Fever:

February 1, 2004
Robert E. Morrison, MD
Robert E. Morrison, MD

,
James B. Lewis Jr, MD
James B. Lewis Jr, MD

A methodical approach to diagnosis usually reveals the cause of fever. In patients with simple fever, a careful history taking and physical examination combined with basic laboratory and imaging studies (complete blood cell count with differential, urinalysis, and possibly a chest film and blood cultures) usually yield the diagnosis. In patients with prolonged fever whose cause remains undiagnosed after extensive examination (fever of unknown origin), repeat the history taking and physical examination; also order routine laboratory studies, an HIV test, a tuberculin skin test, 3 sets of blood cultures, and chest films. In addition, abdominal CT scanning is often useful. Further testing at this point may include fluorodeoxyglucose positron emission tomography, technetium-tagged white blood cell scanning, transesophageal echocardiography, liver biopsy, bone marrow examination, and/or temporal artery biopsy. Exploratory laparotomy is rarely indicated.

Fever is one of the most commonly encountered symptoms in the practice of medicine (Box). Often its cause is immediately apparent. However, when it is not, the dozens of different diseases and conditions associated with fever can make diagnosis challenging.

Here we present a systematic approach to the workup. We also discuss how to proceed when an initial comprehensive examination has failed to yield a diagnosis and a fever is classified as a "fever of unknown origin" (FUO).

WORKUP OF A PATIENT WITH FEVER

Initiate evaluation of fever in any patient with an oral temperature of 38.3°C (101°F) or higher. In immunosuppressed or obviously septic patients, consider an evaluation at lower temperatures. The cause of a fever is usually discerned based on symptoms, clues from the history, physical findings, and/or results of basic laboratory and imaging studies. Fevers in ambulatory patients are usually caused by respiratory tract infections, urinary tract infections, or chronic viral infections (ie, those caused by HIV, Epstein-Barr virus, herpesvirus, or hepatitis virus type B or C).

History. A detailed history-with a complete review of systems-is essential. Ask about:

  • Previous procedures.
  • Travel.
  • Animal exposures.
  • Tuberculosis exposure.

Be sure to assess:

  • Height and duration of the fever.
  • Fever pattern.
  • Relationship of fever to pulse.

A shaking chill, or rigor, suggests bacteremia, respiratory distress, pneumonia, or sepsis and hypotension. In febrile patients with an intravenous catheter-or any type of catheter- assume that a catheter-related infection is the cause of fever until proved otherwise.

Height and duration of fever. Evaluate patients with sustained temperature elevations below 38.8°C (102°F) for malignancy, cirrhosis, viral hepatitis, tuberculosis, legionnaire disease, or 1 of the zoonoses. Fever can be associated with almost any malignancy; it does not appear to be limited to liver or brain metastases. Malignancies do not usually cause high-grade fever, with the exception of renal cell carcinoma (the most common solid tumor cause of FUO) and lymphomas. In lymphomas, a Pel-Ebstein fever (which disappears and reappears over a period of several days to several weeks) may occur.

Temperatures higher than 38.8°C (102°F) with no obvious cause mandate evaluation for an abdominal, GI, or pelvic abscess. Pancreatitis in patients with a temperature of 38.8°C (102°F) or higher is usually the result of a pancreatic abscess or an infected pancreatic pseudocyst. Extreme hyperpyrexia (temperature higher than 41.1°C [106°F]) is usually associated with a noninfectious cause, such as heat stroke or CNS disease.

Fever patterns. Fever curves usually have no diagnostic significance except in uncommon conditions. Hectic, septic fever (fever that spikes several times a day)in hospitalized patients is usually associated with peritonitis; in-tra-abdominal, pelvic, or renal abscesses; or over-zealous use of antipyretics.

A single temperature spike below 38.8°C (102°F) very rarely has diagnostic significance-unless it occurs in an immunocompromised patient. Single temperature spikes above 38.8°C (102°F) usually result from manipulative or invasive procedures that induce transient bacteremias or from the infusion of blood products.

Double-quotidian fevers (fevers that spike twice in a 24-hour period) that are not related to antipyretic use are uncommon. However, such a fe-ver may be the only clue that points to adult-onset Still disease, juvenile rheumatoid arthritis, or culture-negative gonococcal endocarditis.

