I ABSTRACT 16 ACADEMIC EMERGENCY MEDICINE JAN 1996 VOL 3/NO 1

Transcription

1 16 ACADEMIC EMERGENCY MEDICINE JAN 1996 VOL 3/NO 1 Sensitivity of New-generation Computed Tomography in Subarachnoid Hemorrhage Thomas A. Sames, MD, Alan B. Storrow, MD, Jeff A. Finkelstein, MD, Michael R. Magoon, MD I ABSTRACT , ,.,..,..,.,, Objective: To determine the sensitivity of the initial new-generation CT (NGCT) scan interpretation for detection of acute nontraumatic subarachnoid hemorrhage (SAH) and to decide whether lumbar puncture (LP) should follow a normal NGCT scan. Methods: A retrospective chart review was performed of patients admitted between March 1988 and July 1994 with proven SAH. Exclusion criteria were age <2 years, diagnosis other than acute SAH, lhistory of head trauma within 24 hours before symptom onset, NGCT scan not done before diagnosis, and records not available. Patients were placed into two groups: symptom duration <24 hours (group 1) and :.24 hours (group 2) prior to CT scan. The resolution of each NGCT scanner was recorded. An NGCT scanner was defined as a third-generation scanner or more recent. Results: Of 349 SAH patients, 181 met inclusion criteria. The sensitivity of NGCT scans for SAH was 93.1% for the group 1 patients (n = 144) and 83.8% for the group 2 patients (n = 37). The overall sensitivity was 91.2%. All the patients who had SAH not detected by NGCT scans were diagnosed by LP. There was no significant relationship between NGCT scanner resolution and sensitivity for SAH. Conclusion: Initial interpretation of NGCT scans to detect SAH does not approach 100% sensitivity. A normal NGCT scan does not reliably exclude the need for LP in patients who have symptoms wggestive of SAH. Key words: subarachnoid hemorrhage; computed tomography; intracranial hemorrhage; cerebrovascular - accident; imaging; sensitivity; lumbar puncture Acad. Emerg. Med. 1996; 3: Acute nontraumatic subarachnoid hemorrhage (SAH) is a diagnosis often considered in the ED. Aggressive pursuit of the diagnosis is necessary because the overall mortality associated with the disease is approximately 50%. Computed tomography (CT) replaced lumbar puncture (LP) as the initial screening study for sus- From the Joint Military Medical Centers, San Antonio, TX, Emergency Medicine Residency (TAS, ABS, JAF, MRM). Received: May ; revision received: Augusi I, 1995; accepted: August 4, 1995; updated: August 31, Prior presentation: Society of Air Force Clinical Surgeons annual meeting, Dayton, OH. April 1995; and SAEM annual meeting, San Antonio, TX, May Disclaimer: The opinions and assertions contained herein are those of the aulhors and should not be considered as official or as representing the opinions of the Departments of the Army or Air Force. Address for correspondence and reprints: Thomas A. Sames, MD, Wilford Hall Medical CenterlPSAE, 2200 Bergquist Drive, Suite 1, Lackland A FB, TX pected SAHs in the 1970~.~.~ However, LP is considered to be the criterion standard for diagnosis because blood within the cerebrospinal fluid (CSF) should be present in all cases.4 Current standard of care dictates that an LP is indicated for suspected SAHs in patients who have normal CT scans because the sensitivity of detecting SAH by older-generation CT scanners is approximately 90%. Blood disperses in the CSF over time, thus, the sensitivity of CT to detect SAH declines as symptom duration increases.6 Older-generation CT scanners use a translate-rotate design, whereas newer-generation CT (NGCT) scanners use a rotate-rotate (third-generation) or a rotate-stationary (fourth-generation) design. These newer designs along with improved computer hardware and software allow for superior performance and better resolution.* The low-contrast resolution (LCR) characteristic of each model of CT scanner is a multiple (object size x Hounsfield units) defining the ability to differentiate the attenuation coefficients of adjacent pieces of tissue (i.e., blood vs brain parenchyma).8 A lower LCR number represents better discrimination of

