Diagnosis of Subarachnoid Hemorrhage

January 2017



Mrs. Z. is a healthy 30-year-old female who presents to your emergency department 2-hours after onset of “the worst headache of my life” which peaked within 1 minute of onset but was not “thunderclap”. She describes the headache as diffuse with associated nausea, but no photophobia, neck stiffness, vomiting, focal motor-sensory deEicits, or fever. She reports no recent viral or febrile illness, blunt trauma, travel history, or sick contacts. Nobody else at home or work has had a headache recently. Her immunizations are up-to-date and she denies any signiEicant PMH including no history of migraine (or other) headache disorders, cerebral aneurysm, stroke, HIV or other immunocompromising disorder, pseudotumor cerebri, malignancy, bleeding diatheses, or meningitis. She denies recent alcohol use or any history of intravenous or illicit substance abuse. She has not tried any medications for her headache. She is a secretary and lives at home with her husband and 3 year old child.

In the ED, her vitals are BP 106/55, P 58, RR 18, T 37.2°C, and 100% oxygen saturation on room air. She is in no acute distress, and her exam is unremarkable.

As you consider the diagnosis of subarachnoid hemorrhage and await her CT, you weigh the value of a post-CT lumbar puncture. You recall a recent meta-analysis discussed on your favorite “podcast” Emergency Medical Abstracts (listen here) as well as an episode of McMaster’s Textbook of Internal Medicine (watch here) and wonder how other researchers and specialties interpret the best evidence. Realizing that there are two distinct questions for the ED diagnosis of SAH, you develop two PICO questions.

PICO Question #1

Population: ED patients with sudden onset severe headache concerning for aneurysmal SAH (aSAH)

Intervention: Computed tomography within 6 hours of symptom onset

Comparison: No comparator

Outcome: CT sensitivity, specificity, positive/negative likelihood ratio for aSAH

PICO Question #2

Population: ED patients with sudden onset severe headache concerning for aneurysmal SAH (aSAH)

Intervention: Lumbar puncture following unremarkable CT

Comparison: No comparator


LP sensitivity, specificity, positive/negative likelihood ratio for a SAH

Search Strategy
Using these PICO questions, you devise a PubMed search using Clinical Queries (diagnosis/broad) and the search term “aneurysmal subarachnoid hemorrhage” which identifies 2891 studies (see http://tinyurl.com/zrvgc3s) from which you select the 4 studies below from last year:


Article 1: Sensitivity of early brain computed tomography to exclude aneurysmal subarachnoid hemorrhage: A systematic review and meta-analysis, Stroke 2016; 47:750-755).  (http://pmid.us/26797666) ANSWER KEY

Article 2:  False-negative interpretations of cranial computed tomography in aneurysmal subarachnoid hemorrhage, Acad Emerg Med 2016: 23: 591-598 (http://pmid.us/26918885)

Article 3:  Spontaneous subarachnoid hemorrhage: A systematic review and meta-analysis describing the diagnostic accuracy of history, physical examination, imaging, and lumbar puncture with an exploration of test thresholds, Acad Emerg Med 2016; 23: 963-1003 (http://pmid.us/27306497)

Article 4: Determination of a testing threshold for lumbar puncture in the diagnosis of subarachnoid hemorrhage after a negative head computed tomography: A decision analysis, Acad Emerg Med 2016; 23: 1119-1127. (http://pmid.us/27378053)

Bottom Line

Needle in a Haystack

Headaches represent about 2% of emergency department (ED) visits annually. While severe headache patients presenting with altered mental status, fevers, or associated trauma usually generate sufEicient concern to justify further diagnostic evaluations, most other sudden onset headache cases ultimately prove to result from benign, non-life threatening causes like migraine headache. The myriad causes of sudden onset headaches include cough, exertion, and post-coital, but can also include potentially life-threatening conditions like sinus thrombosis, vascular dissection, intracerebral hemorrhage, vasospasm, and aneurysmal subarachnoid hemorrhage. [Landtblom 2002, Delasobera 2012, de Bruijn 1996, Pascual 1996, Dodick 1999] Observational studies indicate that migraine headaches are at least 50-times more common than SAH amongst ED headache patients, so SAH represents a needle in a haystack for a very common chief complaint. [Edlow 2003] Missed SAH diagnosis occurs between 12%-53% of cases with ED providers estimated to miss 5% of them. [Edlow 2000, Vermeulen 2007]

