Two randomized studies (with 106 and 157 patients, respectively)

which unfortunately did not include BMD measurements at specific sites did not find differences in whole-body BMD between patients treated with PI-containing regimens and those treated with regimens not containing PIs [18,28]. We were not able to analyse the potential influence of nonnucleoside reverse transcriptase inhibitors (NNRTIs) as this class was represented in both arms. Randomized studies with NNRTI-sparing arms have shown similar or more pronounced BMD decreases following HAART in this arm compared with the NNRTI-containing HDAC inhibitor arm, which argues against the possibility that the process is driven mainly by

NNRTIs [6,17]. The long follow-up allowed us to document not only BMD changes immediately after HAART initiation but also longer term consequences. We were able to measure BMD changes beyond the transient bone remodelling stage that occurs after interventions with an effect on bone metabolism. Long-term follow-up may be crucial for separating changes during the initial remodelling periods from bone changes that may ensue thereafter [29]. We measured BMD at two specific sites known to be valid surrogate markers for fracture risk and for evaluating effects of medical treatment [10]. The randomized design with two class-sparing arms allowed us to examine the influence of two different Selumetinib concentration drug classes on BMD. The evolution of BMD was not the primary endpoint of the study and consequently there are no power calculations. The study included a limited number of patients, although it was comparable in size to other studies of BMD changes [6,18]. A large proportion of patients switched one or more drugs during the study period. In the NRTI-sparing arm in particular, the changes led

to a switch of drug class, and consequently only half of the patients randomized Amine dehydrogenase to the NRTI-sparing arm were still on an NRTI-sparing regimen at the end of the study period. Thus, any specific drug or drug class effects may have been attenuated and not detected by our measurements. However, the on-class analyses did not suggest any detrimental effect of PIs on BMD. It is well known that side effects may not be class specific, and a large randomized study suggested that tenofovir caused a greater initial decline in BMD than stavudine. Similarly, lipoatrophy is mainly ascribed to the thymidine analogues stavudine and zidovudine but not to other NRTIs such as abacavir or tenofovir [7,30,31]. Thus, our results may not be generalizable to other PIs or NRTIs. The on-treatment analyses corroborated the pattern seen in the ITT analyses, indicating that the stabilization of BMD was not caused by switching to drugs less BMD-toxic than lopinavir/ritonavir or zidovudine/lamivudine.