Nucleoside reverse transcriptase inhibitors (NRTIs) differ in the type and severity of adverse effects resulting from mitochondrial abnormalities. mtDNA in peripheral blood mononuclear cells (PBMCs) was measured during the first 12 months of different NRTIs combinations and its association with clinical lipodystrophy was estimated. Extended follow-up of a randomized trial, ALBI-ANRS 070, including antiretroviral naive patients was conducted. Total DNA was extracted from available cryopreserved PBMCs at baseline and months 6 and 12. Nuclear and mitochondrial genes were amplified using a real-time PCR assay. Clinical lipodystrophy was assessed 30 months after randomization using a standardized questionnaire. A logistic regression analysis assessed the value of mtDNA to predict lipodystrophy. Mean mtDNA level (copies/cell) significantly decreased from 5847 at baseline to 3176 at month 12 (p < 0.0001). In the zidovudine + lamivudine (ZDV + 3TC) arm (n = 37), the mean mtDNA was 6098, 6807, and 3725 copies/cell for baseline, month 6, and month 12, respectively. In the stavudine + didanosine (d4T + ddI) arm (n = 40), the mean values were 5616, 5731, and 2648 copies/cell, respectively. The proportion of patients in the lowest quartile of mtDNA (<1421 copies/cell) at month 12 was higher in 18 patients with lipodystrophy (44%) than in 28 without lipodystrophy (7%) (p = 0.008). At 12 months, a larger reduction of mtDNA from baseline was observed in those started on the d4T + ddI arm. Furthermore, a low mtDNA level at month 12 was associated with the subsequent development of lipodystrophy. This marker may be of value for the early prevention of lipodystrophy in treated HIV-infected patients.