Up to 80% of SAH cases result from a ruptured cerebral aneurysm. Other causes of SAH include low-pressure perimesencephalic bleeds, cerebral amyloid angiopathy, vasculitis, sickle cell disease, and cocaine or amphetamine abuse. [Carpenter 2016] One-fourth of aneurysmal SAH victims die within one-day and 50% of SAH survivors never return to work. Correctly identifying SAH early can reduce these adverse outcomes if subsequent neurosurgical interventions (coiling or clipping) occur emergently. [Schievink 1997] Therefore, the possibility of aneurysmal SAH must be considered in ED patients presenting with severe headaches. Understanding the SAH diagnostic evidence available for bedside evaluation, advanced imaging, and the role for lumbar puncture (LP) is therefore essential and the landscape is shifting.

Non-Contrast Cranial CT

Computed tomography (CT) became widely available for the evaluation of headache patients about 30 years ago. Early CTs were 4-slice and radiologists’ interpretative learning curves were steep. These early CTs were imperfect (sensitivity ~90%) for identifying small amounts of blood in the subarachnoid space, so textbooks, guidelines, and several generations of emergency medicine trainees advised against a CT-only approach to rule-out aneurysmal SAH. Instead, SAH could only be ruled out when a negative CT was followed immediately by a LP demonstrating cerebrospinal Eluid (CSF) without either red blood cells or (at least 12 hours postheadache onset) xanthochromia. However, in 2017 there are three problems with that approach: (1) contemporary CTs are far better at identifying blood in the subarachnoid space; (2) LP’s frequently identify blood that is not in the subarachnoid space (traumatic LPs); and (3) CSF xanthochromia is not an accurate
diagnostic test for SAH.

A rule of thumb to deEine sufEicient diagnostic accuracy to rule-in a disease as positive is likelihood ratio (LR+) > 10 or negative likelihood ratio (LR-) < 0.1 to ruleout a disease. Both Perry 2011 and Backes 2012 demonstrate acceptably safe CT accuracy for SAH when imaging is obtained within 6-hours of headache onset (summary LR+ 235, summary LR- 0.01). Beyond 6-hours CT is still accurate to rulein SAH (summary LR+ 223) but less impressive to rule out SAH (summary LR- 0.07, 95% CI 0.01-0.61). These studies used neuro-radiologists to interpret the CT, which opens the question as to whether general radiologists’ accuracy would equate these studies. Recent studies indicate that general radiologists’ are just as accurate for the diagnosis of SAH as neuro-radiologists.

Post-CT LP– Quantifying BeneXits and Harms

Most European hospitals use spectrophotometry to evaluate the presence or absence of xanthochromia, whereas 99% of North American hospitals use visible
inspection. Recent meta-analyses indicate that visible xanthochromia can rule-in SAH (LR+ 25) but is less accurate to rule it out (LR- 0.22). At least four spectrophotometric methods to evaluate xanthochromia exist and most ED headache studies report different methods, but spectrophometry using any method
doesn’t appear to be signiEicantly better than visible xanthochromia to rule-in or rule-out SAH. [Carpenter 2016, Chu 2014] At a threshold of 2000 x 106 red blood
cells per liter in the Einal tube of CSF collected, the LR+ is 10.3 and the LR- is 0.07 (95% CI 0.01-0.49). [Perry 2015] Importantly, up to 30% of patients experience
worsening post-LP headache, in addition to the risks of post-LP back pain, epidural bleeding, and introduction of skin Elora into the central nervous system. [Seupaul 2005] Patients are not the only reluctant participants in this exercise; physician barriers to routine post-CT LP include inadequate time in the busy ED and
expectation of normal or non-diagnostic results in most patients (low diagnostic yield). In fact, up to 1 in 6 LPs are traumatic (blood detected from skin or superEicial soft tissue, not from the subarachnoid space). Consequently, LP is not performed in over half of acute headache patients in whom a CT is obtained. [Perry 2010, Perry 2013]

