Petersdorf and Beeson in 1961 proposed that the term fever of unknown origin (FUO) be reserved for persons with an illness persisting for 3 or more weeks and accompanied by temperatures higher than 38.4°C (101.2°F) on at least several occasions. They further specified that the cause of the fever should remain undetermined after at least 1 week of investigation in the hospital. Although this definition was arbitrary, it was useful at that time, when many of the diagnostic tests now in routine use were unknown. The purpose of their precise definition was to explore the cause of fever in this select group of adult patients and to permit comparison of data from different investigations. This exacting definition probably never was applied rigorously in pediatric practice. In children, the term fever of unknown origin should be reserved for fever of at least 8 days’ duration and for which no diagnosis is apparent after the initial workup in the hospital or as an outpatient.
Many investigators use the term fever without source (FWS) for fever of recent onset with no adequate explanation determined by the history or physical examination. The distinction between FUO and FWS is of more than academic interest for several reasons. First, although overlap exists, the differential diagnoses of these clinical conditions are distinct, and the most frequent causes of one can be different from the most frequent causes of the other. Second, a child with fever of recent onset generally warrants more immediate evaluation than does a child with FUO. The latter usually does not occur as an emergency and requires timely, but not urgent, diagnostic or therapeutic intervention. Third, although expectant antibiotic treatment of children with FUO generally is not indicated, expectant treatment of infants with FWS is recommended in certain cases.
Fever Without Source
A convenient definition of FWS is the occurrence of fever for 1 week or less in a child in whom a careful history and physical examination fail to reveal a probable cause of the fever. An estimated 14% of children with fever have no localizing signs or symptoms. Most children with fever of recent onset have acute infectious diseases, the majority of which are self-limited. A few of these patients have serious acute infectious diseases, including meningitis, bacteremia, and urinary tract infection, and a very few have acute noninfectious diseases or chronic disorders. For example, an occasional patient with FWS is discovered to have a disorder such as heat illness, drug poisoning, Kawasaki disease, malignancy, or connective tissue disease. However, these disorders occur infrequently. A physician faced with a child with FWS should consider the possibility of a noninfectious cause or the onset of a chronic disease, but unless a clinical clue suggests one of these entities, investigation in this direction is not warranted.
Many children with FWS are in the prodromal stages of an acute infectious illness, and evidence of a specific infection, such as pharyngitis, otitis media, or pneumonia, develops within hours to days of first being evaluated by a physician. Fever can precede the appearance of specific signs and symptoms by as long as 3 days, as in measles, Rocky Mountain spotted fever, and leptospirosis. In some infections, such as roseola, viral hepatitis, infectious mononucleosis, typhus, and typhoid fever, the interval between the onset of fever and the appearance of specific findings often is more than 3 days.
Occult Bacteremia
One concern regarding a young child with FWS is the possibility that the child has occult bacteremia. The patient does not appear ill, is judged clinically well enough to be managed as an outpatient, and does not have an infection commonly associated with bacteremia, such as pneumonia, but the blood culture yields pathogenic bacteria such as Streptococcus pneumoniae, Neisseria meningitidis, Haemophilus influenzae, Escherichia coli, Salmonella, group A streptococcus, or Staphylococcus aureus . Before introduction of the pneumococcal conjugate vaccines, the incidence of occult bacteremia in children with FWS was approximately 3% to 5%. More recent data report an incidence of less than 1%, although this figure may vary depending on vaccine coverage rates in the population studied. A large prospective study of 3571 febrile children younger than 3 years of age found that no patients who had received at least one vaccination of heptavalent pneumococcal vaccine had pneumococcal bacteremia. With the introduction of the 13-valent pneumococcal conjugate vaccine, further decreases in occult bacteremia caused by pneumococci are anticipated.
