Worst Headache’ and Subarachnoid Bleeding

ABSTRACT & COMMENTARY

Source: Morgenstern LB, et al. Worst headache and subarachnoid hemorrhage: Prospective, modern computed tomography and spinal fluid analysis. Ann Emerg Med 1998;32:297-304.

Morgenstern and colleagues sought to determine the frequency of subarachnoid hemorrhage (SAH) not detected by modern CT imaging. All patients presenting to an urban ED were asked at triage if they had a headache and, if so, if it was "the worst headache of your life" or a headache they would rate as 10 out of 10 in severity. Those who answered either question affirmatively were eligible for the study. Patients with recent head trauma, fever, cancer or brain masses, focal neurologic deficits, or coagulopathy were excluded. Study patients underwent a CT of the head, and a lumber puncture (LP) was performed if the CT was negative for SAH. The LP was considered diagnostic of SAH if the red cell count in tube 1 was greater than 1,000 and did not decline by greater than 25% in tube 4, and the fluid showed xanthochromia (either visually or by spectrophotometer), or a D-dimer assay was positive.

Over a 15-month period, 455 patients presented with headaches, of which 170 (37%) had "worst headaches" and 107 agreed to the study. Eighteen patients (17%) had a SAH on CT. Of the remainder, 79 consented to an LP, which was positive for SAH in two (2.5%). Morgenstern et al conclude that CT can reliably exclude SAH in 97.5% of patients presenting with "worst headache" symptoms.

Commentary by David J. Karras, MD, FACEP

A quick skim of this article could lead a reader to assume that the CT detects 97.5% of subarachnoid bleeds. This figure is, in fact, the negative predictive value of the CT (the chance that a negative CT predicts the absence of SAH) and its high magnitude simply reflects the fact that most patients with severe headaches do not have bleeds. The study’s limitations make even this conclusion tenuous.

The sample is highly biased, and those enrolled are probably not representative of all emergency department patients with "worst headaches." Complaints were prompted by the triage nurse. Patients who agreed that they had a "10 out of 10" headache were included, along with those reporting the "worst headache of [their] life"—we don’t know how many study patients simply had a recurrent severe headache. One-third of patients presenting to the ED with a headache met their inclusion criteria. Of this group, 38% declined to participate in the study, and further analysis showed that those who did not consent had less severe symptoms than those who did participate.

To draw conclusions about the diagnostic yield of the CT, one would need to perform a gold standard (or at least standard of care) test in all study patients—in this case, an LP. Only 74% of study patients underwent this test. The criteria for a positive LP are drawn from a number of different sources and do not reflect any widely-accepted standard for LP diagnosis of SAH. Indeed, 17 patients undergoing LP (22%) met some of the criteria for SAH even though they had trivial red cell counts. There was very poor diagnostic agreement between the criteria used to assess LP results, calling into question the validity of these measures.

Surprisingly, Morgenstern et al do not state what I believe to be the single most important study finding: The CT missed two of the 20 patients diagnosed with SAH, yielding a CT sensitivity of only 90%. This rate is consistent with prior studies and does not warrant a change in our practice. No other conclusions regarding the diagnostic utility of the CT can be inferred from this study.