Researchers advance ovarian cancer detection using cfDNA fragmentomes and protein biomarkers

A new study combines cfDNA fragmentomes and protein biomarkers to vastly improve early detection of ovarian cancer, offering a cost-effective and robust tool for mass screening.

Challenges in ovarian cancer diagnosis

Ovarian cancer causes over 200,000 deaths and 300,000 new cases each year throughout the world. In the United States alone, 19,600 cases and 12,700 deaths are predicted for 2024.

Five-year survival at diagnosis for a woman with stage I invasive epithelial ovarian cancer (iEOC) is 93%, whereas 75% of stage II and IIA1 ovarian cancer cases with regional lymph node involvement will survive five-years after diagnosis. However, in more invasive stages of ovarian cancer, which include stage III other than IIIA1 or stage IV, the five-year survival rate is reduced to 31%.

Detecting early-stage ovarian cancer remains difficult due to nonspecific symptoms and the absence of reliable screening tests. Nevertheless, several studies have suggested that specific biomarkers, including mucin-16, otherwise known as CA-125, and human epididymis protein 4 (HE4), may facilitate the early diagnosis of ovarian cancer.

Various screening models incorporate biomarker measurement with patient age, menopausal status, the Risk of Malignancy Index (RMI), and sonographic criteria like the International Ovarian Tumor Analysis (IOTA) to screen for ovarian cancer. However, these methods are limited in their accuracy, thus preventing their clinical application.

In the current study, researchers examine the cfDNA fragmentome, which is defined as the genome-wide collection of cfDNA fragments in the circulation. The cfDNA fragmentome provides insights into not only the genome but also DNA methylation, chromatin, and transcriptional status in both normal and cancer cells.

Cancer-specific changes

The current study included 94 women with ovarian cancer, 203 with benign adnexal masses, and 182 without known ovarian masses. All study participants were prospectively screened or diagnosed in the Netherlands or Denmark.

For validation, the researchers analyzed samples from 40 and 50 patients with ovarian cancer and benign ovarian masses, respectively, and 22 without known ovarian lesions.

Samples obtained from women with ovarian cancer exhibited significant heterogeneity across and within genomes of individual patients, while those from women without cancer or with benign masses were more homogeneous.

Using machine learning, the researchers analyzed cfDNA fragmentome profiles, changes in chromosome arms, and protein biomarkers CA-125 and HE4. This integrated approach, referred to as DELFI Protein (DELFI-Pro), was also applied in earlier studies to detect lung and liver cancers.

DELFI-Pro generated a score between zero and one to classify ovarian cancer status. This score was not significantly associated with other chronic or cardiometabolic illnesses, nor with age, unless ovarian disease was present.

Detecting ovarian cancer

In women without ovarian disease, the median score was 0.07 as compared to 0.93 in stages I and II and one in stages III and IV ovarian cancer, irrespective of the presence of symptoms and menopausal status.

When stratified by stage, the area under the receiver operator characteristic (AUROC) was 0.96, 0.94, 0.99, and 1.00 for stages I, II, III, and IV, respectively, although the small cohort sizes in later stages should be noted as a limitation.

The area under the curve (AUC) for other ovarian cancers ranged from 0.94 to 0.99, with the lowest AUC observed in clear cell tumors at 0.84. For asymptomatic patients, DELFI-Pro achieved an AUC of 0.99, demonstrating its robustness even in the absence of symptoms.

The ichorCNA test, based on copy number changes, had an AUC of 0.71, compared to only 0.59 for screening based on median cfDNA fragment lengths.

At specificity levels above 99%, DELFI-Pro's sensitivity for stages I, II, III, and IV was 72%, 69%, 87%, and 100%, respectively. For high-grade serous ovarian cancer (HGSOC), DELFI-Pro reached an overall detection rate of 90%, ranging from 83% to 100% across stages. By comparison, CA-125 alone detected significantly fewer cases in stages I to III, with detection rates of 34%, 62%, and 63%, respectively.

These results were validated in the external cohort comprising 62 American patients with different ovarian cancer subtypes or cancer-free.

Distinguishing malignant vs benign disease

Fragmentome profiles were different in patients with malignant as compared to benign ovarian disease. The AUC ranged from 0.82 to one for stages I and IV ovarian cancer, whereas the median score for benign disease was 0.17.

Serous carcinomas and endometrioid cancers had higher AUCs of 0.84-0.96 and 0.91, respectively, as compared to mucinous or clear cell subtypes with AUCs of 0.65 and 0.77.

In the Discovery Cohort, DELFI-Pro scoring, at 80% specificity, identified 95% of HGSOC patients, highlighting its potential utility in distinguishing between benign and malignant masses. A higher DELFI-Pro score was also correlated with larger tumor burden.

At over 99% specificity, DELFI-Pro demonstrated a median positive predictive value (PPV) of 23.6%, compared to 9.2% for all other modalities. The false positive rate was substantially lower at 0.95%, compared to rates between 3.1% and 20.6% for other screening methods.

Conclusion

The approach discussed in the current study performed well in detecting early ovarian cancer, particularly HGSOC, which benefits most from timely treatment.

Journal reference:

• Medina, J. E., Annapragada, A. V., Lof, P., et al. (2024). Early detection of ovarian cancer using cell-free DNA fragmentomes and protein biomarkers. Cancer Discovery. doi:10.1158/2159-8290.CD-24-0393.