Preseptal and Orbital Cellulitis: Thirteen Years of Experience and a Novel Scoring System (SNIPPED Score) for Differentiation

We aimed to evaluate clinical and laboratory characteristics of children with preseptal cellulitis (PC) or orbital cellulitis (OC) and to determine whether easily-accessible parameters could be used to predict OC. The data of children diagnosed with PC or OC between January 2008 and December 2020 were evaluated. Patients aged between 1 month and 18 years who were treated with intravenous antibiotics were included. Logistic regression analysis was performed to identify possible parameters in differentiating between PC and OC. A beta coecient-based method was used to derive the scoring system. A total of 375 patients [202 (53.9%) boys], of whom 35 (9.3%) had OC, were evaluated. Median age was 44 (min-max: 1-192) months. Compared to those with PC, patients with OC were older (p = 0.001), had fever and upper respiratory tract infection (URTI) symptoms more frequently, and demonstrated prolonged symptom and hospitalization times (p(cid:0)0.001 for all). Signicant differences between groups were observed for numerous parameters; however, logistic regression analysis revealed only ve parameters independently associated with OC. The SNIPPED score variables and weights were as follows: sinusitis (2 points), neutrophil-to-lymphocyte ratio > 6.78 (3 points), platelet count > 420.5 x10 3 /mm 3 (2 points), proptosis (4 points) and duration of symptoms ≥ 4 days (4 points). A cut-off of ≥ 7 points for OC diagnosis was found to have 91.4% sensitivity, 96.2% specicity, 71.1% PPV, 99.1% NPV and 95.7% accuracy. Conclusion: In addition to showing previously known properties of OC versus PC, our study demonstrated that easily-accessible parameters could be used for the identication of OC. A novel SNIPPED score [sinusitis (2 points), neutrophil-to-lymphocyte ratio >6.78 (3 points), platelet count >420.5 x10 3 /mm 3 (2 points), proptosis (4 points) and duration of symptoms ≥ 4 days (4 points)] may be used to identify OC in pediatric patients. ≥ 7 points from the SNIPPED score was found to have 91.4% sensitivity, 96.2% specicity, 71.1% PPV, 99.1% NPV and 95.7% accuracy to distinguish OC from PC.


Introduction
Orbital infections may be seen in all age groups, but greater frequency is observed in the pediatric population [1]. In cases where the infection penetrates the orbital septum (which forms a barrier between preseptal and post-septal spaces) the condition is called orbital cellulitis (OC). Orbital infections may occur due to spread from the sinuses and by organisms that originate in the upper respiratory tract or skin. Other inciting factors include insect bites, odontogenic infections and trauma [2]. Although preseptal cellulitis (PC) is relatively benign, OC may cause complications including loss of visual acuity, cavernous sinus thrombosis, meningitis, intracranial abscess and septic embolus, if not managed early [3]. Differentiation between PC and OC may be di cult clinically, and there are no laboratory parameters that can directly aid in the discrimination of these two situations.
The aim of this study was to evaluate demographic, clinical and laboratory features, imaging studies, treatment modalities and outcomes of children with orbital infections. Also, we aimed to identify any laboratory parameters that could help differentiate OC from PC, and whether a scoring system could be devised with these parameters.

