The patient's history is by far the most important factor in arriving at the correct diagnosis.47 Questions should be directed to the children and their parents. (Remember, when eliciting information directly from a child, use age-appropriate questions, and interview the adolescent patient privately.) In addition, it is important to bear in mind that children and adolescents respond to pain differently. Younger children may react to headache by crying, rocking, and hiding. Older children, on the other hand, can better perceive, localize, and remember pain. Questions listed in Table 5 (see page 7) should assist in the assessment of headache and help identify patients who warrant further diagnostic testing.47
Emergency physicians are trained to view the self-described "worst headache of my life" as extremely suspicious for significant intracranial pathology. A prospective study that evaluated children presenting to a pediatric ED with abrupt onset of severe headache found that the intensity of the pain was not a discriminating factor for significant pathology. Ninety-eight percent of all these patients selected the most severe, tearful, sad face among a group of images, or 5 out of 5 on a pain severity scale.2
Location/Quality of Headache
The location of the pain is most often nonspecific. However, at times the location can be useful for diagnosis. (See Table 6.) In a study by Lewis et al, the headache location was found to be discriminating in patients presenting to the ED with occipital pain. In this 150-patient study, 2 children had occipital headache, and both had brain tumors. Thus 2 out of the 4 total patients with brain tumors complained of occipital headache.2 In the same study, both the quality of pain and the activity scales (in the absence of severity measurement) were found to be discriminating if the child could not describe the pain or if the child was still able to run and play. All the children with headache due to upper respiratory tract infections could easily assign a describer, whereas half of the children with brain tumors and 2 of the 3 with VP shunt malfunctions could not describe the pain. Experts may argue that, even though these results were overall statistically significant, the subset of patients with brain tumors in the study was very small, preventing validation of these conclusions. In addition, only children with URI-related headache indicated that they were still able to play during their headache.
It is important to determine whether the patient has other symptoms, besides headache. Neurological and visual complaints are crucial. In 1991 the Childhood Brain Tumor Consortium published a report of 3291 children diagnosed with brain tumors.36 Overall, 62% of the children with brain tumors experienced chronic or frequent headaches prior to the diagnosis, and 98% of patients with headache had at least 1 other associated symptom or abnormality on the neurological examination. More than 50% of the children had at least 3 or more other symptoms or abnormal signs present at the time of diagnosis. The most frequent symptoms were nausea or vomiting, upper extremity weakness, visual symptoms, difficulty walking, and changes in personality (a sure clue is an adolescent patient who suddenly and unexpectedly requires anger management classes!), academic performance, or speech.
Other associated symptoms, such as neck pain or light sensitivity, particularly when fever is present in patients with severe headache, should raise suspicion for meningitis. In addition, sore throat and abdominal pain, particularly in school-aged patients, ought to lead the clinician to consider streptococcal pharyngitis as a cause of the headache.
Always ask if anyone in the family suffers from migraine headaches. Other avenues to explore include a family history of hypertension, substance abuse, allergy, collagen vascular disease, psychiatric disorders, epilepsy, tumor, chronic infection, and neurocutaneous disorders.
Past Medical History
Inquire about the patient's history regarding sickle cell disease (which can cause stroke or SAH), coagulopathy, and human immunodeficiency virus (HIV).
Prior Headaches/Changes in Patterns
A gradual increase in frequency and severity of headache is the most worrisome pattern in children, usually leading to the most serious diagnosis, including brain tumors, idiopathic intracranial hypertension, hydrocephalus, chronic subdural hematoma, brain abscess, AVM, lead toxicity, medication side effects, and intracranial malformations (eg, Chiari malformations, Dandy-Walker syndromes).
Ask about pregnancy, labor and delivery, prematurity, growth and development, previous head injuries, and operations (particularly VP shunt).
Review of Systems
It is important to ask about associated symptoms, such as vomiting, fever, stiff neck, vertigo, nasal congestion, cough, alteration in vision, toothache, and jaw pain. Of note, vomiting is the most common symptom associated with intracranial pathology (74%).1 (See Table 7.)
A history of HIV could prompt concerns for intracranial infections, such as toxoplasmosis, herpes simplex virus, cytomegalovirus, atypical myobacteruim, fungal infections, tumor, or lymphoma.
Social and Emotional Factors
These factors are critical, particularly in adolescents, to rule out depression presenting as severe headache. Elicit any changes in behavior (tearfulness, withdrawal from usual activities, hopelessness) and inquire about depression, encounters with the law, truancy, divorce or separation in the family, or death of a close friend or relative.
