Introduction
Non-traumatic acute convexity subarachnoid haemorrhage (cSAH)—also known as acute cortical subarachnoid haemorrahge—is confined to the subarachnoid space over the cortical hemispheric convexities of the brain, and does not extend into the parenchyma, sylvian fissures, ventricles, or basal cisterns [
1]. The aetiology and symptoms of cSAH are diverse, but have been classified by age of presentation [
1‐
7]; in younger patients (usually <60 years), cSAH often occurs in association with reversible cerebral vasoconstriction syndrome (RCVS), with recurrent thunderclap headache as the predominant presenting symptom [
8]. RCVS generally has a favourable outcome with very low recurrence risk, although in the acute phase both ischaemic stroke and intracerebral haemorrhage (ICH) can occur in a minority of patients [
9]. By contrast, in older patients, cSAH is associated with cerebral amyloid angiopathy (CAA) [
10], often presenting with transient focal neurological episodes (TFNE), often consisting of unilateral spreading sensory or motor symptoms. In CAA there is a substantial risk of recurrent symptomatic ICH (sICH) after presentation with ICH [
11], but knowledge about ICH risk following cSAH in patients with suspected CAA is limited, comprising only small cohort studies which individually contain very few clinical outcome events. A pooled meta-analysis of event rates from multiple prospective cohort studies allows a more precise estimate of sICH risk than is available from any individual study. This information is important in guiding clinical decision making with regard to prognosis and antithrombotic treatment, a common management dilemma in older patients at risk of vaso-occlusive events, for example those with atrial fibrillation or ischaemic heart disease.
We therefore investigated the risk of sICH following cSAH in: (1) a cohort of patients from our comprehensive stroke service presenting with acute symptomatic cSAH due to suspected CAA and (2) a pooled analysis which included our cohort and all other available published data on cSAH in suspected CAA cohorts.
Discussion
In our hospital cohort study the risk of symptomatic ICH following cSAH in patients with probable CAA was 4% per year; this risk was unrelated to prior ICH, as only 2 of 20 patients (5%) had a history of prior ICH (and they did not have a sICH during follow up). Most previous data suggesting a high future sICH risk in patients with suspected CAA (of around 9–16% per year [
11,
30]) are from populations of patients who presented with ICH. In our pooled analysis the annual risk of symptomatic ICH after cSAH due to suspected CAA was similarly high, at 19% per year (95% CI 13–27%) for patients with probable CAA, although lower, at 7% (95% CI 3–15%), for patients who did not meet the modified Boston criteria for probable CAA.
We also found that those with cSAH nearly always had cSS, with a high prevalence of disseminated cSS and severe centrum semiovale MRI-visible perivascular spaces, lobar CMBs and severe WMH (leukoaraiosis). The almost universal presence of cSS (often disseminated) and high risk of ICH following cSAH in our cohort is consistent with evidence that siderosis-predominant CAA is a distinct subtype, with a high risk of intracranial (and convexity subarachnoid) haemorrhage compared to CAA with predominant CMBs or leukoaraiosis [
31]. Notably, in our cohort, the prevalence of cSS extending beyond that secondary to the index cSAH was higher than in typical CAA-related ICH populations [
32]. A recent study suggested that active leptomeningeal CAA (with contrast enhancement) is the likely cause of CAA-related cSS [
27]. Thus, the high prevalence of disseminated cSS in our cohort suggests active and widespread leptomeningeal CAA, causing repeated previous cSAH leading to cSS.
Although our study suggests that the risk of future sICH is highest in patients who fulfil the Boston criteria for probable CAA, it should be noted that eligibility for these criteria can change over time; indeed, four of nine patients with possible CAA or non-CAA who had subsequent ICH at follow-up then “converted” to probable CAA, but were still considered to be “non-probable CAA” for the purpose of the pooled analysis.
The pooled event rate of 19% per year for future ICH in patients with probable CAA is higher than the future risk of ischaemic stroke reported after ICH, which has been reported to be between 1.3 and 2.9% per year following deep ICH and 2.5–14.3% per year following lobar ICH [
33]. Data on the risk of future ischaemic stroke following cSAH is scarce [
21,
28], but our findings suggest that the risk of ICH is very likely to be greater than the risk of cerebral ischaemia. This finding is potentially clinically relevant when making decisions about antithrombotic drug use. Although data are not available on whether or how antithrombotic drug use might affect the future risk of sICH or ischaemia in patients with cSAH at risk of future vaso-occlusive events (e.g. those with atrial fibrillation or other vascular risk factors), our findings suggest that antithrombotic drugs might be best avoided after cSAH unless there is a very compelling clinical indication. Since the TFNE commonly associated with cSAH can closely mimic the typical symptoms of TIA (although they are more likely to be positive and of spreading onset) the diagnosis of cSAH is crucial in patients with suspected TIA. Although CT may detect cSAH, MRI might be preferable because it is sensitive to small areas of sulcal haemorrhage and can identify cortical superficial siderosis and CMBs, allowing a more confident diagnosis of cSAH related to CAA.
Our study has strengths. We used multiple overlapping methods of case ascertainment to identify all patients with cSAH presenting to our comprehensive stroke service. All visual rating of neuroimaging was undertaken by a trained observer blinded to outcome. Our systematic search followed established PRISMA guidelines and pooled all available studies using mixed effects Poisson regression to account for studies with no ICH events.
Our study also has limitations. Despite conducting a systematic search and pooling event rates in a meta-analysis, there were still few sICH events. Furthermore, as all studies are hospital-based there is potential for significant selection bias and detection bias (due to potentially selective follow up). We were unable to adjust for confounders such as hypertension, age, antithrombotic treatment, previous ICH, which can all contribute to sICH risk. Follow-up times were variable from study to study, but we did not have access to the data needed for a time-to-event analysis. We did not obtain individual patient data, and so were unable to explore whether imaging findings such as cSS, CMBs and leukoaraiosis modify the risk of ICH.
In summary, patients who present with an acute cSAH and suspected CAA have a substantial risk of symptomatic ICH (16% per patient-year overall). The future sICH risk might be highest in those with neuroimaging findings consistent with probable CAA. These estimates of sICH event rates should be of value to clinicans when counselling patients regarding their future risk, and emphasise the importance of offering treatments aiming to reduce this risk. We suggest that antithrombotic drugs (antiplatelet and anticoagulant agents) are probably best avoided wherever possible after proven cSAH, unless there is a compelling indication, since the future sICH risk is high, and likely to be increased by exposure to such drugs. In patients with cSAH and a strong indication for antithrombotic drugs (e.g. recent symptomatic ischaemic heart disease or peripheral vascular disease), the likely benefits on reducing vaso-occlusive events must be carefully weighed against the potential for increasing the sICH risk. In patients with cSAH it seems reasonable to reduce blood pressure, which is effective for the long-term secondary prevention of ICH, including that attributed to probable CAA [
34]. Although this data currently presents the best available evidence the small sample sizes and limited information on other risk factors remains a limitation; future prospective studies with individual patient data might help consolidate our observations.