Date: October 27th, 2015

Guest Skeptic: Dr. David Sayer is a physician completing his general practice training in the United Kingdom.

Case Scenario: A 34-year-old woman presents with acute onset of headache peaking in 30 minutes with no recent trauma, focal deficits and a normal neurologic examination.

Background: Headaches represent around 2% of all emergency department visits. Of these presentations 1-3% turn out to be a subarachnoid hemorrhage (SAH)  (EdlowVermeulenPerryMorgenstern).

About 5% of SAH are misdiagnosed on the first emergency department assessment (Vermeulen). This is partly because 50% of SAH present with no neurologic deficit (Weir).

Dr. Jeff Perry and his team have tried to create a clinical decision tool to rule out SAH for acute headaches (SGEM#48). The Ottawa SAH Tool contains six variables to decide if a CT scan is necessary.

Applying the tool could decrease the miss rate of SAH from about five percent down to almost zero with a slight increase in utilization. However, the tool needs further evaluation in implementation studies before it is ready for “prime time”.

Traditional methods of working up a SAH has been non-contrast CT followed by a lumbar puncture (LP). Dr. David Newman has questioned this dogma on his SMART-EM podcast. He suggested LPs are not always needed after a negative CT scan.

Dr. Newman calculated the number needed to LP to identify one SAH for which an intervention was indicated to be 700, prompting the question “are you part of the ‘700 Club’?” Should any of us be part of the 700 Club?

Clinical Question: In emergency department acute headache patients, how frequently does LP diagnose SAH after unremarkable CT scan?

Reference: Sayer et al. An Observational Study of 2248 Patients Presenting with Headache, Suggestive of Subarachnoid Hemorrhage, that Received a Lumbar Puncture Following a Normal CT Head.  Acad Emerg Med Nov 2015

  • Population: Adult patients (2248 total, >17 years old) presenting to one of six urban EDs in the United Kingdom with acute headache suspicious for subarachnoid hemorrhage, who had both a negative CT and a lumbar puncture performed.
  • Intervention: LP to achieve identification of CT-negative SAH patients
  • Comparison: None
  • Outcome: Incidence of positive LP (defined only by spectrophometric detection of bilirubin, not by any CSF RBC count) and proportion with cerebral aneurysm identified

Authors Conclusions: In patients presenting to the emergency department with acute, non-traumatic severe headache, LP to diagnose or exclude SAH after negative head CT has a very low diagnostic yield, due to low prevalence of the disease and uninterpretable or inconclusive samples. A clinical decision rule may improve diagnostic yield by selecting patients requiring further evaluation with LP following non-diagnostic or normal non-contrast CT brain imaging.

Quality Checklist for A Diagnostic Study:

  1. checklistThe clinical problem is well defined. YES. Headache represents 1-2% of all ED visits and SAH is one of the most devastating etiologies of headache, yet considerable controversy exists about whether an LP is essential to rule-out the diagnosis after a “negative” CT using contemporary high-resolution scanners.
  2. The study population represents the target population that would normally be tested for the condition (ie no spectrum bias). YES. Emergency department patients with concerning headaches.
  3. The study population included or focused on those in the ED. YES
  4. The study patients were recruited consecutively (ie no selection bias). NO.  Retrospective chart review.
  5. The diagnostic evaluation was sufficiently comprehensive and applied equally to all patients (ie no evidence of verification bias). NO. Although every patient had a CT and LP, neither of these tests is the gold standard for SAH, which would be CT angiogram, traditional cerebral angiogram, MRI/MRA, or 3-month follow-up.
  6. All diagnostic criteria were explicit, valid and reproducible (ie no incorporation bias). NO
  7. The reference standard was appropriate (i.e. no imperfect gold-standard bias). YES
  8. All undiagnosed patients underwent sufficiently long and comprehensive follow-up (i.e. no double gold-standard bias). NO
  9. The likelihood ratio(s) of the test(s) in question is presented or can be calculated from the information provided. NO  There was no 2×2 data provided.
  10. The precision of the measure of diagnostic performance is satisfactory. NO

Key Results: The population was 45% male with a mean age of 41 years. Using the authors’ spectrophotometric criteria for the total population evaluated (2,248 patients), the LP results broke down as follows:

  • 4% positive
  • 13% inconclusive
  • 16% un-interpretable
  • 67% negative

Of the 92 “positive” LP results, 9 were identified with an aneurysm (9/2248 = 0.4%), which equates to 250 LPs to identify one aneurysm.

