Can Computed Tomography Angiography of the Brain Replace

Home , Lumbar puncture

SPECIAL CONTRIBUTION Can Computed Tomography Angiography of the Brain Replace Lumbar Puncture in the Evaluation of Acute-onset Headache After a Negative Noncontrast Cranial Computed Tomography Scan? Robert F. McCormack, MD, and Alan Hutson, PhD

Abstract Objectives: The primary goal of evaluation for acute-onset headache is to exclude aneurysmal subarachnoid hemorrhage (SAH). Noncontrast cranial computed tomography (CT), followed by lumbar puncture (LP) if the CT is negative, is the current standard of care. Computed tomography angiography (CTA) of the brain has become more available and more sensitive for the detection of cerebral aneurysms. This study addresses the role of CT ⁄ CTA versus CT ⁄ LP in the diagnostic workup of acute-onset headache. Methods: This article reviews the recent literature for the prevalence of SAH in emergency department (ED) headache patients, the sensitivity of CT for diagnosing acute SAH, and the sensitivity and specificity of CTA for cerebral aneurysms. An equivalence study comparing CT ⁄ LP and CT ⁄ CTA would require 3,000 + subjects. As an alternative, the authors constructed a mathematical probability model to determine the posttest probability of excluding aneurysmal or arterial venous malformation (AVM) SAH with aCT⁄ CTA strategy. Results: SAH prevalence in ED headache patients was conservatively estimated at 15%. Representative studies reported CT sensitivity for SAH to be 91% (95% confidence interval [CI] = 82% to 97%) and sensitivity of CTA for aneurysm to be 97.9% (95% CI = 88.9% to 99.9%). Based on these data, the posttest probability of excluding aneurysmal SAH after a negative CT ⁄ CTA was 99.43% (95% CI = 98.86% to 99.81%). Conclusions: CT followed by CTA can exclude SAH with a greater than 99% posttest probability. In ED patients complaining of acute-onset headache without significant SAH risk factors, CT ⁄ CTA may offer a less invasive and more specific diagnostic paradigm. If one chooses to offer LP after CT ⁄ CTA, informed consent for LP should put the pretest risk of a missed aneurysmal SAH at less than 1%. ACADEMIC EMERGENCY MEDICINE 2010; 17:444–451 ª 2010 by the Society for Academic Emergency Medicine Keywords: subarachnoid hemorrhage, acute onset headache, computed tomography, computed tomography angiography, lumbar puncture, diagnosis

eadache is a frequent complaint in the emer- a SAH is a noncontrast cranial computed tomography gency department (ED). When a patient com- (CT), followed by lumbar puncture (LP) if the initial CT plains of an acute-onset headache, the scan is negative. This diagnostic strategy has recently H 1 possibility of a subarachnoid hemorrhage (SAH) must be been validated in a study by Perry et al. CT scan fol- entertained. The standard diagnostic strategy to exclude lowed by LP shows a sensitivity of 100% and negative predictive value of 100% in excluding SAH. However, LP is invasive and time-consuming. It can be technically dif- ﬁcult due to body habitus, poor patient cooperation, and From the Department of Emergency Medicine (RFM) and the patient anxiety. Postdural puncture headache occurs in Department of Biostatistics (AH), University at Buffalo School up to 40% patients2 and can add to suffering. There is of Medicine and Biomedical Sciences, Buffalo, NY. signiﬁcant controversy regarding what constitutes a Received June 29, 2009; revisions received August 25 and Octo- positive tap versus a traumatic tap.3 An ‘‘equivocal’’ LP ber 2, 2009; accepted October 3, 2009. occurs 15% to 20% of the time4 and often mandates Address for correspondence and reprints: Robert F. McCor- further workup. Recent advances and availability of mack, MD; e-mail: [email protected].

