Fever of unknown origin

• Fever of unknown origin (FUO) is a febrile illness that persists for more than three weeks, reaches at least 38.3 degrees C on several occasions, and remains unexplained after a standard minimum set of investigations in an immunocompetent adult.
• The cause is usually not exotic: in most series infections account for 16% to 55% of cases, malignancy for 20% to 30%, inflammatory disease for 10% to 30%, and miscellaneous causes for 10% to 20%; up to 51% of cases are never given a specific diagnosis. ^[2][3]
• The single most useful strategy is to hunt for potential diagnostic clues from the history, examination, and basic tests, because one such clue leads to a diagnosis in only about 20% of cases but is still the best available approach. ^[1]
• Modern guidance supports using 18F-fluorodeoxyglucose positron emission tomography combined with computed tomography (FDG-PET/CT) early once FUO criteria are met, especially when no clues are present, and reserving empiric antimicrobial or anti-inflammatory therapy for the very few seriously ill patients in whom all other approaches have failed. ^[2]
• For most patients the outcome is good: up to three-quarters will have spontaneous resolution of fever and a favourable long-term course.

By the end of this page you should be able to:

1. State the classic, nosocomial, neutropenic and HIV-associated definitions of FUO and contrast their temperature, duration and investigation thresholds.
2. Classify the causes of FUO into the five modern categories and name common examples in each, with the typical regional and age-related differences.
3. Explain how fever itself is generated at the level of pyrogenic cytokines and prostaglandin E2, and why fever patterns alone are rarely diagnostic.
4. Take a structured history and perform a targeted examination that efficiently narrows the differential diagnosis of prolonged fever.
5. Outline the stepwise investigation of the immunocompetent adult with FUO, including the minimum standard tests, the role of potential diagnostic clues, and the place of PET-CT, cross-sectional imaging and biopsy.
6. Say when empiric treatment is justified and when it must be withheld.
7. Apply these principles in a low-resource setting.

FUO, also called pyrexia of unknown origin, is one of the few clinical problems in which the doctor is expected to work without the comfort of a localising symptom. The term is older than most practicing clinicians: it was coined in 1961 when Petersdorf and Beeson reported 100 patients whose fever exceeded 38.3 degrees C on several occasions, had lasted more than three weeks, and remained unexplained after at least one week of hospital investigation.^[1] The original purpose of the category was to single out a subset of prolonged fevers that resisted ordinary diagnosis so they could be studied; today the label is often applied too casually to any fever that has not yet been explained.^[4]

The condition is uncommon but not rare, and it carries real morbidity. Two retrospective cohorts from Japan and Italy that together included 9,830 hospitalized patients meeting classic FUO criteria reported a prevalence of 2.0 to 2.9%.^[1] Between 16% and 55% of classic FUO cases turn out to be infection, which means the possibility of a curable and occasionally life-threatening infection such as tuberculosis or infective endocarditis is always present. Equally important, much of the disability comes from the diagnostic process itself: without a standard evaluation, patients accumulate dozens of normal tests while the true cause is missed. Modern consensus work has therefore focused as much on standardising how these patients are assessed as on identifying the underlying disease.^[2]

Start with the normal. Fever is not simply a number above 37 degrees C; it is a regulated elevation of the hypothalamic set point produced by the innate immune system. In the best-studied example, bacterial lipopolysaccharide is recognized by Toll-like receptor 4 on macrophages and dendritic cells, which release pyrogenic cytokines including interleukin-1, interleukin-6 and tumour necrosis factor. Interleukin-6 drives endothelial production of prostaglandin E2 in the preoptic area, and prostaglandin E2 triggers vasoconstriction, brown fat thermogenesis and shivering. That biology matters for two reasons: it explains why rigors point toward infection, and it explains why suppressing fever with an antipyretic can obscure the fever curve without treating the cause.^[4]

The modern position is that the FUO label should be used only after an objective, standard minimum evaluation has been completed. The 1961 definition asked for more than one week of hospital study; Durack and Street's 1991 revision shortened this to three days in hospital or three outpatient visits; later work by de Kleijn introduced a minimum set of investigations performed in either setting. The 2024 Delphi consensus brought these strands together and recommended that classic FUO be defined as more than three weeks of fever without explanation, despite completing a set of minimal standard diagnostic tests, in an immunocompetent adult patient, with fever taken as more than 38.3 degrees C (100.9 degrees F) on more than three occasions.^[1][2] This is a higher and more precise bar than the simple statement that a fever has not yet been explained, and it protects patients from being labelled prematurely in settings where investigations are limited.^{[2] [4]}

