The effect of variation of applicator geometry on classic parameters of dose specification was studied in 90 Fletcher-Suit radium applications in women with cervical cancer. Five statistically significant factors were identified by multiple linear regression analysis and quantified: mg of radium in the colpostats, mg of radium in the tandem, lateral displacement of the colpostats in the frontal plane, vertical separation between the colpostat and tandem sources, and anterior-posterior displacement of the colpostats relative to the tandem. These factors were used to predict "traditional" (i.e., bladder, rectum, and point A) and ICRU dose specification parameters as a function of applicator geometry. Applicator geometry had little effect on the product of ICRU volume specification components, but it greatly affected the individual ICRU components and "traditional" calculation points. Linear regression equations were generated to predict the magnitude of such changes so that the effects of variations in applicator geometry at various prescription points could be predicted radiographically, with subsequent confirmation by computerized dosimetry. Although such regression equations do not replace 3-dimensional computerized dosimetry, they aid applicator placement and treatment planning. © 1990.