Place of the 18F-FDG-PET/CT in the Diagnostic Workup in Patients with Classical Fever of Unknown Origin (FUO)

Objective: To explore the diagnostic contribution of the 18F-FDG-PET/CT in a population of patients with classical fever of unknown origin (FUO), to pinpoint its place in the diagnostic decision tree in a real-life setting, and to identify the factors associated with a diagnostic 18F-FDG-PET/CT. Method: All adult patients (aged ≥ 18 years) with a diagnosis of classical FUO who underwent an 18F-FDG-PET/CT in the University Hospital of Montpellier (France) between April 2012 and December 2017 were included. True positive 18F-FDG-PET/CT, which evidenced a specific disease causing FUO, were considered to be contributive. Results: Forty-four patients with FUO have been included (20 males, 24 females; mean age 57.5 ± 17.1 years). Diagnoses were obtained in 31 patients (70.5%), of whom 17 (38.6%) had non-infectious inflammatory diseases, 9 had infections (20.5%), and 3 had malignancies (6.8%). 18F-FDG-PET/CT was helpful for making a final diagnosis (true positive) in 43.6% of all patients. Sensitivity and specificity levels were 85% and 37%, respectively. A total of 135 investigations were performed before 18F-FDG-PET/CT, mostly CT scans (93.2%) and echocardiography (59.1%), and 108 after 18F-FDG-PET/CT, mostly biopsies (including the biopsy of a temporal artery) (25%) and MRIs (34%). In multivariate analysis, the hemoglobin level was significantly associated with a helpful 18F-FDG-PET/CT (p = 0.019, OR 0.41; 95% CI (0.20–0.87)), while the CRP level was not associated with a contributive 18F-FDG-PET/CT. Conclusion: 18F-FDG-PET/CT may be proposed as a routine initial non-invasive procedure in the diagnostic workup of FUO, especially in anemic patients who could be more likely to benefit from 18F-FDG-PET/CT.


Introduction
Fever of unknown origin (FUO) [1,2] is one of the most challenging diagnostic situations, with about 200 possible underlying diagnoses. Classical FUO refers to a prolonged febrile illness for at least 3 weeks, with a body temperature >38.3 • C on several occasions, and without an established etiology despite extended investigations in non-immunocompromised patients [3]. Causes of classical FUO can be stratified into four categories: noninfectious inflammatory diseases (NIID), infections, malignancies, and miscellaneous causes [4,5].
The diagnostic workup of FUO includes a search for potential diagnostic clues (PDCs) that could help to ascertain a specific diagnosis, based on a complete medical history and symptom-taking, a physical examination, and several investigations, including cultures and standard imaging [6]. However, between 10 and 50% of patients remain undiagnosed. The main difficulty lies in the fact that diagnostic workup should be based on limited investigations with the highest diagnostic yield, those that are non-invasive and cost-effective as far as possible. Particularly in patients without PDCs, 18 F-fluorodeoxyglucose (FDG) positron-emission tomography/computed tomography (PET/CT) is now considered a promising technique in the management of patients with FUO [6]. Indeed, FDG accumulates in cells with a high rate of glycolysis, such as activated leukocytes, in inflammatory conditions. Although giving a lower performance for assessing the urinary tract, gastrointestinal tract, and myocardium, 18 F-FDG-PET/CT has been demonstrated to be of great interest in managing FUO [6]. The helpfulness of 18 F-FDG-PET/CT to identify the underlying cause of FUO was described in several studies, ranging from 38 to 75% [6][7][8]. However, these studies were heterogeneous, and the place and performance of 18 F-FDG-PET/CT in the FUO diagnostic workup remain yet to be determined.
The present study aimed to explore the diagnostic contribution of the 18 F-FDG-PET/CT in a population of patients with classical FUO, to pinpoint its place in the diagnostic decision tree in a real-life setting, and to identify those factors associated with a diagnostic 18 F-FDG-PET/CT.

