The authors have declared that no competing interests exist.
Conceived and designed the experiments: MS EMWK AFB JBFH. Performed the experiments: MS RK JBFH. Analyzed the data: MS RK HG JBFH. Contributed reagents/materials/analysis tools: RK. Wrote the paper: MS EMWK CVH AFB EH JBFH.
Early NEC symptoms are non-specific and diagnostic tests lack discriminative power. Intestinal fatty acid-binding protein (I-FABP), mainly located in small bowel enterocytes, is released into the blood following NEC-associated enterocyte disruption. Aim of this prospective cohort trial was to determine the diagnostic value of I-FABP measured in plasma (I-FABPp) and urine (I-FABPu) for the presence of NEC, to evaluate I-FABP levels during NEC development, and to assess its prognostic value for the progression from suspected to complicated disease.
Between 2010 and 2012 we prospectively enrolled neonates with suspected NEC. We measured I-FABP levels eight-hourly from onset of suspected NEC for at least 48 hours, or until surgery. NEC diagnosis was confirmed radiologically or during operation. We defined NEC as complicated if it resulted in surgery and/or death. We determined disease course and diagnostic I-FABP cut-off points.
The study comprised 37 neonates (24M, 13F), gestational age 28 (24–36) weeks, birth weight 1190 (570–2,400) grams. We found significantly higher I-FABPp and I-FABPu levels in NEC patients (n = 22) than in patients with other diagnoses (n = 15). Cut-off values for diagnosing NEC were 9 ng/mL I-FABPp and 218 ng/mL I-FABPu, with corresponding likelihood ratios (LRs) of 5.6 (95% CI 0.89–35) and 5.1 (95% CI 0.73–36), respectively. I-FABP levels were highest in the first eight hours after symptom onset and gradually decreased over time. Cut-off values for complicated disease were 19 ng/mL I-FABPp and 232 ng/mL I-FABPu, with LRs of 10 (95% CI 1.6–70) and 11 (95% CI 1.6–81), respectively.
Both plasma and urinary I-FABP levels specifically identify NEC in preterm infants prior to appearance of diagnostic radiological signs suggestive for NEC. Moreover, serial I-FABP measurements accurately predict development of complicated disease.
Early symptoms of necrotizing enterocolitis (NEC) are often non-specific, such as abdominal distension, bloody stools, or gastric retention [
Because NEC is characterized by loss of bowel wall integrity, intestinal fatty acid-binding protein (I-FABP) is one of the more promising biomarkers. This small cytosolic protein, located mainly in enterocytes of the small intestine, is released into the blood stream after cell disruption [
While several studies investigated the discriminative power of I-FABP for NEC at onset of disease, no study evaluated changes in I-FABP levels during its development. As NEC is often a progressive disease, consecutive measurements might offer more detailed information about the disease course than a single measurement at first symptoms.
Our aim was threefold: to determine the diagnostic value of plasma and urine I-FABP for the presence of definite NEC, to evaluate plasma and urine I-FABP levels during NEC development, and to assess the prognostic value of plasma and urine I-FABP for the progression from suspected to complicated disease (as defined by disease resulting in a surgical intervention and/or death).
The study has been approved by the Institutional Review Board of the University Medical Centre Groningen, the Netherlands, and has been conducted according to the principles expressed in the Declaration of Helsinki. Written informed consent was obtained from (the parents of) all participants.
As part of the prospective cohort trial ‘NoNEC’ (number NTR3239 in the Dutch Trial Registry), all consecutive patients with suspected NEC, admitted to the neonatal intensive care unit (NICU) of University Medical Centre Groningen between October 2010 en October 2012, were included [
All data were prospectively collected and included demographics (gender, gestational age, and birth weight), other patient data (type of delivery, medication, and co-morbidity), maternal data, and disease characteristics.
