Next steps in the early detection of ovarian cancer

A recent ovarian cancer screening trial found no reduction in mortality, despite increased detection of early stage disease. Here, we discuss these findings and examine next steps to develop more effective approaches for the early detection of ovarian cancer.

While PARP-inhibitor maintenance therapy is generally prescribed for advanced stage (III/IV) disease, patients with screendetected homologous repair deficient early stage (I/II) ovarian cancer might also benefit. Novel agents, including SIK2 inhibitors 3 , are being developed to enhance primary treatment with carboplatin and paclitaxel potentially improving care for both early and late stage disease.
Furthermore, the magnitude of the stage shift observed in the UKCTOCS may not have been sufficient to reduce mortality. The fraction of stage I/II patients in the UKCTOCS increased from 28.4% with no screening to 38.1% with CA125 followed by TVS. The fraction of patients with stage IV disease decreased from 20.7 to 15.1%. A much greater stage shift was observed in the single arm Normal Risk Ovarian Screening Study (NROSS) that has been conducted, in parallel, over the last 19 years in post-menopausal women at average risk in the United States, using the identical twostep multi-modal screening plan with the CA125 based ROCA followed by TVS. Among the 7597 women screened, 16 epithelial ovarian cancers have been detected-2 were borderline and 14 invasive-with 11 (69%) in early stage (I or II) (updated from ref. 4 ). One of 16 cases (6%) was detected in stage IV. Both trials confirmed that adequate specificity could be attained with the twostep strategy, requiring no more than 2-4 operations to detect each case of ovarian cancer. The reason for a greater stage shift in the smaller trial is not clear. This could reflect statistical variation with the smaller size of the NROSS. Difficulties were, however, encountered with TVS imaging in the UKCTOCS. In a retrospective review of 1000 archived cases, ovaries and fallopian tubes could be identified in only 50% of cases 5 . TVS imaging could have been more reliable in the NROSS. Another difference between the trials relates to processing of blood for measurement of CA125. In the NROSS, blood was drawn in glass tubes without gel, serum was separated and frozen on the same day, while in the UKCTOCS blood was drawn in gel separation tubes, shipped at ambient temperature and separated after up to 56 h. A modest systematic reduction in CA125 levels in the UKCTOCS could have decreased the ability to detect early stage disease. In addition, particular care was taken in the NROSS to follow elevations of CA125 with repeated TVS and to minimize time to surgical intervention.
Next steps in the early detection of ovarian cancer. The negative outcome of the UKCTOCS calls into question, where we should go next to find an effective strategy for early detection of ovarian cancer. While some might suspend attempts to detect early stage ovarian cancer awaiting a novel and disruptive technology, the twostep screening strategy has already achieved adequate specificity and a clear stage shift, although sensitivity is not yet adequate. There are opportunities for improvement both in serum biomarkers and in imaging. Only 80% of ovarian cancers express CA125 and serum levels of CA125 are elevated in only 70% of stage I/II cancers. A recent review identified 35 biomarkers that complement CA125 and could potentially improve sensitivity of the initial step in screening 6 . A combination of CA125, HE4, and CA72.4 detects 16% of cases missed by CA125 7 . Through a collaboration sponsored by the NCI Early Detection Research Network (EDRN), CA125 detected 72% of early stage cases at 98% specificity, whereas a combination of CA125, HE4 antigen-autoantibody complexes 8 and osteopontin 9 detected 89% at 94% specificity 10 . A second-generation ROCA algorithm is being developed and can be tested prospectively for specificity in the NROSS cohort.
In addition to detecting a greater fraction of early stage patients, panels of biomarkers could improve lead time with detection of cancers at longer intervals before clinical presentation. Autoantibodies could arise in response to very small volumes of early disease, which would be particularly important for high grade serous lesions arising from the fallopian tube. Anti-p53 autoantibodies have been detected in more than 20% of patients with early and late stage ovarian cancer 11 . Assaying serum samples from the UKCTOCS, titers of anti-p53 autoantibodies increased 8 months before elevation of CA125 and 22 months prior to clinical presentation in patients who did not exhibit increases in serum CA125 12 . This is the first of >120 biomarkers tested by our group that increased lead time over CA125. Among 19 promising autoantibodies tested, anti-p53, anti-CTAG1, and anti-IL-8 detected the greatest fraction of early stage ovarian cancer patients 11 .
A variety of additional biomarkers are being developed to detect ovarian cancer including ctDNA, methylated DNA, and miRNAs. Alterations in cervical and peripheral blood ctDNA can complement CA125 in detecting early stage disease 12 . While ovarian cancer has been included in DNA-based pan-cancer screening strategies 13,14 , detecting stage I/II disease has proven challenging. Future research should optimize the integration of DNA and protein biomarkers.
Whatever the biomarker panel chosen, screening could be performed more frequently. In patients at high risk, largely related to germ-line BRCA1/2 mutations, screening with the ROCA every 3 months proved more effective than annual screening 15,16 , While it is difficult to imagine more frequent screening for ovarian cancer alone in patients at conventional risk, blood might be drawn every six months to screen for multiple cancers in women over 50. Ovarian cancer screening could be paired with DNA-based pan-cancer screening strategies or combined with site-specific blood tests that are being developed to detect colorectal adenomas, and breast and pancreatic cancers 17 .
Imaging, the second step in two-stage screening, poses perhaps the greatest unmet need. As a single modality, TVS lacks adequate sensitivity and specificity for early detection of ovarian cancer. The majority of high-grade serous cancers probably arise in the fallopian tubes. Even in expert hands, fallopian tubes could not be imaged in 23% of 549 healthy women 18 . CT, PET-CT, and MRI also have problems with sensitivity, specificity, exposure to radiation and cost for screening 7 .
One possible solution in patients with rising serum biomarkers and negative TVS is falloposcopy, where a fiberoptic scope is threaded through the uterus and fallopian tube to visualize the fimbriae and ovary. This would be particularly relevant for women with BRCA1/2 mutations who are delaying risk reducing surgery. The EDRN is currently evaluating the feasibility of this approach. Another technology that is being developed is superconducting quantum interference detection (SQUID), which is a sensitive method for detecting faint magnetic fields. Anti-CA125 antibodies have been conjugated with ferritin nanospheres. Only antibody conjugated nanospheres bound to cells are detected by magnetic relaxation. Ex vivo, 10 6 ovarian cancer cells (0.1 mm) can be detected 7 . If uptake of antibody-coated nanospheres can be optimized in xenografts, this approach might be utilized to detect recurrent ovarian cancer and then tested in healthy women with rising biomarkers and negative TVS.

Conclusions
Given the specificity of the two-step screening strategy, opportunities to improve both phases and the impressive stage shift in the CA125-based NROSS trial, further development of this approach appears worthy of pursuit. The potential benefit for ovarian cancer patients is substantial. Computer simulations suggest that an effective strategy for early detection could reduce mortality by 10-30%, a dramatic improvement over our current attempts to improve therapy 7 .