Long-term metabolic and safety profiles of tenofovir alafenamide and tenofovir disoproxil fumarate in ART-naive people living with HIV: a multicenter retrospective study using a mixed-effects model

Tenofovir disoproxil fumarate (TDF) and tenofovir alafenamide (TAF) differ in their effects on renal, bone, and lipid parameters. However, long-term comparative data using robust statistical methods in real-world settings—especially among antiretroviral therapy (ART)-naive people living with HIV (PLWH)—are limited. This study compared the long-term metabolic and safety profiles of TDF versus TAF in ART-naive PLWH, using mixed-effects modeling to inform clinical decisions. A multicenter retrospective cohort study was conducted across 10 HIV centers in Türkiye from 2018 to 2022. The study included 540 ART-naive PLWH aged ≥ 18 years, treated with TDF- or TAF-based regimens for ≥ 12 months, with ≥ 24 months follow-up. Primary endpoints were changes in renal function (eGFR), bone mineral density (T-scores), and lipid profiles over 48 months. Mixed-effects regression models adjusted for baseline demographics, comorbidities, and concomitant medications. Both TAF (n = 197) and TDF (n = 343) showed similar virological suppression and CD4 + recovery, with no significant differences in HIV RNA (p > 0.24) or CD4 + counts (p > 0.54). At 24 months, TAF led to greater increases in total cholesterol, non-HDL cholesterol, LDL cholesterol, and triglycerides (p ≤ 0.039), though differences were not significant at 48 months. Phosphorus levels were similar (p > 0.2), but eGFR declined more with TAF at 48 months (p < 0.001). Overall changes in bone mineral density were comparable (p > 0.18). GGT increases at 24 months were greater with TAF in both sexes; the clinical significance is uncertain and warrants monitoring. Over 48 months, TAF and TDF achieved equivalent virological and immunological responses. In this real-world retrospective cohort, TAF was linked to short-term lipid increases and a greater eGFR decline, while bone density changes were similar overall. Given baseline imbalances, unmeasured confounding (including concomitant medications and regimen components), and multiple testing, these findings should be interpreted cautiously and considered hypothesis-generating. Antiretroviral choice should be individualized.