On first presentation of symptoms, all patients were treated according to hospital protocol: bowel rest (i.e. nil by mouth and decompression by nasogastric suction), broad-spectrum antibiotics (after cultures were obtained), pain management, and respiratory and/or cardiovascular support if necessary. Regular diagnostic laboratory and radiological studies, such as an abdominal X-ray, were performed. The latter was regarded as onset of symptoms, because the time of the abdominal X-ray was well documented in the digital hospital information system and it was performed immediately after suspicion of NEC arose. Specifying the exact time of symptom onset was considered essential information for the purpose of delineating the course of disease. If a patient was transferred to our centre because of suspected NEC, time of abdominal X-ray in the referring hospital was considered as time of symptom onset. Following as soon as possible the first study samples of blood and urine were obtained.
Diagnosis of NEC was defined as the presence of pneumatosis intestinalis on abdominal X-ray (and/or peroperatively confirmed) and corresponding to Bell’s stage II or higher [
Together with every routine diagnostic blood analysis an extra sample of 100 μL was obtained in an EDTA tube for study purposes. Urine samples were collected at regular intervals by placing cotton in the patient’s nappy. Once it was saturated with urine, the urine was gently squeezed into a sterile syringe and then pressed in a standard urine tube. In patients with an indwelling catheter, urine was collected directly from the catheter. Blood and urine samples were put on ice and sent to the laboratory without delay. Ideally, both blood and urine samples were collected every eight hours. If, however, the clinical condition of the patient dictated otherwise, only a urine sample was obtained, thus minimizing the burden for the patient.
Sampling blood and urine for study purposes ended either if ten consecutive samples of blood and/or urine had been obtained during at least 48 hours, if the patient underwent a surgical intervention (because such an intervention would of itself probably cause a non-physiologic rise in I-FABP levels), or if the patient died.
In the laboratory, a blood sample was fractioned by centrifuging it for 10 minutes at approximately 2,000 x
Because this study was essentially exploratory in nature, we refrained from a power analysis.
Statistical analysis was performed using the Statistical Package for the Social Sciences (IBM SPSS Statistics 21, IBM Corp., Armonk, New York, USA). All data are presented as median with (range), unless specified otherwise. Two-tailed
To determine the clinical use of plasma and urine I-FABP levels as a diagnostic test for suspected NEC, positive and negative likelihood ratios were computed after constructing receiver operating characteristic curves. Cut-off points were determined on the basis of the optimum sum of sensitivity and specificity. To avoid erroneously establishing the diagnosis of NEC or unnecessary operative interventions based on false positive I-FABP values, a high specificity was preferred. To determine the clinical use of plasma and urine I-FABP levels as a prognostic test for complicated disease, similar methods were applied.
For the purpose of delineating the course of disease, I-FABP measurements were classified in six time slots, i.e. 0–8 h, 8–16 h, 16–24 h, 24–36 h, 36–48 h, and >48 h after symptom onset, based on the number of hours that had elapsed since symptom onset (identified by the time of first abdominal X-ray). Samples were classified according to these time slots.
During the study period 53 consecutive patients with suspected NEC were admitted to the NICU and eligible for inclusion. Thirteen patients could not be included: 12 because informed consent was withheld and one because of the already terminal nature of the prognosis on presentation (
In
N = 37 | |
---|---|
Gestational age | 28 wk (24–36) |
Birth weight | 1,190 g (570–2,400) |
Gender: male/female | 24/13 |
Postnatal age at first symptoms | 9 d (3–34) |
Diagnosis of NEC: | 22/37 (59%) |
Final Bell’s stage IIA | 9/22 (41%) |
IIB | 2/22 (9%) |
IIIA | 2/22 (9%) |
IIIB | 9/22 (41%) |
NICU stay | 27 d (8–102) |
Values are expressed as median (range). NICU = neonatal intensive care unit.
The remaining 15 patients were diagnosed with ileus caused by sepsis e.c.i. (e causa ignota; n = 3), delayed passage of meconium (n = 2), bloody stool e.c.i. (n = 2), CPAP belly (n = 2), (viral) gastroenteritis (n = 2), spontaneous intestinal perforation (SIP; n = 1), and sigmoid volvulus (n = 1). In two patients no definite diagnosis could be made.
