Targeted treatment for acute stroke: development of prognostic models and decision support tools

Stroke, the world‘s second biggest killer, is most often due to a blockage of a brain blood vessel by blood clot. Drugs that dissolve clots or prevent them from forming improve the chance of recovery after stroke, but they also increase the chance of bleeding; when bleeds occur into the brain they are particularly severe. Most stroke patients need to take drugs that affect blood clotting at some point in their illness, ranging from potent ‘clot busting‘ drugs to less risky drugs like aspirin.

If a doctor could reliably select which patients were helped rather than harmed by these drugs, this would be a big step forward. I hope to develop a method to predict which stroke patients are more likely to be harmed (by bleeding) than benefit (by avoiding forming blood clots) from medicines that affect clotting. I will use statistical methods to make predictions, and then explore the best way for doctors and patients to use them. This will lead to better decisions for each individual patient, and better treatment policies. This study will use the best information from existing methods as well as from newer technologies (e.g. blood tests and brain imaging) to ensure effective and personalised decisions for each stroke patient

Background
Although thrombolytic, antithrombotic and antiplatelet drugs are used for the treatment of acute stroke, they can have severe side effects. Better targeting of these therapies could increase their net population health gain by reducing avoidable haemorrhagic and thrombotic events.

Aim
To develop decision support tools that provide clinically useful guidance and improve patient outcome by better application of existing treatments for stroke. These could be incorporated into the NHS electronic record should they show significant health benefits.

Objectives
1. Develop and validate statistical models to predict haemorrhagic and thrombotic arterial and venous events after stroke with data from a number of large cohorts.
2. Determine whether prediction of haemorrhagic events, based on existing clinical and imaging data, has a significant interaction with the treatment benefits of intravenous thrombolysis, heparin and aspirin in large randomised controlled trials.
3. Create a clinical decision tool, based on the prediction models for haemorrhage and arterial and venous thrombosis developed in parts 1-3
4. Pilot the application of the decision tool in stroke practice in NHS Lothian with a view to establishing a larger scale randomised multi-centre evaluation study.

Design/Methodology
I will develop and validate models for the prediction of haemorrhagic and thrombotic events after stroke in prospective hospital- and community-based cohort studies and randomised controlled trials, from (i) the Stroke Complications and Outcomes Prediction Engine collaboration based in Edinburgh (ii) the Kadoorie Study of Chronic Disease in China and (iii) the Virtual Internet Stroke Archive based in Glasgow
To examine the interaction between the results of models predicting haemorrhagic and thrombotic events after stroke and the effect of treatment I will use data from the MRC International Stroke Trial of aspirin and heparin, the Third International Stroke Trial of thrombolysis, and the Chinese Acute Stroke Trial of aspirin.
Using tested technology, in collaboration with a group experienced in health informatics, I will develop a tool to present the results of prognostic models as a clinical decision tool. I will evaluate its feasibility in practice, with a view to a larger study.

Medical and Scientific Opportunities
After stroke, the benefits of antithrombotic and thrombolytic drugs are offset to some extent by an increase in the risk of haemorrhage from the treatment. Better selection of patients for these commonly used treatments could - by focussing treatments on patients most likely to derive net benefit - improve outcome in individual stroke patients, with potentially great population health gains.