Effects of oxygen consumption in photodynamic therapy (PDT) are considered theoretically and experimentally. A mathematical model of the Type II mechanism of photooxidation is used to compute estimates of the rate of therapy-dependent in vivo oxygen depletion resulting from reactions of singlet oxygen (1O2) with intracellular substrate. Calculations indicate that PDT carried out at incident light intensities of 50 mW/cm2 may consume 3O2 at rates as high as 6-9 microM s-1. An approximate model of oxygen diffusion shows that these consumption rates are large enough to decrease the radius of oxygenated cells around an isolated capillary. Thus, during photoirradiation, cells sufficiently remote from the capillary wall may reside at oxygen tensions that are low enough to preclude or minimize 1O2-mediated damage. This effect is more pronounced at higher power densities and accounts for an enhanced therapeutic response in tumors treated with 360 J/cm2 delivered at 50 mW/cm2 compared to the same light dose delivered at 200 mW/cm2. The analysis further suggests that the oxygen depletion could be partially overcome by fractionating the light delivery. In a transplanted mammary tumor model, a regimen of 30-s exposures followed by 30-s dark periods produced significantly longer delays in tumor growth when compared to the continuous delivery of the same total fluence.