YJPSU-58873; No of Pages 6

Journal of Pediatric Surgery xxx (xxxx) xxx–xxx

Contents lists available at ScienceDirect

Journal of Pediatric Surgery

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

A novel noninvasive appendicitis score with a urine biomarker☆,☆☆

Te-Lu Yap a,⁎, Jing Dan Fan b, Yong Chen a, Matt Fat Ho b, Candy SC Choo a, John Allen c, Yee Low a,
Anette Sundfor Jacobsen a, Shireen Anne Nah a
a
Department of Paediatric Surgery, KK Women's & Children's Hospital, Singapore
b
National Dental Centre, Singapore
c
Duke-NUS Graduate Medical School, Singapore

a r t i c l e i n f o a b s t r a c t

Article history: Purpose: The aim of our study was to develop an appendicitis score incorporating a urine biomarker, Leucine rich
Received 17 September 2018 alpha-2-glycoprotein (LRG), for evaluation of children with abdominal pain.
Accepted 1 October 2018 Methods: From January to August 2017 we prospectively enrolled children aged 4–16 years old admitted for
Available online xxxx suspected appendicitis. Urine samples for LRG analysis were obtained preoperatively and quantified by
enzyme-linked immunosorbent assay (ELISA) after correction for patient hydration status. The diagnosis of ap-
Key words:
pendicitis was based on operative findings and histology. Logistic regression was used to identify prospective
Pediatric
Diagnosis
predictors.
Appendicitis score Results: A total of 148 patients were recruited, of which 42(28.4%) were confirmed appendicitis. Our Appendicitis
Imaging Urinary Biomarker (AuB) model incorporated urine LRG with 3 clinical predictors: ‘constant pain’, ‘right iliac
Urine LRG fossa tenderness’, ‘pain on percussion’. Area under the ROC curve for AuB was 0.82 versus 0.78 for the Pediatric
Appendicitis Score (PAS) on the same cohort of patients. A model-calculated risk score of b 0.15 is interpreted
as low risk of appendicitis. Sensitivity for the AuB at this cutoff was 97.6%, specificity 37.7%, negative predictive
value 97.6%, positive predictive value 38.3%, and negative likelihood ratio 0.06.
Conclusion: The noninvasive AuB score appears promising as a diagnostic tool for excluding appendicitis in chil-
dren without the need for blood sampling.
Type of study: Study of diagnostic test.
Level of evidence: Level III.
© 2018 The Authors. Published by Elsevier Inc. This is an open access article under the CC BY-NC-ND license
(http://creativecommons.org/licenses/by-nc-nd/4.0/).

Despite being the most common surgical emergency in children, ac-
curate diagnosis of acute appendicitis (AA), or its exclusion, remains ar-
duous and often costly [1]. This is due to the considerable overlap in the
clinical features of AA and other common pediatric ailments. In addition,
the presentation of AA can be heterogeneous and atypical especially in
younger patients [2]. This is evidenced by a misdiagnosis rate of 28%–
57% at initial presentation [3–5]. The massive denominator of patients
☆ Author contributions: • Study conception and design: TL Yap, Yong Chen, MF Ho, SA
Nah, John Allen. • Data acquisition: TL Yap, JD Fan, CSC Choo, Yong Chen, MF Ho, SA Nah.
• Analysis and data interpretation: TL Yap, JD Fan, CSC Choo, Yong Chen, MF Ho, SA Nah,
John Allen. • Statistical analysis and construction of the AuB score: John Allen. • Drafting
of the manuscript: TL Yap. • Critical revision of manuscript: TL Yap, John Allen, CSC Choo,
Yong Chen, Yee Low, AS Jacobsen, MF Ho, SA Nah.
☆☆ How this paper will improve care: The Appendicitis Urinary Biomarker model is the
first noninvasive appendicitis score combining three clinical variables with urinary Leu-
cine rich alpha-2-glycoprotein. Its promising performance as a negative predictor of ap-
pendicitis may aid in optimizing imaging usage and reducing hospital admissions.
⁎ Corresponding author at: Department of Paediatric Surgery, KK Women's & Children's
Hospital, 100 Bukit Timah Road, Singapore 229899.
E-mail address: yap.te.lu@singhealth.com.sg (T.-L. Yap).
with abdominal pain, which is a leading cause of Emergency
Department's attendance, presents a formidable challenge. Each year
in the United States, 800,000 children aged 15 years and less presented
at EDs with abdominal pain, but only 72,000 – less than 10% – were
eventually confirmed with appendicitis [6,7].
Traditionally, patient history, physical examination, laboratory
hematological markers and the Pediatric Appendicitis Score (PAS)
have been utilized in diagnosing AA, but with limited accuracy
[8,9]. This has led to heavy reliance on radiological modalities of ul-
trasound, computed tomography (CT) scan and even magnetic reso-
nance imaging. [10–12]. However, there is rising concern regarding
exposure of children to radiation from CT scans, while diagnostic ac-
curacy of ultrasound is heavily operator dependent. In addition, ex-
tensive use of diagnostic imaging has been cited as a major factor
in the increased cost burden for hospitals [13]. Therefore, a risk-
stratifying strategy employing a noninvasive ‘screening’ tool to iden-
tify patients with low likelihood of AA would allow hospitalization
and imaging resources to be effectively concentrated on a selected,
smaller patient population.

