Ethics statement

The trial and informed consent forms were approved by all applicable ethics and regulatory committees: The Protection of Human Subjects Committee of FHI360, the US Food and Drug Administration, the ethics committee (EC) of the University Hospital of Antwerp, the Institutional Review Board of ITM, the Kenyatta National Hospital/University of Nairobi Ethics and Research Committee, the Republic of Kenya Ministry of Health (MoH- Pharmacy & Poisons Board), Kilimanjaro Christian Medical Center Research EC of Tanzania, National Institute for Medical Research of Tanzania, London School of Hygiene and Tropical Medicine EC, Tanzania Food and Drugs Authority, Medunsa Campus Research and EC of South Africa, University of the Free State EC and Medicines Control Council of South Africa. Written informed consent was obtained from all participants prior to conducting any study procedures.

Context: FEM-PrEP study

The details of the FEM-PrEP study have been described elsewhere [11]. The trial was approved by all applicable ethical and regulatory committees and participants provided written informed consent before any procedures were performed. Women were recruited from Pretoria and Bloemfontein, South Africa; Bondo, Kenya; and Arusha, Tanzania. Eligibility criteria required that women be HIV antibody negative, 18 to 35 years old, in general good health, hepatitis B virus surface antigen (HBsAg) negative and have no evidence of abnormal hepatic or renal function at baseline (i.e., serum creatinine < 1.5 mg/dl, creatinine clearance ≥ 60 ml/min estimated by the Cockcroft-Gault method, ALT and AST <2× local upper limit of normal [ULN] and serum phosphorus levels above the lower limit of the normal range and below DAIDS grade 3). Participants attended study visits at screening, enrollment (2–4 weeks later), and at 4-week intervals thereafter for up to 60 weeks (52 weeks on product followed by eight weeks off product). Plasma and upper layer packed cell (ULPC) aliquots were stored for TDF-FTC concentration testing at monthly visits; hepatic and renal (AST, ALT, creatinine and phosphorus) parameters were evaluated at weeks 4, 12, 24, 36, 52, 56 and when clinically indicated [13].

Laboratory analysis

AST, ALT, phosphorus and serum creatinine assays were performed according to manufacturer procedures. Samples were processed within two hours after collection. The VITROS DT II instrument (Ortho-Clinical Diagnostics, Inc., Johnson & Johnson, Buckinghamshire, UK) was used in Kenya until May 2010, thereafter the VITROS 250 (Ortho-Clinical Diagnostics, Inc., Johnson & Johnson, Buckinghamshire, UK), the CX5 Beckman Chemistry Analyzer (Beckman Coulter, Inc, Fullerton, CA, USA) was used at both South African sites and the Cobas Integra 400 Plus (Roche Diagnostics GmbH, Mannheim, Germany) was used at the Tanzanian site.

Site-specific laboratory normal ranges (LNR) were used due to unknown population similarities/differences and different instrumentation used across sites. Manufacturer recommended, or nationally/regionally established LNRs were used during the first two to three months of the trial while site-specific LNRs were developed. The local LNR was identified and verified, using Sigma Diagnostic guidelines, by the central laboratory (Institute of Tropical Medicine, Antwerp, Belgium) [14]. Details of the verification process will be described in a separate manuscript. Locally verified ranges were used for final grading of all biochemistry toxicities in Kenya and South Africa. In Tanzania, due to the delay in study initiation and early closure of the trial, local LNR were not verified; manufacturer LNRs were used instead.

The central laboratory provided on-site training in good clinical laboratory practice and technical training for all protocol-required laboratory tests, and conducted quality assurance activities including co-development of site standard operating procedures (SOPs) and analytical plans, assay and equipment validation, external quality assessment (EQA) schemes, and equipment calibration and maintenance. In addition, central lab staff conducted, at a minimum, annual site visits to verify proper conduct of laboratory procedures and review source data.

