Keywords
necrotizing enterocolitis - neonate - very low birth weight - prematurity - natural
history
Necrotizing enterocolitis (NEC) is a life-threatening disease characterized by inflammatory
necrosis of the neonatal intestine.[1]
[2] Extremely premature and very low birth weight (VLBW) neonates are at greatest risk
for NEC due to immaturity of intestinal anatomy, physiology, and immune function,
as well as abnormal bacterial colonization of the gastrointestinal tract.[1]
[2]
[3]
[4] Interventions to prevent NEC target these risk factors and include antibiotic stewardship,
promoting use of human milk, protocolized feeding, and avoidance of medications that
alter the intestinal microbiome.[5]
[6]
[7] Despite these advances in neonatal intensive care, NEC remains a common cause of
surgical intervention, long-term morbidity, and mortality in preterm and VLBW neonates.[8]
[9]
[10]
[11]
[12]
[13]
[14]
[15]
Understanding the natural history of NEC is essential to the development of novel
interventions to prevent the disease or improve outcomes in affected neonates. Natural
history studies can illustrate how patient characteristics, biomarkers, and care processes
influence disease onset, progression, and resolution, and can frame this information
in a relevant, contemporary context.[16]
[17] To achieve these aims, a natural history study ideally would leverage a robust source
of data that reflects the full spectrum of a given disease and the general experience
of specialists who treat that disease.[18]
The goal of this study was to characterize the natural history of NEC in preterm and
VLBW neonates in a contemporary population cohort. By analyzing data from a large,
national database, we sought to describe the relationship between key patient characteristics
and the onset of NEC; provide new information about the stage-specific treatment and
progression of NEC; and report all-cause mortality, biomarkers, and various patient-important
outcomes after the diagnosis of NEC.
Materials and Methods
Study Design
We performed a historical cohort study utilizing data from the Pediatrix Clinical
Data Warehouse (CDW). The CDW includes data on more than 1 million neonates cared
for by the Pediatrix Medical Group (PMG). Data for the CDW are extracted directly
from the electronic health record (EHR) of each patient. The same propriety EHR software
system (BabySteps; MEDNAX, Inc., Sunrise, FL) is utilized at most PMG neonatal intensive
care units (NICUs). To improve validity, data extraction occurs at the end of each
patient's hospitalization, allowing providers numerous opportunities to review and
verify documentation. This study was deemed exempt by the Mayo Clinic Institutional
Review Board (Rochester, MN) as the data provided for the study was deidentified.
Study Setting and Population
We included all neonates who were admitted to a PMG NICU on day of life (DOL) 0 or
1 after being born at 23 to 29 weeks' estimated gestational age and weighing <1,500 g
at birth between 2004 and 2019. Inborn neonates who died in the delivery room were
not included in the CDW tables that we queried.
Patient Characteristics and Outcomes
We obtained baseline maternal and neonatal characteristics for each patient, including
early care processes required by the patients. The primary outcome was the maximum
stage of NEC as clinically diagnosed by the PMG neonatologists, categorized as suspected,
medical, or surgical NEC in daily clinical notes.[4] Patients with more than one NEC stage diagnosed on the day-of-onset (NEC day 0)
were classified by the highest NEC stage on that day. Patients who were diagnosed
with intestinal perforation in the absence of a diagnosis of NEC were included among
patients categorized as having no NEC, thus, patients diagnosed with spontaneous intestinal
perforation would not be misclassified as having been diagnosed with surgical NEC.[4] To build a contemporary cohort for further analysis, we used Cox's proportional
hazards regression to compare the incidence of medical and surgical NEC (together
considered “definite NEC”) across years, and characterized antibiotic utilization
among these patients on NEC day 0 or 1.
Among contemporary patients diagnosed with NEC, we assessed stage-specific progression
(i.e., progression from suspected NEC to medical or surgical NEC or progression from
medical NEC to surgical NEC), as well as survival following the diagnosis of the maximum
stage of NEC. We also characterized the antibiotic regimens and blood culture results
at the onset of NEC (NEC day 0 or 1); vasopressor and inhaled nitric oxide use on
NEC day 0 or 1; and biomarkers of post-NEC renal and hepatic function at 1-week intervals
till NEC day 28 (creatinine, aspartate aminotransferase [AST], and alanine aminotransferase
[ALT]). Last, among patients with NEC we also quantified the rates of clinically diagnosed
post-NEC stricture, post-NEC intestinal ostomy creation, and cholestasis, as well
as the rate of deoxycholate prescription.
Among all patients in the contemporary cohort, we assessed change in weight z-score between the day of birth and 36 weeks' corrected gestational age (CGA), as
well as the rate of postnatal growth restriction (defined as <3rd percentile for weight
at 36 weeks' CGA based on Olsen's growth curves).[19] We also quantified the rates of retinopathy of prematurity requiring treatment (tROP),
chronic lung disease (CLD; defined as requirement for oxygen or other respiratory
support among those still hospitalized at 36 weeks' CGA), and periventricular leukomalacia
(PVL; among those examined for PVL). Finally, among discharged patients, we measured
the presence of a gastrostomy tube and length of stay (LOS; defined as days between
birth and day of discharge).