Four categories of FUO are recognised, all built on the same temperature threshold of above 38.3 degrees C on several occasions but differing in the host context and the required intensity of investigation.^{[4] [1]}

A few points follow from the categories. The classic category includes only immunocompetent adults, and the definitions of immunocompromise used to exclude patients are explicit: neutropenia below 0.5 times 10 to the 9 per litre for at least one week within three months before the fever; receipt of immunosuppression for solid organ or haematopoietic stem cell transplantation; known hypogammaglobulinaemia; prednisone 10 mg or equivalent for two weeks or more within three months before the fever; uncontrolled HIV infection with CD4 below 200 cells per microlitre; or treatment with biologic therapies such as anti-tumour necrosis factor monoclonal antibodies.^[1] The neutropenic category in particular is a medical emergency rather than a diagnostic puzzle, because serious bacterial and invasive fungal infections are common and empiric therapy is expected. HIV-associated FUO in a patient who has newly progressed to AIDS is most often due to an underlying opportunistic infection or malignancy rather than to HIV itself. In a newly HIV-infected person, fever rarely lasts longer than three weeks and the illness mimics influenza.^[1][3]

The minimum standard diagnostic tests that must be completed before applying the classic label form the standard protocol recommended by the Delphi panel. Laboratory tests are a complete blood count with differential, a comprehensive metabolic panel including calcium and liver function tests, erythrocyte sedimentation rate, C-reactive protein, ferritin, thyroid-stimulating hormone, rheumatoid factor, antineutrophil cytoplasmic antibodies and antinuclear antibodies. Microbiology tests are HIV-1/2 serology, urinalysis with microscopy, a Mycobacterium tuberculosis skin test or interferon-gamma release assay, and at least three sets of blood cultures. Imaging must include abdominal ultrasonography together with either a posteroanterior and lateral chest radiograph or chest, abdominal and pelvic computed tomography. Where computed tomography of the chest, abdomen and pelvis is available, the Delphi panel preferred it over chest radiography and abdominal ultrasonography.^[2]

Common and uncommon cases alike are usually atypical presentations of common diseases, not rare ones. That single fact shapes the whole approach: the clinician should think about how an ordinary infection, tumour or inflammatory disorder can present without its usual localising features rather than hunt for an obscure zebra. Etiology also varies by geography and by patient age. A comparison of 19 prospective studies involving 2,667 patients found tuberculosis the most commonly reported infection within four of six World Health Organization regions, accounting for 34.3% of infections, with no studies available from Africa or the Americas. Southeast Asian series report a higher proportion of infections and European series a higher proportion of inflammatory disease than the overall figures. In one series of patients aged 60 years or more, infections and neoplasms together accounted for 72.6% of cases.^[1] In older adults generally, infections fall to second or third place as a cause. Noninfectious inflammatory disorders are significantly more common in high-income than in middle-income countries, and more common when the febrile illness has lasted longer and produced more symptoms.^[1]

The modern classification uses five disease categories, which the Delphi panel explicitly endorsed for both FUO and inflammation of unknown origin: infection, non-infectious inflammatory disorder, oncologic diseases, miscellaneous conditions and undiagnosed illness.^[1][2] Within these categories, the relative frequencies from published series are roughly as follows: infection 16% to 55% in recent series, and 20% to 40% in the classic American Family Physician grouping; malignancy 20% to 30%; non-infectious inflammatory disease 10% to 30%; miscellaneous 10% to 20%; and undiagnosed 7% to 51%, with the higher figure in recent European cohorts and lower figures in some older series. In a 2023 retrospective study spanning 21 countries of differing economic status and 788 participants, infection accounted for 51.6% of causes.^{[1][3] [4]}

