An exclusive human milk-based diet in extremely premature infants reduces the probability of remaining on total parenteral nutrition: a reanalysis of the data

Abstract

Background: We have previously shown that an exclusively human milk-based diet is beneficial for extremely premature infants who are at risk for necrotizing enterocolitis (NEC). However, no significant difference in the other primary study endpoint, the length of time on total parenteral nutrition (TPN), was found. The current analysis re-evaluates these data from a different statistical perspective considering the probability or likelihood of needing TPN on any given day rather than the number of days on TPN. This study consisted of 207 premature infants randomized into three groups: one group receiving a control diet of human milk, formula and bovine-based fortifier ("control diet"), and the other two groups receiving only human milk and human milk-based fortifier starting at different times in the enteral feeding process (at feeding volumes of 40 or 100 mL/kg/day; "HM40" and "HM100", respectively). The counting process Cox proportional hazards survival model was used to determine the likelihood of needing TPN in each group.

Results: The two groups on the completely human-based diet had an 11-14 % reduction in the likelihood of needing nutrition via TPN when compared to infants on the control diet (p = 0.0001 and p = 0.001, respectively for the HM40 and HM100 groups, respectively). This was even more pronounced if the initial period of TPN was excluded (p < 0.0001 for both the HM40 and HM100 groups).

Conclusions: A completely human milk-based diet significantly reduces the likelihood of TPN use for extremely premature infants when compared to a diet including cow-based products. This likelihood may be reduced even further when the human milk fortifier is initiated earlier in the feeding process.

Trial registration: This study was registered at http://www.clinicaltrials.gov reg. # NCT00506584.

Figures

Figure 1

An example of how event data was compiled using counting process theory. The graph demonstrates the event times of a single subject; each subsequent event adds to the number of days that the subject has been on TPN. The Cox PH model treats each interval as an independent observation, thus allowing for the occurrence of multiple events.