Estimated Hospital Costs
Using the above LOS data in concert with published information regarding the mean
costs of NICU care,[20] hospital cost indices,[21] and U.S. vital statistics,[22] we estimated the percase and system-level costs of medical and surgical NEC. First,
we determined the mean per-day cost of providing care to patients <28 weeks' gestation,
irrespective of NEC status, according to Russell et al ($1,555 in 2001).[20] While a small percentage of these patients would have been diagnosed with NEC, we
considered this subgroup as being “without NEC” for use in subsequent calculations,
noting that doing so would underestimate the impact of NEC on overall cost of care.
Compared with the benchmark subgroup above, Russell et al found that patients with
NEC incurred 1.52-fold higher mean cost of hospitalization.[20] Thus, we estimated the per-day cost of providing care to patients with NEC to be
$2,364 in 2001. To determine the current per-day cost of caring for patients with
and without NEC, we adjusted for the change in the Personal Health Care-Hospital Care
price index between 2001 and 2019 (a factor of 1.65 increase).[21] We then multiplied these adjusted per-day costs by the median LOS that we observed
in patients without NEC and those with medical or surgical NEC. The difference between
these values was our estimated per-case cost of NEC. To estimate the annual cost of
NEC in the U.S. health care system, we multiplied this estimated per-case cost of
NEC by the number of live births <1,500 g as per the National Vital Statistics Report
(2019),[22] and by the combined incidence of medical and surgical NEC, we observed in our contemporary
cohort.
Data Analysis
Continuous data are summarized using means and standard deviations (SD) or medians
and interquartile ranges (IQRs) for continuous data; categorical data are summarized
using frequencies and percentages. The Aalen–Johansen method was used to calculate
the rates of suspected, medical, and surgical NEC, with patients who were transferred
out of the NICU being censored at time of transfer and in-hospital mortality being
considered a competing risk.[23] Cox's proportional hazards regression was used to assess risk factors for each NEC
stage. Proportional hazards assumptions were checked, and all assumptions were met.
In these models, we also assessed for interactions between gestational age and birth
weight.
Similar survival methods were used to assess the secondary post-NEC outcomes of stage-specific
progression, in-hospital mortality, cholestasis, ursodeoxycholate use, and post-NEC
stricture. The remaining secondary outcomes were compared between groups (based on
the maximum NEC stage diagnosed during hospitalization), using a Chi-square test for
categorical data, analysis of variance (ANOVA) for normally distributed continuous
data, and a Kruskal–Wallis test for nonnormally distributed continuous data. All tests
were two-sided, and p-values of ≤0.05 were considered statistically significant. All analyses were performed
using SAS version 9.4 software (SAS Institute, Inc.; Cary, NC) and R version 4.0.3
(R Core Team, R Foundation for Statistical Computing, Vienna, Austria).
Results
Incidence and Initial Antibiotic Treatment of Definite Necrotizing Enterocolitis:
2004–2019
To characterize the natural history of NEC, we first sought to determine a recent
epoch during which the incidence of definite NEC was fairly stable. After reaching
a peak in 2007, the annual incidence of definite NEC declined considerably, with a
stable range of 4.8 to 6.1% between 2015 and 2019 ([Fig. 1A]). Compared with 2019, there was no significant difference in the incidence of definite
NEC from 2015 to 2018 (p ≥ 0.05); however, prior to 2015, the incidence of definite NEC was significantly
higher than it was in 2019 (p < 0.001), so we limited the data for our natural history study to this contemporary
5-year period.
Fig. 1 Annual incidence of confirmed NEC within the first 90 days of life and the frequency
of antibiotic use the time of NEC onset (NEC day 0 or 1). NEC, necrotizing enterocolitis;
Pip, piperacillin; Taz, tazobactam.
The profile of antibiotics prescribed at the onset of NEC also changed between 2004
and 2019 ([Fig. 1B]). Cefotaxime use declined (21.3–1.1%), while use of piperacillin and tazobactam
increased (5.3–21.6%), though we observed only small increases in the use of cefepime
and ceftazidime ([Fig. 1B]). The rates of clindamycin and metronidazole prescription decreased and increased,
respectively, by approximately 15% over this 16-year period. In contrast, between
2015 and 2019, the frequency with which each antibiotic was used changed by less than
10%, the lone exception being vancomycin ([Fig. 1B]).
