Eur J Pediatr DOI 10.1007/s00431-016-2724-8

ORIGINAL ARTICLE

The β-glucosidase assay: a new diagnostic tool for necrotizing enterocolitis. Sensitivity, specificity, and predictive values José Luis Gómez-Chaparro Moreno 1 & Alejandro Rodríguez Torronteras 2 & María Dolores Ruiz González 3 & Lucía Izquierdo Palomares 4 & Daniel Bonilla Valverde 5 & Julia Ruiz Laguna 5 & Alfonso Delgado Rubio 6 & Juan López-Barea 5

Received: 6 August 2015 / Revised: 12 April 2016 / Accepted: 18 April 2016 # Springer-Verlag Berlin Heidelberg 2016

Abstract We aimed to establish the utility of serum cytosolic β-glycosidase (CBG) assay as a NEC diagnosis tool. CBG activity has been compared in 192 NEC-free (NEC−) and 13 NEC-affected (NEC+) neonates, with modified Bell’s stages II/ III, born at Reina Sofia University Hospital; additional blood hematology, microbiology, and biochemical parameters have been assayed. NEC+ neonates have higher serum CBG activity, 26.4 ± 12.4 mU/mg; 95 % CI (18.8–33.9), than NEC− infants, 11.0 ± 6.6 mU/mg; 95 % CI (10.1–11.9) (p < 0.0001). The CBG cutoff value in the ROC curve, 15.6 mU/mg, discriminates NEC+/NEC− infants with 84.6 % sensitivity, 85.9 % specificity, 37.9 positive predictive value and 98.2 negative predictive value, 6.11 positive likelihood ratio and 0.18 negative likelihood

Communicated by Patrick Van Reempts * José Luis Gómez-Chaparro Moreno [email protected]

1

Experimental Unit. Córdoba Health District. Andalusian Health Service, C/ Isla de Lanzarote s/n Edificio 2, 1ª Planta, 14011 Córdoba, Spain

2

Department of Epidemiology. Córdoba Health District, Andalusian Health Service, C/ Isla de Lanzarote s/n Edificio 2, 1ª Planta, Córdoba 14011, Spain

3

Neonatology Unit, Pediatrics Service, RSUH. Andalusian Health Service, Avda. Menendez Pidal s/n., Córdoba 14004, Spain

4

Pediatrics Radiology Section, Radiodiagnostic Service, RSUH, Andalusian Health Service, Avda. Menendez Pidal s/n., Córdoba 14004, Spain

5

Department of Biochemistry and Molecular Biology of Córdoba University, Severo Ochoa Building. A4 Highway, Km 396a, Rabanales Campus, Córdoba 14071, Spain

6

Department of Pediatric of San Pablo-CEU University, School of Medicine. Monteprincipe Campus. Alcorcón, Madrid 28925, Spain

ratio, 33.61 DOR, and 0.89 AUC. A combined panel [CBG + aspartate aminotransferase + C-reactive protein] shows a 0.90 AUC value in multiple linear regressions. Conclusions: The serum CBG level is a good NEC diagnosis test and a novel NEC biomarker which may become a screening tool. What is known: •NEC affects ∼2.5 % of infants at NICU, ∼90 % of them weighing <1500 g. •NEC requires a careful differential diagnosis, being lethal if not diagnosed and treated. What is new: •CBG assay will be useful to determine infants without NEC and preventing unnecessary treatment. •CBG assay could discriminate NEC better than other gut-specific sera protein biomarkers.

Keywords Necrotizing enterocolitis . Cytosolic beta-glucosidase . Premature infants . Humans . Diagnostic . Biological markers

Abbreviations ALT Alanine aminotransferase AST Aspartate aminotransferase AUC Area under curve BUN Blood urea nitrogen CBG Cytosolic β-glycosidase DOR Diagnostic odds ratio ICC Intraclass correlation coefficient LH+ Positive likelihood ratio Negative likelihood ratio LH− MLR Multiple logistic regression NEC Necrotizing enterocolitis

Eur J Pediatr

NEC+ NEC− NPV NICU CRP PPV PRBT ROC RSUH S SLR SP

NEC-affected neonates Neonates without NEC Negative predictive value Neonatology intensive care unit C-reactive protein Positive predictive value Packed red blood transfusion Receiver operating characteristic Reina Sofia University Hospital Sensitivity Simple logistic regression Specificity

Methods Study design This prospective study has been conducted by the Experimental Unit of Córdoba Health District, the Neonatology Unit and the Pediatrics Radiology Section of Reina Sofia University Hospital (RSUH), and the Department of Biochemistry and Molecular Biology of Córdoba University, and was previously approved by the RSUH Research Committee. All neonates were born at the RSUH and were consecutively recruited (January 2006– December 2009) after the parents signed an informed agreement. Study population

