Summary A study was made of the prognostic value of pretreatment measurements of tumour radiosensitivity (surviving fraction at 2 Gy, SF2) in 128 patients with stage I-Ill carcinomas of the uterine cervix undergoing radiotherapy. The median follow-up time was 47 months. In a univariate analysis stratifying patients according to the median value, radiosensitivity was a significant prognostic factor for overall survival, local control and metastasis-free survival. The 5-year survival rate for tumours with SF2 values below the median was 81% and was significantly greater than the rate of 51% for those with SF2 values above the median. In bivariate analyses, SF2 was shown to be independent of disease stage, tumour grade, patient age, colony-forming efficiency and tumour diameter. In a multivariate analysis, radiosensitivity was the most important variable and, after allowing for this, only stage was a significant independent predictor of treatment outcome. These data indicate that, in carcinoma of the cervix treated with radiotherapy, pretreatment tumour intrinsic radiosensitivity is an important prognostic parameter and contributes to prognosis independently of other established and putative parameters.Keywords: predictive assay; intrinsic radiosensitivity; SF2; cervix cancer; radiotherapy There are considered to be three important radiobiological factors that determine how well a tumour responds to radiotherapy: intrinsic radiosensitivity, hypoxia and proliferation. The clinical relevance of these parameters is currently receiving considerable attention, and studies have been published suggesting the potential of all three as prognostic factors for radiotherapy (West, 1994).In carcinoma of the cervix, our own studies have indicated that tumour radiosensitivity is an important determinant of treatment outcome (West et al, 1991(West et al, , 1993. In this work, radiosensitivity is measured using a soft agar clonogenic assay as surviving fraction after 2 Gy in vitro irradiation (SF2). Some support for our finding has come from work on head and neck tumours, which has shown that radiosensitivity, measured using a growth assay as the initial slope of radiation survival curves (a), significantly influenced patient outcome when a high (above median) value was used to stratify data (Girinsky et al, 1994). Other smaller studies using a variety of assays have also suggested that radiosensitive tumours are more responsive to therapy (Hinkley and Bosanquet, 1992;Ramsay et al, 1992;Vaughan et al, 1993;Shibamoto et al, 1994). Studies that have shown no relationship between tumour radiosensitivity and treatment outcome either involved the establishment of cell lines before assay (Allalunis-Turner et al, 1992;Schwartz et al, 1992;Taghian et al, 1993) or treatment with surgery plus radiotherapy (Brock et al, 1992), both of which may be confounding influences.This paper forms an update of a previous study (West et al, 1993) Correspondence to: C West 88 patients. In this report, for the first time, a multivariate analysis has been ...