Long-term anti-epileptic drug (AED) therapy is associated with increased fracture risk. This study tested whether substituting the newer AED levetiracetam has less adverse effects on bone than older AEDs. An open-label randomized comparative trial. Participants had "failed" initial monotherapy for partial epilepsy and were randomized to substitution monotherapy with levetiracetam or an older AED (carbamazepine or valproate sodium). Bone health assessments, performed at 3 and 15 months, included areal bone mineral density (aBMD) and content at lumbar spine (LS), total hip (TH), forearm (FA), and femoral neck (FN), radial and tibial peripheral quantitative computed tomography and serum bone turnover markers. Main outcomes were changes by treatment group in aBMD at LS, TH, and FA, radial and tibial trabecular BMD and cortical thickness. 70/84 patients completed assessments (40 in levetiracetam-and 30 in older AED group). Within-group analyses showed decreases in both groups in LS (-9.0 %; p < 0.001 in levetiracetam vs. -9.8 %; p < 0.001 in older AED group), FA (-1.46 %; p < 0.001 vs. -0.96 %; p < 0.001, respectively) and radial trabecular BMD (-1.46 %; p = 0.048 and -2.31 %; p = 0.013, respectively). C-terminal telopeptides of type I collagen (beta CTX; bone resorption marker) decreased in both groups (-16.1 %; p = 0.021 vs. -15.2 %; p = 0.028, respectively) whereas procollagen I N-terminal peptide (PINP; bone formation marker) decreased in older AED group (-27.3 %; p = 0.008). The treatment groups did not differ in any of these measures. In conclusion, use of both levetiracetam and older AEDs was associated with bone loss over 1 year at clinically relevant fracture sites and a reduction in bone turnover.