Fever-pulse relationship.Relative bradycardia (a pulse-temperature deficit) is associated with typhoid fever, legionnaire disease, chlamydial pneumonia, and possibly dengue fever and sandfly fever. Unfortunately, it is not possible to correlate fever and relative bradycardia with a specific diagnosis in any given patient.1,2

Physical examination. Examine the skin for dermatologic manifestations of infection:

  • Petechiae, associated with meningococcemia (Figure) and Rocky Mountain spotted fever.
  • Pustules, seen in gonococcemia and staphylococcal disease.
  • Conjunctival petechiae, splinter hemorrhages, Osler nodes, and Janeway lesions-any of which may be seen in endocarditis.
  • Ecthyma gangrenosum, which is associated with Pseudomonas aeruginosa infections.
  • Extensive bullae over the extremities, which may occur in Vibrio vulnificus infections.

Percuss the paranasal sinuses for tenderness, which indicates sinusitis. Examine the optic fundi and posterior chambers of the eye for Roth spots, retinitis, and choroiditis. Any of these may be seen in endocarditis or in viral, mycobacterial, or fungal diseases. Inspect the tympanic membranes for evidence of otitis media.

Palpate the thyroid gland. A tender gland in the setting of a sore throat and fever may represent subacute thyroiditis.

Carefully palpate the cervical, axillary, supraclavicular, inguinal, and epitrochlear lymph nodes. Nodes larger than 1 cm in diameter are often pathologic; consider lymph node aspiration in patients with HIV infection or a previous history of malignancy. Lymph node biopsy is useful in the diagnosis of lymphoma, tuberculosis, and a variety of other infections.

Examine the breasts, chest, and heart. Patients with heart disease may respond to fever with high-output angina or even heart failure. Crackles may indicate pneumonia, and dullness can be a sign of pleural effusion.

Perform a musculoskeletal examination that includes careful examination of the joints. Tenderness over the spine on percussion may be associated with infection or malignancy.

During the neurologic examination, be alert for meningismus and for changes in mentation. In older patients, fever may cause inability to concentrate, confusion, delirium, or stupor.

Laboratory and imaging studies. Basic laboratory tests for a patient with fever consist of a complete blood cell (CBC) count with differential and urinalysis. Also obtain a chest radiograph if signs or symptoms of respiratory disease are present. Order blood cultures in hospitalized patients with a temperature of 38.3°C (101°F) or higher and in any patient with fever in whom you suspect a serious bacterial infection, such as pneumonia, osteomyelitis, abscess, or meningitis. Two sets of blood cultures are indicated when the likelihood of bacteremia is low. Three sets of cultures are indicated when a continuous bacteremia is suspected, as in infective endocarditis.

FEVER OF UNKNOWN ORIGIN

Petersdorf and Beeson3 first defined FUO. Based on their 5-year study of 100 patients, they established the following criteria for FUO:

  • Illness of more than 3 weeks' duration.
  • Temperature higher than 38.3°C (101°F) on several occasions.
  • Diagnosis still uncertain after 1 week of study in the hospital.

During the past 40 years, a number of modifications to this definition have been proposed. More recent studies have suggested that the hospital stay might be replaced by a 3-day intensive inpatient evaluation or 3 outpatient evaluations.4 Because the label "FUO" is often cavalierly applied to even very brief fevers that have no apparent cause, Bryan5 has argued for use of the term "prolonged FUO," which he defines as fever that persists for 3 or more weeks.

Durack and Street4 have subdivided FUOs into classic, neutropenic, nosocomial, and HIV-related. Characteristics of the last 3 types are:

  • Neutropenic FUO: fever episodes in which the temperature is 38.3°C (101°F) or higher, neutrophil counts or anticipated counts of less than 500/µL, and an uncertain diagnosis after 3 days' investigation.
  • Nosocomial FUO: no infection on admission, temperature higher than 38.3°C (101°F) on several occasions during a hospital stay, and an uncertain diagnosis after 3 days' investigation.
  • HIV-associated FUO: positive HIV serology, fever episodes in which the temperature is 38.3°C (101°F) or higher, fever lasting 4 weeks or more in an outpatient setting or 3 days or more in an inpatient setting, and an uncertain diagnosis after 3 days' investigation.

Classic FUOs-which we discuss here-are those that meet the criteria described by Petersdorf and Beeson (or the slightly modified criteria that have subsequently been proposed).

