Development and validation of ¹⁸F-FDG PET/CT-based multivariable clinical prediction models for etiological types of fever of unknown origin

Introduction

Classic fever of unknown origin (FUO) is typically defined as a fever exceeding 38.3 ℃ that occurs on multiple occasions over a period of at least three weeks, with an unclear diagnosis after one week of hospitalization and evaluation (1). Common causes of classic FUO include four categories: infections, noninfectious inflammatory diseases (NIIDs), cancers, and other causes (Figure 1). Quickly identifying the category is crucial for deciding the appropriate treatment in patients with FUO.

Figure 1 Illustrative examples of neoplastic disease (A), infectious disease (B) and noninfectious inflammatory disease (C). (A) A 26-year-old male diagnosed with Hodgkin lymphoma had high FDG uptake in the lymph nodes, liver, spleen and bone marrow; (B) a 73-year-old male diagnosed with tuberculosis infection presented focal high FDG uptake in the bone marrow and lymph nodes; (C) a 51-year-old male diagnosed with adult-onset Still’s disease presented diffuse FDG uptake in the liver and spleen. FDG, fluorodeoxyglucose.

In the quest to identify the cause of FUO, patients may undergo comprehensive investigations and medical treatment. Some of these may be invalid and unnecessarily risky. Previous studies have shown that it is difficult to use clinical features to directly diagnose the etiological type of FUO (2,3). In contrast to the conventional diagnostic work-up, the diagnosis of underlying disease may be improved by ¹⁸F-fluorodeoxyglucose positron emission tomography/computed tomography (¹⁸F-FDG PET/CT) (4-6) because of its whole-body visual field imaging and possible discovery of incidental lesions.

However, earlier studies primarily consisted of small case series, with patient numbers ranging from 24 to 376, and the characteristics of PET/CT in patients with different etiologies have not been fully explored (7-9). There are variations in the metabolic characteristics of the spleen, liver, bone marrow, and lymph nodes among patients with different causes of FUO (10), which may serve as clues for etiological diagnosis. Therefore, we investigated the PET/CT characteristics of different types of FUO and further developed novel predictive models for identifying the different categories of etiological disease by integrating PET/CT with clinical variables. We present this article in accordance with the TRIPOD reporting checklist (available at https://qims.amegroups.com/article/view/10.21037/qims-2025-690/rc).

Methods

Study populations and standard diagnostic work‑up

The study was carried out in accordance with the Declaration of Helsinki and its subsequent amendments. This study received approval from the Ethics Committee of West China Hospital (No. 2023-954), and informed consent was waived due to the retrospective nature of the study. The records of patients with FUO (aged ≥14 years) who underwent ¹⁸F-FDG PET/CT at our department from January 2017 to December 2023 were reviewed retrospectively. FUO was defined as (11) (I) lasting more than 3 weeks; (II) having a temperature exceeding 38.3 ℃ on more than three occasions, or (III) having an unclear diagnosis despite appropriate investigations after at least three outpatient visits or at least one week in the hospital. Patients with no diagnosis and with a history of various diseases were excluded because of the small sample size and the lack of commonalities among these diseases. Ultimately, 456 patients with classic FUO were selected for analysis (Figure 2).

Figure 2 Workflow of patient selection. A total of 607 ¹⁸F-FDG PET/CT-screened patients diagnosed with FUO were enrolled in our study. According to our inclusion and exclusion criteria, 142 patients with malignancy-associated FUO and 314 patients with non-malignancy-associated FUO were ultimately included. ¹⁸F-FDG PET/CT, ¹⁸F-fluorodeoxyglucose positron emission tomography/computed tomography; FUO, fever of unknown origin.

The standard diagnostic procedures involved a comprehensive patient history questionnaire, careful physical examination, obligatory investigations, and second-level investigations (CT with contrast, magnetic resonance imaging, nuclear medicine techniques, etc.). The time interval between ¹⁸F-FDG PET/CT imaging and definitive diagnosis was limited to 4 weeks.

