Randomized outcome trial of human milk fortification and developmental outcome in preterm infants1

Transcription

1 Randomized outcome trial of human milk fortification and developmental outcome in preterm infants1 Alan Lucas, Mary S Fewtrell, Ruth Morley, Penny J Lucas, Nicholas J Bishop ABSTRACT Despite potential benefits, human milk may fail to meet preterm infants nutrient requirements. We tested the hypothesis that fortified breast milk, fed alone or with preterm formula, would improve neurodevelopment and growth at 1 8-mo follow-up without adverse short-term clinical or biochemical consequences. Two hundred seventy-five preterm infants from two medical centers (birth weight < 1850 g; mean gestation 29.8 ± 2.7 wk) whose mothers chose to provide breast milk were randomly assigned to receive for a mean of 39 d a multinutrient fortifier or control supplement containing phosphate and vitamins. Breast milk comprised 47.6% and 46.4% of enteral intake in fortified and control groups, respectively; preterm formula supplements were used when insufficient breast milk was available. Overall, there were no significant growth advantages with fortification; although, when breast milk exceeded 50% of intake, fortification promoted faster weight gain (an advantage of 1.6 g - kg - d ; 95% CI: 0.1, 3.1; P < 0.05). Compared with control infants, the fortified group showed 1) higher plasma urea from week 2 (P = 0.04), 2) higher plasma calcium (mean 2.34 ± 0.01 compared with 2.27 ± 0.02 mmolll; P = 0.003), 3) a greater rise in alkaline phosphatase by week 6 (P = 0.04), 4) more clinical infections (suspected plus proven; 43% compared with 3 1%, P 0.04), 5) a nonsignificantly increased incidence of necrotizing enterocolitis (5.8% compared with 2.2%, P = 0. 12), and 6) higher white cell and platelet counts. Developmental scores at 18 mo were slightly but not significantly higher in the fortified group. This study confirmed that breast milk fortifiers can improve short-term growth (when breast milk intakes are high); but beneficial effects on long-term development remained unproven. Future research is required to evaluate potential adverse consequences and explore more optimal fortification strategies. Am J C/in Nuir l996;64: KEY WORDS Breast milk fortifiers, preterm infants, neurodevelopment, growth, plasma biochemistry, infection, necrotizing enterocolitis INTRODUCTION Recent evidence shows that the way preterm infants are fed in the early weeks after birth has a major effect on long-term outcome (1-4). The importance of meeting the increased nutritional needs of preterm infants, compared with those of term infants, was emphasized by our previous study that compared preterm infants randomly assigned to a standard term formula Brian A Baker, Gill Lister, and or preterm formula during, on average, the first 4 postnatal weeks (1). Those fed the standard formula had major developmental disadvantages at 18 mo and our unpublished data now show a persistence of neurocognitive deficits at y. However, there is now also increasing evidence that human milk, which fails to meet the estimated needs of preterm infants for several key nutrients, may, nevertheless, confer clinical benefits, perhaps because of its unique nutrient or nonnutrient properties. Thus, the use of human milk was observed to result in improved enteral feed tolerance (compared with formula), reduced need for intravenous nutritional support, and less necrotizing enterocolitis (5, 6). Previously, we showed that when preterm neonates fed their own mother s milk were compared with those who were not, their intelligence quotients were 8.3 points higher at y, even after social and educational confounding factors were adjusted for (2). Although this latter study remains potentially confounded by sociobiological factors associated with the choice to provide breast milk, in a recent study of a more demographically homogeneous group of infants whose mothers had elected not to provide breast milk, those fed banked donated breast milk rather than a standard term formula had higher psychomotor scores at I 8 mo, despite the lower protein and energy contents of the donor milk (3). These findings led us to hypothesize that the use of fortified human milk in neonatal care might optimize outcome for preterm infants by providing them with the potential advantages of human milk together with those advantages conferred by meeting their calculated increased nutrient requirements. Numerous studies have shown that breast milk fortification with human milk or bovine protein, minerals, and vitamins can influence short-term outcomes, including growth (7, 8), nutrient retention (8, 9), bone mineralization (10, 1 1), amino acid profiles (12), and plasma protein status (9). However, no study I From the Medical Research Council Childhood Nutrition Research Center, Institute of Child Health, London; the Dunn Nutrition Unit and University Department of Pediatrics, Cambridge; Norfolk and Norwich Hospital, Norwich, United Kingdom. 2 Supported by Mead Johnson (Evansville, IN), which also supplied the fortifier. MSF was supported by an MRC Training Fellowship. 3 Address reprint requests to A Lucas, MRC Childhood Nutrition Center, Institute of Child Health, 30 Guildford Street, London WC1N IEH, United Kingdom. Received November 27, Accepted for publication March 26, Am J C/in Nutr 1996;64:142-5l. Printed in USA. to 1996 American Society for Clinical Nutrition