Among 832 infections reported in a recent meta-analysis, the most common organisms and sources were Mycobacterium tuberculosis complex at 34.3%, brucellosis at 9.7%, infective endocarditis at 7.5%, abscess at 7.3%, herpesvirus infection including cytomegalovirus and Epstein-Barr virus at 7.2%, pneumonia at 6.5%, lower and upper urinary tract infection at 6.5% and enteric fever at 4.8%. Common infections always include extrapulmonary tuberculosis, culture-negative endocarditis, cytomegalovirus and Epstein-Barr virus; the uncommon list runs through bartonellosis, brucellosis, occult abscess, salmonellosis, and several arboviral and viral causes.^[1] A dental abscess can be the whole explanation, which is why dentition is examined carefully.^[3]

Malignancy causes 20% to 30% of adult FUO. Lymphoma is the single largest subgroup, accounting for 70.1% of malignancy-associated cases, followed by leukaemia at 25.2%, multiple myeloma at 3.4% and myelodysplastic syndrome at 1.3%. Within the broader neoplastic list are renal cell carcinoma, melanoma, colonic adenocarcinoma, gastric carcinoma, mesothelioma and Castleman disease. A markedly elevated lactate dehydrogenase suggests lymphoma or leukaemia, and an elevated ferritin level, though not specific, is seen with myeloproliferative disease.^[1][3]

Non-infectious inflammatory disease is the category that most often exploits the absence of localising signs. Adult-onset Still disease, systemic lupus erythematosus and polymyalgia rheumatica are the three common presentations of connective tissue disease, representing 58.5%, 26.7% and 5.6% respectively in one recent composite. Other frequent entities are giant cell arteritis (68.9% of vasculitis-associated FUO), polymyalgia rheumatica, sarcoidosis (61.9% of non-infectious granulomatous disease), Crohn disease, and the systemic vasculitides. Eastern Mediterranean cohorts record more systemic lupus erythematosus than other regions, and Kikuchi disease is seen more often in the Western Pacific and Southeast Asia.^[1][3]

Miscellaneous causes are the group most likely to be missed by an infection-focused workup: drug fever (18.4% of this category, hence the need for a complete medication history), thyroid disease (16.5%), habitual hyperthermia (10.7%), venous thromboembolism (2.9%), Addison disease (1.9%) and Dressler syndrome (1.9%). Drug fever is not exotic either; common culprits are antimicrobials, anticholinergics, urate-lowering agents such as allopurinol, non-steroidal anti-inflammatory drugs, antiarrhythmics, first-generation anticonvulsants and antidepressants, and the mechanism can be hypersensitivity, altered thermoregulation or pharmacologic action. Other miscellaneous causes include subacute thyroiditis, pulmonary embolism, haematomas, necrotizing lymphadenitis, periodic (genetic) fevers and haemophagocytic lymphohistiocytosis. Factitious fever belongs here, has been reported in up to 9% of cases, and is suspected with temperatures over 41.1 degrees C, discrepancy between simultaneous oral and rectal readings, loss of diurnal variation, rapid defervescence, absence of tachycardia and a history suggestive of Munchausen syndrome; it is more common in medical personnel and when the febrile illness has lasted longer than six months.^[1][3]

Nothing in the evaluation of FUO is more valuable than a thorough history and physical examination, and nothing is cheaper. A comprehensive history and examination is recommended for patients with prolonged fever who have no localising signs or symptoms, and although examination findings alone are present in as many as 97% of patients, they lead to a diagnosis in only 62%.^[3] The search is for potential diagnostic clues: aspects of the history, examination, basic laboratory tests or imaging that direct the next step, and each clue yields a diagnosis in roughly one in five patients. In one prospective study an average of fifteen clues were recorded per patient, of which only 19% contributed to the final diagnosis while 81% were misleading. That sobering figure is not an argument for giving up on the examination; it is an argument for using every clue as a hypothesis to be tested rather than a conclusion.^[1]

The history should cover the exact pattern of fever and any fever-free intervals; rigors and shaking chills, which occur more often with infection; unintentional weight loss and early anorexia, which lean toward malignancy; travel and geographic residence; contact with animals, pets, unusual foods, sick people and workmates; work and home environment including soil, dust and water; tobacco, alcohol and recreational drugs; all medications and their duration; prior febrile illnesses and known conditions such as lymphoma, inflammatory bowel disease or rheumatic fever; and ethnic background or family history of periodic fever syndromes, notably familial Mediterranean fever. The Delphi panel emphasised that travel history and geographic residence should be incorporated into every FUO investigation.^[1][2][3]

