Introduction
Fever is defined as an elevated body temperature of 38.0–38.3 °C or higher (
1,
2). It is one of the most common symptoms in children presenting at the hospital. In fact, fever is the chief complaint of children who visit the hospital in 16 to 30% of cases (
3‐
5). The differential diagnosis of fever is broad. The most common cause is infection, with autoimmune disease and malignancy as second and third causes, respectively (
6,
7). In approximately 50% of infants up to 3 years old with fever, no cause is found (
8). In all children with FUO, no definitive cause of fever is found in 10 to 20% of cases, despite extensive history taking, physical examination, and laboratory testing (
9).
There is no widespread consensus on the definition of fever of unknown origin (FUO), and various definitions have been applied throughout the years (
2,
10‐
12). Currently, fever in children is considered FUO after 8 or more days of febrile illness, when a careful and thorough history taking, physical examination, and laboratory workup failed to reveal a probable cause of the fever (
7,
9,
13). If these criteria apply, but children have fever for less than 8 days, they are sometimes considered to have “fever without source” (FWS) (
7,
9).
In the workup of febrile patients, various diagnostic tests may be used to identify the cause of fever. These investigations often include chest radiography, abdominal ultrasonography, laboratory tests such as C-reactive protein (CRP), leukocyte count, urinalysis, and microbiologic tests. More advanced or invasive investigations may include computed tomography (CT), magnetic resonance imaging (MRI), immunologic or organ-specific laboratory investigations, cytologic punctures, and histologic biopsies.
In adults,
18F-fluoro-2-deoxy-D-glucose (FDG) positron emission tomography (PET)/CT is known to be helpful for diagnosing infectious or inflammatory foci in patients with FUO (
12,
14‐
20). FDG-PET/CT is already the first imaging modality in several infectious and inflammatory diseases (
21).
In children with FUO however, literature about the value of FDG-PET/CT is scarce. In a study by Jasper et al. (
22), the results of FDG-PET/CT in 17 children with FUO were reported, and in a study by Blokhuis et al. (
23), 28 FDG-PET/CT scans of children with FUO were described. Because of these relatively low numbers of included patients, the role of FDG-PET/CT in children with FUO remains unestablished.
Therefore, the aim of this study was to assess the value of FDG-PET/CT in finding the cause of fever in a large group of children with FUO and FWS.
Methods
Study design and patients
The electronic patient database of the University Medical Center Groningen was searched for all patients aged 0 to 18 years who underwent FDG-PET/CT between 2010 and 2019. All children who underwent FDG-PET/CT for the evaluation of fever without a known cause were potentially eligible for inclusion. Fever was defined as a body temperature of ≥ 38.3 °C. FUO was defined as febrile illness of multiple days, during which careful history taking, physical examination, and laboratory workup did not reveal a cause of fever. Because a minimum threshold of 8 days fever is not used in the clinic to diagnose FUO, children who had fever for multiple consecutive days, but less than 8 days, were also included in this study.
Children who underwent FDG-PET/CT but did not have fever, or had an already established focus of infection, inflammation, or malignancy, were excluded. When children had follow-up FDG-PET/CT scans, only the first FDG-PET/CT scan was included.
Patient data review
The medical files of all children potentially eligible for inclusion were first reviewed for the inclusion and exclusion criteria. When children were eligible for inclusion, their medical files were further reviewed. Age, gender, medical history, duration of fever, physical examination and history taking, laboratory values (hemoglobin, mean corpuscular volume, thrombocyte count, hematocrit, CRP, leukocyte count and differentiation, erythrocyte sedimentation rate), imaging results and procedures, treatment, final diagnosis, and follow-up data were retrieved from the medical files of all included patients.
FDG-PET/CT acquisition
All scans were performed using an integrated PET/CT system (Biograph mCT 40 or 64 slice PET/CT, Siemens Medical Systems, Knoxville, TN, USA) with 3 min per bed position. Low-dose unenhanced CT was performed for attenuation correction and anatomic mapping with 100 kV and 30 mAs. Some patients (n = 13) underwent concomitant full-dose contrast-enhanced CT with a constant tube potential of 80–120 kV and automatic adjustment of mAs in the z-direction. Full-dose CT would be applied when there was a suspicion of inflammation or infection in a certain organ, and low-dose CT would be likely not to offer a resolution high enough to accurately diagnose the disease.
Patients had to fast for at least 6 hours, but some newborns fasted for only 4 h because of risk of hypoglycemia. Patients with a (slight) suspicion of a cardiac focus of fever also had to adhere to a diet low in carbohydrates the day before FDG-PET/CT was performed. For example, this diet allowed eating meat, fish, eggs, vegetables, and clear soup, but did not allow potatoes, rice, bread, or milk.
