International Journal of Pediatric Otorhinolaryngology 171 (2023) 111629

Contents lists available at ScienceDirect

International Journal of Pediatric Otorhinolaryngology

journal homepage: www.elsevier.com/locate/ijporl

Abscess volume as a predictor of surgical intervention in children

hospitalized with orbital cellulitis: A multicentre cohort study
Morgyn F. McKerlie a, Nikolaus E. Wolter b, c, Sanjay Mahant b, d, e, f, Jessica Cichon e,
Elysa Widjaja b, d, Olivier Drouin g, h, i, Catherine Pound j, Julie Quet j, Gita Wahi k, Ann Bayliss l,
Gemma Vomiero m, Jessica L. Foulds n, Ronik Kanani o, Mahmoud Sakran p, q, Anupam Sehgal r,
Cornelia M. Borkhoff b, d, e, f, Eleanor Pullenayegum e, s, Arun Reginald b, t, Patricia C. Parkin b, d, e, f,
Peter J. Gill b, d, e, f, *, on behalf of thePeriorbital and Orbital Cellulitis (POC), Multicentre Study
Group and Canadian Pediatric Inpatient Research Network (PIRN)
a
Queens School of Medicine, Faculty of Health Sciences, Queen’s University, Kingston, ON, Canada
b
The Hospital for Sick Children, Toronto, ON, Canada
c
Department of Otolaryngology, Head & Neck Surgery, University of Toronto, Toronto, ON, Canada
d
Department of Pediatrics, University of Toronto, Toronto, ON, Canada
e
Child Health Evaluative Sciences, SickKids Research Institute, Toronto, ON, Canada
f
Institute of Health Policy, Management and Evaluation, University of Toronto, Toronto, ON, Canada
g
Department of Pediatrics, Université de Montreal, Montreal, QC, Canada
h
Division of General Pediatrics, Department of Pediatrics, CHU Sainte-Justine, Montreal, QC, Canada
i
Department of Social and Preventive Medicine, School of Public Health, Université de Montréal, Montreal, QC, Canada
j
Department of Pediatrics, University of Ottawa, Children’s Hospital of Eastern Ontario (CHEO), Ottawa, ON, Canada
k
Department of Pediatrics, McMaster University, McMaster Children’s Hospital, Hamilton, ON, Canada
l
Trillium Health Partners, Department of Paediatrics, University of Toronto, Mississauga, ON, Canada
m
Department of Pediatrics, University of Calgary, Alberta Children’s Hospital, Calgary, AB, Canada
n
Department of Pediatrics, University of Alberta, Stollery Children’s Hospital, Edmonton, AB, Canada
o
North York General Hospital, University of Toronto, Toronto, ON, Canada
p
Department of Paediatrics, Queens University, Lakeridge Health, Oshawa, ON, Canada
q
Research, Lakeridge Health, Oshawa, ON, Canada
r
Kingston Health Sciences Centre, Queen’s University, Kingston, ON, Canada
s
Dalla Lana School of Public Health, University of Toronto, Toronto, ON, Canada
t
Department of Ophthalmology and Visual Sciences, University of Toronto, Toronto, ON, Canada

A R T I C L E I N F O A B S T R A C T

Keywords: Importance: Orbital cellulitis with subperiosteal or orbital abscess can result in serious morbidity and mortality in
Orbital cellulitis children. Objective volume criterion measurement on cross-sectional imaging is a useful clinical tool to identify
Periorbital cellulitis patients with abscess who may require surgical drainage.
Pediatrics
Objective: To determine the predictive value of abscess volume and the optimal volume cut-point for surgical
Observational study
intervention.
Design: We conducted an observational cohort study using medical records from children hospitalized between
2009 and 2018.
Setting: Multicentre study using data from 6 children’s hospitals.
Participants: Children were included if they were between 2 months and 18 years of age and hospitalized for an
orbital infection with an abscess confirmed on cross-sectional imaging.
Exposure: Subperiosteal or orbital abscess volume.
Main outcome and measures: The primary outcome was surgical intervention, defined as subperiosteal and/or
orbital abscess drainage. Multivariable logistic regression was performed to assess the association of abscess
volume with surgery. To determine the optimal abscess volume cut-point, receiver operating characteristic
(ROC) analysis was performed using the Youden Index to optimize sensitivity and specificity.

* Corresponding author. Peter Gilgan Centre for Research and Learning, 686 Bay Street, Toronto, ON, M5G 0A4, Canada.
E-mail address: peter.gill@sickkids.ca (P.J. Gill).

https://doi.org/10.1016/j.ijporl.2023.111629
Received 11 March 2023; Received in revised form 4 June 2023; Accepted 7 June 2023
Available online 8 July 2023
0165-5876/© 2023 Elsevier B.V. All rights reserved.
M.F. McKerlie et al. International Journal of Pediatric Otorhinolaryngology 171 (2023) 111629

Results: Of the 150 participants (mean [SD] age, 8.5 [4.5] years), 68 (45.3%) underwent surgical intervention. On
multivariable analysis, larger abscess volume and non-medial abscess location were associated with surgical
intervention (abscess volume: adjusted odds ratio [aOR], 1.46; 95% CI, 1.11–1.93; abscess location: aOR, 3.46;
95% CI, 1.4–8.58). ROC analysis demonstrated an optimal abscess volume cut-point of 1.18 mL [AUC: 0.75 (95%
CI 0.67–0.83) sensitivity: 66%; specificity: 79%].
Conclusions and Relevance: In this multicentre cohort study of 150 children with subperiosteal or orbital abscess,
larger abscess volume and non-medial abscess location were significant predictors of surgical intervention.
Children with abscesses >1.18 mL should be considered for surgery.

