Risk of diagnosis of ovarian cancer after raised serum CA 125 concentration: a prospective cohort study

SUBJECTS

In all, 22 000 volunteers were registered with the study between 1 June 1986 and 1 May 1990. Volunteers were recruited via media publicity and several companies' occupational health departments and from the age-sex register of collaborating general practitioners in England, Scotland, and Wales. Eligibility criteria for the study were age >/=45 years and postmenopausal status (more than one year of amenorrhoea). Exclusion criteria were bilateral oophorectomy and ovarian cancer. After the prevalence screen 21 961 volunteers remained eligible, of whom 11 085 selected by computerised random number allocation were invited to participate in annual incidence screens for a further three years. Of these women, 9558 (86.2%) were willing to participate, and on review of their returned datasheets 9344 (84.3%) were still eligible to do so. The 10 876 volunteers who were not randomly selected for further screening were contacted by post, and those who were eligible and did not wish to withdraw from the study (10 120) were followed up annually by postal questionnaire. Women randomised to annual incidence screening were followed up with the questionnaire in years when they did not attend for screening.

CA 125

Venepuncture was performed either in the ovarian cancer screening unit at The Royal London Hospital or in an occupational health department or general practice collaborating with the study. Blood samples taken outside the unit were sent by first class post. All CA 125 assays were performed in the laboratory at the screening unit. Blood samples were centrifuged at 3000 rpm for 10 minutes and serum separated in 300 μl aliquots for storage at −20°C. CA 125 radioimmunoassays were performed with a commercial kit (Abbott Laboratories, Chicago) according to the manufacturer's instructions. The selection of a CA 125 cut off was a compromise between the objective of high sensitivity and the limits of assay performance. A cut off at 30 U/ml was chosen as this was the lowest CA 125 concentration at which satisfactory assay reproducibility could be achieved at the outset of the study.

SCREENING TESTS

The screening strategy has been described previously.5 Briefly, the primary screening test entailed venepuncture for serum CA 125 measurement: volunteers with a concentration of >/=30 U/ml were recalled for ultrasonography. If ovarian volume was abnormal (>8.8 ml) referral for surgery was arranged. Details of the results of the multimodal screening strategy will be reported separately. This report relates only to the serum CA 125 results measured at the prevalence screen and at the three or fewer annual incidence screens.

CANCERS

Volunteers were contacted annually either through attendance for screening or by postal questionnaire. The questionnaire asked for information about any hospital attendance during the previous year. When a reply suggested a possible gynaecological malignancy further enquiries were made with the general practitioner and hospital involved and a surgical and a histopathological report obtained. Index cancers were defined as primary invasive epithelial carcinomas of the ovary and fallopian tube. Fallopian tube cancers were included for several reasons. Firstly, stage for stage these tumours have a similar prognosis to epithelial ovarian cancers. Secondly, evidence exists that advanced stage fallopian tube carcinomas may be misdiagnosed as ovarian cancers, and these cancers would therefore be classified as false negative cases. Thirdly, we have previously reported the detection of fallopian tube carcinomas using this screening protocol.5 10

ANALYSIS

The expected risk of index cancers was calculated from cancer incidence tables of the Office of Population Censuses and Surveys (now the Office for National Statistics)11 on the basis of the age distribution of the study population as previously described.5 The observed risk of index cancers among all study participants measured from the date of the prevalence screen was calculated by dividing the number of index cancers that had developed after a given time interval by the total number of individuals at risk. The risk of index cancers at time interval t after a serum CA 125 concentration satisfying a given criterion (for example, >/=30 U/ml) was calculated from (a) the number of index cancers developing within time t of a serum CA 125 concentration satisfying the criterion, divided by (b) the total number of individuals with any serum CA 125 concentration satisfying the criterion. When individuals who developed an index cancer had more than one CA 125 screen result the earliest result that satisfied the criterion was used in this calculation.

Cumulative risk curves were constructed by plotting the calculated risk values against time. The relative risk of cancer at a given time interval was calculated from (a) the calculated risk value for a specified CA 125 criterion (for example, >/=30 U/ml) at time t, divided by (b) the observed risk for the entire population at time t. Confidence intervals for the log risk estimates and log relative risk estimates were calculated using the Taylor series method.12 Sensitivity was defined as the percentage of volunteers known to have developed an index cancer after a stated length of follow up after their last screen who were screen positive. Specificity was defined as the percentage of volunteers without an index cancer who were screen negative. Positive predictive value was defined as the percentage of screen positive volunteers who had an index cancer.