After blood glucose concentrations were ensured to be below 11 mmol/L, 3 MBq/kg FDG was administered intravenously. PET/CT imaging was performed 60 min after FDG administration. Data acquisition and reconstruction were in accordance with EANM/EARL (European Association of Nuclear Medicine/ResEARch 4 Life) guidelines (
21,
24). Some of the children had to be sedated or anesthetized when FDG-PET/CT was performed to prevent excessive movement during the scan. Scans were performed from skull to mid-thigh or to toes based on the presence or absence of complaints in the lower extremities.
Interpretation of FDG-PET/CT results
All FDG-PET/CT scans were prospectively interpreted by nuclear medicine physicians as part of routine clinical care, using Syngo.via software (Siemens Healthineers, Erlangen, Germany). All scans with inconclusive findings (not clear-cut positive or negative for a cause of fever) were re-evaluated by another nuclear medicine physician (AWJM) who was blinded to original FDG-PET/CT interpretations, other imaging results, clinical, laboratory, and microbiologic tests.
Diagnostic reference standard
The final diagnosis at hospital discharge was used as reference standard for FDG-PET/CT results. This diagnosis was based on results from all examinations performed during hospital admission (including clinical examination, laboratory analysis, microbiologic cultures, histologic biopsies, imaging results), response to treatment, and clinical or outpatient follow-up. FDG-PET/CT scans that did not identify a cause of fever were classified as true negative if patients were diagnosed with FUO at hospital discharge and fever had resolved spontaneously, without a definitive cause of fever found on any additional testing during hospital stay or during follow-up of at least 3 months. Some children with FUO died because of other underlying diseases. Their fever sometimes did not resolve, but their FDG-PET/CT scans could still be considered true negative if their cause of death was another disease not causing the fever. FDG-PET/CT scans were regarded as true positive when the cause of fever found on FDG-PET/CT corresponded with the diagnosis at hospital discharge.
Statistical analysis
Continuous variables were checked for normal distribution using Kolmogorov-Smirnov tests. Normally distributed data were presented as mean ± standard deviation, and non-normally distributed data as median with interquartile range (IQR). Sensitivity, specificity, positive predictive value (PPV), and negative predictive value (NPV) of FDG-PET/CT for identifying a cause of fever in children with FUO were calculated, along with 95% confidence intervals (CIs).
Age, gender, medical history, duration of fever, laboratory values (CRP, leukocyte count), and CT type (low-dose unenhanced or full-dose contrast-enhanced) were analyzed with univariate logistic regression as independent variables and FDG-PET/CT outcome as dependent variable. The dependent variable was either categorized as true positive or not true positive (i.e., false positive, false negative, or true negative). Corresponding odds ratios (ORs) and 95% CIs were calculated, and P < 0.05 was considered statistically significant. Variables with P ≤ 0.10 on univariate analysis were included in the stepwise multivariate logistic regression model. All statistical analyses were performed using IBM Statistical Package for the Social Sciences (SPSS) version 26 (SPSS, Chicago, IL, USA).
Discussion
This study shows that FDG-PET/CT can play a valuable role in identifying a cause of fever in children with FUO.
In 68 of all 110 patients with fever (62%), a definite cause of fever was identified. This cause was identified on FDG-PET/CT in 53 of all 110 patients (48%).
Previous research on the value of FDG-PET/CT in children with FUO is limited. In a study by Jasper et al., FDG-PET/CT was performed in 17 children with FUO. They reported a sensitivity of FDG-PET/CT of 100% for finding a focus of fever, but a specificity was not reported (
22). It was not specified which patients were included in the sensitivity and specificity analyses, although 43% of included FDG-PET/CT scans were performed for finding an inflammatory focus in patients without FUO. Eighteen percent of FDG-PET/CT results were considered helpful because they excluded differential diagnoses, 24% were considered helpful because they allowed targeted evaluation, and 59% were considered not helpful.
In a study by Blokhuis et al., FDG-PET/CT was performed in 28 children. An infection was found in 7% of patients, malignant disease in 7% of patients, non-infectious inflammatory disease in 32% of patients, and in 54% of patients, no cause of fever was found. Blokhuis et al. reported a sensitivity of 80% and specificity of 78% for finding a cause of fever on FDG-PET/CT (
23). A definitive cause of FUO was established in 16 patients, but this cause was found on only 8 FDG-PET/CT scans and 2 FDG-PET scans.
In our patient population, FDG-PET/CT achieved a sensitivity of 85.5% and specificity of 79.2% for identifying a cause of fever. Three out of 10 false positive FDG-PET/CT results were suggestive of lymphoma. FDG-PET/CT has a high sensitivity for diagnosing lymphoproliferative diseases such as lymphoma, but low specificity. As lymphadenopathy is a relatively common finding in patients with FUO, the use of FDG-PET/CT to distinguish benign or reactive from malignant lymphadenopathy is challenging (
25,
26). However, FDG-PET/CT is an excellent tool to identify an easily accessible FDG avid lymphoid lesion for diagnostic biopsy.