Stage-Specific Onset and Risk Factors
Among 34,032 patients in the 2015 to 2019 cohort, PMG neonatologists diagnosed 1,256
patients with suspected NEC, 1,150 patients with medical NEC, and 543 patients with
surgical NEC (cumulative incidence by DOL 90 4.2, 3.9, and 1.8%, respectively). [Supplementary Table S1] (available in the online version) presents the baseline maternal and neonatal characteristics
according to maximum NEC stage, including no NEC. For each stage of NEC, the temporal
pattern of disease onset varied according to gestational age and birth weight ([Fig. 2]). As expected, the stage-specific risk of NEC was inversely related to gestational
age and birth weight ([Fig. 2]; [Supplementary Fig. S1], available in the online version). [Supplementary Fig. S1] (available in the online version) also displays the hazard ratios and confidence
intervals (CIs) for other relevant characteristics and their association with each
NEC stage. Among all these risk factors, multivariable analysis revealed that only
four characteristics were independently associated with an increased risk of both
medical and surgical NEC: gestational age 23 to 24 weeks, birth weight <1,000 g, male
sex, and outborn birth status ([Table 1]).
Table 1
Stage-specific multivariable Cox's proportional hazards regression models
|
Suspected NEC
|
Medical NEC
|
Surgical NEC
|
|
Hazard ratio (95% CI)
|
p-Value
|
Hazard ratio (95% CI)
|
p-Value
|
Hazard ratio (95% CI)
|
p-Value
|
Singleton birth
|
1.20 (1.04–1.38)
|
0.014
|
1.09 (0.94–1.27)
|
0.24
|
1.08 (0.87–1.34)
|
0.48
|
Smoking reported
|
1.05 (0.84–1.30)
|
0.67
|
1.04 (0.83–1.31)
|
0.75
|
1.61 (1.21–2.14)
|
0.001
|
Diabetes
|
0.77 (0.62–0.96)
|
0.021
|
1.00 (0.81–1.22)
|
0.97
|
1.23 (0.92–1.64)
|
0.17
|
Antenatal steroids
|
1.44 (1.20–1.72)
|
<0.001
|
1.13 (0.94–1.36)
|
0.19
|
1.01 (0.79–1.29)
|
0.94
|
PROM (>5 days)
|
0.88 (0.76–1.02)
|
0.096
|
1.08 (0.93–1.25)
|
0.32
|
0.85 (0.67–1.07)
|
0.16
|
Chorioamnionitis
|
0.84 (0.65–1.08)
|
0.18
|
0.76 (0.57–1.00)
|
0.052
|
0.51 (0.32–0.83)
|
0.006
|
Preeclampsia
|
0.84 (0.71–1.00)
|
0.055
|
0.98 (0.82–1.16)
|
0.79
|
0.97 (0.74–1.26)
|
0.81
|
Gestational age (wk)
|
|
|
|
|
|
|
23–24
|
1.47 (1.22–1.76)
|
<0.001
|
1.38 (1.14–1.68)
|
<0.001
|
2.44 (1.84–3.21)
|
<0.001
|
25–26
|
1.08 (0.92–1.26)
|
0.37
|
1.16 (0.99–1.37)
|
0.067
|
1.55 (1.21–2.00)
|
<0.001
|
27–29
|
Reference
|
|
Reference
|
|
Reference
|
|
Gender–male
|
1.18 (1.05–1.32)
|
0.004
|
1.13 (1.00–1.27)
|
0.048
|
1.27 (1.07–1.51)
|
0.006
|
Race
|
|
|
|
|
|
|
White
|
Reference
|
|
Reference
|
|
Reference
|
|
Asian
|
1.08 (0.78–1.49)
|
0.66
|
0.90 (0.62–1.31)
|
0.58
|
0.65 (0.34–1.23)
|
0.19
|
Black
|
1.03 (0.90–1.18)
|
0.67
|
1.17 (1.02–1.35)
|
0.029
|
0.96 (0.78–1.19)
|
0.72
|
Hispanic
|
1.13 (0.97–1.32)
|
0.11
|
1.16 (0.99–1.37)
|
0.071
|
1.32 (1.06–1.66)
|
0.015
|
Other
|
0.94 (0.76–1.17)
|
0.57
|
1.00 (0.79–1.25)
|
0.97
|
1.08 (0.79–1.48)
|
0.63
|
Outborn
|
1.44 (1.23–1.68)
|
<0.001
|
1.20 (1.01–1.44)
|
0.041
|
1.38 (1.09–1.73)
|
0.007
|
Birth weight (g)
|
|
|
|
|
|
|
< 1,000
|
1.36 (1.09–1.70)
|
0.006
|
1.49 (1.18–1.87)
|
<0.001
|
2.62 (1.67–4.11)
|
<0.001
|
1,000–1,249
|
1.00 (0.80–1.24)
|
0.98
|
1.06 (0.84–1.33)
|
0.61
|
1.53 (0.97–2.41)
|
0.068
|
1,250–1,499
|
Reference
|
|
Reference
|
|
Reference
|
|
Major anomaly
|
1.27 (1.11–1.45)
|
<0.001
|
1.46 (1.27–1.68)
|
<0.001
|
1.16 (0.94–1.42)
|
0.16
|
On vent DOL 0–2
|
1.14 (0.99–1.31)
|
0.078
|
1.07 (0.92–1.23)
|
0.38
|
1.24 (0.98–1.58)
|
0.070
|
Vasopressors DOL 0–2
|
1.10 (0.95–1.27)
|
0.20
|
1.04 (0.89–1.22)
|
0.63
|
1.16 (0.94–1.42)
|
0.16
|
PDA DOL 0–2
|
1.22 (1.08–1.38)
|
0.001
|
1.10 (0.97–1.26)
|
0.13
|
0.98 (0.82–1.17)
|
0.82
|
Severe IVH DOL 0–2
|
1.19 (0.96–1.46)
|
0.11
|
1.03 (0.81–1.30)
|
0.83
|
1.70 (1.32–2.19)
|
<0.001
|
Abbreviations: CI, confidence interval; DOL, day of life; iNO, inhaled nitric oxide;
IVH, intraventricular hemorrhage; NEC, necrotizing enterocolitis; PDA, patent ductus
arteriosus; PROM, prolonged rupture of membranes.