Between 0–8 h after symptom onset, I-FABPp (17 samples) and I-FABPu (16 samples) values were 35 (2.2–370) and 491 (7.5–2 116) ng/mL in NEC patients (n = 22), versus 4.6 (0.41–19) and 26 (3.3–218) ng/mL in patients without NEC. Only I-FABPp values reached statistical significance when both groups were compared (
Cut-off points to differentiate between groups are represented by dotted lines (after logarithmic transformation: cut-off points at 3.2 and 6.4 ng/mL correspond to cut-off points at 9 and 218 ng/mL, as mentioned in tables
A. Based on I-FABPp | |||||||
---|---|---|---|---|---|---|---|
h | n | Median values (range) | Cut-off values (ng/mL) | Sensitivity (%) | Specificity (%) | LR+ (95% CI) | LR- (95% CI) |
0–8 | 17 | 8.6 (0.41–370) | 9 | 80 | 86 | 5.6 (0.89–35) | 0.23 (0.06–0.84) |
8–16 | 20 | 11 (0.44–3748) | 11 | 69 | 86 | 4.8 (0.76–31) | 0.36 (0.15–0.85) |
16–24 | 19 | 7.3 (1.1–809) | 10 | 64 | 88 | 5.1 (0.77–34) | 0.42 (0.18–0.94) |
24–36 | 27 | 3.4 (0.25–113) | 7 | 53 | 90 | 5.3 (0.78–36) | 0.52 (0.31–0.88) |
36–48 | 25 | 2.0 (0.79–35) | 2 | 67 | 86 | 4.7 (0.74–29) | 0.39 (0.19–0.78) |
>48 | 30 | 2.8 (1.1–141) | 3 | 67 | 75 | 2.7 (0.95–7.5) | 0.44 (0.22–0.90) |
I-FABPp = intestinal fatty acid-binding protein in plasma; LR+ = positive likelihood ratio; LR- = negative likelihood ratio; CI = confidence interval. |
B. Based on I-FABPu | |||||||
---|---|---|---|---|---|---|---|
h | n | Median values (range) | Cut-off values (ng/mL) | Sensitivity (%) | Specificity (%) | LR+ (95% CI) | LR- (95% CI) |
0–8 | 16 | 69 (3.3–2116) | 218 | 57 | 89 | 5.1 (0.73–36) | 0.48 (0.20–1.2) |
8–16 | 23 | 36 (3.2–23,337) | 90 | 58 | 91 | 6.4 (0.93–44) | 0.46 (0.23–0.91) |
16–24 | 24 | 51 (1.9–18,267) | 52 | 73 | 89 | 6.6 (1.0–43) | 0.30 (0.13–0.71) |
24–36 | 32 | 26 (1.3–1 620) | 85 | 28 | 93 | 3.9 (0.51–30) | 0.78 (0.58–1.1) |
36–48 | 30 | 16 (1.8–857) | 34 | 53 | 91 | 5.8 (0.85–39) | 0.52 (0.32–0.85) |
>48 | 34 | 17 (0.56–3596) | 96 | 21 | 93 | 3.2 (0.39–25) | 0.85 (0.66–1.1) |
I-FABPu = intestinal fatty acid-binding protein in urine; LR+ = positive likelihood ratio; LR- = negative likelihood ratio; CI = confidence interval. |
I-FABPp and I-FABPu levels were highest in the first 24 hours after symptom onset (as defined by the first abdominal X-ray) and tended to decrease gradually over time (