The physical examination for these patients must be quite thorough. To ensure timely care, it is imperative to decide quickly whether there are any signs of toxicity— including meningeal signs — or signs of stress, such as diaphoresis, prior to completing the entire examination.49 (For an extensive listing of physical exam findings and their possible significance, see Table 8.) Note that the neurological examination has demonstrated a high sensitivity for intracranial pathology. In a review of more than 3000 children with brain tumors, 98% had abnormalities on a focused neurological examination that covered mental status, coordination, deep tendon reflexes, sensory, motor, and eye movements, and fundoscopic exam; the implication is that a detailed neurological and eye examination could exclude a brain tumor in up to 98% of cases.36
For patients with a CSF shunt, one must palpate the burr hole, the extracranial proximal catheters, the valve, and the distal catheter tubing all along the track. Assess for fluid collections, tenderness, lack of tubing integrity, and warmth. Fluid collections could be present in 36% of cases of shunt malfunctions.50 Pumpable valves are present in the majority of shunts, though pumping of shunt valves has limited sensitivity (39%) and specificity (85%) for shunt obstruction. Damage of the shunt or valve mechanisms, as well as entrapment of the choroids plexus resulting in obstruction or even retrograde flow into the ventricles, have all been noted with repeated pumping.51
Consensus is lacking on the role of diagnostic testing, including routine laboratory testing, CSF examination, neuroimaging with CT or MRI, and EEG. This is largely due to the dearth of well-designed prospective studies involving sufficient numbers of patients with specifically defined headache types. Such studies are sorely needed to address these issues. To complicate matters further, the information is even more limited for children and adolescents who present with acute headache in the ED.
The value of laboratory testing depends on its specific relationship to the patient's clinical condition. For example, a rapid strep antigen detection test could lead to identification of the cause for the fever, sore throat, and headache in a school-age patient, or a bedside blood glucose test may reveal hypoglycemia in a patient with headache and altered mental status. Consider other tests, such as serum lead, carbon monoxide, thyroid survey, and toxicology screen, as deemed clinically necessary. In addition, hypertensive children with headache require extensive workup, including electrolytes, urine analysis, and electrocardiogram.
Cerebrospinal Fluid Testing
Indications for LP include suspicion for meningitis (fever, stiff neck, vomiting, irritability, and/or lethargy in the younger age group), subarachnoid hemorrhage (SAH), and/or idiopathic intracranial hypertension. It is wise not to wait for meningeal signs to consider meningitis, since these are in fact late signs in the pediatric population. LP is contraindicated in patients with coagulopathy, overlying cellulitis, or substantial risk of herniation. Also consider omitting the LP in patients with CSF shunts and a suspected shunt infection.51 Instead, CSF should be obtained from the shunt by a neurosurgeon. Note that delayed LP after presumptive antibiotics, resuscitation, and stabilization in patients who appear too ill or unstable for an immediate LP is appropriate.
Measure CSF pressure in the lateral decubitus position with legs relaxed in extension, though it is difficult to ask the younger children to remain long enough in that position. Avoid the fetal position, since it can falsely elevate pressure. Send CSF for cell count, differential, glucose, protein measurements, and bacterial cultures. An LP for SAH will yield the best results if performed within at least 6 hours of the onset of headaches and is most sensitive with spectrophotometric determination of xanthochromia.52 Patients who do not have altered mental status, focal neurological exam findings, papilledema, or a clinical impression of impending herniation do not require a screening CT prior to LP. Although there is controversy about whether an LP is causative, or even temporally related, the fact is that herniation can occur in as many as 5% of children with meningitis (patients with bacterial meningitis (ie, those with purulent CSF) are particularly susceptible), and herniation can occur with a normal head CT scan, also.53-57 Traumatic lumbar punctures are all too common in acute pediatric practices worldwide and can lead to uncertainty about management. Mazor et al reported results from 57 patients with blood-stained CSF samples, with an emphasis on predicting patients without meningitis. Using the white blood cells/red blood cells ratio and the observed/predicted ratio seems to be helpful in determining who does not have a positive CSF culture. Unfortunately, the authors failed to point out an underlying condition found both in the literature and in their own data. Patients with a positive culture from blood CSF may not necessarily have meningitis; rather, they may have bacteremia. It would be reassuring to know whether the peripheral blood culture was negative (assuming an adequate sample was drawn) before determining whether a positive CSF culture represented meningitis.58 On the other hand, treatment of a few patients without meningitis is acceptable, provided that no cases of meningitis were missed.
