Plasma lactate measurements are typically performed in real time, limiting their usefulness in multicenter or longitudinal studies. To determine the stability of lactate specimens, blood was drawn in sodium fluoride/potassium oxalate tubes from 13 volunteers before and after 5 min of handgrip exercise to intentionally increase lactate concentrations. Plasma was stored at -70 degrees C. Aliquots were assayed in real time and after 1, 3, 6, 9, 12, 18, and 24 months. Real-time lactate concentrations measured at baseline ranged from 0.52 to 2.23 mmol/L before and from 2.91 to 11.04 mmol/L after handgrip exercise. Using a linear mixed model, the estimated change from baseline at month 24 was 1.67% (95% confidence interval, -0.70% to 4.03%) for pre-exercise samples and 0.39% (95% CI, -1.13% to 1.91%) for post-exercise samples. Stored serial specimens from 232 HIV-infected subjects in a multicenter trial of antiretroviral therapy were also assayed centrally. Among those, median plasma lactate increased from baseline to 64 weeks by 0.4 mmol/L with zidovudine+lamivudine treatment and by 0.6 mmol/L with didanosine+stavudine (each p<0.001 from baseline; p=0.04 for difference between groups over time). When performed as in this study, frozen storage with central batch lactate analysis is appropriate for prospectively collected samples in multicenter trials.