The Subarachnoid Haemorrhage in the Emergency Department (SHED) Study

The diagnosis of subarachnoid haemorrhage in the emergency department is a constant source of anxiety. How do we make sure we pick up everyone with a bleed, but not overinvestigate others?…

A momentous day occurred in early September 2024. The initial concept project for the UK Trainee Emergency Research Network (TERN) was published as an online first in the EMJ. This concludes a process involving creation of the network in 2016, democratic selection of a project, successful funding application, regulatory approval, delivery of the study across a community of 88 Emergency Departments and finally data cleaning, independent statistical analysis, write up, peer review and publication. And with several waves of COVID in the middle. No wonder it took us around 8 years….

But 8 years later TERN has A LOT to be proud of, including this manuscript. The concept of a trainee network has grown from a fledgling idea within the RCEM research committee to a now fully funded entity delivering innovative acute care research, in its 4th iteration of ‘TERN Fellow’ and about to undertake its 6th project supported by experienced academics. TERN has pioneered a novel opt-out consent methodology for delivery of this and other projects in acute care; several fellows have gone on to apply for PhD funding or secure future academic career placements; over 100 EM trainees have engaged and trained to the level of NIHR Principle Investigator; and if you are into numbers – TERN studies have thus far recruited over 10,000 participants to the NIHR Trauma and Emergency Care portfolio. Already a legacy to be proud of. Well done everyone, you’ve smashed it. 

But what about the actual research? TERN studies should of course be held to the same rigorous standards of critical appraisal that we use for any published studies. So, let’s get to it, with a bit of insider info…

The Podcast

The Study

This manuscript summaries the Subarachnoid Haemorrhage in the Emergency Department (SHED) project, evaluating the diagnostic approach to acute severe headache. This was a big issue when we originally formed TERN and selected this project, based on high ranking in the initial EM James Lind priority setting research partnership exercise and trainee consensus. [1] The question at the time was whether any strategy using a clinical decision rule or CT only approach was safe enough to rule out subarachnoid haemorrhage (SAH) in patients attending the ED with acute severe headache. During delivery of the study, we saw further work on this topic published by Jeff Perry, [2] a new iteration of NICE guidelines on aneurysmal SAH (note guidance for aneurysmal SAH only) which suggested not routinely offering lumbar puncture if a CT could be completed within 6h of headache onset [3] and a fairly lukewarm neurovascular response to this work. [4] We also saw NICE research recommendations asking for further work on the diagnostic test characteristics of CT brain imaging for SAH up to 24h from symptom onset. [3]

As such, our original research question appears to have stood the test of time. There is general acceptance that after a SAH bleed, the blood becomes harder to see on CT brain imaging as time passes, due to CSF dilution and blood breakdown. [5] The real questions here are whether:

  1. Clinical teams can exclude SAH based on clinical features or a prediction rule alone
  2. Non-specialist radiologists can accurately report small amounts of SAH blood and 
  3. At what timepoint does it become unsafe to rely solely on CT brain imaging for exclusion of SAH? 

As this study took us 8 years, due to new evidence from New Zealand, a new question of whether the identification of blood on CT brain was better for aneurysmal against non-aneurysmal SAH also became an important consideration. [6] The abstract is below and the full article can be found here. Please give it a read yourself and we encourage your critical appraisal, in addition to our comments below.

The Abstract

Background: People presenting to the ED with acute severe headache often undergo investigation to exclude subarachnoid haemorrhage (SAH). International guidelines propose that brain imaging within 6 hours of headache onset can exclude SAH, in isolation. The safety of this approach is debated. We sought to externally validate this strategy and evaluate the test characteristics of CT-brain beyond 6 hours.

Methods: A prospective, multicentre, observational cohort study of consecutive adult patients with non-traumatic acute headache presenting to the ED within a UK National Health Service setting. Investigation, diagnosis and management of SAH were all performed within routine practice. All participants were followed up for 28 days using medical records and direct contact as necessary. Uncertain diagnoses were independently adjudicated.

