Big Data in Cancer Treatment

Randomised controlled data (RCTs) are considered the gold standard in cancer medicine when performing comparative research between different interventions to assess their effectiveness. RCTs are thought to be grossly free from bias and confounding; they account for unmeasured variables as they tend to have a robust design and provide mostly accurate patient-reported outcomes. However, poorly designed RCTs can have the opposite effect and mis-inform the medical community of the real benefit of an intervention. RCTs of poor quality can be very costly, lead to confusion in the oncology community, demotivate researchers, and can result in loss of trust to the intervention by clinicians.

Real world data (RWD) and clinical registries (CRs) are increasing in popularity due to their ease of collection, inexpensive collection methods, and due to being readily available with every patient care record. If they are systematically collected across a range of different providers or a network of centres belonging to the same provider, RWD and CRs can provide a valuable repository of clinical information, including minimum cancer datasets, clinical characteristics, imaging and biochemical data, genomic data, which can be linked to cancer treatments provided, patient-reported outcomes and long-term clinical outcomes. Although population data cannot give comparative effectiveness between treatments, they can be used to compare patient results between different patient datasets, patients who may differ in specific characteristics (e.g., lifestyle behaviours, genomic markers). As such, they can lead to the development of a correlation between particular patient characteristic(s)and patient clinical outcome(s). In the case of imaging data, the use of artificial intelligence can help the prediction of patient outcomes in the presence of specific imaging characteristics. If systematically collected and analysed, RWD can support the development of predictive algorithms of treatment response, such as in the case of radiotherapy treatments.

Systematically collected patient registries combined with advanced analytical strategies can be of value to researchers who are keen to validate data links in more significant studies and even RCTs. They can also be of value to private medical insurers in guiding reimbursement of new radiotherapy technologies and techniques when RCTs are not possible. In the case of a new radiotherapy technology (e.g., MRI guided radiotherapy) or radiotherapy techniques (e.g., VMAT IMRT), where preliminary research has suggested benefit or non-inferiority, a time-consuming and expensive RCT may not be the right thing to do. Good quality data is a prerequisite to a robust and meaningful analysis and is more critical than data quantity.

Some argue that if a large amount of population data in the form of registries are collected systematically, they can be as informative as randomised controlled trials (Tsai, 2019 and DeWees, 2019). Real world data can often supplement RCTs by providing information around treatment toxicities and post-market utilisation of cancer treatments, which may not have been collected during RCTs. They may also lead to the initiation of Quality Improvement programmes such as in the case of patient pathway improvements, as well as optimise health promotion if certain patient lifestyle behaviours lead to worse cancer outcomes.

Big Data in Radiotherapy

At GenesisCare UK, we are investing heavily in world-class radiotherapy technologies (e.g., MR guided Linac, Gamma Knife radiosurgery, Stereotactic ablative radiotherapy), we are continually improving our treatment techniques (e.g., prostate hypofractionation, breast simultaneous integrated boost, accelerated partial breast radiotherapy) and we are introducing automated tools to improve time efficiencies for clinicians, planners and patients (e.g., volume auto-contouring, physics QA). The role of real-world observational data and registries is of great value for analysing trends of patient-reported outcomes and treatment outcomes as well as predicting the quality of radiotherapy plans and QA outcomes. We use innovative collection tools, well-trained data management teams, and experienced analysts to ensure data robustness. Aligning with the global business in terms of data content and collection methodologies means that we can quickly build a shared database of real world data to use for service and quality improvements as well as for further research. There are complex processes in radiotherapy and if data are systematically collected during each process over time, there is an excellent opportunity for process refinement, process standardisation resulting in risk minimisation and risk prevention. One example is the atlas-based automation of tumour target volume delineation by oncologists. There is often considerable variation in target volume contouring between clinicians and real-world data can help identify and correct outlier practices hence improving radiation treatment outcomes.