The Paper of the Month April28 Apr 2021
A new WSA feature highlighting original research in the field of neurology, selected by WSA Editor-in-Chief Gustavo Saposnik.
Development of imaging-based risk scores for prediction of intracranial haemorrhage and ischaemic stroke in patients taking antithrombotic therapy after ischaemic stroke or transient ischaemic attack: a pooled analysis of individual patient data from cohort studies’ https://doi.org/10.1016/S1474-4422(21)00024-7
Gustavo Saposnik, MD, PhD
This article is a commentary on the following
“The quality of your life is in direct proportion to the amount of uncertainty you can comfortably deal with”. Tony Robbins (American author, coach, speaker, and philanthropist)
How much tolerance to uncertainty do we have as clinicians? How about the uncertainty related to the therapeutic efficacy or side effects of an agent or intervention when applied the evidence to an individual? Most internists, geriatricians, neurologists, stroke neurologists and other colleagues face the challenge of initiating and continuing antithrombotics for stroke prevention among patients with high-risk of bleeding, specially those with microhemorrhages.
In the present study, a joint team from the Microbleeds International Collaborative Network (MICON) conducted a patient-level analysis of 15,784 individuals with ischaemic stroke or transient ischemic attack to create a risk score for developing an ischemic stroke and symptomatic intracranial hemorrhage (sICH). They enrolled patients at 38 hospitals in 18 countries. Participants had MRI scans at baseline that included the presence of microhemorrhage. The MICON-intracranial haemorrhage (MICON-ICH) risk score assigned the greatest weight to the number of cerebral microbleeds (up to 9 points), previous intracranial haemorrhage (5 points), and patient age (up to 4 points). A previous ischaemic stroke, and treatment with either an antiplatelet or vitamin K antagonist agent had lower points (1 point each). By incorporating MRI detection of microhemorrhages, the MICON-ICH risk score had better discrimination compared to other existing scales (e.g. HASBLED). The nearly exponential increased risk of sICH with the number of microhemorrhages is a reflection of the association and underlying pathophysiological mechanism. As concluded by the authors, external validation is needed. However, this study provides another tool for clinicians to better estimate (and perhaps ameliorate some uncertainties) the risk of sICH when the use of antithrombotic agents are warranted.
Jonathan Best, PhD, and David Werring, PhD, FRCP
WHAT DID YOU SET OUT TO STUDY?
We investigated whether using magnetic resonance imaging markers of cerebral small vessel disease could improve the prediction of symptomatic intracranial haemorrhage and ischaemic stroke in patients taking antithrombotic medication for secondary prevention of ischaemic stroke or transient ischaemic attack. We were particularly interested in cerebral microbleeds – small ovoid hypointense lesions assessable using blood-sensitive MRI sequences, thought to indicate previous extravasation of blood and therefore vascular fragility – as our previous work had found these to be associated with both ischaemic stroke and intracranial haemorrhage risk in this population. We hoped combining cerebral microbleed burden with simple clinical variables could produce a prediction tool suitable for use in clinical practice.
WHY THIS TOPIC?
Antithrombotics are the mainstay of secondary stroke prevention, prescribed to nearly all patients with previous ischaemic stroke or transient ischaemic attack. We know they increase the risk of intracranial haemorrhage – by around one-quarter for antiplatelets, one-half for direct oral anticoagulants, and two-fold for vitamin K antagonists – but our ability to predict this potentially-devastating complication is limited. Existing clinical risk scores – such as HASBLED, ATRIA and ORBIT – were developed to predict major bleeding in general, and show poor to moderate performance in predicting intracranial haemorrhage specifically. Clinicians are likely to weight the risk of intracranial haemorrhage much more heavily than that of extracranial bleeding in their decision-making, so an improved prediction tool for intracranial haemorrhage would be particularly valuable.
WHAT WERE THE KEY FINDINGS?
We developed a simple intracranial haemorrhage risk score – MICON-ICH – including age, patient population (East Asian or other), cerebral microbleed burden , MRI sequence type used for assessment, previous intracranial haemorrhage, previous ischaemic stroke, and antithrombotic treatment type. In optimism-adjusted internal validation, the MICON-ICH score appeared to show superior discrimination (c = 0.73) to existing clinical risk scores, and accurately stratified patients into lower, intermediate, and high risk groups. We also found that although including cerebral microbleed burden substantially improved intracranial haemorrhage prediction, it made a limited contribution over clinical variables alone to ischaemic stroke prediction – our companion MICON-IS score showed moderate discrimination, comparable to existing ischaemic stroke risk scores.
HOW MIGHT THESE RESULTS IMPACT CLINICAL PRACTICE?
We hope that the MICON-ICH risk score will help clinicians identify patients who are at the highest risk of future intracranial haemorrhage. These patients might benefit from closer follow-up, more intensive blood pressure monitoring and treatment, and lifestyle interventions such as minimising alcohol intake. Of course, ideally all patients taking antithrombotic medication should be offered this care, but in real-world clinical practice, targeting our efforts at those most likely to benefit is often necessary. Our vision is that the risk scores will be used routinely by clinicians on the stroke unit and in clinics to inform prognostic discussions with patients.
WHAT SURPRISED YOU MOST?
We found little evidence of an association between antithrombotic treatment type and intracranial haemorrhage risk in our study population. We interpret this as being partly due to treatment bias, reflecting the observational data used in our study. While we accounted for many potentially-confounding variables in our model development process, more complex attributes such as frailty – and perhaps even a clinician’s subjective assessment of ‘suitability’ for anticoagulation – are difficult to parameterise in a prediction model. Because of this limitation, we advise against using our scores to decide between antithrombotics. Instead, they should be used to assess prognosis once the intended antithrombotic has been selected.
WHAT’S NEXT FOR THIS RESEARCH?
As we used all available data in our model development process – in order to develop the best prediction model we could – externally validating the MICON-ICH score would be valuable, and we encourage other research groups with access to suitable data to test it. It would be interesting to see if the MICON scores have value in other populations, for example those without previous stroke, or survivors of intracerebral haemorrhage. We are also interested in the potential use of the MICON-ICH score in research. The incidence of intracranial haemorrhage is low compared to that of ischaemic stroke, making research into prevention strategies challenging. The ability to select and recruit participants at highest predicted ICH risk could reduce the sample sizes required, and, with the ever-increasing use of MRI in routine clinical practice, using our score do this should be feasible.
IS THERE ANYTHING YOU’D LIKE TO ADD?
The MICON-ICH and MICON-IS scores are available as a web calculator on the NeuroMind website (www.neuromind.cc/n/60451) and on the NeuroMind mobile App. We are grateful to Pieter Kubben for this. Our sincere thanks also go to our many co-authors and collaborators in this work, which was a truly international effort, to our funders, and to the participants in the 38 research studies included, whose willingness to take part in research made this work possible.
DO YOU HAVE ANY CONFLICTS/DISCLOSURES?
Jonathan Best has no conflicts of interest or disclosures. David Werring reports personal fees from Bayer, Alnylam, and Portola, outside of this work.