The sensitivity of the test ranged from 98% in ovarian cancer to 33% in breast cancer, but the specificity was greater than 99% with only 7 of 812 control participants having a positive result. “We could not be certain that the few ‘false positive’ individuals identified among the healthy cohort did not actually have an as-yet undetected cancer, but classifying them as false positives provided the most conservative approach to classification and interpretation of the data,” the authors wrote.
Based on cancer stage, sensitivity for stage I cancers was 43%, for stage II 73%, and for stage III 78%. Again, sensitivity varied depending on cancer type, with 100% sensitivity for stage I liver cancer and 20% sensitivity for stage I esophageal cancer.
When tumor tissue samples from 153 patients with statistically significant ctDNA levels were analyzed, identical mutations were found in the plasma and tumor in 90% (138) of all cases.
The protein markers in the CancerSEEK test might also be able to anatomically locate malignancies. Using machine learning to analyze patients testing positive with CancerSEEK, the results narrowed the source of the cancer to two possible anatomical sites in approximately 83% of patients and to one anatomical site in approximately 63% of patients. Accuracy was highest for colorectal cancer and lowest for lung cancer.