https://doi.org/10.1016/j.jpedsurg.2018.10.025
0022-3468/© 2018 The Authors. Published by Elsevier Inc. This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/).

Please cite this article as: Yap T-L, et al, A novel noninvasive appendicitis score with a urine biomarker, J Pediatr Surg (2018), https://doi.org/
10.1016/j.jpedsurg.2018.10.025
2 T-L. Yap et al. / Journal of Pediatric Surgery xxx (xxxx) xxx–xxx
Leucine rich alpha-2 glycoprotein (LRG) is a novel protein biomarker
recently identified in the appendices, serum and urine of patients with
appendicitis [14,15]. Its precise physiological function is still largely un-
known but it functions as an acute-phase protein expressed by differen-
tiating neutrophils, hepatocytes and venules of the mesentery such as
the mesoappendix [16]. It is upregulated during inflammatory condi-
tions especially bacterial infection.
Therefore, we hypothesized that urine LRG when combined with ap-
propriate clinical variables would be a useful negative predictor of acute
appendicitis in children. The aim of our study was to develop a noninva-
sive appendicitis score comprising urine LRG and clinical variables to
rule-out appendicitis in children with acute abdominal pain.
1. Materials and methods
1.1. Study design and setting
This was a prospective, blinded, observational cohort study in chil-
dren admitted to our hospital for suspected appendicitis. The study
was conducted in an urban, tertiary children's hospital from January
2017 to August 2017. We obtained our hospital Institutional Research
Ethics Board's approval prior to the study initiation. Parents or legal
guardians of all participating patients provided written informed con-
sents and assents were obtained from children 7 years and older.
1.2. Study population
All patients between the ages of 4 and 16 years, admitted from our
Children's Emergency department to our surgical wards, with acute ab-
dominal pain and signs and symptoms suggesting of possible appendi-
citis were eligible for enrollment. The exclusion criteria were previous
appendectomy, other abdominal surgery, pregnancy, chronic medical
(e.g. inflammatory bowel disease) or malignant conditions. Eligible pa-
tients were identified by the on duty residents and approached by our
study research coordinators primarily during weekdays, from 8 AM to
8 PM.
1.3. Study protocol
The patients were assessed and managed by the attending surgeons
according to standard departmental protocol for suspected appendicitis.
Routine blood and urine tests and, if deemed necessary, ultrasound or
CT scan, or both were requested. The patients' information collected in-
cluded: demographic characteristics (age and gender), duration of ill-
ness, symptoms (characteristics and location of pain, aggravation of
pain by movement or coughing, anorexia, nausea, vomiting, fever,
stool frequency, urinary symptoms) and clinical signs (temperature, lo-
calized tenderness, rebound tenderness, pain on percussion/coughing/
hopping, Psoas sign, Rovsing's sign, guarding, bowel sound characteris-
tics). Laboratory investigations included total white cells counts (WBC),
neutrophil percentage (% Neu), and C-reactive protein (CRP). Radiolog-
ical examinations included ultrasound and/or CT scans of the abdomen.
1.4. Urine LRG collection and analysis
Midstream, clean catch urine specimens for LRG analysis were col-
lected preoperatively or within 24 h of admission. The samples were im-
mediately placed on ice and transported to the laboratory within 30 min
of collection. The pH, color, and turbidity of each urine specimen were
recorded. Urine samples were centrifuged at 3000 × g for 10 min at
4 °C to remove cell debris. The supernatant was collected and protease
inhibitors (Complete Mini, Roche, Basel, Switzerland) added to prevent
protein degradation. All the samples were stored at − 80 °C until
analysis.
Quantitative analysis of urinary LRG levels was performed using
enzyme-linked immunosorbent assay (ELISA) (IBL International,