2 ~~ CT for Subarachnoid Hemorrhage, Sames et al. 17 blood from CSF and thus better sensitivity. Some articles report that NGCT scanners have an improved sensitivity for detecting SAH. - However, for the clinician, two important questions remain: 1) how sensitive is the NGCT scan for SAH and 2) is there still a need to do an LP when the NGCT scan is read as normal by a radiologist? I METHODS Study Design A retrospective, descriptive study was performed to determine the sensitivity of head CT scans for SAH with the NGCT scanners. Population and Setting Patients diagnosed as having acute nontraumatic SAH were identified from four neurosurgical centers in San Antonio, Texas. We reviewed medical records of all patients admitted between March 1, 1988, and July 31, 1994, with a diagnosis of acute nontraumatic SAH (ICD- 9, 430).12 Inclusion criteria were: age >2 years; diagnosis of SAH made by either CT, CSF examination, angiography, surgery, or autopsy; no history of head trauma within 24 hours prior to symptom onset; NGCT scan done prior to diagnosis; and medical records available for review. Patients transferred to one of the neurosurgical centers with an acute SAH were included if they otherwise met these inclusion criteria. Measurements Medical records were reviewed for the following patient information: age, sex, scanner type, symptom duration prior to scan, radiologist s interpretation of initial CT scan, and results of LP, angiography, surgery, or autopsy (if done). The CT scanner run logs were used to validate the actual time of a scan and the type of scanner used. The CT film and the accompanying report also were used to verify information about the I TABLE I Subdivision of Patients with Subarachnoid Hemorrhage Corresponding to the New-generation CT Scanner Class, the Symptom Duration, and the Initial Radiologist s Interpretation Symptoms <24 Hr Symptoms >24 Hr Scanner Class +Scan -Scan +Scan -Scan Class A (LCR < 0.75) Class B (LCR ) Class C (LCR > 1.2) Unknown *LCR = low-contrast resolution radiologist s interpretation, type of scanner, and time of scan. The first documented radiologist s reading for the initial CT scan was used as the study CT interpretation. The initial radiologist was either a neuroradiologist, a general radiologist, or a radiology resident on call after duty hours in one of the teaching hospitals. An NGCT scanner was defined as a third-generation scanner or higher. The published manufacturer s product specifications defined LCR for each NGCT scanner used in this study.i3 New-generation CT scanners were divided into classes corresponding to LCR: class A (GE HiLight Advantage class), LCR < 0.75; class B (GE 9800 class), LCR ; and class C (GE 8800 class), LCR > 1.2. Data Analysis The sensitivity for detection of SAH by NGCT scanning was determined for groups of patients defined by symptom duration at the time of the scan: <24 hours (group 1) or >24 hours (group 2). The two groups were compared using chi-square analysis (EpiInfo, CDC, version 5.01b, Atlanta, GA, 1991). The significance level was set at p < Confidence intervals for overall sensitivity were calc~lated. ~ Groups 1 and 2 were subdivided corresponding to NGCT class (groups la, lb, lc, 2A, 2B, and 2C), and the sensitivity of each group was determined. The groups were compared using Fisher s exact test to determine whether a difference in sensitivity existed between scanners with differing LCRs. The power to detect a 25% sensitivity difference in groups was calculated (NCSS Power Analysis and Sample Size, version 1.0, Kaysville, UT, 1993). I RESULTS Of 349 records identified, 256 were available for review. Seventy-five patients were excluded for the following