Patients And Methods
This retrospective study was conducted by evaluating the data of pediatric patients diagnosed with PC or OC between January 2008 and December 2020 at the Pediatric Infectious Diseases Department of our hospital. A medical database search was made with the following ICD-10 (International Classi cation of Diseases-10) codes: in ammation of the eyelids (H01), disorders of eyelids (H02.8, H02.9), ocular involvement in infectious diseases (H03.1), and orbital disorders including cellulitis and abscess (H05). Patients aged between 1 month and 18 years who were admitted to the hospital and treated with intravenous (IV) antibiotics were included in the study. Patients who were admitted with symptoms similar to PC with different nal diagnosis and patients followed with outpatient oral antibiotic therapy were excluded. The demographic characteristics of patients, symptoms at admission, duration of oral antibiotic therapy before admission, affected eye (unilateral/bilateral and right/left), the duration of symptoms and treatment, the type (preseptal/orbital) and etiological cause (trauma, insect bite, sinusitis, conjunctivitis, odontogenic infections, dacryocystitis) of orbital infection were recorded from patient history and medical les. Laboratory parameters including complete blood count [white blood cell count (WBC), absolute neutrophil count (ANC), absolute lymphocyte count (ALC), and platelet count, mean platelet volume (MPV)], acute phase reactants [C-reactive protein (CRP) and erythrocyte sedimentation rate (ESR)], and culture results (blood, cerebrospinal uid, and conjunctival swab cultures) were also recorded. Hematological indices, such as neutrophil-to-lymphocyte ratio (NLR), MPV-to lymphocyte ratio (MPVLR), platelet-to-lymphocyte ratio (PLR), were calculated. Finally, indications and ndings of imaging studies [orbital computerized tomography (CT), orbital magnetic resonance imaging (MRI)] and whether surgical intervention was performed were recorded from the electronic medical database.
The clinical diagnosis of PC was made according to physical examination ndings when there was just eyelid edema with warmth and tenderness without restriction of ocular movements, proptosis or loss of vision. If there were ndings of proptosis, limitations in ocular motility with or without additional signs (such as conjunctival in ammation with chemosis, orbital pain, loss of vision and afferent pupillary defect), OC was diagnosed in a preliminary fashion and was con rmed according to radiological evidence showing involvement of the orbital soft tissue posterior to the orbital septum [4]. Blood culture had been obtained from all patients. Lumbar puncture (LP) had been performed in infants who had signs of systemic infection and when meningitis could not be excluded.
Ophthalmology and otorhinolaryngology consultations were requested when needed. Imaging studies had been ordered when there was a suspicion of OC or when patients had proptosis, ophthalmoplegia or when physicians had been unable to open patients' eyelids due to extreme eyelid edema (preventing evaluation of ocular movements). If patients needed orbital surgery, they were referred to an ophthalmology department. Beta-lactam/beta-lactamase inhibitors had been commenced empirically in all patients, but treatments were adjusted or changed with respect to culture, antibiotic sensitivity and imaging results. When clinical and laboratory improvement had been achieved, patients were discharged for outpatient follow-up by switching to alternant oral antibiotic therapy. Total duration of treatment (IV plus oral) was at least 10 days in PC and 14 to 21 days in OC.

Statistical Methods
The statistical analyses of this study were performed via the SPSS software (version 22.0) for the Windows operating system. Normality of distribution in continuous variables was assessed with Q-Q plots and the Shapiro-Wilk test. Given the presence of normal distribution, comparisons between groups were performed with the Student's t-test; whereas, the Mann-Whitney U test was used to compare variables without normal distribution. The distributions of categorical data in the two groups were compared with Chi-squared tests. Multivariate regression via the backward conditional method was performed to determine factors that were independently effective on OC diagnosis. All parameters that demonstrated signi cant difference between patients with PC and OC in univariate analysis were included in the model, as well as parameters that were known risk factors of OC (ophthalmoplegia and proptosis). Receiver operating characteristic (ROC) analyses were performed at two stages: the rst was to identify cut-off values for continuous variables showing signi cance in multivariable analysis (for categorical transformation), the second was to identify cut-off values for the scoring system (detailed below). Any p value lower or equal to 0.05 was accepted to demonstrate statistical signi cance.

Deriving The Scoring System
As described above, we identi ed factors that were independently associated with OC diagnosis by performing logistic regression. ROC curves were created for parameters that were found to be signi cant in multivariable logistic regression and cut-off values were identi ed with the Youden J method. Next, in order to be able to derive a score system based on categorical factors, we dichotomized continuous variables with cut-off values obtained from initial ROC analyses. A second logistic regression analysis for OC diagnosis was conducted via the enter method with the dichotomous variables identi ed in the previous step.
Weights were assigned to each signi cant parameter by rounding the beta coe cients of regression analysis (not odds ratio) to the nearest integer [5]. Finally, scores were calculated for each patient, ROC analysis was performed, and two relevant cut-off points for the score were identi ed. Sensitivity, speci city, positive predictive value (PPV), negative predictive value (NPV), and accuracy were calculated.