Number Needed to Tap (NNTap) of 250 to diagnose one aneurysm not picked up on CT scan.

Screen Shot 2015-04-25 at 3.11.12 PMThis was a retrospective study including acute, non-traumatic adult headache patients with suspected SAH presenting to one of six urban training EDs in the UK between 2006 and 2011. Eligible subjects had a non-diagnostic head CT and had a lumbar puncture performed.

Dr. David Sayer

Dr. David Sayer

Five Limitations/Questions (listen to podcast for Dr. Sayers responses):

  1. Spectrophotometry to Evaluate CSF: The authors’ note using appropriate chart review methods and evaluated only spectrophotometric CSF analysis, not CSF RBC counts or visual xanthochromia. This outcome may be problematic since 99% of North American hospitals use visible xanthrochromia rather than spectrophotometry to evaluate for CSF bilirubin. Emergency department providers at centers that lack spectrophotometry would benefit from understanding the sensitivity, specificity, and likelihood ratios and interval likelihood ratios for CSF RBCs, although that becomes problematic with traumatic taps. However, traumatic LPs are a real-world problem and these authors had access to LP results that could have been analyzed as a secondary outcome.
  2. Differential Verification Bias (Double Gold Standard): This occurs when the test results influence the choice of the reference standard. So a positive index test get an immediate/gold standard test whereas the patients with a negative index test get clinical follow-up for disease. This can raise or lower sensitivity/specificity. Since only LP-positive patients routinely underwent additional neuroimaging (CTA or MRA), this study is at risk for differential verification bias that lowers estimates of sensitivity and specificity for disease processes that only become apparent during periods of follow-up (Understanding the Direction of Bias). Un-interpretable LPs were only evaluated at two (out of six) sites and only 5/28 (18%) and 17/56 (30%) had further imaging at those sites so unrecognized cerebral aneurysms probably occurred. Prospective studies would have the benefit of routine criterion standard testing for all patients or alternatively a period of follow-up to ensure that “negative” LP patients were not false-negatives with subsequent symptomatic SAH at a later date.
  3. Chart Review Methods: The authors’ reference chart review methods (Gilbert et al), but they do not describe them explicitly in their methods. Who abstracted the data from the medical records? How were these individuals trained and monitored? Were they blinded to the study hypothesis? Was a standardized abstraction form used? Was inter-rater reliability of chart abstraction assessed for key variables? Without understanding the authors’ specific chart abstraction methods, it is not possible to meaningfully evaluate the possibility of bias in this study.
  4. More Details on Time to CT and Time to LP: Since CT is less accurate for SAH beyond 12-hours after the onset of headaches, additional details about the average delay between headache and imaging is important to understand. Also, theoretically at least 12-hours must elapse between headache onset (sentinel bleed) and formation of CSF bilirubin, so the timing of headache onset and LP should also be reported.
  5. Temporal Bias: Diagnostic tests that rely upon evolving imaging technology are sometimes at risk for temporal bias in which improved ability to obtain high-quality images or finer anatomical cross-sections yield more accurate results (Evidence-Based Emergency Care: Diagnostic Testing and Clinical Decision Rules, 2nd Edition, 2013, pages 54- 64). Since CT technology was evolving between 2006 and 2011, readers should interpret these results conservatively in 2015 and beyond. In other words, when multi-slice CTs (64-slice, 128-slice) are used in your ED today they probably detect CSF blood with even higher resolution than in 2006, resulting in higher sensitivity of the initial CNS imaging, and an even higher number needed to LP than this study suggests.

