Randomized trial of exclusive human milk versus preterm formula diets in extremely premature infants


Cristofalo, EA., et al. Randomized trial of exclusive human milk versus preterm formula diets in extremely premature infants. The Journal of Pediatrics. 2013; 163; 1592-5. PMID 23968744


Nicholas Embleton
Consultant Neonatal Paediatrician
Newcastle Hospitals NHS Foundation Trust

Jemma Cleminson
Specialty Doctor (Paediatrics)
University of York


Prevention and treatment


In a randomized evaluation:

  1. Do infants fed an exclusively human milk-based diet have better short-term health outcomes, designated as fewer days of parental nutrition and less morbidity, than infants fed only bovine milk-based diet?
  2. Do infants fed an exclusively human milk-based diet experience the same rates of growth as infants fed only bovine milk-based diet?


  • Design: Multi-centre trial total of 7 NICUs (6 units in USA and 1 in Austria).
  • Allocation:
    • Randomization was performed separately for each of the 7 study sites, though as no further details of the randomization process are provided , it is unclear whether stratified by site.
    • It is not clear at what point the randomization occurred (as soon as enrolled or when they were ready for enteral feeds)
  • Blinding: Triple blinded (investigators, patient care providers, and families) to group assignment.
    • Blinding was handled by independent hospital personnel not associated with the care of the study infant, and milk was prepared away from the bedside, typically used masked feeding syringes with a colored covering.  It is not clear how effective the masking was.
  • Follow-up period: Participation was ended when one of the following was reached:
    • 91 days of age
    • Discharge from hospital
    • Attainment of 50% oral feeding (i.e. 4 completed oral feeds/day)
  • Patients:
    • Inclusion criteria
      • Infants of birth weight 500 – 1250g whose mothers did not intend to provide milk or were unable to, and
        • who received parental nutrition within 48 hours after birth
        • received enteral feeds before 21 days of age (the protocol suggests similar feeding guidelines were in place for all the centres and recommended trophic feeds starting 1-4 days after birth).
      • Exclusion criteria:
        • Major congenital malformations
        • Transferred to a study site after 48 hours
        • Had a high likelihood of transfer to a non-study site during the study period
        • Were already participants in another study that affected nutritional management
    • Intervention:
      • Intervention commenced when the infant started enteral feeds
      • Randomised to receive either bovine milk-based preterm formula (BOV) or an exclusive fortified human milk diet (HUM) using a pasteurized donor human milk (DHM) product with a human milk based fortifier
      • The BOV diet consisted initially of 20Kcal/Oz preterm formula and subsequently 24Kcal/Oz strength formula. This is standard practice in many North American units.
      • The HUM diet consisted of pasteurized DHM (20Kcal/Oz), to which a pasteurized human milk fortifier derived exclusively from donor human milk-based was added, in order to meet the nutritional needs for human milk fortification. Fortifying with additional nutrients in this way is considered standard practice. It is not clear whether the fortifier was added to all HUM feeds at the start, or whether fortification occurred at a later stage, but whilst this may impact on growth it is not likely to affect the incidence of NEC.
      • Similar nutrition and feeding guidelines were followed across study sites (i.e. initiation and weaning of parenteral nutrition, rates of enteral feed volume increases etc). The goals for enteral nutrition were 150ml/kg/d and 120kcal/kg/day.
    • Outcomes:
      • Primary
        • Duration of TPN
      • Secondary
        • Growth (weight, length, HC)
        • Duration of hospital stay
        • Days of mechanical ventilation
        • Days of oxygen therapy
        • Incidence of late-onset sepsis
        • Incidence of necrotizing enterocolitis (Bell stage II or higher)
        • Incidence of retinopathy of prematurity
    • Analysis and Sample Size:
      • Intention to treat analysis
      • Sample size: The study size (n=26 per group) was determined based on an estimate that PN duration would be decreased by 50% in the HUM group, using available data suggesting a typical duration of 35 days (SD 22 days), with a 2-sided alpha error of 5% and power of 90%.
      • Kaplan-Meier estimates for distribution of PN days were compared using the log rank test
      • The Wilcoxon rank-sum test was used for uni-variate group comparisons
      • Chi-squared test was used for categorical data with p value determined by StatXact 7 software
      • Independent statistician AND by the study group. It is unclear how the study group interacted with the independent statistician i.e. which analyses were performed.
    • Patient follow up: For those patients that were transferred to another hospital before completion of the study, it was determined whether they went on to develop NEC or not.
    • Funding: E.C., R.S., C.B., S.S., R.T., U.K.-K., G.D., and S.A. received financial support from Prolacta Bioscience for the conduct of the study. D.R. and M.L. are employees of Prolacta Bioscience. A.L. is a paid Consultant of Prolacta Bioscience. This would imply that funders also conducted statistical analyses.