reasons: 64 did not have acute SAH, eight were suspected of head trauma within 24 hours of symptom onset, two CT films or reports were not found, and one NGCT scan was not done prior to the diagnosis. Of the 181 patients included, 114 were female and 67 were male. The average age was 53, with a range of years. Fifty-one (28%) of 181 patients died before hospital discharge and 79 (43%) were discharged with some type of neurologic impairment. The overall sensitivity of the NGCT scan interpretation for SAH was 91.2% (95% CI = 87.1%-95.3%). The sensitivity of NGCT scans was 93.1% for group 1 (n = 144) and 83.8% for group 2 (n = 37). There was not a significant difference in sensitivity between groups 1 and 2 (p = 0.08). The power to detect 25% difference in sensitivity between these groups was 89%. Seven of

3 18 ACADEMIC EMERGENCY MEDICINE JAN 1996 VOL 3/NO 1 1 TABLE 2 Manufacturer, Types of Scanners Used, and Corremanding Low-contrast Resolution (LCR) Manufacturer Scanner LCR General Electric Medical Systems 3000 Grandview Blvd. Waukesha, WI Siemens Medical Systems, Inc Northwoods Parkway Norcross, GA 3071 Philips Medical Systems 710 Bridgeport Ave. Shelton, CT Picker International 595 Miner Rd. Highlander Heights, OH Toshiba Medical Systems 2441 Michelle Dr. Tustin, CA Technicare (out of business) GE HiLight Advantage 0.63 GE GE Siemens Somotome Plus S 0.75 Siemens Somotome Plus 0.75 Siemens High Q 0.85 Siemens DRH 1.00 Siemens Somotome CR 1.20 Siemens DR Siemens Somotome Philips SR Philips 60TX 1.50 Philips Tomoscan CX 1.50 Picker 1200 Picker 600 Toshiba TCT600XT oo Technicare Deltascan ten patients in group 1 with normal scans were found to have aneurysms by angiography. For the ten scans in group 1 interpreted as normal, six were scanned and transferred from a smaller hospital, three were scanned at one of the neurosurgical centers after normal working hours when a radiology resident provides the initial interpretation of the scan, and one was read by i 10 0 Hours a neuroradiologist. The patient whose scan WiiS read as normal by the neuroradiologist gave a history of sudden severe headache while deer hunting, and SAH was determined by LP (394 red blood cells in the fourth CSF tube). This patient subsequently had a norrnial angiogram and was discharged from the hospital without neurologic impairment. The CT scan results categorized by NGCT class and patient group (based on symptom duration) are listed in Table 1. The sensitivities to detect SAH for groups 1A-IC were 96%, 90%, and 96%, respectively. The sensitivities for groups 2A-2C were 63%, E16%, and 92%, respectively. The 17 different NGCT scanners (ten different LCRs, range ) used on patients in the study are listed in Table 2. A lower LCIl number should provide better discrimination of blood from CSF and thus better sensitivity. A significant association between NGCT class and sensitivity was not found when groups 1A-1C and 2A-2C were compared (Fig. 1). The p-values for groups 1A-1C and 2A-2C were 0.60 and 0.13, respectively (power < 0.8). I DISCUSSION Patient presentations consistent with SAH are frequently encountered by an emergency physician (EP). Making the diagnosis of SAH is essential because the overall mortality is approximately 50% and tlhe potential for rebleeding is high. ls.l6 Computed tomography was introduced in 1973, and its accuracy for detecting SAH has been controversial. In 1974 Scott et al. reported a 50% sensitivity for detection of SAH. Scotti et al., in 1977, after reviewing 46 patients who had SAH, assumed that CT was 100% sensitive if performed within five to seven days, thus, suggesting that LP may be avoided-relegating the LP to a historical procedure in the diagnosis of SAH.IX LCR Class CLASS A I -CLASS B CLASS c I FIGURE 1. Comparison of sensitivities corresponding to new-generation CT scanner class. LCR = low-contrast resolution.