Please cite this article as: Yap T-L, et al, A novel noninvasive appendicitis score with a urine biomarker, J Pediatr Surg (2018), https://doi.org/
10.1016/j.jpedsurg.2018.10.025
Takara, Japan) according to the manufacturer's recommendations. The
urinary LRG concentration was normalized to urinary creatinine con-
centration. Urinary creatinine was determined by Jaffe reaction using a
commercially available parameter assay kit (R&D Systems, Minneapolis,
MN). The quantity of LRG was reported as LRG/urine Creatinine (μCr)
ratio (g/mol).
1.5. Measurements
WBC and neutrophil counts were measured according to our hospi-
tal standard laboratory methods using a Sysmex XE-5000 analyzer. CRP
levels were measured by the Abbott Architect c8000 analyzer.
Operative findings at appendectomy (presence of fecolith, perfora-
tion, abscess or mass) as well as histological diagnosis of the appendec-
tomy specimens were documented.
Patients' clinical features at admission were initially entered into a
study form and subsequently collated with laboratory and progress in-
formation from the electronic record system. Research coordinators
then entered the data into an Excel format for further statistical analysis.
All data were double-checked for accuracy by one author (TLY).
1.6. Clinical outcomes
The diagnosis of appendicitis was based on histological findings of
mucosal and transmural inflammatory infiltrates. The presence of
fecolith either at surgery or in the resected appendix alone did not con-
stitute appendicitis. The diagnosis of perforated appendicitis was re-
served for appendixes which on histology were gangrenous with
perforation or the operative findings of abscess and appendiceal mass.
The severity of inflammation was classified by intraoperative and histo-
logical findings into acute or perforated.
In nonoperated patients, the diagnosis of nonappendicitis was made
after patients were monitored in the ward for a period of at least 24 h or
until resolution of the symptom of abdominal pain. This was done per
our department's practice protocol. In addition, our research coordina-
tors would follow-up with a telephone consult at 2 weeks postdischarge
to ensure that patients had not been diagnosed with AA in another
hospital.
All surgeons managing the patients and pathologists examining the
appendix specimens were blinded to the results of the AuB score. In ad-
dition, the scientists analyzing the urine LRG were also blinded to the
final diagnosis of the patients.
1.7. Statistical analysis
For statistical analysis, normally distributed baseline continuous var-
iables were summarized as mean and standard deviation, and
nonnormal variables as median and range. Categorical variables were
expressed as absolute values and percentages. Appendicitis and
nonappendicitis groups were compared on baseline variables using
the Mann–Whitney U test or t-test as appropriate for continuous base-
line variables and Fisher's exact test for categorical variables. A p-value
of b0.05 was considered statistically significant.
Stepwise logistic regression with significance levels to enter and stay
of 0.20 and 0.25, respectively, was used to identify predictors in devel-
oping the AuB score model. Receiver operator characteristic (ROC) anal-
ysis and area under the curve (AUC) were used to assess diagnostic
performance. Youden's rule was used to identify the cutpoint for the
AuB risk score. Given a diagnostic test cutpoint, sensitivity, specificity,
negative predictive value, positive predictive value, negative and
positive likelihood ratio with 95% confidence interval (95% CI) were ob-
tained from 2 × 2 tables of test performance frequency counts (diagno-
sis vs actual outcome). Data analysis was performed using SPSS (SPSS
Inc., Chicago, IL) and SAS V9.4 (SAS Inc., Cary, NC USA) software.
The diagnostic performance of the AuB score was compared with the
PAS score at a cutoff of 4 (b4 diagnosed as low risk). Based on recent
 T-L. Yap et al. / Journal of Pediatric Surgery xxx (xxxx) xxx–xxx 3

Table 1
Demographic and clinical characteristics.