  • 53 infants were recruited (24 BOV, 29 HUM). There were differences in growth rates (only significant for length) with infants in the BOV group showing higher rates. There were significant differences in the primary outcome, duration of PN (BOV 36 days v HUM 27 days, p=0.04). There were also statistically significant differences in the rate of NEC (Bell’s stage 2 or greater BOV 21% v HUM 3%) and surgical NEC (BOV 17% v HUM 0%). The 6 cases of NEC occurred on average at 29 days of age (range, 16-55 days), and the average amount of enteral feeding up to the diagnosis of NEC was 1762 mL (range, 14-3462 mL). After controlling for race, receipt of antenatal steroids, Apgar score, and age at first enteral feeding, a multivariate regression identified only exclusive human milk diet as significantly associated with a decreased incidence of NEC (P < .01).
  • Characteristics of study infants:
    No. 24 29
    Birth weight, g 983 ± 207 996 ± 152
    Gestational age, wk 27.5 ± 2.4 27.7 ± 1.5
    Small-for-gestational age, n 2 (8%) 3 (10%)
    Male sex, n 11 (46%) 12 (41%)
    Apgar score at 5 minutes <6, n 3 (13%) 0
    Duration of hospital stay, d 82 (62, 124) 72 (64, 104)
    Mean enteral intake, mL/kg/d 82 ± 32 98 ± 29
    First feeding, d 6.5 (3, 8) 4.0 (2, 7)
    Time to full enteral feed, d 29.3 ± 14.7 24.6 ± 11.7
    Feeding intolerance events, n 1.2 ± 1.3 1.1 ± 1.3
    Weight gain, g/d 17 ± 7.1 15 ± 5.8
    Head growth, cm/wk 0.88 ± 0.18 0.78 ± 0.26
    Length gain, cm/wk 1.12 ± 0.28 0.84 ± 0.21
    • Values are mean±SD or median (25th, 75th percentiles).
    • ∗: P = .03; †: P = .006; ‡: Removing 1 infant with hydrocephalus.
  • Clinical outcomes of study infants:
    Parenteral nutrition, d 36 (28, 77) 27 (14, 39)
    Late-onset sepsis, n 5 (21%) 4 (14%)
    NEC, n 5 (21%) 1 (3%)
    NEC surgery, n 4 (17%) 0
    NEC and/or death, n 5 (21%) 1 (3%)
    Mechanical ventilation, d 24 (10, 75) 17 (2, 38)
    Oxygen therapy, d 28 (21, 61) 20 (5, 32)
    Retinopathy of prematurity, n 5 (21%) 8 (28%)
    Death, n 2 (8%) 0
    • Values are median (25th, 75th precentile).  Late-onset sepsis rates updated from original table following a published correction
    • ∗: P = .04; †: P = .08; ‡: P = .036.


The authors conclude that in extremely preterm infants given exclusive diets of formula (BOV) vs human milk (HUM) there was a significantly greater duration of PN and a higher rate of surgical NEC in infants receiving the BOV diet. The authors state “based on NEC surgery results, the number of infants needed to be fed an exclusive human milk diet to prevent 1 case of surgical NEC is 6 infants”. They conclude that an exclusive human milk diet to nourish extremely preterm infants is warranted.


