Human Milk Intake and Retinopathy of Prematurity in Extremely Low Birth Weight Infants
OBJECTIVES. Our goal was to analyze the association between human milk intake and severe retinopathy of prematurity in extremely low birth weight infants.
PATIENTS AND METHODS. This study is a secondary analysis of data collected for a trial of glutamine supplementation in extremely low birth weight infants (birth weight <1000 g). Among the 1433 participants in that trial, data are available regarding human milk intake and the occurrence of severe retinopathy of prematurity (defined in this study as retinopathy of prematurity treated surgically) for 1057 infants. The volume of human milk intake was expressed as the mean volume (milliliters per kilogram per day) and the mean proportional volume (proportion of total nutritional intake) from birth to discharge or transfer. Using logistic regression, we estimated odds ratios and 95% confidence intervals for any human milk intake and, among infants who received human milk, for each 10 mL/kg per day and each 10% increase in volume.
RESULTS. Of the 1057 infants included in this cohort, 788 infants (75%) received at least some human milk. Among these milk-fed infants, the median volume of human milk intake was 30 mL/kg per day (interquartile range: 6–83 mL/kg per day), and the median proportional volume of human milk intake was 0.18 (interquartile range: 0.03–0.66). One hundred sixty-three infants (15%) developed severe retinopathy of prematurity.
CONCLUSIONS. In extremely low birth weight infants, human milk intake was not associated with a decreased risk of severe retinopathy of prematurity.
Retinopathy of prematurity (ROP) is a vascular disorder of the retina affecting infants born prematurely. The reported incidence of severe ROP (stage 3) in infants born <1000 g varies from 14% to 40%,1–3 representing 96% of all cases of stage 3 ROP. After cryotherapy4 or laser surgery,5 vision is severely impaired in 44% of children and 18% of eyes, respectively. In addition, myopia, strabismus, and amblyopia occur frequently.6,7 The pathogenesis of ROP is incompletely understood, but factors that have been implicated include exposure of the developing retina to abnormal oxygen levels8 and cytotoxic reactive oxygen species,9 the premature infants’ reduced antioxidant defenses,10,11 and decreased ability to synthesize long chain polyunsaturated fatty acids (LCPUFAs).12–14
Because it contains LCPUFAs15,16 and antioxidant enzymes, human milk (HM)17–19 might influence the development of ROP. The objective of this study was to analyze the association between HM intake and the development of severe ROP, defined in this study as ROP that was treated surgically. We hypothesized that infants fed HM would be less likely to develop severe ROP and that higher levels of HM intake would be associated with a lower incidence of severe ROP.
PATIENTS AND METHODS
The cohort for this observational study was derived from the 1433 infants enrolled in a multicenter, randomized clinical trial of parenteral glutamine supplementation, conducted by the National Institute of Child Health and Human Development (NICHD) Neonatal Research Network (NRN) from October 1999 to September 2001.20 The trial included infants born at 14 study centers with birth weights between 401 and 1000 g. Excluded from the trial were infants with major congenital anomalies, congenital nonbacterial infection, or hypoxia with bradycardia for >2 hours or blood pH <6.80, and those for whom full medical support was not provided. For our study, we also excluded 104 infants who were never enterally fed (Table 1).
All data, except those about HM intake, were obtained from the NRN Generic Database or the NRN 18–22 Month Follow up Study database, for which data are collected from medical charts by research nurses using prespecified definitions as described in previous publications.21,22 Data about HM intake were collected as part of the aforementioned clinical trial of parenteral glutamine.20 During that trial, the type and volume of nutritional intake was recorded for all subjects from the day of birth until the earliest of the following events: discharge, transfer, death, or the 120th day of life. Data were recorded daily until the infant reached full enteral feedings. Once an infant had been fully enterally fed for 72 hours, data collection was reduced to 3 days per week: Monday, Wednesday, and Friday (MWF). If enteral feedings were discontinued or parenteral supplementation was reintroduced, daily data collection was restarted, continuing until the infant had again achieved 72 hours of full enteral feedings. The decision to use MWF data collection was made to relieve the site from collecting largely redundant data on subjects whose basic feeding patterns did not change daily. The MWF feeding values and infant weight measurements were used to interpolate the data for the other days to construct a full stream of data for each infant. Thus, feeding values for Tuesday, Thursday, Saturday, and Sunday were constructed as a weighted average of the appropriate neighboring values. A comparison between the raw data and the “enhanced” data showed they were very highly correlated.