Clinical data collection

A standardized form was used to document clinical parameters, including age, sex, and laboratory features, including white blood cell (WBC) count, absolute neutrophil count (ANC), hemoglobin (HGB), platelet count (PLT), C-reactive protein (CRP), procalcitonin (PCT), alanine aminotransferase (ALT), aspartate aminotransferase (AST), lactate dehydrogenase (LDH), serum ferritin (SF), interleukin-2 receptor (IL-2R), interleukin-6 (IL-6), Epstein‒Barr virus (EBV), interferon gamma release (IGRA), antinuclear antibody (ANA), and antineutrophil cytoplasmic antibody (ANCA) levels. Data on laboratory features were collected prior to the initial therapy and within 14 days before or after the ¹⁸F-FDG PET/CT scan. Missing laboratory values were addressed through median imputation for variables with ≤5% missing data, while variables exceeding this threshold were excluded.

Imaging protocol and analysis

¹⁸F-FDG was synthesized at the Nuclear Medicine Department of West China Hospital. All patients were required to fast for 6 hours prior to intravenous administration of ¹⁸F-FDG, with a median activity of 5.18 MBq/kg. We ensured that the patients’ blood sugar levels were maintained below 7.0 mmol/L. The CT scan parameters were set at 120 kV and 40 mAs, with a slice thickness of 5.0 mm and 512×512 matrices. The PET scan was conducted 60±5 minutes after tracer administration, with a duration of 2.5 minutes per bed position. All patients underwent PET/CT scans with one of the following systems: Gemini GXL (Philips Corp., Eindhoven, the Netherlands), DISCOVERY 710 (GE, Waukesha, USA) or uMI780 (United Imaging, Shanghai, China). Patients scanned on different machines followed the same preexamination procedures. We used the acquired CT data to perform attenuation correction on all the PET images.

Two nuclear medicine physicians who were blinded to the etiological diagnosis reviewed the ¹⁸F-FDG PET/CT images. The imaging features of the lymph nodes (high FDG uptake, long and short diameters, maximum standardized uptake value (SUVmax), ratio to mediastinal SUVmax, and asymmetry), bone marrow (focal or diffuse increased FDG uptake, SUVmax, and ratio to mediastinal SUVmax), spleen (focal or diffuse increased FDG uptake, SUVmax, ratio to mediastinal SUVmax, and splenomegaly) and liver (focal or diffuse increased FDG uptake, SUVmax, ratio to mediastinal SUVmax, and hepatomegaly) were recorded. In the absence of hypermetabolic lymph nodes, a right axillary lymph node was selected for documentation. In the absence of hypermetabolic bone marrow, an ellipsoidal region of interest was defined around lumbar vertebra 4 to evaluate the SUVmax of the bone marrow (12). For reference, the SUVmax of the mediastinum was determined by placing a 1-cm-diameter sphere in the aortic arch (13). FDG uptake levels above the surrounding background are considered high uptake.

Statistical analysis

We utilized IBM SPSS statistics software (version 27.0; SPSS Inc.) and R (version 4.0.3) software to perform the statistical analyses. A P value of less than 0.05 was considered statistically significant. Continuous variables were compared via the Mann-Whitney U test, whereas categorical variables were assessed via the χ² test. Statistically significant predictors from preliminary analyses were included in stepwise multivariate logistic regression to identify independent associations. For multiple comparisons, we applied the Bonferroni correction to control for type I error, reporting adjusted P values. Collinearity diagnostic analyses were performed. Predictive models were constructed via logistic equations and simplified into a nomogram. The etiology of FUO can be divided into neoplastic diseases (NDs) and nonneoplastic diseases (NNDs), and NNDs can be further divided into infectious diseases (IDs) and NIIDs. The patients were randomly assigned to training and validation groups at a 7:3 ratio (319 and 137 patients, respectively) (14). This empirical distribution balances the data needed for model training and assessment, preventing underfitting from insufficient training data while maintaining validation stability. The discriminative ability of the models was evaluated via the area under the receiver operating characteristic curve (AUC), with statistical comparisons performed via DeLong’s test. Calibration curves were created to increase the prediction accuracy of the nomogram. Decision curve analysis was employed to evaluate the clinical utility of the models.