Historically occult bacteremia was found to occur more commonly in children with FWS than in febrile children of the same age with infections such as pharyngitis, otitis media, or upper respiratory tract infection. In the 1970s, investigators reported the incidence of bacteremia in febrile children without an obvious source of infection to be 9.9%, as opposed to 3.3% in children with otitis media, upper respiratory tract infection, or a flulike syndrome. During the same era, Teele and colleagues found a 3.9% incidence of bacteremia in children with FWS and a 1.5% incidence in comparably febrile children with otitis media or pharyngitis.
The risk for occult bacteremia developing in a child with FWS is age related, with the greatest risk occurring in the first few months of life. Numerous studies have demonstrated a higher risk for bacteremia or other serious bacterial infections in febrile infants younger than 3 months old. In a group of children with clinically unsuspected meningococcemia, all 12 patients who initially looked well enough to be treated as outpatients were younger than 24 months of age. In the 1990s, Bonadio and colleagues found the incidence of positive bacterial cultures to be 12% in febrile infants younger than 4 weeks and 6% in those between 4 and 8 weeks of age. A more recent retrospective study of bacteremia in infants 1 week to 3 months of age found 2% of all blood cultures were positive for pathogens. The risk for having occult bacteremia and other serious bacterial infections increases with the severity of fever. In a prospective study in which blood was obtained for culture from all febrile children younger than 2 years seen in a walk-in clinic, no positive blood cultures were found in 44 children with FWS and rectal temperatures lower than 38.9°C (102°F), whereas five (3.9%) positive blood cultures were found in 129 children with FWS and rectal temperatures of 38.9°C (102°F) or higher. Other studies have reported similar findings. Several series of children with fevers of 41°C (105.8°F) or higher found a relatively high prevalence of bacteremia and other serious bacterial infections, especially meningitis and pneumonia. However, even in this group of children with very high fever, most of those older than 2 or 3 months of age who looked well did not have a serious bacterial infection.
The white blood cell (WBC) count has been studied extensively as a potential tool in the diagnosis of occult bacteremia. On the basis of a study of hospitalized children, Todd reported that the absolute number of polymorphonuclear leukocytes and the absolute number of nonsegmented polymorphonuclear leukocytes were more sensitive than was the total WBC count, the percentage of polymorphonuclear leukocytes, or the percentage of nonsegmented polymorphonuclear leukocytes. However, whether information based on hospitalized children—presumably all of whom had serious localized infections or looked ill enough to warrant hospitalization—can be applied to children with FWS who look well enough to be treated on an ambulatory basis is questionable.
Considerable debate continues over the usefulness of the WBC count in febrile children evaluated in the outpatient setting. McCarthy and associates concluded that a WBC count of 15,000/mm 3 or greater was helpful in identifying patients at greatest risk for the development of bacteremia. Dershewitz found a direct relationship between the total leukocyte count and the prevalence of bacteremia and stated that “knowledge of the count was a helpful but limited predictor of patients with positive blood cultures.” Other investigators reported that the incidence of bacteremia increased with an increased WBC count and that bacteremia most commonly occurred in patients with counts of 20,000/mm 3 or higher.
Other studies have examined the utility of the WBC count specifically in children with FWS who look well enough to be treated on an outpatient basis. One such study found a sensitivity of 1.0 and a positive predictive value of 0.11 for a total WBC count of 15,000/mm 3 . Using a WBC count of 20,000/mm 3 would have decreased the sensitivity to 0.4 while increasing the positive predictive value to only 0.13. In another series, the sensitivity for a WBC count of 15,000/mm 3 was 0.87, with a specificity of 0.73. Although a total WBC count of 15,000/mm 3 does not accurately predict which child is or is not bacteremic, it is helpful in dividing the population of children with FWS into high- and low-risk groups. However, the value of the 15,000/mm 3 level has been substantially reduced in the studies conducted since the pneumococcal conjugate vaccine was introduced.
