Hong Kong Journal of Emergency Medicine Nuchal pain predicts subarachnoid haemorrhage in severe headache patients CT Lui, KL Tsui, CW Kam Objective: To find out predicting symptom(s) of non-traumatic subarachnoid haemorrhage in conscious patients without neurological deficit presenting with the primary complaint of severe headache. Design: Case control study. Setting: Regional public hospital, Hong Kong. Subjects: Non-traumatic subarachnoid haemorrhage (ICD 9 coding 430) patients, who were conscious and with normal neurological examination, presenting to the accident and emergency department with severe headache between April 2005 to March 2007 were searched from the hospital database. The control group was recruited prospectively from the accident and emergency department during the period from November 2007 to April 2008. Results: A total of 33 cases were compared to 57 controls to study on five selected symptoms and the likelihood to predict risk of subarachnoid haemorrhage. All five symptoms were significant by univariate analysis at significance level of 25%: nuchal pain (p<0.001), transient syncope (p=0.137), dizziness (p=0.026), vomiting (0=0.016) and presence of precipitating events (coitus, defecation or quarrel) (p=0.132). Only nuchal pain was statistically significant in the multivariate logistic regression analysis (OR=11.44; 95% CI=2.14 to 61.19) with age and gender controlled. Conclusions: The presence of nuchal pain as a presenting symptom significantly increases the likelihood of non-traumatic subarachnoid haemorrhage by ten times among severe headache patients with normal neurological examination. Other symptoms like transient syncope, dizziness and vomiting are non-specific and do not predict a higher chance of occurrence of subarachnoid haemorrhage. (Hong Kong j.emerg.med. 2008;15:212-217) 2005 4 2007 3 9 430 2007 11 2008 4 33 57 5 5 25% p<0.001 p=0.137 p=0.026 p=0.016 p=0.132 11.44 95% 2.14 61.19 Keywords: Case-control studies, intracranial haemorrhages, neck pain, odds ratio Correspondence to: Tsui Kwok Leung, FRCSEd, FHKAM(Emergency Medicine) Tuen Mun Hospital, Accident & Emergency Department, Tsing Chung Koon Road, Tuen Mun, N.T., Hong Kong Email: kl12judy@yahoo.com.hk Lui Chun Tat, MBBS(HK), MRCSEd Kam Chak Wah, FRCSEd, FHKAM(Emergency Medicine)
Lui et al./subarachnoid haemorrhage 213 Introduction Headache is a common chief complaint to the accident and emergency department (AED) and it accounted for about 4.5% of total attendances in the United States. 1 Non-traumatic subarachnoid haemorrhage (SAH) accounted for 1% of all headaches presenting to the AED. 1,2 Of patients with acute onset of severe headache and normal neurological examination, about 10% have SAH. 3 For SAH patients presenting to the AED with focal neurological deficit, impaired consciousness or coma, the diagnosis is unlikely to be missed nowadays because of the high availability and low threshold to request computed tomography (CT) scan assessment for this group of patients. The real challenge to emergency physicians is to identify those conscious SAH patients with normal neurological examination, i.e. the "walking SAH". Knowing that the volume of patients with headache is enormous, it is impractical and not cost-effective to pursue CT scan for all cases. Therefore, finding a case of "walking SAH" can be ascribed as "finding a needle in a haystack". It is classically taught in old medical textbooks that SAH patients would present with sudden onset, severe headache described as the "worst headache of my life" or "thunderclap headache" and it is frequently associated with vomiting and transient syncope. 4 However, we are not sure how true this statement is. Obviously, pain is a fairly subjective symptom. Various means to quantify pain including visual analogue scales have been invented but the reproducibility and reliability are variable and unsatisfactory. In a prospective study of 148 patients presenting with sudden and severe headache, SAH was present in 25% overall and in 12% of those, headache was the only symptom. 5 Similar findings were noted in another report in which 20 of 107 patients with the "worst headache of my life" had SAH. 6 In a review, 97% of the conscious patients who were diagnosed to have SAH complained of severe headache. 