Introduction Necrotizing enterocolitis, a common gastrointestinal condition in NICUs, is among the main causes of neonatal mortality and morbidity [33]. This disease has an incidence of 1–3 NEC cases per 1000 neonates and affects 2–2.5 % of infants admitted to NICUs [15]. Low birth weight (LBW) is a major risk factor, since 90 % of the NEC+ neonates weigh below 1500 g [43]. The incidence and onset of NEC are inversely related to gestational age (GA) [9]. Its mortality decreases from 42–100 % in infants weighing below 750 g to 15 % in those weighing 1250–1500 g [49]. NEC has various clinical appearances with divergent symptoms which require a careful differential diagnosis [29], being lethal if not diagnosed and treated early. The initial staging tool has been Bell’s stages lately modified by Walsh et al. [48]. Treatment of neonates categorized as modified Bell’s stages I/IIA depends on medical therapy, and stages IIB/III are treated surgically [19]. Cytosolic β-glucosidase, which metabolizes flavonoid glycosides, belongs to the ubiquitous GH1 hydrolase family [3] containing a β/α-barrel layout and a conserved catalytic site [14]. Whereas in lower mammals CBG has a detoxifying role, its function in humans is unknown [43]. It is located mainly in the intestine (over 80 %), being cytosolic because it lacks transmembrane tracts [28]. The number of infants admitted to NICUs with low GA and LBW is steadily rising, with a parallel increase in NEC risk [9, 15, 43]. Although there were various serum and urine biomarkers for NEC [12, 31, 40], Dimmitt and Morris [4, 28] reported that β-glucosidase increased in NEC-model animals, a finding subsequently confirmed by Dongmei et al. [5] and Benkoe et al. [1] in NEC human neonates, although in short series. The present study establishes the serum CBG levels in NEC+ neonates at modified Bell’s stages II–III compared to a reference group and shows that CBG is a promising NEC biomarker.

The NEC+ neonates had symptoms of modified Bell’s stages II–III [48], although those infants suspected to be modified Bell’s stage I were excluded, to avoid confounding NEC cases. The NEC− reference group included neonates born at the RSUH displaying pathologies different from NEC. In the NEC+ neonates, serum CBG was assayed at the time of NEC clinical diagnostic and in the NEC−, upon admission to the NICU, when senior neonatologists excluded a NEC diagnostic. The CBG activity was compared in neonates with or without NEC (NEC+, NEC−). To avoid confounding effects when comparing NEC+ and NEC−, infants over 35-week gestation were not included as controls, because it is well established that NEC incidence rises in infants with low GA [21]. Neonates with congenital abnormalities, chromosomopathy, spontaneous intestinal perforation (SIP), prenatal risk factors, or derived from pathological pregnancies were excluded. The sample size for comparing the CBG values of NEC+ and NEC− neonates was derived from animal model studies and from our initial results1. For a 4 mU/mg difference between NEC+ and NEC−, with 2.5 mU/mg SD, 0.05 α-error, and 90 % potency, ten infants were needed for each branch. Nevertheless, more reference cases were included, over the minimum estimated, to increase the precision of the CBG values. Notice that seven infants with modified Bell’s stage I were excluded (Fig. 1). Infant identities were masked by a numeric code, and the assays were blinded. Blood was drawn with minimal injury; 2 ml blood was drawn into Vacutainer EDTA tubes (B&D, USA) and the red, white, and platelet series were subsequently analyzed in PENTRA120 (Horiba ABK, France) and ADVIA120 (Bayer Co, USA) counters. Biochemical parameters were assessed after clot formation of 2 ml blood drawn in Vacutainer® SST-II tubes (B&D, USA) using an Architect Systems c8000 device (Abbott 1

Gómez-Chaparro Moreno JL, et al. (2005, Oct). Determinación de los niveles de la enzima β-glucosidasa citosólica en sangre. Poster presented at the XIX BCongreso Nacional de la Sociedad Española de Pediatría Extrahospitalaria y Atención Primaria^. Madrid.

Eur J Pediatr NICU (212)

NEC +

NEC -

(20)

Bell Stage I (7)

Bell Stages II/III (13)

(192)

G.A <35 w (128)

G.A >35 w (64)

Fig. 1 Flowchart of the total BGC study

Labs, USA). Additional laboratory tests were made on purpose for each infant included, following the standard protocols used at the RSUH.

variables was assessed by the correlation matrix and the linearity of logit (p) and independent quantitative variables. No significant variable was excluded from the final model. Nagelkerke R2 and Hosmer-Lemeshow tests were performed to test model adequation. Adjusted odds ratios and 95 % CI were obtained, a receiver operating characteristic (ROC) curve and an area under the curve (AUC) were generated, and the values of the sensitivity (S), specificity (SP), positive predictive value (PPV), negative predictive value (NPV), positive likelihood ratio (LH+), negative likelihood ratio (LH−), and diagnostic odds ratio (DOR) parameters were determined. Statistical analyses were conducted using SAS software v9.3 (SAS Institute, Cary NC). The graphical assessment of MetaDisc v1.4 was used to compare other serum biomarkers of NEC with CBG [51].

β-glucosidase (CBG) Sera were frozen at −86 °C until triplicate CBG assays were performed. From each neonate, 20 μl of serum was added to 0.2 M Na-citrate buffer, pH 6.0, containing 5 mM of 4methylumbe-lyferyl-β-D-glucopiranoside in 0.1 ml. The mixture was incubated for 30 min at 37 °C, and the reaction was terminated with 2.9 ml of ammonia/glycine, pH 10. The fluorescence of 4-methylumbelyferone was measured in an LB-50 device (Perkin-Elmer, USA) at 360-nm excitation and 515-nm emission wavelengths [4]. One CBG unit is the amount of enzyme that transforms 1 nmol of substrate per minute. Solutions in Milli-Q water were made with Sigma-Aldrich (Madrid, Spain) reagents. Bell’s stages, clinical variables, and laboratory variables The epidemiological, clinical diagnostics and laboratory variables (observed by senior neonatologists) and the radiographic findings (observed by a pediatric radiologist) were used to establish the NEC stage according to the modified Bell’s staging criteria [48] (Table 1). Statistical analysis Descriptive statistics are shown as the mean ± SD, and the categorical variables as the value and percentage. A 95 % CI was calculated for each test, all being two-sided. All CBG assays were performed in triplicate and averaged. The intraclass correlation coefficient (ICC) was calculated for triplicate CBG assays. Statistical significance was defined as p < 0.05. CBG values of the NEC+ and NEC− neonates were compared with the Student’s t test. ANOVA and post hoc analysis with Bonferroni correction were used to compare more than two groups. An association of CBG with other quantitative variables was studied by the Pearson correlation coefficient test. A chi-squared test was used with qualitative variables, and the Fisher’s exact test was used if the result was n < 5. Multivariate analysis used logistic regression; covariate selection in multiple models used stepwise regression. The colinearity of independent

