As expressed mother's milk (MM) is known to be colonized by microbial species, it is occasionally considered as a source of infection in premature infants, prompting some clinicians to obtain milk bacterial culture results before infant feeding. To determine whether serial microbial cultures of MM predict infection in premature infants.
Milk microbial flora was determined by plate counts from aliquots of MM obtained from 161 mothers of infants born <30 weeks gestation (n=209). Pathogens isolated from the same infant were tabulated.
Milk samples (n=813) yielded 1963 isolates. There were no relationships between microbial counts and maternal age, ethnicity, education, skin-to-skin contact and infant infection. In 64 infants, milk and pathological isolates had presumptively the same Gram-positive organism, yet the odds of infection before or after exposure to milk containing that Gram-positive organism were not significant (1.18; 95% confidence interval=0.51, 2.76). In eight infants, milk and pathological isolates had presumptively the same Gram-negative organism, which appeared sporadically in milk, either before or after isolation in the infant.
Results of initial milk cultures do not predict subsequent culture results. Random milk cultures, even if obtained at any time during hospitalization, are not predictive of infection in premature infants. The sporadic nature of the appearance of certain isolates, however, suggests common exposure of both mother and infant. Routine milk cultures do not provide sufficient data to be useful in clinical management.
As the benefits increasingly have been recognized, more mothers are providing their breastmilk for their extremely premature infants in the NICU (neonatal intensive care unit).1 In the United States, mothers use electric breast pumps to facilitate milk expression and collection.2 Protocols for the collection, storage and feeding of expressed human milk have been reported.2 Because expressed milk comes in contact with foreign surfaces it may contain common skin bacterial flora and microbial species introduced via the collection apparatus.3, 4 Although case reports suggest that milk may be a source of infection, fewer infections and infection-related events are reported in extremely premature infants if fed their mothers’ milk (MM).5, 6, 7, 8, 9, 10 Nevertheless, some clinicians order bacterial cultures of MM and await the results of the screening before allowing the milk to be fed.11 We determined serial microbial cultures of expressed MM. We hypothesized that the presence of specific microbial flora in expressed MM was correlated with subsequent infection in the infant.
Mothers who delivered premature infants <30 weeks gestation and who intended to provide expressed milk using an electric breast pump were enrolled. Milk collection, transport and storage protocols were similar for all mothers, as published previously.2 Aliquots of fresh milk (expressed within 24 h) were collected weekly and evaluated for microbial colony counts and presumptive identification of microbial species. All pathogens isolated from the infant's blood, cerebrospinal fluid or urine were identified and tabulated. The study was approved by the Baylor Institutional Review Board for Human Subject Research. Informed written consent was obtained from parents before enrollment.
The milk sample was diluted 10-fold and inoculated onto Columbia CNA agar (Fisher Scientific, Pittsburgh, PA, USA) for enumeration of Gram-positive bacteria and MacConkey agar (Fisher Scientific) for enumeration of Gram-negative bacteria and incubated at 35 °C for 24 to 48 h. Based on standard microbiology quality-control procedures, plates containing between 30 and 300 colonies were selected for final estimation of the number of colony forming units (counts per ml). The laboratory staff was unaware of the clinical status of the infant and prior milk culture results from the same mother. Identification of microbial species was performed by standard methods of the American Society of Microbiology.12 Staphylococcus species and Streptococcus species were identified by the catalase reaction. Gram-negative bacteria were identified using the Vitek system (BioMerieux, Hazelwood, MO, USA).
Milk cultures were grouped as initial milk culture (similar to a culture obtained by some clinicians before feeding the infant) and subsequent milk cultures (any cultures after the initial culture). Whether an initial milk culture predicted the results of subsequent cultures was determined by the measure of agreement between initial and subsequent cultures (κ statistic), and the computation of the positive predictive value and the negative predictive value. The relative risk estimate was calculated to determine whether an initial milk culture predicted that the same organism appeared as a pathogen in the infant. Conversely, if a pathogen was isolated from an infant, the odds ratio that the same organism be present in the milk was tabulated.
Milk samples were obtained from 161 mothers (age 28.3±7.1 years, mean±s.d.) of 209 infants (birth weight 974±251 g, gestational age 27.2±1.6 weeks). Forty-seven percent of mothers were Caucasian, 31% Black and 22% Hispanic. Seventy-five percent of mothers had graduated college, and a yearly household income of >$100 000 was reported for 11% of the population. Skin-to-skin contact was practiced by 77% of the mothers.
There were 813 milk samples cultured for microbial species (bacteria and/or yeast) and 1963 microbial isolates were reported. The distribution of isolates was 60% Gram positive, 39% Gram negative and 1% yeast. Thirty-five (4.3%) milk samples had no microbial species isolated (Table 1). Most cultures had low colony counts of Gram-positive organisms (Table 1). There were no relationships between milk colony counts and maternal age, ethnicity/race, education, economic status and skin-to-skin care. There were no significant relationships between total or specific microbial colony counts and feeding tolerance, necrotizing enterocolitis, surgery for necrotizing enterocolitis, duration of antibiotic use, time to full feeding and hospital duration.
