SAH Update: Are New CT Scanners Good Enough to Obviate the LP?
At our last Journal Club, Tom presented a 2005 paper from JEM on new CT scanners in the evaluation of SAH. The paper was called Subarachnoid Hemorrhage Diagnosis By Computed Tomography and Lumbar Puncture: Are Fifth Generation CT Scanners Better at Identifying SAH? by Boesiger and Shiber, and it appears in Journal of Emergency Medicine (2005: Vol. 29, No. 1 pp23-27).
The article is motivated by the fact that 1% of headache patients in the ED have SAH. Most are traumatic, but those that aren’t are usually from Circle-of-Willis aneurysm ruptures, which often kill or disable otherwise healthy people. EM physicians hate that sort of unsettling risk, and the situation is further complicated by the 20-50% of SAHers who present with a sentinel bleed. So there’s a real opportunity to help some potentially moribund patients — but if you ask most interns, they’ll say they’re shoving too many needles into the backs of people who probably just needed some exedrin.
Maybe we can change our practice, based on recent upgrades in CT scanner technology. These authors were the first to look at the new scanners with an eye toward sensitivity in SAH diagnosis. More below…
We were all curious about this ‘fifth generation’ — it turns out if you google “fifth generation CT” you get a wikipedia entry and … this very paper! The wikipedia article is lacking in citation but states:
Although numbered sequentially, the 3rd and 4th generation designs developed at approximately the same time. The concept of electron beam CT, which some authors have called 5th generation, followed later. Some authors have described up to 7 generations of CT design. However, it is only generations one to four that are widely, and consistently, recognised.
The authors used GE Lightspeed 2.X scanners, which don’t use electron beams, as far as I can tell by wading through old press releases (the LightSpeed is different from the GE eSpeed, which does use EBT). Perhaps it’s just easier to say they’re using a multidetector CT, instead of parotting this confusing ‘fifth-generation’ terminology.
The authors retrospectively looked at patient reconrds from 2002 at a rural Level-1 Trauma Center ED (a prospective study would’ve been a tough sell to the IRB). If anyone in their ED with a headache went on to have a CT and LP to evaluate for SAH, they were included in this study (except for those under age 17, and those with recent trauma or neurosurgery). The records needed to mention something about ruling out SAH to be included. Headache did not need to be the chief complaint, but needed to be a reported symptom.
They didn’t exclude anyone with anemia, which might’ve caused some false negatives (if blood is not brighter than brain, SAH is harder to see). And they didn’t record the patient’s report of symptom onset, which would’ve been interesting.
The CTs had 5-mm cuts through the cerebrum and posterior fossa, and were read by attending radiologists. Why just 5mm slices, when they could cut thinner? Well, as it turns out, 5 mm was sensitive enough, and maybe cutting thinner would’ve caused more false positives.
Their gold standard, LP, was positive if tube 1 had at least 400 RBCs, and the RBCs did not clear by tenfold in Tube 4. LPs could also be considered positive if xanthochromia was visually observed. SAH was further evaluated by CT angiography, or telephone followup.
Of the 569 LP’s they performed in 2002, 177 were listed as part of an evaluation for SAH. Of those 177, 159 had negative CTs and negative LPs. Eleven CTs were negative but had positive LP. Six CT scans in 2002 were positive for SAH — one LP was done in this group and this was also positive. And, notably, there was one positive CT that had a negative LP (it was read as questionable for intraventricular blood, and the patient had no further complications, so this was the one false positive they noted.)
Looking through their ED database, they saw 31 SAH in 2002 — 25 were traumatic (81%, but they say it’s 77%), and all traumatic and nontraumatic SAHs were seen on CT.
None of the 170 patients who had a negative CT scan for SAH was found to have one by LP or telephone followup. Of the 11 patients with negative CT but positive LP, none had further headache complications on phone followup, and none had xanthochromia, so these were called ‘LP false positives’.
From this data, they calculated a sensitivity of CT for SAH of 100%, which 95% confidence interval ranging from 61.0-100%. And the specificity of CT for SAH was 99.4% (95%CI was 96.8-99.9%).
The authors don’t calculate likelihood ratios, which we use to augment our pretest probability assignments for various diagnoses. If they did calculate LR, well, a positive CT scan would have a huge LR — even if you took the low end of the confidence interval, LR would be 19 — overwhelmingly likely to make your diagnosis.
However, if you use the low end of their confidence intervals, a negative CT scan only has a negative LR of 0.4, which is not good enough to budge your pretest probability into a diagnosis of SAH.