Results

Patient characteristics

Among the 456 patients, 236 were female (236/456, 51.8%), with a median age of 48 years (range, 14–88 years). NDs, IDs, and NIIDs were reported in 142 (31.1%), 171 (37.5%), and 143 (31.4%) patients, respectively. Lymphoma was the most common type of tumor (103/142, 72.5%), whereas bacterial infections (mostly tuberculosis infection) accounted for the largest proportion of infections (124/171, 72.5%). More than one-third of patients with NIIDs (57/143, 39.9%) had adult-onset Still’s disease (AOSD). No statistically significant differences were found between the training and validation groups (P>0.05) (Tables S1,S2). Between 1% and 60% of the laboratory data were missing, with ≤5% of the missing data (WBC, ANC, HGB, PLT, ALT, AST, LDH) imputed by median values and measures exceeding 5% of the missing data (CRP, PCT, SF, IL2-R, IL-6, EBV, IGRA, ANA, ANCA) removed.

Comparison between NDs and NNDs in the training cohort

As shown in Tables 1,2, no significant differences in age or sex were observed between the ND and NND groups (P=0.075 and P=0.108, respectively). Most PET/CT variables were significantly different between NDs and NNDs. Notably, ND patients presented a significantly greater prevalence of asymmetric hypermetabolic lymph nodes, hepatic hypermetabolism, and focal liver lesions (all P<0.001), whereas NND patients more frequently presented with splenic hypermetabolism (P=0.002). Additionally, the ND group also presented higher SUVmax values in the bone marrow, spleen and liver but lower rates of splenomegaly and hepatomegaly than did the NND group (all P<0.05). No differences were observed between the groups regarding hypermetabolic lymph nodes or bone marrow, the long or short diameters of the lymph nodes, the SUVmax values of the lymph nodes, or focal lesions of the bone marrow or spleen. With respect to laboratory parameters, the WBC count (P<0.001), ANC count (P<0.001), HGB level (P=0.004), and PLT level (P<0.001) were lower, whereas the LDH level (P=0.004) was greater in patients with NDs than in those with NNDs.

Comparison between IDs and NIIDs in the training cohort

As presented in Tables 3,4, the ID cohort presented a significantly greater proportion of older patients than the NIID cohort did (P=0.003), but there was no significant sex difference between the groups (P=0.380). PET/CT analysis revealed limited significant differences between the groups, with focal spleen lesions (P=0.035) being more common in the ID group, whereas splenic hypermetabolism (P=0.006) predominated in the NIID group. Moreover, the ID group presented significantly larger short-diameter lymph nodes than did the NIID group (P=0.046). No intergroup differences were observed for hypermetabolic lymph nodes, long diameters of the lymph node, bone marrow hypermetabolism or focal lesions, liver hypermetabolism or focal lesions, SUVmax values (lymph nodes, bone marrow, spleen, liver), or the prevalence of splenomegaly and hepatomegaly. Laboratory analysis revealed significantly lower WBC, ANC, and PLT values in the ID group than in the control group (all P<0.001). Conversely, no significant intergroup differences were found for HGB (P=0.071), ALT (P=0.064), AST (P=0.096) or LDH (P=0.324) levels between ID patients and NIID patients.

Multivariate logistic regression models

Variables that showed significant associations in the above analyses were incorporated into a multivariate logistic regression model to identify independent predictors. Finally, five PET/CT variables (asymmetrical distribution of hypermetabolic lymph nodes, bone marrow SUVmax, spleen SUVmax, splenomegaly, and focal liver lesion) and three laboratory variables (WBC, PLT and LDH levels) were selected to construct the ND predictive models. Three PET/CT variables (short diameter of the lymph node, hypermetabolic spleen, and focal spleen lesion), one laboratory variable (WBC level) and age were selected to construct the ID predictive models. The P values of these variables are shown in Tables 1-4. Logistic models were used to create nomograms: one nomogram was based solely on clinical variables (clinical model), while the other included PET/CT and clinical information (combined model) (Figures 3,4).

Figure 3 Nomograms for predicting NDs (A,B) and their calibration curves in the training (C) and validation (D) cohorts. (A) Nomogram of the clinical model. (B) Nomogram of the combined model. The black dotted lines on the diagonal are the reference lines indicating that the predicted value is equal to the actual value. The blue and red lines are the calibration curves representing nomograms (A) and (B), respectively. LDH, lactate dehydrogenase; NDs, neoplastic diseases; PLT, platelet count; SUVmax, maximal standardized uptake value; WBC, white blood cell.