2 HUMAN MILK FORTIFICATION IN PRETERM INFANTS 143 to our knowledge has been large enough or specifically designed to test the hypothesis that human milk fortification improves long-term neurodevelopmental outcome. We tested this hypothesis in a randomized trial involving 275 preterm infants whose mothers chose to provide their own breast milk. The principal targeted outcome responses were developmental quotient at 9 mo corrected age and Bayley mental and psychomotor development indexes at 18 mo corrected age (13). However, because previous studies have generally been too small to provide adequate short-term clinical and biochemical safety data on human milk fortification, we also incorporated this objective in our trial. SUBJECTS AND METHODS Subjects and trial design We studied 275 preterm infants with a mean gestation of 29.8 ± 2.7 wk (range: wk) and a mean birth weight 1284 ± g. The infants were from two centers, Cambridge and Norwich, in East Anglia, United Kingdom, and were studied with informed consent and approval from the human ethics committees of the two trial hospitals and the local Committee of the Medical Research Council. Entry criteria for the study were as follows: 1) the mother chose to provide breast milk, 2) birth weight was < 1850 g, 3) gestation was < 37 wk, and 4) the infant survived to be assigned to a study group between 48 and 72 h of age. The only two exclusion criteria were that parents were not resident in the United Kingdom (which might preclude follow-up), and the presence of major congenital abnormalities (eg, trisomy) known to influence neurodevelopment (the principal outcome). Apart from these exclusions, the study was designed to include all eligible infants born in the two centers during the recruitment period of There were only three refusals to enter the trial: two in Norwich and one in Cambridge. Infants were randomly assigned to one of two diets at h. Random assignments were stored in sealed opaque envelopes at each center and the randomization sequence was determined by permuted blocks of randomized length. The randomization was stratified by birth weight, < 1200 g (n = 1 1 1) and g (n = 164), to assist balanced distribution between groups of the smaller, more high-risk infants. Infants were studied until they were discharged or reached 2000 g body weight, whichever was sooner. Follow-up was at 9 and 18 mo corrected age. Trial diets and feeding regimens The trial compared infants fed human milk fortified with either a multinutrient fortifier or with a control supplement containing only sodium and potassium phosphate and vitamins. Giving no supplement to the control infants was considered unethical because the low phosphorus content of human milk would greatly increase the risk of bone disease (14). The study was not intended to test the hypothesis that vitamin fortification influenced outcome (because vitamin supplements are used routinely in preterm infants); so both the multinutnent fortifier and control supplement contained the same vitamin concentrations, which were designed to meet the needs of preterm infants. The multinutrient fortifier used was manufactured by Mead Johnson (Enfamil human milk fortifier; Mead Johnson, Evansville, IN) in the form of a powder in packets. One packet was added to each complete 25 ml breast milk provided to the baby. The nutrient content of the fortifier that would be added to each 100 ml milk (four packets of fortifier) is shown in Table 1. The control supplement was prepared as a liquid, 0.25 ml, to be added to each complete 25 ml breast milk. The contents of 1 ml solution (to be added to each 100 ml) is shown in Table 1 and provided 15 mg phosphate/100 ml (1 50 mg/u) together with the same multivitamin content as the multinutrient fortifier. It was not possible to blind the study without giving a significant amount of an inert substance to the control infants; this was considered ethically undesirable. The study was intended to reflect common clinical practice and was therefore not designed to include only those infants fed breast milk exclusively. Thus, if a mother chose to provide breast milk but subsequently stopped doing so, or was unable to express the full volume required by the infant, a preterm formula (Osterprem; Farley Health Products Ltd, Kendal, United Kingdom) was used to bring total feed volume to the prescribed amount, reaching 180 ml - kg - d when full enteral feeds were established. Enteral feeds were initially increased according to feed tolerance and illness of the infant; TABLE 1 Composition of breast milk fortifier and control human milk supplement; amounts given are what would be added to each 100 ml breast milk Protein (g)3 Fat (g) Carbohydrate Energy (g) Fortifier Control supplement Preterm formula (Id) 58 IS 332 (kcal) 14 IS 80 Calcium (mg) Phosphorus (mg) Zinc (mg) Copper (zg) Magnesium (mg) 1-5 Sodium (mg) Potassium (mg) Vitamin A (.tg RE) Vitamin D (j.tg cholecalciferol) Vitamin E (mg TE) Vitamin C (mg) Folic acid (;.Lg) Thiamine (.tg) Riboflavin (,.Lg) Niacin (ILg) Pyridoxine (/Lg) Biotin (Lg) Pantothenic acid (pg) Phylloquinone (pg) Vitamin B-12 (.tg) IS, insignificant; RE, retinol equivalents; TE, tocopherol equivalents. 2 Osterprem (Farley Health Products Ltd. Kendal, United Kingdom): per 100 ml ready-to-feed. -I Whey:casein, 60:40 by vol.