Some investigators have found the erythrocyte sedimentation rate to be no more useful than the WBC count in predicting bacteremia in ambulatory febrile patients. Others have reported that the serum concentration of C-reactive protein may be more accurate than the complete blood count or erythrocyte sedimentation rate in distinguishing bacterial from viral infections. However, recent studies reported conflicting data regarding the utility of the C-reactive protein test in screening children 3 to 36 months old for occult bacteremia. An elevated serum procalcitonin level has been found by some investigators to be at least as sensitive and specific as C-reactive protein in predicting serious bacterial infection in children with fever and no localizing signs. A meta-analysis of 881 children from developed countries with FWS found that procalcitonin in addition to WBC count and C-reactive protein tests was more accurate for detecting early meningococcal disease than WBC and C-reactive protein testing alone.
Other hematologic findings that suggest bacteremia include thrombocytopenia, Döhle inclusion bodies, toxic granulations, and vacuolization of neutrophils. In one study, peripheral blood smears of children younger than 24 months with acute febrile illnesses were reviewed the following day by a single investigator to determine whether vacuolization and toxic granulations were present; when both abnormalities were present, the positive predictive value for bacteremia was 0.76. The presence of these findings should be considered when estimating the risk for bacteremia.
Several studies have examined the response to acetaminophen and found no difference in the rate of reduction of temperature or improvement in clinical appearance between bacteremic and nonbacteremic children. Mazur and associates, however, found that febrile children aged 2 months to 6 years who did not respond to a dose of acetaminophen by a reduction in temperature of at least 0.8°C in 2 hours had a statistically significant increased risk for having occult bacteremia in contrast to those who did respond.
The most important aspects of assessment of a febrile child are a careful history and physical examination. Laboratory data are secondary and should be ordered on the basis of the clinical assessment. By definition, a child with FWS has no localizing signs to explain the fever or indicate a site of infection. Many physicians suggest that a general impression can indicate whether the child has occult bacteremia. Some physicians have suggested that careful clinical judgment, based on extensive experience, can identify most, if not all, children with serious illnesses. McCarthy and colleagues, in a series of carefully designed studies, elucidated the variables of history and observation that were most useful in assessing febrile children. They found that observation of the variable playfulness had the strongest correlation with overall assessment. However, they observed that even an experienced attending pediatrician could identify only 57% of seriously ill children by initial impression before performing a full physical examination. Dershewitz found that private pediatricians were no more accurate than pediatric residents in identifying children with occult bacteremia and that, in the private office, pediatricians were no better at predicting bacteremia in familiar patients than in first-time patients.
In a study of 292 consecutive febrile children seen in an emergency department, Waskerwitz and Berkelhammer identified a subgroup of patients who had no localizing signs and who looked so well that they were predicted not to have bacteremia. The physicians were assisted in their assessment by a functional scale that gave 0 to 2 points for the child’s eating, drinking, sleeping, and play activities, with a best possible score of 8. The group of patients who had functional scores of 5 or greater, with no localized infection and predicted clinically not to have bacteremia, were free of bacteremia, whereas 14 of 202 patients with functional scores of 4 or less were bacteremic. In this study, the physicians were not able to identify which patients had bacteremia and which did not; rather, they were able to identify one subgroup at high risk for having bacteremia and another at very low risk. Teach and Fleisher found that although Yale Observation Scale scores were higher in patients with bacteremia than in those without, the difference was not clinically useful in detecting bacteremia in well-looking febrile children without a discernible focus of infection. The clinician’s overall assessment of the degree of illness of the child appears to be a valuable but not infallible tool in estimating the risk for occult bacteremia in children with FWS.
Clinical Management of Fever Without Source
The clinician should not be dogmatic about the management of children with FWS. One reasonable approach, based on a careful history, thorough physical examination, and overall clinical impression, is to classify these children as being at low or high risk for the presence of occult bacteremia and other serious bacterial infections. For the low-risk group, no laboratory investigation is required routinely. For the high-risk group, a complete blood count and blood culture should be obtained, especially in young infants and children who are incompletely immunized. Many studies have shown that the most common serious bacterial infection in children with FWS is urinary tract infection. Therefore, urinalysis and urine culture should be performed in febrile infants and children younger than 24 months of age. Lumbar puncture and chest radiography are considered on an individual basis. If the patient appears ill, admission to the hospital may be justified even if all test results are negative. When high-risk children look well enough to be sent home, they are reasonable candidates for expectant antibiotic therapy pending the outcome of blood and urine cultures. For patients clinically considered to be at moderate risk (not clearly high or low risk), the physician has the option of obtaining a WBC count and using the results to decide whether to obtain blood and urine for culture and prescribe antibiotics expectantly.
