Enteral human IgG for prevention of necrotising enterocolitis

Document Sample
Enteral human IgG for prevention of necrotising enterocolitis Powered By Docstoc
					Enteral human IgG for prevention of necrotising enterocolitis:
a placebo-controlled, randomised trial
Gregor Lawrence, David Tudehope, Kathryn Baumann, Heather Jeffery, Andrew Gill, Michael Cole, John
Drew,Andrew McPhee, John Ratcliffe, Graham Reynolds, John Simes, Cheryl Swanson, David Cartwright,
Peter Davis,Ian Humphrey, Andrew Berry
Queensland Institute of Medical Research, PO Royal Brisbane Hospital, Australia
Lancet 2001; 357: 2090–94

Necrotising enterocolitis causes great morbidity and mortality in preterm infants in the first weeks of life, and is the
most common condition requiring emergency surgery in neonatal intensive care. 5–10% of very-lowbirthweight infants
develop definite necrotizing enterocolitis, with mortality around 20%. 1 The pathogenesis of necrotising enterocolitis is
not well understood, and there are probably various mechanisms responsible for the development of the
patchy gut wall necrosis seen in necrotizing enterocolitis.1 Prematurity and low birthweight are associated with the
disorder, as is enteral feeding. The successful use of infection control measures to contain outbreaks, and the
association of epidemic cases with specific organisms, indicate the importance of bacteria in the pathogenesis of
necrotising enterocolitis, but no consistent association with specific organisms has been identified. Reports of
prevention of necrotizing enterocolitis by use of oral non-absorbed antibiotics,2–4 and by gastric and food acidification,5
support the importance of luminal bacteria in its pathogenesis. Prevention by oral antibiotics has not been widely
adopted, probably because of concerns about the development of resistant bacteria in nursery populations. Gut
immaturity6 and feeding practices7 are also elements that can contribute to the disorder. Prenatal maternal steroid
administration reduces the frequency of necrotising enterocolitis,6 possibly via the induction of mucosal maturation.
Lucas and Cole8 confirmed a part protective effect of breast milk and of delaying the introduction of formula feeds.
Eibl and colleagues9 reported the effects of giving preterm infants a mixture of human IgA and IgG enterally in the first
weeks after birth. Their trial was stimulated by a previous report that oral human IgG, given to prevent or modify
rotavirus infection in neonates, might have prevented the disorder. 10 Six of 91 control infants but none of the treated
infants in Eibl and co-worker’s9 trial developed necrotizing enterocolitis. Those treated received 600 mg/kg daily of
an immunoglobulin mixture containing 73% IgA and 26% IgG. Treatment in the trial was randomised but not blinded.
Fast and Rosegger3 compared the same dose of IgA/IgG mixture with oral gentamicin in a study of the same design.
Again there were only six cases of definite necrotising enterocolitis, but this time five were in the IgA group, leaving
the status of treatment unresolved. Intravenous administration of IgG does not seem to protect against necrotising
enterocolitis.11 We planned to do a three way comparison of oral IgA mixture, oral IgG, and placebo. However, the IgA
group was not practical because this preparation is not available in Australia, and the registered immunoglobulin
production process could not be modified to obtain IgA without great expense and delay. Therefore our aim became to
ascertain whether or not enteral administration of human IgG could prevent development of necrotising enterocolitis in
preterm infants.

We enrolled preterm infants (birthweight 1500 g), from 11 Australian nurseries, between Sept 9, 1991, and Dec 31,
1994. All newborn babies were about to begin enteral feeding. Infants who had been enterally fed for more than 24 h
were ineligible; this group mainly consisted of infants transferred to tertiary care units from regional centres. The ethics
committees of all the involved hospitals approved the study, and written informed consent was obtained from all