Results
There were a total of 375 patients of whom 202 (53.9%) were boys. Thirty-ve (9.3%) children had OC and 340 (90.7 %) had PC. Overall median age was 44 (min-max: 1-192) months. Median age of patients with OC was higher than the patients with PC (p 0.001).
Symptoms were similar in the PC and OC groups, except for fever and upper respiratory tract infection (URTI) symptoms which were signi cantly more common in the OC group (p 0.001 for both). Demographic, clinical, laboratory ndings, indications and results of imaging studies, and antimicrobial treatment of patients are depicted in Table 1.  Three hundred and twenty-eight (87.5%) patients underwent ophthalmology consultation and 46 (12.3%) underwent otorhinolaryngology consultation. Sixty (16%) had undergone orbital imaging (54 had orbital CT and six had orbital MRI). The indications for imaging studies were: inability to evaluate eye movements in 37 (61.6%) patients, proptosis in 21 (35%), ophthalmoplegia in four (6.6%), and history of trauma in ve (8%). Thirteen patients (7 boys, 6 girls) with OC were diagnosed as having subperiosteal/orbital abscess with orbital imaging studies and ten of them needed surgical intervention. Mean abscess diameter was 20.3 ± 13.6 millimeters. Median age of the patients with subperiosteal abscess was 120 (min-max: 36-156) months and other characteristics are detailed in Table 3.

