Postdischarge Nutrition in Preterm Infants

For 35 years, research focused on undernutrition after discharge in infants born preterm has been reflected in numerous randomized trials, systematic reviews, and meta-analyses that address whether nutritional intervention has benefits, notably for growth or neurodevelopment. Despite this extensive research effort, there are major gaps in the evidence, including insufficient trial evidence on neurodevelopmental outcomes and lack of research attention to common subgroups, notably infants who have residual respiratory or feeding problems. This includes those who cannot regulate their feeding volume intake. Importantly, there is a paucity of interventional trials in preterm infants after discharge who are breastfed, the recommended mode of feeding in this population. There are different interpretations of the evidence and significant differences in global practices. One goal of this review is to revisit the extensive background evidence regarding postdischarge nutrition, with the additional goal of helping identify gaps in research.

After completing this review, readers should be able to:

1. Describe the evidence and clinical science of postdischarge nutrition growth in infants born preterm.

2. Explain whether postdischarge nutritional interventions can improve neurodevelopment in preterm infants.

3. Compare the strengths and limitations of existing evidence of postdischarge nutrition on growth and neurodevelopment.

Over 35 years ago, amid increasing interest in the importance of early nutrition in hospitalized preterm infants, there emerged an awareness that these infants were being discharged from the hospital in a suboptimal nutritional state. At that time, preterm infants were discharged often close to term equivalent, yet might have had a body weight little more than half that of a term born infant, with growth restriction that persisted well into childhood. (1) This matter was highlighted in the provoking title of Embleton’s study over 20 years ago: “Postnatal malnutrition and growth retardation: an inevitable consequence of current recommendations in preterm infants?” (2) This study quantified the substantial estimated deficits in body protein and energy as preterm infants entered the postdischarge period. Indeed, a range of other deficits was described in body stores of specific nutrients including calcium, phosphorus, zinc, copper, and iron. (3) Reduced body stores of calcium and phosphorus may have implications for bone health. (4)(5) Yet, at the time that these problems with nutritional status and growth were being identified, common practice was to stop the nutritional supplementation used for hospitalized preterm infants at discharge and change to diets suitable for full-term infants.

These initial concerns about preterm infants after discharge provoked questions relevant to current practice. Can we, and should we, attempt to improve growth and nutritional status in this population with nutritional supplementation? Arguably of most importance, is the postdischarge period a critical window during which enhancing nutrition and growth could favorably “program” later cognitive function?

The attempt to address these questions is reflected in a body of published work over the past 35 years, notably randomized controlled trials (RCTs) of postdischarge nutritional interventions in infants born preterm using nutrient-enriched formulas (6) or supplements given to breastfed infants. (7) Observational findings have also contributed to this field, notably studies that test for a relationship between postdischarge somatic or head growth and later neurodevelopment. (8) However, the evidence from modern systematic reviews has proved far from easy to interpret. Also, a further question arose concerning the safety of enhancing postdischarge growth given that early growth promotion has been linked with adverse programming of the metabolic syndrome. (9)

In the United States, widespread practice is to routinely use nutrient-enriched postdischarge dietary regimens for breast- and formula-fed infants born preterm. In other countries, postdischarge practice is more variable. In the United Kingdom for instance, not all NICUs routinely use enriched postdischarge formula (PDF), with a move away from this in recent years; such formulas are more generally used for those with residual clinical problems and very-low-birthweight infants. In the United Kingdom, infants born preterm who are breastfed after discharge most often do not receive fortifiers (N.E. Embleton, personal communication).

Those who have appraised the literature vary in their conclusions, some arguing that there is little evidence that nutritional interventions have an established role in promoting growth and in particular, neurodevelopment. (6)(8)(10) Our purpose in this review is to revisit the underlying clinical science to help determine its strengths and weaknesses in underpinning clinical practice in this area. We emphasize that while this review examines evidence on postdischarge nutrient supplementation in breast- and formula-fed preterm infants, breastfeeding after discharge is greatly recommended (11) and may confer benefits beyond the scope of this review.

