Fever of unknown origin is still considered one of the most difficult clinical scenarios in children. The current study included 110 children with FUO, the most common diagnostic categories were collagen vascular diseases (30.9%; n = 34), infectious (28.2%; n = 31), miscellaneous (19.1%; n = 21), and malignancy (17.3%; n = 19). In 4.5% of cases (5 cases), the diagnosis was not reached. These findings are contrary to the results of other published studies, which found infectious causes to be the most common [8,9,10,11,12,13]. A previous study conducted at the Infectious Disease Unit at Mansoura University Children's Hospital, from 2006 to 2011, found that the infectious category was the most common (36.22%) [14]. Matched results detected by Kasai et al. in Japan as collagen vascular diseases were detected in 58% followed by infectious diseases (23%) [15].

Several factors contribute to the diversity in the causes of FUO. Several factors contribute to the diversity in the causes of FUO. Firstly, infants less than one year old, who are more susceptible to infections, account for only a small percentage (7.3%) of studied cases Secondly, most cases of FUO are referred to specialized fever hospitals. Thirdly, 85.5% of cases had received antibiotics before being hospitalized, which might interfere with the growth of microorganisms in cultures. Finally, the impact of the COVID-19 pandemic could be a cause of increased other categories as a cause of fever of unknown origin [16].

In the infectious category, Acute schistosomiasis (Katayama fever) was the most common cause (16.1%; n = 5), followed by urinary tract infection (UTI) and brucellosis which were equally represented, (12.9%; n = 4) each. The five cases of acute schistosomiasis were diagnosed based on history, clinical presentation (cercarial dermatitis, acute enterocolitis, fever, malaise, hepatomegaly, and splenomegaly), laboratory assays (eosinophilia, schistosome serology, and or presence of eggs in stool). In one case colonoscopy and biopsy were performed due to suspicion of inflammatory bowel disease, revealing Schistosoma colitis.

Different findings were detected by Chien Yl et al. in Tawain [10], Sumathi Sri R et al. in India [13], and Atia et al. in Egypt [14]. The former two studies found that respiratory infections were the most common, while the latter one reported that enteric fever was the most common. In the current study, UTI and brucellosis were the second most common causes (12.9% (n = 4) each) consistent findings were detected by Reddy PA et al. [11] and Attia et al. [14]. On the contrary, tuberculosis and enteric fever were the leading causes in other studies [12, 13, 17, 18].

Two cases of Epstein-Barr virus (EBV) were identified, making it the most common viral pathogen [6.5% (n = 2)]. The diagnosis of these cases was based on the classical presentation of mononucleosis-like illness, accompanied by positive serology of EBV VCA IgM or rising titter of IgG. In addition, HIV and COVID-19 infections were each detected in 3.2% of infectious cases. Matched results were also detected [12, 17]. Geographical considerations, referral patterns, time of the study, and patient age may all contribute to the differences in the infectious category's aetiologies. In the current study, connective tissue diseases accounted for 30.9% of cases. Systemic lupus erythematosus (SLE) (47.1%) and systemic-onset juvenile idiopathic arthritis (38.2%) were the most common causes. This was like studies carried out by Chien et al. [10] and Lodhi et al. [18]. Meanwhile, juvenile idiopathic arthritis was the most common in other studies [9, 12, 14].

In the miscellaneous group, post-COVID-19 Multisystem inflammatory syndrome in children (MIS-C) was the most common cause accounting for 52.4% (n = 11) of cases, this was not reported in previous studies as the current study was conducted following the COVID-19 pandemic in 2019. The diagnosis of MIS-C was based on the World Health Organization (WHO) case definition [19]. The second most common diagnosis in this category was hemophagocytic lymphohistiocytosis (HLH) which accounted for 23.8% of cases (5 cases).

The possibility of HLH should be kept in mind in all febrile patients with organomegaly and cytopenia involving one or more cell lines. Similar findings were observed in other studies [6, 12]. Acute lymphoblastic leukemia [68.4%(n = 13)] and lymphoma [26.3% (n = 5)] were the most common in the malignant category. Additionally, there was only one case of neuroblastoma detected with positive flow cytometric immunophenotyping from a bone marrow aspirate. Similar findings were matched with other studies [10, 14, 20].

Though, a comparison between collagen vascular, infectious, and malignancy categories regarding demographic categories found a difference in the distribution of age this was found to be statistically insignificant (P = 0.575). In the infectious category, preschool children (1–5 years) were the most common (38.7%), while school-age children were the most common age group in the collagen and malignancy categories [ (55.9%) and (52.6%) respectively]. Additionally, adolescent children were mainly presented in the collagen category (17.6%).

In contrast to age, sex was significantly different (P = 0.007); females (64.7%) were common in the collagen vascular category, while males were more common in the infectious (61.3%), and malignancy (78.9%) categories. Comparable findings were also detected in studies done in Taiwan and Korea [9, 20].

Comparing the duration of fever; though the median duration of fever before admission [45.0 (21.0–68.0) days] was longer in the malignant category this difference was not statistically significant (P = 0.390) similar findings were detected by Reddy et al. and Joshi et al.; who also detected longest duration of fever in malignancy too [11, 21].

Regarding the clinical presentations, children in the infectious category showed higher rates of hepatomegaly (41.9%) and splenomegaly (35.5%), but these differences were not statistically significant (p-values of 0.24 and 0.06, respectively). On the other hand, lymphadenopathy was significantly more prevalent in the collagen (63.6%) and malignant (78.9%) categories. Additionally, arthritis (38.2%) and bone pain (68.4%) were significantly more common in the collagenic and malignant categories, respectively.

In addition to clinical presentation, laboratory data can offer important clues for making a differential diagnosis. Children in the infectious category significantly had higher neutrophil count [5.76 (2.28–7.92) × 10^3/µl] and lymphocytic count [ 4.2 (2.04–5.91) × 10^3/µl]; (P < 0.001 and < 0.010 respectively). In the collagen category the median lymphocytic count was lower [1.95 (1.47–2.73) × 10^3/µl] with significant (P < 0.010) higher neutrophil/ lymphocyte ratio [2.30 (1.53–3.91)]. Other laboratory markers showed no significant difference between the three groups. Therefore, a higher neutrophil and lymphocytic count can predict infectious causes in children with FUO while a low lymphocytic count could predict collagen diseases. Similar results were documented by Ching-Yi Cho in Tawain [20].

Though inflammatory markers, including C-reactive protein (CRP), and ferritin, were higher in the collagen category, and erythrocyte sedimentation rate (ESR) was the highest in the malignant categories; these differences were statistically insignificant. Different results were detected by Bing et al. in China where WBC, CPR, ESR, and PCT in the inflammatory group were significantly higher than those in other groups (p < 0.05) [12]. Moreover Ching-Yi Cho in Tawain [20] found a significant elevation of CRP levels in the infectious group (p < 0.01) and hemoglobin level was significantly lower (12 g/dL) in autoimmune and malignancy (p < 0.01).

First, it is one tertiary-centre study that receives more complicated cases that may not fully explain the pattern of FUO in children. Multicentre studies are needed in the future to validate these results.

Second, given the limited number in each category, though there were significant differences in some clinical presentations and laboratory parameters the current data can’t be used statistically to identify predictors for the diagnosis of FUO.