Results : Between March 2020 and February 2023, 3663 eligible patients were enrolled from 88 EDs (mean age 45.8 (SD 16.6), 64.1% female). 3268 patients (89.2%) underwent CT-brain imaging. There were 237 cases of confirmed SAH, a prevalence of 6.5%. CT within 6 hours of headache onset (n=772) had a sensitivity of 97% (95% CI 92.5% to 99.2%) for the diagnosis of SAH and a negative predictive value of 99.6% (95% CI 98.9% to 99.9%). The post-test probability after a negative CT within 6 hours was 0.5% (95% CI 0.2% to 1.3%). The negative likelihood ratio was 0.03 (95% CI 0.01 to 0.08). CT within 24 hours of headache onset (n=2008) had a sensitivity of 94.6% (95% CI 91.0% to 97.0%). Post-test probability for SAH was consistently less than 1%. For aneurysmal SAH, post-test probability was 0.1% (95% CI 0.0% to 0.4%) if the CT was performed within 24 hours of headache onset.

Conclusion: Our data suggest a very low likelihood of SAH after a negative CT-brain scan performed early after headache onset. These results can inform shared decision-making on the risks and benefits of further investigation to exclude SAH in ED patients with acute headache.

Subarachnoid haemorrhage in the emergency department (SHED): a prospective, observational, multicentre cohort study. Emergency Medicine Journal . Published Online First: 12 September 2024. doi: 10.1136/emermed-2024-214068

Tell me about the study aims

The aims of the study were to externally validate both the Ottawa SAH clinical decision rule and a diagnostic strategy using negative CT brain imaging within 6 hours of headache onset to exclude SAH. If it is safe to reassure a certain proportion of patients with severe headache based on CT only, it is clearly in everyone’s interest that we do so. The risks of LP include hospital admission, overnight stay, nosocomial complications and the complications of the LP itself such as failure, infection, post LP headache and inconclusive result.

Who were the patients?

We preselected broad inclusion criteria, of adult patients (18 and above) who were awake and fully orientated (GCS 15) and presented with non-traumatic acute headache reaching maximal intensity within one hour. We have been asked about these inclusion criteria at peer review and by neurovascular colleagues already, as some clinicians feel these criteria do not fit their internal definition of a ‘thunderclap’ headache and suspected SAH case. This is an interesting but refutable point; criteria were selected based on prior high quality research on this topic; thunderclap headache has a fairly arbitrary and fluctuant definition in the literature; not all SAH cases present with classic thunderclap headache (so it was important for us to be inclusive); and when we actually look at our patient group in the results, we found that >85% of our sample had a headache which reached peak within 10 mins of onset (table 1). 

We excluded patients if they met any of the following criteria: direct head trauma in the previous 7 days; returning for reassessment of the same headache within the recruitment period; established prior diagnosis of SAH; known brain neoplasm; known ventricular shunt or hydrocephalus prior to attendance at the ED; focal neurological deficit; headache with onset >14 days prior to attendance; recurrent headaches (three or more headaches of similar character and intensity as presenting headache); transfer from another hospital with confirmed SAH; prisoners and patients currently detained under the Mental Health Act. We registered the study prospectively at ISCTRN and the protocol is freely available on line.     

Tell me about the methods?

This was not a randomised trial or implementation study. We selected a prospective observational cohort design, allowing clinical teams to evaluate and manage patients with acute headache in accordance with their routine practice. We introduced an additional data collection timepoint to ensure clinical headache features were accurately documented in a prospective fashion (prior to brain imaging), as our previous work clarified that subjective features and headache timings are poorly recorded. In particular, we ensured that all features of the Ottawa SAH rule were captured prior to investigation. We also introduced an additional defined follow up period of 28 days. Participants were followed up through case note review, regional imaging resources, direct contact and primary care contact as required. The aim was to identify a reference standard outcome for all patients by 28 days from presentation. 

What outcome measures were used?

This is where it got slightly interesting – did you know there is no universally accepted definition for SAH? We didn’t at the start. We therefore looked around and decided to use the criteria suggested by Perry et al within previous research, including the usual suspects such as SAH reported on CT brain, SAH on CTA or MRA, visible xanthocromia on LP or blood in the CSF and an aneurysm detected. However, we also slightly adapted the outcome definition to include spectrophotometry cerebrospinal fluid (CSF) findings consistent with SAH according to UK clinical biochemistry guidance. [7] The latter criteria for diagnosis is a little debatable, so if anything our criteria could be accused of being over diagnostic (classing more cases as SAH than other studies would). 

For those cases where the diagnosis was unclear after 28 day follow up, we convened an independent panel including neurosurgeons, neurointensivists, acute physicians and emergency physicians. The panel discussed each case and allocated a reference standard diagnosis of SAH positive or SAH negative. 