RISK FACTORS FOR OVARIAN CANCER

Raised serum CA 125 concentration at a screen in this study was associated with a substantially increased risk of developing ovarian cancer. The risk after a moderately raised concentration (>/=30 U/ml) was one order of magnitude greater than after a concentration <30 U/ml, while the risk after a concentration >100 U/ml was more than two orders of magnitude greater. The increased risk for diagnosis of ovarian cancer occurred largely within a year of the raised serum CA 125 concentration. Several factors associated with risk of ovarian cancer have been identified by epidemiological and genetic studies. These include age, parity, oral contraceptive use, family history, and mutation of the BRCA1 gene. The increase in risk associated with age, parity, and a family history involving one relative is small and probably no greater than a threefold to fourfold increase in lifetime risk. The level of risk conferred by a strong family history of ovarian cancer is much greater. For example, in a family with two affected first degree relatives the probability that ovarian cancer is due to autosomal dominant inheritance of a gene such as BRCA1 is 66%. The daughter of an affected relative therefore has a 33% probability of inheriting the predisposing genetic abnormality, and as the penetrance of BRCA1 for ovarian cancer is 44%13 her lifetime risk of ovarian cancer is approximately 0.15. This lifetime risk of ovarian cancer is similar to the level of risk documented in our study during the year immediately after a serum CA 125 concentration of >/=100 U/ml (0.149). These results establish that raised serum CA 125 concentration is a powerful index of ovarian cancer risk in postmenopausal women.

Two features of the study design should be emphasised. Firstly, the study involved asymptomatic and postmenopausal women. The prevalence of conditions causing false positive raised serum CA 125 concentrations is higher in premenopausal women and the incidence of ovarian cancer lower than in postmenopausal women. For these reasons the level of risk observed in postmenopausal women is unlikely to apply to premenopausal women with a raised serum CA 125 concentration. Secondly, this study was an interventional screening study. The study design explains the observation that the increase in the risk of diagnosis of ovarian cancer after a raised serum CA 125 concentration was largely within a year of the screen. If the study had been observational the increase in risk would probably have spread over several years. This interpretation is supported by the shape of the curves for expected and observed cumulative risk (fig 1). The total number of observed cases seven years after screening was close to the number expected, but because of the lead time of detection by screening, the slope of the observed cumulative risk was steeper initially. Surgical intervention also explains the lower mean number of screens in volunteers who developed an index cancer (1.5) compared with the rest of the population (2.2) as no further screens were performed after diagnosis of an index cancer.

IMPORTANCE OF RAISED CA 125 IN NON-OVARIAN DISEASE

The importance of a raised serum CA 125 concentration for subsequent diagnosis of disease other than ovarian cancer has not been well documented to date. Of the 767 volunteers with a serum CA 125 concentration >/=30 U/ml, only 49 developed cancer of the ovary or fallopian tube. We have previously reported that on serial sampling, serum CA 125 concentrations in most individuals with initially raised concentrations fall to <30 U/ml.14 A minority of postmenopausal women, however, have persistently raised concentrations. Not all of these women will develop ovarian cancer, and the cause of the raised concentration in those who do not is currently unclear. The volunteers with one or more serum CA 125 concentrations >/=30 U/ml are therefore being followed up closely to establish whether raised concentration is associated with subsequent development of benign or malignant disease other than an index cancer. There is no doubt that raised serum CA 125 concentration occurs in the advanced clinically apparent stages of several malignancies—for example, breast, lung, and colorectal cancers—but whether raised concentrations can indicate the preclinical stages of these cancers is uncertain. We have not yet identified an association between a raised serum CA 125 concentration and the preclinical stages of any cancers apart from those of the ovary and fallopian tube.

The role of serum CA 125 measurement in screening for ovarian cancer is not yet established. It is clear that serum CA 125 concentration is raised before clinical diagnosis in a significant proportion of ovarian cancers and that a raised concentration is associated with a high risk of ovarian cancer. Satisfactory specificity can be achieved by combining CA 125 with ultrasonography in a sequential screening strategy.4 5 Recent evidence suggests that it may also be possible to improve sensitivity by estimating the risk from the temporal pattern of CA 12515 and by the use of tumour markers complementary to CA 125.16 Ultimately, the value of CA 125 in screening will depend on whether the high risk associated with raised serum CA 125 concentration occurs sufficiently early in the course of the disease for effective intervention to be possible. This issue will be resolved only when the results of the large scale randomised controlled trials that are under way in our own unit and elsewhere are available.

We are grateful for the help of Mrs Mary Butcher throughout this study and for the cooperation of British United Provident Association (BUPA) and the occupational health departments at Marks and Spencer and the National Westminster Bank. The study would not have been possible without the collaboration of the following general practitioners: Drs V Bali (Rhondda), D Bannatyne (Harrogate), M E Barnard (Edgware), J Blacklin (Greasby), C J Browne (Tipton), C E Cock (Choppington), M Crick (Hemsby), M Davies (Cookham), V M de-Hoxar (Heswall), G Dowling (Biggin Hill), E Eaton (Markfield), J C Ewart (Bletchley), C Gilbert (Colchester), C Goldwyn (Twickenham), M C N Henchy (Bretton), N Higson (Hove), R D Last (Somerset), R Liesching (Great Holm), A Martin (Ayr), M Monks (Warrington), N Naidoo (Oxford), I Nelemans (Bournemouth), J M Phillips (Swindon), D J Pilling (Blackpool), D C Rawlings (Coleford), C D Sansom (Thingwall), L Singer (Southend), A R Snead (Shropshire), A Thake (Handsworth), and G J Tyler (Letchworth).