To examine the effect of the exposure measures C(max) (peak exposure), AUC(E) (early exposure) and AUC (total exposure) on the bioequivalence of two sustained release formulations of bupropion (i) in the fasted state and (ii), after a high fat meal. The ratio C(max)/AUC (sensitive to rate of absorption) was also evaluated. A two-formulation, two-sequence, four-period replicate design study was performed in 29 healthy men and women after an overnight fast. Similarly, a two-period study was performed in 20 healthy men and women after ingestion of a high fat breakfast. Plasma concentrations of bupropion were measured by HPLC/MS/MS and the data were analyzed (SAS PROC MIXED) by the Schuirmann-Sattersthwaite procedure (four-period study) and by the two one-sided test procedure (SAS PROC GLM) (two-period study). Standard bioequivalence limits of 80 - 125% were applied to all measures including AUC(E) and C(max)/AUC. In the fasting study, the mean plasma concentration vs. time curves from (including over the first 24 hours) following the two administrations of each formulation were similar although there was a significant difference in median t(max) between formulations. This may have contributed to a low estimate of geometric mean ratio (GMR) for AUC(E) (69%) which was judged to have failed bioequivalence. There was also rather low GMRs for Cmax (88%) and C(max)/AUC (89%) but these measure passed because the within-subject variabilities (WSV) were relatively low (19.6% and 11.2%, respectively). Total exposure (AUC(last)) met standard bioequivalence limits of 80 - 125% easily. The raw data from the two-period fed also showed differences in the shapes of the plasma concentration vs. time curves around C(max) although there was no difference in median t(max). The WSV at median t(max) was high (34%) as was the GMR (117%) for AUCE which failed, as did C(max) (GMR 112%). The WSV was very high at early time points before settling into a "plateau" at about 11%. There was no "spike" in the plasma concentration vs. time profiles up to median t(max) or beyond and therefore there was no evidence of dose dumping of the test formulation in either fasted or fed states. No bioequivalence limits have been set for AUC(E) but the application of standard BE limits of 80 to 125% meant that the fed study was clearly underpowered given the high WSV at early time points. More research is needed on the interesting concept of early exposure. The WSV is often high at median t(max) which means that standard bioequivalence limits of 80 - 125% may be inappropriate. Despite the lack of dose dumping, application of AUC(E) to the fasting study, would have resulted in failure to declare bioequivalence since the GMR for this measure was only 69.5%. Application of a 90% confidence interval to AUC(E) to the fed study would have required powering to cope with the fact that this measure was highly variable.