Figure 4 Nomograms for predicting IDs (A,B) and their calibration curves in the training (C) and validation (D) cohorts. (A) Nomogram of the clinical model. (B) Nomogram of the combined model. The black dotted lines on the diagonal are the reference lines indicating that the predicted value is equal to the actual value. The green and pink lines are the calibration curves representing nomograms (A) and (B), respectively. FDG, fluorodeoxyglucose; IDs, infectious diseases; WBC, white blood cell.

For discriminating NDs from NNDs, the clinical model showed acceptable diagnostic accuracy, with an AUC of 0.729 [95% confidence interval (CI): 0.667–0.791]. The specificity of the model was 94%. Notably, the combined model had a greater AUC of 0.791 (95% CI: 0.734–0.849) (P=0.006). The specificity of the model was 95%. The decision curve analysis indicated that the combined model demonstrated superior clinical utility for predicting malignancy compared with the clinical model (Figure 5).

Figure 5 Evaluation of the models in the training cohort. (A) Receiver operating characteristic curves of the two models for differentiating NDs from NNDs; (B) receiver operating characteristic curves of the two models for differentiating IDs from NIIDs; (C) decision curve analysis for the abovementioned models for differentiating NDs from NNDs; (D) decision curve analysis for the abovementioned models for differentiating IDs from NIIDs. AUC, area under the receiver operating characteristic curve; IDs, infectious diseases; NDs, neoplastic diseases; NNDs, nonneoplastic diseases; NIIDs, noninfectious inflammatory diseases.

For the differentiation of IDs and NIIDs, the clinical model demonstrated acceptable accuracy (AUC =0.725, 95% CI: 0.658–0.792; sensitivity =77%). The combined model showed comparable performance (AUC =0.745, 95% CI: 0.680–0.809; sensitivity =71%), without statistically significant improvement over the clinical model (P=0.389). Decision curve analysis revealed that the combined model demonstrated greater clinical utility for predicting infection than did the clinical model (Figure 5).

Validation of the models in the validation cohort

We implemented our logistic models on the validation cohort (Table 5), and the models had an acceptable discrimination ability (AUC >0.65). Compared with the clinical model, the combined model did not significantly improve the discrimination between NDs and NNDs [AUC =0.742 (95% CI: 0.649–0.835) vs. 0.680 (95% CI: 0.584–0.775), P=0.08; specificity =92% vs. 91%]. Similarly, for IDs vs. NIIDs, the combined model [AUC =0.737 (95% CI: 0.638–0.840), sensitivity =75%] did not significantly improve the clinical model [AUC =0.678 (95% CI: 0.566–0.791), sensitivity =84%; P=0.111]. Decision curve analysis revealed that the combined models performed better than the clinical models did (Figure 6).

Figure 6 Evaluation of the models in the validation cohort. (A) Receiver operating characteristic curves of the two models for differentiating NDs from NNDs; (B) receiver operating characteristic curves of the two models for differentiating IDs from NIIDs; (C) decision curve analysis for the abovementioned models for differentiating NDs from NNDs; (D) decision curve analysis for the abovementioned models for differentiating IDs from NIIDs. AUC, area under the receiver operating characteristic curve; IDs, infectious diseases; NDs, neoplastic diseases; NNDs, nonneoplastic diseases; NIIDs, noninfectious inflammatory diseases.

Discussion

This study suggests that ¹⁸F-FDG PET/CT may be a practical technique for identifying the etiological types of FUO. However, incorporating ¹⁸F-FDG PET/CT features did not significantly enhance the discriminative ability of clinical models for FUO etiology. Further validation in future studies is warranted.

Patients with FUO are clinically challenging due to nonspecific symptoms and many differential diagnoses and may require extensive tests and even diagnostic treatment. The three most common etiologies of FUO are IDs, NDs, and NIIDs. In this study, IDs were the most common cause of FUO (37.5%), followed by NDs (31.4%) and NIIDs (31.1%). This etiological distribution is in concordance with previous studies (15,16). Prompt identification of the underlying category is crucial for enhancing treatment strategies for patients with FUO (17-19).