Dominant in the overall strategy for postdischarge nutrition is encouraging and assisting mothers to breast-feed, as emphasized, for instance, by the American Academy of Pediatrics (AAP). (11) However, historically, much of the research in postdischarge nutritional supplementation was conducted in formula-fed infants born preterm. Before reviewing evidence on nutrient supplementation in breastfed infants born preterm, the evidence from those fed formula will be considered because this establishes principles relevant to all infants after discharge.

The late 1980s and early 1990s saw a cluster of small RCTs that began to examine postdischarge diet, growth, and development in infants born preterm. (6) The first to use a diet specifically for postdischarge nutrition was conducted in 1992 by Lucas and coworkers as a pilot study in non-breastfed infants. (12) This tested whether feeding a PDF versus a standard term formula (TF) in former preterm infants up to 9 months’ corrected age (CA) could affect growth, bone mineralization, and volume intake. The PDF was enriched in protein, calcium, phosphorus, zinc, copper, and selected vitamins, and provided minimal additional energy. This enriched formula, which no longer exists, was nevertheless an initial template for further development of postdischarge diets. This RCT, described below, helped define key hypotheses subsequently tested in RCTs that form the substance of this review. In this trial, Lucas et al obtained biweekly measurements of weight, length, and occipitofrontal head circumference (OFC) from discharge until 9 months’ CA. (12) Each body measurement was the same in both groups at randomization, but thereafter all 21 measurements for each growth parameter were higher in the PDF group. In longitudinal analyses that used all data points, the PDF group had significantly enhanced weight and length gain and a strong trend toward faster head growth. (12) Growth data, when superimposed on a contemporary growth chart, showed substantial dietary effects. Thus, body weight at 9 months CA was close to the 25th percentile in the PDF group but between the 3rd and 10th percentile in the TF controls. The study showed bone mineral content (BMC) was low at randomization, with faster catch-up seen in the PDF group.

Milk volume intake was measured by supplying the formulas in preweighed bottles from which the infant fed. (13) After each feeding, the bottle was recapped and retained to be reweighed—a technique previously validated by stable isotope kinetics accounting for regurgitations. No trial had previously measured formula volume intake in ad libitum–fed infants continuously for 9 months. Two physiologic observations emerged. Firstly, there was no difference in volume intake between randomized groups. Previous data showed that when energy is supplemented, infants adjust their volume intake to equalize energy intake of intervention and control groups. However, in this study, the energy intake difference between groups was trivial (6%), suggesting that infants may not regulate volume intake based on protein content, which differed significantly between formulas. (12)(13) Secondly, the formula volume intake by ad libitum–fed preterm infants after discharge was much higher than in term infants. At 1 to 4 weeks, mean volume intake was 230 mL/kg per day, so that a 4-week-old infant after discharge received an estimated mean protein intake of 3.3 g/kg per day from TF and 4.3 g/kg per day from PDF, the latter, typical of intakes given to hospitalized extremely preterm infants.

This trial (12)(13) raised the hypothesis that the findings would hold up in larger studies and that growth effects would persist beyond the intervention. The faster growth and trend toward larger OFC also led to the hypothesis that postdischarge supplementation could influence neurodevelopment. Our data indicated that actual nutrient intakes were highly influenced by volume intake regulation in ad libitum–fed infants. The impact on BMC indicated that quality of growth including bone mineral accretion and body composition needed further exploration. These hypotheses are explored here using evidence from numerous RCTs.

By 2012, 25 randomized trials of nutrient-enriched formulas had been conducted in preterm infants after discharge who are not breastfed. (6) The most recent meta-analyses of these were in the 2016 Cochrane review by Young et al, with very few new studies thereafter. (6) Young et al excluded 9 of 25 trials that failed the eligibility criteria used for composition of the trial diets, for instance, those comparing isocaloric formulas, which may now be viewed as eligible. Of the 16 remaining trials, 5 used preterm formula (PTF)(14)(15)(16)(17)(18) and 1 used PDF, of which 7 were included in meta-analyses (note: PDF and PTF are discussed further later in this section). (1)(12)(19)(20)(21)(22)(23)

Young et al concluded that, given inconsistently significant growth differences between PDF and TF, PDF did not convincingly affect growth. (6) However, inspection of the Young et al meta-analyses showed that PTF, generally only licensed at that time for in-hospital use, more often, but not consistently, increased body weight, length, and OFC at periods up to 18 months’ CA. The growth data, together with a failure to detect developmental benefits using the Bayley scales in 1 PDF and 2 PTF trials, led Young et al and other groups to conclude that the data did not support prior recommendations to use PDF. (6)(10)

We suggest further consideration before such a conclusion can be reached. Neurodevelopmental aspects will be considered later, and growth considered here. Young et al showed much variation in the growth effects of nutritional intervention. (6) This could relate to variation in formula composition. If an enriched diet had a higher energy content than the TF, the expected differential volume intake adjustment to equalize energy intake between groups could greatly diminish the difference in protein intake and reduce potential growth advantage in the nutrient-enriched group.