Results All CBG assays had an ICC of 0.95 (95 % CI = 0.93 to 0.96). Figure 1 shows that our study included 192 neonates who did not qualify for modified Bell’s stages (NEC−) when admitted to the NICU of the RSUH for different reasons. The CBG activity of these NEC− infants was 10.99 ± 6.62 mU/mg. Of the 20 NEC+ neonates, 7 were considered modified Bell’s stage I and excluded from the study. The 13 remaining NEC+ neonates, with Bell’s stages II–III, had a CBG activity of 26.36 ± 12.44 mU/mg; 9 of them had surgical treatment, and other 4 medical treatment. Table 1 shows the general variables and diagnostic features of the NEC+ and NEC− groups. To compare them while avoiding body weight as confounding factor, 64 infants with a GA ≥35 weeks were excluded from the NEC− group [20, 21]. As shown in Fig. 1, CBG differences were compared between the 13 NEC+ neonates with Bell’s stages II–III and the 128 NEC− neonates with GA <35 weeks. Table 2 shows the demographic data and CBG activity of the 13 NEC+ neonates. Table 3(A) shows 25 quantitative differences compared in NEC− and NEC+ groups. The red series had significant differences (erythrocytes, hemoglobin, and hematocrit, p = 0.0005, p < 0.0001, and p < 0.0001, respectively). NEC+ neonates had fewer platelets than NEC− (p = 0.0004). The aspartate aminotransferase (AST) and alanine aminotransferase (ALT) values differed significantly between NEC+ and NEC− (both, p = 0.0001). NEC+ neonates had lower serum Na+ and K+ than NEC− (p = 0.0001 and p = 0.002, respectively). NEC+ have higher glycemia (p = 0.04) and 5-fold higher CRP level (p < 0.0001) than NEC−. Table 3(B) shows that from the 12 qualitative variables compared, only packed red blood transfusion (PRBT) showed significant difference between NEC− and NEC+ (p < 0.001). The following variables failed to show significant differences between both groups (p > 0.05): gender, delivery type, hypoxia-ischemia, venous/arterial umbilical catheter, respiratory distress, preeclampsia,

Eur J Pediatr Table 1 Epidemiological characteristics and diagnostics features of the NEC+ and NEC− neonates

NEC−

NEC+

Variables

n

Mean (SD)

n

Mean (SD)

Gender (male/female)

6/7

–

72/56

–

Delivery mode (vaginal/C-section)

6/7

–

95/33

–

Gestational age (weeks) Birth weight (g)

13 13

28.53 (2.5) 1128.85 (458)

128 128

29.54 (2.4) 1255 (382.58)

Apgar score 1 min

13

5.7 (2.5)

128

6.23 (1.75)

Apgar score 5 min Other diagnostic

13 n

7.4 (2.5) %

128 n

8.05 (1.28) %

Exitus Minor malformations

3 0

23.3 0

10 7

7.8 5.4

Cardiocirculatory pathology

4

30.7

19

14.8

Digestive pathology

2

15.3

15

11.7

Endocrine-metabolic pathology Hematologic pathology

0 2

0 15.3

7 9

5.4 7

Infectious pathology Neurologic pathology

9 2

69.2 15.3

40 15

31.2 11.7

Oftalmologic pathology Respiratory pathology No pathology

2 3 0

15.3 23.07 0

6 56 31

4.6 43.7 24.2

SD standard deviation Minor malformation: syndactyly (n = 4), metatarsus adductus (n = 2), foot amniotic band (n = 1)

prenatal corticoids, hemoculture, ductus arterious, polycythemia, and feeding type. Via simple logistic regression (SLR, Fig. 2a), CBG was established as the only predictive variable of NEC, according to the following equation: logit (p) = – 5.07 + 0.16 * CBG

CBG was statistically significant (p < 0.0001). The Hosmer-Lemeshow test revealed a good fit (χ2 = 7.31; p = 0.50), and the Nagelkerke R2 effect size (R2 = 0.39) a good predictive model. The ROC curve for the intersect cutoff values was 15.6 mU/mg CBG. The AUC was 0.89. A stepwise method was used for multiple logistic

Table 2 Demographics data and CBG activity of the 13 NEC+ neonates

Neonate

Gender

BW (g)

GA (weeks)

APGAR 1 min

APGAR 5 min

Analytical date (days alive)

CBG (mU/mg)