In 5 out of 9 false negative results, the diagnosis at hospital discharge was endocarditis that was not identified on FDG-PET/CT. Four patients had artificial valve endocarditis, and 1 patient had native valve endocarditis. FDG-PET/CT is known to have a low sensitivity for native valve endocarditis, but a very high sensitivity for artificial valve endocarditis (
27). In 2 out of 4 cases of artificial valve endocarditis, the diagnosis was based on clinical signs and positive blood cultures. However, these two patients had not followed a diet low in carbohydrates to suppress physiologic FDG uptake of the myocardium, which may have masked pathologic FDG uptake of the cardiac valves.
As is illustrated by these false negative results, it is important to keep in mind that not all cases of FUO can be diagnosed with FDG-PET/CT. Because of various reasons, the underlying cause of fever may not be visible on FDG-PET/CT. Several precautions can be taken to avoid unnecessary false negative results. Most importantly, these include adhering to a diet low in carbohydrates, especially when a focus of fever is suspected in tissues with high metabolic activity such as the heart, and reducing dosage of corticosteroid treatment to a minimum, especially when vasculitis is suspected. Likewise, prolonged use of antibiotic treatment may reduce the chance of finding a focus of infection (
28). When FDG-PET/CT is unable to detect a cause of fever, but there is still clinical suspicion of a certain disease, additional imaging or testing is still warranted.
In 53% of patients, treatment was modified after FDG-PET/CT. Because of the retrospective nature of this study, it is difficult to relate treatment modifications directly to FDG-PET/CT outcome, especially with regard to changes within the same class of medication (e.g., antibiotics and anti-inflammatory agents). However, in 16 out of 53 patients with a true positive cause of fever, a new type of treatment was started, while this was only the case in 4 out of 38 true negative patients. Thus, a new type of treatment was three times more likely to be started in patients with a true positive cause of fever on FDG-PET/CT than in patients with true negative FDG-PET/CT results.
On univariate logistic regression, only leukocyte count was significantly associated with FDG-PET/CT outcome. On multivariate logistic regression, CRP level (positively) and leukocyte count (negatively) were significantly associated with FDG-PET/CT outcome (OR of 1.01 and 0.91 per unit increase, respectively). Although CRP and leukocyte count are sometimes regarded as infection and inflammatory parameters that are jointly elevated, these results illustrate that this is not always the case. CRP levels can reach very high levels in autoinflammatory diseases such as vasculitis and systemic juvenile idiopathic arthritis, while leukocyte count may be normal or only moderately elevated (
29). Likewise, leukopenia instead of leukocytosis is sometimes seen in patients with sepsis or systemic lupus erythematosus (
30,
31). In autoimmune disease and immunodeficiency, higher erythrocyte sedimentation rates are seen with lower levels of CRP. The different ways in which CRP and leukocyte count are associated with disease activity in different types of diseases likely explain the discrepancy between a positive association for CRP and negative association for leukocyte count in finding a focus of fever.
Aside from CRP level and leukocyte count, no other clinical or demographic factors were significantly associated with FDG-PET/CT outcome. Therefore, it remains challenging to prospectively identify those patients with FUO in whom it would be more or less likely to find a cause of fever with FDG-PET/CT.
In the study by Jasper et al., CRP, neutrophilic granulocytes, and thrombocytes correlated significantly with “positive” FDG-PET or FDG-PET/CT results. However, the definition of “positive scans” also included all scans with non-specific FDG uptake (such as elevated FDG uptake in the bone marrow, which is common in fever), and indeed, 49% of all positive scans were retrospectively regarded as “not helpful” (
32). Because stand-alone FDG-PET results were also included, it is also unclear which final diagnoses were based on FDG-PET and which were based on FDG-PET/CT.
In the study by Blokhuis et al., CRP level was not associated with a true positive focus of fever on FDG-PET(/CT), which might be due to the low number of included patients. Also, by maintaining very strict criteria for FUO, it is questionable how well their study results can be extrapolated to the whole population of children with FUO (
33).
Our study has some limitations. First, due to its retrospective design, there may have been selection bias. In most children with FUO, conventional diagnostics such as radiography, ultrasonography, or routine blood tests were performed before FDG-PET/CT. Only when these tests fail to identify a definite cause of fever, children might be scheduled for FDG-PET/CT. Therefore, the results of this study may not apply to all children with FUO. Second, the diagnoses at hospital discharge and results from patient follow-up were used as reference standard for FDG-PET/CT results. Although all diagnostic test results, clinical signs, response to treatment, and clinical follow-up were considered in the reference diagnosis, the diagnosis at hospital discharge was partly based on FDG-PET/CT itself and therefore might have caused verification bias.
Publisher’s note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.