Data about the occurrence of ROP surgery were obtained from the NRN Generic database for infants who underwent ROP surgery before NICU discharge, transfer, or death and from the NRN 18–22 Month Follow up Study database or the Glutamine Study 30 Month Follow up database for those whose ROP surgery occurred after discharge or transfer. Data about ROP surgery in the NRN Generic database are based on a review of medical charts, whereas in the NRN Follow up Study databases, these data are based on parental report.
HM intake was defined in 3 ways. In analyses involving the entire cohort, HM intake was defined as a dichotomous variable (at least some HM or no HM feedings). In analyses involving only those infants who received HM, the magnitude of the intake was expressed in 2 ways: (1) the mean volume of HM intake, expressed as milliliters per kilogram per day, from birth to discharge or transfer, and as (2) the proportional volume of all recorded nutritional fluid intake (parenteral nutrition and enteral nutrition, but not other glucose-containing solutions) that consisted of HM, from birth through discharge or transfer. At the time this study was conducted, preterm formulas did not contain supplemental omega-3 LCPUFAs, and none of the infants received donor milk.
Because this study was an observational study using previously collected data, no efforts were made to influence the HM intake or any other facet of nutritional management of the infants in this study.
Surgical therapy for ROP was used as a surrogate for severe or threshold ROP because the data set available to us did not contain information about the number of “clock hours” of involvement, which is needed to classify ROP as “threshold,” “prethreshold,” or “less than prethreshold” ROP.23 The data set included information about whether surgery was performed for ROP for all infants through discharge or transfer (whichever occurred first), and whether ROP surgery occurred beyond this time, for those infants who returned for a follow-up clinic visit at 18 and/or 30 months’ adjusted age. Infants were classified as to whether they had developed ROP that was treated surgically. However, infants who did not undergo ROP surgery before discharge or transfer and did not return for follow-up at either 18 or 30 months’ adjusted age were classified as “lost to follow-up,” meaning that data about the outcome of interest (ROP treated surgically) were missing for these infants. In addition, data were not included in the analysis for infants who died before reaching an age at which they could be classified by ROP outcome.
Although this study did not determine the criteria for performing ROP surgery, these criteria are generally well accepted and based on those defined in the CRYO-ROP study.1
In the NRN Generic database, prenatal care is defined as ≥1 prenatal care visits, and antenatal steroids as maternal treatment with corticosteroids before delivery. Mothers are classified as having hypertension if ≥1 of the following is listed in the mother's chart: diagnosis of hypertension, systolic blood pressure >140 mmHg, or a diastolic blood pressure >90 mmHg. Maternal education was categorized as less than high school, high school, or greater than high school. For this study, early onset sepsis (suspected or confirmed) is defined as treatment with antibiotics for ≥5 days, beginning in the first 72 hours after birth, regardless of blood culture results. An infant was classified as having respiratory distress syndrome if he/she had both “clinical features of respiratory distress” and an “abnormal chest radiograph within 24 hours of birth.” For the current study, the fetal growth ratio-10 (FGR) was used as an indicator of fetal growth. This was determined by dividing the birth weight of the infant by the 10th percentile of birth weight for the appropriate gestational age and gender from growth charts developed by Alexander et al.24 The FGR is a more informative indicator of fetal growth than dichotomous measures, such as birth weight less than the 10th percentile.25 Although other investigators based the FGR on the gestational age–specific median birth weight, the reference values provided by Alexander et al24 do not include gestational age–and gender-specific medians but do contain these values for the 10th percentiles of birth weight. Therefore, we standardized the birth weights in this study to the 10th percentiles of these curves.
Our analytical approach was to test whether any HM intake was associated with severe ROP and then to look at whether the level of HM intake was associated with severe ROP among those subjects who received HM by using 2 continuous measures of HM intake. Unadjusted odds ratios (ORs) and 95% confidence intervals (CIs) were estimated for the association of HM intake (any versus none, mean volume and proportional volume) and severe ROP.