3 144 LUCAS ET AL adjustments were made by the clinical team responsible for the infant, not by the researchers. Intravenous electrolyte and glucose solutions with or without intravenous amino acids and lipids were used in the early phase as required to meet fluid and nutrient needs. As soon as human milk was tolerated, the multinutrient fortifier or control supplement was added in proportion to the volume of breast milk fed. The fortifier or control supplement was added to the breast milk available for that day with careful attention to mixing. The milk was then dispensed into individual feeds as prescribed. Breast milk feeds (fortified or control supplemented) were given first and any formula feeds required were given toward the end of the 24-h period. Formula and breast milk were never mixed. Breast milk collection and storage Mothers provided breast milk by manual or mechanical expression. They were advised to undertake at least six expressions per day, and were guided by nursing staff on techniques to increase output. Despite this, a proportion of mothers did not persist in providing breast milk for the whole study period or did not provide their baby s required volume. Thus, the average proportion of breast milk received by the infant was just < 50% of enteral intake (Table 2). This figure was predicted from our previous trials ( 15). We elected at the outset to poststratify the infants so that a separate subgroup analysis (see below) would be performed on infants who received between 50% and 100% of their diet as breast milk. Freshly expressed breast milk was refrigerated for 48 h before use, or was frozen until used. Unlike the usual policy for banked breast milk (from unrelated donors) it was not the policy ofeither unit in to pasteurize a mother s own milk or to screen it for pathogenic organisms. Data collection Extensive demographic, clinical, anthropometric, and biochemical data were obtained by trained personnel at the two centers during the study period. Despite the impracticality of blinding, there were no differences between groups in the frequency of biochemical and clinical monitoring. For the main targeted outcome, long-term neurodevelopment, the investiga- TABLE 2 Demographic characteristics of the preterm infant study population Characteristic.. Fortified group (n 137) Control group (n 138) Birth weight (g) 1306 ± ± 26 Gestation (wk) 29.8 ± ± 0.2 Males (n) 65 [47]2 65 [471 Receiving antenatal steroids (ii) 33 [ [24.6] Days requiring partial or full parenteral nutrition 9.4 ± ± 0.9 Receiving dexamethasone (n) 9 [71 9 [7] Receiving diuretics (n) 79 [58] 76 [55] Days in study ± I ± 1.6 Belonging to higher social class (n) 67 [50] 61 [47] : ± SE. 2 in brackets. 3 Body weight 2000 g or discharge from the hospital. 4 Registrar General s classes I + II + III (nonmanual). tors were blind to the original dietary assignment. Anthropometry was performed as described previously (15). Body weight was obtained daily and body length, head circumference, and triceps and subscapular skinfold thickness measurements were obtained twice a week. Feeding Daily intake volumes of human milk and formula were recorded together with intake of intravenous nutrients. Daily feed tolerance data included gastric aspirate volumes, vomiting, stool frequency and consistency (assessed clinically on a three-point scale, see below), and abdominal distension. The milk fortifier (or control supplement) used was recorded to check that appropriate amounts were fed. Blood samples Blood samples were routinely analyzed by the hospital laboratories, initially daily and at a minimum weekly, for hematologic indexes, urea, electrolytes, albumin, calcium, and acidbase status. In the main center, Cambridge, weekly blood samples were also analyzed in our own laboratory for total protein, alkaline phosphatase, phosphorus, and amino acids (as previously described; 16). Clinical and demographic data Detailed obstetric and daily clinical data were collected and the mother s educational attainment was recorded on a fivepoint scale (ranging from no qualifications to higher professional qualifications) as described previously ( 17). Social class was based on the Registrar General classification with class III subdivided into manual and nonmanual classes ( I 7). Outcome data The subjects were invited to two follow-up appointments. At 9-mo postterm, examination included anthropometry and Knobloch et al s (18) developmental inventory (scores calculated as a quotient). At 18-mo postterm, further anthropometric measurements were performed together with the Bayley Scales of Infant Development (Mental and Psychomotor Scales; 13) and the Vineland test of social maturity (19). Developmental follow-up staff (RM in Cambridge, GL in Norwich) were blind to the original dietary assignments. Statistical analyses Analyses were made by Students t test, chi-square test, and regression analysis. Weight gain data were calculated from a best linear fit of data points beyond the time when birth weight was regained, and expressed as g - kg gained. d. Hospital biochemical and hematologic data were averaged for each week to reduce any bias incurred by individual differences in sampling frequency. Summary measures (either the mean or peak value, or the rate of change of the variable during the study period) were then used to test differences in hematologic and biochemical data between the dietary groups, to avoid the problems inherent in multiple comparisons of serial measurements of the same individuals (20). The rate of change of a measurement was examined by using multiple-regression analysis with the value of the variable at the end of the study as the dependent variable, and the baseline value (with or without the

4 HUMAN MILK FORTIFICATION IN PRETERM INFANTS 145 addition of intermediate values) plus dietary supplement as the independent variables. Initial sample size (141 infants per group) was calculated to detect a difference of one-third SD between groups (at 5% significance and 80% power) in developmental quotient; this would correspond to a five-point difference in Bayley scales between groups in a normal population (for which the SD = 15), but a six-point difference in those born preterm, in whom we found that the SD of the Bayley score is higher. Recruitment (n = 138 and 137 in the two groups) was thus on target. The major planned subgroup analysis was on babies who received either 50% or > 50% of their intake as breast milk; in the subgroup that received > 50% breast milk (in which the greatest dietary effect would be predicted), our study was planned to detect a < 0.5-SD difference in developmental score between groups (at 5% significance and 80% power). This was considered a plausible difference because a difference approaching 1 SD between feed groups in Bayley psychomotor scores at 18 mo was found in our previous study comparing standard with preterm formula (1). The biochemical and safety monitoring inevitably resulted in multiple response testing, which would increase the chance of identifying a dietary difference at the 5% significance level. No statistical correction was made for this and values close to the 5% significance level should be interpreted with this in mind. RESULTS Main characteristics of the two