A. Based on I-FABPp. B. Based on I-FABPu.
The specificity for I-FABP to determine whether patients would develop complicated NEC gradually increased over time, ranging from 75% at onset of disease to 91% after 48 hours (
A. Based on I-FABPp | |||||||
---|---|---|---|---|---|---|---|
h | n | Median values (range) | Cut-off values (ng/mL) | Sensitivity (%) | Specificity (%) | LR+ (95% CI) | LR- (95% CI) |
0–8 | 10 | 35 (2.2–370) | 53 | 67 | 75 | 2.7 (0.45–16) | 0.44 (0.12–1.6) |
8–16 | 13 | 53 (0.54–3748) | 19 | 88 | 80 | 4.4 (0.74–26) | 0.16 (0.02–1.1) |
16–24 | 11 | 27 (1.1–809) | 26 | 86 | 75 | 3.4 (0.61–19) | 0.19 (0.03–1.4) |
24–36 | 16 | 11 (0.86–113) | 29 | 43 | 90 | 4.3 (0.55–33) | 0.63 (0.33–1.2) |
36–48 | 18 | 2.9 (0.89–1620) | 4 | 63 | 90 | 6.3 (0.90–43) | 0.42 (0.17–1.0) |
>48 | 18 | 3.6 (1.1–141) | 9 | 43 | 91 | 4.7 (0.60–37) | 0.63 (0.33–1.2) |
I-FABPp = intestinal fatty acid-binding protein in plasma; LR+ = positive likelihood ratio; LR- = negative likelihood ratio; CI = confidence interval. |
B. Based on I-FABPu | |||||||
---|---|---|---|---|---|---|---|
h | n | Median values (range) | Cut-off values (ng/mL) | Sensitivity (%) | Specificity (%) | LR+ (95% CI) | LR- (95% CI) |
0–8 | 7 | 491 (7.5–2116) | 687 | 67 | 75 | 2.7 (0.41–17) | 0.44 (0.08–2.6) |
8–16 | 12 | 161 (3.2–23,337) | 232 | 71 | 80 | 3.6 (0.58–22) | 0.36 (0.10–1.3) |
16–24 | 15 | 103 (1.9–18,267) | 103 | 88 | 86 | 6.1 (0.98–38) | 0.15 (0.02–0.9) |
24–36 | 18 | 32 (1.3–1620) | 222 | 57 | 91 | 6.3 (0.87–45) | 0.47 (0.20–1.1) |
36–48 | 19 | 34 (1.8–857) | 177 | 38 | 91 | 4.1 (0.52–33) | 0.69 (0.39–1.2) |
>48 | 19 | 19 (0.56–3596) | 46 | 90 | 91 | 5.5 (0.75–40) | 0.55 (0.27–1.1) |
I-FABPu = intestinal fatty acid-binding protein in urine; LR+ = positive likelihood ratio; LR- = negative likelihood ratio; CI = confidence interval. |
Within eight hours after symptom onset, I-FABPp and I-FABPu values were higher in patients with complicated disease (n = 12: 11 infants with complicated NEC and one infant with a spontaneous intestinal perforation, SIP) than in patients without complicated disease, although statistical significance was not unequivocal for both I-FABPp and I-FABPu (
Between 8–16 h after symptom onset, I-FABPp and I-FABPu values were 103 (10–3,748) and 517 (3.2–23,337) ng/mL in patients with complicated disease versus 4.0 (0.44–70) and 22 (3.2–232) ng/mL in patients without complicated disease with associated
A. Based on I-FABPp | |||||||
---|---|---|---|---|---|---|---|
h | n | Median values (range) | Cut-off values (ng/mL) | Sensitivity (%) | Specificity (%) | LR+ (95% CI) | LR- (95% CI) |
0–8 | 17 | 8.6 (0.41–370) | 53 | 67 | 91 | 7.3 (1.0–52) | 0.37 (0.12–1.2) |
8–16 | 20 | 11 (0.44–3748) | 19 | 88 | 92 | 10 (1.6–70) | 0.14 (0.02–0.87) |
16–24 | 19 | 7.3 (1.1–809) | 26 | 86 | 92 | 10 (1.5–69) | 0.16 (0.03–0.97) |
24–36 | 27 | 3.4 (0.25–113) | 29 | 43 | 95 | 8.6 (1.1–70) | 0.60 (0.32–1.2) |
36–48 | 25 | 2.0 (0.79–35) | 4 | 56 | 94 | 8.9 (1.2–65) | 0.47 (0.23–0.99) |
>48 | 30 | 2.8 (1.1–141) | 10 | 50 | 95 | 11 (1.4–84) | 0.52 (0.26–1.1) |
