Risk of Ovarian Cancer Algorithm to Screen for Ovarian Cancer

Abstract & Commentary

By Robert L. Coleman, MD, Associate Professor, University of Texas; M.D. Anderson Cancer Center, Houston, Texas, is Associate Editor for OB/GYN Clinical Alert.

Dr. Coleman is on the speaker’s bureau for GlaxoSmithKline, Bristol-Myers Squibb, and Ortho Biotech.

Synopsis: An OCS strategy using the ROC algorithm is feasible and can achieve high specificity and PPV in postmenopausal women. It is being used in the United Kingdom Collaborative Trial of Ovarian Cancer Screening and in the United States in both the Cancer Genetics Network and the Gynecology Oncology Group trials of high-risk women.

Source: Menon U, et al. Prospective study using the risk of ovarian cancer algorithm to screen for ovarian cancer. J Clin Oncol. 2005;23:7919-7926.

Population-based ovarian cancer screening programs have been difficult to recommend and implement because poor sensitivity and positive predictive value characteristics accompany expensive and inefficient testing methodology and triage algorithms. Menon and colleagues approached this problem by evaluating a new prospectively based algorithm among a population-based screening program currently underway in the United Kingdom. The population cohort used to evaluate the new screening strategy comprised of 13,582 menopausal women 50 years of age or older with at least one ovary, of whom 6532 completed a first screen; the remainder served as controls. The screening strategy was a staged process by which each CA125 drawn underwent a calculation for risk of ovarian cancer. The calculation is based on the patient’s age and CA125 value relative to their personal baseline. In this trial, estimated risk less than 1 in 2000 was considered normal, while a risk of greater than 1 in 500 was considered elevated; those in between were considered intermediate and required repeat testing.

Those not considered normal were referred for a second stage of screening which incorporated a transvaginal ultrasound (TVS) and a repeat CA125. TVS was considered normal, abnormal, or equivocal based on ovarian volume and morphology. Based on the combination of CA125 risk estimation and TVS, a follow-up recommendation was made which could be gynecologic oncology referral, repeat CA125 and/or TVS testing, or annual screening. Among the screened group, nearly 80% continued with annual screening; 91 (1.4%) were considered at elevated risk. Among the intermediate group, repeat testing was normal in 92%, leaving 188 (2.9% of initial population) who were to undergo second-stage evaluation. Of the 144 who stayed in the program, 95 were returned to annual screening based on CA125 and TVS findings; 6 were found with non-gynecologic malignancies, 43 were referred to a gynecologic oncologist of whom 27 were returned to annual screening and 16 women who underwent surgery. From this group, 5 ovarian cancers were identified (4 malignant epithelial and 1 borderline); 11 remaining women had benign ovarian neoplasms. Compared to Menon et al’s previous algorithm based on flat CA125 values (normal ≤ 30 U/mL), the new process referred less than half to secondary screening. They concluded that the new algorithm increases screening precision. Its effect on cost and survivorship are to be determined when the trial completes its accrual of 200,000 women anticipated in 2011.


Most clinicians even remotely familiar with the story of ovarian cancer screening have at best mixed emotions for its feasibility. This has been rooted largely in the repeated lack of overwhelming success of a number of different strategies even among "high-risk" populations. While the potential for a major impact in the dismal outcome of newly diagnosed ovarian cancer patients looms, enthusiasm is tempered by the frequent observation that a large proportion (1 in 20 to 1 in 5) of patients with positive screening need to undergo surgery to identify cancer—most of which are already metastatic. The principal reason for the underwhelming success of these programs is the low prevalence of the disease in the general population. This places a high burden on individual testing characteristics. Coupled with the fact that a reproducible and detectable preinvasive state has yet to be delineated makes the prospect for perfect success improbable. In addition, population-based screening demands that the tool or tools utilized are inexpensive and acceptable to the population for widespread utilization. Even a standard test such as a CA125 or a TVS is too expensive to use routinely in all women for this purpose. However, the impact must not be understated, as therapeutic advances over the last 20 years have not significantly altered mortality due to disease. A good example of a model and its potential impact is the Pap smear, now chiefly recognized for reducing the mortality due to cervical cancer.

Fortunately, this track record has not kept investigators from studying new and existing technologies with the intent of impacting outcome. Several recent advances in biomarkers, proteomics, genomics and risk identification have started to mature into exploratory projects with the intent of being applicable to a mass audience. These tools would help to identify patients at risk for future cancer development based on particular signals. While exciting and promising in their initial evaluation, they are undergoing validation and are currently not quite ready for "prime-time."

While prediction of disease in otherwise asymptomatic healthy patients is an ultimate goal, the differential in survival between early and advanced-stage ovarian cancer makes even identification of localized cancer a "win" scenario. This has been the focus over the last 20 years from the UK team and their collaborators. In the current report, the positive predictive value (ie, the proportion of positive screened patients in whom cancer was subsequently diagnosed) of this new algorithm is still about 1 in 5 patients taken to surgery, the number of patients referred onto secondary and tertiary evaluation is being reduced, making the overall cost for cancer detection lower. Based on this group’s previous work, identified cases are in a different stage distribution than those identified by conventional means and have better long-term survival. The cohort reported in this trial is a subset of the many thousands of patients currently being screened in the ongoing prospective clinical trial so, in a sense, is only a taste of how the methodology has been improved. We can only hope that this will translate into lasting benefit for our patients.

Suggested Reading

  1. Jacobs I, et al. Prevalence screening for ovarian cancer in postmenopausal women by CA125 measurement and ultrasonography. BMJ. 1993;306:1030-1034.
  2. Jacobs IJ, et al. Screening for ovarian cancer: a pilot randomised controlled trial. Lancet. 1999;353:1207-1210.
  3. van Nagell JR, Jr., et al. The efficacy of transvaginal sonographic screening in asymptomatic women at risk for ovarian cancer. Gynecol Oncol. 2000;77:350-356.
  4. Mills GB, et al. Future for ovarian cancer screening: novel markers from emerging technologies of transcriptional profiling and proteomics. J Natl Cancer Inst. 2001;93:1437-1439.
  5. Zhu W, et al. Detection of cancer-specific markers amid massive mass spectral data. Proc Natl Acad Sci USA. 2003;100:14666-14671.