The Rate of Bone Loss Slows After 1–2 Years of Initial Antiretroviral Therapy

Final Results of the Strategic Timing of Antiretroviral Therapy (START) Bone Mineral Density Substudy

A Carr; B Grund; AV Schwartz; A Avihingsanon; S Badal-Faesen; JI Bernadino; V Estrada; A La Rosa; PWG Mallon; S Pujari; D White; N Wyman Engen; K Ensrud; JF Hoy


HIV Medicine. 2020;21(1):64-70. 

In This Article


Participant Characteristics

The present analysis included 411 (96.9%) of the 424 BMD substudy participants (immediate group, n = 201; deferred group, n = 210); we excluded 13 participants with no analysable DXA scan at baseline or during follow-up (Figure S1). Baseline characteristics of these two groups were well matched (Table S1). The racially diverse population had a median age of 32 years; 26% were female and 80% non-Caucasian.

Participants were followed for a mean 4.5 [standard deviation (SD) 0.5] years; data completeness was > 90% at each visit (Table S2). In the immediate group, > 96% used ART each year up to year 5. In the deferred group, 16%, 29%, 58%, 84% and 94% used ART at years 1–5, respectively (Figure S2). Initial ART contained TDF for 82.8% of participants, efavirenz for 78.1% and a protease inhibitor for 13.0%. No tenofovir alafenamide (TAF) was used.

Changes in BMD

Mean per cent changes in BMD from baseline and annual rates of BMD change at the spine and total hip are summarized in Table 1. BMD had declined more in the immediate than in the deferred group by year 1 [estimated difference −1.7% (95% confidence interval (CI) −2.3, −1.2) at the spine and −1.6% (95% CI −2.2, −0.9) at the total hip; Figure 1a,b). BMD values in the immediate and deferred groups had largely converged by years 3–4 as deferred group participants progressively initiated ART. In the immediate group after year 1, rates of change in BMD were stable at the spine, but continued to decline by about 0.5% per year at the total hip.

Figure 1.

Mean per cent change in bone mineral density (BMD) with 95% confidence intervals (CIs) from baseline to follow-up. Panels (a)–(c) show the intent-to-treat comparison between the immediate and deferred antiretroviral therapy (ART) groups. In panels (d)–(f), follow-up in the deferred group was censored at ART start, and in the immediate group, participants who did not start ART in the first year were excluded. (a, d) Lumbar spine; (b, e) total hip; (c, f) femoral neck. *In the immediate group, five participant who did not start ART in the first year were excluded.

When follow-up was censored at ART start in the deferred group, the immediate minus deferred treatment difference at year 1 was −2.0% (95% CI −2.6, −1.3) at the spine, and −2.0% (95% CI −2.6, −1.4) at the hip. After year 1, the annual rates of BMD change in the immediate group and those remaining off ART in the deferred group were not significantly different (Figure 1d–f).

In subgroup analyses, the treatment difference in spine BMD was larger among those with baseline CD4 counts ≥ 650 cells/μL compared with those with lower CD4 counts (−1.7% and −0.3%, respectively; P = 0.008 for heterogeneity). BMD losses were less steep among smokers than among nonsmokers (spine −0.7% and −2.2%, respectively; hip −0.7% and −2.8%, respectively; P = 0.05 for heterogeneity at both the hip and spine) (Figure S3). There was no evidence for heterogeneity of the treatment effect among subgroups defined by age, sex, race, geographical region, baseline HIV RNA level and TDF in the pre-specified ART regimen.

Predictors of BMD Decline

In the immediate ART group, no baseline variable consistently predicted greater loss of BMD at both the hip and spine (Table S3). Higher HIV viral load was associated with steeper decline of total hip BMD. A higher CD8 count was associated with steeper decline of spine BMD, and smoking was associated with steeper BMD loss at the hip. Within the deferred group (censored at ART start), a lower CD4 count was associated with steeper BMD decline at the spine and femoral neck (Table S4).

Incidence of Fractures

In the parent START study, 182 (3.9%) of the 4684 participants experienced a fracture: 91 (0.86 per 100 person-years) in the immediate group and 91 (0.85 per 100 person-years) in the deferred group (hazard ratio 1.01; 95% CI 0.76, 1.35; P = 0.98) (Table S5, Figure S4).