Bone Mineral Density, Kidney Function, Weight Gain and Insulin Resistance in Women Who Switch From TDF/FTC/NNRTI to ABC/3TC/DTG

F Ibrahim; A Samarawickrama; L Hamzah; R Vincent; Y Gilleece; L Waters; S Kegg; B Barbini; L Campbell; FA Post


HIV Medicine. 2021;22(2):83-91. 

In This Article


Bone Evaluation in women who Switch from TDF + 3TC/FTC + NNRTI to Triumeq (BESTT) is an ongoing 96-week, open-label, phase IV, randomized, controlled, multicentre clinical trial (EudraCT 2015-005297-37). HIV-positive women who were HLA-B5701-negative and ≥ 40 years old with viral load < 50 HIV RNA copies/mL on an ART regimen containing TDF/FTC (or TDF/3TC) and a NNRTI [efavirenz (EFV), nevirapine (NVP) or rilpivirine (RPV)] for ≥ 12 months were recruited from nine HIV clinics in London and Brighton, UK. All participants gave written, informed consent, and study approval was granted by the Health Research Authority and National Health Service Research Ethics Committee.

Women were excluded if they had hepatitis B (surface antigen-positive), hepatitis C (RNA-positive), moderate or severe hepatic impairment, HIV resistance mutations to any of the study drugs, screening laboratory abnormalities (AIDS Clinical Trials Group grade 3 or 4) or abnormal serum alanine aminotransferase (≥ three times the upper limit of normal) or bilirubin levels (≥ 1.5 times the upper limit of normal with > 35% direct bilirubin), creatinine clearance < 50 mL/min (by Cockcroft-Gault), or drank > 3 units of alcohol daily. Women of child-bearing potential had to take measures to avoid pregnancy for the duration of the trial.

While the presence of osteopenia or osteoporosis per se was not an exclusion criterion, those who were currently using or planning to use bisphosphonates were excluded. We performed a fracture risk (FRAX) assessment and followed the National Osteoporosis Guideline Group recommendations ( to identify individuals who required DXA scanning prior to their baseline visit to confirm equipoise for continued TDF exposure, and thus whether to randomize these participants.

Hypothesis and Sample Size

We hypothesized that BMD improves in HIV-positive women who switch ART from TDF/FTC/NNRTI to ABC/3TC/DTG. With 1:2 randomization and 90 participants, the study had 82% power to detect a 2% difference in total hip BMD at 48 weeks [mean percentage change (SD) = 0.02 (0.03) vs. 0.001 (0.03)] at week 48.


Eligible participants were randomized 1:2 using randomly permuted blocks, stratified by age (< or ≥ 50 years), to continue their current regimen or switch to ABC/3TC/DTG for 96 weeks.


Participants were screened for eligibility and randomized within 28 days. Post-baseline visits occurred at 4 (switch arm only), 12, 24, 48, 72 and 96 weeks. Safety assessments included symptom-directed physical examinations, kidney and bone profiles, and urine dipstick analysis at each visit. In addition, HIV RNA, CD4 cell count, fasting lipids and glucose, and pregnancy tests (women of child-bearing potential only) were performed every 6 months. Aliquots of plasma, serum and urine were snap-frozen and stored at −70°C for biomarker analyses.

Bone mineral density (expressed in g/cm2) was assessed using dual-energy X-ray absorptiometry (DXA) scans of the lumbar spine (L1–L4) and the (non-dominant) total hip and femoral neck at baseline, and weeks 24 and 48. To ensure comparability, each subject was scanned using the same DXA scanner and the same scanning parameters and positioning as that used during the baseline visit. To ensure the reproducibility of the DXA scans, each image from the follow-up scan was visually compared with the image from the first visit (baseline) during each follow-up acquisition. If necessary, the subject was repositioned, and the scan repeated. To ensure comparability between study sites and to check the relative calibration of the BMD results from each site, quality control was performed using a single bona fide spine phantom at all study sites at the start and end of the study.

We evaluated bone turnover at baseline and weeks 24 and 48 by quantification of alkaline phosphatase (ALP), type I collagen cross-linked C-telopeptide (CTX) and procollagen type 1 N-terminal propeptide (P1NP), weight and renal function at baseline and weeks 12, 24 and 48 by eGFR (Chronic Kidney Disease Epidemiology Collaboration[22]), serum cystatin C, albumin/creatinine ratio (ACR), protein/creatinine ratio (PCR), retinol-binding protein/creatinine ratio (RBPCR)[20] and fractional excretion of phosphate (FE-PO4), 25-hydroxy-vitamin D [25(OH)D], parathyroid hormone (PTH), and insulin at baseline and week 48 in central laboratories. Insulin resistance was assessed by Homeostatic Model Assessment of Insulin Resistance (HOMA-IR),[23] Quantitative Insulin Sensitivity Check Index (QUICKI)[24] and Insulin Resistance index,[25] and metabolic syndrome using the Adult Treatment Panel diagnostic criteria.[26]


The primary endpoint was change in total hip BMD between the two study arms at week 48. Additional prespecified secondary endpoints were changes in BMD of the lumbar spine and femoral neck, and markers of bone turnover and kidney function through week 48. We conducted post hoc exploratory analyses of weight gain at 12, 24, 48 weeks and, in non-diabetic participants with fasted blood samples, changes in insulin resistance and prevalence of metabolic syndrome at baseline and week 48.

Statistical Analysis

Analyses were by intention-to-treat. Baseline characteristics were summarized by randomization group as means and standard deviations (continuous normally distributed variables), medians and interquartile ranges (IQRs, non-normally distributed variables), and frequencies and percentages (categorical variables). Linear regression was used to estimate the mean difference in BMD from baseline to week 48 between the two study arms, with adjustment for age, ethnicity, time on TDF, BMD and body mass index (BMI) at baseline. Baseline demographics were included as a covariate to minimize the impact of differences in baseline characteristics between the study arms on outcomes.[27] Missing observations were imputed regardless of the reason(s) they were missing. Predictive mean matching (with five nearest neighbours assuming unobserved measurements were missing at random) was used to impute primary and secondary outcomes.[28] Secondary longitudinal outcomes, measured at baseline and weeks 12, 24 and 48, were analysed using repeated measures mixed-effects models with an unstructured variance-covariance matrix to assess mean difference between study arms.

The study was not powered to assess virological efficacy. To assess efficacy, an United States Food and Drug Administration snapshot analysis was performed at week 48. Virological failure was defined as HIV RNA ≥ 50 copies/mL on two consecutive visits. All analyses were conducted using STATA version 12 (StataCorp LLC, College Station, TX, USA).