The physical examination should be meticulous and repeated, because new findings appear as the disease evolves. At a minimum it includes the skin and nails, oropharynx with attention to dentition, eyes and fundi for Roth spots, heart for new murmurs, abdomen, lymph node groups, genitalia, joints, muscles and spine, and the vascular tree for asymmetric pulses or blood pressures. The findings that suggest specific aetiologies are worth memorising rather than just reading once, because each redirects the whole workup: a new heart murmur suggests infective endocarditis or an atrial myxoma; splenomegaly suggests cytomegalovirus, Epstein-Barr virus, miliary tuberculosis or a lymphoproliferative disorder; lymphadenopathy supports sarcoidosis, systemic lupus erythematosus, rheumatoid arthritis or lymphoma; spinal tenderness points to vertebral osteomyelitis; Roth spots to endocarditis; oral ulcers to Behcet disease or systemic lupus; and an epididymal nodule to extrapulmonary tuberculosis or sarcoidosis. Fever peaks in the morning with typhoid fever, polyarteritis nodosa and Whipple disease. Twice-daily fever peaks occur with malaria, miliary tuberculosis, Still disease and visceral leishmaniasis. Fever with relative bradycardia, the Faget sign, is seen in typhoid fever, yellow fever, central nervous system malignancy and lymphoma. An asymmetric pulse or blood pressure between extremities suggests Takayasu arteritis. A Roth spot is a retinal haemorrhage with a pale centre seen in infective endocarditis.^[3]

Fever patterns deserve a measured view. A few have real diagnostic value: the tertian and quartan periodicity of malaria, the Pel-Ebstein pattern of some Hodgkin lymphoma, typhus inversus in some disseminated tuberculosis, pulse-temperature dissociation in typhoid fever and twice-daily spikes in malaria. Most patterns, however, are rarely observed and are insufficient as sole identifiers of any disease, and liberal use of antipyretics and non-steroidal anti-inflammatory drugs makes them even less useful. The naproxen test, in which the fever of a neoplastic process is presumed to fall with naproxen, does not reliably distinguish neoplastic from non-neoplastic fever and should not be relied on. Recent data show that fever is usually continuous rather than episodic; episodic fever with fever-free intervals of at least two weeks is associated more often with remaining undiagnosed, and recurrent fever carries an undiagnosed rate as high as 47.6% to 50% compared with 20.0% to 25.9% for continuous fever.^[1]

Fever must also be verified before the workup begins. The diurnal rhythm of body temperature, with a morning nadir between 6 and 8 am and a late-afternoon peak between 4 and 6 pm, means that a single normal reading is not reassuring; patients without symptoms or signs of an underlying illness should keep a daily log of temperatures above 38.3 degrees C, noting the measurement site and the instrument used. Factitious fever should be excluded. Body temperature varies by site, contact thermometers at the forehead, mouth, axilla or rectum have traditionally been used, and non-contact infrared devices measuring the forehead surface are influenced by human, environmental and equipment variables that can affect accuracy and relationship with core temperature. Whatever device is used, the anatomic site, instrument and threshold should be recorded.^[1][2]

There is no evidence-based guideline for the evaluation of FUO, and approaches are largely based on expert opinion. The current emphasis is on a systematic, stepwise, multidisciplinary approach involving infectious diseases, internal medicine, rheumatology and oncology as needed, with radiology and procedural specialists involved when biopsies are required. Evaluation should be coordinated rather than chaotic, because uncoordinated testing wastes time and drives false-positive results.^[1][3]