7 Sudden onset of severe headache could be a sensitive symptom of SAH, but it is non-specific. It is observed that SAH associates with other symptoms like vomiting, dizziness, transient syncope, nuchal pain, and may be precipitated by events like coitus, defecation or quarrel. In this study we aimed at comparing different symptoms on the likelihood to predict SAH in patients who were fully conscious and with normal neurological examination, presenting to the AED for acute onset severe headache. Methods This case-control study was conducted at the AED of a regional hospital in Hong Kong. The regional hospital serves a population of above a million people living in the New Territories region of Hong Kong and has a diversified clientele with daily AED attendance of about 600 patients. All adult patients of age 18 or above who presented to the AED during a study period of 2 years from 1st April 2005 to 31st March 2007 because of severe headache and diagnosed to have nontraumatic SAH were included. The eligible cases were searched from the hospital electronic record database by the Clinical Data Analysis & Reporting System (CDARS), which is a computerised data retrieval system of the Hospital Authority of Hong Kong. We searched the diagnosis of non-traumatic subarachnoid haemorrhage (ICD 9 coding 430) under the categories of principal and secondary diagnoses. All case records were retrieved and then reviewed by one of the authors. All paediatric cases were excluded. All cases which presented to the AED with impaired consciousness (Glasgow Coma Scale less than 15) or with focal neurological deficit were excluded. The control group representing the study variables in the source population was recruited at the same study site in a period of 6 months starting from 1st November 2007 to 30th April 2008. The control group was made up of non-sah adult patients presenting with a chief complaint of sudden onset of severe headache not associated with a recent head injury or obvious pathology (e.g. frontal sinusitis) and had normal neurological findings on examination. A standard data collection sheet was used to collect the target information and was filled in by the case emergency physician immediately after the consultation. Severe
214 Hong Kong j. emerg. med. Vol. 15(4) Oct 2008 headache is difficult to define as there is no validated scoring system to reflect the diversified nature and characteristics of headache. As a logical way, we defined severe headache in this study as a headache that by itself, exceeded the threshold of the attending emergency physician to pursue a CT scan examination for the subject. The symptom variables (predicting factors) to examine were nuchal pain (defined as neck pain below the occipitoatlantal junction), transient syncope, dizziness, vomiting and the presence of selected precipitating events (coitus, defecation or quarrel). All variables were first screened through univariate analysis by nonparametric hypothesis test (Chi square or Fisher's exact test) at the 25% level of significance. All factors which passed the univariate analysis screening were included in the model of unconditioned logistic regression analysis. Age and gender were also included as confounding variables. Likelihood ratio statistic was used to assess the goodness of fit of the model and level of significance was taken at 5%. The statistical package employed was NLOGIT 3.0. There was no epidemiology study in the literature to provide suitable reference on sample size planning. We therefore planned our sample size by assuming a power of 80%, a significance level of 5%, baseline response probability as 0.3 to detect a moderate effect size of odds ratio=3 or higher. The total sample size (case plus control) required would be 95 as calculated by PASS 2008 Software by NCSS, USA. Results A total of 86 adult SAH cases were retrieved by CDARS. After reviewing the case notes, 9 cases of traumatic SAH due to wrong coding were excluded. Seventeen cases with significant incomplete documentation on symptoms (headache or other) were excluded. Twenty-seven cases were associated with abnormal neurological findings (in coma or focal neurological deficit) on presentation and were also excluded. All the remaining 33 cases presented with severe headache and were included in the final analysis. For the control group, a total of 230 datasheets were collected. After reviewing the datasheet and case records, 28 paediatric cases were excluded. Five cases had focal sign, impaired consciousness or intra-cerebral pathology other than SAH and therefore were excluded. One hundred and forty cases did not have CT scan done in the AED and were also excluded. Finally, a total of 57 control subjects were included for analysis. All the 57 control cases were checked to have no subsequent admission due to