ISSN 1069-6563 ª 2010 by the Society for Academic Emergency Medicine 444 PII ISSN 1069-6563583 doi: 10.1111/j.1553-2712.2010.00694.x ACAD EMERG MED • April 2010, Vol. 17, No. 4 • www.aemj.org 445 computed tomography angiography (CTA) raise the pos- METHODS sibility of an alternative diagnostic strategy. Can CTA replace LP after a negative CT to exclude SAH due to After a MEDLINE search (key words: subarachnoid aneurysm or arterial venous malformation (AVM)? hemorrhage, tomography x-ray computed, spinal punc- It is difficult to prove that these are equivalent diag- ture, angiography, angiography digital subtraction; nostic strategies because of the strength of the initial from 1996 to present) and review of article references, diagnostic test (CT) and a near zero margin for error we selected the most recent and relevant studies with in diagnosis. It is well established in the literature that regard to SAH prevalence in ED patients complaining patients who suffer a missed diagnosis of SAH do of acute-onset headache, the sensitivity of CT for the poorly.5–7 A missed diagnosis is more likely to occur diagnosis of SAH, and the sensitivity and specificity of with subtle presentations and often results in death or CTA for cerebral aneurysm. permanent disability. In those patients where the ‘‘sentinel’’ hemorrhage is detected, the outcome is SAH Prevalence much more favorable. Because of this opportunity to There is no published meta-analysis of SAH prevalence greatly influence overall outcomes, it is clinically in ED patients with an acute-onset headache and a nor- imperative to find all patients with SAH due to an mal neurologic exam. Based on the current literature, aneurysm. the prevalence is between 3% and 16% (Table 1).1,19–22 Approximately 85% of SAH are due to aneurysm or We have conservatively estimated the prevalence of AVM.8–11 SAHs that are not due to cerebral aneurysm SAH in an ED patient complaining of an acute-onset or AVM tend to have better prognoses.12 Approximately headache with an intact neurologic exam as 15%. 10% will display a perimesencephalic pattern of bleed- Although the actual prevalence may be lower, assuming ing and carry a good prognosis.13 The small remainder a higher prevalence in a probability model will maxi- will be due to vasculitis, tumors, coagulation disorders, mize the number of potential missed cases and reveal and a variety of other causes and typically carry a bet- weaknesses in the model. ter prognosis than those caused by aneurysms.9,14 Patients with acute-onset headache and abnormal Aneurysms without SAH are often undetected in the mental status or neurologic deficits are presumed to current standard of care. A ‘‘symptomatic aneurysm’’ have a higher prevalence of SAH. Currently, there are without actual hemorrhage has been postulated as a no adequate prevalence data for SAH in this subgroup cause of acute-onset headache. However, on long-term of patients and therefore they should not be included in follow-up with the current diagnostic strategy, those this model. patients who do not have SAH but may have an aneurysm even in the presence of acute headache do CT Sensitivity well.1,15–19 A noncontrast cranial CT has a sensitivity for SAH The cases a physician cannot afford to miss are those between 82% and 100%, depending on the timing of the with both SAH and an aneurysm or AVM present. study after the onset of the headache (Table 2).20,23–27 Missing either diagnosis alone when the other diagno- The 95% confidence intervals (CIs) substantially lower sis is not present is not optimal, but should be relatively these sensitivities. Most clinicians estimate the sensitiv- acceptable due to a better prognosis. CT is sensitive for ity to be above 90% in the first 24 to 48 hours after SAH, and CTA is very sensitive for aneurysm, but both headache onset. Sensitivity continues to decrease with remain less than perfect. However, the performance of time due to diffusion of blood away from the site of the two tests is additive when the goal is to exclude the hemorrhage. The studies on sensitivity of CT scan are concurrent presence of both diseases. We constructed limited by selection bias, retrospective design, and the a mathematical model to assess the posttest probability challenge of keeping pace with advancements in tech- of both SAH and aneurysm being present when both nology. There is no published meta-analysis. Compari- CT and CTA are negative. We hypothesize that in select son of the results shows performance to be similar patients, CT ⁄ CTA can replace CT ⁄ LP as a diagnostic between studies. We have selected the publication by strategy for aneurysmal SAH. Byyny et al.26 as a representative study for the purpose

Table 1 Selected Studies Reporting Prevalence of SAH in ED Headache Patients

Study Data Prevalence Comments Bo et al.22 433 patients ⁄ 71 SAH 16% Norway, majority referred by GPs, possible selection bias Perry et al.1 592 patients ⁄ 61 SAH 10% CT ⁄ LP veriﬁcation study Carstairs et al.21 116 patients ⁄ 5 SAH or aneurysm 4.3% CT ⁄ CTA for aneurysm or SAH 116 patients ⁄ 3 SAH 2.5% Boesiger and Shiber20 177 patients ⁄ 6 SAH 3.4% Retrospective, found pt by LP ⁄ CT ⁄ ﬁnal dx Landtblom et al.19 137 patients ⁄ 23 SAH 11.3% Sweden, University ED

CT = computed tomography; CTA = computed tomography angiography; dx = diagnosis; GP = general practitioner; LP = lumbar puncture; SAH = subarachnoid hemorrhage. 446 McCormack and Hutson • CAN CT ⁄ CTA REPLACE CT ⁄ LP IN THE EVALUATION OF ACUTE-ONSET HEADACHE?