The most common microbial isolate was Staphylococcus epidermidis (Table 2). When initial and subsequent weeks’ cultures were compared, there were no differences in the pattern of colonization and no relationship was observed between initial and subsequent cultures, the κ statistic being non-significant for any isolate. For example, the κ statistic for S. epidermidis was −0.04 and Enterococcus faecalis was 0.32. There was a large variation in the positive predictive value, the probability of isolating the same organism in initial and subsequent milk cultures. In general, initial milk cultures did not predict the results of subsequent milk cultures (Figure 1). The negative predictive value, the probability of not finding an organism in subsequent milk cultures if it is absent in the initial milk culture, indicates that only a few, rarely isolated organisms, have a high NPV (Figure 1).
The overall proportion of isolates from milk and from the infant differed (Table 3). The odds ratio was used to determine whether infection in the infant was associated with milk exposure to the same organism at any time during hospitalization. The odds of developing infection before or after exposure to milk containing that organism were not significant for most Gram-negative organisms (Klebsiella, Enterobacter, Escherichia coli). However, the odds of infection before or after exposure to milk containing Serratia marcescens were significant. However, Serratia accounted for infections in only two infants and when evaluated before infection, the initial cultures did not identify this organism. Of ∼34 Gram-negative bacterial infections, there were only eight instances where the same organism was isolated from the milk and from the same infant. The eight Gram-negative organisms sporadically appeared in milk, either before or after the infection, and usually in combination with other Gram-negative organisms.
Of ∼109 Gram-positive bacterial infections, there were only 64 instances where milk and pathological isolates had presumptively the same organism. The odds of infection before or after exposure to milk containing that organism were not significant for any Gram-positive organism. Overall, despite the large number of milk cultures, few organisms were associated with pathogens in the infant. Most associations were sporadic and based on a very small numbers of isolates.
We investigated whether an initial milk culture could predict the risk of infection in the infant with the same organism as found in the milk. The relative risk of infection was negligible with the exception of E. coli, which was statistically significant if the initial culture was positive for that organism (Table 4). However, there were only two cases where milk and pathogenic isolates were similar for that organism. In one case, milk containing that organism was found as a subsequent culture only on day 14 and not on day 10, 20 or 28, but the infant had the organism isolated as a pathogen on day 67. In the second case, the organism was present on day 7 and not on 7 subsequent cultures. The organism was isolated from the infant on day 11.
Similar to other investigations, we found that nearly all expressed milk samples had microbial colonization.3 It has been reported that maternal hygiene affects milk colonization rates.11 The method of milk expression and collection also may affect colonization patterns. Mechanical pumping tends to be associated with greater contamination rates than manual expression.13 In this study, all milk was collected via protocol using mechanical expression at home and none of the usual demographic variables was useful in predicting bacterial colony counts.
Contrary to common beliefs, we found that initial milk cultures did not predict the flora in subsequent milk cultures. For example, Group B Streptococcus was isolated in five milk samples but only in one initial culture. Thus, despite the potential for unusual pathogens to be transmitted to the premature infant via breastmilk, a single early culture would not be helpful in predicting such an exposure. Indeed, these data as well as those of others question the continued practice of obtaining routine milk cultures.14
Overall, these data suggest that routine milk cultures are not useful in identifying potential infection issues. We found that the distribution of microbial species differed between milk and pathological (blood, cerebrospinal fluid and urine) isolates. For example, Coagulase-negative Staphylococcus, E. coli and Klebsiella were common isolates from the infant but uncommon isolates from milk samples, while Acinetobacter and Stenotrophomonas were frequent isolates from milk but never isolated from study infants. Other investigators have commented that although pathogen exposure in milk is concerning, few infants are affected adversely.4, 15 The antimicrobial properties of the milk most likely accounts for this protection.16
There are reports of recurrent infections in premature infants associated with isolation of the same organism in the milk.7, 9, 17 Our data suggest that simultaneous occurrence of positive cultures in both milk and infant may signify common exposure of infant and mother to that organism. It is difficult to presume only unidirectional transmission from mother to infant. However, our data suggest that such associations are unusual and are not predicted in advance.
Thus, our data do not suggest that random milk cultures, obtained early in the postpartum period or at any time during hospitalization, are predictive of infections in extremely premature infants. Uncommon pathogenic organisms appear sporadically in milk; the nature of their isolation in milk and in the infant suggests common exposure of both mother and infant. Routine milk cultures do not provide sufficient data to be useful in clinical management.
We thank Pat Wilson MT (ASCP) for the microbiology laboratory work. We appreciate the assistance of Dr Martin Lesser of the Feinstein Institute for Medical Research of the North Shore-Long Island Jewish Health System. This study was supported by the National Institute of Child Health and Human Development, Grant No. RO-1-HD-28140 and the National Institutes of Health General Clinical Research Center, Baylor College of Medicine, Grant No. MO-1-RR-00188. The contents of this publication are solely the responsibility of the authors and do not necessarily represent the official views of the National Institute of Child Health and Human Development or the National Institutes of Health.