Causes of FUO. Prolonged fe-ver that constitutes a true FUO can be caused by any of 150 or more diseases (Tables 1 and 2). These can be grouped into 4 broad categories:

  • Infections.
  • Neoplasms.
  • Noninfectious inflammatory diseases.
  • Miscellaneous illnesses.

Studies that attempt to quantitate the various causes of FUO are influenced by geographic factors, referral patterns, the state of diagnostic testing at the time of the study, the experience of the investigators, and the age of the patients studied. In a review of data on FUOs from the 1950s to the 1990s, Mourad and colleagues6 found the following distribution of causes: infections (28%), inflammatory diseases (21%), malignancies (17%), miscellaneous (15%), and undiagnosed (19%). The most recent series from 2003 breaks down the causes of diagnosed FUOs into infections (30%), neoplasms (15%), noninfectious inflammatory diseases (35%), and miscellaneous (20%).7

Infections. Splenic abscesses can cause clinically silent bacteremia. Cytomegalovirus infections may cause nonspecific constitutional symptoms with generalized lymphadenopathy, mild hepatosplenomegaly, and fever. Osteomyelitis of the spine with tuberculosis (Pott disease) or other bacterial infection may cause low-grade FUO accompanied by weakness and back pain.

Neoplasms. Malignancies of reticuloendothelial origin (leukemia and lymphoma) may cause FUO. Renal cell carcinoma frequently causes FUO-sometimes without accompanying hematuria. Atrial myxomas are histologically benign neoplasms that often cause FUO.

Noninfectious inflammatory diseases. In this category, Still disease, giant cell arteritis, and inflammatory bowel disease are among the more common causes of FUO.

Miscellaneous causes. These include noninflammatory illnesses such as pulmonary embolism, Addison disease, deep venous thrombosis, cholesterol emboli, and drug-induced fever.

Yesterday's FUO is today's instantaneous diagnosis. For example, sinusitis was a common cause of FUO in patients in ICUs who had been nasally intubated; now it can be quickly diagnosed by CT scan followed by nasal puncture.8 Because of the widespread use of sophisticated diagnostic techniques, the number of patients whose illness is initially categorized as FUO is shrinking (instead, they receive a prompt diagnosis).7 However, because many patients who formerly were first categorized as having an FUO eventually received a diagnosis, the percentage of patients with true FUO is rising.

Evaluation. A methodical approach to diagnosis is often successful.This consists of a careful history taking and physical examination, routine initial laboratory and imaging studies, abdominal CT, and (if needed) more invasive diagnostic testing guided by clues already uncovered. An evidence-based approach to FUO is detailed in the Algorithm.

History and physical examination. The same approach used initially in patients with fever applies in FUO. Do not skip questions or parts of the examination because "they were covered before." It is useful to repeat the basic history taking and physical examination; these will sometimes yield a clue the second time around that leads to a diagnosis.

If any potential causative medications are discovered in the history, stop them for 72 hours to see if defervescence occurs. Drug-induced fever that is not accompanied by severe rash or evidence of vasculitis usually disappears 72 to 96 hours after the offending medication has been stopped. (However, drug-induced fever caused by anticonvulsants may not disappear this quickly.) Rash, eosinophilia, and a relative bradycardia can also help in the diagnosis of drug-induced fever, but they are seen in a minority of cases.9

Initial laboratory and imaging studies. These include a CBC count, urinalysis and culture, liver and renal function tests, an HIV antibody test, measurement of erythrocyte sedimentation rate (ESR), a tuberculin skin test and control, 3 sets of blood cultures, and posteroanterior and lateral chest radiographs. Some experts might also include measurement of antinuclear antibodies and rheumatoid factor.

More advanced imaging studies. Ultrasonography and CT have more than a 90% sensitivity for hepatobiliary disease, abdominal abscesses, and kidney tumors. Because recent research has encouraged prompt use of noninvasive studies (such as ultrasonography, MRI, and CT), it is unlikely that an intra-abdominal abscess will remain undiagnosed after a week of study. In the 1980s, a prospective French study noted that in 50% of patients with FUO, a diagnosis would have been established earlier if abdominal CT scanning had been routinely performed.10 A 1992 community hospital-based study showed that in patients with FUO, a percutaneous CT-guided procedure yielded diagnostic results in 77% of those with infections and 23% of those with tumors.11

The newest imaging study used for evaluation of FUO is [18F]fluorodeoxyglucose positron emission to- mography (FDG-PET). In a 2001 study, FDG-PET contributed useful diagnostic information in 41% of patients with FUO-which compares favorably with the results achieved with gallium scanning in the same population.12

The role of radionuclide scanning in FUO is controversial because of the high false-positive rate and nonspecificity of findings. However, a technetium-tagged white blood cell scan may be appropriate if initial imaging studies are negative.

