Large-scale observational prospective cohort study of a multi-cancer early detection (MCED) test in symptomatic patients referred for cancer investigation.

Abstract

1501 Background: Delays in cancer diagnosis can lead to increased mortality, leading to specialized diagnostic pathways for symptomatic patients when cancer is suspected. Identification of circulating tumour DNA can stratify individuals into those more or less likely to have cancer and predict the cancer origin. This could both expedite cancer diagnosis and reduce harm and inefficient investigation in those who do not have cancer. We evaluated the performance of a next-generation sequencing MCED test (GRAIL, LLC) using cell-free DNA (cfDNA) isolated from whole blood in symptomatic patients referred to cancer diagnostic clinics by their primary care physician. Methods: Patients referred for urgent imaging, endoscopy or other diagnostic modalities to investigate symptoms suspicious for cancer were invited to take part. Participants with non-specific symptoms or being tested for lung, gynaecological, upper gastrointestinal (GI) or lower GI cancers gave a blood sample on the day they attended for urgent standard of care investigations. cfDNA was isolated from blood samples and stored until the MCED test was performed in batches, blinded to clinical outcome. The test’s predictions (cancer signal detected yes/no; predicted signal origin) were compared with the diagnosis obtained by standard care to establish the positive and negative predictive value (PPV & NPV), sensitivity and specificity across the whole study population, by referral pathway, and by presenting symptom. In addition, sensitivity was analysed by cancer site and stage. Results: 6238 participants were recruited over 5 months from the 5 clinical pathways at 44 hospital sites in England and Wales. 5461 individuals had an MCED test result and diagnostic outcome and were evaluable. Mean age was 62.1 years (SD, 13.8), 3609 (66.1%) female, 2533 (46.4%) current or former smokers. 368 (6.7%) evaluable patients had a cancer diagnosed. The MCED test detected a cancer signal in 323 cases, 244 in whom cancer was diagnosed and 79 where it was not, yielding a PPV of 75.5% (95% CI 70.5-80.1%), NPV 97.6% (97.1-98.0%), sensitivity 66.3% (61.2-71.1%), and specificity 98.4% (98.1-98.8%). Sensitivity increased with increasing age and cancer stage, from 24.2% (16.0-34.1%) in Stage I to 95.3% (88.5-98.7%) in Stage IV. Sensitivity 80.4% (66.1-90.6%) and NPV 99.1% (98.2-99.6%) were highest for patients referred with symptoms qualifying for the upper GI pathway. Conclusions: This is the first large-scale prospective evaluation of an MCED test in a symptomatic population. These data provide the basis for a prospective, interventional study in patients presenting to primary care with non-specific signs and symptoms with low, but higher than background, probability of being due to cancer. Study data are also being used to enhance the negative predictive value of the current MCED classifier for a symptomatic population. Clinical trial information: ISRCTN10226380 .