Specificity of multimodal screening superior to ultrasound alone in ovarian and tubal cancers

Issue: April 10, 2009

The specificity of multimodal screening for ovarian cancer was superior to that of ultrasound screening alone; however, both screening modalities showed encouraging results, according to data from the UK Collaborative Trial of Ovarian Cancer Screening. Combined, the two screening modalities were able to identify 48.3% of invasive cancers while they were stage I/II.

The study included 202,638 postmenopausal women aged 50 to 74 years. Participants were randomly assigned to a control group (n=101,359), multimodal screening with annual CA125 and transvaginal ultrasound (n=50,640) or annual transvaginal ultrasound alone (n=50,639). Repeat tests were performed on participants with abnormal screens; those with persistent abnormal screens underwent clinical evaluation and surgery, when appropriate.

Multimodal ultrasound was performed in 98.9% of the cohort and ultrasound was performed in 95.2% of the cohort.

Repeat tests were required for 8.7% of women in the multimodal screening cohort and 12.0% of those in the ultrasound alone cohort; 0.3% of women in the multimodal screening cohort required clinical evaluation compared with 3.9% of women in the ultrasound alone cohort. Surgery was performed in 0.2% of women in the multimodal screening cohort compared with 1.8% from the ultrasound alone cohort.

Multimodal screening detected 42 ovarian and tubal cancers and ultrasound alone detected 45; 28 were borderline tumors (MMS=8, USS=20). With multimodal screening, for all primary ovarian and tubal cancers, the sensitivity was 89.4%, the specificity was 99.8% and the positive-predictive value was 43.3%. With ultrasound screening alone, sensitivity was 84.9%, specificity was 98.2% and positive-predictive value was 5.3%. – by Stacey L. Adams

Menon U. Lancet Oncol. 2009;doi:10.1016/S1470-2045(09)70026-9.

To date ovarian cancer continues to be the gynecological malignancy with the highest mortality rate. The main reason for this is that approximately 75% of patients are diagnosed at stages IIIc and IV when there is a large amount of tumor or it has spread beyond the intraperitoneal compartment. To date there is no effective screening strategy to identify ovarian cancer in the general population.

The current study represents the largest clinical trial ever attempted to study strategies for ovarian cancer screening, and builds upon more than 15 years of research by the lead and senior authors. They show that screening for ovarian cancer is possible in a large scale (general population that is asymptomatic) even when the strategies are relatively complex in terms of establishing the level of risks and the subsequent reevaluations required. A combined modality that incorporates CA125 serum determinations and ultrasound based on the calculated risk established from the level of CA125 seems to have the best positive predictive value (43.3%). More than 20% positive predictive value is considered to be a good predictor.

The other important finding is that the specificity is very high (99.8%). This is the most difficult figure to achieve in screening studies. It means the ability of a negative test to be truly negative. At the population level this is the more relevant number because these are the people that will be missed since they will be told that there is nothing wrong in their test. Compared to screening based on ultrasound, multimodality screening was associated with a lesser number of surgeries required to find a case of ovarian cancer (2.3 surgeries per case of actual invasive ovarian cancer diagnosed). This has been the major obstacle to ultrasound based screening strategies that will detect benign ovarian lesions and therefore the number of surgeries performed to find one case of actual invasive cancer is too high.

The best study in the United States was the one by Van Nagell (2007) and their rate of surgeries was 9.3 procedures per case of ovarian cancer. The strategy reported in The Lancet is clearly better from this point of view. The other important finding is that it looks like the strategy was able to identify cancers in earlier stages. The goal of any strategy is to identify cancers in early stage when treatment is easier and associated with higher cure or survival rates. They report that 48.3% of the cancers were stages I and II. I calculated that 41.2% were stage I, which is associated with at least 95% survival in five years without the need for postoperative chemotherapy. Of course the main question really is whether the screening strategy is able to reduce mortality from ovarian cancer. This is the main outcome measure for the study and the data is not available yet. Also it must be recognized that there is no data to make comparisons with the control group either. So we are not ready to implement the strategy at any level yet. The other question is whether strategies that incorporate other molecular markers or panels of markers could outperform the sensitivity, specificity and positive predictive value of the strategy utilized by this group. The reality is that this large randomized trial will be difficult to outpace in the near future.

– Wilberto Nieves-Neira, MD

The Cancer Institute of New Jersey
Assistant Professor, Department of Obstetrics, Gynecology and Reproductive Sciences,
Robert Wood Johnson Medical School, New Brunswick, N.J.