The pathogenesis of some components of the lipodystrophy (LD) syndrome might be linked to the use of nucleosides. Earlier reports did not compare treatment regimens according to the nucleoside backbone. We studied a cohort of individuals who did not switch between stavudine and zidovudine. LD was defined to be present if one of three criteria was met: self-report by the patient, observation by an investigator who had known the patient since commencement of highly active antiretroviral therapy (HAART), or examination by a physician masked to therapy. The mean duration of therapy was 101 weeks (range: 26-234 weeks). Overall prevalence of LD was 48.7%. Lipoatrophy and lipohypertrophy occurred in 33.9% and 28.7% of patients, respectively. Logistic regression showed four parameters to be significantly associated with lipoatrophy: HAART longer than 2 years (p =.002, odds ratio [OR] = 4.4, 95% confidence interval [CI]: 1.608-11.965), baseline viral load >100,000 copies/ml (p =.004, OR = 4.3, CI: 1.726-11.197), age >40 years (p =.016, OR = 3.2, CI: 1.247-8.373), and white ethnicity (p =.041, OR = 5.4, CI: 1.070-28.184). Cholesterol levels of >200 mg/dl at baseline were associated with a risk reduction (p =.047, OR = 0.36, CI: 0.130-0.987). Use of lipohypertrophy as a dependent variable resulted in a significant association with HAART duration (p = 0.028, OR = 2.7, CI: 1.2-6.5) and protease inhibitor use (p =.014, OR = 3.8, CI: 1.3-11.2). LD prevalence is similar with both backbones using stavudine or zidovudine. This is the first time that baseline cholesterol was shown to be significantly associated with lipoatrophy.