Really, this size of this study was small, and it was underpowered to provide meaningful values for the sensitivity and specificity. A larger trial, involving rural and nonrural centers, might produce a confidence interval narrow enough to give a meaningful likelihood ratio.
An article in Postgrad Medical Journal (2005;8:470-473) cites the familiar statistic, that CT is 98% sensitive in diagnosing SAH within 12 hours, 95% sensitive within 24 hours, and dips to 75% sensitivity if the SAH is between 24-72 hours old. (That data seems to come from a paper by Adams, in Neurology, 1983: Vol. 33, pp981-988). LP, on the other hand, doesn’t hit 100% sensitivity until after 12 hours – so the combo of CT/LP has been a powerful tool for emergency physicians. It will stay that way, for now — it’s far too soon to consider abandoning LP.
More practice resources, from the 2002 ACEP guidelines on the approach to headache:
Level B recommendations. Patients presenting to the ED with headache and abnormal findings in a neurologic examination (ie, focal deficit, altered mental status, altered cognitive function) should undergo emergent noncontrast head CT scan. Patients presenting with acute sudden-onset headache should be considered for an emergent head CT scan. HIV-positive patients with a new type of headache should be considered for an urgent neuroimaging study.
Level C recommendations. Patients who are older than 50 years presenting with new type of headache without abnormal findings in a neurologic examination should be considered for an urgent neuroimaging study.
Level C recommendations. Patients with a thunderclap headache who have negative findings in a head CT scan, normal opening pressure, and negative findings in CSF analysis do not need emergent angiography and can be discharged from the ED with follow-up arranged with their primary care provider or neurologist.
(additionally, the ACEP guidelines and extensive (but now outdated) lit review revealed response to therapy can’t be the sole basis of a headache diagnosis (Level C), and LPs can be performed without prior neuroimaging in patients lacking signs of increased intracranial pressure (Level C).
EMCrit.org cites a variety of sources in the EM and neurology literature to provide some useful, evidence-based recommendations in evaluating SAH and the results of your LP.
Posted
on Monday, May 7th, 2007 at 5:55 pm by Nick. Filed under
Stroke / TIA, Headache, Procedures, Journal Club, Radiology.
You may post a comment.
Nick raises some excellent points in the discussion of this article.
This trial could have easily been done prospectively; it is a diagnostic study not a treatment study. You would just need to enroll all of the pts with cts ordered for SAH and then look at the LP results. If the clinicians decided for whatever reason not to perform an LP then you would need an alternative critierion standard such as follow-up.
What many of these studies do not mention is that SAH is a disease whose tests exhibit a spectrum effect.
The patients we feel definitely need an LP after their CT are the ones in which the CT is likely to be positive in the first place–the ones with striking, severe presentations.
The patients with a subtle or crappy stories are the ones we often feel we can get away with skipping the LP after a negative CT. These patients are the ones the CT is likely to miss, because they may have only a scant amount of blood from a sentinel bleed.
Spectrum effect messes with Bayesian reasoning b/c our pretest probability is not independent of the test. This, along with the reasons Nick mentioned, is why an LP still must follow your CT if you suspect SAH. At least, until more compelling data than this small study emerges.
s
Comment by Scott on May 8th, 2007 at 5:25 am
One think I don’t really understand, is how they got the lower CI for sensitivity to be 61%.
Calculating CI when the incidence is low is tricky, and you can’t use the same biostatistic calculation as one does for CI for greater numbers. A review of this is discussed in Annals EM, 30(3), 1997, pp 301-306.
In short, 170 negative LPs after CT would indicate a lower limit sensitivity of 98.6%, (approx 3/171… see the article).
Of note, the original studies that determined that “CT and LP” is safe to r/o SAH is based on “only” 199 patients total spread out over 4 different studies; the largest study is 77 patients. This data is prospectively collected and outcome based, but raises the same questions regarding the small size of the study.
In the end, I’m not so concerned about the size of the study, but the fact that it wasn’t prospectively collected. As Scott pointed out, the syndrome can easily be a spectrum which can affect our results, so I would be particularly interested in those patients who got CTs and did not get an LP for some reason, in order to make sure they didn’t have any bad outcomes.
Until then, I generally offer an LP to every r/o SAH headache patient, and discuss the currently available data with them. Patient autonomy can best balance the small risk of a SAH with the prospect of an LP.
Just make sure you document the discussion. If the patient turns out to have a SAH, there’s a decent chance he/she won’t remember the conversation….
Comment by Dave P on May 11th, 2007 at 8:27 am