The first phase is a comprehensive history and physical examination followed by the minimum laboratory panel detailed above, three sets of blood cultures and a chest radiograph or CT. Additional infectious evaluation includes serology for Epstein-Barr virus, cytomegalovirus and hepatitis A, B and E, and HIV screening; tuberculosis skin testing or interferon-gamma release assay with chest imaging to exclude pulmonary tuberculosis, since a negative skin test or interferon-gamma result does not exclude active disease. Procalcitonin of 0.5 ng per dL or more supports severe bacterial infection but is not diagnostic of a focus. Echocardiography is useful when endocarditis is suspected, particularly in younger adults where injection drug use has increased the incidence. Abdominal ultrasonography or computed tomography looks for abscess; CT has a sensitivity of 92% for abdominal CT in determining the cause of FUO and a specificity of 60% to 70%, while ultrasonography avoids radiation and cost. For malignancy, computed tomography of the chest, abdomen and pelvis is recommended, and a markedly elevated lactate dehydrogenase supports lymphoma or leukaemia. For inflammatory disease, complete blood count, erythrocyte sedimentation rate and C-reactive protein are measured, with C-reactive protein and sedimentation rate having a high negative predictive value for inflammatory causes; ferritin greater than 1,000 ng per mL suggests adult Still disease; elevated ferritin with creatine kinase warrants evaluation for temporal arteritis, in which a normal sedimentation rate largely excludes giant cell arteritis; rheumatoid factor and antinuclear antibody point toward rheumatoid arthritis or systemic lupus; and chest CT findings can reveal the hilar adenopathy of sarcoidosis. Stool lactoferrin or calprotectin is considered when inflammatory bowel disease remains possible. Echocardiography for endocarditis, venous duplex studies of the lower limbs to exclude venous thromboembolism (the cause in 6% of one series), and thyroid-stimulating hormone (raised or suppressed) for thyroiditis complete the initial panel.^[3]

Imaging beyond the first line is chosen to match the remaining suspicion. Nuclear imaging is generally sensitive but not specific; technetium-based scans are insensitive but 93% to 94% specific and can localise an infectious or inflammatory focus at lower radiation than CT. Magnetic resonance imaging excels for focal problems such as an epidural abscess but may be inaccurate early in inflammatory or infectious disease. Whole-body magnetic resonance imaging has shown promise in detecting inflammatory foci in a large number of patients in small retrospective studies and needs prospective validation. FDG-PET/CT allows whole-body screening for infective, inflammatory and neoplastic processes and appears to be cost-effective when used earlier because it reduces the need for repeated expensive invasive investigations and shortens hospitalisation. The Delphi panel endorsed 18F-FDG-PET/CT as an important early diagnostic test once a patient meets FUO criteria with minimal tests, especially in the absence of potential diagnostic clues, but explicitly did not recommend incorporating PET results into the diagnostic criteria themselves; another expert group, the Society of Nuclear Medicine and Molecular Imaging, reached similar conclusions. Note that this panel endorsed early use even though it strongly disagreed with making PET part of the FUO definition.^[1][2][3]

Biopsies remain the invasive test of choice when non-invasive tests are unrevealing, and they are guided by clues rather than exploratory. Temporal artery biopsy is useful in patients older than 55 years as a low-risk investigation to exclude giant cell arteritis, although FDG-PET/CT is gaining ground as a quick, accurate and non-invasive alternative. Liver biopsy carries a diagnostic yield of 14% to 27% with low harm. Bone marrow biopsy is high-yield when clinical and laboratory findings suggest haematologic disease: early biopsy showed bone marrow abnormalities in 96.2% of patients with FUO and provided a specific diagnosis in 23.7%. Bone marrow cultures have very low yield in immunocompetent patients and are not recommended. Lymph node biopsy is useful with generalised lymphadenopathy but not when adenopathy is confined to the cervical or inguinal regions. Skin or skin-muscle biopsy has a fair yield in areas that look suspicious, bone marrow biopsy is rapid in patients with co-infection with HIV and FUO, and exploratory laparoscopy has been reported as a relatively safe final step in small studies since non-invasive imaging has mostly replaced open laparotomy.^[3]

Empiric antimicrobial therapy has not been shown to be effective in FUO and can delay identification of the real cause, so it should be withheld in stable patients. Two exceptions matter. First, neutropenic, immunocompromised or critically ill patients should receive broad-spectrum antibiotics, with antifungals considered in neutropenic fever because of the high rate of serious bacterial and invasive fungal infection in that group. Second, an empirical trial against tuberculosis, brucellosis, tularemia or suspected infective endocarditis can be used in selected situations, and the resolution of fever after disease-specific therapy can itself provide evidence that supports a presumptive diagnosis. Anti-inflammatory or corticosteroid trial therapy is likewise reserved for seriously ill patients or for those in whom every other approach has failed, because untargeted steroids can mask the very disease they are meant to treat. Corticosteroids should not be introduced casually, and molecular diagnostics and genetic testing are kept for late in the evaluation: molecular methods when tissue biopsy shows granulomatous pathology or targeted pathogen testing is negative, and genetic testing for relapsing, recurring or periodic fever suggesting autoinflammatory disease. Symptomatic treatment including antipyretics is reasonable for comfort and quality of life only once the patient is stable and genuinely reassured that withholding diagnosis-driven therapy is safe.^[2][3]