1. Introduction 2009, and December 31, 2018, and had a physician diagnosis of severe
orbital infection. Patients were identified using the International Clas­
Orbital cellulitis can result in serious morbidity and mortality in sification of Diseases, Tenth Revision, Canada diagnostic code of H05.0
children [1,2]. Extension of orbital cellulitis into adjacent anatomical (acute inflammation of the eye). The study protocol and main results
areas can lead to abscess formation including subperiosteal abscesses have been previously published [5,17]. The research ethics boards of
(SPA), collections of purulent fluid deep to the periosteum of the each participating site approved the study. The study was reported ac­
ethmoid, frontal, or maxillary bones, or orbital abscesses in the intra­ cording to the Strengthening the Reporting of Observational Studies in
conal space [3,4]. Early diagnosis and aggressive management of these Epidemiology (STROBE) checklist [18].
severe infections are essential to avoid vision- or life-threatening com­
plications including permanent vision loss, cavernous sinus thrombosis, 2.2. Study population
meningitis, and death [3]. Children with orbital infections often require
hospitalization and intravenous (IV) antibiotics, with some requiring For this analysis, children were included if they underwent cross-
surgical intervention such as abscess drainage and functional endo­ sectional imaging with a computerized tomography (CT) scan which
scopic sinus surgery [5]. In a recent multicentre cohort study of 1579 confirmed the presence of a subperiosteal or orbital abscess. Only pa­
children hospitalized with severe orbital infections, of whom 12% (n = tients with abscess measurements reported in three CT planes (anterior-
189) received surgery, several risk factors for surgical intervention were posterior, transverse, and craniocaudal) were included. Children were
identified, including the presence of subperiosteal or orbital abscess on excluded if they had a primary diagnosis of orbital tumor or pseudotu­
cross-sectional imaging [5]. mor; herpes simplex or herpes zoster; previous craniofacial or ocular
Predictors of surgical intervention in hospitalized patients with surgery; craniofacial abnormality; trauma, laceration, or recent surgery;
orbital cellulitis have been examined by several authors [2,4,6–16]. In acquired or congenital lesion; or immunodeficiency or immunocom­
2000, a seminal study by Garcia and Harris recommended a more con­ promised state. Patients were excluded if they had an abscess in another
servative, nonsurgical approach with IV antibiotics for patients <9 location (e.g., eyelid) or if the CT report did not include abscess mea­
years, provided certain surgical criteria were absent: (1) “large” SPA surements in all three planes. Only patients hospitalized at children’s
size, (2) frontal sinusitis, (3) non-medial abscess location, (4) suspicion hospitals were included, as surgery was not routinely performed at
of anaerobic infection, (5) recurrence, (6) chronic sinusitis, (7) optic community hospitals.
nerve compromise, and (8) infection of dental origin [6]. In subsequent
years, the predictive value of these criteria has been studied with 2.3. Predictors
inconsistent results [2,7–11,14]. Numerous studies have attempted to
elucidate an abscess volume cut-point above which surgical intervention The main predictor was abscess volume, defined as volume in mL.
would be required, ranging widely from 0.5 mL to 3.8 mL on Abscess volume was calculated using the ellipsoid formula used in
cross-sectional imaging [2,7–11]. Three studies reported results of previous studies [2,7]: 4/3 x π x abc where a, b, and c correspond to
receiver operating characteristic (ROC) analysis with sensitivity ranging length in each dimension. For children with >1 abscess identified on
from 71% to 76% and specificity ranging from 80% to 85% [2,7,8]. imaging, the largest abscess was included. If children had >1 CT scan,
These cut-points were derived in single-centre studies conducted at measurements from the first CT were included.
tertiary children’s hospitals with sample sizes ranging from 66 to 108
children. None of these volume cut-points have been evaluated in 2.4. Outcomes
different populations from which they were derived, raising questions
about external validity. The primary outcome was surgical intervention, defined as sub­
Using data from a larger multicentre observational cohort study, we periosteal and/or orbital abscess drainage, or medical intervention only.
sought to determine the predictive value and optimal cut-point of ab­ Both open and endoscopic surgical intervention were included. Surgical
scess volume to predict surgical intervention, defined as abscess intervention is an important marker of disease severity and assessing
drainage, in children hospitalized with an orbital infection and an ab­ need for surgery is the main indication for cross-sectional imaging [5].
scess confirmed on cross-sectional imaging. Our secondary objective Surgical intervention is also the main determinant of length of hospital
was to evaluate the diagnostic performance of previously suggested stay (LOS) [5]. Secondary outcomes included LOS, intensive care unit
abscess volume cut-points in our heterogenous cohort. (ICU) admission, and complications.

2. Methods
2.5. Clinical characteristics
2.1. Study design
Other characteristics collected on all patients included: chronic dis­
ease, antibiotics prior to emergency department (ED) visit, Canadian
This was a secondary analysis of data from a multicentre cohort
Triage and Acuity Scale (CTAS) score in ED which is a marker of disease
study, using hospital records as the data source, from 10 hospitals
severity at initial presentation [19], transfer from community hospital,
participating in the Canadian Pediatric Inpatient Research Network
maximum temperature in ED, clinical signs at initial presentation
(PIRN – www.pirncanada.com). Children were included if they were
(proptosis, eye swollen shut, painful extra-ocular movements, subjective
between 2 months and 18 years of age, hospitalized between January 1,
abnormal vision, ophthalmoplegia), admitting service, and clinical