Table 2 Selected Studies Reporting Sensitivity of CT for SAH

Study Data Sensitivity, % (95% CI) Comments Lourenco et al.27 61 SAH: 97 (84–100) Did not factor time to CT, 60 CT, 1 LP retrospective, allowed abnormal mental status and exam Byyny et al.26 149 SAH: 93 (88–97) Did not factor time to CT, HA, normal exam HA, normal neuro exam retrospective, subgroup of HA 139 CT, 10 LP 91 (82–97) with normal neuro exam Boesiger et al.20 177pts ⁄ 6 SAH: 100 (64–100) Did not factor time to CT, wide CI, 6 CT, 0 LP retrospective Sidman et al.24 140 patients Overall 92 Retrospective, 100% prevalence, 80 ⁄ 80 CT <12 hours 100 allowed abnormal mental 49 ⁄ 60 CT >12 hours 82 status ⁄ exam Sames et al.25 114 < 24 hours HA 93 Retrospective, 100% prevalence, 37 > 24 hours HA 84 allowed abnormal mental status ⁄ exam Van der Waal et al.23 175patients 119 SAH: 98 CT <12 hours, high prevalence 117 CT, 2 LP (69%) of disease

CT = computed tomography; HA = headache; LP = lumbar puncture; SAH = subarachnoid hemorrhage.

of our model because it best represents ED headache numbers of patients would be needed to prove that patients with acute-onset headache and normal neuro- CT ⁄ CTA is equivalent to CT ⁄ LP. For example, assume logic exam. The study by Lourenco et al.27 study was that the CT ⁄ LP strategy has a sensitivity of 99.9%. If a excluded because it allowed altered mental status and study were to compare the sensitivity rates in a ran- neurologic abnormalities, which could skew test sensi- domized two-arm equivalence trial of CT ⁄ LP versus tivity. The study by Boesiger and Shiber20 was excluded CT ⁄ CTA, and assuming equivalence is defined as a one- because of the wide CIs due to study size and design. sided 0.5% window within the target sensitivity of The studies by Sidman et al.,24 Sames et al.,25 and van 99.9% (i.e., within the range of 99.4%–99.9%), then der Wee et al.23 were felt to use outdated CT technol- roughly 3,000 subjects would need to be accrued to the ogy and are limited by patient selection bias. The speci- respective trial. Due to the low prevalence of disease, it ficity of CT for SAH is not reported in most studies, but would take significant time and resources to conduct a is assumed to be 100%. study with sufficient power. The only study on SAH in ED patients to enroll 3,000 patients took 8 years and 12 LP Sensitivity centers to accrue.33 Note that a more refined sample LP shows excellent sensitivity. The recent study by size calculation would depend on the respective investi- Perry et al.1 demonstrates 100% sensitivity. However, gator’s specific set of input parameters. the specificity is low, at 67%. RESULTS CTA Sensitivity Numerous studies have examined the performance of Patients who present to an ED with a chief complaint CTA for aneurysms in the setting of SAH. CTA has of acute-onset headache and have an intact neurologic been shown to be a reliable substitute for the tradi- exam have a conservatively estimated pretest probabil- tional criterion standard of digital subtraction angiogra- ity of SAH of 15%. Patients with loss of consciousness phy (DSA) for aneurysm diagnosis in a number of associated with acute headache, those with menigismus recent studies. Sixty-four-detector CTA of the brain has or cranial nerve findings, or those with significant risk been shown to be 98% sensitive and 100% specific for factors (first-degree relative with aneurysm, connective the detection of aneurysms in the setting of SAH.28 tissue disease, polycystic kidney disease) may have a Multiple other studies of 16-, four-, and single- detector higher pretest probability that has not been adequately CTA indicate similar sensitivities and specificities (see defined and should be excluded from this model. Table 3).28–32 All of these studies include patients with Review of the literature on CT for SAH shows fairly variable levels of mental status and neurologic exam consistent results over the past 15 years. The study by findings who have a diagnosis of SAH by CT or LP. We Byyny et al.26 showed sensitivity of 91% with a 95% selected the study by Agid et al.28 because it utilized 64- CI = 82% to 97% and a specificity of 100% in neurologi- detector technology. There is no formal published cally intact patients with an acute-onset headache. CT meta-analysis, and the most recent review in the litera- sensitivity is well known to diminish with prolonged ture is now technologically obsolete. time from ictus. For the purpose of this model, time It would be very difficult to prove a new diagnostic from ictus to CT should be less than 48 hours. strategy that could match the sensitivity of the current The sensitivity and specificity of CTA for diagnosis of CT ⁄ LP strategy. Because of the low prevalence of dis- cerebral aneurysm in the setting of SAH have been ease and excellent sensitivity of the initial study, large remarkably consistent with multidetector CT. Agid et al.28 ACAD EMERG MED • April 2010, Vol. 17, No. 4 • www.aemj.org 447

Table 3 Selected Studies Reporting Sensitivity of Multidetector CTA for Cerebral Aneurysm

Study Data Sensitivity, % Speciﬁcity, % Comments Kokkinis et al.32 198 patients SAH: 97.9 100 Single detector, 179 w ⁄ AN, 15 AVM found all AVMs El Khaldi et al.31 104 patients, 98.8 100 16-detector CT 84 w ⁄ AN, 0 AVM Yoon et al.30 85 patients SAH 71 w ⁄ AN, 100 for leaking AN 100 16-detector, 0 AVM 92.5 all AN 93.3 caught 100% of culprit AN Agid et al.28 65 patients w ⁄ SAH, 97.9 for AN 100 64-detector CT 47 w ⁄ AN, 0 AVM (95% CI = 88.9%–99.9%)