Children, adolescents, immunocompromised adults and patients with HIV-associated FUO are outside the classic FUO definition and need their own pathways. In the patient who reaches the end of the investigation without a diagnosis, the news is usually better than expected: up to 75% will have spontaneous resolution of the fever and an excellent long-term prognosis. A markedly better outcome in the undiagnosed group does not mean the search should stop early, but it allows honest reassurance. Hodgkin lymphoma can present with the Pel-Ebstein pattern; malaria may produce the twice-daily peaks; typhoid and disseminated tuberculosis can produce pulse-temperature dissociation and typhus inversus respectively, but these are anchors rather than rules. Patients with undiagnosed FUO, episodic fever or recurrent fever need prolonged follow-up because some diagnoses declare themselves only with time, and referral to an expert centre is reported to have high diagnostic and therapeutic value. Most patients with undiagnosed classic FUO have a low mortality.^{[1][3] [2]}

The principles do not change when advanced imaging is unavailable, but the sequence must. A transthoracic echocardiogram, available in many emergency and district hospitals, is the single test that can transform a case of suspected endocarditis or atrial myxoma, and it should be requested in patients whose mitral stenosis does not fit, in young stroke, or in persistent fever with weight loss. Where CT or FDG-PET/CT is unavailable, the investigation leans more heavily on repeated clinical examination, targeted plain radiography, ultrasonography, serology and tissue diagnosis, and the minimum standard tests with locally available tools. Tuberculosis remains the most common identifiable infection across several regions and should be excluded in every patient, and brucellosis, malaria, enteric fever and disseminated infection should be considered where geography suggests them. Drug history, including traditional medicines, should be taken carefully. Tissue biopsy, especially liver, lymph node and bone marrow in HIV-endemic settings, retains a central role because it is cheap relative to repeated imaging. An HIV test belongs in the minimum panel everywhere. Referral to a centre with CT or a surgical biopsy service should be made early when the patient is deteriorating or when no diagnosis is reached. The Delphi panel's preference for chest, abdominal and pelvic CT over plain radiography and ultrasonography is an aspiration for resource-rich settings, not a prerequisite for sensible care.^[1][2][3]

• Fever of unknown origin is a regulated, cytokine-driven process: interleukin-1, interleukin-6 and tumour necrosis factor raise the hypothalamic set point through prostaglandin E2, and rigors more often accompany infection.
• Classic FUO is fever above 38.3 degrees C on more than three occasions, lasting more than three weeks, without explanation despite a standard minimum set of tests in an immunocompetent adult. The nosocomial, neutropenic and HIV-associated categories share the same temperature threshold but differ in host context and the required intensity of investigation. ^[4]
• Most cases are atypical presentations of common disease, so the workup should start from a careful history and examination rather than from rare diagnoses.
• Infection is the largest category, with tuberculosis the most common identifiable cause across several regions, followed by endocarditis and abscess; lymphoma and leukaemia dominate malignancy-associated FUO.
• Non-infectious inflammatory causes, including adult Still disease, systemic lupus erythematosus, giant cell arteritis and sarcoidosis, are more common in high-income countries and with prolonged illness.
• Miscellaneous causes are easily missed: drug fever, thyroid disease, habitual hyperthermia, venous thromboembolism and factitious fever.
• Potential diagnostic clues are the engine of diagnosis, but most are misleading, so each is treated as a hypothesis.
• Interferon-gamma release assay is preferred to tuberculin skin testing where available, and a negative result does not exclude tuberculosis.
• Computed tomography of the chest, abdomen and pelvis and 18F-FDG-PET/CT are the workhorse imaging studies once the minimum panel is complete.
• Tissue biopsy, especially temporal artery in patients over 55, liver and bone marrow, is the high-yield invasive next step.
• Empiric antimicrobial therapy is ineffective and should be avoided except in neutropenic, immunocompromised or critically ill patients.
• As many as 75% of patients with undiagnosed FUO will have spontaneous resolution and a good long-term prognosis.