¹⁸F-FDG PET/CT can reportedly be used to successfully localize the source of fever in 42% to 58% of patients with FUO (20,21). However, directly establishing a definitive etiological diagnosis can be challenging. Many patients exhibit FDG uptake at multiple sites, with the spleen, liver, bone marrow, and lymph nodes being the most common (22-24). However, there are differences in the metabolic characteristics of these organs and tissues among different etiologies of FUO (10), which can be used as clues. In our study, the asymmetrical distribution of hypermetabolic lymph nodes and the SUVmax of the bone marrow and the spleen allowed us to distinguish between NDs and NNDs. These results were associated with the high prevalence of lymphoma among patients with malignant diseases, which was also noted by Zhang et al. (25) and Wang et al. (26). In addition, consistent with several studies (26-28) reporting that focal FDG uptake is indicative of underlying malignancy, our findings suggest that focal liver lesions may similarly indicate malignancy. However, few studies have reported the volume of the spleen to date. The present study demonstrated that the spleen volume in the NND group was greater than that in the ND group. In the identification of IDs and NIIDs, our results indicated that focal spleen lesions were closely linked to infection, which is consistent with the findings of a previous report (29). AOSD, which typically presents as diffuse elevated FDG uptake in the spleen (30), accounts for the largest proportion of NIIDs, which may explain why low FDG uptake in the spleen is helpful for the diagnosis of infections.

In addition to PET/CT findings, clinical factors can assist in determining the cause of FUO. Age is a significant predictor, with older patients being more prevalent in the ID group, which is consistent with past results (17). Owing to a high proportion of missing laboratory data, only common hematological and biochemical parameters were included in the final analysis. In NDs, impaired immune regulation may lead to abnormal laboratory values, such as elevated LDH levels (31) and decreased PLT counts (32). Conversely, the WBC count serves as a marker of infection and inflammation (33). Notably, elevated WBC and PLT counts, along with reduced LDH levels, are independent predictors of NND. In the identification of IDs and NIIDs, leukopenia is helpful for the diagnosis of infection, possibly because bacterial infection can lead to sepsis (34), and it is an independent predictor of infection. In addition, previous studies have demonstrated that elevated ALT levels contribute to the diagnosis of NIIDs, possibly because AOSD can lead to hepatitis (35). However, there was no significant difference between IDs and NIIDs in our study. The heterogeneity of the study population may partially account for this result, and the predictor should also be evaluated in larger cohorts.

Previous studies emphasized that ¹⁸F-FDG PET/CT or clinical features alone are insufficient to make an accurate differential diagnosis for FUO (2,3,36). Therefore, a comprehensive model is needed. However, incorporating ¹⁸F-FDG PET/CT imaging features with clinical features did not significantly enhance model discrimination performance, regardless of whether NDs were differentiated from NNDs (AUCcombined =0.742 vs. AUCclinical =0.680; P=0.080) or IDs from NIIDs (AUCcombined =0.737 vs. AUCclinical =0.678; P=0.111). Previous research on the value of combined models in the etiological classification of FUO is limited. The model developed by Chen et al. (10) focused only on the SUVmax of the hottest lesion, and no tools for visualization have been established for clinical use. Our research focused on variations in the metabolic characteristics of the lymph nodes, liver, spleen, and bone marrow among patients with different etiologies of FUO and established nomograms. Moreover, the models were evaluated via calibration curves and decision curves, which revealed good consistency and clinical utility of the combined model.

There are several limitations in this study. First, this study was limited by its single-center, retrospective design and small sample size. More large-sample, multicenter and prospective studies are needed to determine the value of ¹⁸F-FDG PET/CT for identifying the etiological types of FUO. Second, the absence of external validation limits the generalizability of our findings. Future studies should include independent patient cohorts to confirm the diagnostic utility of ¹⁸F-FDG PET/CT in this context. Third, increased uptake was assessed visually rather than quantitatively. However, owing to inconsistent background uptake in the lymph nodes, spleen, liver, and bone marrow, defining elevated uptake using uniform SUVmax thresholds is challenging. Future studies will employ alternative methods based on larger datasets to establish more reliable criteria. Finally, we did not report patients with no diagnosis or miscellaneous diseases because of the limited sample size and the absence of common characteristics among those diseases, which is what we aim to investigate.

Conclusions

While our pretreatment ¹⁸F-FDG PET/CT-based model demonstrated utility in differentiating FUO etiologies, its diagnostic performance did not significantly surpass that of clinical models. These findings highlight the need for multicenter, externally validated prospective studies to further validate and strengthen the evidence.