The ultimate objective is not to understand test accuracy; instead, the goal is to deliver the appropriate care to the right patients. How is a busy ED physician supposed to interpret all of these numbers and communicate with patients meaningfully to allow shared decision making? One approach to balance the harms
and beneEits of post-CT LP is to hypothesize test- and treatment-thresholds. The test-threshold describes the probability of a diagnosis (aneurysmal SAH) below
which continuing to test for the diagnosis will harm more patients than it will help, whereas above the threshold additional testing will beneEit more patients than will be harmed. This threshold is derived from the test accuracy (CSF xanthochromia or RBC), risk of the test (post-LP headache, infection, or epidural bleed), beneEit of the treatment for those with disease, and harms of the treatment for those without disease (since false-positives will receive the treatment and have no possibility of beneEit since they don’t actually have the disease). Based upon one recent diagnostic meta-analysis, the threshold at which post-CT LP would beneEit patients is quite narrow (2-4% for CSF RBC or 2%-7% for visible xanthochromia). Doubling the risk of beneEit and halving the risk of harm was used to further evaluate these thresholds and the thresholds didn’t change signiEicantly. Still not convinced? Thinking about the beneEit of LP alone (and neglecting the potential for harms associated with LP), the Number Needed to LP (NNLP) to identify one CNS infection in acute onset headache patients is 227 [Brunell 2013] while the NNLP to identify one additional aneurysmal SAH upon which neurosurgical intervention occurs and which was missed by CT within 6 hours ranges from 250 [Perry 2011, Sayer 2015] to 15200! [Blok 2015]

Journal Club Discussion Between Neurosurgery & Emergency Medicine

Here is the synopsis of the Journal Club discussion between Neurosurgery and
Emergency Medicine, including Dr. Sivilotti the senior author of the PGY-III metaanalysis.

CT within 6 hours of thunderclap headache onset is sufEiciently sensitive to rule out SAH without LP, if the CT is 16 slice or greater with cuts < 5mm.

Non-contrast head CT is an excellent test for ED headache patients in whom SAH is a concern, but like every other test that exists it is not 100% sensitive. Scenarios in which false negative CTs occur include imaging more than 6 hours after the onset of the headache, less severe bleeds (as measured by Hunt-Hess classiEication), anemia (Hg <10), and non-neuro-radiologists readers. In those scenarios, patient shared decision making about additional testing (CTA or LP) is warranted but should include informed consent about anticipated beneEits and potential harms.

Less than one in 10 headache patients concerning for SAH are ultimately diagnosed with SAH in recent studies. While certain symptoms and signs increase or decrease the likelihood of SAH, no single characteristic on history/physical exam is sufEicient to rule in or rule out SAH. Within 6 hours of symptom onset, noncontrast cranial CT is highly accurate, while a negative CT beyond 6 hours substantially reduces the likelihood of SAH. Based on anticipated treatment beneEit of 80% with immediate aneurysmal SAH (aSAH) diagnosis, LP risk of 1% (including signiEicant post dural headache, discomfort, and infection), and total (cancer + other) risk for angiography of 2%, LP would only be helpful in a pre-LP probability of aSAH range of 2-7% (if using visible xanthochromia) or 2-4% (if using CSF RBC > 1000). Therefore, LP appears to beneEit relatively few patients within a narrow pretest probability range. Scenarios in which false negative CTs occur include imaging more than 6 hours after the onset of the headache, less severe bleeds (as measured by Hunt-Hess classiEication), anemia (Hg <10), and non-neuro-radiologists readers. In those scenarios, patient shared decision making about additional testing (CTA or LP) is warranted but should include informed consent about anticipated beneEits and potential harms. With improvements in CT technology and an expanding body of evidence, test thresholds for LP may become more precise, obviating the need for a post-CT LP in more acute headache patients. Existing SAH clinical decision rules await external validation, but offer the potential to identify subsets most likely to beneEit from post-CT LP, angiography, or no further testing.

Decision analysis using accepted methods but sparse, low quality evidence-basis demonstrates post-CT LP test threshold that is signi>icantly higher than the majority of ED headache patients in whom SAH is considered. Nonetheless, the similar testthreshold point-estimate for post-CT LP in comparison to Carpenter et al (which used very different methods to derive a test-threshold), indicates that current textbook and guideline recommendations (which advocate post-CT for everyone) merit revision.