AN = cerebral aneurysm; AVM = arterial venous malformation; CT = computed tomography; CTA = computed tomography angiography; LP = lumbar puncture; SAH = subarachnoid hemorrhage.

demonstrated CTA sensitivity for aneurysm was 97.9% For notational purposes we have the following: with a 95% CI = 88.9% to 99.9%, and specificity was P(+,+)= the joint probability of a positive criterion 100%. This study group is different than our target group standard test for CT and CTA, in that all of these patients had SAH and varying clinical P1(+|+) = sensitivity of CT (probability of a positive presentations. We are looking for aneurysms in patients CT, given a true SAH), with or without SAH. However, we will never be able to P1()|))= specificity of CT (probability of a negative obtain true sensitivity and specificity of CTA in this CT, given no true SAH), patient setting because it would be impractical to conduct P2(+|+) = sensitivity of CTA (probability of a positive CTA and DSA testing on ED headache patients in the CTA, given a true aneurysm), absence of SAH. Eighty-five percent of SAHs are caused P2()|)) = specificity of CTA (probability of a negative by aneurysms or AVM. We assume a baseline prevalence CTA, given no true aneurysm). of aneurysm without SAH as 2.5% based on the literature Note also that P()|+) = 1 – P(+|+) and that P(+|))=1– (Table 4).34–36 P()|)). Computed tomography looks for blood in the sub- Under the independence assumption between the CT arachnoid space. CTA looks for cerebral aneurysms. and CTA tests, the expected proportion of subjects who Assuming that these two tests are independent of one are diagnosed as CT+ and CTA+ is given by the product another, the expected proportion of cases falling into a of three probabilities P(+,+)P1(+|+)P2(+|+). Similarly, the particular diagnostic category can be calculated as a expected proportion of subjects who are diagnosed as function of the joint probability of each test being either CT+ and CTA+ given CT was truly ‘‘+’’ and CTA was positive (+) or negative ()) times the respective sensitiv- truly ‘‘)’’ is given by the product of three probabilities ity (+|+) or 1 – sensitivity ()|+) and specificity ()|))or P(+,+)P1(+|+)P2(+|)) and so on as described above. 1 – specificity (+|)) depending on the + ⁄ ) test combina- Using the clinical assumptions made (see Table 4) and tions. results from the selected studies, Table 5 illustrates the various test result probabilities. This model makes a probabilistic argument for CT ⁄ CTA in patients with a pretest probability of SAH Table 4 of 15%. Using the clinical assumptions above and given Probability Model Clinical Assumptions that the two diagnostic tests are independent, the CI of the posttest probability of both an aneurysm and an • 15% of ED patients with acute-onset headache and normal SAH being truly present in the face of a two negative neurologic exam will have a SAH. results ranges from 0.0% to 0.25%. In other words, • 85% of SAH will be due to aneurysm or AVM; 15% are due there is a posttest probability of 99.98% (95% CI = to other causes of which approximately 10% are 99.75% to 100%) that both diseases are not present perimesencephalic hemorrhage. assuming study independence between CT and CTA. • Therefore, 12.75% of ED acute-onset headache patients will have SAH due to aneurysm or AVM, and 2.25% of ED acute-onset headache patients will have SAH without Test Dependence aneurysm or AVM. This probability model has assumed that CT and CTA • 2.5% prevalence of aneurysm in patients without risk are independent tests looking for different diseases factors for aneurysm and without SAH. • CT: 91% sensitivity with a 95% CI = 82% to 97% and a (SAH vs. aneurysm). In support of the tests being inde- specificity of 100%. pendent are that the factors that decrease the sensitivity • CTA: 97.9% sensitivity with a 95% CI = 88.9% to 99.9% and of CT include time from headache onset, hematocrit a specificity of 100%. concentration, and volume of hemorrhage. These factors do not have an adverse effect on the sensitivity of AVM = arterial venous malformation; CT = computed tomog- CTA. Factors that would decrease the sensitivity of CTA raphy; CTA = computed tomography angiography; SAH = subarachnoid hemorrhage. such as aneurysm size, motion artifact, and intravenous contrast injection technique would not adversely affect 448 McCormack and Hutson • CAN CT ⁄ CTA REPLACE CT ⁄ LP IN THE EVALUATION OF ACUTE-ONSET HEADACHE?