There are, however, statistical reasons why growth effects might not have been detected in some meta-analyses from Young et al. (6) Their understandable problem was to unify the studies to combine them in meta-analyses. They chose a cross-sectional approach of doing meta-analyses of data available at 3, 6, 9, 12, and 18 months’ CA. This approach eliminates RCTs from meta-analyses when no data were collected at a particular time point. With a reduced number of trials in the meta-analyses at any point, and attrition of RCTs in later follow-ups, statistical power to detect a difference would be reduced. Also, variation in which RCTs contributed to successive time points could have produced spurious variability in the dietary effect over time.

However, a dominant factor causing major loss of statistical power is the complete separation of meta-analyses into 2 groups, PDF interventions and PTF interventions, resulting in as few as 1 or 2 RCTs in some meta-analyses. (6) It might seem reasonable to analyze effects of different types of formula separately, but we elected to combine the PDF and PTF trials for several reasons. A key priority is to establish robustly whether we can improve postdischarge growth of preterm infants with nutrient-enriched formula, statistically favored by combining PDF and PTF interventions. Indeed, there is no formal distinction between PTF and PDF; PTF used previously often had lower protein (eg, 2.0 g per 100 mL) than PDF used now (eg, 2.1 g per 100 mL), which is like the PTF used in the RCT by Cooke et al (2.2 g/100 mL). (15) RCTs that do and do not show a positive effect on growth are found in both PDF and PTF trial groups. A key factor for growth, regardless of whether PDF or PTF is used, is receiving a high protein to energy ratio, noted by Teller et al; thus, Yu et al performed the only RCT of PDF versus highly enriched PTF but with equal protein to energy ratios, and both groups had the same growth rates. (24)(25)

For the previously described reasons, we present a new combined meta-analysis of the studies identified in the 2016 Cochrane review of Young at al. (6) We have not altered their trial eligibility criteria, but we included the eligible PDF versus TF trial by Carver et al, excluded from their meta-analysis principally because of absent n values, which we retrieved. (26)

The effect sizes are summarized in Table 1 (for reference, the forest plots are in the Fig). No significant early effects (3–4 months’ CA) were seen for weight, length, or OFC but effects began to emerge at 6 months for OFC and are then seen for weight, length, and OFC from 9 months onwards. Of the 9 meta-analyses at 9, 12, and 18 months’ CA, 6 showed significant positive effects for the enriched diet and there was a trend to a positive effect in another meta-analysis (Table 1). Thus, the data indicate predominantly that enriched formula improves postdischarge growth.

Interestingly, in 8 of 13 trials used in our meta-analyses, the intervention lasted 2 to 6 months, (6) yet the major impact on growth was at 9 to 18 months. Indeed, in 4 individual trials, an enriched diet significantly increased 1 or more of weight, length, or OFC beyond the period of intervention by 6 to 10 months. (1)(15)(17)(18) There is little evidence that growth effects persist beyond 18 months’ CA but a key question is whether dietary effects on OFC and therefore brain volume during this critical period could affect development, even if later catch-up in OFC occurred.

However, these growth effects cannot be interpreted fully without considering the impact of a single influential trial, discussed below.

