Introduction Ovarian cancer (ovarian-/tubal-/peritoneal cancer) give dull symptoms why early diagnosis is challenging. Endogenous Volatile Organic Compounds (VOC) are products of metabolic activity in cancer and elevated glycolysis leads to increases in lactate, fumarate, and other metabolites. VOC analyses in plasma and urine have shown to indicate early cancer.

Methods With highly sensitive gas sensors, preoperative plasma from 87 women with stage I-IV ovarian cancer was examined and compared to that from 26 healthy control women. Data analyses were performed using feature extraction from 32 gas sensors per sample. The dataset has been processed by principal component analysis (PCA) for dimensionality reduction and feature reduction (9 principal components were kept retaining 95% of the original information in the features-observations dataset). A support vector machine model was then trained towards algorithmic binary classification: positive (cancer) or negative (no cancer). To avoid overfitting while not losing any observations, 5-fold cross validation was used during training of the classification algorithm.

Results The analysis of VOCs revealed positive results in 85 out of 87 ovarian cancer patients, yielding a sensitivity of 97.7% (95% confidence interval [CI] 91.9 – 99.7%). Out of the healthy controls 22 were negative and 4 showed positive results (specificity 84.6% 95% CI 65.1 – 95.6%). Positive predicted value 95.5% (95 CI: 88.9 - 98.8%) and accuracy of 94.7% (95% CI: 88.8 - 98.0%).

Conclusion/Implications VOC analyses in plasma show very high sensitivity to distinguish ovarian cancer patients’ stage I-IV from healthy controls.