Nonappendicitis Appendicitis P-value

Total, n 106 (71.6%) 42 (28.4%)

Gender, Male 51 (48.1%) 30 (71.4%) 0.011
Age (year) mean (S.D) 10.3 (3.2) 10.3 (2.9) 0.915
Clinical characteristics
Fever (≥38 °C), n 32 (30.2%) 13 (31.0%) 0.928
Anorexia, n 40 (37.7%) 26 (61.9%) 0.008
Nausea, n 55 (51.9%) 27 (64.3%) 0.173
Migration of pain, n 23 (21.7%) 17 (40.5%) 0.21
RIF tenderness, n 57 (53.8%) 35 (83.3%) b0.001
Pain on percussion, n 5 (4.7%) 12 (28.6%) b0.001
Constant pain, n 24 (22.6%) 27 (64.3%) b0.001
WBC count (×109/L),median (range) 9.6 (4.9–29.0) 13.7 (5.3–23.4) b0.001
Neutrophil %, median (range) 68.7 (4.3–95.0) 77.5 (30.0–91.0) 0.013
CRP (g/L), median (range)a 10.6 (0.2–400.8) 23.9 (0.2–176.5) 0.015
Urine LRG ×102(g/mol), median (range) 4.10 (0.03–328.55) 22.18 (0.03–719.58) 0.014
AuB score, median (range) 0.152 (0.065–0.907) 0.477 (0.066–0.881) b0.001
PAS, median (range) 3.5 (0–9) 6 (2–10) b0.001
a
2 missing CRP values.

literature, PAS b 4 was used most commonly as the low risk cutoff to ex- model for the linear predictor of the AuB score is expressed as:
clude AA [17].
y ¼ −2:669 þ 1:604∙constant pain þ 0:943∙Right iliac fossa tenderness
2. Results þ 1:540∙Pain on percussion þ 0:384∙LRG=Creatinine ðg=molÞ

2.1. Demographic and clinical characteristics of study population
Over the 8-month period, 168 patients were approached for partici-
pation in our study of whom 20 (11.9%) declined or were excluded
based on our study criteria. The final study population was 148 patients
with a mean age of 10.3 years for both the AA and non-AA group
(Table 1). There were 30 (71.4%) males in the AA group and 51
(48.1%) in the non-AA group. There were a total of 42 (28.4%) patients
with appendicitis of whom 9 (21.4%) were perforated and we had 5
(10.6%) negative appendectomies. The demographic, clinical character-
istics and diagnosis of the patients were shown in Tables 1 and 2.
In terms of diagnostic performance of individual laboratory markers,
urine LRG by itself was comparable to ‘WBC’, ‘% Neu’ and ‘CRP’ by
area under the receiver operating characteristic curve (AUC) (Fig. 1
and Table 3).
In analysis of the radiological examination data, 126 (85.1%) of the
148 patients had at least one modality of imaging studies. 120 (81.1%)
patients underwent ultrasound studies alone of which 39 (48.1%)
were confirmed appendicitis. 6 (4.1%) patients had initial ultrasound
followed by CT scan as US was inconclusive. 3 out of the 6 were eventu-
ally diagnosed with appendicitis.
For each of the three clinical variables, its presence is indicated by ‘1’
and absence by ‘0’.The value y calculated from the linear predictor is
transformed to the predicted probability of appendicitis by the formula,
p = exp {y}/(1 + exp {y}).
The diagnostic threshold for the AuB score, determined on the clini-
cal basis of an acceptable value for negative predictive value was se-
lected as p=0.15. Below this cutoff, patients are deemed to be at ‘low
risk’ and would result in AA being ruled out. Patients with AuB score
p ≥ 0.15 are considered ‘equivocal’ for AA.
2.3. Diagnostic performance of the AuB score
The median AuB score for patients with appendicitis was 0.47 and
was distinctively elevated when compared to the 0.15 of non-AA pa-
tients (p b 0.001). The median AuB score was even higher, 0.75, for pa-
tients with perforated appendices. Receiver operating characteristic
(ROC) curves were generated for the AuB score and PAS to determine
their diagnostic capability and for comparison (Fig. 1). The area under

2.2. Construction of the appendicitis urine biomarker (AuB) score

We performed stepwise multivariable logistic regression analysis

with diagnosis of AA (0 = N, 1 = Y) as the dependent variable. Predic-
tors selected as the best discriminators were presence/absence (P/A)
of constant pain (p = 0.0002), P/A of right iliac fossa tenderness
(p = 0.0634), P/A of pain on percussion/ hopping/ coughing (p =
0.0162) and LRG/creatinine (g/mol) levels (p = 0.0588). Model coeffi-
cients corresponding to predictors are given in Table 4. The AuB score

Table 2
Final diagnosis of non-AA patients.