groups are shown in Table 2. There were no significant differences in birth weight, gestation, sex ratio, requirement for ventilation, days requiring parenteral nutrition, or social class. Short-term growth Breast milk comprised 47.6% of enteral intake in the fortifled group and a similar intake, 46.4%, in the control infants (Table 3). Short-term growth rates for the whole trial were not significantly different for weight, length, or head circumference gain. For greater comparability with those fortification studies in which the subjects were fed all or predominantly human milk, the cohort was substratified into those who received < 50% of their intake as breast milk (n = 141) and those (n = 134) who received 50% (Table 4). In the group with low breast milk intake (mean intake 20 ± 15%), the advantage for the fortified group was at or close to zero for the three growth variables. However, in the group with high breast milk intake (mean intake 81 ± 17%), there was a significant advantage in weight gain (1.6 g kg. d ), although the advantage in head growth seen in the fortified group was not significant. The mean percentage intake of breast milk and preterm formula in these infants was not significantly different between the groups (breast milk: 80.3% compared with 82.1%; preterm formula: 19.7% compared with 17.9%, respectively). Short-term hematologic and biochemical findings Hematologic and biochemical variables are expressed here as means for each week from weeks 1 to 6. Beyond the first month there was an inevitable and significant attrition in the number of subjects due to discharge from the hospital; those remaining in the study were the smaller or sicker infants. The number of infants who provided samples in successive weeks did not differ significantly between groups (maximum number: 129, 115, 103, 88, 77, and 56 in the fortified group, and 130, 1 16, 98, 80, 69, and 59 in the control group for weeks 1-6). Because our analysis showed that there was no significant difference between trial groups in attrition rate, the comparison of groups for hematologic and biochemical findings remains valid throughout the study period. Hematologic indexes There were no differences between groups in hemoglobin or red cell indexes, but the fortified group had a higher total white cell count over the period from weeks 1 to 6 after the baseline white cell count was adjusted for by using regression analysis (P = 0.01; Figure 1). The platelet count was also higher in the fortified group, the mean (± SE) peak value was 560 ± 17 X i09/l for the fortified group compared with 498 ± 16 X l09/l for the control infants (P 0.01). Urea electrolytes and acid-base status No longitudinal differences in plasma concentrations for sodium or potassium were found between groups, but urea values diverged and were consistently higher from week 3 in the fortified group (Figure 1). Mean blood urea values were often raised in the first 2 wk, but equally so in both groups. However, beyond 2 wk those fed the fortifier were more than twice as likely as the control infants to have urea concentrations > 5 mmol/l, and the rate of fall in plasma urea from weeks 2 to 6 was significantly lower in the fortified group (P 0.04; Figure 1). A greater proportion of infants from the control group had a plasma urea concentration that fell below 1 mmol/l (the lower limit of the reference range), 60 of 138, compared with 41 of 137 in the fortified group (P = 0.02). Base deficit, obtained by venous sample, was greater in the fortified group throughout the study after values in week 1 were adjusted for by using regression analysis (P 0.01). However, significant metabolic acidosis (base deficit < 8 TABLE 3 Short-term growth of infants fed breast milk supplemented with a multinutnient fortifier or a control supplement Or 0 wth in d cx Fortified (n 137) group Control group (n 138) Advantage for fortified group (95% CI) Percentage of enteral intake consumed as breast milk (%) 47.6 ± 3.1 ( ) 46.4 ± 3.0 ( ) Weight gain (g. kg - d) 15.6 ± ± (-0.3, 1.6) Length gain (mm/d) 1.33 ± ± (-0.18,0.11) Head circumference gain (mm/d) 1.44 ± ± (-0.04, 0.20) I SE; range in parentheses.

5 146 LUCAS ET AL TABLE 4 Short-term growth in infants fed breast milk supplemented with a fortifier or a control supplement, according to breast milk consumed as a proportion of total intake Breast milk < 50% of feed intake (n = 141) Breast milk 50% of feed intake (n = 134) Proportion of intake as breast milk (%) 20 ± ± 17 Short-term growth advantage for fortified group2 Weight gain (g - kg - d ) 0.0 (-1.26, 1.26) 1.6 (0.1, 3.1) Length gain (mm/d) (-0.31, 0.05) 0.09 (-0.17, 0.35) Head circumference gain (mmld) 0.06 (-0.10, 0.22) 0.15 (-0.03, 0.33) : ± SD CI in parentheses. P < mmolll) was not different between groups at any stage (Table 5). Plasma protein and amino acids The mean plasma protein concentration was not significantly different between groups, although further analysis indicated that in weeks 3-4, total plasma protein values were significantly higher in the fortified group (Figure 1). However, the incidence of hypoproteinemia (plasma protein < 40 mgfl) was not significantly different between groups. Plasma albumin, generally regarded as a more specific marker of nutritional status, was not significantly different between groups. Plasma amino acids (Table 6) were measured principally to ensure that human milk fortification did not result in significant hyperaminoacidemias, which were not observed. However, when weekly plasma amino acid values were averaged over the hospital stay, the fortified group had mean plasma concentrations of taurine (P = 0.01), proline (P = 0.02), glycine (P = 0.002), and phenylalanine (P = 0.01) that were lower than those of the control group. Indeed, only 1 of the 18 amino acid.j c 14.0 E 4.0 R E13.O a, C) CO 12.0 E3. a) U, C) CO J 6.0 concentrations measured (histidine) was higher (though not significantly so) in the fortified than in the control group. Plasma calcium, phosphorus, and alkaline phosphatase There were no significant differences in plasma phosphorus between groups at any age (Figure 2). Plasma calcium, however, was significantly higher in the fortified group from week 2 onward and remained higher thereafter (2.34 ± 0.01 compared with 2.27 ± 0.02 mmolll, P = 0.003). Moreover, hypercalcemia (plasma calcium > 2.85 mmolil) occurred in a greater, though nonsignificant, proportion of infants in the fortified group (P = 0.09; Table 5). Plasma alkaline phosphatase values (Figure 2) began to diverge after 3 wk and the rate of increase from 3 to 6 wk was significantly greater in the fortified group than in the control group (P = 0.02). Feed tolerance Summary data for vomiting, volume aspirated from the stomach, stool consistency, and abdominal distension are shown in Table 7. There were no significant differences be-.1 ;si #{176} 49 CO I I C I I Postnatal age (wk).j I I I I I I I C C Postnatal age (wk) Postnatal age (wk) 45L FIGURE 1. White blood cell count and plasma urea and plasma total protein concentrations for infants according to dietary group at from 1 to 6 wk of age. ± SE. For the fortified group #{149}. n = 129 in week 1, n = 56 in week 6: for the control group #{149}, n 130 in week I, n = 59 in week 6 (see text).