If endocarditis is strongly suspected, order a transesophageal echocardiogram and hold cultures for up to 3 weeks to detect the HACEK (Haemophilus,Actinobacillus,Cardiobacterium,Eikenella, and Kingella) group of organisms.13

A recently published evidence-based review of FUO suggests the use of lower extremity venous ultrasonography to detect deep venous thrombosis in puzzling cases.6

Invasive diagnostic procedures. These are less commonly used now to evaluate an FUO than they were in the past. Exploratory laparotomy, which was often performed several decades ago, is rarely indicated today. However, consider a temporal artery biopsy in an elderly patient with FUO and an elevated ESR when the initial workup is negative for infection.14 Bone marrow aspiration and biopsy are particularly useful in patients with FUO and abnormalities in the CBC count. Consider a liver biopsy when the results of liver function tests are abnormal-even if an abdominal CT scan is normal. If the patient's condition is deteriorating, a recent evidence-based review of FUO advocates liver biopsy-regardless of the physical findings or the liver profile.6

Prognosis of undiagnosed FUO. In 5% to 20% of patients with FUO, even a thorough evaluation proves nondiagnostic. However, if such patients remain stable, the prognosis is usually excellent, with eventual defervescence and no subsequent emergence of serious disease. A therapeutic trial of corticosteroids or antibiotics is generally not indicated.

References:

REFERENCES:

1. Ostergaard L, Huniche B, Anderson PL. Relative bradycardia in infectious diseases. J Infect. 1996;33: 185-191.

2. Wittesjo B, Bjornham A, Eitrem R. Relative bradycardia in infectious diseases. J Infect. 1999;39:246-247.

3. Petersdorf RG, Beeson PB. Fever of unexplained origin: report on 100 cases. Medicine(Baltimore). 1961;40:1-30.

4. Durack DT, Street AC. Fever of unknown origin-reexamined and redefined. In: Remington JS, Swartz MN, eds. Current Clinical Topics in Infectious Diseases. Boston: Blackwell Science; 1991:17-52.

5. Bryan CS. Fever of unknown origin. The evolving definition. Arch Intern Med. 2003;163:1003-1004.

6. Mourad O, Palda V, Detsky AS. A comprehensive evidence-based approach to fever of unknown origin. Arch Intern Med. 2003;163:545-551.

7. Vanderschueren S, Knockaert D, Adriaenssens T, et al. From prolonged febrile illness to fever of unknown origin. The challenge continues. Arch Intern Med. 2003;163:1033-1041.

8. Marik PE. Fever in the ICU. Chest. 2000;117:855-869.

9. McDonald M, Sexton DJ. Drug fever. UpToDate. 2003;11(2).

10. Lebas J, Pariente D, Bletry O, et al. Fièvres prolongées et syndromes inflammatories inexpliqués. Presse Med. 1985;14:577-581.

11. Kazanjian PH. Fever of unknown origin: review of 86 patients treated in community hospitals. Clin Infect Dis. 1992;15:968-973.

12. Blockmans D, Knockaert D, Maes A, et al. Clinical value of [18F]fluorodeoxyglucose positron emission tomography for patients with fever of unknown origin. Clin Infect Dis. 2001;32:191-196.

13. Bor DH. Etiologies of fever of unknown origin. UpToDate. 2003;11(2).

14. Hirschman JV. Fever of unknown origin in adults. Clin Infect Dis. 1997;24:291-302.

15. Mackowiak PA, Worden G. Carl Reinhold August Wunderlich and the evolution of clinical thermometry. Clin Infect Dis. 1994;18:458-467.

16. Mackowiak PA, Wasserman SS, Levine MM. A critical appraisal of 98.6°F, the upper limit of the normal body temperature and other legacies of Carl Reinhold August Wunderlich. JAMA. 1992;268: 1578-1580.

17. Arnow PM, Flaherty JP. Fever of unknown origin. Lancet. 1997;350:575-580.