Selection of Potential Confounding Variables
Factors that were considered as potential confounding variables met the following criteria: (1) previous studies suggested that the factor was associated with either HM or ROP, (2) the factor was ascertained before the feedings were initiated, so it could not be involved in a pathway linking HM intake and ROP, and (3) data were available for the factor in the NRN Generic database. For factors that met these criteria, we analyzed associations between the factor and severe ROP and between the factor and HM. If the P values for both of these associations were ≤0.2, the factor was regarded as a potentially confounding variable and was included in the multivariate analyses, as described below.
Specification of Multivariate Regression Models
The potential confounding factors, identified in the manner described above, were classified as either prenatal or postnatal factors, according to the time of their occurrence (factors listed in Table 2). In the first step of multivariate analysis, HM and the prenatal factors that were identified as potential confounders were entered into a logistic model as independent variables with ROP (no ROP/ROP not requiring surgery versus ROP treated with surgery) as the outcome variable. The prenatal variables that were found to be independently associated with ROP outcome, at a significance level of P < .1, were retained in the model. In the second step, the postnatal variables that were identified as potential confounders were entered into the model. Then, those postnatal variables that were not found to be independently associated with ROP outcome at a significance level of P < .1 were eliminated from the model. In the final step of multivariate analysis, ORs and 95% CIs were estimated and adjusted for the retained potential confounding factors. For the analyses in which HM intake was expressed as the mean volume or the proportional volume, ORs were reported for each 10 mL/kg per day and each 10% increase in intake, respectively. The generalized estimating equation method was used to account for the inclusion of infants born of multiple gestations (twins and triplets).
Analysis of Possible Effect Modifiers
To explore whether certain factors influence the strength of association between HM and the severity of ROP (ie, effect modification) we estimated ORs for strata, which were defined in terms of 4 factors that previous research suggests might be effect modifiers: maternal educational level (a surrogate measure of socioeconomic status), ethnicity, FGR-10 (as a measure of the adequacy of fetal growth), and postmenstrual age at first feeding (as an indicator of the timing of initial exposure to HM).
Given the estimated SEs of the ORs from this study for the HM (any or none) and HM (mL/kg per day), the available sample sizes were sufficient to detect a reduction in the OR to ≤0.52 for HM (any or none) and to ≤0.92 for HM (mL/kg per day) with a power of 80% and a 2-tailed significance level of .05. The OR for HM (mL/kg per day) is the reduction in the OR per 10 mL/kg per day.
Of 1433 infants enrolled in the NRN trial of glutamine supplementation, 1329 infants were enterally fed. Of these, 272 had missing data for the outcome (126 died, 5 did not have an eye examination, and 141 did not come for follow-up at 18 or 30 months’ adjusted age). Thus, there were 1057 infants with data for both the exposure and outcome of interest (Table 1). The mean (SD) day of life and mean (SD) postmenstrual age at the time of discharge or transfer were 87.9 (25) days and 38.7 (2.9) weeks, respectively. Eighty-three (7%) infants were transferred to another hospital before discharge home at a mean (SD) day of life and mean (SD) postmenstrual age at transfer of 55.5 (26.3) days and 34.5 (3.1) weeks, respectively.
Description of ROP and HM Feeding
Of 1057 infants for whom complete data were available, 163 (15%) developed severe ROP. Seven hundred eighty-eight infants (75%) received HM. Among these infants, the median (interquartile range) of the volume of HM intake was 30 (6–83) mL/kg per day and the median (interquartile range) for the proportion of total nutritional fluids (enteral nutrition plus parenteral nutrition) consisting of HM was 0.18 (0.03–0.66).
The unadjusted OR (95% CI) for the association between any HM intake and severe ROP was 1.41 (0.94–2.13; P = .1). Among infants who received any HM, increasing volume of intake (mL/kg per day) was associated with a decreased unadjusted odds of developing severe ROP (OR [95% CI] per 10 mL/kg per day increase in intake: 0.95 [0.91–1.00]; P = .05). The unadjusted OR (95% CI) for developing severe ROP with increasing proportional volume of HM intake per 10% increase was 0.96 (0.91–1.02; P = .2).