Fig. 2 Cumulative incidence of neonates diagnosed with NEC (suspected, medical, and surgical),
accounting for the competing risk of death, by gestational age (23–24, 25–26, and
27–29 weeks) and birth weight (<1,000, 1,000–1,249, and 1,250–1,499 g). NEC, necrotizing
enterocolitis. NEC, necrotizing enterocolitis.
Stage-Specific Progression of Necrotizing Enterocolitis
Sixty-six patients initially diagnosed with suspected NEC progressed to either medical
or surgical NEC, with 4.5% (95% CI: 3.2–5.7%) progressing within 7 days of diagnosis
of suspected NEC. One hundred and three patients with medical NEC progressed to surgical
NEC (7-day rate of progression = 8.5%, 95% CI: 6.6–10.2%). We did not observe a difference
in rate-of-progression among either the gestational age or birthweight subgroups.
Please refer to [Supplementary Tables S2] and [S3] (available in the online version) for additional information on progression rates
in gestational age and birth weight subgroups.
Table 2
Stage-specific treatment and outcomes of NEC
|
Suspected (n = 1,095)
Mean ± SD)/n (%)
|
Medical (n = 990)
Mean ± SD)/n (%)
|
Surgical (n = 543)
Mean ± SD)/n (%)
|
p-Value
|
Lowest hemoglobin within 7 days prior to NEC[a]
|
10.26 ± 1.90
|
10.23 ± 2.07
|
9.64 ± 1.92
|
<0.001
|
Anemia (hemoglobin < 8) within 7 days prior to NEC[a]
|
77 (8.7)
|
69 (8.5)
|
72 (16.1)
|
<0.001
|
PDA diagnosis before NEC
|
677 (61.8)
|
511 (51.6)
|
284 (52.3)
|
<0.001
|
PDA ligated before NEC
|
72 (10.5)
|
80 (15.0)
|
34 (11.5)
|
0.034
|
IVH diagnosis before NEC
|
|
|
|
<0.001
|
Missing
|
81
|
127
|
96
|
|
0
|
622 (61.3)
|
545 (63.2)
|
243 (54.4)
|
|
1
|
134 (13.2)
|
159 (18.4)
|
54 (12.1)
|
|
2
|
111 (10.9)
|
66 (7.6)
|
52 (11.6)
|
|
3
|
59 (5.8)
|
37 (4.3)
|
35 (7.8)
|
|
4
|
88 (8.7)
|
56 (6.5)
|
63 (14.1)
|
|
Cholestasis prior to NEC
|
132 (12.1)
|
154 (15.6)
|
110 (20.3)
|
<0.001
|
Ursodeoxycholate prior to NEC
|
49 (4.5)
|
72 (7.3)
|
36 (6.6)
|
0.021
|
Positive culture on DOL 4 to NEC day 1 (out of no. with culture done)
|
182/536 (34.0)
|
203/552 (36.8)
|
108/286 (37.8)
|
0.47
|
Antibiotics on NEC day 0/1
|
|
|
|
|
Ampicillin
|
247 (22.6)
|
259 (26.2)
|
89 (16.4)
|
<0.001
|
Gentamicin
|
712 (65.0)
|
715 (72.2)
|
287 (52.9)
|
<0.001
|
Tobramycin
|
41 (3.7)
|
44 (4.4)
|
30 (5.5)
|
0.25
|
Cefepime
|
43 (3.9)
|
40 (4.0)
|
32 (5.9)
|
0.15
|
Ceftazidime
|
51 (4.7)
|
36 (3.6)
|
36 (6.6)
|
0.030
|
Clindamycin
|
83 (7.6)
|
128 (12.9)
|
96 (17.7)
|
<0.001
|
Metronidazole
|
176 (16.1)
|
264 (26.7)
|
204 (37.6)
|
<0.001
|
Piperacillin + tazobactam
|
173 (15.8)
|
188 (19.0)
|
149 (27.4)
|
<0.001
|
Meropenem
|
70 (6.4)
|
64 (6.5)
|
86 (15.8)
|
<0.001
|
Nafcillin
|
64 (5.8)
|
62 (6.3)
|
15 (2.8)
|
0.010
|
Vancomycin
|
587 (53.6)
|
546 (55.2)
|
309 (56.9)
|
0.44
|
Vasopressors NEC day 0/1
|
112 (10.2)
|
144 (14.5)
|
317 (58.4)
|
<0.001
|
iNO NEC day 0/1
|
20 (1.8)
|
11 (1.1)
|
24 (4.4)
|
<0.001
|
