Cancer is one of the leading causes of death worldwide. Advances in therapies and surgery are not effective in significantly lowering cancer deaths, as prognosis is still extremely poor for tumours diagnosed in later stages. A simple, rapid, accurate and cost-effective technology enabling earlier detection would represent a turning point in the battle against cancer, by increasing survival rates and patients’ quality of life. Dxcover® Cancer Liquid Biopsy (Dxcover Ltd., UK) utilises attenuated total reflection - Fourier transform infrared (ATR-FTIR) spectroscopy combined with machine learning techniques to generate predictions of disease status from the analysis of human blood serum samples. This thesis focuses on the development and adaptability of this technology for the earlier detection of multiple cancers, evaluating diagnostic ability and pre-analytical requirements for clinical translation.
Firstly, investigations on blood collection, serum extraction and sample preparation were performed. Three different brands of blood collection vials were tested, showing no significant spectral variance. In addition, centrifugation procedures for serum extraction were examined and it was found that, upon successful separation of serum from the blood fraction, the spectral response is independent from time and speed of extraction. Deposition techniques were also investigated to understand the most suitable deposition method for the Dxcover® Sample Slide (Dxcover Ltd., UK), concluding that 3 ± 1 μL of serum spotted at the centre of the well with homogenous spreading across the corners was deemed the best practice.
In order to break the barriers to clinical translation, novel diagnostic techniques must meet requirements that go beyond the pre-analytical development. It is necessary that
vi
they also prove their economic value. A cost evaluation for pancreatic cancer was conducted, showing that the implementation of our test for the earlier detection of pancreatic cancer could save the National Health Services (NHS) over £8 million/year.
Pancreatic tumours are extremely difficult to detect in early stages due to absence of specific symptomatology and accurate diagnostic tests. A study with 235 patients was conducted to investigate the use of our technology for the earlier detection of pancreatic cancer, generating promising results. Partial least squares - discriminant analysis (PLS-DA) reported 92% sensitivity and 88% specificity with a receiver operating characteristic area under the curve (ROC-AUC) of 0.95 for cancer versus healthy controls. For cancer versus symptomatic controls, 70% sensitivity and 85% specificity with ROC-AUC of 0.83 were obtained, showing the potential great impact of our test in a real-life scenario.
To conclude, the technology was employed in a large-scale multi-cancer early detection (MCED) study with 2094 patients. An effective MCED test must be able to accurately identify cancer patients with and without specific symptoms. By differentiating cancer versus non-cancer (including symptomatic), the sensitivity-tuned model achieved 90% sensitivity and 61% specificity, with detection rates of 93% for stage I, 84% for stage II, 92% for stage III, and 95% for stage IV cancers. Cancer versus non-cancer asymptomatic classification resulted in 98% sensitivity and 58% specificity for the sensitivity-tuned model, with detection rates between 97 and 99% for all cancer stages. For organ specific cancer classifiers, ROC-AUC obtained were brain (0.90), breast (0.75), colorectal (0.91), kidney (0.91), lung (0.91), ovarian (0.86), pancreatic (0.85) and prostate (0.86), highlighting the performance of this MCED technology for earlier detection of cancer and its future impact in a clinical scenario.
Date of Award | 16 Sept 2022 |
---|
Original language | English |
---|
Awarding Institution | - University Of Strathclyde
|
---|
Sponsors | University of Strathclyde |
---|
Supervisor | David Palmer (Supervisor) & Matthew Baker (Supervisor) |
---|