1 2 3 4 5 6 7 8 9 10 11 12 13

F F F F F F F M M M M M M

960 1460 1450 1010 1100 685 1095 1860 2090 735 790 720 720

29 31 30 26 28 26 28 32 33 30 26 26 26

7 8 4 3 1 8 3 9 7 8 5 4 7

8 9 7 6 1 9 5 10 9 10 7 6 9

27 3 43 9 5 41 54 14 7 4 8 54 43

15.77 16.0 28.5 29.47 31.44 35.92 43.4 8.46 12.83 15.73 19.3 42.67 43.25

F female, M male

11 764 (8 085) 5 839 (3 013) 5 222 (4 561) 643 (858) 341 (705) 5 (19) 27 (172) 3 603 (1 936) 4 750 (3 662) 2 200 (1 973) 1 542 (1 588) 178 (105) 248 (216) 103 652 (99 726) 217 592 (110 766) 101.6 (105.08) 48.2 (45.1) 36.1 (53.3) 16.5 (26.6)

− + − +

− + − + − + − + − + − + − + −

Creatinin (micromol/L)

BUN (mmol/L)

+ − +

+ − + −

K+ (mmol/L)

Glucose (mmol/L)

+ −

Na+ (mmol/L)

ALT (UI/L)

AST (UI/L)

Platelets (×10E9/L)

Eosinophyls (×10E9/L)

Monocytes (×10E9/L)

Lymphocytes (×10E9/L)

Blasts (×10E9/L)

Band neutrophyl (×10E9/L)

21.28 (9.02) 55.8 (40.1)

17.7(13.7)

3.26 (1.05) 4.14 (0.92) 7.33 (3.11) 5.36 (3.00)

129.2 (7.4) 134.9 (4.8)

0.38 (0.08) 11 592 (6 412)

− +

Leukocytes (×10E9/L)

Segmented neu. (×10E9/L)

12.89 (2.78) 0.29 (0.07)

− +

Hematocrit (L/L)

Hemoglobin (g/dL)

Erythrocytes (×10E12/L)

9.42–25.95 9.49–12.06 31.54–79.98

2.63–3.90 4.01–4.28 5.45–9.21 4.34–5.20

124.77–133.69 134.20–135.56

195–397 6–17 6–43 2 433–4 772 4 615–5 658 1 007–3 391 1 209–1 661 114–242 223–284 43 387–163 915 211 944–243 479 37.05–178.23 38.68–57.68 3.28–74.90 10.85–22.07

11 300–13 602 4 017–7 659 5 093–6 392 124–1162

0.40–0.42 7 716–15 467

13.16–13.95 0.25–0.33

95 % CI 18.84–33.88 10.05–11.93 2,650,328–3,366,594 3,731,768–3,933,958 8.52–11.22

0.67

0.07

0.04

0.002

0.0001

<0.0001

<0.0001

0.0004

0.22

0.22

0.14

0.7

0.2

0.63

0.94

<0.0001

0.0001

0.0005

p <0.0001

Feeding type

Ductus arterious

PRBT

Hemoculture

Corticoids

Preeclampsia

Respiratory distress

Arterial catheter

Venous catheter

Hypoxia-ischemia

Delivery type

Gender

Variables

Mean (SD) 26.36 (12.44) 10.99 (6.62) 3,008,461 (592,467) 3,634,735 (710,186) 9.87 (2.24)

Variables CBG (mU/mg)

NEC group + − + − +

B

Differences between quantitative (A, t test) and qualitative (B, Fisher’s test) variables in NEC+ and NEC− cases

A

Table 3

No Yes No Yes Negative Positive No Yes No Yes Parenteral nutrition Formula milk Breastfeeding exclusive Breastfeeding partial

No Yes No Yes

No Yes

No Yes

Female Male Cesarean Vaginal

113 (88.28) 15 (11.72) 27 (21.09) 3 (23) 78 (75.73) 25 (24.27) 105 (82.03) 23 (17.97) 112 (87.50) 16 (12.50) 51 (39.84) 46 (35.94) 27 (21.09) 4 (3.13)

45 (35.16) 83 (64.84) 34 (66.98) 92 (73.02)

49 (38.28) 79 (71.72)

104 (81.25) 24 (18.75)

NEC (%) No 56 (43.75) 72 (56.25) 33 (25.78) 95 (74.22)

13 (100) 0 (0.00) 10 (76.92) 3 (23.08) 6 (60) 4 (40) 5 (38.46) 8 (61.54) 10 (76.92) 3 (23.08) 6 (46.15) 6 (46.15) 1 (7.69) 0 (0.00)

4 (30.77) 9 (69.23) 3 (23.08) 10 (76.92)

6 (46.15) 7 (53.85)

10 (76.92) 3 (23.08)

Yes 7 (53.85) 6 (46.15) 7 (53.85) 6 (46.15)

0.7

0.38

0.001

0.27

1

0.36

1

1

0.57

0.71

0.19

0.56

p

Eur J Pediatr

Eur J Pediatr

regression (MLR, Fig. 2b), and no colinearity of independent variables existed. The Hosmer-Lemeshow test revealed a good fit ( χ 2 = 12.84; p = 0.12), and the Nagelkerke R2 effect size (R2 = 0.50) indicated a good predictive model, according to the following equation: logitðpÞ ¼ 5:97 þ 0:14*CBG þ 0:012*AST þ 0:014*CRP

MLR identified statistically significant differences in CBG, AST, and CRP (p < 0.0002, p < 0.0026, and p < 0.051, respectively). The AUC increased from 0.89 in SLR to 0.90 in MLR. Figure 2c shows the probability tree diagram of CBG in the NEC+ vs NEC− study.

