We conducted a prospective quasiexperimental pre–post intervention study. The intervention consisted in the set of measures described in the nutrition protocol for VLBW infants in our unit, which we present in Table 1. Our definition of high risk for NEC was not limited to low weight, but also included asphyxia and diagnosis by the obstetrician of intrauterine growth restriction (IUGR) with umbilical artery and/or foetal middle cerebral artery Doppler abnormalities (whether or not postnatal weight was below the 10th percentile). The protocol was introduced in the unit in April 2012, although some of its measures had already been implemented in the preceding months, which is why these months were not included in the study. In the pre-intervention period analysed in the study, there was no written protocol on how to administer enteral nutrition.

There were two additional changes in the management of patients with haemodynamically significant patent ductus arteriosus. Thus, during the first period, omeprazole was usually added to treatment with ibuprofen, and enteral feeds were maintained based on digestive tolerance. In comparison, during the second period enteral feeding was withdrawn and omeprazole no longer given. There were no other significant changes in care between the two periods.

Enfamil Fortifier® was the fortifier used during both periods. It was added when the mother brought in her breast milk, and the milk was then separated into bottles in the volume to be administered in each feeding and stored under refrigeration until used. Its administration was initiated at a dose of 1g% when the infant reached an intake of 100mL/kg/day, and was increased by 1g each day until reaching 4g/100mL.

Group I (pre-intervention): born in year 2011.

Group II (post-intervention): born between May 1, 2012 and April 30, 2013.

We included all newborns with birth weights of less than 1500g that received care in our unit, including those born in other facilities that were transferred within one week from birth. We excluded newborns that had been referred exclusively to undergo a specific diagnostic test or treatment and subsequently transferred back to the original facility.

Primary outcome: incidence of NEC (Bell stage II or higher).

Secondary outcomes: focal intestinal perforation, overall mortality and mortality attributed to NEC (cases in which NEC was included in the sequence of events that led to death), nosocomial sepsis, weight at 28 days and 36 weeks of corrected age, percentage of newborns with a weight below the 10th percentile at the time of hospital discharge, and length of stay in days (excluding newborns that died or were transferred).

The definitions of all the variables were consistent with the definitions used in the SEN 1500 database.6 Furthermore, for the primary outcome of the study, NEC, and for patients diagnosed with focal intestinal perforation, the clinical history and imaging tests for all cases were reviewed by peers.

For quantitative variables, we first tested the normality of the distribution by means of the Kolmogorov–Smirnov test. Based on its results, we used a parametric test (Student's t) for variables with normal distributions, and otherwise a nonparametric test (Mann–Whitney U).

We summarised length of stay as median and interquartile range to minimise the influence of outliers.

For qualitative variables, we compared percentages for independent samples by means of the chi square test. When the proportion of any result was less than 5%, we used Fisher's test. We defined statistical significance as an alpha level of less than 0.05.

We performed the analysis with the EPIDAT statistical software (version 4.1).

Early introduction of enteral nutrition seems to be beneficial.14 However, the timing of its introduction does not seem to influence the incidence of NEC, at least in preterm infants that are haemodynamically stable. Based on our protocol, feeding was usually initiated in the first or second day, but was not introduced if the patient showed signs of haemodynamic instability. In this regard, we ought to note that almost all trials that compared enteral fasting with initiation of enteral feeding (even if it consisted of trophic feeding) specifically excluded haemodynamically unstable infants, so it is not known whether it is or is not safe in these patients. It is likely that in our pre-intervention cohort, trophic feeding was initiated in some infants that were still haemodynamically unstable, a practice that we now consider inappropriate. We believe that under these circumstances, there is blood flow redistribution at the skin level (evinced by poor peripheral perfusion) but also in other regions, such as the bowel (which would promote the development of NEC).