Table 5 Posttest Probability: Study Results

Test Result ‘‘True’’ Aneurysm + CT ) CT + CT ) CT ‘‘True’’ SAH + SAH + SAH ) SAH ) SAH + CTA + aneurysm 0.1137 0.0023 <0.0001 0.0227 ) CTA + aneurysm 0.0112 0.0002 <0.0001 0.0022 + CTA ) aneurysm <0.0001 <0.0001 <0.0001 0.0008 ) CTA ) aneurysm 0.0220 0.0004 0.0008 0.8234

CT sensitivity 0.91 and CTA sensitivity 0.98. CT = computed tomography; CTA = computed tomography angiography; SAH = subarachnoid hemorrhage. the sensitivity of CT for SAH. However, the indepen- posttest probability would be 99.43% (95% CI = 98.86% dence assumption may not be realistic. These tests are to 99.81%) that both diseases are not present. done in immediate sequence using the same technology. The tests are interpreted at the same time by the same DISCUSSION examiner, which could threaten their independence. The most concerning dependent relation is that the If it is acceptable that a small number of SAHs that are same examiner interprets both studies. This may have not caused by an aneurysm or AVM are missed upon an additive negative effect if the examiner is not profi- initial evaluation, then the CT ⁄ CTA strategy may be an cient or is distracted. However, the dependence of the alternative to the current paradigm in select patients. two studies may have a positive effect if a subtle finding SAH due to perimesencephalic bleeding or in the set- on one scan causes increased scrutiny when reading ting of normal cerebral arteries does not carry the same the other scan. risk for poor outcome. In patients who are neurologi- Although the dependent effect of the two studies is cally intact and whose headaches are controlled, dis- difficult to determine, the following mathematical con- charge should have no ill effect. struct attempts to assess the effect. Using basic proba- If the patient has an aneurysm that has not leaked bility relationships, we can drill down into the (i.e., SAH is not present), the current CT ⁄ LP strategy is calculation of the proportion of missed cases assuming only 98% sensitive.1 There is still a potential to miss some structure with respect to the dependence between these patients with this CT ⁄ CTA strategy. However, CT and CTA in the specific case where both tests are CTA has been shown to be very sensitive detecting negative when in fact there was an SAH or an aneu- aneurysms, especially those greater than 3 mm. As the rysm. This may be carried forth by utilizing the specific likelihood of rupture is directly related to aneurysm calculation of the expected proportion of missed cases, size,37 missed small aneurysms have a very low risk to which is now the product of three probabilities subsequently bleed. The excellent outcome on long- P(+,+)P1()|+)P2|1(()|+)|()|+)), where now P2|1(()|+)|()|+)) term follow-up of patients who underwent CT ⁄ LP with corresponds to the conditional probability of the event negative results shows this to be a safe discharge. that CTA was a false negative given that the CT was a A definitive study would involve a CT ⁄ CTA ⁄ LP strat- false negative. Note that the sensitivity of CTA is egy. All patients who had a negative CT ⁄ CTA would go ‘‘embedded’’ within P2|1(()|+)|()|+)). To examine the on to have LP. This could prove that the CT ⁄ CTA strat- dependence structure, we varied the values of egy is sufficient and that LP offers no additional infor- P2|1(()|+)|()|+)) from 0. 25 to 0.75, where the level of mation. There are some real-world problems with 0.25 implies a low degree of correlation and the level conducting this study. Most importantly, the number of 0.75 implies a high degree of correlation between the subjects would need to be very large (>3,000 patients). tests with respect to jointly missing a true SAH and a Second, when LP results are inconclusive, CTA is often true aneurysm. Although the true dependence is done as a more definitive study. This research strategy unknown, we have conservatively assumed a 0.5 level would reverse that logic. Finally, obtaining informed of correlation for the purpose of this model based on consent is one of the biggest obstacles in obtaining an the previously mentioned factors. This implies that 50% LP in these patients. Currently, after a negative CT in a of the time that one study is misread, it will negatively patient with a headache of less than 48 hours, the phy- affect the other study’s sensitivity. sician can inform the patient that there is approximately Note that the sensitivity for CT scans of 91% (0.91) a 5%–10% chance of an undetected SAH. When both based on the literature corresponds to P1()|+) = 0.09 in the CT and the CTA are negative, informed consent Table 6. Under the independence scenario of the CT should estimate the probability of a missed SAH result- and CTA tests in terms of jointly arriving at false nega- ing from an aneurysm or AVM as less than 1%. This tives, we saw from Table 5 that expected proportion of could significantly reduce the number of patients who missed SAH and an aneurysm was calculated to be are willing to undergo an LP after negative tests. 0.0002, or 0.02%. From Table 6 we see that between the In a patient with a pretest probability of less than two tests, the expected proportion of missed SAH and 15%, a CT ⁄ CTA may be a less invasive and more spe- aneurysm with a 25, 50, and 75% study dependence cific diagnostic strategy. As pretest probability would be 0.0029, 0.0057, and 0.0086, respectively. In decreases, the likelihood of two false-negative tests also other words, assuming a 50% study dependence, the decreases. In those cases where LP is technically impos- ACAD EMERG MED • April 2010, Vol. 17, No. 4 • www.aemj.org 449