Study protocol
1200 mg/kg daily of IgG, given from the start of enteral feeding, was agreed upon as a pragmatic therapeutic
intervention for testing. We randomly assigned infants to receive a solution that was either 16% IgG or 30%
albumin (placebo), at least four times daily for up to 28 days. We graded intake up to 8 mL/kg daily (1200 mg/kg IgG)
as quickly as possible. The higher concentration of albumin in the placebo solution was used to simulate the viscosity
of the IgG solution, but later in the trial we reduced the concentration of the albumin solution to 15% because of
concern that the protein load was excessive (after 946 recruits). The protein content of both solutions increased the
protein intake of the infants, especially in larger infants, who were growing rapidly and ingesting at least 150 mL/kg
daily of solution. To compensate for this extra intake of protein, when these infants reached a weight of more than 1 kg,
were growing rapidly, and were receiving full feeds and doses of trial solution, we chose a formula with lower protein
content (Wyeth S26 Low Birthweight, Wyeth Nutritionals, PA, USA). One of us (CS) prepared the solutions, neither of
which contained preservatives. The IgG solution was coloured with approved food colourings to match the darker
colour of the albumin solution. The colourings, with the dose received at 8 mL/kg daily, and the proportion of
acceptable daily intake per kg of bodyweight were as follows: Food Red 16185, 0·036 mg/kg, 4·8%; Food Yellow
47005, 0·152 mg/kg, 1·5%; and Food Blue 42090, 0·09 mg/kg, 0·07% (CSL, Parkville, Australia). Acceptable daily
intakes are based on data for older children and adults. The trial solutions were packaged in identical glass snap-top
vials. After opening, vials were kept refrigerated and covered with a sterile cap. Opened vials were discarded after 24 h.
We designed the trial to have 90% power to detect a 50% reduction in the number of infants with definite necrotising
enterocolitis, with a two-sided 5% level test. On the basis of experience with necrotising enterocolitis in the
participating nurseries for the years 1985–87, we assumed an event rate of 6–7%, requiring a trial population of 1500.
We set up an independent safety and data monitoring committee (SDMC) to do interim analyses and to advise on
adverse events. We planned to do four interim analyses on the primary outcome of definite necrotising entercolitis by
allocated treatment after 20, 30, 50, and 75 diagnoses of the disorder. A p value less than 0·003 was used as a criterion
for early stopping of the trial.12 In February, 1992, we notified the SDMC of several diagnoses of Heinz body
haemolytic anaemia from one centre in the trial, and asked them to assess whether this anaemia was related to
treatment. The method of analysis chosen (selected in advance) to assess the data was a stratified 22 exact test 13 because
of the small number of patients diagnosed from each hospital. The analysis was not adjusted for multiple interim
analyses of the trial and interpretation of the p values took this factor into account. Allocation envelopes were produced
from a list of random numbers with five IgG and placebo patients in each block of ten envelopes. When an infant was
entered into the trial, the hospital pharmacists provided ampoules of trial solution identified only by name.
We entered study data onto a form attached to the clinical record of every infant. We then transferred the data into
computer files in Brisbane. 200 computer records were selected randomly and checked for accuracy, and a low error
rate of 4·5% was recorded; no errors that affected primary outcomes were identified. The same form was used for
infants whose parents declined to allow them into the trial and for those infants in the weight group that were not, for
various reasons, randomly assigned. The frequency of necrotising entercolitis in this group of nontrial infants was also
recorded. All necrotizing entercolitis cases in the 28 days after randomisation were reported. We reviewed by mail
infants transferred away from tertiary care before completing 28 days of treatment, to ascertain whether necrotising
entercolitis developed in this period. Definite necrotising entercolitis was defined as necrotising entercolitis diagnosed
at surgery or by necropsy, or on radiological or clinical grounds. The radiological diagnostic criteria were the presence,
in an infant with a clinical history consistent with necrotizing entercolitis, of pneumatosis intestinalis or portal vein
gas, or the presence of a fixed dilated loop of bowel on serial examinations. The radiographs of all infants diagnosed
with the disorder were reviewed independently by two radiologists (JR and one other), who were unaware of the
treatment group or identity of the infants. Films of infants in which there was diagnostic discordance between the two
radiologists were reviewed jointly and a consensus reached. The clinical criteria for necrotising entercolitis without
radiological or necropsy confirmation were a history consistent with the disorder and the presence of a palpable
abdominal mass associated with overlying abdominal wall cellulitis. Infants classified as having suspected necrotising
entercolitis had a clinical history and presentation suggesting the disorder, but no confirmatory radiological, surgical, or
necropsy results. Criteria for diagnosis of definite necrotising entercolitis, apart from in the exclusively clinically
diagnosed group, satisfied Bell’s criteria.14