Table 63.1 lists risk factors for the development of occult bacteremia. Current information is not sufficient to warrant the use of scoring systems except as part of investigational series. In the final analysis, the clinician’s judgment, taking into account all available clinical and laboratory data about each patient, is the guide to selecting which children require a diagnostic workup and expectant therapy with antibiotics.
| Factor | High Risk | Low Risk |
|---|---|---|
| Age | ≤3 months | >3 months |
| Immunization status a | Incomplete | Complete |
| Magnitude of fever | ≥40°C (104°F) | ≤39.4°C (103°F) |
| White blood cell count | ≥15,000/mm 3 | <15,000/mm 3 |
| Peripheral blood smear | Toxic granulation or vacuolization of polymorphonuclear leukocytes, thrombocytopenia | Unremarkable |
| Underlying chronic disorder | Sickle-cell disease, immunodeficiency, malnutrition | None |
| History of contact with bacterial disease | Contact with Neisseria meningitidis or Haemophilus influenzae | None |
| Clinical appearance | Appears ill, toxic, or unhappy; inconsolable; irritable or lethargic; not eating or drinking enough | Looks well, playful, eating normally, not irritable |
a Vaccination status is incomplete if the child has received fewer than three doses of Haemophilus influenzae vaccine and fewer than three doses of conjugated pneumococcal vaccine and is complete if the child received at least three doses of each vaccine.
If the physician elects to prescribe antibiotics while awaiting the results of blood culture, such antibiotic therapy should provide adequate coverage for S. pneumoniae, N. meningitidis, and H. influenzae, although the frequency of H. influenzae and S. pneumoniae has decreased dramatically with current immunization practice, and N. meningitidis is uncommon. A single injection of ceftriaxone 50 to 75 mg/kg given while awaiting the results of blood culture has been successful in resolving fever, clearing bacteremia, and preventing meningitis and was found to be superior to oral regimens in several series. Children with a positive blood culture should be recalled for re-evaluation even if they are afebrile. Children who have been immunized against both H. influenzae and S. pneumoniae should be considered to be at relatively low risk for the development of occult bacteremia and may require less workup unless they appear ill or have a very high fever.
Infants younger than 90 days pose a special problem because they have an increased risk for developing a serious bacterial infection, clinical evaluation is more difficult, and a broader spectrum of invading organisms (e.g., group B streptococcus, E. coli, and Listeria monocytogenes ) exists. Baker and coworkers showed the safety of managing selected low-risk infants (i.e., normal WBC count, urinalysis, lumbar puncture, and chest radiograph, and, if diarrhea was present, negative smear for fecal leukocytes) 30 to 90 days of age on an outpatient basis without antibiotics. Jaskiewicz and coworkers found the Rochester criteria (WBC count of 5000 to 15,000/mm 3 , band count <1500/mm 3 , spun urine specimen <10 WBCs/high-power field, stool specimen [if diarrhea] <5 WBCs/high-power field) to have a 98.9% negative predictive value in well-appearing, previously healthy infants younger than 90 days with no focal infections. One reasonable practice guideline for managing infants with FWS is to hospitalize and treat all who appear toxic and all younger than 28 to 30 days. Those between 30 and 90 days of age can be managed as outpatients if they look well and the blood count, urinalysis, and cerebrospinal fluid analysis (if empiric antibiotics are to be prescribed) are within normal limits.