In our forest plots, one RCT by Koo and Hockman (20) appears visually as an outlier RCT showing significant, large negative effects on growth for PDF versus TF, whereas other trials either showed no effect or a beneficial one. Teller et al (25) reviewed 30 postdischarge RCTs and the trial by Koo and Hockman alone showed a significant negative effect. (20) Interestingly, Koo and Hockman (20) used the same PDF as Carver et al, (26) but with opposite effects on growth (Fig). We considered if PDF could suppress growth by protein overload, but this is not supported in other trials, including those studying even higher protein contents. Koo and Hockman (20) suggest that their RCT lacked demographic confounding seen by others, but most other studies had no significant imbalance of baseline demographic factors yet did not show that enriched formula suppressed growth. The apparent adverse effects of PDF in the study by Koo and Hockman were not confined to growth rate: 18 RCTs (25) examined body composition, 7 using dual x-ray absorptiometry or air displacement plethysmography and only Koo and Hockman reported an adverse effect, with lower lean and fat mass. BMC was examined in 5 trials (25)(27)—the nutrient-enriched group had no benefit in 2 trials, but 2 trials (one of breast-milk fortification) showed a positive effect, yet Koo and Hockman alone saw a significantly unfavorable effect. It is hard to see how this modestly enriched PDF could significantly depress growth, quality of growth, and BMC; our calculations indicate that this is most unlikely and these anomalous results have attracted attention from others. (25)

Unless an outlier RCT reflects a demonstrable error, the purpose of identifying it is not to eliminate it from meta-analyses, particularly if entry criteria were met, because this could negate the objectivity of the process and add bias. However, it is important to identify and report outliers if they have disproportionate influence. Also, we and others (6) note unexpectedly marked heterogeneity in postdischarge meta-analyses. We speculated that the Koo and Hockman study might have been contributory. Sources of heterogeneity need identifying because this may affect robustness of the findings. We decided to explore these matters.

One recommended approach to identifying studies with major influence is a “leave-one-out analysis.” (28) We report a pragmatic sensitivity analysis simply with and without the study by Koo and Hockman. At 4 periods (3, 6, 9, and 12 months’ CA), Koo and Hockman provided data for meta-analyses of weight, length, and OFC. Thus, 12 meta-analyses could be used for our analysis. Table 2 shows the highly influential impact of the Koo and Hockman RCT on estimated effect sizes.

In 11 of 12 of these meta-analyses including the trial by Koo and Hockman, significant heterogeneity was seen (Higgins I² >50%); however, when the Koo and Hockman RCT was deleted from our analysis, no meta-analysis showed significant heterogeneity (I² <50%) with 6 of 12 meta-analyses showing heterogeneity values of only 0% to 15%. Thus, 1 outlier study largely explained the observed heterogeneity.

When a single study has such a disproportionate impact on effect size (Table 2) and explains so much of the heterogeneity, some doubt is cast on the validity of the estimated effect. Extrapolating from this, clinical conclusions heavily influenced by the impact of a single study would also lack robustness. Thus, the conclusion that PDF does not affect growth overall needs revisiting. Indeed, our analysis (not shown in this review) of the original subset of trials that solely used PDF, which were used to inform that conclusion, was highly influenced by the data of Koo and Hockman, which markedly reduced effect sizes and significance.

The explanation of the findings of Koo and Hockman remains unclear, so we cannot exclude them, but we demonstrate their disproportionate impact on effect size, and importantly, practice recommendations.

In Table 1, our meta-analyses (including the study from Koo and Hockman) reveal that an enriched diet in infancy may significantly increase infant weight by about 300 g; body composition studies show this weight accretion includes good quality lean tissue. Using the enriched diet, length was increased by 1 cm but the impact on head growth was more notable. OFC is measured as an indicator of brain volume but because OFC is a linear measurement and brain volume a cubic one, a very small change in OFC reflects a counterintuitively large change in brain volume. This principle can be crudely illustrated as follows: Martini et al showed in healthy 4- to-12-month term infants that OFC had 78% predictive value for intracranial volume (ICV; an indicator of brain size), though it is less reliable in those with abnormally shaped skulls, which may include some preterm infants. (29) We can only very crudely estimate ICV from OFC alone. Martini et al showed that between 4 and 12 months, healthy full-term infants increased their ICV by 405 mL and their OFC by 3.5 cm (35 mm). (29) Therefore, the 3.3-mm OFC advantage for the enriched formula group at 6 months’ CA (P<.003) shown in 11 RCTs could reflect a very crudely estimated advantage of 40 mL of ICV.