Study Population and Design
This retrospective study included all adult patients (age ≥ 18 years) with a diagnosis of classical FUO who underwent an 18 F-FDG-PET/CT in the University Hospital of Montpellier (France) between April 2012 and December 2017.
We retrospectively screened for eligible patients from the medical files of the nuclear medicine department to avoid referral selection bias, using prespecified search terms for the target condition (e.g., "prolonged fever", "long-term fever", "fever of unknown origin", "recurrent fever"). Inclusion criteria were a diagnosis of FUO [2], defined as a febrile illness of more than 3 weeks' duration, with an established body temperature >38.3 • C (>101 • F), without a diagnosis after history-taking, complete physical examination, and classical biological and radiological investigations that were left to the discretion of the physician, in addition to an 18 F-FDG-PET/CT examination during the diagnostic process. Importantly, the temperature had to be measured and recorded in the medical file, to confirm fever and avoid declarative data alone. A recurrent fever [9] was defined as repeated episodes of FUO with fever-free intervals of at least 2 weeks and the apparent remission of symptoms. The exclusion criteria were non-classical FUO, i.e., immunocompromised patients with neutropenia (leukocyte count < 1.0 × 10 9 /L and/or granulocyte count < 0.5 × 10 9 /L, known human immunodeficiency virus (HIV) infection, known hypogammaglobulinemia (Ig < 50% of the normal value), or the use of the equivalent of more than 10 mg of prednisone for at least 2 weeks, as well as nosocomial fever [2]. This study is registered (French Commission Nationale Informatique et Liberté (CNIL) number 216910). The protocol of this study was approved by the Institutional Review Board of Montpellier University Hospital (IRB-MTP_2021_04_202100784).

18 F-FDG-PET/CT Acquisitions
All patients were instructed to fast for at least 6 h before 18-FDG injection. Serum glucose levels were measured using the hexokinase method. Whole-body emission and transmission scans were acquired in the 3D mode, 60 min after the intravenous administration of 3.5 MBq/kg 18-FDG. Nondiagnostic-quality non-contrast-enhanced CT images were acquired before the acquisition of the PET data. 18 F-FDG-PET/CT was not per-formed for patients with hyperglycemia >2 g/L, for whom clinicians proposed to postpone 18 F-FDG-PET/CT acquisition until after normalization of their glucose level. Quantitative assessment was performed for each metabolic target lesion using the standard uptake value (SUVmax).

Outcomes and Variables
Demographic and clinical characteristics were assessed, including age, gender, past medical history, and several accompanying symptoms such as night sweats, unintentional weight loss, or any specific organic symptom that could potentially offer diagnostic clues. The C-reactive protein (CRP) level, hemoglobin, leukocyte count, lactate dehydrogenase (LDH), ferritin, fibrinogen and gammaglobulins levels were determined. The times between the beginning of fever and hospitalization, and between the beginning of fever and 18 F-FDG-PET/CT acquisition, as well as the time between an 18 F-FDG-PET/CT and final diagnosis, were also recorded. Moreover, in order to determine the sequential investigations that had been realized, before and after 18 F-FDG-PET/CT, were reported for each patient, including CT scanning of the chest, abdomen and pelvis, magnetic resonance imaging (MRI), digestive endoscopy, bronchial endoscopy, bone scintigraphy, transthoracic or transoesophageal echocardiography, and invasive procedures such as a bone marrow biopsy, temporal artery biopsy, liver biopsy or other biopsies.

Final Diagnosis and 18 F-FDG-PET/CT Helpfulness
As classically reported [9], diagnostic results were grouped by infectious diseases, non-infectious inflammatory diseases, malignancy, other diseases, or no diagnosis. To assess the helpfulness of 18 F-FDG-PET/CT, results were stratified into four categories: true positive, true negative, false positive, and false negative. 18 F-FDG-PET/CT was categorized as a true positive or helpful when it revealed a specific disease to be causing FUO that could be confirmed with a biopsy or be clinically ascertained. Thus, only a true positive 18 F-FDG-PET/CT that led to a final diagnosis was considered helpful. 18 F-FDG-PET/CT was categorized as a true negative when it was normal or with some aspecific tracer uptakes that were not contributory to any diagnosis, with no cause of FUO further identified despite extensive procedures. It was categorized as a false positive when the detected results were considered to be unrelated to the condition causing the FUO, and a false negative when the 18 F-FDG-PET/CT was normal but a particular disease was further evidenced with other tests. The helpfulness of 18 F-FDG-PET/CT was independently assessed according to these four categories by two different members of the study team (one senior and one fellow).