Fever of Unknown Origin
The exact definition of FUO is a subject of considerable disagreement, and series in the pediatric literature differ in their criteria for inclusion. Brewis defined FUO in children as a temperature of 38.3°C (101°F) or higher for 5 to 7 consecutive days without localizing signs or symptoms. In sharp contrast, McClung and Lohr and Hendley considered children with fever for at least 3 weeks on an outpatient basis or 1 week in the hospital to have FUO. Pizzo and associates, however, required only that the fever be present for 2 weeks, with no distinction made between outpatient or in-hospital status. A reasonable working definition of FUO for clinical purposes is the presence of fever for 8 or more days in a child for whom a careful and thorough history, physical examination, and preliminary laboratory data fail to reveal a probable cause of the fever.
Most cases of FUO in children are caused by relatively common diseases. In four series of FUO totaling 418 children, only five patients would be considered to have rare disorders (i.e., Behçet syndrome, ichthyosis, variant of “blue diaper” syndrome, diencephalic seizure disorder, and “possible chronic lead and/or arsenic intoxication”). The adage that FUO is more likely to be caused by an unusual manifestation of a common disorder than by a common manifestation of a rare disorder certainly is true in pediatrics. The three most common discernible causes of FUO in children, in order of decreasing frequency, are infectious diseases, connective tissue diseases, and neoplasms. In approximately 10% to 20% of cases, a definitive diagnosis never is established.
In the United States, the systemic infectious diseases diagnosed most frequently in children with FUO include tuberculosis, brucellosis, tularemia, salmonellosis, cat-scratch fever and infections caused by rickettsia, spirochetes (e.g., leptospirosis), Epstein-Barr virus, cytomegalic inclusion virus, human immunodeficiency virus (HIV), hepatitis viruses, and other viruses. The most common causes of localized infection are upper respiratory tract infections (e.g., sinusitis, otitis, tonsillitis), urinary tract infection, osteomyelitis, and occult abscesses, including hepatic and pelvic abscesses.
The connective tissue disease most commonly manifested as FUO in children is juvenile idiopathic arthritis, which accounts for more than 90% of connective tissue diseases in most series, followed by systemic lupus erythematosus and then by undefined vasculitis. Frequently a definitive diagnosis of juvenile idiopathic arthritis can be made only after an extended period of observation because physical examination may yield few findings and the results of specific serologic studies generally are normal or negative.
Malignancy is a less frequent cause of FUO in children than in adults and usually is the third-largest group, after infectious diseases and connective tissue diseases. Malignancy accounted for 7% of FUO cases in the series reported by Pizzo and associates and for 13% in the Lohr and Hendley series. Leukemia and lymphoma are responsible for most cases of cancer manifested as FUO in children. Other tumors less commonly reported as causing FUO include neuroblastoma, hepatoma, sarcoma, and atrial myxoma.
The prognosis for children with FUO is better than that for adults, and most children with FUO have treatable or self-limited disease. Historically mortality rates for children with FUO were 6% to 9%. However, more contemporary studies indicate that mortality is quite low.
Diagnostic Approach to a Child With Fever of Unknown Origin
A child with FUO may be admitted to the hospital for more than simply laboratory investigation. Hospitalization provides an opportunity to observe the child, repeat the history and physical examination, analyze all available data, and investigate every potential diagnostic lead. In the Lohr and Hendley series of 54 children with FUO, an incomplete history delayed establishing the diagnosis in nine cases, and physical findings that were ignored delayed rendering the diagnosis in four cases. In McClung’s report of 99 pediatric cases of FUO, errors in the history or physical examination obscured the correct diagnosis for at least 10 patients. Failure to use existing laboratory data correctly is another common factor preventing early determination of the diagnosis in children with FUO.
Clinical Evaluation
The first and most important step in the diagnostic workup of a child with FUO is obtaining a complete and detailed history and conducting a physical examination. The clinical evaluation must be thorough and careful, and it must be repeated frequently. Often a patient or parent eventually recalls information that was omitted or forgotten when the initial history was obtained. Physical findings change, and abnormalities not originally present can appear subsequently. Lohr and Hendley noted that in more than 25% of children admitted to the hospital with FUO, significant physical findings developed that were not present at the time of admission.