Irrespective of whether HM intake was expressed as a dichotomy (any HM intake versus none) or as a continuous variable (mean volume of intake or mean proportional volume), the following factors were identified as potential confounders: antenatal steroids, maternal education, ethnicity, mode of delivery, gestational age, treatment for proven or suspected early onset sepsis, and pneumothorax (Table 2). In addition, the following factors were identified as potential confounders when HM intake was expressed as a continuous variable: occurrence of labor, maternal hypertension, outborn status, low 5-minute Apgar score, and multiple birth. Lastly, infant gender and receipt of surfactant were identified as potential confounders only when HM intake was express as mean volume. The group to which infants were randomized in the glutamine supplementation trial (from which the nutritional data for this study were obtained) did not influence either HM intake or ROP outcome. The diagnosis of patent ductus arteriosus or ultrasonographic evidence of severe intraventricular hemorrhage or white matter injury were not associated with HM intake and, therefore, were not included as confounders.
Although variables associated with chronic lung disease (eg, duration of mechanical ventilation and supplemental oxygen, and receipt of postnatal steroids) were associated with ROP and HM intake in this study, they were not included as confounders in the analysis because they are likely intermediates in the presumed pathway linking HM and the development of ROP.
ORs for HM, and for the variables which, in multivariate models, had a statistically significant association with severe ROP, are depicted in Figs 1 and 2. The adjusted ORs (95% CI) for developing severe ROP for each of the models containing the 3 HM exposure variables and the confounders listed above are as follows: any HM intake versus none: 1.47 (0.94–2.32), P = .09 (Fig 1); proportional volume of HM intake: 0.98 (0.91–1.05), P = .52 (Fig 2); and mean volume of HM intake (mL/kg per day) was 0.98 (0.92–1.04), P = .43 (not shown). The variables included in the model containing HM intake by mean volume were pneumothorax, ethnicity, maternal hypertension, day to first enteral feeding, antenatal steroids, gestational age at birth, and maternal education.
There was no evidence of a statistically significant dose response relationship between increasing HM intake and development surgical ROP or a threshold of HM intake above which the probability of developing severe ROP declined (Fig 3).
The strength of the association between HM and ROP was not influenced by maternal education, ethnicity, FGR-10, or postmenstrual age at first enteral feeding. In addition, analyses in which “study center” was included as a potential confounder did not produce significantly different results from those in which this variable was not included; therefore, this variable was not retained in the final regression models.
Our study was undertaken to investigate whether 1 or more constituents of HM are protective against severe ROP. Our hypothesis was based on the fact that HM contains immunomodulatory substances, such as secretory immunoglobulin A, lactoferrin, lysozyme, cytokines, oligosaccharides, antioxidant enzymes and cellular components.18,19,26 These factors are thought to influence immune defenses of the infant, which may explain the lower risk of necrotizing enterocolitis and sepsis among HM-fed infants.27–29 HM also contains docosahexaenoic acid (DHA), which has an important role in the developing brain and retina.16,30 Contrary to our hypotheses, in the sample of extremely low birth weight infants whom we studied, neither receipt of HM nor increasing intake of HM were associated with a decreased risk of developing severe ROP.
The point estimate for the OR for any HM (versus none) was >1, whereas the ORs for the 2 continuous HM variables were <1. The most likely explanations for this apparent paradox are bias attributable to unmeasured confounders and sampling bias because of the fact that the analysis of any HM intake (versus none) included all infants in the cohort (n = 1057), whereas the analysis of increasing HM intake included only HM-fed infants (n = 788).
The association of HM and ROP risk has been examined previously, with conflicting results. Hylander et al31 found that HM (any versus none) was associated with a decreased risk of ROP, even after controlling for gestational age, duration of supplemental oxygen therapy, 5-minute Apgar score, and ethnicity (OR: 0.46; 95% CI: 0.19–0.93), but no dose response was detected with increasing levels of HM intake; no association between HM and ROP was found in a study by Furman et al.27 Schanler et al32 reported that infants who received only HM, compared with those who were fed HM in addition to either donor HM or premature formula, were at lower risk for ROP, suggesting that higher intake of HM is associated with a lower risk of ROP.