SAH within three months after the current attendance. The occurrence of the study parameters for both case and control groups are shown in Table 1 and the most frequent symptoms for the SAH group were dizziness (76%), vomiting (67%), nuchal pain (33%), transient syncope (15%) and precipitating events (6%) respectively in a descending order. For univariate analysis, as shown in Table 1, all five factors were significant at the level of 25%. For multivariate analysis, the factor of precipitating event was not examined because no control case was found to have a precipitating event and therefore, this factor could not be analysed by the logistic regression model. As shown in Table 2, a total of six factors, namely dizziness, vomiting, nuchal pain, transient syncope, age and gender were entered into logistic regression analysis. Nuchal pain was found to be the only factor significantly (p=0.004) associated with SAH (OR= 11.44, 95% CI=2.14 to 61.19). Syncope was possibly associated with SAH (OR=4.70, 95% CI=0.79 to 27.42) but only at a level of 10% significance (p=0.089). The regression model passed the goodness of fit test at p<0.001 and well fitted with the observed score. Discussion We set out to explore the role of symptoms other than headache to predict the likelihood of non-traumatic SAH in conscious patients with normal neurological
Lui et al./subarachnoid haemorrhage 215 Table 1. Univariate analysis for subarachnoid haemorrhage Predictive symptom SAH group Control P value (two-tailed) (total n=33) (total n=57) number (%) number (%) Nuchal pain <0.001 Yes 11 (33.3%) 2 (3.5%) No 22 (66.7%) 55 (96.5%) Syncope 0.137 Yes 5 (15.2%) 3 (5.3%) No 28 (84.8%) 54 (94.7%) Dizziness 0.026 Yes 25 (75.8%) 29 (50.9%) No 8 (24.2%) 28 (49.1%) Vomiting 0.016 Yes 22 (66.7%) 22 (38.6%) No 11 (33.3%) 35 (61.4%) Precipitating event 0.132 Yes* 2 (6.1%) 0 (0%) No 31 (93.9%) 57 (100%) * One or more of the following events: coitus, quarrel or defecation. findings. Out of the many associated symptoms, we have shown that only nuchal pain is significantly associated with a higher likelihood of SAH. In contrast to previous belief, the presence of vomiting, dizziness or transient syncope does not increase the likelihood of SAH in a patient with severe headache. In other words, they may not play a significant role in making the decision to pursue CT scan investigation for severe headache patients. The largest single-institution case series study on SAH reported that the most common misdiagnoses for SAH were migraine and tension headache; and good grade (Hunt and Hess grade 1 and 2) cases were more commonly misdiagnosed. 8 The missed diagnosis rate of SAH attributed to the AED alone was about 5%. 8,9 Notably, failure to obtain a CT scan has been ascribed as the most important cause of delayed or missed diagnosis of SAH especially in so-called "minor leaks" or normal neurological status SAH. 10 This has serious implication because the evidence suggested that the better grade SAH patients with better prognosis were jeopardised by the failure of the physicians or their lower tendency to purse CT scan for this "most needed" group of patients. Perry et al suggested that the approach to patients with acute severe headache appeared ready for a clinical decision rule. The area under the receiver operating characteristic (ROC) curve for SAH was 0.85 (95% CI=0.08 to 0.91) and just slightly less than the figure of 0.88 in the Ottawa's ankle rule. They also indicated that emergency physicians discriminated moderately well between headaches due to SAH and other types of headaches despite the absence of any decision rule. Using a pretest probability of 2% or greater as the threshold to use diagnostic tests yielded a sensitivity of clinical suspicion of 93% and specificity of 49%. 1 Based on the result of our study, presence of nuchal pain as an associated symptom will increase the odds to have SAH by 11 times and theoretically help to sharpen the clinical suspicion and push the ROC curve to the left i.e. to reduce missed diagnoses as well as negative diagnostic investigations. Further to that, nuchal pain quite frequently occurred in our SAH cases (about 33%), and this ascertain its property as a useful discriminating symptom in clinical practice. In this study, we observed a few cases of SAH after coitus, defecation, or quarrel. They represent events that may cause soaring in blood pressure to trigger
