The Quiet Malignancy with a Loud Blood Signal

Ovarian cancer kills approximately 13,000 Americans each year. Caught while still localized, the five-year survival rate is 92 percent. Once it has spread to distant organs, fewer than 32 percent of patients survive five years.¹ Yet only 20 percent of ovarian cancers are diagnosed at the localized stage. The disease is called a silent killer, but its blood signature is not silent at all.

92% survival when caught early, 32% once spread. Only 20% are found early. The blood signal exists. The question is whether anyone is reading it.

The Signature in the Blood

The most specific pre-diagnostic blood signal for ovarian cancer is thrombocytosis. A New England Journal of Medicine study of 619 patients established the mechanism: ovarian tumors secrete interleukin-6 (IL-6), which stimulates hepatic thrombopoietin synthesis, causing paraneoplastic platelet elevation.² A prospective cohort of 160,000 United Kingdom patients found that thrombocytosis (>400 × 10⁹/L) on a routine complete blood count (CBC) was associated with ovarian cancer risk increasing 23-fold, with platelet counts rising slowly over 18 months, then sharply in the 6 months prior to diagnosis.³

The CA-125 trajectory compounds the signal. The Risk of Ovarian Cancer Algorithm (ROCA) demonstrated in a trial of 202,638 postmenopausal women that longitudinal CA-125 velocity—the rate of change, not just the absolute level—detected significantly more cancers than a single-threshold rule, based solely on serial routine blood draws.⁴

The Machine Learning Case

A random forest classifier trained on 32 routine pretreatment blood parameters in 334 epithelial ovarian cancer patients achieved an accuracy of 92.4 percent and an area under the curve (AUC) of 0.968 for distinguishing ovarian cancer from benign tumors.⁵ A second multicenter study on 11,000 patients trained 20 AI classifiers on 52 laboratory variables, achieving AUC 0.949 (internal) and 0.882 to 0.884 (external validation), outperforming CA-125 and HE4 individually, especially for early-stage disease.⁶

What This Proves

Ovarian cancer’s blood signature is among the most extensively studied and algorithmically validated in this series. The paraneoplastic thrombocytosis mechanism is established, the longitudinal CA-125 trajectory is validated in a 200,000-person trial, and machine learning models achieve near-diagnostic accuracy from routine blood inputs. The science is established, but the primary care screening algorithm has not yet been deployed.

Endnotes

1.National Cancer Institute. “Cancer Stat Facts: Ovarian Cancer.” SEER, 2024.

SEER data reporting five-year relative survival of 91.7% for localized ovarian cancer versus approximately 31.8% for distant-stage disease, with only about 20% of cases diagnosed at the localized stage. Establishes the critical detection gap that routine blood surveillance could address.

2.Stone RL, Nick AM, McNeish IA, et al. “Paraneoplastic Thrombocytosis in Ovarian Cancer.” N Engl J Med. 2012;366(7):610–618.

NEJM study of 619 epithelial ovarian cancer patients establishing that tumor-derived IL-6 stimulates hepatic thrombopoietin synthesis, causing paraneoplastic thrombocytosis. Anti-IL-6 treatment reduced platelet counts in both mouse models and patients, confirming the mechanistic basis of the CBC platelet signal.

3.Giannakeas V, Narod SA. “Incidence of Cancer Among Adults With Thrombocytosis in Ontario, Canada.” JAMA Netw Open. 2021;4(8):e2120633.

Ovarian cancer risk was increased 23.3-fold in patients with thrombocytosis on routine CBC. Platelet counts showed a slow rise for 18 months then a rapid rise in the 6 months before diagnosis, establishing the temporal pattern directly applicable to a CBC-based screening algorithm.

4.Menon U, Ryan A, Kalsi J, et al. “Risk Algorithm Using Serial Biomarker Measurements Doubles the Number of Screen-Detected Cancers Compared With a Single-Threshold Rule in the United Kingdom Collaborative Trial of Ovarian Cancer Screening.” J Clin Oncol. 2015;33(18):2062–2071.

In 202,638 postmenopausal women in the UKCTOCS trial, the ROCA algorithm using serial CA-125 velocity detected significantly more cancers than a single 35 U/mL threshold rule. Demonstrates that longitudinal trajectory of a routine blood marker is the key algorithmic input, not a static value.

5.Kawakami E, Tabata J, Yanaihara N, et al. “Application of Artificial Intelligence for Preoperative Diagnostic and Prognostic Prediction in Epithelial Ovarian Cancer Based on Blood Biomarkers.” Clin Cancer Res. 2019;25(10):3006–3015.

Random forest classifier on 32 routine blood parameters in 334 ovarian cancer patients plus 101 benign controls achieved 92.4% accuracy and AUC 0.968 for cancer versus benign distinction. Demonstrates that routine blood inputs alone yield near-diagnostic machine learning performance for ovarian cancer.

6.Cai G, Huang F, Gao Y, et al. “Artificial intelligence-based models enabling accurate diagnosis of ovarian cancer using laboratory tests in China: a multicentre, retrospective cohort study.” Lancet Digit Health. 2024;6(3):e176–e186.

Multicenter study of 11,000 patients using 20 AI classifiers on 52 laboratory variables. AUC 0.949 (internal) and 0.882–0.884 (external validation), outperforming CA-125 and HE4 individually for early-stage ovarian cancer diagnosis from routine blood inputs.