2
M.F. McKerlie et al. International Journal of Pediatric Otorhinolaryngology 171 (2023) 111629

diagnosis (periorbital cellulitis, orbital cellulitis). Additional data 3. Results

collected included blood culture, other bacterial cultures (ocular
discharge, sinus fluid, abscess aspirate), Chandler classification (III 3.1. Characteristics of the study population
subperiosteal abscess, IV orbital abscess), number of CT and magnetic
resonance imaging (MRI) scans, type of surgical drainage (open, endo­ Of the 412 patients with orbital cellulitis and abscess on CT scan, 150
scopic, both), complications (e.g., thrombosis), ICU admission, death, were included across 6 hospitals (Supplementary Fig. 1). Children
LOS (hours), discharge disposition, return to ED within 30 days, and excluded due to incomplete volumetric measurements (n = 245) were
readmission to hospital. similar to children in the final sample (n = 150). Notable differences in
excluded patients included an older age, fewer transfers from commu­
2.6. Covariates nity hospitals, and fewer children presenting with abnormal clinical
exam findings (Supplementary Table 1). Table 1 describes the study
Patient demographics included child sex (male/female), and age population, including those who received medical intervention only (n
(years, continuous). A male predominance in children presenting with = 82) and surgical intervention (n = 68). The mean age of the total
orbital cellulitis has been well-documented in the literature [2]. Studies cohort was 8.5 years (SD: 4.5), with a male predominance (65.3%) and
have found that simple aerobic infections were more common in chil­ 15.3% (n = 23) had a chronic disease.
dren <9 years of age while there is a trend towards more complex pol­ Children who underwent surgery, compared with those who received
ymicrobial and anaerobic infections with increasing age above 9 years medical intervention only, were older (mean age 9.7 vs. 7.6 years, p =
[6,10]. Patient age was used as a surrogate to control for patient size
relative to abscess volume. We also included abscess location, defined as Table 1
medial or non-medial. Sinus involvement, defined as frontal or Baseline characteristics of children hospitalized with subperiosteal or orbital
non-frontal, was also included as prior studies have suggested frontal abscess.
sinus involvement is associated with surgery, thought to be due to an Variable Total (N Medical (N Surgical (N p-
increased risk of intracranial extension [6,10]. Clinical characteristics = 150) = 82) = 68) value
included proptosis (yes/no), ophthalmoplegia (yes/no), subjective Age in Years, mean (SD) 8.5 ± 4.5 7.6 ± 4.2 9.7 ± 4.5 0.004
abnormal vision (yes/no), and white blood cell count (WBC, categori­ Age (Category), n (%)
cal). Proptosis and ophthalmoplegia are indicators of disease severity 2 months to 5 years 39 (26.0) 25 (30.5) 14 (20.6) 0.17
[16]. Abnormal vision, suggesting potential optic nerve involvement, is 5 to <9 38 (25.3) 28 (34.1) 10 (14.7) 0.01
9 to <14 52 (34.7) 22 (26.8) 30 (44.1) 0.03
a proposed criterion for surgical intervention by Garcia and Harris [6]
14 to 18 21 (14.0) 7 (8.5) 14 (20.6) 0.03
and studies have suggested urgent drainage in the case of visual dete­ Male Sex, n (%) 98 (65.3) 53 (66.2) 45 (66.2) 0.84
rioration [24]. WBC count is a useful objective indicator of inflamma­ Presence of Chronic Disease, 23 (15.3) 11 (13.4) 12 (17.6) 0.48
tion, is a predictor of surgical intervention, and has previously been n (%)
included in other studies [9,11,14]. Respiratory 12 (8.0) 6 (7.3) 6 (8.8) 0.83
Neurologic 1 (0.7) 0 (0.0) 1 (1.5) 0.33
Genetic or Metabolic 1 (0.7) 1 (1.2) 0 (0.0) 0.30
2.7. Statistical analysis Other 12 (8.0) 6 (7.3) 6 (8.8) 0.83
Antibiotic Prior to ED Visit, 97 (64.7) 50 (61.0) 47 (69.1) 0.30
Continuous variables were summarized as medians with inter­ n (%)
CTAS Score in ED, median 3 (2.00, 3.00 (2.00, 2 (2.00,
quartile ranges (IQRs) for variables that were non-normally distributed
(IQR) 3.00) 3.00) 3.00)
or as means with standard deviation (SD). Categorical variables were 1, n (%) 0 (0.0) 0 (0.0) 0 (0.0)
summarized as counts with percentages. P values were generated using a 2, n (%) 45 (43.7) 21 (37.5) 24 (51.1) 0.17
two-sample t-test for means and a two-proportion t-test for counts. 3, n (%) 49 (47.6) 28 (50.0) 21 (44.7) 0.60
We conducted multivariable logistic regression to identify predictors 4, n (%) 6 (5.8) 4 (7.1) 2 (4.3) 0.55
5, n (%) 1 (1.0) 1 (1.8) 0 (0.0) 0.36
of surgical intervention including the covariates described above. Re­
Missing, n (%) 2 2 0
sults were reported as adjusted odds ratios (aORs) with 95% confidence N/A a
47 26 21
intervals (CIs). The model included hospital as a fixed effect to account Transferred from 78 (52.0) 35 (42.7) 43 (63.2) 0.01
for hospital site. Multiple imputation by chained equations was used for Community Hospital, n
(%)
variables with missing data (WBC and abscess location); modelling was
Antibiotics started before 72 (92.3) 33 (94.3) 39 (90.7) 0.55
conducted on each of the 20 imputed datasets and the results were transfer
pooled to generate the final model outputs [20]. We also conducted a Temperature in ED ( C),
◦
37.7 ± 37.8 ± 1.0 37.7 ± 0.8 0.5
sensitivity analysis restricting the model to children with a subperiosteal mean (SD) 0.9
abscess only, excluding those with orbital abscess. ≥ 38 C, n (%) 52 (35.1) 26 (32.5) 26 (38.2) 0.47
Missing 2 2 0
We then conducted a receiver operating characteristic (ROC) anal­
Clinical Signs at Presentation, n (%)
ysis to determine the optimal abscess volume cut-point as a predictor of Proptosis 83 (55.3) 39 (47.6) 44 (64.7) 0.04
surgical intervention. The optimal cut-point was determined using the Eye Swollen Shut 65 (43.3) 32 (39.0) 33 (48.5) 0.24
Youden Index which maximizes both sensitivity and specificity by Painful Extra-Ocular 77 (51.3) 50 (61.0) 27 (39.7) 0.01
Movements
finding the maximal value of (sensitivity + specificity - 1). Youden Index
Abnormal Vision 46 (30.7) 18 (22.0) 28 (41.2) 0.01
values range from zero (poor accuracy) to 1 (perfect accuracy) [21]. We Ophthalmoplegia 61 (40.7) 29 (35.4) 32 (47.1) 0.15
also calculated the area under the curve (AUC), positive predictive Admitting Service, n (%)
value, and negative predictive value with 95% CIs. The sensitivities and Pediatric Medicine 117 72 (87.8) 45 (66.2) 0.002
specificities of abscess volume cut-points previously suggested in the (78.0)
Otolaryngology 26 (17.3) 7 (8.5) 19 (27.9) 0.002
literature were calculated to evaluate diagnostic accuracy in our cohort.
Ophthalmology 1 (0.7) 0 (0.0) 1 (1.5) 0.26
Statistical analysis was performed using R 4.0.2 (R Core Team, 2020) Neurosurgery 2 (1.3) 0 (0.0) 2 (2.9) 0.12
and p-values <0.05 were considered significant. Intensive Care Unit 2 (1.3) 2 (2.4) 0 (0.0) 0.20
Other 2 (1.3) 1 (1.2) 1 (1.5) 0.87