It is worth highlighting here I think that we were reliant on diagnoses being achieved in standard practice, which mainly occurs by CT, so our study design introduced a level of verification bias; if CT is your index test, but also forms part of your reference standard diagnosis, this can be an issue. It is worth highlighting here that we followed up all patients for 28 days to provide assurance on the accuracy of early negative CT imaging and that although it would be ideal to send all patients for digital subtraction angiography to determine whether a) they have bled and b) they have an aneurysm, this would clearly be unethical. 

Any clever stats to talk about?

Not really – the whole idea of this project was to keep things simple. We wanted to know the presenting features for people attending the ED with acute headache, and whether they went on to achieve a reference standard diagnosis of SAH within the next 28 days. We then wanted to look principally at the diagnostic test characteristics for different strategies, including the Ottawa SAH rule, CT within 6h of headache onset, CT within 12h of onset, CT within 24h of onset. We used an independent statistician to analyse the data and report diagnostic test characteristics, including likelihood ratios, with confidence interval estimates. 

Worth also mentioning that we excluded patients with a serum Hb <100g/L from the primary analysis. Interestingly, previous studies have raised concerns about the ability to detect bleeds when there is a low red cell concentration in the serum. 

Sample size is always tricky in diagnostic studies. We assumed a prevalence of 5% for SAH and powered the study looking to achieve a lower limit of the 95% CI for sensitivity as above 98%. This is clearly a very high bar indeed, which is why the original sample size came out at 9000. 

Drum roll please….. and the results?

Headline figures first – we recruited 3663 eligible patients and 89.2% had a CT brain scan ordered within routine practice – this speaks volumes about the ED approach to acute severe headache in a contemporary setting and shows how we have moved forwards over the last few decades. A reference standard diagnosis of SAH was seen in 6.5%, a little higher than our original estimates, which speaks to the validity of our protocol. 87.8% of SAH diagnoses were made on initial CT. 9.7% were diagnosed by LP. Interestingly, only half of SAH diagnoses (56.1%) were consequently judged to be aneurysmal, the remainder classed as non-aneurysmal or unclear. 

772 (23.9%) of patients presented and underwent CT brain imaging within 6 hours of headache onset. CT imaging was 97% sensitive (95% CI 92.5 to 99.2%) for the diagnosis of SAH in this group, conveying a negative likelihood ratio of 0.03 and a post test probability of 0.5% (95% CI 0.2 to 1.3). 3 cases were reported in the study database as missed by CT imaging – two were diagnosed by LP and classified as non-aneurysmal SAH after further imaging, with no requirement for intervention. The other case was a complex and slightly unusual approach to diagnosis, described further in the paper. 

A lot of lumbar punctures were done in this cohort. Out of 3021 participants with a negative CT brain scan, 1034 LPs were performed, with 25 positive results. Thus, if you were referred for LP after a negative CT brain in the current climate, the chances of a positive result were 2.4% 

The sensitivity of CT brain imaging for any SAH decreased over time up to 24h, as expected. However, the post test probability for aneurysmal SAH after a negative CT brain was 0.1% (95% CI 0.0 to 0.4) up to 24h. 

Finally – although the Ottawa CDR had a high sensitivity for the diagnosis of SAH, specificity was dreadful and the rule conveyed minimal impact on post test probability. The actual utility of a rule that very rarely rules out disease, and when it does can still occasionally be wrong, is pretty questionable. 

Any interesting subgroups?

Well, aneurysmal vs non aneurysmal SAH is the one that potentially stands out for us. In these cases of non-aneurysmal disease diagnosed in practice, we saw no downstream neurosurgical interventions. Patients were managed conservatively and usually discharged after a short period of observation, with a future clinic appointment. There is potential for this diagnosis to impact DVLA driving rules and patients may well be interested to know of course, but there is a real question about diagnostic implication to be asked here – if we are recommending hospital admission and LP to mainly diagnose a condition which is managed conservatively without further intervention, are the risks of investigation really justified?  