Cristofalo et al. are to be congratulated for an important study. It is the only RCT in the last 30 years (apart from a 2010 study funded by the same company with a different study design1) to compare an exclusive human milk diet with sufficient nutrients to meet recommended intakes in preterm infants, with the more typical standard diet that necessitates the use of BOV products either as formula or fortifier. NEC and sepsis are important outcomes and responsible for more deaths after the first week of life, that any other common pathologies in preterm infants.2 Since NEC was reviewed in a blinded fashion by a panel of the study investigators, this certainly adds an element of rigour to the diagnosis of NEC as compared to many other studies. However it would be of interest to know further details on how this was conducted. If any of the study investigators were also members of the medical/nursing team then radiographs and clinical details might inadvertently unmask the cases.

An intervention that significantly decreases surgical NEC is of relevance to NICUs across the world. Whilst DHM is widely promoted, the current evidence of benefit comes largely from trials conducted in the 1980s when common neonatal practices (antenatal steroids, surfactant etc,) in use today were not commonplace. In addition, most of those trials did not provide sufficient nutrients to meet current recommendations i.e. they did not use a breast milk fortifier, and growth would be considered sub-optimal by most clinicians. New trials are currently being performed and results of those will help to further guide research and practice (Domino Study O’Connor, D at NCT02137473; The MILK study Colaizy, T at NCT01534481 ; PREMFOOD Mills, L. at NCT01686477 ).

The main outcome of this study was duration of parenteral nutrition, which is a reasonably objective, quantifiable surrogate of feeding tolerance and neonatal morbidity. There is limited evidence that use of a HUM diet reduces time to full feed tolerance, and the reduction in use of PN may reflect the decrease in NEC in the BOV group. Even if duration of PN was decreased [without any effect on NEC] this would still be an important outcome because of the risks of central line associated blood stream infection (CLABSI) and the economic costs of PN. There were no differences in rates of infection between the BOV and HUM groups (note that a correction of infection rates in table II was published in the same journal BOV 21% v HUM 14%, ns).

The study size was determined based on an estimate that PN duration would be decreased by 50% using available data suggesting a typical duration of 35 days (SD 22 days), with a 2-sided alpha error of 5% and power of 90%. The trial results actually show a 25% reduction from 36 days (BOV) to 27 days (HUM) (p=0.04). It seems unlikely that any intervention such as milk type in well managed and resourced NICUs will reduce key outcomes (especially those that have a multi-factorial origin like PN duration) by much more than 20%, but even this reduction would be important. If it was estimated that PN duration would be decreased by just 7 days (20%) then sample size estimates would increase from ~29 per group to ~169 per group i.e. the entire study population would need to be almost 5 times larger to power a study to determine effects on PN duration.

NEC is a multi-factorial disease and there is no gold standard diagnosis. As such, definitions are open to question in all neonatal trials, especially where a tissue diagnosis (obtained at surgery or post-mortem) is not available. In this study however, differences in NEC surgery were also significantly different (NEC surgery BOV 17% v HUM 0%). This result is unlikely to have been influenced by inadequate blinding or masking of the intervention (if indeed that occurred). Nevertheless, although this was a high risk population (median gestation 27.6 weeks gestation) background rates of surgical NEC in the ‘control’ group (BOV 17%) are higher than seen in many other NICUs.

In most health systems, key interventions in medical practice nowadays require a contemporaneous cost or health economic analysis. Arguably, this may be less important for outcomes such as surgical NEC or death, but most practitioners will still need to present a ‘business case’ for introducing new potentially costly interventions or innovations to their hospital management boards. It can be argued that economic analyses are complex and require additional resources and independent statisticians to analyse the data. However major new interventions merit such analyses. Although no costs are included in this study, it seems likely that the adoption of humanized milk products is associated with significant costs.