The administration of small volumes of milk (≤20–24mL/kg), without advances or with advances that remain within this limit for a period of five to seven days, is as safe as maintaining preterm newborns under enteral fasting for this duration, and shortens the time elapsed to full enteral feeding.15 On the other hand, daily advances from the beginning or advancing too quickly could foster the development of NEC.16,17 Before the introduction of the protocol, enteral feeds were advanced from the beginning or soon after introduction based on feeding tolerance in order to discontinue parenteral nutrition early and reduce the risk of catheter-related sepsis. Conversely, in the post-intervention period, trophic feeds were maintained for a period of five to seven days. The pre-post intervention studies mentioned above consistently report a trophic feeding period, of variable but protracted duration, prior to advancement of enteral feedings.9,10

After a period of trophic feeding, increments of 30–35mL/kg/day do not increase the risk of NEC compared with volumes of 20mL/kg/day18 and shorten the time to establishment of full enteral feeding. In the pre-intervention period, we advanced volumes by 20mL/kg/day, while in the post-intervention period we maintained the 20mL volume in patients with birth weights of less than 1000g, but raised it to 30mL/kg/day in patients with greater weights, and this was not associated with an increased incidence of NEC. Another difference is that in the second period, the increases were made at two different times during the day, which made us re-evaluate feeding tolerance before each increase. It is possible that this approach allowed us to detect problems earlier than before.

Constipation has been proposed as a possible risk factor for NEC because it can promote bacterial overgrowth in paretic bowel loops. Pietz et al. published a large case series with a very low incidence of NEC (0.8%) and emphasised how important they found this factor; they used glycerine if bowel movements were absent for 24h.8 In our study, we gave patients normal saline enemas. The only change in this regard was that in the post-intervention period, the enemas were routinely prescribed if the patient did not have any bowel movements in a 24-h period, whereas their use was rather discretionary in the first period.

Measurement of gastric residual volume has been suggested as a means to monitor feeding tolerance, but its routine use is the subject of debate.19 This practice was only occasional in the pre-intervention period, while in the post-intervention period it was proposed as an assessment method during the trophic feeding or advancement periods, but its actual implementation was left to the judgement of the physician in charge. In general, residuals were not checked in the absence of abdominal distension. Our decision tree (Appendix A) was based on the algorithm published by Hanson et al.20 It is possible that suspicion of NEC before the patient develops pneumatosis intestinalis could prevent its progression to more advanced stages if enteral feeding is discontinued and the patient is given antibiotics, as is proposed for Bell I stage.

Based on the existing literature,21–23 this protocol encouraged the use of breast milk from the beginning of enteral feeding. Maternal colostrum was used preferentially, and if it was not available, banked donor milk. In the pre-intervention period, “breast milk or artificial formula” was usually prescribed, and this often led to initiation of enteral feeding with formula, so that bacterial colonisation of the intestine was probably not ideal. The availability of banked donor milk allowed initiation of enteral nutrition with fresh colostrum or donor milk in nearly all patients. We believe that the benefits derived from breast milk are greatly enhanced if it is used from the beginning; in fact, in the years preceding the study, the rate of nutrition with the infant's own mother's milk at the time of discharge had been increasing in our hospital, reaching 74% by 2011 (44.5% for exclusive feeding with breast milk), but this increase had not been associated with a reduction in the incidence of NEC.

A recent study on enteral feeding practices in preterm infants showed considerable variability between countries, hospitals and neonatologists working in the same hospital.4 It is possible that this variability accounts at least partly for the also considerable variability found in the NEC incidence in different hospitals.7 It is possible that the reduction of intracentre variability resulting from the implementation of a standardised protocol could have an impact in and of itself.8,9,24

In the first period, it was customary practice to add omeprazole for gastric protection from initiation of treatment with ibuprofen. The use of ranitidine had already been discontinued due to its association with an increased incidence of sepsis and death in VLBW infants.25 Since we did not observe any benefits associated with this practice, and studies were being published that pointed at similar risks associated with the use of omeprazole, we decided to restrict its use to infants with active gastrointestinal bleeding, discontinuing treatment as soon as bleeding resolved.

Supplementation with a fortifier is associated with an improvement in short-term growth in infants fed with breast milk; it has not been found any increase in the incidence of NEC.26 The fortifier was administered in a similar way during the two periods under study, so we exclude the possibility of it having influenced the outcomes.

The diagnostic criteria for NEC and focal intestinal perforation were the same during the entire period under study, and all cases were documented prospectively for submission to the SEN-1500 database. It is possible that the definition of NEC cases in the second period was more stringent than in the first period, and that some cases of focal intestinal perforation were diagnosed as NEC in the first period. To avoid this potential bias, we retrospectively reviewed the diagnosed cases of NEC and focal intestinal perforation, re-evaluating the imaging tests and the histological examinations in surgical cases, and included only those cases that were confirmed. The reduction in the incidence of cases with radiographic evidence of pneumatosis intestinalis (Bell II stage) and, concurrently, of forms of NEC that required surgery, confirmed the actual reduction in the incidence of disease.