6 HUMAN MILK FORTIFICATION IN PRETERM INFANTS I 47 TABLE 5 Incidence of adverse events according to feed group Adverse events Fortified group Control group Biochemical events n/iota I 1%] High blood urea (> 5 mmol/l) 22/137 [161 10/ Metabolic acidosis (base deficit 8 mmolil) 40/137 [29] 49/ Hypercalcemia (plasma calcium > 2.85 mmol/l) 33/1 19 [28] 28/ Hypophosphatemia (plasma phosphate < 3.0 mg) 12/72 [ 17] 8/ High alkaline phosphatase (> 800 UIL) 12/72 [17] 7/ Clinical events Death 7/ /138 [6.51 Ventilation > 7 d 30/137 [22] 29/138 [211 Necrotizing enterocolitis 8/137 [5.8] 3/138 [2.21 Intraventricular hemorrhage None 1l2[84.0] Grade I 7 [5.3] 8 [6.01 Grade 2 7 [5.3] 6 [4.5] Grade3 2[l.5] 3[3.8l Grade 4 5 [3.8] 4 [3.0] Systemic infections All3 59/137 [43] 43/138 t3ll Confirmed by hematologic and bacteriologic evidence 14/137 [10] 8/138 [61 Necrotizing enterocolitis or confirmed systemic infection 22/137 [16] 10/ After 2 wk of age. 2.4 Significantly different from fortified group: 2 p 0.02, p = Includes those based on clinical and hematologic criteria. tween groups except that hard stools were reported twice as commonly in the fortified group (P = 0.007) by routine nursing staff (all stools on the two units were routinely recorded as loose, normal, or hard ). TABLE 6 Plasma amino acid concentration of preterm infants during the study period while consuming trial formulas. Amino. acid Fortified (n75) group pmol/l Control (n78) Alanine 217 ± ± 11 Arginine 59 ± 4 64 ± 4 Aspantate 21 ± I 23 ± 2 Glutamine 242 ± ± 18 Glycine 200 ± ± 82 Histidine 97 ± 7 89 ± 3 Isoleucine 48 ± 2 49 ± 2 Leucine 95 ± ± 4 Lysine 154 ± ± 6 Methionine 25 ± 1 27 ± 2 Ornithine 68 ± 2 70 ± 3 Phenylalanine 62 ± 1 71 ± 33 Proline 193 ± ± 10 Serine 134±4 143±5 Taurine 75 ± 3 89 ± 53 Threonine 210 ± ± 12 Tyrosine 101 ± ± 9 Valine 217 ± ± I I ± SE. 2,3 Significantly different from fortified group: 2 p < 0.01, - p < o.os. group Adverse events Our analysis of whether there was a differential incidence between the groups of adverse biochemical and clinical events is summarized in Table 5 (for biochemical events, see sections above). There were no differences between groups in the death rate, requirement for mechanical ventilation, or incidence of intraventricular hemorrhage or patent ductus. The number of infants with necrotizing enterocolitis (as defined previously) was 8 of 137 (5.8%) in the fortified group and 3 of 138 (2.2%) in the control group (NS). All systemic infections, based on clinical and hematologic criteria ( suspected infection, clinical suspicion with high or low white cell count, or low platelets) plus infection, also based on bacteriologic criteria ( confirmed infection ), were significantly more common in the fortified group (P = 0.04); but the observed numerical excess of the most stringently confirmed systemic infections (clinical, hematologic, plus bacteriologic evidence) in the fortified group was not significant. Long-term development and growth There were no significant differences between groups in growth or developmental indexes at 9 or 1 8 mo (Table 8). The 18-mo test scores were higher in the fortified group by 2.2 points for the Bayley Mental Development Index, by 2.4 points for the Psychomotor Development Index, and by 3. 1 points for social maturity, but these differences were not significant. Our trial was not large enough to detect differences of this magnitude. Further analysis did not provide evidence that those with the highest intakes of breast milk showed greater differences between groups (not shown in the tables). There was some evidence (not significant) for a greater advantage for fortifier in

7 148 LUCAS ET AL a, 2.45 i CU) 0. E C E 0. E235.. a,500 C ID = o CO al.80 B U) CO CO 0 E E 22 U, 0. U) CO CO E 22 Cl) I I Postnatal age (wk) Postnatal age (wk) Postnatal age (wk) FIGURE 2. Plasma calcium, plasma phosphate, and plasma alkaline phosphatase concentrations for infants according to dietary group at from 1 to 6 wk of age. ± SE. For the fortified group U, n = 129 in week 1, n = 56 in week 6; for the control group #{149}. n = 130 in week 1, n 59 in week 6 (see text). males, eg, the Bayley Mental Development Index at 18 mo was ± 3.4 compared with 96.0 ± 2.9 in the control infants. However, because there was no significant interaction between sex and diet, presentation here of a separate analysis according to sex was not considered justified. DISCUSSION Because human milk does not meet the estimated nutritional needs of preterm infants (21) and yet increasing scientific clinical evidence supports its value in neonatal care (2, 3, 6), it is logical that human milk fortifiers have been developed over the past y. However, fortification of human milk theoretically poses complex biological problems. First, expressed breast milk varies greatly between individuals in its composition, which is influenced by the volume of milk expressed, the type of milk obtained (foremilk or hindmilk), and the stage of lactation. Thus, our own unpublished data on 6000 samples from 24-h collections of human milk show that adding a fixed amount of protein to breast milk early in lactation, say 7 g/l, would give a proportion of babies more than the recommended TABLE 7 Feed tolerance data by group Tolerance index Fortified group (n 137) Control group (n 138) Abdominal distension (n) 47 [34] 44 [32] Vomiting (n) 94 [69] 94 [69] Hard stools (n) 35 [27] 18 [1312 Stools in study period 96 ± ± 4 Volume of gastric aspirate for whole study (ml) 140 ± ± 17 I % in brackets. 