Statistical Analyses
Descriptive statistics are presented as mean ± SD or number (%) where appropriate. The Mann-Whitney U test (or Student's t-test) was used to compare continuous variables and chi-square test (or Fisher's exact test) to compare the categorical variables. The p values < 0.05 were considered significant, and all statistical tests were two-sided. Furthermore, we calculated the sensitivity and specificity (and their 95% confidence interval), as well as the positive predictive value (PPV) and negative predictive value (NPV) of the 18 F-FDG-PET/CT. All statistical analyses involved the use of an SAS V7.12 (SAS Institute, Cary, NC, USA).

Study Sample and Patients' Characteristics
Based on the screening of 2021 18 F-FDG-PET/CT tests from the database of the nuclear medicine department between April 2012 and December 2017, according to the inclusion criteria we selected 72 patients who underwent 18 F-FDG-PET/CT to explore an FUO. After excluding 28 patients who did not have a classical FUO (e.g., nosocomial FUO), or the workup was without any temperature measurement reported in the medical file, we analyzed the data from 44 patients.

Position of the 18 F-FDG-PET/CT in the Diagnostic Process
The mean number of exams per patient was 5 ± 2.74. A total of 135 investigations of different kinds (e.g., CT scans, MRIs, echocardiography, endoscopy, a biopsy of the temporal artery, bone marrow biopsy, liver biopsy, or other biopsies) were performed before ( Figure 2A) and 108 after 18 F-FDG-PET/CT ( Figure 2B). In addition, 47.7% of patients were referred to 18 F-FDG-PET/CT after undergoing 2 exams, mostly CT scans (93.2%) and echocardiography (59.1%). After 18 F-FDG-PET/CT, a mean of 2 investigations was performed in 59.9% of patients, mostly biopsy of the temporal artery (25%) or other biopsies, and MRI (34%).

Diagnostic Contribution of 18 F-FDG-PET/CT
Five patients were lost from follow-up after undergoing 18 F-FDG-PET/CT without any diagnosis. Out of the 39 remaining patients, 18 F-FDG-PET/CT was helpful for their final diagnosis (true positive) in 17 patients (43.6% of all patients and 54.8% of patients having a diagnosis) (Figure 3). Seven patients (17.9%) were found to have a true negative 18 F-FDG-PET/CT, with no final diagnosis after workup. Twelve patients (30.8%) had a positive 18 F-FDG-PET/CT with no relationship shown with the final diagnosis or with no final diagnosis (false positive). In 3 patients, we had false-negative findings (7.7%) while a final diagnosis could be made thanks to further investigations or in response to specific treatments. The sensitivity and specificity calculated were 0.85 (95% CI 0.69-1.00) and 0.37 (95% CI 0.15-0.58), respectively. PPV was 0.58 (95% CI 0.41-0.76) and NPV was 0.7 (95% CI 0.41-0.98). 18 F-FDG-PET/CT showed a greater yield in continuous FUO (50%) compared to periodic FUO (22.2%) although this difference was not statistically significant (p = 0.25). Moreover, out of the 31 final diagnoses, 18 F-FDG-PET/CT was contributive to the diagnoses of 41.2% of the NIID, 66.7% of the infectious diseases, and of all malignancies.

Diagnostic Contribution of Other Investigations in the Diagnostic Process
Out of the 31 patients with an FUO diagnosis, 11 chest and abdominal CT scans were found to be helpful for diagnosis, only one echocardiography (pericarditis), two MRIs (biliary tract infection), one bone CT scan (bone metastases), and one bronchial endoscopy (sarcoidosis). None of the 21 digestive endoscopies contributed to FUO diagnosis.