I-FABPp = intestinal fatty acid-binding protein in plasma; LR+ = positive likelihood ratio; LR- = negative likelihood ratio; CI = confidence interval. |
B. Based on I-FABPu | |||||||
---|---|---|---|---|---|---|---|
h | n | Median values (range) | Cut-off values (ng/mL) | Sensitivity (%) | Specificity (%) | LR+ (95% CI) | LR- (95% CI) |
0–8 | 16 | 69 (3.3–2116) | 687 | 67 | 92 | 8.7 (1.1–67) | 0.36 (0.07–1.8) |
8–16 | 23 | 36 (3.2–23,337) | 232 | 71 | 94 | 11 (1.6–81) | 0.30 (0.09–0.99) |
16–24 | 24 | 51 (1.9–18,267) | 103 | 88 | 94 | 14 (2.1–95) | 0.13 (0.02–0.84) |
24–36 | 32 | 26 (1.3–1620) | 222 | 57 | 96 | 14 (1.9–108) | 0.47 (0.19–1.1) |
36–48 | 30 | 16 (1.8–857) | 177 | 38 | 95 | 8.3 (1.0–68) | 0.65 (0.38–1.1) |
>48 | 34 | 17 (0.6–3 596) | 96 | 44 | 96 | 11 (1.4–87) | 0.58 (0.32–1.0) |
I-FABPu = intestinal fatty acid-binding protein in urine; LR+ = positive likelihood ratio; LR- = negative likelihood ratio; CI = confidence interval. |
When the analysis is being limited to (all) NEC patients, 11 patients with complicated NEC versus 11 patients with uncomplicated NEC showed statistically significant different I-FABP levels, whether measured in plasma (103 [10–3,748] ng/mL versus 4.1 [0.54–70] ng/mL; p = 0.019) or urine (517 [3.3–23,337] ng/mL versus 35 [3.2–232] ng/mL; p = 0.042) at 8–16 h after onset of disease (
Cut-off points to differentiate between groups are represented by dotted lines (after logarithmic transformation: cut-off points at 3.9 and 6.5 ng/mL correspond to cut-off points at 19 and 232 ng/mL, as mentioned in tables
Ten out of 37 patients (27%) underwent a surgical intervention (i.e. laparotomy; no peritoneal drainage was performed in this patient cohort). Nine (90%) patients were operated because of NEC. This intervention occurred median 55 hours (3–823) after symptom onset. The patient who underwent a laparotomy after 823 hours was initially conservatively treated but turned out to have suffered a NEC totalis. He died shortly after the operation. None of the patients were unable to undergo surgical intervention, e.g. because of contra-indications, and subsequently succumbed. Three patients underwent surgery because of free intra-peritoneal air and six patients because of clinical deterioration despite maximal conservative therapy. In four of the latter patients a covered perforation was found during surgery. Median I-FABP levels in plasma and urine were not statistically different between patients with and without a radiologically proven perforation. The tenth patient was operated because of SIP 177 hours after symptom onset. The initial plasma and urine I-FABP levels of this particular patient were 0.79 ng/mL and 4.5 ng/mL, respectively, and highest preoperative I-FABP levels were 13 ng/mL and 63 ng/mL, respectively.
Patient outcome is shown in
Our study shows that in neonates classified as Bell’s stage I, I-FABP measured in plasma at symptom onset can identify those patients in whom non-specific symptoms will evolve into definite NEC. A cut-off point for I-FABPp was determined at 9 ng/mL, associated with a positive LR of 5.6. The corresponding high specificity minimizes the chance of erroneously establishing the diagnosis of NEC.