Overall, whilst this study is a well-conducted RCT the primary challenges for clinicians in interpreting this study are:

  1. Study power – the key outcome of interest is NEC. Using typical estimates of surgical NEC in infants <1250g of ~7.5%, and estimating that the intervention will half the rate of NEC, a study will need to recruit >1000 infants.
  2. Generalisability – surgical NEC rates in this study (17%) are much higher than in many other reports. In addition, the median age at starting feeds (~5 days) and time to full feeds (~27 days) are later/longer than seen in many NICUs especially in Europe.
  3. Health economic assessment – this is likely to be an expensive intervention so a full consideration of costs is likely to be required by hospital/health management organisations before introduction into routine practice.

Given the high rates of death and serious long-term morbidity as a consequence of NEC, interventions that aim to reduce its occurrence are extremely important. However, further studies will be required to better understand the mechanisms of actions of interventions, and larger studies will be needed to more precisely determine efficacy and cost-effectiveness.


  1. Sullivan S, Schanler RJ, Kim JH, et al. An exclusively human milk-based diet is associated with a lower rate of necrotizing enterocolitis than a diet of human milk and bovine milk-based products. J Pediatr 2010;156:562-7 e1.
  2. Berrington JE, Hearn RI, Bythell M, Wright C, Embleton ND. Deaths in preterm infants: changing pathology over 2 decades. J Pediatr 2012;160:49-53 e1.
  3. Abrams SA, Schanler RJ, Lee ML, Rechtman DJ. Greater mortality and morbidity in extremely preterm infants fed a diet containing cow milk protein products. Breastfeed Med 2014;9:281-5.
  4. Arslanoglu S, Corpeleijn W, Moro G, et al. Donor human milk for preterm infants: current evidence and research directions. J Pediatr Gastroenterol Nutr 2013;57:535-42.
  5. Klingenberg C, Embleton ND, Jacobs SE, O’Connell LA, Kuschel CA. Enteral feeding practices in very preterm infants: an international survey. Arch Dis Child Fetal Neonatal Ed 2012;97:F56-61.
  6. Quigley M, McGuire W. Formula versus donor breast milk for feeding preterm or low birth weight infants. Cochrane Database Syst Rev 2014;4:CD002971.
  7. Stoll BJ, Hansen NI, Bell EF, et al. Neonatal outcomes of extremely preterm infants from the NICHD Neonatal Research Network. Pediatrics 2010;126:443-56.
  8. Menon G, Williams TC. Human milk for preterm infants: why, what, when and how? Arch Dis Child Fetal Neonatal Ed 2013;98:F559-62.


  1. Thank you to Dr. Cristofalo and colleagues for performing this important study and to Drs. Embleton and Cleminson for a thoughtful review.

    Despite setting itself apart from older trials of donor human milk by providing an “appropriately fortified” human milk based diet, the point estimates for growth in this trial fall in line with those of older trials, summarized in a Cochrane review (Quigley& McGuire. 2014; CD002971). Collectively, these suggest worse short term growth for infants on human milk based diets. Could this negatively impact neurodevelopment? Insufficient data exist for this outcome. While the accumulating evidence shows less NEC with human milk based diets, how should one balance this benefit against an uncertain but possible developmental risk?

    Should NICUs be implementing policies of exclusive human milk based diets, or is the jury still out? What is the practice at your own NICU?

  2. I’d argue that the jury is still out.
    It is often reported that smaller studies report larger effects. The small sample size (n=53) and reported large effect (NNT ≈ 6) in this RCT could be related to “the “small study effect”

    Drs. Embleton and Cleminson comment on statistical power: to detect a 50% reduction in NEC under the asumption that the baseline risk is about 7.5% (5% sign-level, 80% power), about a trial including 1200 infants would be needed.

    We (and most units in Sweden) use almost entirely breast milk for feeds to preterm infants (and have an infrastructure for milk donation, milk banking, individual fortification etc). However, I believe that firm conclusions on NEC prevention needs to be based on studies including >53 infants.

Comments are closed.