Developing a risk score for early detection of bowel cancer using routine blood test data
<p><strong>Introduction</strong></p> <p>Bowel cancer is the fourth most common type of cancer in the UK. Approximately 55% of patients are diagnosed with a late-stage tumour (Stage 3 and 4), where likelihood of survival is reduced: five-year survival is 93% at Stage 1 v...
| Main Author: | |
|---|---|
| Other Authors: | |
| Format: | Thesis |
| Language: | English |
| Published: |
2021
|
| Subjects: |
| _version_ | 1826307699776684032 |
|---|---|
| author | Virdee, PS |
| author2 | Birks, J |
| author_facet | Birks, J Virdee, PS |
| author_sort | Virdee, PS |
| collection | OXFORD |
| description | <p><strong>Introduction</strong></p>
<p>Bowel cancer is the fourth most common type of cancer in the UK. Approximately 55% of patients are diagnosed with a late-stage tumour (Stage 3 and 4), where likelihood of survival is reduced: five-year survival is 93% at Stage 1 vs. 10% at stage 4. Early detection is crucial. The full blood count (FBC) blood test is commonly performed in primary care. It may help but currently there is limited evidence on how to utilise this test for early detection.</p>
<p><strong>Methods</strong></p>
<p>I performed four research objectives. I reviewed relevant research studies in the literature to learn about the relevance of the FBC and eliminate some key uncertainties in existing approaches for early detection. I performed a data quality and validation assessment of electronic health record (EHR) patient data. I then performed longitudinal analyses to identify patterns in the FBC over time, which could help identify presence of bowel cancer. I subsequently incorporated my findings into a dynamic prediction model that could facilitate earlier detection and improve survival rates.</p>
<p><strong>Results</strong></p>
<p>My systematic review identified many relevant FBC parameters for bowel cancer detection, such as haemoglobin, mean corpuscular volume, and platelets, and that all existing FBC-based prediction models are flawed. Using primary care data from the Clinical Practice Research Datalink, I identified relevant trajectories in many FBC parameters in patients with bowel cancer, not observed in patients without. Relevant differences in FBC levels occurred for years before diagnosis, suggesting the FBC could facilitate earlier detection. I incorporated the trajectories in haemoglobin, mean corpuscular volume, and platelets into multivariate joint models of longitudinal and time-to-event data to identify two-year risk of diagnosis. My joint models perform well based on my initial testing.</p>
<p><strong>Discussion</strong></p>
<p>My joint models use routinely available data and would not require additional work for healthcare workers or patients. These models perform well based on my testing, but further testing is required before use in general practice. My models could facilitate earlier detection and improve survival rates.</p> |
| first_indexed | 2024-03-07T07:07:01Z |
| format | Thesis |
| id | oxford-uuid:f65c7f99-7abf-4f67-9b1b-527ffac217ed |
| institution | University of Oxford |
| language | English |
| last_indexed | 2024-03-07T07:07:01Z |
| publishDate | 2021 |
| record_format | dspace |
| spelling | oxford-uuid:f65c7f99-7abf-4f67-9b1b-527ffac217ed2022-05-10T08:23:12ZDeveloping a risk score for early detection of bowel cancer using routine blood test dataThesishttp://purl.org/coar/resource_type/c_db06uuid:f65c7f99-7abf-4f67-9b1b-527ffac217edPrimary care (Medicine)Cancer researchMedical statisticsEpidemiologyEnglishHyrax Deposit2021Virdee, PSBirks, JHolt, TPatnick, JWatkinson, P<p><strong>Introduction</strong></p> <p>Bowel cancer is the fourth most common type of cancer in the UK. Approximately 55% of patients are diagnosed with a late-stage tumour (Stage 3 and 4), where likelihood of survival is reduced: five-year survival is 93% at Stage 1 vs. 10% at stage 4. Early detection is crucial. The full blood count (FBC) blood test is commonly performed in primary care. It may help but currently there is limited evidence on how to utilise this test for early detection.</p> <p><strong>Methods</strong></p> <p>I performed four research objectives. I reviewed relevant research studies in the literature to learn about the relevance of the FBC and eliminate some key uncertainties in existing approaches for early detection. I performed a data quality and validation assessment of electronic health record (EHR) patient data. I then performed longitudinal analyses to identify patterns in the FBC over time, which could help identify presence of bowel cancer. I subsequently incorporated my findings into a dynamic prediction model that could facilitate earlier detection and improve survival rates.</p> <p><strong>Results</strong></p> <p>My systematic review identified many relevant FBC parameters for bowel cancer detection, such as haemoglobin, mean corpuscular volume, and platelets, and that all existing FBC-based prediction models are flawed. Using primary care data from the Clinical Practice Research Datalink, I identified relevant trajectories in many FBC parameters in patients with bowel cancer, not observed in patients without. Relevant differences in FBC levels occurred for years before diagnosis, suggesting the FBC could facilitate earlier detection. I incorporated the trajectories in haemoglobin, mean corpuscular volume, and platelets into multivariate joint models of longitudinal and time-to-event data to identify two-year risk of diagnosis. My joint models perform well based on my initial testing.</p> <p><strong>Discussion</strong></p> <p>My joint models use routinely available data and would not require additional work for healthcare workers or patients. These models perform well based on my testing, but further testing is required before use in general practice. My models could facilitate earlier detection and improve survival rates.</p> |
| spellingShingle | Primary care (Medicine) Cancer research Medical statistics Epidemiology Virdee, PS Developing a risk score for early detection of bowel cancer using routine blood test data |
| title | Developing a risk score for early detection of bowel cancer using routine blood test data |
| title_full | Developing a risk score for early detection of bowel cancer using routine blood test data |
| title_fullStr | Developing a risk score for early detection of bowel cancer using routine blood test data |
| title_full_unstemmed | Developing a risk score for early detection of bowel cancer using routine blood test data |
| title_short | Developing a risk score for early detection of bowel cancer using routine blood test data |
| title_sort | developing a risk score for early detection of bowel cancer using routine blood test data |
| topic | Primary care (Medicine) Cancer research Medical statistics Epidemiology |
| work_keys_str_mv | AT virdeeps developingariskscoreforearlydetectionofbowelcancerusingroutinebloodtestdata |