Post-NEC stricture[b]
|
8 (0.8)
|
18 (2.3)
|
4 (0.8)
|
0.050
|
Cholestasis after NEC (among those without it prior to NEC)[b]
|
132 (13.3)
|
165 (18.1)
|
160 (37.5)
|
<0.001
|
Ursodeoxycholate after NEC (among those without it prior to NEC)[b]
|
97 (6.3)
|
118 (6.1)
|
110 (8.2)
|
<0.001
|
Abbreviations: DOL, day of life; iNO, inhaled nitric oxide; IVH, intraventricular
hemorrhage; NEC, necrotizing enterocolitis; PDA, patent ductus arteriosus; SD, standard
deviation.
a Available in 2,148 (888 with suspected NEC, 814 with medical NEC, and 446 with surgical
NEC).
b Rates are calculated using survival methods within 30 days after NEC, p-values are from Cox's proportional hazards regression model.
Table 3
Stage-specific growth and patient-important outcomes of NEC
|
None
Median (IQR)/ mean ± SD)/n (%)
|
Suspected
Median (IQR)/ mean ± SD)/n (%)
|
Medical
Median (IQR)/ mean ± SD)/n (%)
|
Surgical
Median (IQR)/ mean ± SD)/n (%)
|
p-Value
|
Weight at 36 weeks' CGA, [a]
|
2,207 ± 360
|
2,073 ± 360
|
2,076 ± 369
|
2,014 ± 421
|
<0.001
|
Weight Z-score at birth
|
−0.1 ± 1.0
|
−0.2 ± 1.1
|
−0.2 ± 1.1
|
−0.1 ± 1.0
|
<0.001
|
Weight Z-score at 36 weeks' CGA
|
−1.1 ± 0.7
|
−1.4 ± 0.8
|
−1.3 ± 0.8
|
−1.5 ± 0.9
|
<0.001
|
Change in weight Z-score
|
−1.0 ± 0.8
|
−1.2 ± 0.8
|
−1.1 ± 0.8
|
−1.3 ± 1.1
|
<0.001
|
≤3rd percentile
|
3,032 (12.7)
|
209 (24.4)
|
167 (23.7)
|
86 (33.5)
|
<0.001
|
ROP requiring treatment[d]
|
937 (3.8)
|
57 (6.3)
|
56 (7.3)
|
54 (18.0)
|
<0.001
|
Chronic lung disease[e]
|
7,388 (35.1)
|
341 (44.1)
|
319 (49.1)
|
135 (56.0)
|
<0.001
|
Periventricular leukomalacia[f]
|
1,099 (3.9)
|
57 (5.3)
|
50 (5.2)
|
42 (7.9)
|
<0.001
|
Gastrostomy at discharge[g]
|
54 (0.2)
|
5 (0.7)
|
4 (0.7)
|
4 (2.0)
|
<0.001
|
LOS for survivors[g]
|
75 (58, 97)
|
89 (69, 114)
|
95 (74, 118)
|
122 (101, 152)
|
<0.001
|
Abbreviations: CGA, corrected gestational age; IQR, interquartile range; LOS, length
of stay, NEC, necrotizing enterocolitis; ROP, retinopathy of prematurity, SD, standard
deviation.
a Available in 23,811 without NEC, 859 with suspected NEC, 705 with medical NEC, and
257 with surgical NEC.
b Available in 20,637 without NEC, 719 with suspected NEC, 591 with medical NEC, and
207 with surgical NEC.
c Available in 21,889 without NEC, 794 with suspected NEC, 651 with medical NEC, and
229 with surgical NEC.
d Available in those with ROP evaluated (24,915 without NEC, 905 with suspected NEC,
767 with medical NEC, and 300 with surgical NEC).
e Available in those still in the neonatal intensive care unit at CGA ≥ 36 weeks (21,023
without NEC, 774 with suspected NEC, 750 with medical NEC, and 241 with surgical NEC).
f Available in those examined (28,158 without NEC, 1,069 with suspected NEC, 964 with
medical NEC, and 531 with surgical NEC).
g Among those discharged alive (23,241 without NEC, 766 with suspected NEC, 612 with
Medical NEC, and 197 with surgical NEC).