4 CT for Subarachnoid Hemorrhage, Sumes er ul. 19 However, the Cooperative Aneurysm Study (1,157 patients who had SAH from January 1981 to February 1982), reported sensitivities of 90.5% for symptom duration <24 hours and 84.2% if scanned after 24 hours.5 They concluded that although CT should be the initial diagnostic study, a normal CT scan should not preclude LP for patients with suspected SAH. The type of scanners used in this study was not reported, but most likely second-generation scanners were used, because use of third-generation scanners did not become widespread until Although the sensitivity of the NGCT scan has not been defined in the literature, several authors claim a higher sensitivity with NGCT One author claims that the sensitivity of the NGCT scan to detect SAH approaches loo%, and discourages the use of LP to make the diagno~is. ~ This author, like Scotti et al., appears to have made an assumption without data to support the conclusion. This unsubstantiated belief may lead to the misdiagnosis of a patient having SAH. The test performance characteristic of interest to EPs is the initial CT reader s ability to discriminate blood from CSF. In part this is related to the quality of the image produced by the CT scanner, thus giving rise to claims of a higher sensitivity with NGCT scans. The training and skill of the reader also affect the sensitivity of the initial interpretation. We used the initial radiologist s interpretation as the criterion standard for CT interpretation because EPs are not always afforded the luxury of second or third interpretations by other or more qualified radiologists. We sought to determine the sensitivity of NGCT scan interpretations for SAH and to ascertain whether EPs could forgo performing LP on a patient with symptoms suspicious for SAH and a normal CT interpretation. Although LP is a minor procedure, it costs money, causes patient discomfort, and increases patient waiting time. On the other hand, a missed small SAH is likely to result in a subsequent SAH, with permanent neurologic deficits.6 Those patients diagnosed by CSF examination will benefit greatly from further diagnostic testing (i.e., angiography) and possibly surgical intervention. Thus, CT scan sensitivity must approach 100% before one can recommend omitting the LP. We found that the sensitivity of NGCT scans for SAH does not approach 100% sufficiently to negate the need for an LP in the face of a negative NGCT scan. Our data support the results of the Cooperative Aneurysm Study. Interestingly, we found no association between the sensitivity of NGCT scanners and their LCRs, although the power of this analysis was low. Factors that were not specifically studied but that might explain our results include: 1) radiologist experience and specialty training, 2) technique used to do the scan, and 3) human physiologic factors (in other words, has the evolution of the image quality reached the point where the educated human eye cannot detect a difference?). I LIMITATIONS AND FUTURE QUESTIONS Our retrospective study design has a number of inherent limitations. A large number of patients were excluded for lack of available records. The role of physical examination signs was not studied; it would be interesting to know how many of the SAH patients with normal head CT scans had neurologic findings. If they all had focal findings, then clinical judgment would prevail and LP may have been performed anyway. Also, our power for comparison of patient subgroups (e.g., 1A- 1C and 2A-2C) was limited due to small numbers of patients in each group. Radiologist training and experience were not studied and could impact initial interpretation of the CT scan and resultant NGCT scan sensitivity; thus, our results may not be applicable to all ED settings. Furthermore, some missed CT readings might have been correctly interpreted with additional review(s) of the CT scan. Several important questions remain: 1) what is the likelihood of SAH for a neurologically intact patient presenting with the worst headache of my life and a normal CT scan, when the headache disappears spontaneously or after administration of an analgesic?; 2) which criteria can be used to divide patients into highand low-probability groups prior to obtaining the CT scan?; 3) can LP be withheld for neurologically intact patients who have low pretest (i.e., prescan) probabilities and are subsequently found to have normal scans?; 4) do these patients need follow-up?; 5) if so, what follow-up is appropriate?; and 6) would there be a detectable difference in classes of scanners with differing LCRs if larger groups were compared? I CONCLUSION We found the overall sensitivity of NGCT to be 91.2% (95% CI = 87.1%-95.3%). The sensitivity decreases as the interval between symptom onset and initial scan increases. We recommend performing LP on patients with symptoms suspicious for SAH who also have normal CT scans. Aggressive pursuit of the diagnosis is necessary to identify those individuals with SAH, since therapy targeted at the source of the SAH can reduce morbidity and mortality. The authors express appreciation to Stanley Fox, PhD, physicist, General Electric Company, for technical advice, and Gregory Osborne, AlC, USAF, for administrative assistance.