These effect sizes may be underestimated firstly because of the possible disproportionate blunting by an outlier RCT (Table 2), and secondly, the potential to optimize growth using postdischarge diets with high protein to energy ratios, as highlighted in the analysis by Teller et al. (25)

The effect sizes here are estimated population means but specific subgroups may show greater or lesser effects. A well-studied factor is sex: 9 of 14 trials showed a greater effect of diet in males, and 5 RCTs showed no difference. (25) Two RCTs showed a sex-related favorable impact on quality of growth that was not seen in 2 others.

Some data had indicated that preterm infants born small for gestational age (SGA) may be more resistant to postdischarge interventions to promote growth, suggesting their postnatal growth trajectory was programmed in fetal life, but others have not found this.

Carver et al (26) found a significant interaction between lower birthweight (<1,250 g) and postdischarge diet, indicating that smaller infants, perhaps of lower gestation, were more responsive to an enriched diet in all growth parameters, noted also for OFC in the breast milk fortification trial of O’Connor et al. (27)

Because of inconsistency in the data on the sensitivity of subgroups of infants to nutrient-enriched postdischarge diets, it is difficult to predict which individual infants would benefit more than average from such a diet. This might favor a population approach rather than targeting selected subpopulations. Nevertheless, we recognize that sick or poor-growing preterm infants after discharge will need individual attention.

A recurring theme in this review is the hypothesis that after discharge, infants born preterm regulate volume intake in relation to the energy content of the feeding rather than the protein content. This hypothesis is supported by data from 21 RCTs. (25) Volume adjustment made by the infant to equalize energy intake between randomized groups tends to reduce the difference between groups in protein intake, yet protein is critical for growth because it provides nitrogen, required for protein synthesis and hence new tissue accretion. Recognizing this, Teller et al showed that those formulas with a high protein to energy ratio promoted increased weight and in particular, length and OFC. (25) However, the ratio itself may prove less important than simply supplying adequate extra protein to counteract the effect of volume downregulation.

Vermont Oxford Network data from 2017 indicated 56.8% of preterm infants after discharge are still receiving human milk. (11) Historically, there was concern about postdischarge nutritional status in breastfed infants without added nutritional support. Our trial included growth data on 65 breastfed reference infants after discharge and 229 infants randomized to TF or PDF. (1) A comparison of infants who were PDF-fed versus breastfed found that by 6 weeks’ CA, infants who were almost exclusively breastfed were already 0.5 kg lighter and 1.6 cm shorter. (1) Differences were maintained at 6 and 9 months’ CA when the breastfed group had generally stopped breastfeeding. Such evidence underpinned recommendations for fortification. (30) However, the paucity of RCTs on breastfed preterm infants after discharge is regrettable: until 2008, no RCTs had explored whether fortification of human milk promoted postdischarge growth (considered here) or improved neurodevelopment (considered later).

In 2008, the pilot RCT by O’Connor et al of 39 preterm infants (27) compared a predominantly breastfed control group with a group of infants whose mothers expressed 50% of their milk, which was fortified to the level of a typical PDF and then bottle fed to the infant. Fortification, adjusted to the infant’s weight, was given for 3 months. In 2011, a much larger RCT was conducted by Zachariassen et al of 207 preterm infants, randomly assigned to a postdischarge control group that had received any human milk or an intervention group receiving an extra 1.4 g protein and 17.5 kcal daily for 4 months as fortifier powder mixed in 20 to 50 mL of expressed milk and then bottle fed to the baby. (31) Both trials examined weight, length, and OFC at 12 months’ CA, which was 8 to 9 months after the intervention ceased.

These 2 trials were combined in a meta-analysis in a Cochrane review. (7) Young et al (7) used a different analytic approach than the O’Connor et al study (27) to combine 2 different studies; Young et al reported that O’Connor et al found significant positive effects for the multinutrient-fortified group for body weight at 12 months’ CA, and length and OFC at both 3 and 12 months’ CA. The Zachariassen et al trial showed no significant effects of fortification on growth at 4 or 12 months. (7)(31) However, meta-analysis of both trials revealed a significant positive effect on body length at 12 months.