SD = Standard Deviation; IQR, interquartile range; ED = Emergency Depart­

ment’ CTAS = Canadian Triage Acuity Score.
a
Not applicable as patients were not seen in the emergency department.

3
M.F. McKerlie et al. International Journal of Pediatric Otorhinolaryngology 171 (2023) 111629

0.004), more likely to be transferred from a community hospital (63.2% volume compared to those in the medical group (2.6 mL vs. 1.1 mL, p <
vs. 42.7%, p = 0.01), have proptosis (64.7% vs. 47.6%, p = 0.04), have 0.001) and were less likely to have a medial abscess location (45.6% vs.
painful extra-ocular movements (61% vs. 39.7%, p = 0.01), and have 75.6%, p < 0.001). Of the non-medial locations, superior was the most
abnormal vision (41.2% vs. 22%, p = 0.01) at initial presentation. There common in the surgical group (n = 14, 20.6%) followed by super­
were no differences in rates of receiving antibiotics prior to ED pre­ omedial (n = 8, 11.8%). There were no differences in rates of reported
sentation (69.1% vs. 61%, p = 0.30), eye swollen shut (48.5% vs. 39%, p sinus involvement on CT scan between groups. Seventeen patients had
= 0.24), and ophthalmoplegia (47.1% vs. 35.4%, p = 0.15) between an MRI scan completed, most of whom were in the surgical group
groups. Characteristics of the 18 children who presented with abnormal (20.6% vs. 3.7%, p = 0.001).
vision and who did not undergo surgery are described in Supplementary
Table 2.
3.2. Predictors of surgical intervention
Table 2 outlines laboratory and imaging characteristics of included
patients. Children in the surgical group had a larger mean abscess
The univariable and multivariable analyses for predictors of surgical
intervention are shown in Table 3. In multivariable analysis, only ab­
scess volume (aOR 1.46, 95% CI 1.11–1.93) and abscess location (aOR
Table 2 3.46, 95% CI 1.4–8.58) were associated with an increased risk of sur­
Laboratory testing and diagnostic imaging in children hospitalized with sub­
gical intervention. Findings were unchanged when sensitivity analysis
periosteal or orbital abscess.
was restricted to children with subperiosteal abscess only, excluding
Variable Total (N = Medical (N Surgical (N p-value those with orbital abscess.
150) = 82) = 68)