It would be difficult to conduct a randomized, controlled trial of HM using infants’ own mother's milk. Observational studies such as this one may be confounded by factors that influence the mother's decision to provide HM and the length of time mothers choose to provide HM, such as socioeconomic status, ethnicity, maternal heath status, and infant health status. If these same factors also influence the risk of ROP, inconsistency in the results of studies of HM and ROP risk could be attributable, at least in part, to variation in the extent to which confounding has been controlled.
Other possible sources of variation across studies of HM and ROP include the amount of HM fed to infants, the age at which enteral feedings are initiated, the rapidity with which feedings are advanced, and the composition (eg, DHA content) of the HM. The DHA content of HM from women in the United States is very low, presumably because of low fish intake.16 In addition, in the present study, intake of HM-fed infants was relatively low, comprising only ∼15% of their total nutrition throughout their hospitalization. Although this is a very low level of intake, it is likely representative of the current intake of HM in extremely low birth weight infants in the United States. Also in this study, feedings were initiated after the first week of life in a majority of study infants. If HM does lower the risk of severe ROP, it is more likely to be effective when the antiinflammatory and antioxidant components of HM are provided earlier in life during a time when infants are exposed to high levels of oxidative stress and inflammatory cytokines. Averaging HM feedings over the entire hospitalization, as was done in this study, may not be sensitive enough to detect a potential effect of HM on ROP that operates over a specific limited time span.
Three limitations of our study that could have masked an association of HM and ROP should be noted. First, because the majority of infants were discharged before 42 weeks’ postmenstrual age and no data were collected about the results of eye examinations performed after discharge, we were not able to fully classify infants as to the development of ROP or its degree of severity, unless they developed disease severe enough to be treated surgically. Combining infants with no ROP and infants with ROP not treated surgically would be expected to attenuate an association between the level of HM intake and ROP, if such an association exists. It may be that HM intake is protective against the development of less severe forms of ROP. Second, data were missing for slightly >10% of the sample who did not require surgery for ROP before discharge and did not return for follow-up visits at 18 or 30 months’ corrected age. Infants with missing data for ROP outcome were less likely to have received HM and more likely to have had a pneumothorax. The latter was associated with a higher risk of severe ROP, suggesting that the bias resulting from losing infants to follow-up after discharge would have attenuated a potential “protective” effect of HM. A third limitation of this study is the absence of defined criteria for performing surgery for ROP across the study centers. However, we feel this is largely a theoretical concern as criteria for performing peripheral retinal ablation surgery are generally well accepted and based on those defined in the CRYO-ROP study.1
Despite our finding of no association between HM and severe ROP, HM has been found to have several other associated benefits, including a reduced risk of late onset infection and necrotizing enterocolitis,28 improved tolerance of enteral feedings,32 and possible beneficial effect on neurodevelopmental outcomes.33 Future research should be directed toward investigating the association between HM and ROP in an exclusively HM-fed group with complete follow-up of infants after discharge to at least 42 weeks’ postmenstrual age. Lactation support and initiation of enteral feedings as early as possible could be used to maximize HM intake in HM-fed infants.
We thank the children and their parents who participated in the randomized trial of glutamine supplementation and made this study possible.