Diagnoses (N = 106) No. of patients (%)

Gastritis and gastroenteritis 62 (58.5)

Mesenteric adenitis 9 (8.5)
Constipation 4 (3.8)
Nonspecific abdominal pain 15 (14.1)
Ovarian cyst 3 (2.8)
Fig. 1. ROC curves for detection of appendicitis by AuB, PAS, urine LRG and hematological
Others 13 (12.3)
biomarkers (WBC, %Neu and CRP).

Please cite this article as: Yap T-L, et al, A novel noninvasive appendicitis score with a urine biomarker, J Pediatr Surg (2018), https://doi.org/
10.1016/j.jpedsurg.2018.10.025
4 T-L. Yap et al. / Journal of Pediatric Surgery xxx (xxxx) xxx–xxx

Table 3
Performance characteristics of AuB score, PAS and three hematological markers (WBC, %Neu and CRP).

Sensitivity % Specificity % (95% CI) PPV % NPV % Positive likelihood ratio (95% CI) Negative likelihood ratio (95% CI)
(95% CI) (95% CI) (95% CI)

AuB Score b0.15 98 (87–100) 38 (29–48) 38 (35–42) 98 (85–100) 1.57 (1.34–1.83) 0.06 (0.01–0.44)
PAS b4 81 (66–91) 50 (40–60) 39 (34–45) 87 (78–93) 1.62 (1.27–2.06) 0.38 (0.20–0.73)
WBC ≥10,000 81 (66–91) 52 (42–62) 40 (34–46) 87 (78–93) 1.68 (1.32–2.15) 0.37 (0.19–0.70)
%Neu ≥75% 57 (41–72) 63 (53–72) 38 (30–47) 79 (72–84) 1.55 (1.08–2.23) 0.68 (0.46–0.99)
CRP ≥24a 48 (32–64) 66 (56–75) 36 (27–46) 76 (70–81) 1.41 (0.93–2.15) 0.79 (0.57–1.09)
a
2 missing CRP values.

the ROC curve (AUC) for AuB score was 0.82 (95% CI: 0.75–0.89) and
outperformed PAS with AUC of 0.78 (95% CI: 0.69–0.86) in our cohort.
At the cutoff of AuB b0.15, sensitivity and negative predictive value
(NPV) of the AuB score were 97.6% with the negative likelihood ratio
(LR −) of 0.06. Specificity was 37.7%, positive predictive value (PPV)
38.3% and positive likelihood ratio (LR +) 1.57. Therefore, values of
AuB b 0.15 identify patients at low risk of AA. In our study, an AuB
score b 0.15 correctly identified 40 out of 41 (97.5%) non-AA patients
with only one false negative. Patients with AuB score ≥ 0.15 were cate-
gorized as ‘Equivocal’ as the positive predictive value of AA at the 0.15
cut-point is low. In our cohort, there were 66 (61.6%) non-AA patients
and 41 (38.4%) AA patients in this group (Fig. 2a). The strength of the
AuB score was therefore as a negative predictor of AA. Within this low
probability group, ultrasonography was requested for 27 (65.8%)
patients.
On the other hand, PAS at a cutoff of less than 4, demonstrated 80.9%
sensitivity, 86.8% NPV, 50% specificity and 39% PPV. The LR− and LR+
of PAS were 0.38 and 1.62 respectively. When compared with AuB,
PAS demonstrated a lower sensitivity, NPV and LR−, comparable PPV
but a higher specificity. Although the AUC of the AuB score was higher
than that of the PAS 0.82 vs. 0.78, it was not statistically significant
(p = 0.52). Nonetheless, when we compare the AuB at a cutoff of 0.15
with the PAS at a cutoff of 4: the PAS incorrectly identified eight (19%)
false negative patients with AA into the low risk group (Fig. 2b). In ad-
dition, the PAS was only able to correctly categorized 86.8% (53/61) of
the non-AA patients. As ours was only a pilot study, we bench marked
our score with the most widely acceptable PAS.
There were 5 patients in the study who had negative appendecto-
mies. Three of these had definite intraabdominal pathologies: cecal di-
verticulitis, helminthic infestation and presence of fecolith within
lumen of the appendix. Of the remaining two patients with normal ap-
pendix on histology, one of them actually had an AuB score in the low
probability group while the other had a marginally raised score of 0.157.
and the level of urine LRG. To our knowledge, it is the first and only ap-
pendicitis score to employ a noninvasive urine biomarker.
In our study population, AuB scores below the ‘test’ cut-point thresh-
old of 0.15 were superior to PAS at its low cutoff of 4. The performance of
CPR is assessed universally by its sensitivity and negative likelihood
ratio (LR −) [18]. A high-performing CPR should have a sensitivity of
greater than 0.95 and negative likelihood ratio of less than 0.1. The
AuB score satisfied both of these criteria.
The AuB score was able to classify 40 out of 106 non-AA patients, or
37.7%, correctly into the low probability category. This group of patients
may then be discharged, eliminating continued ED observation or hos-
pital admission. Furthermore, they may not require initial imaging stud-
ies. The single patient in our study with a false negative AuB score had
an acutely inflamed, and not perforated, appendix. From a safety stand-
point, the patient would have been reviewed in a planned follow up
12 hours later.
The use of urine samples for diagnosis is exceptionally attractive in
pediatric practice due to its noninvasive nature. Previous concerns of al-
leged instability and low substrate quantum have been addressed in re-
cent publications [20]. Instead urine proteome has proved to be quite