The discussion amongst the Journal Club attendees focused around two issues.

1) Some neurosurgeons expressed concerns about the risk of aneurysmal rebleeding within 24 hours, which has been recognized to be about a 6% risk since observational studies were conducted in the 1950’s. Some of the neurosurgeons also noted that no test is perfect, so their traditional approach has been that more testing is generally better for patients and providers. Dr. Sivilotti made the point that attempts to attain diagnostic perfection (0% miss rate) leads to operational inefEiciencies for the hospital and at eventually harms some patients.

2) Some noted uncertainty about how to derive the initial pre-test probability estimate by which to guide post-CT decision-making about the beneEits vs. harms of LP. It was noted that the PGY-III meta-analysis included weighted average prevalence of SAH of 7.5% in recent prospective studies of headache patients in whom SAH is a concern. This meta-analysis prevalence is one evidence-based estimate of pre-test probability. The point was then made that the 7.5% probability of SAH falls within the 2%-7% range for which LP beneEits outweigh risks, but the meta-analysis authors noted that the 7.5% prevalence estimate is pre-CT. Whether a CT is obtained less than 6-hours post-headache onset (summary LR- = 0.01 so post-CT probability of SAH = 0.1%) or later than 6-hours (summary LR- = 0.07 so post-CT probability of SAH = 0.6%). Either way, the post-CT probability of SAH is far below the lower margin of beneEit (2%). The issue with CT diagnostic accuracy beyond 6-hours is the 95% ConEidence Interval which extends to 0.61, so these situations require more careful contemplation and perhaps offer a role for CT
angiography – a proposal to which neurosurgery eagerly agreed requires additional research at Washington University.

 Future Research Priorities
Carpenter et al. was grounds for practice-change for many ED diagnosticians, but has also received criticism – mainly based on philosophical grounds. Is the role of emergency medicine to deEinitively rule-in or rule-out particular diagnoses or simply to exclude (or minimize) life-threats without putting a label of certainty on symptom etiology? Based upon the personal observations that thunderclap headache represents a potential life-threat that is not SAH in over half of cases and that a CT/ LP approach alone will fail to identify other life-threats without consulting Neurology and/or Neurosurgery with additional advanced imaging at the discretion of these consultants, some argue that diagnostic research based upon ruling-in or ruling out SAH alone is inherently Elawed and not worthy of altering the longstanding practice paradigm of CT then LP if CT is non-diagnostic. However, this approach neglects to consider the potential harms of LP, the number of additional studies required to save one additional life or avert one bad outcome, or the role for shared decision-making. It is worth noting that the perspective that thunderclap headache more commonly represents a non-SAH life-threatening diagnosis than SAH or migraine is counterintuitive to most ED research. This phenomenon likely reflects spectrum bias in that sicker patients with more likelihood of life-threatening disease and poor outcomes are being labeled as “thunderclap headache” than is the norm in published research and common practice. Indeed, a subset of acute headache patients who do beneEit from post-CT LP almost certainly exists, including those with hemoglobin less than 10 g/dL (increased risk of false-negative CT), signiEicant delay between headache onset and CT, and those in whom diagnoses like CNS infection or pseudotumor cerebri are being contemplated. The art of medicine is identifying the individual patients who are more likely to beneEit from additional testing rather than applying a one-size Eits all that mandates routine post-CT LP.

Many questions remain unanswered. Can general radiologists rule-out aneurysmal SAH as well as neuro-radiologists? Will clinical decision aids provide an additional risk-stratiEication mechanism by which to identify subsets more likely to beneEit from post-CT LP? Is there a role for CT angiography (CTA) following a nondiagnostic non-contrast CT? If so, which patients would beneEit from CTA?
Bottom Line: The decades old dogma that acute headache patients in whom SAH is a consideration must uniformly undergo an LP following a nondiagnostic
CT appears is unnecessary for a large subset of these patients and may leads to harms via additional downstream testing that results from the imperfect, non-speciXic Xindings in CSF.