New guidelines for diagnostic testing in patients with nonacute headache have been released by the U.S. Headache Consortium and endorsed by the American Academy of Family Physicians, the American Academy of Neurology, the American Headache Society, the American College of Emergency Physicians, the American College of Physicians-American Society of Internal Medicine, the American Osteopathic Association, and the National Headache Foundation.
The U.S. Headache Consortium is a group of 14 physicians (mainly academic neurologists) whose aim is to “develop scientifically sound, clinically relevant practice guidelines on chronic headache in the primary care setting.” The authors’ definition of nonacute headache includes all headache syndromes that have occurred for at least four weeks in the patient’s lifetime. The recommendations are based on a comprehensive review and meta-analysis of scientific evidence. The full text is available online athttp://www.aan.com/professionals/practice/guideline/index.cfm.
|Level I||Independent, blind comparison with a “gold standard” of anatomy, physiology, diagnosis, or prognosis among a large number of consecutive patients suspected of having the target condition|
|Level II||Independent, blind comparison with a “gold standard” among a small number of consecutive patients suspected of having the target condition|
|Level III||Independent, blind comparison with a “gold standard” among non-consecutive patients suspected of having the target condition|
|Level IV||Studies that did not meet criteria for at least level III evidence|
The majority of headaches are primary headache disorders, such as migraine, cluster headache, and tension-type headache. Secondary headaches (i.e., those with underlying pathology such as tumor, aneurysm, or giant cell arteritis) are less common. Because of the rarity of secondary headaches, there is concern about the diagnostic utility of routine neuroimaging studies, either computed tomography (CT) or magnetic resonance imaging (MRI), to exclude underlying causes of headache.
The authors reviewed the published clinical evidence in the medical literature from January 1966 through August 1998, including all English-language studies that estimated the sensitivity, specificity, or predictive value of neuroimaging in patients presenting with nonacute headache. They also included studies examining observer variation or reproducibility of diagnostic tests for patients with nonacute headache.
In their evaluation of the medical aspects of neuroimaging, the authors focused exclusively on significant abnormalities, defined as abnormalities related to headache that may require further action (e.g., acute cerebral infarct, neoplastic disease, hydrocephalus), and vascular abnormalities such as aneurysm or arteriovenous malformation.
Of 28 studies that met the inclusion criteria, 22 were retrospective, and the six prospective studies lacked blinding. All 28 studies were reviewed and assigned a quality grade. The authors assigned a level IV quality grade to all of the studies. Definitions of evidence levels used by the U.S. Headache Consortium in this guideline are listed in Table 1.
Most of the studies did not characterize the study population in terms of duration of headache disorder, included a discussion of the neurologic examination but did not comment on symptoms, or used populations of patients with acute and nonacute headache.
History and Physical Examination Findings
Of the 28 studies, eight reported sufficient information to relate specific findings on history or physical examination to the occurrence of significant abnormalities on CT or MRI.
Neuroimaging abnormalities in patients with neurologic findings on examination were discussed in five studies. In four of these studies, an abnormal neurologic examination significantly increased the likelihood of finding a significant abnormality on neuroimaging. In all five studies, the absence of findings on the neurologic examination led to a decreased likelihood of finding a significant lesion on neuroimaging.
The accuracy of tests was reported as likelihood ratios (LRs). LRs greater than 1.0 increase the likelihood of disease, while LRs less than 1.0 decrease the likelihood of disease. Values near 1.0 mean that the test results do not significantly alter the likelihood of disease. Abnormal findings on neurologic examination have an LR of 3.0 (95 percent confidence interval [CI], 2.3 to 4.0). Because the probability of a significant intracranial abnormality is low (less than one in 100), the positive predictive value of an intracranial abnormality in a patient with an abnormal neurologic examination still is very low (less than three in 100). A combined negative LR of 0.7 (95 percent CI, 0.52 to 0.93) suggests that a normal neurologic examination reduces the odds of finding a significant intracranial abnormality on neuroimaging by 30 percent.