3. Discussion

Clinical prediction rules (CPRs) or scoring systems for children with

suspected AA have been developed and heavily relied on to guide the di-
agnostic course of action [18]. The most widely used score for children is
the Pediatric Appendicitis Score (PAS) [19]. It was conceptualized ini-
tially for the definitive diagnosis of AA, and not for its exclusion or elim-
ination. Our AuB score was conceived primarily to identify patients with
low probability of AA. It is made up of only three clinical variables: ‘con-
stant pain’, ‘RIF tenderness’, ‘pain on percussion/ coughing/ hopping’

Table 4
Maximum likelihood estimates for construction of AuB score.

Predictor Maximum likelihood Adjusted Adjusted OR (95%

Estimate OR CI)

Constant Pain 1.604 4.97 2.13–11.58

RIF tenderness 0.943 2.56 0.95–6.95
Pain on 1.538 4.65 1.33–16.32
percussion
Fig. 2. a: Frequency distribution of AuB with cutoff at 0.15 b: Frequency distribution of PAS
LRG/Creatinine 0.384 1.00 1.00–1.01
with cutoff at 4.

Please cite this article as: Yap T-L, et al, A novel noninvasive appendicitis score with a urine biomarker, J Pediatr Surg (2018), https://doi.org/
10.1016/j.jpedsurg.2018.10.025
 T-L. Yap et al. / Journal of Pediatric Surgery xxx (xxxx) xxx–xxx
stable and ‘processed’ during its period of ‘storage’ in the bladder by en-
dogenous proteolytic enzymes. But the most pertinent issue in pediatric
practice is that it avoids noxious and often traumatic venipuncture and
intravenous access. It also obviates the need for specialized skills and
presence of phlebotomists. Therefore, the most appealing aspect of the
AuB score is that it is noninvasive.
In addition, the AuB score has only 3 clinical variables and one quan-
titative measurement of urine biomarkers. Thus it is easy to recall and
apply in a busy ED setting. In a multicenter study on interrater variabil-
ity by Kharbanda et al., our 3 clinical variables were noted to be among
those with the highest correlation rates [21].In particular, ‘pain on per-
cussion/ coughing/ hopping’ emerged with the highest interrater reli-
ability. In contrast, variables like ‘migration of pain’, ‘anorexia’, and
‘nausea’ had the lowest reliability. The author thus recommended that
incorporating high interrater reliability predictors into CPR will enhance
its utility.
There are however potential challengers of testing urine samples in a
time-critical clinical area such as the ER. We had experienced delays
with urine collection in patients, who were dehydrated, who were un-
able to collect midstream urine samples or who just refused to void vol-
untarily. The actual feasibility and logistic of analyzing urine biomarkers
in a clinical setting may pose additional obstacles but as our samples
were analyzed in a research context, we do not have experience in
this aspect. However, the time taken for processing of urine LRG in a
clinical care setting had been reported to be 2 hours and at a cost of
26 USD in the study by Salo et al. [22]. This was made possible by the
new generation ‘Ultra-Fast ELISA’ assay.
A Smartphone App for medical calculation may also be adopted to
facilitate the AuB score computation. These features may enable the
AuB score to function as a point-of-care tool. One critique of the AuB
however, as with all other scoring systems, is that it is affected by the
prevalence of AA in the study population; however, the LR- is not.
The discriminatory accuracy of urine LRG as a solo diagnostic tool for
appendicitis has been equivocal so far. Kentsis et al. reported the levels
of urine LRG, when quantified by highly specialized selected ion moni-
toring (SIM) mass spectrometry, to be elevated more than 100-fold in
patient with AA as compared to those without [14]. In their blinded pro-
spective cohort of 49 patients, a near perfect area under the ROC curve
(AUC) of 0.98 and 0.99 was reported. However when the measurement
was repeated using the ELISA technique, its performance was reduced
to an AUC of 0.80, due to an interference effect. This was replicated in
the study by Kharbanda et al. [23], wherein despite achieving a 100%
sensitivity and negative predictive value, the AUC was only 0.63 by
ELISA. In another comparative study employing equal numbers of AA
and non-AA cases, Salo et al. showed improved performance when
urine LRG was used in conjunction with the Pediatric Appendicitis
score [22]. The combination demonstrated superior sensitivity of 95%,
specificity of 90% and AUC of 0.94.
This resonated in our experience as well. In our study, the diagnostic
ability of urine LRG, by itself, was only mildly superior to ‘WBC’ and ‘%
Neu. This may be explained by the fact that LRG is a nonspecific, acute
phase protein which is secreted in response to any inflammatory
stimulus.
The strategy of risk stratification was first described in the develop-
ment of clinical decision rules (CDR). A truly useful CDR would classify
patients into 3 groups – low-, moderate- and high-risk groups – based
on the ‘test’ and ‘treatment’ threshold respectively. For the AuB score,
we focused on the ‘test’ threshold to rule out patients with low pre-
