OBJECTIVE: To compare BMD changes over 96 weeks in adults virologically failing standard first-line therapy, randomised to raltegravir plus lopinavir/ritonavir (RAL+LPV/r) or conventional 2-3 N(t)RTIs+LPV/r second-line therapy. METHODS: Participants underwent DXA at baseline, weeks 48 and 96 to measure total-hip and lumbar-spine BMD. Analyses were adjusted for gender, BMI and smoking. Linear regression was used to compare between-group differences, logistic regression for low BMD (hip or spine Z-score≤-2) incidence, and multivariate linear regression to determine predictors of BMD change. This work represents the extension and final results of the previously published initial 48 weeks of the study. RESULTS: The population included 210 adults from 5 middle-income countries: 52% females, 52% Asians, 43% Africans, mean (SD) age 39(8) years. In the 2-3N(t)RTI group (vs. RAL), BMD reduction was greater at the spine (mean change -4.9% vs. -3.5%; adjusted difference -1.9%; 95%CI -3.3 to -0.5%; p=0.009), and hip (-4.1% vs. -2.2%; -1.9%; 95%CI -3.4, -0.4%; p=0.012). BMD decrease was greatest at 48 weeks with stabilisation to week 96. Overall, low BMD occurred in 15 (7.9%) participants, with no between-group differences. Independent predictors for bone loss included lower BMI (regression coefficient: hip -0.18%, spine -0.26% per 1 kg/m), longer tenofovir exposure (hip: -0.74%, spine: -1.0% per year), greater change in CD4+ to week 12 (hip: -5.11% per 10-fold higher), and higher baseline HIV-RNA (spine: -0.7% per 10-fold higher). CONCLUSION: Over 96 weeks, there was greater BMD decrease with 2-3N(t)RTI+LPV/r compared to RAL+LPV/r; the relative decrease at the spine was greater than the hip. BMD decreases with second-line ART largely occurred in the first 48 weeks with stabilization, but not recovery, thereafter.