Members of the NICHD Neonatal Research Network Trial (and NICHD grants) include University of Cincinnati: Alan Jobe, MD, PhD (chairman and principal investigator); University of California at San Diego (U10 HD40461): Neil N. Finer, MD (principal investigator), Yvonne Vaucher, MD (follow-up principal investigator), Chris Henderson, and Martha Fuller; Case Western Reserve University (U10 HD21364): Avroy A. Fanaroff, MB, BCh (principal investigator), Dee Wilson, MD (follow-up principal investigator), Nancy Newman, RN, and Bonnie Siner, RN; University of Cincinnati (U10 HD27853, M01 RR 08084): Vivek Narendran, MD, MRCP, Jean Steichen, MD (follow-up principal investigator), Marcia Mersmann, RN, and Teresa Gratton, RN; Emory University (U10 HD27851): Barbara J. Stoll, MD (principal investigator), Ira Adams-Chapman, MD (follow-up principal investigator), and Ellen Hale, RN; Indiana University (U10 HD27856, M01 RR 00750): James A. Lemons, MD (principal investigator), Anna Dusick, MD (follow-up principal investigator), Lucy Miller, RN, and Leslie Richard, RN; University of Miami (U10 HD21397): Charles R. Bauer, MD (principal investigator), Shahnaz Duara, MD, and Ruth Everett, RN; National Institute of Child Health and Human Development: Linda L. Wright, MD, Rosemary Higgins, MD, and James Hanson, MD; University of New Mexico (U10 HD27881, M01 RR00997): Lu-Ann Papile, MD (principal investigator), and Conra Backstrom, RN; Research Triangle Institute (U01 HD36790): W. Kenneth Poole, PhD, Abhik Das, PhD, Betty Hastings, and Carolyn Petrie, MS; Stanford University (U10 HD27880, M01 RR 00070): David K. Stevenson, MD (principal investigator), Susan Hintz, MD (follow-up principal investigator), and Bethany Ball, BS; University of Tennessee at Memphis (U10 HD21415): Sheldon B. Korones, MD (principal investigator), Henrietta Bada, MD, and Tina Hudson, RN; University of Texas Health Science Center at Houston (U10 HD21373): Jon E. Tyson, MD, MPH (principal investigator), and Georgia McDavid, RN; University of Texas Southwestern Medical Center (U10 HD40689): Abbot R. Laptook, MD (principal investigator), Roy Heyne, MD, Susie Madison, RN, and Janet Morgan, RN; Wayne State University (U10 HD21385): Seetha Shankaran, MD (principal investigator), Yvette Johnson, MD, Geraldine Muran, RN, and Debbie Kennedy, RN; Women and Infants Hospital (U10 HD27904): William Oh, MD (principal investigator), Betty Vohr, MD, Angelita Hensman, RN, and Lucy Noel, RN; Yale University (U10 HD27871, M01 RR 06022): Richard A. Ehrenkranz, MD (principal investigator and follow-up investigator), Patricia Gettner, RN, and Elaine Romano, RN; and University of Alabama at Birmingham (U10 HD34216): Waldemar A. Carlo, MD (principal investigator), Myriam Peralta, MD, Monica V. Collins, RN, and Vivien Phillips, RN.
- Accepted March 15, 2007.
- Address correspondence to Cherrie D. Heller, MD, MPH, Wake Forest University School of Medicine, Medical Center Boulevard, Winston-Salem, NC 27157. E-mail:
The authors have indicated they have no financial interests relevant to this article to disclose.
- ↵Good WV, Hardy RJ, Dobson V, et al. The incidence and course of retinopathy of prematurity: findings from the early treatment for retinopathy of prematurity study. Pediatrics.2005;116 :15– 23
- ↵O'Connor AR, Stephenson T, Johnson A, et al. Long-term ophthalmic outcome of low birth weight children with and without retinopathy of prematurity. Pediatrics.2002;109 :12– 18
- ↵Poindexter BB, Ehrenkranz RA, Stoll BJ, et al. Parenteral glutamine supplementation does not reduce the risk of mortality or late-onset sepsis in extremely low birth weight infants. Pediatrics.2004;113 :1209– 1215
- ↵Paneth N, Jetton J, Pinto-Martin J, Susser M. Magnesium sulfate in labor and risk of neonatal brain lesions and cerebral palsy in low birth weight infants. Neonatal Brain Hemorrhage Study Analysis Group. Pediatrics.1997;99(5) . Available at: www.pediatrics.org/cgi/content/full/99/5/e1
- ↵Ashraf RN, Jalil F, Zaman S, et al. Breast feeding and protection against neonatal sepsis in a high risk population. Arch Dis Child.1991;66 :488– 490
- ↵Boersma ER, Offringa PJ, Muskiet FAJ, Chase WM, Simmons IJ. Vitamin-E, lipid fractions, and fatty-acid composition of colostrum, transitional milk, and mature milk: an international comparative study. Am J Clin Nutr.1991;53 :1197– 1204
- ↵Schanler RJ, Lau C, Hurst NM, Smith EO. Randomized trial of donor human milk versus preterm formula as substitutes for mothers’ own milk in the feeding of extremely premature infants. Pediatrics.2005;116 :400– 406
- Copyright © 2007 by the American Academy of Pediatrics