216 Hong Kong j. emerg. med. Vol. 15(4) Oct 2008 Table 2. Multivariate logistic regression analysis for subarachnoid haemorrhage Predictive variables % of SAH Ratio of SAH in the group Odds ratio 95% CI P value Age (mean±sd) [yrs] 1.030 0.99-1.07 0.107 SAH: 50.0±12.0 Control: 45.9±15.8 Gender 0.495 Male 37.5% 15/40 0.69 0.23-2.02 Female 36.0% 18/50 1 Nuchal pain 0.004 Yes 84.6% 11/13 11.44 2.14-61.19 No 28.6% 22/77 1 Syncope 0.089 Yes 62.5% 5/8 4.70 0.79-27.42 No 34.1% 28/82 1 Dizziness 0.114 Yes 46.3% 25/54 2.41 0.81-7.19 No 22.2% 8/36 1 Vomiting 0.106 Yes 50.0% 22/44 2.39 0.831-6.898 No 23.9% 11/46 1 Goodness of fit test Number of observations 90 Unrestricted log likelihood -45.89 Restricted log likelihood -59.14 Likelihood ratio statistic 26.50* p-level <0.001 * Passed the internal consistency test at the 95% confidence level rupture of an existing aneurysm or arteriovenous malformation. However, due to the extremely small number of cases in our sample, their significance could not be clearly determined. Transient syncope was found to have a moderate odds ratio of 4.7 and was significant at 10% level (p=0.089). It suggests that it may be useful in predicting SAH to some extent. Notably, transient syncope occurred only in less than 6% of the control group. The low incidence in the sample population has reduced the study power and we are taking a risk to commit a high type 2 error when rejecting transient syncope at a significant level of 5%. This study has a number of limitations. Firstly, the definition of severe headache adopted in this study was imperfect. We defined severe headache as a headache that could alarm the attending emergency physician to pursue for a CT examination. However, the threshold to pursue CT scan varied among different emergency physicians and the variability was not controlled. Secondly, the information of the SAH group was collected retrospectively and misclassification errors were possible. Thirdly, the control group was selected prospectively by collecting the filled-up datasheets. The number of datasheet collected was much smaller than the expected number of severe headache cases that had actually occurred during the recruitment period. The control group was therefore potentially not consecutive cases and it posed a possible selection bias. Furthermore, the case and control groups were recruited from different time periods; again, selection bias could be introduced. Lastly, although we controlled age and gender as possible confounders, there could be other confounders uncontrolled like hypertension or smoking status.
Lui et al./subarachnoid haemorrhage 217 Conclusion In this single-institution, small-scale case control study, we determined that the presence of nuchal pain as a presenting symptom significantly increased (10 times) the likelihood of SAH among severe headache patients with normal neurological examination. Other symptoms like vomiting, transient syncope and dizziness were not specific and did not help to predict a higher chance of occurrence of SAH. References 1. Perry JJ, Stiell IG, Wells GA, Mortensen M, Lesiuk H, Sivilotti M, et al. Attitudes and judgment of emergency physicians in the management of patients with acute headache. Acad Emerg Med 2005;12(1):33-7. 2. Vermeulen M, van Gijn J. The diagnosis of subarachnoid haemorrhage. J Neurol Neurosurg Psychiatry 1990;53(5):365-72. 3. Edlow JA, Malek AM, Ogilvy CS. Aneurysmal subarachnoid hemorrhage: update for emergency physicians. J Emerg Med 2008;34(3):237-51. 4. Linn FH, Wijdicks EF. Causes and management of thunderclap headache: a comprehensive review. Neurologist 2002;8(5):279-89. 5. Kowalski RG, Claassen J, Kreiter KT, Bates JE, Ostapkovich ND, Connolly ES, et al. Initial misdiagnosis and outcome after subarachnoid hemorrhage. JAMA 2004;291(7):866-9. 6. Vermeulen MJ, Schull MJ. Missed diagnosis of subarachnoid hemorrhage in the emergency department. Stroke 2007;38(4):1216-21. 7. Gorelick PB, Hier DB, Caplan LR, Langenberg P. Headache in acute cerebrovascular disease. Neurology 1986;36(11):1445-50. 8. Mayer PL, Awad IA, Todor R, Harbaugh K, Varnavas G, Lansen TA, et al. Misdiagnosis of symptomatic cerebral aneurysm. Prevalence and correlation with outcome at four institutions. Stroke 1996;27(9):1558-63. 9. Linn FH, Wijdicks EF, van der Graaf Y, Weerdesteynvan Vliet FA, Bartelds AI, van Gijn J. Prospective study of sentinel headache in aneurysmal subarachnoid haemorrhage. Lancet 1994;344(8922):590-3. 10. Morgenstern LB, Luna-Gonzales H, Huber JC Jr, Wong SS, Uthman MO, Gurian JH, et al. Worst headache and subarachnoid hemorrhage: prospective, modern computed tomography and spinal fluid analysis. Ann Emerg Med 1998;32(3 Pt 1):297-304.