5 20 ACADEMIC EMERGENCY MEDICINE JAN 1996 VOL 3/NO 1 I REFERENCES 1. Little N. Acute head pain. Emerg Med Clinic North Am. 1987; 5: Leicht MJ. Non-traumatic headache in the emergency department. Ann Emerg Med. 1980; 9: Bell BA, Kendall BE, Symon L. Computed tomography in aneurysmal subarachnoid hemorrhage. J Neurol Neurosurg Psychiatry. 1980; Rowland LP. Blood and CSF in cerebrovascular disease. In: Rowland LP (ed). Merritt's Textbook of Neurology, 7th ed. Philadelphia: Lea & Febiger, 1984, p Adam HP, Kassell NF, Torner JC, Sahs AL. CT and clinical correlation's in recent aneurysmal subarachnoid hemorrhage: a preliminary report of the Cooperative Aneurysm Study. Neurology. 1983; 33: Espinosa F, Weir B, Noseworthy T. Nonoperative treatment of subarachnoid hemorrhage. In: Youmans JR (ed). Neurological Surgery, 3rd ed. Philadelphia: W. B. Saunders. 1990, pp Kelsey CA. Computerized tomography. In: Kelsey CA (ed). Essentials of Radiology Physics. St. Louis, MO: Warren H. Green, pp Wiesen EJ, Miraldi F. Imaging principles in computer tomography. In: Haaga JR, Lanzieri CF, Sartoris DJ, Zerhouni EA (eds). Computed Tomography and Magnetic Resonance Imaging of the Whole Body, 3rd ed. St. Louis, MO: Mosby-Year Book, 1994, pp Fontanarosa PB. Recognition of subarachnoid hemorrhage. Ann Emerg Med. 1989; 18: Henry GL. Headache. In: Rosen P, Baker FJ, Barkin RM, Braen GR, Dailey RH, Levy RC (eds). Emergency Medicine-Concepts and Clinical Practice, 2nd ed. St. Louis, MO: Nlosby-Year Book, 1988, pp Sidman RD, Connolly EM, Lemke TJ. Computed romography (CT) in the diagnosis of subarachnoid hemorrhage: is spinal fluid analysis always needed when the CT is normal? (abs~ract]. Acad Emerg Med. 1995; 2: Index to diseases. In: Jones MK, Brouch KL, Allen MM, et al. (eds). International Classification of Diseases-9th Revision- Clinical Modification Code Book. Alexandria, VA: St. Anthony Publishing, 1992, p CT scanners: comparison of product specifications. Med Electronics Equipment News. 1994; 34(2): Gardner MJ, Altman DG. Calculating confidence intervals for proportions and their differences. In: Gardner MJ,.4ltman DG (eds). Statistics with Confidence, Confidence Intervals and Statistical Guidelines. London: British Medical Journal, 1989, pp Meyer FB, Morita A, Puumala MR, Nicholes DA. Medical and surgical management of intracranial aneurysms. Mayo Clin Proc. 1995; 70: O'Hare TH. Subarachnoid hemorrhage: a review. J Emerg Med. 1987; 5: Scott WR, New PFJ, Davis KR, Schnur JA. Computerized axial tomography of intracerebral and intraventricular hemorrhage. Radiology. 1974; Scotti G, Ethier R, Melancon D, Terbrugge K, Tchang S. Computed tomography in the evaluation of intracranial aneurysms and subarachnoid hemorrhage. Radiology. 1977; 123: White RJ. Subarachnoid hemorrhage: the lethal intracranial explosion. Emerg Med. 1994; May:74-80.