Likewise – if the post-test probability of aneurysmal SAH after a negative CT up to 24h is 0.1%, these data can potentially inform shared decision-making conversations on the risks and benefits of hospital admission for further testing. It’s worth taking a second to think about why we are seeing these results when we know the sensitivity of CT dips up to 24h. Our hypothesis is based on two assumptions – 

1) Aneurysmal SAH is arterial in nature, therefore is more likely to have a larger volume bleed, therefore more likely to be identifiable on CT. Non-aneurysmal SAH is likely to be venous in origin, therefore more likely to be smaller volume bleeds, therefore more likely to be missed on CT but picked up on LP. [8-10]

2) Due to the arterial nature of an aneurysmal bleed and the larger haemorrhage, patients with aneurysmal rupture may be more likely to present earlier with more severe symptoms. The earlier you present, the earlier your CT is performed, and the less time blood has to dilute and breakdown. The concept of self-triage in headache has been well established previously, with those patients presenting to secondary care tending to have far higher rates of pathology than those attending primary care. This is similar logic.  

Interesting – but what does it all mean for our practice?

Well, therein lies the rub. Our data suggest that patients who present early with acute severe headache and have a negative CT brain scan within 6h (and even up to 24h) are highly unlikely to have aneurysmal SAH. It would appear clinical teams have already got this message. Two thirds of our participants with negative CT brain scans were discharged without LP, many likely to be outside a 6h window in this cohort. Our data can at least further inform these conversations and quantify the risks and benefits to patients. They can also potentially help separate the complex issues of aneurysmal and non-aneurysmal SAH in an understandable manner. 

Do these data really support stretching the CT time window? I’m not so sure – although the message is an interesting one, we need further implementation research to evaluate any strategy that deviates from NICE guidance. We report a post test probability of SAH as >1% in patients with a negative CT after 18h, with the upper limit of 95% CI reaching >4% in the 18-24h group. This would clearly be an unacceptable miss rate by anyone’s standards. This data leans heavily towards non-aneurysmal SAH and confidence interval ranges are potentially inflated by our lower sample size, but even so caution should be heavily exercised when applying these findings to practice. Do these data support using the Ottawa SAH rule in an NHS population. We think we can be more conclusive here and say no, not really. These findings fit with other attempted UK validation studies and someone really needs to put forward a powerful argument for use of this tool, if it is to be further explored. 

Any limitations?

We only recruited a third of our original sample size and do not report screening to recruitment conversion rates, therefore we cannot offer any guarantee of a consecutive sample. This occurred due to multiple logistical issues in the main, including several periods of cessation where research activity was prioritised towards the pandemic and the challenge of utilising rotational trainees as principal investigators. 

We have discussed the potential issue of verification bias through incorporation of CT head as both the index test and part of the reference standard. Again, we would highlight that 28 day follow up is an excellent proxy marker for missed SAH and invasive angiography for all would clearly be unethical. 

An observational study such as this does not necessarily accurately evaluate the implications of a standardised 6h or 12h CT rule out strategy in practice. We allowed clinicians to manage cases in keeping with routine care. Thus, clinicians may have referred patients with a negative CT within 6h for LP based on clinical gestalt, or likewise discharged patients presenting at 12-24h without CT imaging. Indeed, just over 10% of our cohort had no CT imaging at all within the context of the study. Although these issues may affect our sensitivity and specificity estimates for different diagnostic strategies, we would argue that pragmatic evaluation was a strength to the study. All patients were followed for 28 days and the adaptive diagnostic approaches to certain patient groups reflects real world practice. Patients with a negative CT within 6h who had ongoing severe symptoms would, we suspect, be more likely to be admitted for further work up. There is nothing wrong with this adaptive approach – even NICE guidelines are aimed at 80% of the population in the main, accepting that clinicians will sometimes deviate from recommendations in specific circumstances. Our work captures some of this adaption, essentially reporting as an ‘intention to manage’ rather than ‘per protocol’ approach.

There are other limitations reported in the manuscript, mainly inherent to observational research methods. Take a look and consider while you are digesting the findings. 

What do you think the neurovascular surgeons will make of all this?

We met with representatives from the Society of British Neurological Surgeons and the British Neurovascular Group last week to discuss this very issue. It was a positive meeting and very civil. They were quick to highlight the catastrophic consequence of missed disease and reiterated their feeling that current NICE guidance is too dismissive of the role for LP in diagnosis. We discussed this at length in light of our data, which actually shows a fairly high rate of LP in this group, a very low yield from LP and a very high sensitivity for aneurysmal SAH. 

