The interpretation of creatine kinase (CK) flux measured by (31)P NMR magnetization transfer in vivo is complex because of the presence of competing reactions, metabolite compartmentation, and CK isozyme localization. In the isovolumic perfused rat heart, we considered the influence of both ATP compartmentation and ATP-P(i) exchange on the forward (F(f): PCr --> ATP) and reverse (F(r)) CK fluxes derived from complete analysis of inversion transfer. Although F(f) should equal F(r) because of the steady state, in both protocols when PCr (inv-PCr) or ATP (inv-ATP) was inverted and the contribution of ATP-P(i) was masked by saturation of P(i) (sat-P(i)), F(f)/F(r) significantly differed from 1 (0.80 +/- 0.06 or 1.32 +/- 0.06, respectively, n = 5). These discrepancies could be explained by a compartment of ATP (f(ATP)) not involved in CK. Consistently, neglecting ATP compartmentation in the analysis of CK in vitro results in an underestimation of F(f)/F(r) for inv-PCr and its overestimation for inv-ATP. Both protocols gave access to f(ATP) if the system was adequately analyzed. The fraction of ATP not involved in CK reaction in a heart performing medium work amounts to 20-33% of cellular ATP. Finally, the data suggest that the effect of sat-P(i) might not result only from the masking of ATP-P(i) exchange.