Over 3 years and 2 months, we enrolled 768 infants to the IgG group and 761 to placebo (figure). Table 1 shows
birthweight, gestational age, days treated, and amount of trial solution given in the two groups. Closely similar numbers
of infants in the IgG and placebo groups had definite necrotising enterocolitis during the 28 day trial period. More
infants died in the IgG group than in the placebo group (table 2). Two infants had two episodes of definite necrotising
entercolitis, occurring 23 and 26 days apart. One episode in each was in the trial period. Although 22 infants with
definite necrotising entercolitis had a palpable abdominal mass with overlying cellulitis, in only one was this sign the
only available diagnostic criterion for definite necrotising entercolitis. 25 infants on IgG and 36 on placebo had
suspected necrotizing entercolitis (p=0·14). 13 infants developed the disease after completion of trial treatment or after
transfer to another nursery; three died. Table 3 shows some clinical and treatment variables in children with necrotising
enterocolitis in the IgG and placebo groups. The time to definite disease, the components of trial solution given, and the
timing of the treatment are much the same between groups. Two infants from each group developed the disorder soon
after cessation of treatment. 20 of the 84 infants with definite necrotising entercolitis arising during the trial had
received no trial solution; nine were in the IgG group and 11 were in the placebo group. Reasons included parents
withdrawing consent, clinical condition of the infants precluding enteral feeds, and mistakes in following the protocol.
97 (6%) randomly assigned infants developed definite necrotising entercolitis, of whom 19 (1%) died. Of the 542
infants not allowed to take part in the study, 39 (7%) developed the disorder; eight (1·5%) died. Table 4 shows the
results of logistic regression analysis of some possible risk factors by treatment group. Gestational age was a
significantly predictive variable for development of necrotising entercolitis, whereas APGAR scores were not.
Although infants in the IgG group who developed necrotising entercolitis had more days of mechanical ventilation than
infants with necrotising entercolitis in the placebo group, 5 or more days of mechanical ventilation did not predict
subsequent necrotising entercolitis when added to the model. Expressed breast milk (EBM) consumption was

Table 1: Comparison of variables between groups
Table 2: Distribution of all patients with necrotizing enterocolitis
Table 3: Details of patients with necrotising enterocolitis
added to the analysis by classification of trial infants as having received EBM on 5 days or more during the trial
or as those given it on fewer days. Fewer days on EBM significantly predicted necrotising entercolitis (odds ratio 52·4,
95% CI 1·40–3·95, p=0·001). However, none of the 104 trial infants not given EBM developed necrotising entercolitis.
The mean birthweight and gestational age of the 104 infants not fed EBM was a little higher than that of the others
(mean 1169 [SD 251] g vs 1077 [247] g, p=0·0006). When the 104 infants given no EBM were removed from the
analysis, the odds ratio associated with fewer days of EBM was 3·20 (95% CI 1·64–6·64, p=0·001). There was no
difference between the treatment groups with respect to EBM feedings, either in amount (total or average) or duration
(p=0·44 for all comparisons). Table 5 shows the distribution of recruitment and the number of infants with necrotising
entercolitis. The initial cluster of Heinz body haemolysis (Heinz bodies present in 5% of erythrocytes) in one nursery,
and all subsequent cases, were reported by the SDMC to the hospital ethics committees of Royal Women’s Hospital,
Brisbane. After four interim analyses, there was a clear excess in the IgG group (13 vs 4 in placebo group; p=0·005).
The proportion of infants given blood transfusions was similar in the IgG and albumin groups (62% vs 70%) as was the
mean number of transfusions given (2·6 vs 2·9).

