Optimal screening and surveillance regimes for early diagnosis of cancer and precision medicine using mathematical modelling

The main rationale for cancer screening is that detecting disease early offers the opportunity to change its prognosis. Compared with symptomatic cancers, the lifetime prognosis is often greatly improved for patients found to have precancerous lesions or small cancers that are detected at an early stage. Therefore, clinicians focus on identifying patients with precancerous change on initial tests (screens), and then after this initial screening may advise these patients to undergo long-term, periodic screening by returning to the clinic at certain intervals for regular examinations (surveillance screens) throughout the course of their lives. However, many precancerous changes will never progress to cancer in the lifetime of the patient. Thus, many patients who undergo regular surveillance will never be diagnosed with cancer in their lifetimes. Overall, many current approaches for prevention by screening and surveillance programs have achieved minimal success in reducing cancer deaths at a high cost to healthcare services, and thus paradoxically yield both under-diagnosis due to inadequate screening and over-diagnosis due to ineffective patient stratification in surveillance protocols.
Herein lies the balancing act performed during risk stratification - identify who is most `at risk' of progressing to cancer and suggest effective surveillance and intervention strategies for the high risk groups. With this motivation, the overall public health goal of the mathematical modelling presented in this research plan is to help improve the efficacy of screening and surveillance for precancerous and cancer lesions. Although clinical endpoints like cancer incidence and prevalence of pre-cancerous change are reported on a population level, many important biological processes on smaller physical and temporal scales occur with significant differences between patients during disease progression from normal tissue to incident cancer. Due to the biological and clinical nature of screening patients at various times during their lives, such details at tissue and cell levels provide vital information to determine screening outcomes obtained by different modalities and protocols. This research plan will use many such levels of data in inventive and rigorous ways to improve personalized healthcare.
Within the MRC priority area Precision Medicine and Diagnostics, mathematical modelling of cancer formation can be used to derive and to optimise the timing of clinical screens so that an individual is screened within a certain "window of opportunity" for intervention when early cancer development may be observed. By using data from epidemiological studies with long-term patient follow-up, current empirical approaches can aid in screening design and may inform cost-effectiveness analyses to compare proposed screening and intervention strategies. However, mechanistic modelling that incorporates a greater level of biological understanding and detail for how and when normal tissues progress to cancer can be used for a more refined screening design than typically implemented in population screening studies.
The aims of this Health Data Research UK research plan are
1) To use mathematical modelling to inform optimal cancer screening recommendations for a population,
2) To perform patient risk stratification by identifying who is low risk versus who is high risk in order to make personalised surveillance regimes
that are more effective than "one-size-fits-all" approaches,
3) To build tools that will assist in precision medicine such as clearly portraying future cancer risk to a patient in visuals aids.

Current screening strategies for early cancer detection aim to identify individuals with premalignant tissue changes that signify an important first step in the progression to cancer, a manifestation of field cancerisation that occurs in many tissues, and when clinical intervention may be possible. However, many current approaches for prevention by screening and surveillance programs have achieved minimal success in reducing mortality at a high cost to healthcare services.
Within the MRC priority area Precision Medicine and Diagnostics, my research project will develop and apply mechanistic mathematical models that describe how normal tissues develop into cancer to predict how surveillance (and thus early detection) may be optimised. Through simulation and mathematical analysis of the pattern and pace of cancer development, I will calculate the likely position of any individual on a developmental pathway towards cancer. By then creating a virtual population of patients, I will be able to judge the effectiveness of a screening/surveillance regime across a population. I will parameterise the biological details of these models with data spanning molecular to population scales from Health Data Research UK and a growing network of expert collaborators. I will then use this framework to predict the optimal times to screen at-risk individuals and to tailor a surveillance regime to individual patients, and validate these predictions with epidemiological data. The ultimate aims are to:
1) Rigorously assess and inform the best screening strategies to benefit public health recommendations
2) Translate this work into risk assessment, biomarker discovery, and patient stratification in the healthcare setting.
My novel methodologies integrate bioinformatics using Big Data, survival analysis using public health record data, stochastic modelling, and classical biostatistics to address cancer prevention problems.