Table 6 The Posttest Probability of False-negative Cases of SAH and Aneurysm by Both CT and CTA Given That a True SAH and Aneurysm Exists with 25, 50, and 75% Study Dependence

CT Sensitivity P1(+|+), 1-Sensitivity CT Dependence Posttest Probability CT()|+),CTA()|+) % (95% CI) P1()|+) P2|1((+|+)|()|+)) (95% CI) 91 (82–97) 0.09 0.25 0.0029 (0.0057–0.0009) 91 (82–97) 0.09 0.50 0.0057 (0.114–0.0019) 91 (82–97) 0.09 0.75 0.0086 (0.172–0.0029)

CT = computed tomography scan; CTA = computed tomography angiogram; SAH = subarachnoid hemorrhage.

sible, or for patients who refuse LP, CT ⁄ CTA may offer to current practice that exists at some institutions but a reasonable alternative diagnostic strategy. has no true supporting literature. Shortfalls of the In those cases with a higher pretest probability due model include the unknown dependence of the two to risk factors, abnormal exam or mental status, or clas- tests and the true sensitivity of CTA in the absence of sic presentation, a new diagnostic strategy of SAH in neurologically intact headache patients. How- CP ⁄ CTA ⁄ LP may be preferred. The true pretest proba- ever, unlike CT where clinical condition is directly bility in these patients is unknown, but estimating it to linked to volume of hemorrhage and a worse clinical be 25% would increase the posttest probability of two condition may positively skew CT performance, the false-negative tests to 0.96% (95% CI = 0.3% to 1.9%), same should not be true of CTA. which may be unacceptably high in a high-risk patient. The level of dependence between the CT and CTA tests In those patients with a CT that is negative for SAH, is a significant assumption. A test of independence (vs. and the CTA shows an aneurysm, an LP will need to be dependence) between the CT and CTA tests could be car- undertaken to distinguish bleeding versus nonbleeding ried forth with a small study of roughly 30 to 130 subjects aneurysm. A nonbleeding aneurysm may be followed depending upon the specific alternatives via a straight- conservatively after specialty consultation. The true forward test involving Kendall’s tau-b test statistic. prevalence of incidental aneurysms in acute headache Clinical care shortfalls include the exposure to intra- patients who seek care in the ED is unknown. Inciden- venous contrast and additional radiation, the lack of tal aneurysm prevalence literature is based on autopsy diagnostic information regarding other conditions and angiography studies and varies between 1 and obtained from LP (benign intracranial hypertension, 6%.35,38 More recent angiology studies show a signifi- meningitis), and an expected increased frequency of cantly lower percentage (1%–2%) of incidental aneu- detection of nonbleeding ‘‘incidental’’ aneurysms that rysms.34,36 Rinkel et al.36 concluded in 1998 that the will require further follow-up. prevalence of incidental aneurysm in adults without risk This is a model and not based on actual study data. A factors for SAH is approximately 2%. The only study on large multicenter study examining acute headache ED patients with headache who underwent CTA patients who present to the ED with normal neurologic showed an aneurysm prevalence of 5% (5 ⁄ 106), with exams who receive CT ⁄ CTA and LP should be con- 3% (3 ⁄ 106) being culprit lesions, one being potentially ducted. These patients should be followed for an symptomatic (1 ⁄ 106), and one being most likely inciden- extended period of time after discharge as a surrogate to tal (1 ⁄ 106).21 The clinical utility of the CT ⁄ CTA approach detect missed aneurysms because adding DSA as the cri- may be determined by the true prevalence of these terion standard to CTA would be impractical. Assuming nonbleeding aneurysms and the implications for patient that the current prevalence, sensitivities, and specificities management and follow-up. Discovering ‘‘incidental’’ hold true, an equivalence study would require at least aneurysms may subject the patient to additional risk 3,000 subjects to be enrolled. The feasibility of such a from subsequent testing and interventions. study is limited. A smaller study could test the depen- In the current clinical setting of readily available CT dence assumption in a similar patient population and and CTA at most emergency care facilities, this diag- offer support to this model and practice. nostic strategy may offer an more consistent and less invasive diagnostic work-up of an acute headache. CONCLUSIONS Patients are often reticent to undergo and physicians are hesitant to conduct an LP. Equivocal LP results In ED patients with a complaint of acute-onset head- often cloud the diagnostic picture further. These issues ache of less than 48 hours’ duration who have an intact may contribute to suboptimal work-ups and may neurologic exam and no significant increased risk of account for the persistence of an basal rate of ‘‘missed aneurysm, it is reasonable to conclude that computed bleeds’’ that was recently demonstrated to be 5% of tomography, followed by computed tomography angi- acute headache patients in Canadian EDs.39 ography, has excluded subarachnoid hemorrhage due to aneurysm or arterial venous malformation with a LIMITATIONS 99% posttest probability. A clinician may choose to spare the lumbar puncture in these cases. At a mini- This paradigm has some shortfalls, but it attempts to mum, informed consent for lumbar puncture should frame the clinical and research questions. It also speaks put the theoretical risk of a missed diagnosis of sub- 450 McCormack and Hutson • CAN CT ⁄ CTA REPLACE CT ⁄ LP IN THE EVALUATION OF ACUTE-ONSET HEADACHE? arachnoid hemorrhage due to aneurysm or arterial 15. Wijdicks EF, Kerkhoff H, van Gijn J. Long-term fol- venous malformation after negative computed tomogra- low-up of 71 patients with thunderclap headache phy ⁄ computed tomography angiography at less than mimicking subarachnoid haemorrhage. Lancet. 