Predictors of High-Yield 18 F-FDG-PET/CT
The clinical and biological characteristics of patients with an 18 F-FDG-PET/CT that was helpful for diagnosis were compared to those with an 18 F-FDG-PET/CT that was not contributive. In the univariate analysis (Table 4), the absence of neurological (p = 0.02) or cutaneous (p = 0.01) symptoms, as well as anemia (p = 0.01), were associated with the diagnostic accuracy of 18 F-FDG-PET/CT. The latter result was confirmed in multivariate analysis with hemoglobin level significantly associated with helpful 18 F-FDG-PET/CT (p = 0.019, OR 0.41 95%CI (0.20-0.87)). In the multivariate analysis, the absence of neurological (p = 0.018, OR 0.09, 95%CI (0.01-0.66)) or ENT symptoms (p = 0.014, OR 0.03, 95% CI (0.00-0.50)) were also associated with a helpful 18 F-FDG-PET/CT. The CRP level was not associated with contributive 18 F-FDG-PET/CT (p = 0.50).

Discussion
FUO is still a condition that is challenging to diagnose in the field of internal medicine, with the inherent risk of missing any serious disease. The spectrum of diseases causing FUO is large and has changed over the last twenty years, with NIID being more prevalent than infectious conditions, especially in Western countries [10]. Nevertheless, between 10 and 50% of FUO remain undiagnosed [11,12]. In our population, a final diagnosis could be reached in 70.5% of patients. Regarding the underlying diseases causing FUO, NIID were involved in 38.6% of cases, infections in 18.2%, malignancies in 9.1%, and miscellaneous diseases in 6.1%. Our results are in accordance with other studies [11,13,14], e.g., Zenone et al., who have reported that 74% of 144 patients received a diagnosis of FUO, of which 35.5% were of NIID, 30.8% of infectious diseases, and 13.5% of malignancies [15]. Among NIID, giant-cell arteritis was the most prevalent cause of FUO (18.2%), before adultonset Still's disease (6.8%) as in other studies [10,15,16]. We found a sensitivity of 85%, respectively, which is consistent with previous studies. Indeed, two meta-analyses, the first including 42 studies and the second including 23 studies, reported a pooled sensitivity of 18 F-FDG PET/CT of 86% and 84%, respectively [17,18].
Until now, there was still no standardized diagnostic algorithm that is recommended in FUO. However, there is now some strong evidence that 18 F-FDG-PET/CT may offer a great contribution to the diagnostic workup in FUO, with the helpfulness of 18 F-FDG-PET/CT shown in 38-75% of patients [6]. In our present study, based on rigorous inclusion criteria and the definition of FUO, we found that 18 F-FDG-PET/CT contributed to the final diagnosis in 43.6% of all patients and 54.8% of patients already having a diagnosis, with a sensitivity of 85% and a specificity of 37%. Our results are consistent with those in previous reports. In one prospective study including 48 patients with FUO, 18 F-FDG-PET/CT showed a diagnostic contribution of 46% [19]. In one of the largest retrospective studies exploring the diagnostic performance of 18 F-FDG-PET/CT, Gafter-Gvili et al. reported that 18 F-FDG-PET/CT could lead to a final diagnosis in 46% of 112 patients with FUO [20]. In their meta-analysis, Bharucha et al. found a global helpfulness of 18 F-FDG-PET/CT of 56% [7]. However, comparing the retrospective studies that have been included seems to be difficult given their heterogeneity, with some studies lacking FUO definition or with a different definition of FUO, others including immunocompromised patients. One prospective study of 240 patients has addressed the diagnostic contribution of 18 F-FDG-PET/CT in both FUO (n = 72) and patients with inflammatory syndrome of unknown origin (IUO), some of whom had no fever [8]. Of all 190 patients with a final diagnosis (79.2%), 18 F-FDG-PET/CT returned a true positive in 136 patients (71.6%; hence, 56.7% of all patients). Moreover, while only 5% were found to be a false negative, 30% were false positive. These findings demonstrated a sensitivity of 91.8% and specificity of 21.7%. Nonetheless, this study did not differentiate FUO from IUO, so that it seems difficult to compare these findings to ours. One remaining concern lies in the definition of a helpful 18 F-FDG-PET/CT [21]. In our study, as in many others [6,21], we considered 18 F-FDG-PET/CT helpful when it was a true positive and directly led to the final diagnosis of FUO. However, one may argue that a true negative 18 F-FDG-PET/CT may also be considered clinically helpful for FUO diagnosis when no cause of FUO was further identified [21].