2 Significantly different from fortified group, P = i ± SE. 300 upper limit for protein, while giving some infants less than desirable intakes, so that the composition of a milk fortifier must at best represent a compromise. Second, human milk is a highly complex secretion (22) with live cells and a wide variety of biologically active factors and has separate fat globule and aqueous phases. Consideration must be given to whether fortification of milk might influence nutrient availability or biological properties, for instance of antiinfective factors (23). Given these problems and possibilities, it is important that major studies on efficacy and safety are undertaken. To date, most studies have been small and have concentrated on physiologic end points. The large outcome study reported here shows the complexity of testing milk fortifiers and of interpreting the findings. Our findings include areas of benefit, areas of possible concern, unexplained outcomes, and important unresolved questions that may require different experimental strategies in the future. Our main hypothesis that breast milk fortification promotes long-term neurodevelopment was not proven in this trial, although a neurodevelopmental benefit smaller than that detectable here remains possible. To confirm an outcome effect of milk fortification most efficiently, the intervention and control groups should be fed fully on human milk, as they were in several small studies (7, 9, 12, 24, 25). We elected not to pursue this design, but rather to enroll all preterm infants weighing < 1850 g whose mothers provided breast milk, regardless of the proportion of enteral feed intake that this eventually constituted. The average proportion of breast milk approached 50% of intake either because of insufficient supply or early cessation. The remainder of the feed was preterm formula. Thus, comparison of fortification with control supplementation of human milk was blunted because a significant proportion of the diet (preterm formula) was common to both groups. The importance of this design, however, is that it tested the outcome of current practice, because in our experience preterm infants are frequently fed on a combination of preterm formula and breast milk (15). Never- 6

8 HUMAN MILK FORTIFICATION IN PRETERM INFANTS 149 TABLE 8 Growth and developmental scores at 9 and I 8 mo corrected age Fortified group (n 125 at 9 and 18 mo) Control group (n = 123 at 9 mo and 120 at 18 mo) Advantage for fortified group (95% CI) 9 mo Developmental Growth scores2 Adaptive ± ± (-2.6,4.7) Grossmotor 102.7± ± (-5.3,5.1) Finemotor 98.7± ± (-1.8,6.2) Language 92.6 ± ± (-2.3, 3.2) Personal, social ± ± (-4.5, 2.5) Overall developmental quotient ± ± ( ) Weight (kg) 8.09 ± ± ( ) Length (mm) 69.8 ± ± (-0.6, 1.5) Head circumference (mm) 45.7 ± ± (-2.3, 5.4) 18 mo Developmental Growth scores3 Bayley Mental Development Index (13) ± ± (-3.4, 7.8) Bayley Psychomotor Development Index (13) 92.3 ± ± (- 1.9, 6.7) Social maturity quotient ± ± (-1.6, 7.9) Weight (kg) ± ± (-0.36, 0.27) Length (mm) 80.0 ± ± (-0.9, 0.8) Head circumference (mm) 48.1 ± ± (-0.4, 0.4) i ± SE. 2 Scores at 9 mo for the developmental inventory of Knobloch et al (18); referenced to a score of Referenced to a score of The Vineland Test (19). theless, we planned at the outset to poststratify for the percentage of breast milk consumed. The stratum receiving > 50% of intake as breast milk had a mean breast milk intake of > 81% of their enteral consumption. Short-term growth The control diet supplied multivitamins and one-third of the phosphorus content of the fortified diet. Thus, compared with the control infants, the fortified group had increased intakes of protein, energy [580 Id (140 kcal)/l], microminerals (zinc, copper), calcium, magnesium, and phosphorus. Perhaps the most important ingredient for short-term growth was the additional 7 g protein/l. For the trial as a whole, fortification of the breast milk component (50% of intake) had a small and nonsignificant effect on growth of any measurement. In the stratum receiving > 50% of intake as breast milk (mean of > 80% of diet as breast milk), a significant advantage 1.6 g - kg - d in weight gain was seen, although the increased head growth was not significant. Our further data modeling predicted that at 100% human milk intake, fortification would increase weight gain by 2.0 g - kg. d. Actual growth rates reported for intake of fortified breast milk vary greatly, as little as g kg d in one study (24); growth rates would be influenced by whether sick babies were included. Our study included a large unselected group in neonatal care. In the stratum of babies receiving the most breast milk (a mean of 81% fortified breast milk and 19% preterm formula), weight gain was 16.1 g - kg - d, slightly less than we reported in preterm infants fed preterm formula alone (17.2 g - kg. d ; 14). We would predict some growth shortfall compared with preterm formula, because in some infants breast milk plus fortifier would still provide a suboptimal amount of protein, resulting in reduced growth, whereas at the other extreme excess protein would not be expected to counterbalance this by promoting excessive growth. Metabolic monitoring Babies fed the fortifier had a greater acid load as evidenced by a greater base deficit, but had no increased incidence of frank metabolic acidosis. A significantly greater proportion of control than fortified infants had a plasma urea concentration < I mmol/l. This finding was associated with suboptimal diets (eg, unfortified donor milk, no longer used) in our previous studies (unpublished). The higher plasma urea in the fortified group reported here and elsewhere (8, 9) combined with the reduced incidence of low plasma urea might therefore be seen as an advantage, indicating improved nutritional status. However, beyond 2 wk we found more than two times the incidence of high blood urea concentrations (> 5 mmol/l) in the fortified compared with the control infants (16% compared with 7%); the clinical risk of such relative azotemia has not been well defined. We emphasize that 60% of the cohort needed yentilation and > 20% needed it beyond 1 wk. This unselected population might be expected to be more metabolically vulnerable than the low-risk subgroups often chosen by investigators for nutritional studies. Bone minerals Paradoxically, despite the very high amounts of calcium and phosphorus added to the fortifier [90 and 45 mg/loo ml, (900