TDF-FTC concentration analysis on stored samples was conducted at the University of North Carolina at Chapel Hill. Tenofovir (TFV) was measured in plasma and tenofovir diphosphate (TFV-DP) in ULPC. Laboratory procedures for TDF-FTC concentration analysis were published previously [11, 13]. Adherence was qualitatively classified on a 6-point scale based on drug concentration data, with the top two adherence scores being “good” and “excellent” (good = plasma TFV levels >10 ng/ml and ULPC intracellular TFV-DP between 100,000-1,000,000 femtomoles/million cells, excellent = plasma TFV levels >10 ng/ml and ULPC intracellular TFV-DP between >1,000,000 femtomoles/million cells) [13].

Management of toxicity and study product interruptions

Protocol-defined toxicities were defined and managed according to clinical assessment, grade and relatedness to the study product as assessed by study clinicians [12]. Study product was temporarily or permanently withdrawn as per protocol-described criteria.

Hepatotoxicity grade 2 was defined as 2.6 -5.0 × upper limit of normal (ULN), if determined to be related to the study product, the study product was interrupted and was restarted only if AST/ALT decreased to ≤ grade 1. Hepatotoxicity grade 3 was defined as 5.1 - 10.0 × ULN and study product was interrupted independent of relatedness assessment. Study product was restarted if AST/ALT decreased to ≤ grade 1. Hepatotoxicity grade 4 was defined as > 10.0 × ULN and study product was permanently withdrawn independent of relatedness.

Creatininemia grade 1 was defined as 1.1 -1.3 × ULN and grade 2 as 1.4 -1.8 × ULN; study product was interrupted for both grades regardless of the relatedness to study product use and was restarted if creatininemia decreased to < grade 1 or < 1.3× baseline. Study product was permanently withdrawn for creatininemia grade 2 determined to be related to the study product, and for any creatininemia grade 3 (>1.8 × ULN)

Study product was also permanently withdrawn when there was hypophosphatemia associated with elevated creatininemia, low creatinine clearance or proteinuria. In case of pregnancy, study product was interrupted and resumption allowed after delivery and completion of breastfeeding or with evidence of pregnancy loss/termination). Study product was immediately discontinued in participants who HIV seroconverted.

Sample selection

The analysis population consisted of all women randomized into the trial, excluding participants who never received study product, returned all product unused, or never returned for a follow-up visit. TDF-FTC concentration data were available for a random sub-cohort of 150 participants assigned TDF-FTC, (50 from each of the three sites where HIV infections took place, i.e., Bloemfontein, Pretoria and Bondo).

Statistical analyses

Cumulative probabilities of grade 1+ creatinine, ALT and AST and grade 2+ phosphorus were plotted with data pooled across sites by treatment group over time using Kaplan-Meier methods. In addition, the number and of percentage of women experiencing toxicities by grade were summarized by baseline HBsAg status and treatment group; fisher exact tests were used to compare differences between groups.

For the random sub-cohort of 150 women in the TDF-FTC arm, we further assessed the association between adherence to TDF-FTC and change from baseline to week 4 in creatinine, phosphorus, ALT, and AST levels using generalized linear models, with adjustment for baseline level, age, body mass index, oral contraceptive (OC) use at enrollment and study site. A similar analysis was done for change from last visit on-product and first visit after product use ceased (at study exit or earlier product withdrawal). SAS 9.3 (SAS Institute, Cary, NC) was used for all analyses; statistical significance was defined as p ≤ 0.05.

Characteristics of enrolled population

The analysis included a total of 2,058 women who were enrolled during June 2009 to April 2011. There were 701 and 707 person-years of follow up in the TDF-FTC and placebo arms respectively. Demographic characteristics, medical history and self-reported sexual behavior at baseline have been reported elsewhere and were generally similar between both groups [11]. At enrolment 20.8% and 21.4% of participants were hepatitis B surface antibody (HBsAb) positive in the TDF-FTC and placebo arms respectively.