PRBT Packed red blood transfusion

7.31 (0.09) −

7.31–7.33

0.66 pH

1.48 (2.66) 1.78 (4.84) 7.32 (0.12) Procalcitonin (ng/mL)

+ − +

0.13–3.09 0.82–2.20 7.25–7.40

0.82

<0.0001 11.5 (26.9) −

6.17–13.84

48.0 (7.5) 55.3(52.7) − + CRP (mg/mL)

46.88–49.04 23.41–87.14

0.07

0.58

Proteins (g/L)

2.44–2.51 37.65–47.27 2.47 (0.27) 42.5 (8.0) − +

Ca2+ (mmol/L)

59.62 (120.8) 117.9 (65.7) 128.6 (84.0) 2.42 (0.48) − + − + Bilirrubin (μmol/L)

A

Table 3 (continued)

51.01–85.41 78.23–157.63 132.31–156.22 2.13–2.71

0.51

B

Discussion The CBG activity was 2.6-fold higher in the NEC+ infants, confirming Dimmit’s results [4] in an animal model. Our results also confirmed those of Dongmei et al. [5] although their paper, reporting 19-fold higher CBG values in NEC+ infants, used an unexisting Bcommercially available^ ELISA assay, and their units were out of the International System. Our results agree also with those of Benkoe et al. [1] although they reported just a 1.4-fold higher CBG activity in NEC+ infants. The differences between both studies could be due to the HPLC-MS/MS assay used in plasma (in our study fluorescence assay in serum) or to the inclusion by Benkoe et al. [1] in the NEC+ group of some neonates with Bell’s stage I. In addition, these two previous papers had very short series of NEC− infants, 41 the study of Dongmei et al. [5] and 18 the study of Benkoe et al. [1]. Notice also that we have measured a neutral/cytosolic CBG form (EC 3.2.1.21), while Benkoe et al. [1] measured an acid/lysosomal BG form (EC 3.2.1.45), leading to unnecessary confusions in neonates affected of lysosomal storage disorders. It should be noticed that neither Dongmei et al. [5] nor Benkoe et al. [1] calculated the S or SP values. Because CBG is found within gut microorganisms, its increase in serum could be due to microbial growth and paracellular translocation from gut lumen [11]. In fact, NEC is closely linked to infection, a necessary but insufficient diagnostic condition [27]. Since CBG is not correlated with hemoculture, its increase derives from the neonatal intestine rather than from bacterial growth. CBG is also found within leukocytes [10], especially immature forms, another NEC+ biomarker [37], although the leukocyte increase in early NEC stages was not significant (p > 0.05). CBG locates in the cytosol of polymorphonuclear leukocytes that are also rich in lysosomes with acid β-glucosidase activity [39]. Nevertheless, the CBG tested in the present study was a soluble form, because sera were previously centrifuged and filtered to remove cell debris.

Eur J Pediatr

Fig. 2 a Receiver operating characteristic of the simple logistic regression (SLR). b Multiple logistic regression (MLR). c Probability tree diagram of CBG and NEC in the NEC+ vs NEC− study

The >5-fold higher CRP level in NEC+ infants, confirmed previous reports [34, 35]. There were lower values in the red

series (erythrocytes, hemoglobin, hematocrit), confirming previous studies [22], and significant differences in PRBT, in

Eur J Pediatr

a)

b)

Fig. 3 a ROC curve to decide which test is better for ruling-in and ruling-out a NEC. Red area: test better for ruling-out. Blue area: test better for rulingin. Yellow area: test with higher discriminating ability. b Comparative table of S, SP, LH+, LH−, and DOR with 95 % CI for the NEC serum biomarkers

Eur J Pediatr

contrast to earlier studies [44]. NEC+ neonates had less platelets confirming early reports [38]. Whereas in our study NEC+ had >2-fold higher AST and ALT values than NEC−, Morini et al. did not find AST differences between NEC+ and NEC− [26]. NEC-related multiorgan failure releases enzymes from the cytosol and extramembrane domains into the blood, as occurs in intestinal/liver sepsis and nonocclusive intestinal ischemia. It is not clear whether the high AST levels derived from liver injury occurs via sepsis [42]. Some authors consider Na+ decreases as a negative prognostic biomarker, thus relevant for the correct treatment of NEC [47]. For the first time, we have detected a decreased serum K+ as relevant to NEC, confirming K+ release as essential in apoptosis [50]. The NEC+ neonates have higher glycemia than NEC−, thus confirming hyperglycemia as a biomarker of mortality or a longer hospital stay [18]. None of the other variables studied had significant differences between both groups. The AUC predicts that a neonate chosen at random from the NEC+ group has a 89 % probability of a higher CBG than another selected at random from the NEC− group (Fig. 2a). The new diagnostic test was validated by its high sensitivity (true positives vs total NEC+ neonates) S = 0.846, and specificity (true negatives vs total NEC− infants) SP = 0.859 (Fig. 2c). Not considering any clinical, radiological, or biochemical criteria and based on CBG assay alone, 11 out of 13 NEC+ and 110 out of 128 NEC− neonates would have been correctly diagnosed. In the daily clinic, when faced with a suspected case of NEC, the relevant question is as follows: with a CBG >15.6 mU/mg, what is the probability that the neonate has NEC? The answer provides clues regarding the accuracy of the diagnostic test. In the case of NEC, where there is a low prevalence (5.4 % at the HURS, 7.1 % in Spain), a CBG <15.6 mU/mg rules out NEC with 98 % accuracy, but a CBG >15.6 mU/mg does not fully confirm a NEC diagnosis (37 %). Figure 2c includes the predictive values (PPV, 37.9 %; NPV, 98.2 %), the likelihood ratios (LH+, 61.1 %; LH− 1.8 %), and the diagnostic odds ratio (DOR, 33.61 %) of this new diagnostic test. The new CBG-based test has no side effects, involves low discomfort, has a low cost ($7.9 of which $5.5 due to the Vacutainer tube), implicates a straightforward extraction, and is carried out close to real time (<2 h from extraction to results). A comparison of the main statistical properties of SLR and MLR models (Fig. 2a, b) indicated the explanatory power of the CBG test that increased only from an AUC value of 0.89 in SLR to a value of 0.90 in MLR. Despite this limited gain in AUC, thus confirming CBG as the main component, the new assay increased from being a Bgood^ diagnostic test, in the SLR model, to an Bexcellent^ diagnostic test, in the MLR model, according to Swets’s classification [46]. Other variables with significant differences between NEC+ and NEC− in the SLR model (erythrocytes, hemoglobin, hematocrit, PRBT, platelets, ALT, Na+, K+, glycemia) were identified by