A detailed history should be obtained regarding contact with infected or otherwise ill persons and any exposure to animals, including pets and wild animals. The number of children with zoonotic infections is increasing each year. History of a cat scratch or exposure to kittens may be a clue for Bartonella henselae . Immunization of domestic animals such as the dog against leptospirosis can prevent canine disease, but it does not prevent carriage, excretion, and transmission of this infection. A history of travel extending back to birth must be elicited. Reemergence of histoplasmosis, coccidioidomycosis, blastomycosis, leishmaniasis, or malaria years after visiting or living in an endemic area can occur. Inquiring about prophylactic immunizations, precautions taken against the ingestion of contaminated food or water, and malarial prophylaxis is important. Questioning should include the possibility that rocks, soil, or artifacts from geographically distant regions may have been brought into the home, as well as the possibility that contact with persons who have visited distant countries has occurred. Even contact with insects can be important. Tick bites can be a clue to Rocky Mountain spotted fever or tick-borne relapsing fever. North American mosquitoes and some ticks carry a variety of arboviruses.
The physician should determine whether the patient has eaten game meat, raw meat, or raw shellfish. A history of pica should be sought routinely. Ingestion of dirt can suggest a diagnosis of visceral larva migrans, toxoplasmosis, or other infectious diseases. A detailed history regarding all medications, including topical agents and nonprescription items, must be elicited. Any history of surgical procedures should be explored.
Questions designed to determine the genetic or ethnic background of the patient can reveal information that specifically suggests or largely excludes diagnoses such as nephrogenic diabetes insipidus (found in Ulster Scots), familial Mediterranean fever (found in Armenians, Arabs, and Sephardic Jews), familial dysautonomia (found in Jews), and Kikuchi-Fujimoto disease, a benign and self-limited histiocytic necrotizing lymphadenitis (found mostly in young Asian females and characterized by fever, lymphadenopathy, and malaise).
The history should be exacting regarding the duration, height, and pattern of the fever, as well as the circumstances under which temperature elevation occurs, whether the child appears ill or any signs or symptoms develop, and how well the fever responds to antipyretic drugs. A history of “fever” occurring only after exercise or late in the afternoon can indicate parental concern about normal variations in body temperature. A history of high fever occurring in the absence of malaise or other generalized signs can be a clue to factitious fever. The physician also should obtain a careful history regarding how well the fever has been documented. Has a thermometer been used, by whom, and in whose presence? A history of sweating and heat intolerance can indicate hyperthyroidism, whereas a history of heat intolerance with the absence of sweating can be a clue to ectodermal dysplasia.
Several investigators found that neither the pattern of fever nor its duration was useful in pointing to or establishing a diagnosis in children with FUO. However occasionally the character of the fever can be helpful. Intermittent fever is characterized by a return of temperature to normal at least once daily. If the peak of fever is high and the rate of defervescence quick, this pattern often is referred to as hectic or spiking . Intermittent fevers suggest pyogenic infections but also occur with tuberculosis, lymphoma, and juvenile idiopathic arthritis. In remittent fever, the temperature fluctuates but does not return to normal. A sustained fever pattern is characterized by persistent fever with little or no fluctuation and can occur in typhoid fever or typhus. Antipyretic agents can make a remittent or sustained fever appear intermittent. Relapsing fever refers to a pattern in which the patient is afebrile for 1 or more days between episodes of fever and can be seen with malaria, rat-bite fever, infection with Borrelia, and lymphoma. Recurrent episodes of fever of more than a year’s duration can suggest metabolic defects, central nervous system abnormalities in temperature control, and immunodeficient states.
The general activity and appearance of the patient should be observed, vital signs checked, and growth parameters measured. Weight loss is an important, though nonspecific, finding. Impairment of linear growth or short stature can be a clue to inflammatory bowel disease, an intracranial lesion involving the pituitary gland, or a long-standing chronic disease. Examining the patient during an episode of fever to observe the presence or absence of sweating, the effect of the fever on the heart and respiratory rate, the presence or absence of malaise or other symptoms, and the appearance of “toxicity” is helpful. The rash of juvenile idiopathic arthritis characteristically is evanescent and may be present only during periods of temperature elevation.
