed to study forms: ART medications/dates, CD4 counts/dates, opportunistic infections/dates and pregnancies/dates.
The primary outcome was change in CD4 count over time, comparing implant users to their non-hormonal controls. Secondary outcomes included the incidence of pregnancy and opportunistic infections. We used generalized growth curve models and Wald chi-square tests to compare changes in CD4 counts across the study groups, which controls for time in study. Prospective CD4 measures were censored (excluded), if any of the following events occurred: change in ART, implant removal and use of hormonal contraception among controls. Analyses were conducted using SAS 9.2 (SAS Institute, Cary, NC, USA).