Complete Blood Count, n 149 (99.3) 81 (98.8) 68 (100.0) 0.37

(%) 3.3. Abscess volume cut-point
WBC Cell Count (x109/L), 14.1 ± 5.0 13.9 ± 4.9 14.4 ± 5.1 0.54
mean (SD) ROC curve analysis demonstrated that abscess volume predicted
Missing 3 (2.0) 1 (1.2) 2 (2.9) surgical intervention with an AUC of 0.75 (95% CI 0.67–0.83) (Fig. 1).
WBC Cell Count (x109/L), n (%)
<5 0 (0.0) 0 (0.0) 0 (0.0)
The abscess volume cut-point with the highest optimized Youden Index
5–12 58 (38.7) 30 (36.6) 28 (41.2) 0.56 was 1.18 mL (Youden index 0.45) with a sensitivity of 66% (95% CI
>12–20 71 (47.3) 41 (50.0) 30 (44.1) 0.47 55%–77%), specificity of 79% (95% CI 70%–88%), positive predictive
>20 17 (11.3) 9 (11.1) 8 (11.8) 0.88 value of 74% (95% CI 61%–83%), and negative predictive value of 74%
Blood Culture, n (%) 109 (72.7) 63 (76.8) 46 (67.6) 0.21
(95% CI 65%–83%). (Supplementary Fig. 2). When the analysis was
Positive, n (%) 10 (9.2) 3 (4.8) 7 (15.2) 0.06
Other Cultures, n (%) 76 (50.7) 18 (22.0) 58 (85.3) <0.001 restricted to only patients with a subperiosteal abscess, the results were
Ocular Discharge 15 (19.7) 6 (33.3) 9 (15.5) 0.10 unchanged.
Positive 9 (60.0) 3 (50.0) 6 (66.7) 0.53 In multivariable analysis using the newly identified abscess volume
Sinus Fluid 19 (25.0) 1 (5.6) 18 (31.0) 0.03 cut-point of 1.18 mL, patients with an abscess ≥1.18 mL had a signifi­
Positive 15 (78.9) 1 (100.0) 14 (77.8) 0.60
cant increased risk of surgical intervention with aOR 7.82 (95% CI
Abscess Aspirate 37 (48.7) 1 (5.6) 36 (62.1) <0.001
Positive 29 (78.4) 1 (100.0) 28 (77.8) 0.60 2.96–20.67) (Supplementary Table 3).
Other 30 (39.5) 11 (73.3) 18 (31.0) 0.01 The diagnostic performance of published abscess volumes as surgical
Number of CT Scans, 1.2 ± 0.5 1.1 ± 0.3 1.4 ± 0.7 0.001 predictors in our cohort is shown in Table 4 with sensitivity and speci­
mean (SD)
ficity values ranging from 25.4% to 85.9% and 41.8%–94.9%,
Number with >1 CT 29 (19.3) 7 (8.5) 22 (32.4) <0.001
scans, n (%) respectively.
Sinus Involvement on CT 147 (98.0) 81 (98.8) 66 (97.1) 0.46
Scan, n (%)a Table 3
Maxillary 104 (70.7) 60 (74.1) 44 (66.7) 0.33
Multivariable analysis of predictors associated with surgical intervention with
Ethmoid 106 (72.1) 61 (75.3) 45 (68.2) 0.34
fixed effects for hospital site.
Sphenoid 48 (32.7) 30 (37.0) 18 (27.3) 0.21
Frontal 44 (29.9) 23 (28.4) 21 (31.8) 0.65 Characteristic Unadjusted Adjusted
Mastoid 10 (6.8) 6 (7.4) 4 (6.1) 0.75
OR 95% CI p-value aOR 95% CI p-
Diffuse Paranasal 37 (25.2) 16 (19.8) 21 (31.8) 0.09
value
Abscess on CT Scan
Volume (mL), median 0.9 (0.5, 0.6 (0.3, 1.7 (0.8, Abscess Volume (mL) 1.68 1.33, 0.01 1.46 1.11, 0.01
(IQR) 2.4) 1.1) 3.9) 2.14 1.93
Volume (mL), mean 1.8 ± 2.2 1.1 ± 1.5 2.6 ± 2.5 <0.001 Sex (Male) 0.97 0.46, 0.80 1.02 0.42, 0.95
(SD) 1.82 2.49
>1 Abscess 7 (4.7) 3 (3.7) 4 (5.9) Age (years) 1.16 1.07, <0.001 1.09 0.97, 0.12
Abscess Location on CT Scan, n (%) 1.27 1.24
Medial 93 (62.0) 62 (75.6) 31 (45.6) <0.001 Medial Abscess 5.16 2.36, <0.001 3.46 1.40, 0.01
Superomedial 11 (7.3) 3 (3.7) 8 (11.8) 0.06 Location 11.27 8.58
Inferomedial 8 (5.3) 4 (4.9) 4 (5.9) 0.79 Frontal Sinus 1.46 0.68, 0.33 0.69 0.26, 0.48
Inferior 2 (1.3) 1 (1.2) 1 (1.5) 0.88 Involvement 3.11 1.89
Superior 20 (13.3) 6 (7.3) 14 (20.6) 0.02 Proptosis 2.52 1.14, 0.02 1.83 0.74, 0.19
Superolateral 7 (4.7) 1 (1.2) 6 (8.8) 0.03 5.01 4.55
Lateral 5 (3.3) 2 (2.4) 3 (4.4) 0.51 Abnormal Vision 3.18 1.46, 0.003 1.66 0.61, 0.31
Other 0 (0.0) 0 (0.0) 0 (0.0) 6.91 4.54
Missing 4 (2.7) 3 (3.7) 1 (1.5) Ophthalmoplegia 1.56 0.68, 0.29 0.72 0.2, 0.54
Chandler Criteria, n (%) 3.54 2.04
III (subperiosteal 123 (82.0) 73 (89.0) 50 (73.5) 0.01 WBC (Ref ≤12 × 109/L)
abscess) >12–20 0.88 0.42, 0.73 0.64 0.26, 0.34
IV (orbital abscess) 27 (18.0) 9 (11.0) 18 (26.5) 0.01 1.83 1.59
MRI Scan, n (%) 17 (11.3) 3 (3.7) 14 (20.6) 0.001 >20 0.85 0.27, 0.77 0.69 0.18, 0.61
2.63 2.78
WBC, white blood cell; SD, standard deviation; CT, Computed Tomography;
MRI, Magnetic Resonance Imaging. OR, Odds Ratio; aOR, adjusted Odds Ratio; CI, Confidence Interval; WBC, white
a
Patients may have had evidence of sinus involvement in multiple locations. blood cell.

4
M.F. McKerlie et al. International Journal of Pediatric Otorhinolaryngology 171 (2023) 111629

MRI. No child died. Three children in the medical group had a return
visit to the ED related to orbital cellulitis within 30 days and two were
readmitted. There were no readmissions in the surgical group.