We discussed some potential audit work they have been undertaking, which again shows a reasonable rate of LP confirmed diagnosis, but does not at present offer information on timings. We discussed the potential role for early CT angiography in this group prior to LP; they are clearly against this strategy due to concerns around incidental aneurysm detection and a potentially increased risk of intervention. We discussed the issue of aneurysmal vs non-aneurysmal haemorrhage and our take on this – we perhaps differed here, as they were clearly more interested in the idea of finding blood in the first place and then trying to work out if it is aneurysmal or non-aneurysmal later, whereas we focussed a little more on the merits of diagnosis.

What are the next steps for research in this area?

That’s a really interesting topic. Have we seen enough here to consider an implementation study of CT rule out at 12, 18 or 24h? Or how about a randomised diagnostic trial of early CTA vs conventional testing strategies? Or how about a qualitative study on shared decision making informed by these risk estimates? NICE research recommendations will continue to attract attention, but we think that with all the focus presently on this condition it may not be long before we see a commissioned call.  

Anything else?

Just a huge thank you to the TERN collaborators, investigators, analysts, sponsor, supporting research teams and writing committee. This is your data and your study and you have achieved in a relatively short space of time what can often take decades in the research world. Hats off to everyone. Now on to the next project!

Best wishes

Tom Roberts (Inaugural TERN fellow and SHED lead) and Dan Horner (SHED Chief Investigator)

References

  1. Jason, S., et al., An Emergency Medicine Research Priority Setting Partnership to establish the top 10 research priorities in emergency medicine. Emergency Medicine Journal, 2017. 34(7): p. 454.
  2. Perry, J.J., et al., Validation of the Ottawa Subarachnoid Hemorrhage Rule in patients with acute headache.CMAJ: Canadian Medical Association journal = journal de l’Association medicale canadienne, 2017. 189(45): p. E1379-E1385.
  3. Nice, Subarachnoid haemorrhage caused by a ruptured aneurysm: diagnosis and management [A] Evidence review for symptoms and signs. 2022.
  4. British Neurovascular, G., Please read the attached statement from @The_SBNS and @BritNeurovasc in conjunction with the guideline on #SAH published today by @NICEComms https://t.co/yRgIwelVl4. 2022.
  5. Edlow, J.A., A.M. Malek, and C.S. Ogilvy, Aneurysmal subarachnoid hemorrhage: update for emergency physicians. The Journal of Emergency Medicine, 2008. 34(3): p. 237-251.
  6. Vincent, A., et al., Sensitivity of modern multislice CT for subarachnoid haemorrhage at incremental timepoints after headache onset: a 10-year analysis. Emergency Medicine Journal, 2022. 39(11): p. 810-817.
  7. Cruickshank, A., et al., Revised national guidelines for analysis of cerebrospinal fluid for bilirubin in suspected subarachnoid haemorrhage. Annals of Clinical Biochemistry, 2008. 45(3): p. 238-244.
  8. Konczalla, J., et al., Non-aneurysmal non-traumatic subarachnoid hemorrhage: patient characteristics, clinical outcome and prognostic factors based on a single-center experience in 125 patients. BMC Neurology, 2014. 14(1): p. 140.
  9. Mohan, M., et al., Subarachnoid haemorrhage with negative initial neurovascular imaging: a systematic review and meta-analysis. Acta Neurochirurgica, 2019. 161(10): p. 2013-2026.
  10. Tarkiainen, J., et al., The clinical course and outcomes of non-aneurysmal subarachnoid hemorrhages in a single-center retrospective study. Acta Neurochirurgica, 2023. 165(10): p. 2843-2853.

Podcast Transcription


Where to listen

You can listen to our podcast in numerous ways, ensuring you never miss an episode no matter where you are or what device you’re using. For the traditionalists, Apple Podcasts and Google Podcasts offer easy access with seamless integration across all your Apple or Android devices. Spotify and Amazon Music are perfect for those who like to mix their tunes with their talks, providing a rich listening experience. If you prefer a more curated approach, platforms like Podchaser and TuneIn specialize in personalising content to your tastes. For those on the go, Overcast and Pocket Casts offer mobile-friendly features that enhance audio quality and manage playlists effortlessly. Lastly, don’t overlook YouTube for those who appreciate a visual element with their audio content. Choose any of these platforms and enjoy our podcast in a way that suits you best!


Cite this article as: Dan Horner, "The Subarachnoid Haemorrhage in the Emergency Department (SHED) Study," in St.Emlyn's, September 26, 2024, https://www.stemlynsblog.org/the-subarachnoid-haemorrhage-in-emergency-department-study/.

Thanks so much for following. Viva la #FOAMed

Scroll to Top