The CA125 tumour marker has an established clinical role in monitoring the course of ovarian cancer during and after therapy (Rustin et al, 1996a(Rustin et al, , 1996b. In addition this marker is of value as a prognostic indicator (Gadducci et al, 1995;Nagele et al, 1995) and in the differential diagnosis of pelvic adnexal masses in symptomatic patients (Jacobs et al, 1990;Tingulstad et al, 1996). CA125 is also currently under investigation as a potential screening test for ovarian cancer (Jacobs et al, 1993). Although the role of CA125 in screening remains controversial, there is evidence that serum elevation is associated with an increased risk of ovarian cancer. We have previously reported that asymptomatic postmenopausal women with a CA125 ≥ 30 U ml -1 have a 36-fold increased risk of diagnosis for ovarian cancer in the subsequent year (Jacobs et al, 1996). This data was derived from a multimodal screening study which used pelvic ultrasound as a secondary test. We have now been able to perform a further analysis to quantify the value of pelvic ultrasound in refining the risk of cancer amongst asymptomatic women with CA125 elevation.
METHODS
DesignThe overall study design has been described in a previous report (Jacobs et al, 1988(Jacobs et al, , 1993. A total of 22 000 post-menopausal women, aged ≥ 45 years underwent serum CA125 screening. Women with a CA125 level ≥ 30 U ml -1 underwent an ultrasound. Transabdominal and/or transvaginal ultrasonography was used to measure the diameter of each ovary in three planes and to document ovarian morphology. Ovarian volume was calculated using the formula for an ovoid (Campbell et al, 1989). Ovarian morphology was classified as normal if the ovaries exhibited uniform hypoechogenicity and smooth outlines. All other morphological appearances were classified as abnormal. Women with abnormal results on scan were referred to a gynaecologist for assessment and further management. All women were followed up annually during the study by a questionnaire that specifically enquired about any illness or hospital visit in the previous year. When information suggested that a study participant may have had a gynaecological malignancy, histopathological and clinical information was obtained from the general practitioner and/or hospital.
Statistical analysisIndex cancers were defined as primary invasive epithelial carcinomas of the ovary and fallopian tube (Jacobs et al, 1996). The risk of index cancer at time interval t from the date of ultrasound in volunteers with scans satisfying a particular ultrasound criterion (abnormal ovarian morphology and volume over a specified cutoff) was calculated by dividing the number of volunteers with scans satisfying that criterion who developed index cancers within time t by the total number of volunteers with scans satisfying the criterion. Cumulative risk curves were constructed by plotting the calculated risk values against time. The observed risk of index cancer for all 22 000 study volunteers as well as for all volunteers with CA125 ≥ 30 U ml...