The usefulness of radiolabeled nitroimidazoles for measuring hypoxia will be clarified by defining the relationship between tracer uptake and radiobiologically hypoxic fraction. We determined the radiobiologically hypoxic fraction from radiation response data in 36B10 rat gliomas using the paired cell survival curve technique and compared the values to the radiobiologically hypoxic fraction inferred from mathematical modeling of time-activity data acquired by PET imaging of [(18)F]FMISO uptake. Rats breathed either air or 10% oxygen during imaging, and timed blood samples were taken. The uptake of [(3)H]FMISO by 36B10 cells in vitro provided cellular binding characteristics of this radiopharmaceutical as a function of oxygen concentration. The radiobiologically hypoxic fraction determined for tumors in air-breathing rats using the paired survival curve technique was 6.1% (95% CL = 4.3- 8.6%), which agreed well with that determined by modeling FMISO time-activity data (7. 4%; 95% CL = 2.5-17.3%). These results are consistent with the agreement between the two techniques for measuring radiobiologically hypoxic fraction in Chinese hamster V79 cell spheroids. In contrast, the FMISO-derived radiobiologically hypoxic fraction in rats breathing 10% oxygen was 13.1% (95% CL 7.9-8.3%), much lower than the radiobiologically hypoxic fraction of 43% determined from the radiation response data. This discrepancy may be due to the failure of FMISO to identify hypoxic cells residing at or above an oxygen level of 2-3 mmHg that will still confer substantial protection against radiation. The presence of transiently hypoxic cells in rats breathing reduced oxygen may also be under-reported by nitroimidazole binding, which is strongly dependent on time and concentration.