Journal Logo

Special Report

Special Report

LP's Low Diagnostic Yield Has Experts Questioning Its Usefulness for SAH

Shaw, Gina

doi: 10.1097/01.EEM.0000480772.37727.07

    Physicians must perform 204 lumbar punctures (LP) to detect one case of subarachnoid hemorrhage (SAH) in patients presenting to the emergency department with acute severe headache whose CT scans had been interpreted as normal. That's the finding from a new British study published in October's Academic Emergency Medicine, which the authors state is the largest retrospective study of its kind to date.

    David Sayer, MD, MRCS, of the department of neurosurgery at Queen's Hospital in Romford, Essex, and colleagues from the emergency medicine and pathology departments, reviewed cases of 2,248 adults who had presented to six urban EDs over a five-year period with acute nontraumatic headache, negative CTs, and LP performed for CSF analysis. (Acad Emerg Med 2015;22[11]:1267.)

    A total of 4.1 percent of all LPs undertaken were positive, and nine patients within that group had vascular abnormalities, equating to 0.47 percent of all LPs undertaken with analyzable CSF, and 204 LPs per identified vascular abnormality.

    Previous studies have also highlighted the low diagnostic yield of LP for SAH diagnosis following a negative CT scan, with findings ranging from one to 7.7 percent. (Emerg Med J 2014;31[9]:720; BMJ 2011;343:d4277), but this study suggests that the yield may be even lower than previously thought.

    Some of that may be because of a high number of uninterpretable results: Only 1,898 of the patients had interpretable LPs, Dr. Sayer said.

    “That can be the case because the most junior staff is delegated to do the LPs and their technique is not quite as good. Some of those samples could have been positive if, say, quality control was a bit better,” he said. Only 22 of the 300-plus uninterpretable studies were investigated further; of those, two patients were ultimately diagnosed with SAH, both of whom received additional imaging because of high clinical suspicion.

    What is the next step for this patient with a severe headache and this CT scan?

    Whatever the precise figure, there is little doubt that you have to do quite a lot of LPs to pick up one SAH, Dr. Sayer said.

    U.S. Rate

    “I think these numbers are pretty representative of the U.S. rate as well,” said Ryan Radecki, MD, an assistant professor of emergency medicine at the University of Texas Health Sciences Center in Houston. “We're worried about liability, and we are doing LPs on patients who are at lower and lower risk. I see a lot more patients who have false-positive findings or who have complications or adverse events from the LP than patients whose SAH is detected with an LP.”

    These worries are not unfounded. The consequences of a missed subarachnoid hemorrhage are dire: The mortality rate from SAH is approximately 45 percent, and 30 percent of survivors have severe disability. And the relative rarity of SAH, combined with its common and often subtle symptoms and signs, make diagnosis a challenge even for experienced emergency physicians. A 2007 study from Toronto's Institute for Clinical Evaluative Sciences found that 5.4 percent of SAH patients — or about one in 20 — are not correctly identified when they first present to the ED, most often because they were initially diagnosed with a migraine or other headache and classified as a low-acuity presentation. Even at teaching institutions, three percent of SAH cases were initially misdiagnosed. (Stroke 2007;38[4]:1216.)

    An older report noted that the combination of a negative CT and negative LP can absolutely rule out SAH. (Ann Emerg Med 2008;51[6]:707.) But how should emergency physicians, who may go years without seeing a true SAH case, determine which “suspicious headache” patients with negative CT findings are in need of an additional diagnostic procedure that is expensive, painful, and has its own risks, including rare but serious ones, such as meningitis or epidural hematoma, that can lead to paralysis?

    Timing of the CT is obviously one factor. “A negative CT scan within six hours of symptom onset has an excellent diagnostic yield,” said Dr. Radecki. Perry and colleagues from the University of Ottawa, writing in BMJ in 2011, found that modern third-generation CT performed perfectly in identifying SAH in patients who were within six hours of symptoms onset, with a sensitivity, specificity, and positive and negative predictive value of 100 percent. (BMJ 2011;343:d4277.) A Dutch study published in Neurology in May found that CTs done within that six-hour window still had a 99.4 percent negative predictive value even when performed by nonsubspecialized staff radiologist in a community hospital setting. (2015;84[19]:1927.)

    Even further out from symptom onset, CT still performed well; the Perry study found a negative predictive value of 99.4 percent and a positive predictive value of 100 percent for CT overall.