MLR as confounding factors and were excluded from this model. The properties of our CBG-based test have been compared using ROC curves with other early NEC diagnostic tests, as reviewed by Evennett et al. [8], although we excluded the studies with fewer NEC+ neonates than ours, and those with less than 30 neonates. Figure 3a, b lists the previous studies of serologic tests in the diagnosis of NEC. Our study has one of the largest sample sizes, as shown by the point diameters in the ROC curve [2, 6, 7, 13, 16, 17, 23–25, 32, 34–37, 41, 45]. Sample sizes near ours have the study of Pourcyrous et al. with CRP [26, 38], that of Rabinowizt et al. with platelet-activating factor (PAF) [36], that of Ng et al. using γ-interferon-inducible protein-10 (IPF-10) [30], all these studies resulted in S/SP, LH+/ LH−, and DOR lower than ours. The study of Pourcyrous in 2005 using CRP [35] resulted in S = 0.92 and SP = 0.48 and in 1993, [24] S = 0.65 and SP = 0.49, clearly below the results of our CBG study. That of Rabinowitz using PAF [36] had a value of S = 1, higher than ours, but SP was 0.83, lower than ours. That of McLachlan et al. [25], with alkaline phosphatase (AP) had values of S = 0.53 and SP = 0.59, clearly below ours. That of Ng et al. with IPF-10 [32] had a value of S = 1, higher than ours, but SP = 0.7. Finally, the study by Ragazzi et al. [37] using leukocyte count had a value of S = 0.31, clearly below the CBG assay in the present study. The present study shows the following strengths: (1) all infants were diagnosed by the same neonatolologists, pediatric surgeons, and pediatric radiologists along the study, (2) the high number of NEC− infants included. All surgical NEC+ cases were confirmed by pathologists. Whereas our study has also limitations, including NEC+ sample size, it agrees with the low NEC prevalence in our population. Our study establishes that the NEC+ infants categorized as modified Bell’s stages II– III have 2.6-fold higher CBG activity than the NEC− controls. CBG has good properties as a single assay because it is a sensitive, specific, and LH− biomarker differentiating NEC+ from NEC−. The CBG assay is inexpensive and has few side effects. It could help neonatologists and pediatric surgeons to determine infants without NEC and preventing unnecessary treatment and days of NPO in infants with suspected NEC. To date, CBG assay discriminates better than other gut-specific sera protein biomarkers previously used for NEC diagnosis. These results warrant future larger series. Authors’ contributions Dr. Gómez-Chaparro Moreno: conceptualized and designed the study; carried out the analyses, analyzed the data, and interpreted the results; reviewed and revised the manuscript. Dr. Rodríguez Torronteras: designed the study, supervised the data collection, and interpreted the results; reviewed and revised the manuscript. Dra. Ruiz Gónzalez: designed the study and carried out the analyses and analyzed the data. Dra. Izquierdo Palomares: conceptualized and designed the study; carried out the analyses and helped in collecting the data; revised the manuscript. Dra. Ruiz Laguna: carried out the analyses and analyzed the data. Dr. Bonilla Valverde: carried out the analyses, analyzed

Eur J Pediatr the data, and interpreted the results. Prof. López-Barea: conceptualized and designed the study; supervised the data collection and interpreted the results; drafted the article and revised critically the manuscript. Prof. Delgado Rubio: revised critically the manuscript. All authors approved the final manuscript as submitted. Compliance with Ethical Standards Conflict of interest The authors declare that they have no conflicts of interest. Funding source This study was funded by grants from (1) the Health Research Fund, Carlos III Institute of Health, Health and Consumption Ministry, Spain (FIS, 05/609) and (2) the Health Agency, Andalusian Government, Spain (146/05). Ethical approval All procedures performed in studies involving human participants were in accordance with the ethical standards of the institutional RSUH research committee and with the 1964 Helsinki declaration and its later amendments. Informed consent was obtained from all neonates’ parents or legal tutor included in the study.

References 1.

2.

3.

4.

5.

6.

7.

8.

9.

10.

11.

12.