Berley et al conducted a prospective study of invasive bacterial infections in children at a rural Kenyan district hospital.59 A total of 515 out of 3668 (14%) LPs performed in children > 1 month of age yielded a blood-stained CSF (erythrocytes 500 cells/μl). In 324 of these (63%), a leukocyte count was possible, and in 191 (37%), the CSF was too bloody to count the cells. The simple CSF leukocyte count was found to be as predictive of culture-proven meningitis as either of the other ratios described in blood-stained CSF samples. In addition, the CSF/blood glucose ratio was found to be highly effective in predicting culture-proven bacterial meningitis, irrespective of blood staining. Among confirmed meningitis cases, a CSF leukocyte count of < 10 x 106/L occurred in 4 of 116 (3%) of those with non–bloodstained CSF, and 1 of 18 (6%) of those with blood-stained CSF where leukocyte counting was possible. These cases would not have been detected by use of either the CSF leukocyte/erythrocyte ratio or the observed/predicted CSF leukocyte ratio. There were 2 out of the 5 who had a CSF/blood glucose ratio < 0.1, including the one with blood-stained CSF. The sensitivity of the combination of tests was 98%, and the specificity was 90%. Of those with uncountable leukocytes, a CSF/blood glucose ratio < 0.1 had a sensitivity of 67% (4 of 6) and a specificity of 99% (152 of 154). (Note that 31 were incomplete, including 2 with meningitis).59 Where doubt exists or the sample is too blood-stained to count leukocytes, consideration should be given to beginning treatment and repeating the LP after 24-48 hours.60
A review of the literature disclosed only a single class III study of 104 children who were being evaluated by a pediatric neurologist. Laboratory studies, including complete blood count, electrolyte levels, liver function profiles, and urinalysis, were performed by the referring pediatrician.61 The laboratory studies were described as "uniformly unrevealing," but the number of patients studied and specific laboratory data were not described. No other reports investigating the role of laboratory studies in the evaluation of recurrent headache in children or adolescents have been published. Therefore, there is inadequate documentation in the literature to support any recommendation as to the value of routine laboratory studies or the performance of routine lumbar puncture in the evaluation of recurrent headache in children. In addition, the AAN has published a parameter on diagnostic and therapeutic indications for performing LP in adults and children, and recurrent headache was not included as an indication.62
The MRI is more sensitive than CT for detecting congenital anomalies, white matter abnormalities, and posterior fossa lesions (the most common location for brain tumors in children). Despite this, the noncontrast CT scan is still the most widely available and useful neuroimaging test available to emergency physicians. In patients with histories suggestive of "vascular events," magnetic resonance arteriography should be performed at the same time as the MRI.63 In patients with suspected venous obstruction, MRI plus magnetic resonance venography can be diagnostic. The indications for obtaining a CT for pediatric patients seen in the ED with headache are listed in Table 9.
Data are available from 6 pediatric studies to address whether the AAN parameters (published in 1994 and 2003) regarding the evaluation of headache in adults with normal neurological exams and the Headache Consortium guidelines are also applicable to children with recurrent headaches.64-66
The recommendations were against routine neuroimaging in patients with recurrent migraine headaches, provided there were no recent changes in pattern, history of seizures, or other focal neurological signs or symptoms. If any of these features were present, such studies might be indicated. The pediatric studies assessing neuroimaging use in children with recurrent headache — with one study reporting EEG data — included 1 class II and 5 class III studies, in which 605 of 1275 children with recurrent headaches who underwent neuroimaging were reviewed.67-72
The patients were collected from different populations, with 5 studies using clinic-based populations67,69-72 and 1 using only children referred for neuroimaging.68 Of these, only 1 specifically focused on clinical subsets (eg, migraine and chronic daily headache); the rest were from mixed populations of headache subtypes. For the entire group of children, the types of headaches included migraine (62%), tension (22%), mixed type (2%), posttraumatic (2%), seizure-related (1%), tumor (1%), psychogenic (<1%), and other and unclassified (11%). None of these studies were conducted on patients seen in the ED. Data on the total 605 of 1275 children from the combined studies of children with recurrent headache who had been examined by a neurologist and who underwent neuroimaging found that only 14 (2.3%) had nervous system lesions requiring surgical treatment. All 14 children had definite abnormalities on examination. No patient with a normal examination had a lesion that required surgical treatment.68,69
CT scans were performed in 116 of these children, MRI in 483, and both modalities in 75. Those not imaged were followed clinically, and no long-term problems were found for the 1- to 2-year follow-up time period reported in several of these studies. Imaging abnormalities were found in 97 children (16%). In 79 of these children, the abnormalities were considered to be incidental — for example, a nonsurgical lesion or one that did not require specific medical management. Nonsurgical abnormalities included: Chiari malformation (n = 24), arachnoid cyst without mass effect (n = 13), paranasal sinus disease (n =13), occult vascular malformations (n = 5), pineal cyst (n = 2), plus a variety of incidental structural abnormalities in 22 (ie, cavum septi, pineal cysts, ventricular asymmetry, and "hyperintense" lesions). Eighteen children (3.0%) had a surgically treatable lesion (n = 14) or a lesion (n = 4) that required medical treatment (eg, pituitary adenoma that resolved spontaneously). Ten children had tumors (2 medulloblastomas; 2 cerebellar astrocytomas; one each of choroid plexus papilloma, sarcoma, primitive neuroectodermal tumor, glioblastoma multiforme, brain stem glioma, and craniopharyngioma). Symptomatic vascular malformations were found in 3 children, and an arachnoid cyst that necessitated surgery was found in 1 patient. Critically, in all 14 children with CT- or MRI-detected lesions considered surgically treatable, abnormalities were described on neurological examination and included papilledema, abnormal eye movements (including nystagmus), and motor or gait dysfunction. In a class III study that accounted for most of the surgical cases, the authors also performed univariate analysis on the 28 children who had surgical and nonsurgical space-occupying lesions. In this study, all patients were examined by a neurologist and 5 predictive variables were determined that helped distinguish patients with space-occupying lesions from those without such lesions.68 In another class III study, 79 of 137 children examined by a child neurologist were scanned, and in those with normal neurological examinations, no surgically remediable lesions were found.72 See Table 10 for variables that predicted the presence of a space-occupying lesion.
