Andrew Vickers

PhD

Memorial Sloan Kettering Cancer Center

Attending Research Methodologist

PLCO (Learn more about this study)

2025-0037

Jun 12, 2025

Glycovariants of CA-125 for early detection of ovarian cancer

Two randomized trials have shown that screening with CA-125 is not of value for reducing ovarian cancer mortality. One possible explanation is that changes in CA-125 occur too late in the disease course to improve the outcome of treatment, such as by enabling surgical cure of early-stage disease. There was no difference in the stage distribution with screening in PLCO; in the UKCTOCS randomized trial, the reduction in stage III/IV disease was only 10%, too small to shift mortality even without the purported countervailing increased risk of mortality in early stage patients in the screening arm.

We recently developed assays for two novel glycovariants of CA-125: recombinant human macrophage galactose-type lectin (“Glycovariant L”) and Sialyl-Thomsen-nouveau (“Glycovariant A”). We used data from a population-based observational cohort, the Malmö Diet and Cancer project, to determine whether the CA-125 glycovariants could detect ovarian cancer several years before clinical diagnosis. This does not appear to be the case for conventional CA-125 but would be critical for any marker used in ovarian cancer early detection. Women diagnosed with incident ovarian cancer up to 10 years after blood draw (n=82) were matched 1:3 with controls. Glycovariant A was associated with subsequent ovarian cancer (coefficient 0.53 / log U / mL; 95% CI 0.19, 0.87; p=0.002). There was no association with CA-125 (p=0.8), HE4 (p=0.5) or Glycovariant L (p=0.12). In the primary analysis restricted to cases within 5 years, the area-under-the-curve for Glycovariant A was 0.68; a moving window analysis suggested it retains its predictiveness out to 7 or 8 years. Just over 25% of cases occurred in the top 5% of Glycovariant A levels (2.3 U / ml), with 49% occurring in the top 20% (manuscript in preparation).

We propose to attempt replicating this research using serum from the PLCO. In particular, the PLCO data set can help us address three limitations from the Malmö study. The first is that due to relatively limited power, we were unable to model the relationship between Glycovariant A and risk, such as whether there is a threshold effect. Second, we only had a single sample and so were unable to model changes in markers over time, such as is done for the Risk of Ovarian Cancer Algorithm. Finally, we did not have patient-level data on histology or stage. It could be Glycovariant A increases before a tumor could be detected on further work-up or, alternatively, only after cancer has spread intraperitoneally. Data from PLCO could help us address all limitations and hence provide critical data on the use of CA-125 Glycovariant A for ovarian cancer early detection.

The investigators previously used data from the PLCO in the development of the 4Kscore, a test used in prostate cancer diagnosis in over 75,000 US men per year. In other words, the investigators have a track record of conducting marker studies on PLCO samples and turning the results into markers that are used clinically to improve patient care.

Specific aims

1. To determine whether 0.25 ml of a single serum sample of CA-125 Glycovariant A can predict incident ovarian cancer diagnosed up to 8 years subsequently. This would be a direct replication of our work on the Malmö cohort and use a similar case-control methodology.
2. To determine whether changes in CA-125 Glycovariant A over time over time better predict subsequent ovarian cancer than a single sample. We would create a model analogous to the Risk of Ovarian Cancer Algorithm except that Glycovariant A of CA-125 would replace conventional CA-125 in the model.
3. To determine the shape of the relationship between CA-125 Glycovariant A and the risk of subsequent ovarian cancer. Our initial research suggested that there might be a threshold effect, such that the level of CA-125 Glycovariant A only mattered above a certain threshold. This is unlike, say, PSA, where for two men with low PSAs, the man with the higher of the two has higher risk. The larger number of cases in the PLCO will allow us to better determine the shape of the relationship between CA-125 Glycovariant A and subsequent risk.
4. To determine whether the predictiveness of CA-125 Glycovariant A for subsequent ovarian cancer is modified by stage or histopathology. Knowing whether cases associated with a high CA-125 Glycovariant A are high-grade serous carcinoma, or are earlier vs. later stage at diagnosis, will give insights into its potential for early detection.