Some special aspects of the physical examination merit mention. Hypohidrosis, anomalous dentition, and sparse hair, particularly involving the eyebrows and eyelashes, suggest anhidrotic ectodermal dysplasia. Palpebral conjunctivitis can be a clue to the presence of infectious mononucleosis, Newcastle disease, or lupus erythematosus, whereas predominantly bulbar conjunctivitis can suggest leptospirosis or Kawasaki disease. Phlyctenular conjunctivitis can signal tuberculosis.
Absence of the pupillary constrictor response can be caused by a deficiency of the constrictor sphincter muscle of the eye. This muscle, derived from ectoderm rather than mesoderm, develops embryologically at the same time that hypothalamic structures and function are undergoing differentiation. Absence of this muscle can suggest that the elevation in temperature is the result of hypothalamic or autonomic dysfunction. Careful funduscopic examination can disclose evidence of miliary tuberculosis, vasculitis, or toxoplasmosis. Lack of tears, absence of corneal reflexes, and a smooth tongue with absence of the fungiform papillae suggest familial dysautonomia.
Purulent or persistent nasal discharge can be a sign of sinusitis. The physician should palpate for tenderness over the sinuses.
Hyperemia of the pharynx, even in the absence of exudate or specific symptoms, can be a clue to the diagnosis of infectious mononucleosis, cytomegalovirus, toxoplasmosis, tularemia, or leptospirosis. Gingival hypertrophy or inflammation and loosening or loss of teeth can indicate leukemia or Langerhans cell histiocytosis.
The bones and muscles should be palpated carefully. Tenderness over a bone can be found in cases of osteomyelitis or marrow invasion by neoplastic disease. The appreciation of a new heart murmur may result from infective endocarditis. Muscle tenderness can be associated with trichinosis, dermatomyositis, polyarteritis, or various arboviral infections.
The search for skin lesions and rash must be careful, extensive, and repeated. Petechiae can indicate endocarditis or other sources of bacteremia but also can occur with viral and rickettsial infections. A seborrheic rash can be a sign of histiocytosis. A small papule present for more than a week may be the inoculation site for cat-scratch disease.
A careful rectal examination can reveal pararectal tenderness or a mass indicative of a pelvic abscess or tumor. A test for occult blood should be performed on stool. Examination of the external genitalia should be completed on patients of all ages, and sexually active adolescent females should undergo a pelvic examination.
Laboratory Evaluation
The extent of laboratory investigation depends on the age of the patient, duration of the fever, and history and physical examination findings. Laboratory studies should be directed, as much as possible, toward the most likely diagnostic possibilities. The tempo of the diagnostic evaluation should be adjusted to the severity of the illness. In a critically ill child, speedy evaluation is important. If the patient is less severely ill, however, the evaluation can proceed more slowly; sometimes the fever can disappear without apparent explanation before a definitive diagnosis can be established and any invasive diagnostic procedures have been undertaken.
A complete blood count and careful examination of the peripheral smear are indicated for all patients. Anemia, thrombocytosis, and thrombocytopenia should be noted. Although mild or moderate changes in the total WBC or differential count usually are of no help, in some series, children with more than 10,000 polymorphonuclear leukocytes or 500 nonsegmented neutrophils/mm 3 were found to have a greater likelihood of having a serious bacterial infection. Atypical lymphocytes generally indicate viral infection, whereas bizarre or immature forms can suggest leukemia. Although the erythrocyte sedimentation rate and C-reactive protein are of no specific diagnostic value, they are a general indicator of inflammation and can help in ruling out factitious fever, determining the need for further evaluation, and monitoring the progress of the disease process.
Blood should be obtained from all patients for aerobic and anaerobic culture. In select cases, media appropriate for the isolation of Francisella organisms, Leptospira, and Spirillum also should be used.