1%, assuming a 50% test sensitivity dependence. 1988; 2:68–70. Patients with classic presentations, meningismus, 16. Markus HS. A prospective follow up of thunderclap abnormal neurologic exam, or significant risk factors headache mimicking subarachnoid haemorrhage. J may warrant computed tomography ⁄ lumbar puncture Neurol Neurosurg Psychiatry. 1991; 54:1117–8. with or without computed tomography angiography. 17. Harling DW, Peatfield RC, Van Hille PT, Abbott RJ. This model attempts to answer the question of: Can Thunderclap headache: is it migraine? Cephalalgia. computed tomography ⁄ computed tomography angiog- 1989; 9:87–90. raphy replace computed tomography ⁄ lumbar puncture 18. Linn FH, Wijdicks EF, van der Graaf Y, Weerdes- in selected patients? The answer is yes. The next ques- teyn-van Vliet FA, Bartelds AI, van Gijn J. Prospec- tion, which will depend on the prevalence of incidental tive study of sentinel headache in aneurysmal aneurysm, may be: should it? subarachnoid haemorrhage. Lancet. 1994; 344:590–3. 19. Landtblom AM, Fridriksson S, Boivie J, Hillman J, References Johansson G, Johansson I. Sudden onset headache: a prospective study of features, incidence and 1. Perry JJ, Spacek A, Forbes M, et al. Is the combina- causes. Cephalalgia. 2002; 22:354–60. tion of negative computed tomography result and 20. Boesiger BM, Shiber JR. Subarachnoid hemorrhage negative lumbar puncture result sufficient to rule diagnosis by computed tomography and lumbar out subarachnoid hemorrhage? Ann Emerg Med. puncture: are fifth generation CT scanners better at 2008; 51:707–13. identifying subarachnoid hemorrhage? J Emerg 2. Evans RW. Complications of lumbar puncture. Neu- Med. 2005; 29:23–7. rol Clin. 1998; 16:83–105. 21. Carstairs SD, Tanen DA, Duncan TD, et al. Com- 3. Shah KH, Edlow JA. Distinguishing traumatic lum- puted tomographic angiography for the evaluation bar puncture from true subarachnoid hemorrhage. of aneurysmal subarachnoid hemorrhage. Acad J Emerg Med. 2002; 23:67–74. Emerg Med. 2006; 13:486–92. 4. Shah KH, Richard KM, Nicholas S, Edlow JA. Inci- 22. Bo SH, Davidsen EM, Gulbrandsen P, Dietrichs E. dence of traumatic lumbar puncture. Acad Emerg Acute headache: a prospective diagnostic work-up Med. 2003; 10:151–4. of patients admitted to a general hospital. Eur J 5. Jakobsson KE, Saveland H, Hillman J, et al. Warn- Neurol. 2008; 15:1293–9. ing leak and management outcome in aneurysmal 23. van der Wee N, Rinkel GJ, Hasan D, van Gijn J. subarachnoid hemorrhage. J Neurosurg. 1996; Detection of subarachnoid haemorrhage on early 85:995–9. CT: is lumbar puncture still needed after a negative 6. Mayer PL, Awad IA, Todor R, et al. Misdiagnosis of scan? J Neurol Neurosurg Psychiatry. 1995; 58:357–9. symptomatic cerebral aneurysm. Prevalence and 24. Sidman R, Connolly E, Lemke T. Subarachnoid correlation with outcome at four institutions. hemorrhage diagnosis: lumbar puncture is still Stroke. 1996; 27:1558–63. needed when the computed tomography scan is 7. Kowalski RG, Claassen J, Kreiter KT, et al. Initial normal. Acad Emerg Med. 1996; 3:827–31. misdiagnosis and outcome after subarachnoid hem- 25. Sames TA, Storrow AB, Finkelstein JA, Magoon orrhage. JAMA. 2004; 291:866–9. MR. Sensitivity of new-generation computed 8. Vermeulen M, van Gijn J. The diagnosis of sub- tomography in subarachnoid hemorrhage. Acad arachnoid haemorrhage. J Neurol Neurosurg Psy- Emerg Med. 1996; 3:16–20. chiatry. 1990; 53:365–72. 26. Byyny RL, Mower WR, Shum N, Gabayan GZ, Fang 9. Rinkel GJ, van Gijn J, Wijdicks EF. Subarachnoid S, Baraff LJ. Sensitivity of noncontrast cranial com- hemorrhage without detectable aneurysm. A review puted tomography for the emergency department of the causes. Stroke. 1993; 24:1403–9. diagnosis of subarachnoid hemorrhage. Ann Emerg 10. van Gijn J, Rinkel GJ. Subarachnoid haemorrhage: Med. 2008; 51:697–703. diagnosis, causes and management. Brain. 2001; 27. Lourenco AP, Mayo-Smith WW, Tubbs RJ, Sidman 124(Pt 2):249–78. R. Does 16-detector computed tomography improve 11. van Gijn J, Kerr RS, Rinkel GJ. Subarachnoid haem- detection of non-traumatic subarachnoid hemor- orrhage. Lancet. 2007; 369:306–18. rhage in the emergency department? J Emerg Med. 12. Suarez JI, Tarr RW, Selman WR. Aneurysmal sub- 2009; 36:171–5. arachnoid hemorrhage. N Engl J Med. 2006; 28. Agid R, Lee SK, Willinsky RA, Farb RI, terBrugge 354:387–96. KG. Acute subarachnoid hemorrhage: using 64-slice 13. Rinkel GJ, Wijdicks EF, Vermeulen M, Hageman multidetector CT angiography to ‘‘triage’’ patients’ LM, Tans JT, van Gijn J. Outcome in perimesence- treatment. Neuroradiology. 2006; 48:787–94. phalic (nonaneurysmal) subarachnoid hemorrhage: 29. Goddard AJ, Tan G, Becker J. Computed tomogra- a follow-up study in 37 patients. Neurology. 1990; phy angiography for the detection and characteriza- 40:1130–2. tion of intra-cranial aneurysms: current status. Clin 14. Carvi y Nievas MN, Archavlis E. Atypical causes of Radiol. 2005; 60:1221–36. nontraumatic intracranial subarachnoid hemor- 30. Yoon DY, Lim KJ, Choi CS, Cho BM, Oh SM, Chang rhage. Clin Neurol Neurosurg. 2009; 111:354–8. SK. Detection and characterization of intracranial ACAD EMERG MED • April 2010, Vol. 17, No. 4 • www.aemj.org 451