Initial Stage-Specific Treatment of Necrotizing Enterocolitis
At the onset of NEC, the frequency with which a given antibiotic was prescribed varied
according to stage of the disease ([Table 2]). Perhaps as anticipated, broad-spectrum and anaerobic coverage most commonly were
provided to patients with surgical NEC. Surgical NEC patients were four times as likely
than those with medical NEC to require support with vasopressors and inhaled nitric
oxide at the time of disease onset ([Table 2]). Among surgical NEC patients for whom procedure type was documented (n = 250), 34% were treated with peritoneal drain placement alone, 51% underwent primary
laparotomy and stoma creation, and 15% first were treated with peritoneal drainage
before subsequent laparotomy and stoma creation. Post-NEC stricture was infrequent
in both definite NEC subgroups (medical NEC = 2.3% and surgical NEC = 0.8%).
Maximum Stage-Specific Survival
Among all patients diagnosed with NEC, the all-cause mortality rate at 30 days of
postonset was greater among patients with higher acuity, maximum-stage disease (suspected
NEC = 6.1%, 95% CI: 4.7–7.5%; medical NEC = 16.4%, 95% CI :14.1–18.7%; and surgical
NEC = 43%, 95% CI: 38.4–47.3%). The median time of death was similar among all three
stages (suspected NEC = 1.5 days, IQR: 0–21 days; and medical and surgical NEC = 1
day, IQR: 0–6 days). The degree of prematurity was less consistently associated with
stage-specific mortality than was birth weight category, with patients <1,000 g at
birth most likely to die in each maximum-stage NEC category ([Fig. 3]).
Fig. 3 Stage-specific in-hospital survival following diagnosis of NEC by gestational age
(23–24, 25–26, and 27–29 weeks) and birth weight (<1,000, 1,000–1,249, and 1,250–1,499 g).
NEC, necrotizing enterocolitis.
Post–Necrotizing Enterocolitis Blood Culture Microbiology
Among NEC patients for whom a blood culture was obtained within 7 days of onset, those
with surgical NEC were twice as likely to have a positive culture as patients with
suspected or medical NEC (suspected NEC = 20.6%, medical NEC = 20.2%, and surgical
NEC = 40.2%; p < 0.001). Details on the organisms seen on positive cultures can be found in [Supplementary Table S4] (available in the online version).
Biomarkers after the Onset of Necrotizing Enterocolitis
Creatinine levels on the day-of-onset (NEC day 0) were the highest among patients
diagnosed with surgical NEC, but after 3 weeks, there were no differences between
the three NEC subgroups ([Supplementary Fig. S2], available in the online version). Both AST and ALT levels likewise were the highest
among surgical NEC patients on NEC day 0, but within 1 week, these liver enzymes were
similar to those of patients diagnosed with suspected and medical NEC ([Supplementary Fig. S2], available in the online version). Interestingly, AST and ALT levels trended up
by NEC day 21 in surgical NEC patients, with a difference in ALT still detected by
NEC day 28. This latter finding is of interest given that the new-onset diagnosis
of cholestasis and requirement for ursodeoxycholate therapy was the highest among
patients diagnosed with surgical NEC as shown in [Table 2].
Clinical Outcomes
Patients diagnosed with NEC had higher rates of postnatal growth restriction and various
clinician-reported, patient-important outcomes than those without NEC ([Table 3]). Patients in each of the NEC subgroups more often required postdischarge enteral
feeding support via gastrostomy tube, as well as prolonged hospitalization ([Table 3]). These descriptive data were derived from analyses in which we did not control
for risk factors, thus they do not indicate independent stage-specific associations
between NEC and each outcome.
Estimated Costs of Definite Necrotizing Enterocolitis
Among patients discharged home following a diagnosis of medical or surgical NEC, the
median LOS was 101 days (IQR: 79–127), or 26 days longer than discharged patients
without NEC (75 days, IQR: 58–97 days; p < 0.001). After adjusting for inflation in the Personal Health Care-Hospital Care
price index,[21] this increased LOS translates to approximately $200,000 per case of NEC. Based on
the 4.8% incidence of definite NEC that we observed in our patients born in 2019,
and given that 51,716 neonates were born at <1,500 g that same year[22] that the total cost of NEC to the American Health Care System could approach or
even exceed $500 million annually.