Comment on authors conclusion compared to SGEM Conclusion: In the setting of acute, non-traumatic headache presenting to the ED, a multi-slice CT (16-64 slice) that does not demonstrate radiographic evidence of SAH is likely sufficient to rule-out a SAH in most patients. In fact, the number needed to tap (NNTap) to detect one aneurysm in this CT-negative population is 250 using a spectrophotometric definition of abnormal LP, which constitutes a significant amount of patient risk and discomfort, expense to healthcare payers, and delays in care for others awaiting treatment with no benefit for 249/250 patients.

Also, an additional 33% of CT-negative/LP equivocal patients with positive, inconclusive, or un-interpretable CSF results would still require additional imaging (CTA or MRA) to truly rule out SAH. Although an accurate, reliable, well-validated clinical decision instrument with a convincing impact analysis would be a useful adjunct to clinical gestalt, such an instrument does not currently exist and was not tested in this study.

Future studies that evaluate the role of LP in CT-negative headache patients with suspected SAH should prospectively assess outcomes similarly in all patients to avoid differential verification bias, while reporting on both visual/spectrophotometric CSF bilirubin and overall CSF RBC diagnostic accuracy (including likelihood ratios and interval likelihood ratios).

SGEM Bottom Line: In this study, one patient would be diagnosed with SAH out of every 250 patients receiving a LP who presented to the emergency department with a headache that did not have their bleed identified on CT scan.

Case Resolution: You are clinically concerned and get a non-contrast CT head that is negative. You discuss the risks and benefits of an LP with the patient. A shared decision is made with the patient not to do an LP. She is discharged home with appropriate analgesia. She is to return to the emergency department if she develops focal neurologic symptoms, pain increases, loss of consciousness, seizure or is otherwise worried.

Clinically Application: Contrary to classic teaching, LP in the setting of acute, non-traumatic headache of <12 hours duration concerning for SAH rarely yields the elusive cerebral aneurysm diagnosis and is often falsely positive or inconclusive.

What do I tell my patient? Sudden onset headaches are clinically concerning for SAH, which can have devastating long-term consequences if misdiagnosed. Unfortunately, migraine headaches outnumber SAH headaches 50:1 in ED settings and do not benefit from CT-imaging of your brain, evaluation of your cerebrospinal fluid, or surgery. Modern CT scanners are very good at identifying blood in your brain from a SAH, if the CT is obtained within 12-hours of when the headache began. Older research indicated that only 9/10 SAH were detected by a CT so evaluation of fluid around your brain (CSF) was recommended to be completely certain that SAH was not the cause of the headache. Newer research indicates that CT scans almost never miss the diagnosis of SAH and that 250 lumbar punctures (LP) are needed to identify one cerebral aneurysm that CT missed. LPs are not benign procedures and can cause post-LP headaches, infections, nerve damage, and bleeding around your spinal cord.

Keener Kontest: Last weeks winner was Gregory Rodgers. He knew that Willem Einthoven was the Dutch physician who designed the first practical ECG in 1903 and won the Nobel Prize in medicine in 1924 for the invention.Listen to the podcast for this week’s question. If you know the answer, then be the first person to email TheSGEM@gmail.com with “keener” in the subject line to win a cool skeptical prize.SGEM-HOP-SAEM-logo-227x300


SGEM-HOP Social Media Feedback: Respond early and often to this podcast for a chance to have your comments published. You can use the SGEM blog, Twitter, or Facebook, but you only have one-week to respond before this carriage turns into a pumpkin. At the end of 1-week, we’ll pick the Top Five Social Media comments to publish in a future edition of AEM.

Remember to be skeptical of anything you learn, even if you heard it on the Skeptics’ Guide to Emergency Medicine.