Benkoe TM, Mechtler TP, Pones M, Prusa AR, KlebermassSchrehof K, Rebhandl W, Kasper DC (2015) The plasma activities of lysosomal enzymes in infants with necrotizing enterocolitis: new promising class of biomarkers? Clin Chim Acta 438(1):279–283 Caplan MS, Sun XM, Hseuh W, Hageman JR (1990) Role of platelet activating factor and tumor necrosis factor-alpha in neonatal necrotizing enterocolitis. J Pediatr 116(6):960–964 De Graaf M, van Veen IC, van der Meulen-Muileman IH, Gerritsen WR, Pinedo HM, Haisma HJ (2001) Cloning and characterization of human liver cytosolic beta-glycosidase. Biochem J 356(Pt 3): 907–910 Dimmitt RA, Glew R, Colby C, Brindle M, Skarsgard E, Moss R (2003) Serum cytosolic beta-glucosidase activity in a rat model of necrotizing enterocolitis. Pediatr Res 54(4):462–465 Dongmei C, Guofeng L, Weilin P (2012) Serum cytosolic betaglucosidase levels in neonatal necrotizing enterocolitis. Iran J Pediatr 22(4):452–456 Edelson MB, Bagwell CE, Rozycki HJ (1999) Circulating pro- and counterinflammatory cytokine levels and severity in necrotizing enterocolitis. Pediatrics 103(4Pt1):766–771 Edelson MB, Sonnino RE, Bagwell CE, Lieberman JM, Marks WH, Rozycki HJ (1999) Plasma intestinal fatty acid binding protein in neonates with NEC: a pilot study. J Pediatr Surg 34(10):1453– 1457 Evennett N, Alexander N, Petrov M, Pierro A, Eaton S (2009) A systematic review of serologic tests in the diagnosis of necrotizing enterocolitis. J Pediatr Surg 44(11):2192–2201 Fitzgibbons SC, Ching Y, Yu D, Carpenter J, Kenny M, Weldon C, Lillehei C, Valim C, Horbar JD, Jaksic T (2009) Mortality of necrotizing enterocolitis expressed by birth weight categories. J Pediatr Surg 44(6):1072–1076 Forsythe RM, Xu DZ, Lu Q, Deitch EA (2002) Lipopolysaccharideinduced enterocyte-derived nitric oxide induces intestinal monolayer permeability in an autocrine fashion. Shock 17(3):180–184 Gareau MG, Silva MA, Perdue MH (2008) Pathophysiological mechanisms of stress-induced intestinal damage. Curr Mol Med 8(4):274–281 Gounaris A, Alexiou N, Costalos C, Daniilidou M, Frangou E, Konstandellou E (1997) Gut hormone concentrations in preterm infants with necrotizing enterocolitis. Acta Paediatr 86(7):762–763

13.

Guthmann F, Borchers T, Wolfrum C, Wustrack T, Bartholomaus S, Spener F (2002) Plasma concentration of intestinal- and liver-FABP in neonates suffering from necrotizing enterocolitis and in healthy preterm neonates. Mol Cell Biochem 239(1-2):227–234 14. Henrissat B (1991) A classification of glycosyl hydrolases based on amino acid sequence similarities. Biochem J 280(Pt 2):309–316 15. Holman RC, Stoll BJ, Curns AT, Yorita KL, Steiner CA, Schonberger LB (2006) Necrotising enterocolitis hospitalisations among neonates in the United States. Paediatr Perinat Epidemiol 20(6):498–506 16. Isaacs D, North J, Lindsell D, Wilkinson AR (1987) Serum acute phase reactants in necrotizing enterocolitis. Acta Paediatr Scand 76(6):923–927 17. Jurges ES, Henderson DC (1996) Inflammatory and immunological markers in preterm infants: correlation with disease. Clin Exp Immunol 105(3):551–555 18. Kao LS, Morris BH, Lally KP, Stewart CD, Huseby V, Kennedy KA (2006) Hyperglycemia and morbidity and mortality in extremely low birth weight infants. J Perinatol 26(12):730–736 19. Kastenberg ZJ, Sylvester KG (2013) The surgical management of necrotizing enterocolitis. Clin Perinatol 40(1):135–148 20. Kliegman RM, Fanaroff AA (1981) Neonatal necrotizing enterocolitis: a nine-year experience. Am J Dis Child 135(7):603–607 21. Kliegman RM, Fanaroff AA (1984) Necrotizing enterocolitis. New Engl J Med 310(17):1093–103 22. Kling PJ, Hutter JJ (2003) Hematologic abnormalities in severe neonatal necrotizing enterocolitis: 25 years later. J Perinatol 23(7): 523–530 23. Kultursay N, Kantar M, Akisu M, Huseyinov A, Coker I (1999) Platelet-activating factor concentrations in healthy and septic neonates. Eur J Pediatr 158(9):740–741 24. Lieberman JM, Sacchettini J, Marks C, Marks WH (1997) Human intestinal fatty acid binding protein: report of an assay with studies in normal volunteers and intestinal ischemia. Surgery 121(3):335–342 25. McLachlan R, Coakley J, Murton L, Campbell N (1993) Plasma intestinal alkaline phosphatase isoenzymes in neonates with bowel necrosis. J Clin Pathol 46(7):654–659 26. Morini F, di Crosta I, Ronchetti MP, Dituri F, Nahom A, Corchia C, Bagolan P (2008) Lactate dehydrogenase activity is increased in plasma of infants with advanced necrotizing enterocolitis. Pediatr Surg Int 24(6):705–709 27. Morowitz MJ, Poroyko V, Caplan M, Alverdy J, Liu DC (2010) Redefining the role of intestinal microbes in the pathogenesis of necrotizing enterocolitis. Pediatrics 125(4):777–785 28. Morris S, Hays W, Enomoto M, Glew R, Feddersen R, Fry D, Morris D (1999) Serum cytosolic beta-glucosidase elevation and early ischemic injury to guinea pig small intestine. Surgery 125(2):202–210 29. Neu J, Walker WA (2011) Necrotizing enterocolitis. N Engl J Med 364(3):255–264 30. Ng PC, Ang IL, Chiu RW, Li K, Lam HS, Wong RP, Chui KM, Hm C, Ng EW, Fok TF, Sung JJ, Lo YM, Poon TC (2010) Hostresponse biomarkers for diagnosis of late-onset septicemia and necrotizing enterocolitis in preterm infants. J Clin Invest 120(8): 2989–3000 31. Ng PC, Chan KYY, Poon TCW (2013) Biomarkers for prediction and diagnosis of necrotizing enterocolitis. Clin Perinatol 40(1): 149–159 32. Ng PC, Li K, Chui KM, Leung TF, Wong RP, Chu WC, Wong E, Fok TF (2007) IP-10 is an early diagnostic marker for identification of late-onset bacterial infection in preterm infants. Pediatr Res 61(1):93–98 33. Patel BK, Shah JS (2012) Necrotizing enterocolitis in very low birth weight infants: a systemic review. ISRN Gastroenterol 2012: 562594