9 150 LUCAS ET AL and 450 mgfl)] compared with only modest phosphorus supplement in the control supplement (15 mg/100 ml or 150 mg/l), the plasma biochemical findings were not those expected if bone mineralization had been improved. Although the higher calcium values seen with the fortifier by us and others (1 1, 26, 27) have been interpreted as improved calcium nutrition (27), the combination of higher plasma calcium, lower plasma phosphorus, and increased alkaline phosphatase, seen here, is the pattern conventionally described in phosphorus depletion with reduced bone mineralization (14). Indeed, the higher calcium in the fortified group was associated with a higher incidence of hypercalcemia (plasma calcium > 2.85 mmol/l), 28%, compared with 19% in the control infants (P = 0.09). Our data need to be reconciled with other studies showing improved bone mineral retention and bone mineralization with fortified breast milk (10, 26, 28). Further studies are needed, especially with sick infants, to explore the clinical significance of our findings. Adverse events None of the adverse biochemical events in Table 5 were significantly different between groups except for the higher proportion of those in the fortified group with raised plasma urea concentration. Clinical monitoring showed no difference in death rate. However, two clinical aspects raised some concern. First, taking all systemic infections into account, both suspected infection and confirmed infection (including, in addition, positive blood culture results), the fortified group was significantly disadvantaged, with 59 infants having one or more infections (43% incidence) compared with 43 cases (31% incidence) in the control infants (P = 0.04). Using the most stringent categorization of confirmed infections alone (dinical, hematologic, plus blood culture evidence), 14 such infective episodes (10% incidence) occurred in the fortified group and 8 (6%) in the control infants-a difference that was not significant at the 5% level. A larger study would have been needed to detect a difference at this level in incidence of such a relatively uncommon event. Such a difference is, however, biologically plausible in the light of the evidence that direct addition of fortifier to human milk resulted in significant reduction of activity of some antiinfective properties (23). Second, although we recognize that it is contentious to combine our subjects who had necrotizing enterocolitis with those with systemic sepsis, such an analysis revealed that in the fortified group 22 of 137 (16%) subjects had either stringently confirmed infection or necrotizing enterocolitis, whereas only 10 of 138 (7%) of the control infants fulfilled this criterion (P = 0.02). The incidence of necrotizing enterocolitis alone was 5.8% in the fortified group compared with 2.2% in control infants. Although not significant, a larger study than any undertaken so far would be required to detect a two- to threefold increase in incidence at the 5% significance level. This suggests a need for postmarketing surveillance with this type of product. Long-term growth and neurodevelopment We found no evidence that breast milk fortification influenced postdischarge growth at 9 or 1 8 mo or influenced overall developmental quotient at 9 mo (a 0.25-SD increase in the higher fine motor subscale scores in the fortified group were not significant). At 18 mo however, the overall trial (substantially blunted by > 50% intake of preterm formula in both groups) showed 2.2-, 2.4-, and 3.1-point higher scores, respectively, for mental development, psychomotor development, and social maturity scores in the fortified group. Our study was not large enough to detect significant differences of this magnitude because it was targeted to exclude a five-point difference in cognitive scores between groups. We cannot reject the hypothesis, therefore, that preterm infants, particularly those fed a greater proportion of their diet as fortified human milk, would have some increase in developmental scores compared with control infants; but a larger study, perhaps with exclusively breast-fed infants would be required. Interestingly, as in our previous studies (1), and indeed as in animal studies (29), there was a tendency for the greatest difference between groups to be found in males. For mental development a 6.2-point advantage in the fortified group was seen in males-although again, our study would need a larger cohort to confirm this. Overview In summary, this randomized outcome trial of milk fortification showed that the fortifier used was well-tolerated (apart from inducing harder stools) and promoted faster weight gain when > 50% of intake was breast milk. Improved nutrient status was indicated by some evidence of higher plasma protein concentrations, together with higher plasma urea. However, we were unable to confirm the expected benefits of the high fortification amounts of calcium and phosphorus in terms of either biochemical evidence of