Responses to provision of personalised cancer risk information: a qualitative interview study with members of the public

Towards the Interpretability of Machine Learning Predictions for Medical Applications Targeting Personalised Therapies: A Cancer Case Survey

Artificial Intelligence is providing astonishing results, with medicine being one of its favourite playgrounds. Machine Learning and, in particular, Deep Neural Networks are behind this revolution. Among the most challenging targets of interest in medicine are cancer diagnosis and therapies but, to start this revolution, software tools need to be adapted to cover the new requirements. In this sense, learning tools are becoming a commodity but, to be able to assist doctors on a daily basis, it is essential to fully understand how models can be interpreted. In this survey, we analyse current machine learning models and other in-silico tools as applied to medicine-specifically, to cancer research-and we discuss their interpretability, performance and the input data they are fed with. Artificial neural networks (ANN), logistic regression (LR) and support vector machines (SVM) have been observed to be the preferred models. In addition, convolutional neural networks (CNNs), supported by the rapid development of graphic processing units (GPUs) and high-performance computing (HPC) infrastructures, are gaining importance when image processing is feasible. However, the interpretability of machine learning predictions so that doctors can understand them, trust them and gain useful insights for the clinical practice is still rarely considered, which is a factor that needs to be improved to enhance doctors' predictive capacity and achieve individualised therapies in the near future.

Development and usability testing of a very brief intervention for personalised cancer risk assessment to promote behaviour change in primary care using normalisation process theory

BACKGROUND

Cancer is the second leading cause of death worldwide. Lifestyle choices play an important role in the aetiology of cancer with up to 4 in 10 cases potentially preventable. Interventions delivered by healthcare professionals (HCPs) that incorporate risk information have the potential to promote behaviour change. Our aim was to develop a very brief intervention incorporating cancer risk, which could be implemented within primary care.

METHODS

Guided by normalisation process theory (NPT), we developed a prototype intervention using literature reviews, consultation with patient and public representatives and pilot work with patients and HCPs. We conducted focus groups and interviews with 65 HCPs involved in delivering prevention activities. Findings were used to refine the intervention before 22 HCPs completed an online usability test and provided further feedback via a questionnaire incorporating a modified version of the NoMAD checklist.

RESULTS

The intervention included a website where individuals could provide information on lifestyle risk factors view their estimated 10-year risk of developing one or more of the five most common preventable cancers and access lifestyle advice incorporating behaviour change techniques. Changes incorporated from feedback from the focus groups and interviews included signposting to local services and websites, simplified wording and labelling of risk information. In the usability testing, all participants felt it would be easy to collect the risk information. Ninety-one percent felt the intervention would enable discussion about cancer risk and believed it had potential to be easily integrated into National Health Service (NHS) Health Checks. However, only 36% agreed it could be delivered within 5 min.

CONCLUSIONS

With the use of NPT, we developed a very brief intervention that is acceptable to HCPs in primary care and could be potentially integrated into NHS Health Checks. However, further work is needed to assess its feasibility and potential effectiveness.

Development and Validation of Lifestyle-Based Models to Predict Incidence of the Most Common Potentially Preventable Cancers

BACKGROUND

Most risk models for cancer are either specific to individual cancers or include complex or predominantly non-modifiable risk factors.

METHODS

We developed lifestyle-based models for the five cancers for which the most cases are potentially preventable through lifestyle change in the UK (lung, colorectal, bladder, kidney, and esophageal for men and breast, lung, colorectal, endometrial, and kidney for women). We selected lifestyle risk factors from the European Code against Cancer and obtained estimates of relative risks from meta-analyses of observational studies. We used mean values for risk factors from nationally representative samples and mean 10-year estimated absolute risks from routinely available sources. We then assessed the performance of the models in 23,768 participants in the EPIC-Norfolk cohort who had no history of the five selected cancers at baseline.

RESULTS

In men, the combined risk model showed good discrimination [AUC, 0.71; 95% confidence interval (CI), 0.69-0.73] and calibration. Discrimination was lower in women (AUC, 0.59; 95% CI, 0.57-0.61), but calibration was good. In both sexes, the individual models for lung cancer had the highest AUCs (0.83; 95% CI, 0.80-0.85 for men and 0.82; 95% CI, 0.76-0.87 for women). The lowest AUCs were for breast cancer in women and kidney cancer in men.

CONCLUSIONS

The discrimination and calibration of the models are both reasonable, with the discrimination for individual cancers comparable or better than many other published risk models.

IMPACT

These models could be used to demonstrate the potential impact of lifestyle change on risk of cancer to promote behavior change.

A randomised controlled trial of the effect of providing online risk information and lifestyle advice for the most common preventable cancers: study protocol

BACKGROUND

Cancer is a leading cause of mortality and morbidity worldwide. Prevention is recognised by many, including the World Health Organization, to offer the most cost-effective long-term strategy for the control of cancer. One approach that focuses on individuals is the provision of personalised risk information. However, whether such information motivates behaviour change and whether the effect is different with varying formats of risk presentation is unclear. We aim to assess the short-term effect of providing information about personalised risk of cancer in three different formats alongside lifestyle advice on health-related behaviours, risk perception and risk conviction.

METHODS

In a parallel group, randomised controlled trial 1000 participants will be recruited through the online platform Prolific. Participants will be allocated to either a control group receiving cancer-specific lifestyle advice alone or one of three intervention groups receiving the same lifestyle advice alongside their estimated 10-year risk of developing one of the five most common preventable cancers, calculated from self-reported modifiable behavioural risk factors, in one of three different formats (bar chart, pictograph or qualitative scale). The primary outcome is change from baseline in computed risk relative to an individual with a recommended lifestyle at three months. Secondary outcomes include: perceived risk of cancer; anxiety; cancer-related worry; intention to change behaviour; and awareness of cancer risk factors.

DISCUSSION

This study will provide evidence on the short-term effect of providing online information about personalised risk of cancer alongside lifestyle advice on risk perception and health-related behaviours and inform the development of interventions.

TRIAL REGISTRATION

ISRCTN17450583. Registered 30 January 2018.