dicted probability of AA. Patients with a score above the threshold of
0.15 are to be designated as ‘equivocal’ relative to risk of AA. All of
these patients would be mandated to undergo imaging studies, admis-
sion or further observation.
If the AuB score were used as a negative predictor for our cohort,
we would have potentially obviated US in 37.7% of the non-AA pa-
tients. In centers where computed tomography (CT) scan is the
first-line imaging modality, the corresponding proportion would
Please cite this article as: Yap T-L, et al, A novel noninvasive appendicitis score with a urine biomarker, J Pediatr Surg (2018), https://doi.org/
10.1016/j.jpedsurg.2018.10.025
5
have escaped the radiation-based investigation. On the other hand,
US findings may also be interpreted in conjunction with the AuB
score. This was recommended by Bachur et al., as in the presence of
a low-risk PAS, a positive US result suggestive of AA must be
interpreted with caution [24].
A number of limitations deserved mention with respect to our study,
the most important being its single center, convenience cohort design
and small sample size. Our recruitment was restricted to weekdays
and evenings, thus a potential source of error may be from patients eli-
gible for recruitment on weekends and overnight. Ours was a derivative
study which has not been validated both internally and externally. Fur-
thermore, the analysis of urine LRG samples was performed in a re-
search laboratory and not a clinical care facility. The strength of the
study was, its prospective design. In addition, all non-AA patients
were followed up in-hospital as well as with phone verification
2 weeks following discharge to ensure no change in diagnosis.
4. Conclusion
With the aim to do away with invasive and noxious blood sampling,
we have developed a novel appendicitis score employing the urine bio-
marker, LRG. The score performed well as a negative predictor for AA in
our sample of patients. Its application may serve to improve the cost effec-
tiveness of diagnostic approaches in children suspected of AA. However,
given the medical and legal implications of missed appendicitis, extensive
validation efforts to ascertain its true diagnostic performance are required.
Acknowledgments
This research was supported by the SingHealth Foundation Research
Grant 2015, Singapore (SHF/FG591P/2015). The authors gratefully ac-
knowledge the contributions of Goh Xin Yu Isis, Tan Hong Yi, Muham-
mad Khidhir Bin Iswanti and Mark Ian Lonio Angus for the collection
of data that were used in this study.
References
[1] Humes DJ, Simpson J. Acute appendicitis. BMJ 2006;333:530–4.
[2] Becker T, Kharbanda A, Bachur R. Atypical clinical features of pediatric appendicitis.
Acad Emerg Med 2007;14:124–9.
[3] Graff L, Russell J, Seashore J, et al. False-negative and false-positive errors in abdom-
inal pain evaluation: failure to diagnose acute appendicitis and unnecessary surgery.
Acad Emerg Med 2000:1244–55.
[4] Reynolds SL. Missed appendicitis in a pediatric emergency department. Pediatr
Emerg Care 1993;9:1–3.
[5] Rothrock SG, Pagane J. Acute appendicitis in children: emergency department diag-
nosis and management. Ann Emerg Med 2000;36:39–51.
[6] National hospital ambulatory medical care survey: 2013 emergency department
summary tables. https://www.cdc.gov/nchs/data/ahcd/nhamcd_emergency/2013_
ed_web_tables.pdf, Accessed date: 28 March 2017.
[7] Hall MJ, DeFrances CJ, Williams SN, et al. National hospital discharge survey: 2007
summary. Natl Health Stat Report 2010;29:1–20.
[8] Benabbas R, Hanna M, Shah J, et al. Diagnostic accuracy of history, physical examina-
tion, laboratory tests, and point-of-care ultrasound for pediatric acute appendicitis
in the emergency department: a systematic review and meta-analysis. Acad
Emerg Med 2017;24:523–5.
[9] Kulik DM, Uleryk EM, Maguire JL. Does this child have appendicitis? A systematic re-
view of clinical prediction rules for children with acute abdominal pain. J Clin
Epidemiol 2013;66:95–104.
[10] Fahimi J, Herring A, Harries A, et al. Computed tomography use among children present-
ing to emergency departments with abdominal pain. Pediatrics 2012;130:e1069–75.
[11] Hryhorczuk AL, Mannix RC, Taylor GA. Pediatric abdominal pain: use of imaging in
the emergency department in the United States from 1999 to 2007. Radiology
2012;263:778–85.
[12] Larson DB, Johnson LW, Schnell BM, et al. Rising use of CT in child visits to the emer-
gency department in the United States, 1995–2008. Radiology 2011;259:793–801.
[13] Lee A, Johnson JA, Fry DE, et al. Characteristics of hospitals with lowest costs in man-
agement of pediatric appendicitis. J Pediatr Surg 2013;48:2320–6.
[14] Kentsis A, Ahmed S, Kurek K, et al. Detection and diagnostic value of urine leucine-
rich alpha-2-glycoprotein (LRG) in children with suspected acute appendicitis. Ann
Emerg Med 2012;60:78–83.
[15] Kentsis A, Lin YY, Kurek K, et al. Discovery and validation of urine markers of acute
pediatric appendicitis using high accuracy mass spectrometry. Ann Emerg Med
2010;55:62–70.
6 T-L. Yap et al. / Journal of Pediatric Surgery xxx (xxxx) xxx–xxx