Supplementation of enteral feeds with IgG did not reduce necrotising enterocolitis. The frequency of necrotising
entercolitis in infants in the same weight group who did not take part in the trial was similar to that in the pooled
trial groups. Was sufficient IgG used in this pragmatic trial? The amount and the volume was judged the maximum
acceptable addition to the intake of infants in the first weeks of life. In practice, the average amount of IgG solution
given to infants diagnosed with necrotizing entercolitis during the 3 days before diagnosis of the disease was 6 mL/kg
daily. This amount was somewhat less than intended, and was the result of irregular increases of dose with increasing
weight and of the many changes in oral feeding brought about by the clinical condition of the very-low-birthweight
infants. When oral intake was stopped for clinical reasons, clinicians were generally reluctant to continue treatment
with oral IgG because of the large volumes involved. Ironically, food intake is generally reduced because of the fear
that necrotising entercolitis might develop, yet preventive treatment is also halted. Eibl and colleagues 9 reported
protection against necrotising entercolitis from enterally administered immunoglobulin. We have not confirmed this
result. Some differences between the trials should be considered. Eibl and co-workers used a mixture of IgA (70%) and
IgG at 600 mg/kg daily, whereas we used IgG alone. IgA from blood might be more stable in the bowel lumen than
IgG, which was one of the reasons we increased the dose of IgG to 1200 mg/kg daily. Blum and colleagues’ 15 results
showed that functioning IgG survived passing through the gut of infants given commercial human serum IgG by
mouth. 4–12% of the ingested IgG was recovered in stools. All their patients, who received 152–1120 mg of
IgG in 24 h, had IgG function detected in the stool. Eibl et al 9 measured antibody concentrations in the stools of
their trial infants and identified immunologically active IgA and IgG. Eibl and colleagues9 began treatment of their
infants from day one, whereas we did so with enteral feeding. Less than 10% of necrotising entercolitis cases arise in
infants who have not been enterally fed.1 During this trial only three infants who were considered for enrolment
developed definite necrotising entercolitis before feeding began, and the mean period between entering the trial and
diagnosis of necrotising entercolitis was 14 days in the IgG group. The delay in beginning treatment was probably
unimportant. However, early presence of immunoglobulin could have interfered with enteral colonisation in some way.
There were no obvious differences in the pattern or intensity of colonisation by anaerobic or aerobic species in seven
pairs of twins who we studied (data not shown). None of Eibl and colleagues’ 9 infants received breast milk, whereas in
our study over 90% of the infants were given some. Breast milk has some antibacterial properties

Table 4: Risk factors for definite necrotising enterocolitis
Table 5: Definite necrotising enterocolitis and recruitment by hospital

and is thought to play a part in prevention of necrotizing entercolitis. 8 As mentioned, results of trials of prophylactic
oral aminoglycoside antibiotics suggest that this strategy reduces the frequency of necrotising entercolitis. 16 In view of
these results, oral immunoglobulin might have been expected to have some effect, dependent on the amount and
specificities of the antibodies present. Our failure to show any overall effect on necrotising entercolitis, by comparison
with the success reported by Eibl et al,9 could indicate a wider range of organisms causing necrotizing entercolitis in our
nurseries, which were spread around a large country. The infants in Eibl and colleagues’ trial9 were all in one nursery,
and protection could have been the result of antibody activity directed against fewer flora. Indeed, the IgA antibody
they used, although obtained from blood, might have contained a more suitable range of enteral specificities than the
IgG we used. Our trial was much larger than Eibl’s, and was blinded, giving weight to the recorded lack of effect with
oral IgG. We have no explanation for the excess number of infants diagnosed with Heinz body haemolysis in the IgG
group. A toxicologist who we consulted did not believe the colouring materials were a likely source of oxidative stress.
Clinically, many of the infants with clinical Heinz body haemolysis developed the disorder about 2 weeks after
commencing the trial, had high levels of Heinz bodies (>50%), and needed a single transfusion. Heinzbody haemolysis
was not judged a major adverse effect in that it was self-limiting. There was no excess of transfusions in the IgG group,
in respect of the number of babies transfused or in the number of transfusions given, to suggest a serious general
adverse effect. We informed the ethics committees about the haemolytic episodes and transfusion amounts after we
identified the first cases, and modified information sheets for parents accordingly. Once Heinz body haemolysis was
being actively sought, cases were found in the clinical records that would normally have gone undetected, which
suggests that this type of haemolysis is probably more common in nurseries than is recognised.
In our nursery population, necrotising entercolitis caused about a third of deaths recorded after the start of enteral
feeding, emphasising the importance of necrotizing entercolitis in neonatal practice. Most deaths in very-
lowbirthweight infants occur in the first week or so of life. Once enteral feeding begins, necrotising entercolitis and
its prevention become important issues. There is no evidence from our trial that enteral IgG, given as
described, offers any protection against necrotizing enterocolitis.


Lingjuan Ma Lingjuan Ma