aneurysms with 16-channel multidetector row CT 34. Atkinson JL, Sundt TM Jr, Houser OW, Whisnant JP. angiography: a prospective comparison of volume- Angiographic frequency of anterior circulation intra- rendered images and digital subtraction angiogra- cranial aneurysms. J Neurosurg. 1989; 70:551–5. phy. Am J Neuroradiol. 2007; 28:60–7. 35. Inagawa T, Hirano A. Autopsy study of unruptured 31. El Khaldi M, Pernter P, Ferro F, et al. Detection of incidental intracranial aneurysms. Surg Neurol. cerebral aneurysms in nontraumatic subarachnoid 1990; 34:361–5. haemorrhage: role of multislice CT angiography in 36. Rinkel GJ, Djibuti M, Algra A, van Gijn J. Preva- 130 consecutive patients. Radiol Med. 2007; lence and risk of rupture of intracranial aneurysms: 112:123–37. a systematic review. Stroke. 1998; 29:251–6. 32. Kokkinis C, Vlychou M, Zavras GM, Hadjigeorgiou 37. Bederson JB, Awad IA, Wiebers DO, et al. Recom- GM, Papadimitriou A, Fezoulidis IV. The role of 3D- mendations for the management of patients with computed tomography angiography (3D-CTA) in unruptured intracranial aneurysms: a statement for investigation of spontaneous subarachnoid haemor- healthcare professionals from the Stroke Council of rhage: comparison with digital subtraction angiog- the American Heart Association. Circulation. 2000; raphy (DSA) and surgical ﬁndings. Br J Neurosurg. 102:2300–8. 2008; 22:71–8. 38. McCormick WF, Nofzinger JD. Saccular intracranial 33. Perry JJ, Steill IG, Sivilotti ML, et al. Early CT with- aneurysms: an autopsy study. J Neurosurg. 1965; out LP reliably excludes subarachnoid hemorrhage 22:155–9. in neurologically intact ED patients with acute 39. Vermeulen MJ, Schull MJ. Missed diagnosis of sub- headache [abstract]. Acad Emerg Med. 2009; 16(4 arachnoid hemorrhage in the emergency depart- Suppl 1):S149. ment. Stroke. 2007; 38:1216–21.