[16] Saito K, Tanaka T, Kanda H, et al. Gene expression profiling of mucosal addressin cell
adhesion molecule-1+ high endothelial venule cells (HEV) and identification of a
leucine-rich HEV glycoprotein as a HEV marker. J Immunol 2002;168(3):1050–9.
[17] Ebell MH, Shinholser J. What are the most clinically useful cutoffs for the Alvarado and
pediatric appendicitis scores? A systematic review. Ann Emerg Med 2014;64:365–72.
[18] Maguire JL, Kulik DM, Laupacis A, et al. Clinical prediction rules for children: a sys-
tematic review. Pediatrics 2011;128:e666–77.
[19] Sameul M. Pediatric appendicitis score. J Pediatr Surg 2002;37(6):877–81.
[20] Albalat A, Mischak H, Mullen W. Clinical application of urinary proteomics/
peptidomics. Expert Rev Proteomics 2011;8(5):615–29.
[21] Kharbanda AB, Stevenson MD, Macias CG, et al. Interrater reliability of clinical find-
ings in children with possible appendicitis. Pediatrics 2012;129:695–700.
[22] Salö M, Roth B, Stenström P, et al. Urinary biomarkers in pediatric appendicitis.
Pediatr Surg Int 2016;32:795–804.
[23] Kharbanda AB, Rai AJ, Cosme Y, et al. Novel serum and urine markers for pediatric
appendicitis. Acad Emerg Med 2012;19(1):56–62.
[24] Bachur RG, Callahan MJ, Moniteaux MC, et al. Integration of ultrasound findings and
a clinical score in the diagnostic evaluation of pediatric appendicitis. J Pediatr 2015;
166:1134–9.

Please cite this article as: Yap T-L, et al, A novel noninvasive appendicitis score with a urine biomarker, J Pediatr Surg (2018), https://doi.org/
10.1016/j.jpedsurg.2018.10.025