Kidney toxicity

The cumulative probabilities of grade 1+ creatinine or grade 2+ phosphatemia toxicities were higher in the TDF-FTC arm, although neither of these differences were statistically significant (Figures 1 and 2).

Six participants developed grade 2+ creatininemia: four (0.4%) in the TDF-FTC arm and two (0.2%) in the placebo arm. Their baseline creatininemia levels were comparable to the average baseline creatininemia at their respective sites. Among the four participants in the TDF-FTC arm, the grade 2+ creatininemia resolved after 1, 2, 16 and 28 weeks respectively; among the two in the placebo arm, one resolved after 6 weeks and decreased to a grade 1 for 10 weeks in the second participant. The latter participant was referred to a nephrologist at the closing of the trial; she later self-reported that her creatinine returned to normal, though results could not be validated by the study team.

None of the six participants with grade 2+ creatininemia reported taking concomitant medication at the time of the event. TDF-FTC drug concentration data were available for the four participants in the TDF-FTC arm: one was classified as having excellent adherence, two as having good adherence, and one as not adherent in the interval prior to the event.

Hepatic toxicity

The cumulative probabilities of grade 1+ ALT and AST toxicities were higher among participants in the TDF-FTC arm than in the placebo arm (p = 0.03 for both) (see Figures 3 and 4).

Sixteen participants developed grade 3+ ALT and/or AST toxicities: eight in the TDF-FTC arm and eight in the placebo arm. Baseline AST/ALT levels among these participants were comparable to the average ALT/AST in their respective sites. In three of these 16 participants, the grade 3+ ALT/AST occurred at the time of HIV seroconversion (one in the TDF-FTC arm and two in the placebo arm), and 11 of the 16 (six in the TDF-FTC arm and five in the placebo arm) events occurred in participants who had received the hepatitis B vaccine. Six participants (two in the TDF-FTC arm and four in the placebo arm) with grade 3+ ALT/AST toxicities reported taking a concomitant medication at the time of the event; medications included acetaminophen, ibuprofen, diclofenac and metronidazole. All grade 3+ ALT/AST toxicities were resolved between two and eight weeks without any difference in time to resolution between TDF-FTC and placebo arms. One participant (in the placebo arm) with grade 3+ ALT at week 12 never returned to the study clinic despite intense tracing efforts. Drug level data were available for seven participants in the TDF-FTC arm with grade 3+ ALT/AST: two were classified as having excellent adherence, one as having good adherence and four as non-adherent in the interval before the event.

At one site – Bondo – the proportion of grade 2+ AST toxicities was significantly higher in the TDF-FTC arm (5.0%) than in the placebo arm (1.9%) (p = 0.02). Also, overall, Bondo observed higher proportions of grade 1+ AST toxicities (43.1% in TDF-FTC arm vs. 36.4% in placebo arm) compared to Bloemfontein (5.7% vs. 3.7% respectively), Pretoria (2.9% vs. 2.4% respectively), and Arusha (0% vs. 3.6% respectively). At baseline, Bondo participants had a higher exposure to hepatitis B.

TDF-FTC concentrations and observed kidney or hepatic toxicities

TDF-FTC concentration data from a sub-cohort of 150 women indicated that very few consistently took the study drug, making it impossible to assess the impact of continuous exposure over longer periods and liver/kidney function. However, the percentage of women with evidence of good adherence in their first four weeks of participation (~40%) was sufficient to assess the impact of short-term drug use. We observed that women with evidence of good adherence had higher mean change from baseline to week 4 in their AST levels (2.90 (0.37, 5.42); p = 0.025) as compared to women with less than good adherence. However, corresponding changes in ALT (1.54 (-1.44, 4.52); p = 0.31), creatinine (0.19 (-2.60, 2.98); p = 0.89) and phosphorus (0.02 (-0.04, 0.08); p = 0.53) were not statistically significant. We did not observe an association between adherence to TDF-FTC and change in ALT, AST, creatinine or phosphorus levels between the final drug-use interval and four weeks after product withdrawal.