One strength of our study is to describe the position of 18 F-FDG-PET/CT in the diagnostic workup process in a real-life clinical setting. Almost 95% of patients underwent a CT scan before 18 F-FDG-PET/CT, mostly a chest and abdomen CT scan (77.3%). Interestingly, the advantages of 18 F-FDG-PET/CT are the non-influence of results by metallic implants, whole-body screening, and the absence of contrast-related side-effects, while having the anatomical input of an integrated CT scan [22]. Moreover, gastroscopy and colonoscopy were never useful, as in Bleeker-Rovers et al.'s study [4]. Histological investigations were useful when guided by 18 F-FDG-PET/CT, especially the temporal artery biopsy and lymph node biopsy, while liver and bone marrow biopsies were frequently non-contributive. All these findings are consistent with those of Bleeker-Rovers et al. [4]. Our data, thus, suggest that 18 F-FDG-PET/CT should be performed quite early in the workup of FUO, after echocardiography, if deemed necessary, and in place of a thoracoabdominal CT scan. Histological biopsies might be performed only when guided by anomalies shown by 18 F-FDG-PET/CT. The indication for digestive endoscopic studies should be carefully weighed.
As FUO is a very heterogeneous condition, identifying subgroups of patients with a higher chance of benefiting from 18 F-FDG-PET/CT is of interest in FUO diagnostic development. In this context, finding that anemia was significantly associated with helpful 18 F-FDG-PET/CT (p = 0.019) is of great interest. Similarly, Crouzet et al. showed that anemia is significantly predictive of 18 F-FDG-PET/CT's contribution in the FUO diagnostic [23]. Low hemoglobin may be considered a marker of inflammation. However, we did not find any association between CRP and the helpfulness of 18 F-FDG-PET/CT. This association remains controversial, some studies reporting a significant association between CRP thresholds and 18 F-FDG-PET/CT contribution [8,22,23], while others did not [16,24,25]. In their prospective FUO/IUO study, Shonau et al. found that 18 F-FDG-PET/CT showed a better performance in patients older than 50 years, or those with CRP >30 mg/L and no fever, and proposed that 18 F-FDG-PET/CT could be performed earlier in these patients [8]. Further studies seem necessary to explore this relationship between CRP levels and the helpfulness of 18 F-FDG-PET/CT in FUO.
Our study has several limitations. First, it has the inherent limitations of a retrospective observational study, with inclusion bias. Secondly, some patients have been treated with antibiotics or corticosteroids (<2 weeks), which may have reduced the 18 F-FDG-PET/CT helpfulness. Nevertheless, this point reflects the therapeutic process in real-life conditions. Third, the clinical performance of the referring physicians, and the entire diagnostic process prior to 18 F-FDG-PET/CT, could also have created a bias in patient selection. Finally, the size of the study population might imply relatively low power. However, this size is consistent with most of the published retrospective studies.

Conclusions
In summary, the present data argue for an 18 F-FDG-PET/CT-based diagnosis of FUO. We showed that 18 F-FDG-PET/CT allowed eventual diagnosis in 43.6% of patients. Therefore, 18 F-FDG-PET/CT may be proposed as a routine initial non-invasive procedure in the diagnostic workup of FUO, especially in anemic patients who could be more likely to benefit from a helpful 18