Discussion
Trends in the Incidence and Initial Treatment of Definite Necrotizing Enterocolitis
Among a large cohort of extremely premature neonates, we observed a recent decline
in the rates of medical and surgical NEC,[10]
[24] with the incidence of disease leveling off between 2015 and 2019. As previously
described for patients in this cohort,[24] favorable trends in the use of human breast milk and avoidance of prolonged early
empiric antibiotic treatment likely contributed to these improved outcomes.[25]
[26]
[27]
[28]
[29]
[30]
[31]
[32] Over the same period of time, we observed shifts in the antibiotics selected to
cover gram-negative and anerobic organisms at the onset of definite NEC. Given that
the incidence of definite NEC and antibiotic prescription patterns were fairly stable
between 2015 and 2019, we studied the stage-specific, natural history of NEC among
patients born in this 5-year period. In the subsections below, we first discuss matters
related to medical and surgical forms of NEC (“definite NEC”), then briefly address
the more challenging topic of suspected NEC.
Risk Factors for Definite Necrotizing Enterocolitis
The risk of medical or surgical NEC was inversely related to gestational age and birth
weight, as expected, with distinctly higher risk among patients born at 23 to 24 weeks
or birth weight <1,000 g. Surprisingly, among patients with definite NEC, there was
no significant interaction between gestational age and birth weight among patients
with definite NEC (p-values >0.65). Our multivariable analysis likewise did not identify small for gestational
age birth size as an independent risk factor for NEC. These findings clarify that
among all VLBW neonates those who are least mature and the lowest birth weight require
special consideration in clinical processes, such as standardized feeding protocols.[24]
Our analysis of other risk factors allowed us to build on our earlier work and compare
our current findings to others' observations. While Fang et al identified a small
increase in the risk of NEC among outborn VLBW neonates born between 2000 and 2014,[33] our present study suggests that this outcome disparity increased over the subsequent
5 years ([Table 1]). The presence of congenital anomalies is another baseline patient characteristic
that others have associated with NEC,[34]
[35] though we observed only an increased risk of medical NEC after adjusting for other
variables. Perhaps due to differences in study design, we did not identify early exposure
to inotropes as an independent risk factor for definite NEC as did Wong et al.[36] Because this care process may influence the risk of NEC, we analyzed the rate at
which at least one inotrope was prescribed DOL 0 to 2 and found that it decreased
over time (2004–2014: 24.4% v. 2015–2019: 18.1%, p < 0.001). Among patients with definite NEC, prior diagnosis of anemia (Hgb <8 g/dL)
within 1 week prior to the onset of NEC were most common among patients with surgical
NEC.[37] Similarly, we identified that preexisting cholestasis may be a diagnostic harbinger
of more advanced forms of NEC to our knowledge, a novel finding.
Stage-Specific Onset, Treatment, and Progression of Definite Necrotizing Enterocolitis
The temporal pattern of disease onset differed considerably between medical and surgical
NEC. The risk of medical NEC increased linearly among all gestational age and birth
weight subgroups over the second and third week of life, at which point the risk began
to plateau for patients born at >26 weeks or >1,000 g. The risk for less mature and
lower birth weight patients did not appreciably level off until nearly 8 weeks of
life. The onset of surgical NEC was characterized by immediate distinctions in risk
among the three gestational age subgroups during the second week of life. Among the
birth weight subgroups, there was a clear difference between patients born <1,000 g
and patients of greater birth weight. Understanding these stage-, gestational age–,
and birth weight–specific patterns of onset could improve the precision with which
clinical neonatologists discuss prognosis with VLBW patients' families.
While the initial antibiotic treatment of definite NEC has stabilized in recent years
([Fig. 1B]), we observed a broad array of drug classes prescribed at the onset of both medical
and surgical NEC ([Table 2]). This variability in antibiotic prescription patterns mirrors recent reports from
single centers,[38]
[39] and likely reflects the lack of clear evidence on which to develop specific guidance.[40]
[41] With this in mind, we note that approximately one in nine patients with medical
NEC progressed to surgical NEC in our cohort, most within 1 week of the onset of medical
NEC. Bacteremia was most common among surgical NEC patients in our cohort as well.[42] Thus it is important to be mindful of VLBW patients' response to initial antibiotic
regimens, with careful attention for signs of progression and surveillance for bacteremia,
as changes in antibiotic coverage may be warranted.
Stage-Specific Survival in Definite Necrotizing Enterocolitis
Among patients with definite NEC most mortality occurred within a few days of disease
onset. There was an inverse, stratal relationship between survival and both gestational
age, and birth weight among patients with medical NEC. This was less, so the case
for surgical NEC. Survival was similar among the least mature and lowest birth weight
subgroups, and patients >1,250 g gestation were notably most likely to survive surgical
NEC. The relatively high risk of mortality among surgical NEC patients appears to
be reflected in the high rates of bacteremia (40.2%) and requirement for vasopressors
(58.4%) that we observed after disease onset. Late-onset sepsis, particularly with
gram-negative organisms, like those we observed, is strongly associated with death
in neonates,[43] while vasopressor treatment has been independently associated with death among patients
with NEC.[14]
Biomarkers of Disease: Relevance to Surgical Necrotizing Enterocolitis
Creatinine levels on the day of diagnosis suggested that surgical NEC patients may
have experienced some degree of acute kidney injury at the time of disease onset.