improved bone mineralization or increased linear growth; indeed, the fortified group may have been biochemically disadvantaged compared with the control group, which received more modest phosphorus supplementation. There was also an increased incidence of azotemia in the fortified group. Of more concern, there was some evidence that infection and necrotizing enterocolitis could be increased with human milk fortification, although this requires further confirmation. Findings of unknown significance included the elevated peak platelet count and increased white cell count. We were unable to confirm our hypothesis that milk fortification promoted neurodevelopment, but the nonsignificantly higher developmental scores in the fortified group, even when on average < 50% of the diet was breast milk, should encourage further work in this area. The use of human milk fortifiers remains a logical option, and some potential benefits have been observed. However, our study draws attention to the need to invest more in the evaluation of this potentially important nutritional strategy in neonatal care and to further define potential adverse consequences as well as potential outcome benefits. At a more fundamental level, examination of the consequences of multinutrient fortification on the biological properties of breast milk itself may result in evolutionary changes to fortification techniques that minimize risk and optimize benefit. U We acknowledge Peter Crowle, Norfolk and Norwich Hospital, for his collaboration. REFERENCES 1. Lucas A, Morley R, Cole TJ, et al. Early diet in preterm babies and developmental status at 18 months. Lancet l990;335:l

10 HUMAN MILK FORTIFICATION IN PRETERM INFANTS Lucas A, Morley RM, Cole TJ, Listen 0, Leeson-Payne C. Breast milk and subsequent intelligence quotient in children born preterm. Lancet 1992;339:26l Lucas A, Morley RM, Cole TJ, Gore SM. A randomized multicentre study of human milk versus formula and later development in preterm infants. Arch Dis Child 1994;70:Fl Lucas A. Role of nutritional programming in determining adult morbidity. Arch Dis Child 1994;7l: Lucas A. AIDS and milk bank closures. Lancet l987;1: Lucas A, Cole TJ. Breast milk and neonatal necrotising enterocolitis. Lancet 1990;336: Ronnholm KAR, Perheentupa J, Siimes MA. Supplementation with human milk protein improves growth of small premature infants fed human milk. J Pediatr 1986;77: Modanlou HD, Lim MO, Hansen JW, Sickles V. Growth, biochemical status, and mineral metabolism in very-low-birth-weight infants receiving fortified preterm human milk. J Pediatr Gastroenterol Nutr l986;5: Kashyap 5, Schulze KF, Forsyth M, Dell RB, Ramaknishnan R, Heird WC. Growth, nutrient retention, and metabolic response of low-birthweight infants fed supplemented and unsupplemented preterm human milk. Am J Clin Nutr l990;52: Greer FR, McCormick A. Improved bone mineralisation and growth in premature infants fed fortified own mother s milk. J Pediatr l988;l 12: Schanler RJ, Abrams SA. Postnatal attainment of intrauterine macromineral accretion rates in low birth weight infants fed fortified human milk. J Pediatr 1995:126: Polberger SKT, Axelsson IE, Raiha NCR. Amino acid concentrations in plasma and urine in very low birth weight infants fed proteinunenniched or protein-enriched human milk. J Pediatr 1990;86: Bayley N. Bayley scales of infant development. New York: Psychological Corporation, Bishop N. Bone disease in preterm infants. Arch Dis Child l989;64: Lucas A, Gore SM, Cole TJ, et al. A multicentre trial on feeding low birth weight infants: effects of diet on early growth. Arch Dis Child 1984;59: Lucas A, Baker BA, Morley RM. Hyperphenylalaninemia and outcome in intravenously fed preterm neonates. Arch Dis Child 1993;68: Morley R, Cole TJ, Powell R, Lucas A. Mother s choice to provide breast milk and developmental outcome. Arch Dis Child 1988:63: Knobloch H, Pasamanick B, Sherard ES. A developmental screening inventory for infants. J Pediatr 1966:38: Doll EA. Vineland Social Maturity Scale. Minnesota: American Guidance Service, Matthews JNS, Altman DO, Campbell MJ, Royston P. Analysis of serial measurements in medical research. Br Med J 1990:300: Lucas A, Hudson 0. Preterm milk as a source of protein for low birthweight infants. Arch Dis Child 1984;59: Lucas A. Human milk and infant feeding. In: Battaglia F, Boyd R, eds. Peninatal medicine. London: Butterworths, 1983: Quan R, Yang C, Rubenstein 5, Lewiston NJ, Stevenson DK, Kerner JA. The effect of nutritional additives on anti-infective factors in human milk. Clin Pediatr l994;33: Hagelberg 5, Lindblad BS, Persson B. Amino acid levels in the critically ill preterm infant given mother s milk fortified with protein from human or cow s milk. Acta Pediatr Scand 1990;79: Moro GE, Minoli I, Fulconis F, Clementi M, Raiha NCR. Growth and metabolic responses in low-birth-weight infants fed human milk fortified with human milk protein or with a bovine milk protein preparation. J Pediatr Gastnoenterol Nutr 1991:13: Schanler RJ, Garza C. Improved mineral balance in very low birth weight infants fed fortified human milk. J Pediatr 1988:1 12: Ehrenkrantz RA, Gettner PA, Nelli CM. Nutrient balance studies in premature infants fed premature formula or fortified human milk. J Pediatr Gastroenterol Nutr 1989:8: Carey DE, Rowe JC, Goetz CA, Horak E, Clark RM, Goldbert B. Growth and phosphorus metabolism in premature infants fed fortified own mother s milk. Am J Dis Child 1987:141: Smart JL. Undernutrition, learning and memory: review of expenimental studies. In: Taylor TO, Jenkins NK, eds. Proceedings of the XIII International Congress of Nutrition. London: Libbey, 1986:74-8.