4. Discussion

We examined 150 children hospitalized across multiple children’s

hospitals with orbital cellulitis with an abscess confirmed on cross-
sectional imaging, one of the most important complications of these
infections. We examined predictors of surgical intervention, defined as
abscess drainage, and sought to determine a predictive volume cut-point
guiding surgical vs. medical management. Using multivariable regres­
sion, adjusting for important covariates and known predictors of surgi­
cal intervention, only larger abscess volume and non-medial abscess
location were associated with surgical intervention. The abscess volume
cut-point with the greatest diagnostic accuracy for surgical intervention
was 1.18 mL.
Despite considerable attention in the literature, the management of
these infections remains controversial, including the indications for
surgical intervention [2,4,6–16]. “Large” abscess size and non-medial
location are well documented predictors of surgical drainage, with
more recent small single centered studies suggesting objective volume
criterion, ranging from 0.5 mL to 3.8 mL [2,7–11], although none have
been externally validated. Our study is the first to use a large cohort of
children from multiple hospitals to derive and validate the predictive
value of abscess volume in patients with a subperiosteal or orbital ab­
scess. An objective volume criterion measurement is a useful clinical tool
Fig. 1. Receiver operating characteristic curve for abscess volume as a pre­ to identify patients with abscess who may require surgical drainage.
dictor of surgical intervention.
Using ROC analysis, we determined an optimal volume cut-point of 1.18
Receiver operating characteristic (ROC) curve analysis demonstrating the as­
mL. Using the Youden Index, sensitivity was optimized at 66% and
sociation between abscess volume and likelihood of surgical intervention. Using
logistic regression, the area under the curve is 0.75. Sensitivity and specificity
specificity at 79%. Prior studies documented higher sensitivity and
are maximized at 66% and 79% respectively at an abscess volume cut-point of specificity values ranging from 71 to 76% and 80–85%, respectively.
1.18 mL. However, when the diagnostic performance of these cut-points was
evaluated in our cohort, there were considerable trade-offs in sensitivity
3.4. Clinical outcomes or specificity. These differences may be explained by the heterogeneity
of our sample compared to smaller, single-centre studies, which are
Surgical intervention and clinical outcomes are shown in Supple­ prone to bias. For example, within our study, we observed a tendency
mentary Table 4 and 5 Of the 68 patients who underwent surgery, 38.2% towards surgical drainage of smaller abscesses at three sites, which may
(n = 26) had orbital abscess drainage, 66.2% (n = 45) had subperiosteal reflect a lower threshold for surgery.
abscess drainage, and 4.4% (n = 3) had both. Most (n = 44, 64.7%) were We used several measures of diagnostic performance to determine
completed using an open technique and 16 (23.5%) had both open and the optimal abscess volume cut-point, including the Youden Index,
endoscopic drainage. Supplementary Table 6 shows type of surgical which weighs sensitivity and specificity equally [21]. Given both the
intervention by abscess location. morbidity/mortality associated with subperiosteal/orbital abscess, as
Children who underwent surgery, compared with those who received well as the risks associated with surgical intervention, we wanted to
medical care only, had a higher rate of complications (10.3% vs. 3.7%) balance false negatives (FN) with false positives (FP). In our cohort, the
and longer median hospital stays (7.7 days vs. 4.6 days). In the surgical positive likelihood ratio was 3.18 and the negative likelihood ratio was
group, the most common complication was intracranial extension (n = 0.43, meaning 17 children with abscesses ≥1.18 mL did not undergo
5, 7.4%) followed by vision loss (n = 2, 2.9%). One child developed surgery (FP) while 23 children with abscesses <1.18 mL did undergo
osteomyelitis (1.5%), one child developed an epidural abscess and tri­ surgery (FN). These cases highlight that abscess characteristics are only
geminal nerve hypoesthesia, one child developed a subdural empyema, one component to surgical planning, and it is important to consider
and one child was found to have non-specific dural enhancement on other clinical characteristics, such as disease course (e.g., response to
antibiotics), and the values and preferences of families (e.g., watchful

Table 4
Diagnostic performance of published abscess volumes in cohort.
Paper Optimal Cutoff (mL) Original Diagnostic Diagnostic Accuracy in Cohort
Accuracy

Sensi-tivity Speci-ficity Accuracy (95% CI)a Sensitivity (95% CI) Specificity (95% CI) PPV (95% CI) NPV (95% CI)

McKerlie 1.18 – – 0.73 (0.66, 0.80) 0.66 (0.55, 0.77) 0.79 (0.70, 0.88) 0.74 (0.61, 0.83) 0.74 (0.65, 0.83)
Aryasit 1.5 0.71 0.80 0.71 (0.64, 0.78) 0.57 (0.46, 0.69) 0.82 (0.73–0.90) 0.72 (0.60, 0.84) 0.70 (0.61, 0.80)
Le 3.8 0.76 >0.85 0.63 (0.55, 0.71) 0.26 (0.16, 0.37) 0.94 (0.89, 0.99) 0.78 (0.61, 0.95) 0.61 (0.52, 0.69)
McCoy 0.51 0.71 0.84 0.61 (0.53, 0.69) 0.85 (0.77, 0.94) 0.40 (0.29, 0.51) 0.54 (0.45, 0.64) 0.77 (0.64, 0.89)
Nation 0.50 N/A N/A 0.60 (0.52, 0.68) 0.85 (0.77, 0.94) 0.39 (0.28, 0.49) 0.54 (0.44, 0.63) 0.76 (0.63, 0.89)
Todman 1.25 N/A N/A 0.73 (0.66, 0.80) 0.65 (0.53, 0.76) 0.79 (0.70, 0.88) 0.74 (0.61, 0.83) 0.73 (0.64, 0.82)

CI, confidence interval; PPV, positive predictive value; NPV, negative predictive value.
a
Accuracy defined as value at the cut-point: ([true positives + true negatives]/total number of patients).

