Cancer survival in the UK is poorer than in other European countries and differs widely between regions and social classes. Despite continuing efforts these problems persist and it is crucial to better understand how they arise to develop more effective strategies for an earlier cancer diagnosis and timely treatment.
Several factors contribute to poor cancer outcomes and survival inequalities, including low symptom awareness, barriers to presentation, referral and diagnostic delays, comorbidities and suboptimal treatments.
To strengthen the current evidence, we propose to use an up-to-date version of linked datasets to study the entire clinical care pathway. Our research is divided into two phases, the pre-diagnostic (evaluating symptomatic presentations and diagnostic investigations before cancer diagnosis) and post-diagnostic phases (evaluating treatments and care after diagnosis). This application focuses on the pre-diagnostic parts. We will submit a second application for the post-diagnostic parts.
We aims here at investigating the pattern of primary care presentations and diagnostic investigations prior to the cancer diagnosis, to identify patient and healthcare factors associated with an increased risk of a diagnosis during an emergency presentation or at an advanced tumour stage, thus poorer survival. We will examine the presence of missed opportunities for earlier diagnosis and access inequalities.
The main research objective is to produce population-based evidence for improving diagnosis and survival for cancer patients. Focusing on the pre-diagnostic phase and, on adult bowel cancer and brain tumours in children and young adults, we aim to evaluate the roles played by primary-care symptomatic presentations on (i) first, the risk of being diagnosed with the tumour during emergency admission, then, on (ii) the extend of the tumour and on survival from the tumour.
We will use an up-to-date version of linked datasets combining the national cancer data repository and CPRD. We will extract from that the relevant cancer patients with at least one year of CPRD records prior to cancer diagnosis.
We will primarily model the association between emergency diagnosis and symptoms/clusters of symptoms using logistic regression analysis, and the association between emergency diagnosis and consultation rates for relevant symptoms, using Poisson regression. We will account for patients’ socio-demographic and clinical characteristics. Random effect will be considered to account for patient-level clustering of repeated symptoms. We will then estimate the associations of the symptoms patterns identified at this step with the extent of the disease and the survival from the disease, using multinomial/logistic regression and excess hazard model, respectively.
Bernard Rachet - Chief Investigator - London School of Hygiene & Tropical Medicine ( LSHTM )
Cristina Renzi - Collaborator - University College London ( UCL )
Krishnan Bhaskaran - Collaborator - London School of Hygiene & Tropical Medicine ( LSHTM )
Michel Coleman - Collaborator - London School of Hygiene & Tropical Medicine ( LSHTM )
Rachael Williams - Collaborator - CPRD
Thomas Chu - Collaborator - London School of Hygiene & Tropical Medicine ( LSHTM )
Timothy Card - Collaborator - University of Nottingham
NCRAS Cancer Registration Data;Other