That tumors may be composed of cells under various tensions of oxygen, and thus with different radiation sensitivities, has been postulated (1, 2), and the implications of such heterogeneity for the radiotherapy of tumors have been discussed repeatedly (1, 3–5). In these theoretical discussions, it has been pointed out that even a tiny proportion of hypoxic cells can increase the dose of radiation required for cure of tumor, or, conversely, decrease the size of a tumor that can be cured by a given dose of radiation. If one assumes that tumor cure requires sterilization of all the tumor cells (or all but a stated number) and therefore is dependent on (a) the total number of cells, (b) the radiation sensitivity of the most resistant cell, and (c) the proportion of resistant cells, the consequences of the presence of a given fraction of anoxic cells in an otherwise oxygenated population can be calculated (Fig. 1). It has been shown that even as small a proportion as 1 per cent of anoxic cells should increase the dose necessary for cure to nearly that required for a completely anoxic tumor (3–5).

The foregoing postulated effects of the presence of anoxic cells derive from compelling evidence demonstrating a difference in radiation sensitivity between anoxic and oxygenated cell populations. Both in vivo and in vitro cell-survival studies have shown that fully oxygenated cells are about 2.3 times as sensitive to x- or gamma-rays as are identical cells irradiated in the absence of oxygen (6–10). That is, curves relating the logarithm of the fraction of cells surviving to the dose of radiation administered, which typically have the form of a straight line preceded by a shoulder, have terminal slopes that are 2.3 times steeper in the presence of oxygen than in its absence.

Theoretical cell-survival curves that have been calculated for mixed populations of cells are characterized by having two or more exponential slopes rather than a single one (3, 4). Data reported for both in vivo and in vitro experiments have demonstrated only a single exponential component, however, even in those cases in which survival was measured to about the 0.1 per cent level (11–15). In particular, such results have been obtained with tumor cell populations irradiated in situ (7, 8), despite the foregoing postulated inhomogeneities in degree of oxygenation. It may be presumed, therefore, that one or more of the following has been pertinent in these studies:

• a. 
  A particular population may be homogeneous with respect to degree of oxygenation (and any other determinant of radiation sensitivity); that is, the postulate of inhomogeneity is unwarranted.
• b. 
  The overall sensitivity of a cell population that is actually heterogeneous with respect to radiation response may, under certain conditions, exhibit relatively simple survival kinetics, such that the heterogeneity is obscured (16–8).