5
M.F. McKerlie et al. International Journal of Pediatric Otorhinolaryngology 171 (2023) 111629

waiting vs early intervention). Funding

We found that larger abscess volume and non-medial abscess loca­
tion were significant predictors of surgical intervention. While abscess The authors do not have any specific grant funding for this research
volume has been reported in other studies, the evidence supporting from any commercial, public, or not-for-profit enterprise.
abscess location has been inconsistent [2,8,11]. While Ryan et al. found
that location was not associated with surgery, according to McCoy et al. Declaration of competing interest
non-medial location is likely an independent risk factor predicting sur­
gical intervention [8,11]. Our multivariable model, adjusting for PJG has received grants from the Canadian Institutes of Health
important patient and clinical characteristics, highlights the importance Research (CIHR), the PSI Foundation, and The Hospital for Sick Chil­
of abscess location in clinical decision making. Both large volume and dren. He has received nonfinancial support from the EBMLive Steering
non-medial abscesses are more likely to fail medical management due to Committee (expenses reimbursed to attend conferences) and the CIHR
poor antibiotic penetration secondary to thicker abscess cavities and Institute of Human Development, Child and Youth Health (as a member
sequestration in avascular regions [10]. Prior studies have also sug­ of the institute advisory board, expenses reimbursed to attend meet­
gested that the larger sinuses with comparatively smaller ostia of older ings). He is a member of the CMAJ Open and BMJ Evidence Based
children harbour more complex, polymicrobial infections and patient Medicine Editorial Board. PCP has received grants from the Hospital for
age ≥9 years seems to be an established indicator for surgical drainage Sick ChildrenFoundation (SP05-602), Canadian Institutes of Health
[10,22,23,25]. However, in our study, there was no association with Research (FRN # 115059), and non-financial support for an
older age or sinus location with surgical intervention. It is possible that investigator-initiated trial for which Mead Johnson Nutrition provides
changes in antibiotic use over time with broader coverage have reduced non-financial support (Fer-In-Sol® liquid iron supplement)
the importance of virulence and therefore the risk stratification of pa­ (2011–2017). OD was supported by a Chercheur Boursier Clinicien
tients according to age. Other variables previously suggested to increase Award, from the Fonds de recherche du Québec – Santé. GW has
surgical risk were not observed in our study. Clinical features at initial received grants from the Canadian Institutes of Health Research and the
presentation, including abnormal vision, ophthalmoplegia, and prop­ Hamilton Health SciencesFoundation.
tosis, which are considered markers of orbital cellulitis, did not increase
the risk of surgery. In patients that have CT confirmed subperiosteal or Appendix A. Supplementary data
orbital abscess, abscess volume and location are key drivers for
decision-making. Supplementary data to this article can be found online at https://doi.
Our study has several limitations. First, our study is limited by its org/10.1016/j.ijporl.2023.111629.
retrospective nature. To calculate abscess volume, we required di­
mensions reported in three CT planes and 245 participants were References
excluded because of incomplete measurements. We compared charac­
teristics of those included in the study with those excluded, and found [1] K.J. Williams, R.C. Allen, Paediatric orbital and periorbital infections, Curr. Opin.
Ophthalmol. 30 (5) (2019) 349–355, https://doi.org/10.1097/
the populations were similar overall. Second, due to inconsistencies in ICU.0000000000000589.
CT reports, we were unable to classify dimensions according to CT plane [2] T.D. Le, E.S. Liu, F.A. Adatia, J.R. Buncic, S. Blaser, The effect of adding orbital
(i.e., anterior-posterior, transverse, and craniocaudal). Additional in­ computed tomography findings to the Chandler criteria for classifying pediatric
orbital cellulitis in predicting which patients will require surgical intervention,
formation about the shape of the abscess relative to the orbit may have J AAPOS 18 (3) (2014) 271–277, https://doi.org/10.1016/j.jaapos.2014.01.015.
provided useful information. Third, as the primary outcome was surgical [3] A. Danishyar, S.R. Sergent, Orbital cellulitis, in: StatPearls, StatPearls Publishing,
intervention defined by abscess drainage, we assumed that abscess 2022. http://www.ncbi.nlm.nih.gov/books/NBK507901/. (Accessed 28 August
2022).
volume was a consideration for surgery, however patient-specific in­
[4] N. Tanna, D.A. Preciado, M.S. Clary, S.S. Choi, Surgical treatment of subperiosteal
dications for surgery were unknown. Finally, we calculated abscess orbital abscess, Arch. Otolaryngol. Head Neck Surg. 134 (7) (2008) 764–767,
volume using the ellipsoid formula used in previous studies, however https://doi.org/10.1001/archotol.134.7.764.
[5] P.J. Gill, O. Drouin, C. Pound, et al., Factors associated with surgery and imaging
measurements may have varied between hospitals and radiologists, and
characteristics in severe orbital infections, J. Pediatr. (2022), https://doi.org/
may not be an ideal representation of actual volume. 10.1016/j.jpeds.2022.05.010.
[6] G.H. Garcia, G.J. Harris, Criteria for nonsurgical management of subperiosteal
5. Conclusion abscess of the orbit: analysis of outcomes 1988-1998, Ophthalmology 107 (8)
(2000) 1454–1458, https://doi.org/10.1016/s0161-6420(00)00242-6.
[7] O. Aryasit, S. Aunruan, N. Sanghan, Predictors of surgical intervention and visual
We studied a large cohort of 150 children hospitalized with orbital outcome in bacterial orbital cellulitis, PMID: 34160383; PMCID: PMC8238295,
cellulitis with subperiosteal and orbital abscess across 6 children’s Medicine (Baltim.) 100 (25) (2021 Jun 25), e26166, https://doi.org/10.1097/
MD.0000000000026166.
hospitals. A total of 82 children were treated conservatively with sys­ [8] J.L. McCoy, R. Dixit, J.E. Dohar, A.B.J. Tobey, Pediatric subperiosteal orbital
temic antibiotics while 68 children underwent surgical drainage. Larger abscess characterization and prediction of size, location, and management, Int. J.
abscess volume and non-medial abscess location were significant pre­ Pediatr. Otorhinolaryngol. 144 (2021), 110693, https://doi.org/10.1016/j.
ijporl.2021.110693.
dictors of receiving surgery. An abscess volume of ≥1.18 mL was the [9] J. Nation, A. Lopez, N. Grover, D. Carvalho, D. Vinocur, W. Jiang, Management of
optimal cut-point for surgical drainage with a sensitivity of 66% and large-volume subperiosteal abscesses of the orbit: medical vs surgical outcomes,
specificity of 79%. Otolaryngol. Head Neck Surg. 157 (5) (2017) 891–897, https://doi.org/10.1177/
0194599817728490.
[10] M.S. Todman, Y.R. Enzer, Medical management versus surgical intervention of
Additional contributions pediatric orbital cellulitis: the importance of subperiosteal abscess volume as a new
criterion, Ophthalmic Plast. Reconstr. Surg. 27 (4) (2011) 255–259, https://doi.
org/10.1097/IOP.0b013e3182082b17.
Additional staff who assisted with the study include: Hospital for Sick
[11] J.T. Ryan, D.A. Preciado, N. Bauman, et al., Management of pediatric orbital
Children, Toronto, ON: Nurshad Begum, Dana Arafeh, Carol Chan; CHU cellulitis in patients with radiographic findings of subperiosteal abscess,
Sainte-Justine, Montreal, QC: Camille Bédard-Gauthier; Trillium Health Otolaryngol. Head Neck Surg. 140 (6) (2009) 907–911, https://doi.org/10.1016/j.
Partners, Mississauga, ON: Yasmeen Razvi, Joanna Matthews; Alberta otohns.2009.02.014.
[12] E.A. Adil, M.E. Muir, K. Kawai, N.D. Dombrowski, M.J. Cunningham, Pediatric
Children’s Hospital, Calgary, AB: Olivia Genereux, Jovey Sharma, Jes­ subperiosteal abscess secondary to acute sinusitis: a systematic review and meta-
sica Cooper, Natalie Czuczman, Rouzbeh Ghadiry-Tavi; and Lakeridge analysis, Laryngoscope 130 (12) (2020) 2906–2912, https://doi.org/10.1002/
Health, Oshawa, ON: Jack Warren, Natascha Kozlowski. lary.28570.
[13] M.A. Dewan, D.R. Meyer, E.J. Wladis, Orbital cellulitis with subperiosteal abscess:
demographics and management outcomes, Ophthalmic Plast. Reconstr. Surg. 27
(5) (2011) 330–332, https://doi.org/10.1097/IOP.0b013e31821b6d79.