While their creatinine levels were similar to those of suspected and medical NEC patients
after 3 weeks, it is conceivable that this biomarker portended an increased LOS among
surgical NEC patients.[44] Our limited transaminase data also suggest that surgical NEC patients experienced
transient liver injury on the day of diagnosis, perhaps followed by the evolution
of more chronic liver disease. Indeed, the prevalence of cholestasis among surgical
NEC patients was twice that of patients with medical NEC, and quite similar to that
of surgical NEC patients in a contemporary, prospectively studied cohort.[42]
Growth and Other Patient-Important Outcomes of Definite Necrotizing Enterocolitis
Consistent with other contemporary studies of growth outcomes among VLBW neonates,[45]
[46] we identified significant postnatal growth failure among patients with medical and
surgical NEC. Severe restriction (<3rd percentile) in weight was most prevalent among
surgical NEC patients. We attribute this finding to the high prevalence of intestinal
stoma in our cohort (17%),[47] but cannot resolve whether this factor per se contributes to the adverse neurodevelopmental
outcomes observed in patients with surgical NEC.[42] Nevertheless, the observed restriction in postnatal growth is especially concerning,
considering the high rates of tROP, CLD, and serious brain injury (severe intraventricular
hemorrhage and PVL) among surgical NEC patients.[48]
[49]
Length of Stay and Estimated Costs of Definite Necrotizing Enterocolitis
In our cohort, the diagnosis of definite NEC was associated with a 26-day increase
in the median LOS. While several multicenter studies have demonstrated similarly prolonged
hospitalizations,[10]
[42]
[44]
[50] the literature regarding the surgical and overall hospital costs of NEC is limited.[20]
[51] While Russell et al provided LOS and mean hospital cost data,[20] we only could estimate and compare the per-day cost of hospitalization for patients
with and without NEC. To understand the cost implications of NEC more clearly, future
studies of the Nationwide Inpatient Sample or other government databases will be essential.[20]
[51] For now, though, it seems as though our approach to cost estimation is valid, when
comparing our patients' likely total cost of care with those reported by Stey et al.[51]
Suspected Necrotizing Enterocolitis
There is virtually no literature describing the risk factors or natural history of
suspected NEC, perhaps because the nonspecific nature of this diagnosis has prevented
or disincentivized investigation into its origins and outcomes.[3] While patients diagnosed with suspected NEC might be affected by other non-NEC diagnoses
(e.g., bacteremia with septic ileus and feeding intolerance of prematurity),[3] we reasoned that these selected patients could serve as a “least acute” comparator
for patients with medical or surgical NEC, a group that may be considered collectively
as those with “definite” forms of NEC.[37]
We were surprised to see many similarities between patients with suspected and medical
NEC. There was substantial overlap in the patient characteristics of these two subgroups,
as well as similar patterns of disease onset, early care requirements, survival, biomarkers,
and outcomes of disease. These similarities could indicate that suspected and medical
NEC exist on a continuum of disease (i.e., infectious inflammation that does not lead
to intestinal perforation) or they could reflect patients' risks for and gastrointestinal
complications of other life-threatening systemic diseases (e.g., bacterial sepsis).
Without better diagnostic and prognostic tools for NEC,[52] it is unlikely that we will resolve this question with certainty.
Limitations
We acknowledge that our analyses are limited by the diagnostic imprecision inherent
to NEC.[53] In our classification system, we applied concepts and methods previously described
for the study of NEC using CDW data[3]
[4] and critically assessed prior studies of NEC outcomes that leveraged these data.[14]
[24]
[33]
[54] We also could not resolve exactly which surgical procedures were required for a
given patient, and recognize that initial placement of a peritoneal drain versus laparotomy
is associated with different clinical outcomes among extremely low birth weight neonates.[42] Lastly, while there is a considerable interest in the role of probiotics might play
in the prevention of NEC, we did not analyze probiotic exposure among patients in
our cohort. Prior study of CDW patients revealed significant variability in probiotic
organisms and no information about dosing,[55] and given recent guidance from the American Academy of Pediatrics, our decision
seemed prudent.[56]
Conclusion
We present the natural history of NEC in large, contemporary cohort of VLBW patients
who received care from neonatologists who practice in more than 300 NICUs in the United
States. Among these patients, the incidence and initial antibiotic treatment were
stable over a period of 5 years. Studying these patients allowed us to characterize
in detail the stage-specific patterns of onset, progression, and survival of NEC,
including novel information about suspected NEC. Perhaps most importantly, we identified
the extent to which all stages of NEC impact the in-hospital outcomes of these patients,
as they represent key opportunities for disease prevention in future clinical studies.