Journal Club Synopsis: Subarachnoid Hemorrhage (SAH): Do We Still Need the LP?
Thanks to Lisa and Ted Toerne for hosting and for the ridiculous Smoque BBQ, and to presenters Lindsay, Beau, Natalie, Pikul, J Cash, and Brian for their thoughtful critiques.
Background: Headache is the chief complaint for 3% of ED visits. Mixed in with all the benign headaches are 3-4% of patients with headache + normal neuro exam who have serious pathology. One elusive serious diagnosis is the “sentinel bleed” of SAH. Traditional teaching states that a negative CT is not definitive, and debates rage on about how many hundreds of LPs we perform to identify a sentinel SAH. We do plenty of other high volume low yield tests in order to identify rare but life threatening disease (may I have another ECG please), but even putting aside the discomfort and time requirements of LP, there are potential downsides (post LP headache, the rare spinal hematoma/abscess, and the risk of false positive “bloody taps” leading to further unnecessary testing). With the publication of the Perry article in 2011, the question was raised yet again: when considering SAH, is current CT technology finally sensitive enough to obviate the need for LP after a negative CT?
Article #1: Presenters: Lindsay Purnell, Beau Willison
Cortnum, Søren MD, et al: Determining the Sensitivity of Computed Tomography Scanning in Early Detection of Subarachnoid Hemorrhage. Neurosurgery: May 2010 - Volume 66 -Issue 5 - p 900–905.
In this Danish retrospective study, medical records were reviewed from 499 patients referred to a neurosurgical unit due to suspicion of/confirmed SAH. All patients with a negative CT had a LP performed. Of 296 patients diagnosed with SAH, 295 were diagnosed on CT, and one was diagnosed on day 6 due to a positive LP. Post-LP headache rate was 7.4%. Of patients receiving LPs, 2% were diagnosed with viral meningitis. Authors concluded that CT had a sensitivity of 99.7% (95% CI 98.1-99.99%).
Sounds good. Unfortunately, their methods were horrible. A positive or negative LP was not defined, although they imply that they relied on xanthochromia. They also didn’t define their gold standard for the diagnosis of SAH. It’s unknown who was responsible for reading the CT scans (neurosurgeons ran the study, and did see the scans, but unknown if they were blinded to the study, and if they used an explicit data abstraction form…doubt it). There’s nothing about followup.
Also, there is the issue of referral bias. These are by definition sicker patients than our population, as they were referred to a neurosurgical center, and 60% were diagnosed with SAH. There is an unknown denominator of possible patients who could have been included, and young, healthy patients with bad headaches may not all get referred. Another way to think of it is in terms of spectrum bias, where a test performs differently in different patient populations. CT sensitivity is likely lower in less sick patients.
Bottom line: not practice changing due to poor methods and the referral/spectrum bias. Remember 2% of patients who received LP were diagnosed with viral meningitis-you may identify something besides blood.
Article #2: Presenters: Natalie Kmetuk, Pikul Patel
Jeffrey J Perry, Ian G Stiell, et al: Sensitivity of computed tomography performed within six hours of onset of headache for diagnosis of subarachnoid haemorrhage: prospective cohort study BMJ 2011;343:d4277.
In this prospective study conducted in 11 tertiary care EDs across Canada, sensitivity of head CT for the diagnosis of SAH was evaluated in 3132 consecutively enrolled neurologically intact adult patients presenting with new acute HA peaking within one hour of onset. SAH rate was 7.7%. Sensitivity of CT for SAH was 92% (95% CI 89-95.5%), with a specificity of 100% (trying to identify specificity is silly here due to incorporation bias-if you define a “positive” using a test you’re evaluating, it has to be 100% specific). The reason this paper has received so much attention is that for the 953 patients receiving CT within 6 hours of HA onset, CT sensitivity was 100% (95% CI 97-100%). All patients were scanned on third generation multi-slice scanners.
Limitations: Only half the patients received LP after negative CT, and the adequacy of follow-up has been questioned. One patient ultimately diagnosed with SAH who had CT within 6 hours of HA onset was sent home after CT was read as negative by the ED physician and a radiology trainee. Additionally, 2292 potentially eligible patients were not enrolled. External validity questioned-will this hold up outside of academic centers?
Bottom line: This article offers support for CT without LP when considering SAH, if CT is performed on a modern scanner within 6 hours of HA onset, and is read by a qualified radiologist. That being said, there are nationally respected emergency physicians with widely divergent opinions on whether this study is enough to change practice.
Article #3: Presenters: Jennifer Cash, Brian Fort
Dustin G. Mark, MD, et al: Nontraumatic Subarachnoid Hemorrhage in the Setting of Negative Cranial Computed Tomography Results: External Validation of a Clinical and Imaging Prediction Rule. ePub2012 / Annals of Emergency Medicine 2012.
This study attempts external validation of a clinical decision instrument identifying high- risk clinical features for SAH in neurologically intact patients with headache. The derivation study was by Perry et al, and published in BMJ in 2010. In that article, they evaluated 1999 patients with max onset headache intensity within one hour (130 cases of SAH), and identified 3 combinations of history and exam findings that were 100% sensitive in identifying SAH. This study by Mark et al attempted to externally validate one of these instruments (investigate patients with any combination of age > 40, neck pain or stiffness, LOC, or HA onset during exertion), as well as the timing of imaging decision instrument described in article #2 (CT within 6 hours of HA onset) using a matched case-control study
of 55 patients over 11 years with normal neuro exam and diagnosis of SAH after (-) CT/ (+) LP. The combination of high-risk clinical features demonstrated a sensitivity of 97.1% (95% CI 88.6-99.7%), specificity of 22.7% (95% CI 16.6-29.8%), and a negative LR of 0.13 (95% CI 0.03-0.61) for the diagnosis of SAH. Twenty percent of SAH cases had negative CT results when performed within 6 hours of headache onset. The authors conclude that the high-risk clinical features may have some Bayesian utility, but that a 6 hour CT cut-off missed 20% of SAH.
Limitations: Retrospective case-control study design, with methodology that started with known SAH patients, rather than with patients presenting with undifferentiated headache. Also, someone was worried enough about these patients to push for an LP after negative CT; it’s not the same as in a prospective study where all patients with negative CT get LP, regardless of how persuasive the clinician was feeling that day.
Bottom line: External validation studies for decision instruments often report lower sensitivities than the initial derivation studies, especially when studies move from academic centers to the community. These decision instruments will need to be re- tested in a number of different practice settings before making definitive conclusions about their accuracy, but the high-risk clinical features may help determine pre/post test probabilities.
Journal Club bottom line:
There was no consensus. Some in the room are ready to tell neurologically normal patients who are imaged for rule-out SAH within 6 hours of HA onset that if their CT is negative, they need no further evaluation. Others are not convinced, and given the high stakes of missing a sentinel bleed, will tell the same patient there is still up to a 1% of missed bleed (based on LRs, pre/post test probabilities-another discussion), and incorporate the patient’s risk tolerance into decisions about LP/further testing.
Certainly, these articles and the general discussion highlight the importance of “maximal intensity at onset” (within a few minutes) as the most important clinical feature of SAH, even more than “worst HA of life”. They also identify useful high risk clinical features (age > 40, LOC, neck pain or stiffness, HA onset during exertion, vomiting, SBP > 160, DBP > 100) that can help with pre/post test probabilities,
and impact which patients really need the hard sell for the LP after a negative CT, potentially even when CT is within 6 hours of HA onset. Final pearl-beware the patient who isn’t feeling better after analgesics and a negative CT-that’s another high-risk feature.