6
M.F. McKerlie et al. International Journal of Pediatric Otorhinolaryngology 171 (2023) 111629

[14] H. Gavriel, E. Yeheskeli, E. Aviram, L. Yehoshua, E. Eviatar, Dimension of [20] Ian R. White, Patrick Royston, Angela M. Wood, Multiple imputation using chained
subperiosteal orbital abscess as an indication for surgical management in children, equations: issues and guidance for practice - white - 2011 - statistics in medicine -
Otolaryngol. Head Neck Surg. 145 (5) (2011) 823–827, https://doi.org/10.1177/ wiley online library. https://onlinelibrary.wiley.com/doi/abs/10.1002/sim.4067.
0194599811416559. (Accessed 19 September 2022).
[15] R. Rahbar, C.D. Robson, R.A. Petersen, et al., Management of orbital subperiosteal [21] Ana-Maria Šimundić, Measures of diagnostic accuracy: basic definitions - PubMed.
abscess in children, Arch. Otolaryngol. Head Neck Surg. 127 (3) (2001) 281–286, https://pubmed.ncbi.nlm.nih.gov/27683318/. (Accessed 19 September 2022).
https://doi.org/10.1001/archotol.127.3.281. [22] G.J. Harris, Age as a factor in the bacteriology and response to treatment of
[16] S.J. Wong, J. Levi, Management of pediatric orbital cellulitis: a systematic review, subperiosteal abscess of the orbit - PubMed. https://pubmed.ncbi.nlm.nih.gov
Int. J. Pediatr. Otorhinolaryngol. 110 (2018) 123–129, https://doi.org/10.1016/j. /8140703/. (Accessed 19 September 2022).
ijporl.2018.05.006. [23] G.J. Harris, Subperiosteal abscess of the orbit. Age as a factor in the bacteriology
[17] P.J. Gill, P.C. Parkin, N. Begum, et al., Care and outcomes of Canadian children and response to treatment, Ophthalmology 101 (3) (1994) 585–595, https://doi.
hospitalised with periorbital and orbital cellulitis: protocol for a multicentre, org/10.1016/s0161-6420(94)31297-8.
retrospective cohort study, BMJ Open 9 (12) (2019), e035206, https://doi.org/ [24] V. Sciarretta, M. Demattè, P. Farneti, M. Fornaciari, I. Corsini, O. Piccin,
10.1136/bmjopen-2019-035206. D. Saggese, I.J. Fernandez, Management of orbital cellulitis and subperiosteal
[18] E. von Elm, D.G. Altman, M. Egger, et al., Strengthening the Reporting of orbital abscess in pediatric patients: a ten-year review, Int. J. Pediatr.
Observational Studies in Epidemiology (STROBE) statement: guidelines for Otorhinolaryngol. 96 (2017 May) 72–76.
reporting observational studies, BMJ 335 (7624) (2007) 806–808, https://doi.org/ [25] F. Tabarino, M. Elmaleh-Bergès, S. Quesnel, M. Lorrot, T. Van Den Abbeele,
10.1136/bmj.39335.541782.AD. N. Teissier, Subperiosteal orbital abscess: volumetric criteria for surgical drainage,
[19] M.J. Bullard, T. Chan, C. Brayman, et al., Revisions to the Canadian emergency Int. J. Pediatr. Otorhinolaryngol. 79 (2) (2015 Feb) 131–135.
department triage and acuity Scale (CTAS) guidelines, Can. J. Emerg. Med. 16 (6)
(2014) 485–489, https://doi.org/10.1017/S148180350000350X.