Eur J Pediatr 34.

Pourcyrous M, Bada HS, Korones SB, Baselski V, Wong SP (1993) Significance of serial C-reactive protein responses in neonatal infection and other disorders. Pediatrics 92(3):431–435 35. Pourcyrous M, Korones SB, Yang W, Boulden TF, Bada HS (2005) C-reactive protein in the diagnosis, management, and prognosis of neonatal necrotizing enterocolitis. Pediatrics 116(5):1064–1069 36. Rabinowitz SS, Dzakpasu P, Piecuch S, Leblanc P, Valencia G, Kornecki E (2001) Platelet-activating factor in infants at risk for necrotizing enterocolitis. J Pediatr 138(1):81–86 37. Ragazzi S, Pierro A, Peters M, Fasoli L, Eaton S (2003) Early full blood count and severity of disease in neonates with necrotizing enterocolitis. Pediatr Surg Int 19(5):376–379 38. Rastogi S, Olmez I, Bhutada A, Rastogi D (2011) Drop in platelet counts in extremely preterm neonates and its association with clinical outcomes. J Pediatr Hematol/Oncol 33(8):580–584 39. Rest RF, Cooney MH, Spitznagel JK (1978) Subcellular distribution of glycosidases in human polymorphonuclear leucocytes. Biochem J 174(1):53–59 40. Richir MC, Siroen MP, van Elburg RM, Fetter WP, Quik F, Nijveldt RJ, Heij HA, Smit BJ, Teerlink T, van Leeuwen PA (2007) Low plasma concentrations of arginine and asymmetric dimethylarginine in premature infants with necrotizing enterocolitis. Br J Nutr 97(5):906–911 41. Scheifele DW (1990) Role of bacterial toxins in neonatal NEC. J Pediatr 117(1 Pt 2):S44–S46 42. Shamir R, Maayan-Metzger A, Bujanover Y, Ashkenazi S, Dinari G, Sirota L (2000) Liver enzyme abnormalities in gram-negative bacteremia of premature infants. Pediatr Infect Dis J 19(6):495–498

43.

Sharma R, Hudak ML, Tepas JJ 3rd, Wludyka PS, Marvin WJ, Bradshaw JA, Pieper P (2006) Impact of gestational age on the clinical presentation and surgical outcome of necrotizing enterocolitis. J Perinatol 26(6):342–347 44. Sharma R, Kraemer DF, Torrazza RM, Mai V, Neu J, Shuster JJ, Hudak M (2014) Packed red blood cell transfusion is not associated with increased risk of necrotizing enterocolitis in premature infants. J Perinatol 21:1–5 45. Sharma R, Tepas JJ 3rd, Hudak ML, Mollitt DL, Wludyka PS, Teng RJ, Premachandra BR (2007) Neonatal gut barrier and multiple organ failure: role of endotoxin and proinflammatory cytokines in sepsis and necrotizing enterocolitis. J Pediatr Surg 42(3):454–461 46. Swets JA (1988) Measuring the accuracy of diagnostic systems. Science 240(4857):1285–1293 47. Tepas JJ 3rd, Sharma R, Leaphart CL, Celso BG, Pieper P, Esquivia-Lee V (2010) Timing of surgical intervention in necrotizing enterocolitis can be determined by trajectory of metabolic derangement. J Pediatr Surg 45(2):310–314 48. Walsh MC, Kliegman RM (1986) Necrotizing enterocolitis: treatment based on staging criteria. Pediatr Clin North Am 33(1):179– 201 49. Walsh MC, Kliegman RM, Hack M (1989) Severity of necrotizing enterocolitis: influence on outcome at 2 years of age. Pediatrics 84(5):808–814 50. Yu SP (2003) Regulation and critical role of potassium homeostasis in apoptosis. Prog Neurobiol 70(4):363–386 51. Zamora J, Abraira V, Muriel A, Khan K, Coomarasamy A (2006) Meta-DiSc: a software for meta-analysis of test accuracy data. BMC Med Res Methodol 6:31