Acknowledgments

None.

Footnote

Reporting Checklist: The authors have completed the TRIPOD reporting checklist. Available at https://qims.amegroups.com/article/view/10.21037/qims-2025-690/rc

Data Sharing Statement: Available at https://qims.amegroups.com/article/view/10.21037/qims-2025-690/dss

Funding: This study was supported by the National Key R&D Program of China (No. 2023YFC2414201-1) and the National Natural Science Foundation of China (No. 81971653).

Conflicts of Interest: All authors have completed the ICMJE uniform disclosure form (available at https://qims.amegroups.com/article/view/10.21037/qims-2025-690/coif). The authors have no conflicts of interest to declare.

Ethical Statement: The authors are accountable for all aspects of the work in ensuring that questions related to the accuracy or integrity of any part of the work are appropriately investigated and resolved. The study was conducted in accordance with the Declaration of Helsinki and its subsequent amendments. The study was approved by the Ethics Committee of West China Hospital (No. 2023-954), and informed consent was waived because this was a retrospective study.

Open Access Statement: This is an Open Access article distributed in accordance with the Creative Commons Attribution-NonCommercial-NoDerivs 4.0 International License (CC BY-NC-ND 4.0), which permits the non-commercial replication and distribution of the article with the strict proviso that no changes or edits are made and the original work is properly cited (including links to both the formal publication through the relevant DOI and the license). See: https://creativecommons.org/licenses/by-nc-nd/4.0/.