Several earlier studies reported on I-FABP measured in plasma and serum [
Urine analysis represents a good alternative for I-FABPp measurements because every blood sample taken means an extra burden for the neonate [
Although no interventions other than the aforementioned conservative measures (i.e. gastric decompression and broad-spectrum antibiotics) exist to prevent the progression of NEC, I-FABP, whether measured in plasma or urine, provides the clinician with a valuable tool: if patients, who eventually do not develop NEC, can be identified at the time when only non-specific symptoms are present, NEC-specific treatment can be avoided and duration of conservative treatment can be limited. Re-introducing enteral feeding and early discontinuation of antibiotic treatment would be among the main benefits. Alternatively, if progression to NEC is suspected, necessary measures, such as an earlier surgical intervention, could be warranted. As already mentioned, cut-off values for I-FABP determined in our study correspond to a high specificity to minimize the chance of erroneously establishing the diagnosis of NEC. By performing serial I-FABP measurements over a short course of time (e.g. every 4–8 hours) the chance of false negative results will be further minimized. If the cut-off values calculated in the first time slot (i.e. 0–8 h after onset of symptoms) would have been applied to all following time slots, results showed very low sensitivity values in combination with specificity values soon equalling 100%, thereby rendering calculations of positive likelihood ratios impossible. Cut-off values for each time slot were therefore calculated.
To the best of our knowledge, our study is the first to demonstrate I-FABP levels in plasma and urine during the course of NEC development. One of the most striking characteristics of I-FABP levels plotted against time is the peak early-on in the course of disease, i.e. within 24 hours after symptom onset. From that moment on, the gradual decline in I-FABP levels is only interrupted by minor fluctuations. A possible explanation for this course in I-FABP levels starts with the fact that I-FABP is expressed in mature enterocytes (rather than in the crypts) and that these cells are located at the most distal point from mucosal blood supply [
Within 8–16 hours after symptom onset, I-FABP, whether measured in plasma or urine, could predict the development of complicated disease with cut-off values of 19 ng/mL I-FABPp and 232 ng/mL I-FABPu, with associated positive LRs of 10 (1.6–70) and 11 (1.6–81), respectively. By classifying patients according to the previously mentioned method, even patients without NEC, but with an indication for laparotomy (e.g. SIP), are being acknowledged. This also holds true when we focus on NEC patients. Patients who develop complicated NEC can also be identified early during the course of disease. It is tempting to speculate that measuring I-FABP levels might help to identify patients undergoing early surgery, thus possibly preventing further clinical deterioration. A specific example from our study cohort clearly stresses the importance of previously mentioned speculation: I-FABP levels of the patient with NEC totalis, who was operated on (after 823 hours) following a protracted conservative treatment (see also paragraph ‘Therapy and outcome’ of the Results), were 91 ng/mL and 2,116 ng/mL, as measured in plasma and urine at onset of symptoms, respectively. Acknowledgment of these very high I-FABP levels would perhaps have justified a much earlier surgical intervention.
In many centres clinical deterioration despite maximal conservative therapy is considered an indication for surgical intervention. In our study, there was no difference in urine or plasma I-FABP levels between patients in whom a radiologically proven perforation was the indication for surgery and patients in whom clinical deterioration despite maximal therapy warranted an intervention. However, numbers are limited and further research is needed, as deciding to take a patient to theatre in absence of signs of a perforation forms one of the major challenges for physicians treating children with NEC.
We recognize some limitations of our study. First of all, it was limited by the relative small number of patients. Secondly, NEC in preterm or in full term infants is considered a distinct entity [
Intestinal fatty acid-binding protein, whether measured in plasma or urine, identifies NEC patients among preterm infants with non-specific Bell’s stage I symptoms. Even more so, in NEC patients I-FABP levels predict complicated disease during the first stages of NEC. By using serial measurements and clearly defined cut-off values, the course of disease can be closely monitored and predicted. Highest I-FABP levels are present during the first 24 hours after symptom onset and followed by gradually diminishing I-FABP levels.
(XLSX)
The authors thank Dr. T. Brantsma—Van Wulfften Palthe for correcting/editing the English manuscript.