Data from 8 studies assessing EEG use in 1148 children with recurrent headaches showed that the EEG was not necessary for distinguishing a diagnosis of primary headache disorder in children from a diagnosis of secondary headache caused by structural disease involving the head and neck, or those of a psychogenic etiology.73-80 Data from published studies on the use of EEG in the evaluation of recurrent headaches, particularly in children, are difficult to interpret.81 groups; lack of comparisons of the study population to Methodological problems included: mixed types of headaches in the patient population; poor definition of headache diagnostic criteria; multiple age age-matched control subjects; unclear definitions of EEG abnormalities; and EEG abnormalities previously considered abnormal in children are currently not considered pathologic.
Previous studies in children, as well as in adults, have suggested that the EEG in patients with migraine is more likely to be abnormal (particularly paroxysmal abnormalities) than EEGs in patients with other types of headaches.82 This has led to the use of EEG for diagnosing migraine, based on the assumption that this would lead to migrainespecific treatments. But the issue is murkier for pediatric patients, because the incidence of paroxysmal abnormalities detected by EEG in healthy children is greater than in adults.81
Pooled data from 219 children with migraine and 929 with all headache types shows no significant difference in EEG abnormalities (slowing, spike activity, or other abnormalities) between children with migraine and the "all headache" group.74-78 The reason for this lack of difference is likely due, at least in part, to the fact that 44% of patients in the "all headache" group were diagnosed with migraine. As previously noted, extraction of information about the patients with migraine compared with other groups was impossible in these studies. Even if some differences were found between these 2 groups (ie, migraine versus nonmigraine), there was no evidence that the EEG findings would be of sufficient specificity or sensitivity in an individual patient to be clinically useful. Therefore, to date, there are no studies that definitively compare the incidence of EEG abnormalities in migraine versus nonmigraine pediatric headache patients. Thus far, the data do not suggest that there are enough differences in EEGs between children with migraine compared with other recurrent headache types to make them diagnostically usefu in the individual patient for determining an etiology or to make a diagnosis of migraine. The diagnosis of migraine and other primary headache disorders is made on clinical grounds, for the most part, based on information gleaned from the history of the patient's symptoms and lack of findings on examination.
Though seizure-related headaches have been recognized in the past,83,84 they remain infrequently diagnosed, and there is lingering controversy as to whether such an entity even exists.81 Data are only available from a single class III study addressing this issue.73 In this study of 215 children, "seizure headaches" were diagnosed in 58 children (27%). A seizure headache was described as a "paroxysmal brief headache" accompanied by nausea, vomiting, or other autonomic signs, followed by postictal lethargy or sleep with "typical epileptiform discharge" on EEG recording. The authors do not define the "typical EEG" features, but describe 36 patients with partial, 3 with generalized, and 5 with multifocal seizures. The authors do not state when the EEG was performed in relation to the epoch of headache. These patients had a much higher incidence of abnormal EEGs that were paroxysmal (75.9%; n = 44) compared with other groups (migraine: 8.3%; psychogenic: < 1%; remaining groups did not show paroxysmal abnormalities). Out of 58 children, 11 had a previous history of seizures. Data from 1 class III study suggest the idea that children may have seizure-related headaches and that, in these children, the EEG is likely to be paroxysmal. The limited available literature indicates that this condition is infrequently diagnosed, and even its existence as a clinical entity is still questioned.