References

1. Mulders-Manders C, Simon A, Bleeker-Rovers C. Fever of unknown origin. Clin Med (Lond) 2015;15:280-4. [Crossref] [PubMed]
2. Jasper N, Däbritz J, Frosch M, Loeffler M, Weckesser M, Foell D. Diagnostic value of (18)F-FDG PET/CT in children with fever of unknown origin or unexplained signs of inflammation. Eur J Nucl Med Mol Imaging 2010;37:136-45. [Crossref] [PubMed]
3. Bleeker-Rovers CP, Vos FJ, de Kleijn EMHA, Mudde AH, Dofferhoff TSM, Richter C, Smilde TJ, Krabbe PFM, Oyen WJG, van der Meer JWM. A prospective multicenter study on fever of unknown origin: the yield of a structured diagnostic protocol. Medicine (Baltimore) 2007;86:26-38. [Crossref] [PubMed]
4. Schönau V, Vogel K, Englbrecht M, Wacker J, Schmidt D, Manger B, Kuwert T, Schett G. The value of (18)F-FDG-PET/CT in identifying the cause of fever of unknown origin (FUO) and inflammation of unknown origin (IUO): data from a prospective study. Ann Rheum Dis 2018;77:70-7. [Crossref] [PubMed]
5. Kouijzer IJE, Mulders-Manders CM, Bleeker-Rovers CP, Oyen WJG. Fever of Unknown Origin: the Value of FDG-PET/CT. Semin Nucl Med 2018;48:100-7. [Crossref] [PubMed]
6. Hung BT, Wang PW, Su YJ, Huang WC, Chang YH, Huang SH, Chang CC. The efficacy of (18)F-FDG PET/CT and (67)Ga SPECT/CT in diagnosing fever of unknown origin. Int J Infect Dis 2017;62:10-7. [Crossref] [PubMed]
7. Wang WX, Cheng ZT, Zhu JL, Xing MY, Zheng CF, Wang SJ, Xie NN. XianYu ZQ, Song JX. Combined clinical parameters improve the diagnostic efficacy of (18)F-FDG PET/CT in patients with fever of unknown origin (FUO) and inflammation of unknown origin (IUO): A prospective study in China. Int J Infect Dis 2020;93:77-83. [Crossref] [PubMed]
8. Wang Q, Li YM, Li Y, Hua FC, Wang QS, Zhang XL, Cheng C, Wu H, Yao ZM, Zhang WF, Hou QY, Miao WB, Wang XM. 18F-FDGPET/CT in fever of unknown origin and inflammation of unknown origin: a Chinese multi-center study. Eur J Nucl Med Mol Imaging 2019;46:159-65. [Crossref] [PubMed]
9. Letertre S, Fesler P, Zerkowski L, Picot MC, Ribstein J, Guilpain P, Le Moing V, Mariano-Goulart D, Roubille C. Place of the (18)F-FDG-PET/CT in the Diagnostic Workup in Patients with Classical Fever of Unknown Origin (FUO). J Clin Med 2021;10:3831. [Crossref] [PubMed]
10. Chen J, Xing M, Xu D, Xie N, Zhang W, Ruan Q, Song J. Diagnostic models for fever of unknown origin based on (18)F-FDG PET/CT: a prospective study in China. EJNMMI Res 2022;12:69. [Crossref] [PubMed]
11. Knockaert DC, Vanderschueren S, Blockmans D. Fever of unknown origin in adults: 40 years on. J Intern Med 2003;253:263-75. [Crossref] [PubMed]
12. Yang X, Lu X, Liu J, Kan Y, Wang W, Zhang S, Liu L, Li J, Yang J. Development and Validation of (18)F-FDG PET/CT-Based Multivariable Clinical Prediction Models for the Identification of Malignancy-Associated Hemophagocytic Lymphohistiocytosis. Korean J Radiol 2022;23:466-78. [Crossref] [PubMed]
13. Boktor RR, Walker G, Stacey R, Gledhill S, Pitman AG. Reference range for intrapatient variability in blood-pool and liver SUV for 18F-FDG PET. J Nucl Med 2013;54:677-82. [Crossref] [PubMed]
14. Chen X, Tang Y, Mo Y, Li S, Lin D, Yang Z, Yang Z, Sun H, Qiu J, Liao Y, Xiao J, Chen X, Wu X, Wu R, Dai Z. A diagnostic model for coronavirus disease 2019 (COVID-19) based on radiological semantic and clinical features: a multi-center study. Eur Radiol 2020;30:4893-902. [Crossref] [PubMed]
15. Zhou G, Zhou Y, Zhong C, Ye H, Liu Z, Liu Y, Tang G, Qu J, Lv X. Retrospective analysis of 1,641 cases of classic fever of unknown origin. Ann Transl Med 2020;8:690. [Crossref] [PubMed]
16. Georga S, Exadaktylou P, Petrou I, Katsampoukas D, Mpalaris V, Moralidis EI, Arvaniti K, Papastergiou C, Arsos G. Diagnostic Value of (18)F-FDG-PET/CT in Patients with FUO. J Clin Med 2020;9:2112. [Crossref] [PubMed]
17. Yan Y, Chen C, Liu Y, Zhang Z, Xu L, Pu K. Application of Machine Learning for the Prediction of Etiological Types of Classic Fever of Unknown Origin. Front Public Health 2021;9:800549. [Crossref] [PubMed]
18. Xu T, Wang L, Wu S, Zhou F, Huang H. Utility of a Simple Scoring System in Differentiating Bacterial Infections in Cases of Fever of Unknown Origin. Clin Infect Dis 2020;71:S409-15. [Crossref] [PubMed]