    “I see more and more people using the six-hour CT to support their decision not to do an LP because they are so demoralized by the lack of an effective way of working up subarachnoids,” says Dr. Radecki. “We're doing this invasive procedure that is really painful, has a small risk of infection, and sometimes makes the headache worse — and 90 percent of the time gives us inconclusive results. But it's the only way to rule out a subarachnoid 100 percent.

    “We're taught to chase this diagnosis, but when you get experience with it, you realize that it's a horrible yield and a difficult and time-consuming procedure,” he said. “Everything is conspiring for physicians to consent patients for LP in such a way that they are more likely to refuse the procedure than to accept.”

    Dr. Sayer said he tells patients he does not recommend an LP with a negative CT within 12 hours of symptom onset. “That may be potentially biasing the patient's decision, but I don't think our job as a physician is only to say, ‘Here's the evidence one way or another.’ I believe we should present evidence and make a recommendation. But I also present the counter-argument and give patients the choice,” he said.

    Can't-Miss Diagnosis

    Many emergency physicians think the counter-argument is a strong one, including Hannah Stewart, MD, an emergency physician at Torbay Hospital in Devon, UK, who in 2014 published a single-center retrospective study that put CT's overall sensitivity significantly lower than Perry's, at 93.8 percent. (Emerg Med J 2014;31[9]:720.) “LP is a nasty procedure, and our study showed that it often fails to be conclusive as well,” she acknowledged. “It has its own risks and cost implications; here, for example, all patients get 12-hour admission for LP.”

    But even with those drawbacks, Dr. Stewart said she believes in the importance of LP for ruling out SAH. “Even if you pick up one patient in 200 that wouldn't have been picked up by other diagnostic strategies, I think that's still worth it. You just don't want to miss these patients because the risk of harm is so high and the consequence of missing a subarachnoid in these patients is massive, as they're often quite young people. Your diagnostic strategy has to be as close to 100% as possible,” she said. The mortality rate from SAH is approximately 45 percent, and 30 percent of survivors will have severe disability.

    And other researchers differ with the Perry study's findings on six-hour CT's reliability. A two-pronged study of adults presenting to Kaiser Permanente emergency departments between 2000 and 2011 found that 11 of 55 patients who were ultimately diagnosed with SAH had negative CT findings within six hours of symptom onset, a CT miss rate of 20 percent. Not all of the missed SAHs were scanned with modern 64-slice scanners, but at least four were. (Ann Emerg Med 2013;62[1]:1.)


    A number of decision rules have also been proposed to help emergency physicians rule out SAH without the use of LP. One of the most recent is the Ottawa SAH rule, developed by Dr. Perry and colleagues, which would exclude SAH in the absence of all the following characteristics: age over 40, complaint of neck pain or stiffness, witnessed loss of consciousness, onset of headache with exertion, thunderclap headache, or limited neck flexion on examination.

    Investigators from Dr. Perry's group and a multicenter group of Canadian universities found that this rule was 100% sensitive and had 100% negative predictive value for SAH, although its specificity — 15.3% — was lower than several other decision rules studied. (JAMA 2013;310[12]:1248).

    An earlier iteration of the Ottawa SAH rule, which omits the thunderclap headache and neck flexion factors, was found to have nearly perfect sensitivity (98.5%), 99.6% negative predictive value, and specificity of 27.65. The shorter decision rule was also assessed in the Kaiser study, where it missed only 2.9 percent of patients with SAH who had a negative CT.

    “The Ottawa SAH Rule is a clinical decision rule that may help to standardize the investigation of acute headache among patients suspected of having subarachnoid hemorrhage without significantly changing current investigation rates,” Dr. Perry's group wrote. “Patients at ultra-high risk (i.e., those with [more than] 50% clinical pretest probability for subarachnoid hemorrhage) still warrant testing with CT, lumbar puncture, or both, given that they would have a posttest probability of greater than 1% for subarachnoid hemorrhage. This may include patients with strong family history or strong risk factors such as adult-onset polycystic kidney disease, although it is highly likely that these patients will have one or more of the high-risk features of the Ottawa SAH Rule.”

    Dr. Stewart suggested that more advanced generations of CT scanners or other imaging technology in the future may achieve true 100 percent sensitivity and specificity and ultimately render this argument moot.

    But in the meantime, emergency physicians are in a tough spot when it comes to mystery headaches and negative CTs, Dr. Radecki said. “This is a very rare condition, and we really need much more specific ways of predicting which patients should undergo any kind of evaluation,” he said.

    Share this article on Twitter and Facebook.

    Access the links in EMN by reading this on our website or in our free iPad app, both available at

    Comments? Write to us at

    Copyright © 2016 Wolters Kluwer Health, Inc. All rights reserved.