The project is suitable for candidates with strong statistical background, ideally with MSc in Medical Statistics, Biostatistics or Statistics.
Evidence synthesis of time-to-event outcomes, for example using network meta-analysis, is a key component in health technology assessment decision-making by agencies such as the National Institute for Health and Clinical Excellence, particularly for new cancer therapies. Traditionally, synthesis of time-to-event outcomes has been conducted using aggregate data from clinical trials. However, clinical trial data for novel cancer therapies is limited and increasingly use of real world evidence, such as from electronic health records (EHRs), is explored to improve the evidence base. Use of real world evidence also allows for a more realistic assessment of the effectiveness of therapies in the general population. In addition, aggregate data from clinical trials typically results from the time-to-event analysis assuming that the treatment effect is constant over time (known as ‘proportional hazards (PH)’). However, in fields such as oncology, there is increasing evidence that this is not always the case especially in real-world data. Individual participant data (IPD) allows consideration of treatment-covariate interactions that can help ameliorate this issue. However, IPD from clinical trials are rarely available and EHRs can provide a suitable source of data to allow for development and application of more appropriate methods for analysis of such complex data on new cancer therapies. Furthermore, joint synthesis of multiple outcomes can improve inferences and better capture uncertainty. In this project, data from EHRs at IPD level will facilitate flexible modelling of non-PH data allowing for more appropriate and realistic assessment of new cancer therapies.
Development of methods for simultaneously synthesising IPD for time-to-event outcomes in the presence of non-PH combining EHR and clinical trial data will:
- Improve the clinical evidence base that contributes to the health technology assessment decision-making process, and
- Identify subgroups of patients who are most likely to respond to treatment, which could change clinical practice resulting in better quality treatment for patients.
These methodological developments will demonstrate optimal use of EHRs in healthcare policy decisions and will be generalizable to a broad range of disease areas.
Through development of a framework for conducting IPD multivariate evidence synthesis, this project will:
- Use EHRs to emulate cancer trials with non-PH and assess the performance of different modelling approaches
- Combine EHRs and clinical trials to form connected networks and assess the performance of different modelling approaches
- Assess the impact of combining EHRs and clinical trials on the health technology assessment process.
For initial methodology development a simulated Systemic Anti-Cancer Therapy (SACT) dataset available freely from Simulacrum will be used with breast cancer as one of the case studies (including over 33,000 patients). The developed methodology will then be applied to actual SACT data for breast cancer patients.
- Bujkiewicz S, et al. NICE DSU Technical Support Document 20: Multivariate meta-analysis of summary data for combining treatment effects on correlated outcomes and evaluating surrogate endpoints. 2019; available from http://www.nicedsu.org.uk
- Freeman SC, Cooper NJ, Sutton AJ, Crowther MJ, Carpenter JR, Hawkins N. Challenges of Bayesian modelling approaches for network meta-analysis of time-to-event outcomes to aid decision making: application to a melanoma network. Statistical Methods in Medical Research. Accepted
- Freeman SC, Sutton AJ, Cooper NJ. Uptake of methodological advances for synthesis of continuous and time-to-event outcomes would maximize use of the evidence base. Journal of Clinical Epidemiology. 2020; 124: 94-105, DOI: https://doi.org/10.1016/j.jclinepi.2020.05.010
- Achana FA, Cooper NJ, Bujkiewicz S, Hubbard SJ, Kendrick D, Jones DR and Sutton AJ. Network meta-analysis of multiple outcome measures accounting for borrowing of information across outcomes. BMC Medical Research. Methodology 2014, 14: 92, DOI: https://doi.org/10.1186/1471-2288-14-92
- Freeman SC, Carpenter JR. Bayesian one-step IPD network meta-analysis of time-to-event data using Royston-Parmar models. Research Synthesis Methods 2017; 8: 451, DOI: https://doi.org/10.1002/jrsm.1253
- Freeman SC, Fisher D, Tierney JF, Carpenter JR. A framework for identifying treatment-covariate interactions in individual participant data network meta-analysis. Research Synthesis Methods 2018; 9: 393-407, DOI: https://doi.org/10.1002/jrsm.1300