19. Efstathiou SP, Pefanis AV, Tsiakou AG, Skeva II, Tsioulos DI, Achimastos AD, Mountokalakis TD. Fever of unknown origin: discrimination between infectious and non-infectious causes. Eur J Intern Med 2010;21:137-43. [Crossref] [PubMed]
20. Takeuchi M, Dahabreh IJ, Nihashi T, Iwata M, Varghese GM, Terasawa T. Nuclear Imaging for Classic Fever of Unknown Origin: Meta-Analysis. J Nucl Med 2016;57:1913-9. [Crossref] [PubMed]
21. Hess S, Hansson SH, Pedersen KT, Basu S, Høilund-Carlsen PF. FDG-PET/CT in Infectious and Inflammatory Diseases. PET Clin 2014;9:497-519. vi-vii. [Crossref] [PubMed]
22. Yang YQ, Ding CY, Xu J, Fan L, Wang L, Tian T, Li TN, Li JY, Xu W. Exploring the role of bone marrow increased FDG uptake on PET/CT in patients with lymphoma-associated hemophagocytic lymphohistiocytosis: a reflection of bone marrow involvement or cytokine storm? Leuk Lymphoma 2016;57:291-8. [Crossref] [PubMed]
23. Shieh AC, Guler E, Smith DA, Tirumani SH, Beck RC, Ramaiya NH. Hemophagocytic Lymphohistiocytosis: A Primer for Radiologists. AJR Am J Roentgenol 2020;214:W11-9. [Crossref] [PubMed]
24. Boddu P, Oviedo SP, Rausch CR, Yam C, Daver N, Kantarjian H, Kadia TM. PET-CT in AML-related hemophagocytic lymphohistiocytosis. Leuk Lymphoma 2018;59:1486-9. [Crossref] [PubMed]
25. Zhang LJ, Xu J, Liu P, Ding CY, Li JY, Qiu HX, Zhang SJ. The significance of 18F-FDG PET/CT in secondary hemophagocytic lymphohistiocytosis. J Hematol Oncol 2012;5:40. [Crossref] [PubMed]
26. Wang J, Wang D, Zhang Q, Duan L, Tian T, Zhang X, Li J, Qiu H. The significance of pre-therapeutic F-18-FDG PET-CT in lymphoma-associated hemophagocytic lymphohistiocytosis when pathological evidence is unavailable. J Cancer Res Clin Oncol 2016;142:859-71. [Crossref] [PubMed]
27. Zheng Y, Hu G, Liu Y, Ma Y, Dang Y, Li F, Xing H, Wang T, Huo L. The role of 18F-FDG PET/CT in the management of patients with secondary haemophagocytic lymphohistiocytosis. Clin Radiol 2016;71:1248-54. [Crossref] [PubMed]
28. Merrill SA, Naik R, Streiff MB, Shanbhag S, Lanzkron S, Braunstein EM, Moliterno AM, Brodsky RA. A prospective quality improvement initiative in adult hemophagocytic lymphohistiocytosis to improve testing and a framework to facilitate trigger identification and mitigate hemorrhage from retrospective analysis. Medicine (Baltimore) 2018;97:e11579. [Crossref] [PubMed]
29. Yang X, Lu X, Liu J, Kan Y, Wang W, Zhang S, Liu L, Li J, Yang J. Development and Validation of (18)F-FDG PET/CT-Based Multivariable Clinical Prediction Models for the Identification of Malignancy-Associated Hemophagocytic Lymphohistiocytosis. Korean J Radiol 2022;23:466-78. [Crossref] [PubMed]
30. Jiang L, Xiu Y, Gu T, Dong C, Wu B, Shi H. Imaging characteristics of adult onset Still's disease demonstrated with 18F-FDG PET/CT. Mol Med Rep 2017;16:3680-6. [Crossref] [PubMed]
31. Li F, Li P, Zhang R, Yang G, Ji D, Huang X, Xu Q, Wei Y, Rao J, Huang R, Chen G. Identification of clinical features of lymphoma-associated hemophagocytic syndrome (LAHS): an analysis of 69 patients with hemophagocytic syndrome from a single-center in central region of China. Med Oncol 2014;31:902. [Crossref] [PubMed]
32. Tabata C, Tabata R. Possible prediction of underlying lymphoma by high sIL-2R/ferritin ratio in hemophagocytic syndrome. Ann Hematol 2012;91:63-71. [Crossref] [PubMed]
33. Pijl JP, Kwee TC, Legger GE, Peters HJH, Armbrust W, Schölvinck EH, Glaudemans AWJM. Role of FDG-PET/CT in children with fever of unknown origin. Eur J Nucl Med Mol Imaging 2020;47:1596-604. [Crossref] [PubMed]
34. Clarke RT, on T, van Hamel Parsons V, King AJ. Neutropenic sepsis: management and complications. Clin Med (Lond) 2013;13:185-7. [Crossref] [PubMed]
35. Priori R, Barone F, Alessandri C, Colafrancesco S, McInnes IB, Pitzalis C, Valesini G, Bombardieri M. Markedly increased IL-18 liver expression in adult-onset Still's disease-related hepatitis. Rheumatology (Oxford) 2011;50:776-80. [Crossref] [PubMed]
36. Yuan L, Kan Y, Meeks JK, Ma D, Yang J. 18F-FDG PET/CT for identifying the potential causes and extent of secondary hemophagocytic lymphohistiocytosis. Diagn Interv Radiol 2016;22:471-5. [Crossref] [PubMed]