Urine analysis and culture should be completed for all patients. In one series of FUO in children, failure to perform urinalysis and failure to investigate pyuria adequately were the most common laboratory errors. Radiographic study of the urinary tract, however, should be performed only when indicated.
All patients should undergo radiographic examination of the chest. Diagnostic imaging of the nasal sinuses, mastoids, and gastrointestinal tract is performed initially only for specific indications but may be done eventually in all children whose fever persists without explanation for a long period. Persistent fever and elevation of the erythrocyte sedimentation rate or C-reactive protein, with or without anemia, abdominal complaints, anorexia, and weight loss, are sufficient indications for radiographic study to evaluate for inflammatory bowel disease.
All patients should have an intradermal tuberculin skin test. Control skin tests with antigens such as Candida are of limited value because the anergy may be specific for tuberculosis rather than universal for all skin-testing materials. A positive control test result and negative tuberculin test result do not rule out tuberculosis.
Bone marrow examination is most useful in diagnosing cancer (especially leukemia), histiocytic disorders, and hemophagocytic disease. It is less useful in determining infection. Hayani and associates reviewed the results of 414 bone marrow examinations for FUO in children. In only one case was an organism ( Salmonella group D) recovered from the marrow that also was not recovered from blood or another source. Noninfectious causes of FUO were found in 8% of specimens: malignancy (6.7%), hemophagocytic lymphohistiocytosis (0.7%), histiocytosis (0.5%), and hypoplastic anemia (0.2%). In most of these cases, the diagnosis had been suspected clinically before the bone marrow was examined.
Patients should undergo a serum test for HIV infection. Other appropriate serologic tests can help establish a diagnosis of cat-scratch disease, brucellosis, tularemia, Epstein-Barr virus infection, cytomegalovirus infection, other viral infections, toxoplasmosis, and certain fungal infections.
Hepatic enzymes and serum chemistry, including electrolytes, urea nitrogen, and creatinine, should be determined in all patients. Serum antinuclear antibody should be measured in those older than 5 years if family history is significant or clinical suspicion warrants it. Serum hepatitis antigens, electrocardiography, electroencephalography, echocardiography, and stool culture and examination for ova and parasites generally should be performed in selected cases. Other tests to be considered for individual patients include ophthalmologic examination by slit lamp, radiographic bone survey, technetium bone scan, abdominal imaging by ultrasonography, computed tomography (CT), or magnetic resonance imaging (MRI). CT scanning and indium-111 scanning can detect inflammatory lesions and tumors. Such scanning procedures offer a relatively noninvasive technique for screening patients with FUO for a variety of disorders. Steele and associates found that radionucleotide scans seldom led to unsuspected diagnoses in children and suggested that they not be used indiscriminately. Gallium scanning has been helpful in diagnosing adult patients with FUO but is not generally recommended in children. Recent evidence suggests that (18)F fluorodeoxyglucose positron emission tomography (FDG PET) imaging may be useful in select adults and children with FUO in whom the etiology of fever remains elusive despite thorough physical examination, laboratory evaluation, and basic imaging techniques. FDG PET may be able to detect causes of FUO such as infection, tumor, and noninfectious inflammation because they all exhibit hypermetabolism of glucose. Identification of abnormalities using FDG PET could guide the next steps in clinical decision making to make the diagnosis. More invasive testing such as lymph node biopsy, liver biopsy, and exploratory laparoscopy are reserved for patients with evidence of involvement of these organs.
In general, antibiotics or other medications should not be administered empirically as a diagnostic measure in children with FUO. Exceptions include the use of nonsteroidal agents in children with presumed juvenile idiopathic arthritis and the use of antituberculosis drugs in critically ill children thought to have disseminated tuberculosis. Empiric trials of broad-spectrum antibiotics generally do more to obscure than illuminate the etiology of FUO and can mask or delay establishing the diagnosis of infections such as meningitis, parameningeal infection, endocarditis, or osteomyelitis.
Examples of disorders that can be manifested as FUO in children are listed in Box 63.1 . A few of these disorders are discussed briefly in the following sections.