Two studies identified individual neurologic signs and symptoms that were related to the presence of significant abnormalities on imaging. The following symptoms increased the likelihood of finding a significant abnormality on neuroimaging: rapidly increasing headache frequency; history of dizziness or lack of coordination; history of subjective numbness or tingling; and history of headache causing awakening from sleep. However, the absence of these findings did not significantly lower the likelihood of finding a neuroimaging abnormality. Contrary to common practice, the same two studies also found that headache accompanied by nausea, a history of syncope, or experiencing the “worst headache” of the patient’s life did not significantly increase the likelihood of finding a significant abnormality on neuroimaging.
The studies also reported the distribution of several continuous variables (e.g., age, duration of headache, headache frequency, ergot consumption) for patients with normal versus abnormal neuroimaging results. A statistically significant difference in the duration of the headache disorder between those with and those without a significant abnormality on neuroimaging was reported in a study of patients with recent onset of headache or a clear change in the character of their nonacute headache during the past year. Therefore, duration of headache disorder may be a useful predictor of significant abnormalities among patients with recent-onset headache symptoms, but not useful among patients with longstanding headache disorders.
Nonacute Headache and Normal Neurologic Examination
Among patients with headaches diagnosed as migraines and normal results on neurologic examinations, the prevalence of significant intracranial abnormalities on neuroimaging ranged from zero to 3.1 percent in 11 studies. These results were combined in a meta-analysis to yield a summary prevalence of approximately 0.2 percent, with an upper 95 percent confidence limit of approximately 0.6 percent, representing a prevalence that is less than the 0.8 percent incidence of arteriovenous malformations and the 2.4 percent incidence of saccular aneurysms found in autopsy in the general population.
Two small studies reported no cases of significant intracranial abnormalities among patients with normal results on neurologic examinations and diagnoses of tension-type headaches.
UNSPECIFIED TYPE OF HEADACHE
In patients with normal neurologic examinations who presented with headache not described as migraine or tension type, the rate of finding significant intracranial abnormalities (range: zero to 6.7 percent) was more variable than among patients with migraine.
Relative Effectiveness of CT and MRI
Three studies compared CT and MRI in patients with headache. No significant abnormalities were detected in patients from these studies, so the relative effectiveness of CT and MRI for detecting significant lesions cannot be determined. The consortium authors conclude that the limited number of studies identified and reviewed suggests that MRI may be more sensitive than CT for identifying clinically insignificant abnormalities. However, MRI may not be more sensitive for identifying clinically significant pathology relevant to the cause of headache.
|Recommendation—neuroimaging should be considered in patients with nonacute headache and an unexplained abnormal finding on the neurologic examination (Grade B†).|
|Recommendation—evidence is insufficient to make specific recommendations regarding neuroimaging in the presence or absence of neurologic symptoms (Grade C†).|
|Migraine and a normal neurologic examination|
|Recommendation—neuroimaging usually is not warranted for patients with migraine and normal neurologic examination (Grade B†). For patients with atypical headache features or patients who do not fulfill the strict definition of migraine (or have some additional risk factor), a lower threshold for neuroimaging may be applied (Grade C†).|
|Tension-type headache and normal neurologic examination|
|Recommendation—data were insufficient to make an evidence-based recommendation regarding the use of neuroimaging for tension-type headache (Grade C†).|
|Effectiveness of CT versus MRI|
|Recommendation—data were insufficient to make any evidence-based recommendations regarding the relative sensitivity of MRI compared with CT in the evaluation of migraine or other nonacute headache (Grade C†).|
Principles of Management
The U.S. Headache Consortium identified three consensus-based (not evidence-based) general principles of management for making decisions regarding neuroimaging in patients with headache: (1) testing should be avoided if it will not lead to a change in management; (2) testing is not recommended if the patient is not significantly more likely than anyone else in the general population to have a significant abnormality; and (3) testing that normally may not be recommended as a population policy may make sense at an individual level, resources notwithstanding. For example, exceptions may be considered for patients who are disabled by their fear of serious pathology, or when the physician is suspicious even in the absence of known predictors of abnormalities on neuroimaging studies (i.e., red flags).
Specific treatment recommendations from the U.S. Headache Consortium are listed in Table 2.