Discussion
Differentiating OC from PC without access to imaging can be very di cult. As such, this study was undertaken with the aim to describe our experience with PC and OC, and to identify easily-accessible parameters that could be used to distinguish OC. The results show that the SNIPPED scoring system, which is comprised of sinusitis, NLR, platelet count, proptosis and duration of symptoms, was a reliable tool that could be used to predict OC.
Most cases of OC are caused by spread of infection from an adjacent infected sinus, especially ethmoid sinuses. Other much less common sources are spread from the globe (panophthalmitis), the eyelids, the lacrimal sac, infected teeth, and orbital foreign bodies. Finally, although rare, penetrating orbital trauma or skin infection of the face, dental infections, surgery involving ocular or dental regions may cause OC [6-8]. In a study from Turkey that evaluated 36 patients with orbital infections (32 PC and 4 OC) admitted in a 2-year period, the most common predisposing factor was sinusitis in 16 patients (12 PC and 4 OC), and all patients with OC had sinusitis as a predisposing factor [9]. A recently published study reported that 68% of pediatric PC and OC cases had respiratory tract infection or sinusitis [10]. In a retrospective cross-sectional study from a tertiary ophthalmology hospital including 93 pediatric and adult patients (39 OC and 54 PC), it was reported that sinusitis was the most common etiology in both groups and was signi cantly more frequent in the OC group [11]. In another retrospective study including 122 pediatric patients with 80.3% PC and 19.7% OC, sinusitis was reported to be associated with OC, whereas dental abscess and trauma were associated with PC [2]. In our study, sinusitis was more common in patients with OC-similar to previous studies. The in ammation around the eyes in PC occurs as a complication of paranasal sinusitis. The pathogenesis is attributable to the venous drainage of the eyelid and surrounding structures. OC occurs when the infection in the sinuses spread to the orbit through the neighborhood or venous spread [6, 12].
Among laboratory parameters, regression analysis only revealed that NLR and platelet count were associated with OC. Although in ammatory parameters are elevated in both PC and OC, the absolute levels of acute phase reactants may be helpful in clinical practice [12]. For instance, higher CRP levels were reported to be associated with post-septal disease in a recent study [10]. NLR is a parameter which is studied in bacterial infections such as neonatal sepsis, urinary tract infection and community acquired pneumonia [13][14][15]. A recent study which included 243 children with orbital infections, of whom 51 (20.6%) had OC, found that an NLR value of > 3.14 could differentiate OC from PC with a sensitivity of 75.5% and a speci city of 77.4% [16]. Our threshold of NLR (> 6.78) in OC patients seems a bit higher compared to the previous study; but these differences may be due to the well-known variations (based on measurement and devices) in NLR values.
Conjunctival cultures are frequently sterile in children with orbital infections. However, the most commonly identi ed organisms include Streptococcus spp., Staphylococcus spp., respiratory gram-negative organisms, and anaerobes [17]. In our study, the most common was Staphylococcus spp. (24%). Similarly, in a 10-year report including 94 patients (67 PC and 27 OC), the most common pathogen was S. aureus (local abscess, eye swab, and blood cultures) in both adult and pediatric patients. The authors also noted that S.aureus and S.pyogenes were the only agents identi ed in pediatric cases, while S.viridans, Pseudomonas spp., Escherichia coli, H.in uenzae, Fusobacterium spp., Peptostreptococcus spp. demonstrated growth in the cultures of the adult group [18]. Another study that evaluated 213 pediatric cases with orbital infections reported that orbital cultures were performed in 54 (25%) children and the most common pathogen was S. aureus [19]. In a previous research including 94 children with OC (in which a pathogen was recovered in 31% of patients), the most commonly identi ed pathogen was S. anginosus group (15%), while S.aureus was identi ed in 9% of all patients [20]. In our study, only three of the cultures yielded S. anginosus group. Therefore, empirical treatment should cover all these pathogens in children with orbital infections [12].
There are a few systematic reviews that focus on the management of pediatric patients with periorbital cellulitis [21,4]. In children with PC, IV antibiotics should be considered for infants and those with signs of serious systemic infection. Because the meninges are susceptible to inoculation in the infant age group, LP should be performed unless the clinical picture precludes meningitis [12]. In our study group, LP was performed in 9.9% of patients, but meningitis was not diagnosed in any subjects.
OC is treated with the administration of IV antibiotics after the patient is hospitalized. If the patient fails to respond to antibiotic treatment within 24 to 48 hours, it is reasonable to perform CT to look for an orbital abscess and/or plan for surgery [7,21]. Orbital abscess development is associated with various factors, such as being older than 3 years old, having peripheral blood neutrophil count greater than 10.000/mm 3 , presence of periorbital edema and having received treatment with antibiotics previously [22]. All patients in our study group who needed surgery were above 3 years old and the most identi ed predisposing factor was sinusitis. The orbital surgery rate in the current study was 28.5% in the OC group and 2.6% overall. In a study which evaluated 175 pediatric cases (36 had OC), 27 patients had subperiosteal abscess and surgery was needed in 31 (1.7%) of all patients [10]. Another study found the rates of surgical intervention as follows: 29% overall (n = 93), 48.7% in the 39 patients with OC and 14.8% in the 54 patients with PC. The higher rates of surgery may be related to the inclusion of adult patients in this study [11]. It was stated that the proportion of patients requiring surgery increased with age in a study which evaluated 40 children treated for subperiosteal abscess [23]. It may be feasible to suggest that older children and those with sinusitis may bene t from being monitored more carefully about the development of subperiosteal abscess. In our study, surgical intervention was required in all but three of the patients with subperiosteal abscess. In these three patients, abscess diameter was small, fever and other symptoms resolved quickly, response was well to antibiotic treatment, and close monitoring was performed. Ultimately, control imaging studies were normal. Patients who underwent surgery did not attend follow-up visits which presents one of the limitations of this study.
The strength of this study is the large number of cases included over a 13-year period. Another strength is that, to our knowledge, this study is the rst pediatric research that has found a novel scoring system to distinguish OC from PC. Previous studies show that imaging is required to differentiate between PC and OC; however, imaging is expensive and, more importantly, may not be available in resource-limited settings. The value of this scoring system comes from the fact that it drives its results from readily available and/or cheaply measured parameters. Although the ≥ 7-point cut-off has very good accuracy for the diagnosis of OC, the second cut-off point (≥ 12) was identi ed to describe a value that could be used for de nite OC diagnosis in settings where imaging is not readily-available. In addition, this threshold can be used to quickly identify the severity of a patient when clinical suspicion is insu cient or imaging may be delayed.
There are some limitations to discuss. First, this is a retrospective study and carries all limitations associated with this design (ascertainment bias, selection bias). Second, although the number of patients included is respectable, all data is from only one tertiary center; thus, it may be feasible to perform prospective multicenter studies to better evaluate and con rm this scoring system by performing strati cation based on various patient-related characteristics, including sex, age, race, and factors affecting the parameters used for scoring. Finally, it must be noted that we include a temporal parameter (duration of symptoms) in the scoring system. It is evident that judging the utility of this variable in patients who apply before 4 days of symptoms will require further studies. In relation, physicians must be aware of the fact that they must repeat score calculation daily until the 4th day when symptoms continue. However, to conclude, we believe that this score may be helpful to distinguish between PC and OC, since the early and promptly diagnosis and treatment of OC may prevent complications.