The difference in effect between the 2 trials was notable. The O’Connor et al RCT with just 39 subjects only had power to detect sizable differences. At 12 months’ CA, advantages for the intervention group, according to the Young et al analysis, (7) were 1.2 kg for weight, 3.8 cm for length, and 1.0 cm for OFC, reflecting a sizeable though crudely estimated increase in intracranial volume (as calculated above). This substantial impact on head growth may relate to the early developmental impact of fortification considered later. Also, O’Connor et al found that BMC was increased at both 4 and 6 months’ CA. (27) Zachariassen et al note that lack of growth effects in their study might signify the low fortification used, reflecting their concern that supplementary nutrition could reduce breastfeeding duration. (31) A further difficulty in studies of this type is postrandomization cessation of breastfeeding leading to a major blunting effect for intention-to-treat analyses; by 4 months’ CA, only 19% of the intervention group were taking their assigned diet. (31) For the same reason, per-protocol analyses were underpowered. Nevertheless, growth advantages for somatic and head growth were seen in female infants. A PTF-fed reference group (31) showed greater growth than with either of the trial diets, indicating the potential for enhancing growth with more breast-milk fortification, which was recommended by Zacchariassen et al. Although no further RCTs have been conducted for a decade, a 2020 AAP publication (11) suggests at least smaller or growth-restricted breastfed infants receive fortification after discharge for at least 12 weeks (see summary and conclusions section).

We return now to a key hypothesis that postdischarge nutrition of infants born preterm affects neurodevelopment. This would accord with a more general view that early undernutrition impairs long-term neurodevelopment, and that nutritional intervention may improve developmental outcomes. However, historically it has been extraordinarily difficult to prove this. After reviewing 165 animal experiments, Smart concluded that early undernutrition adversely affected later behavior and learning. (32) Yet, extrapolating these animal studies, principally on rodents, to human intellect has not been compelling. The large body of past epidemiologic evidence from developing countries linking infant malnutrition with reduced later cognition was addressed in an expert meeting convened by Dobbing in 1987. (33) A consensus was that past study designs were so flawed that it was impossible to conclude that malnutrition affected cognition, and that adverse cognitive outcomes identified may have related to poor social circumstance, poverty, and lack of stimulation However, RCTs in hospitalized preterm infants in the early 1980s began to identify more compelling evidence of an impact of early nutrition on neurodevelopment. Preterm infants randomly assigned to a PTF versus TF as sole diets or as supplements to the mother’s own milk in the neonatal period had superior Bayley psychomotor development index and social quotient values and a lower incidence of moderate developmental impairment at 18 months’ CA. (34) Compared with preterm infants randomly assigned to nonenriched diets, those fed nutrient-enriched diets had higher verbal IQ by 7 points at 7 to 8 years of age, and these same subjects had a similar advantage in verbal IQ at 16 years, when magnetic resonance imaging studies also showed that the enriched diet group had larger caudate nuclei, brain structures linked to cognitive performance. (35)(36) Such trial data are backed by observations showing that neonatal growth in weight or OFC is positively related to longer term cognitive function. (8)(37)(38) Indeed, nutrient supplementation in just the first week after birth in low-birthweight infants has been linked to higher developmental scores at 18 months. (39)

The question here is whether this neonatal window of sensitivity or vulnerability of the brain to early nutrition extends into the postdischarge period.

The systematic review by Teller et al (25) identified 10 RCTs that examined effects of enriched PDFs on cognitive development in infants born preterm (1)(15)(16)(17)(23)(40)(41)(42)(43)(44); an 11th study by Werkman and Carlson (45) was retrieved from conference proceedings. At that time, all these studies examined developmental outcomes within the first 2 years. A further RCT of fortification of breast milk after discharge included developmental testing in the first year. Longer term follow-up data have now emerged at 8, 10, and 11 to 16 years. The interpretation of these RCTs has been challenging.

It might be hypothesized that the postdischarge period for preterm infants is a less sensitive one for nutrition than the predischarge period when brain growth is so rapid. If so, postdischarge trials would need to be larger to detect smaller effect sizes. In the case of a quotient score like IQ, the standard deviation (SD) is 15 points. We might argue that a 5-point difference between groups (0.33 SDs) would be an effect size that we would not wish to miss. However, to detect a 0.33 SD difference in any test score between groups, for instance, at 80% power and 5% significance would require 144 subjects per group for a 2-limb comparison, 288 subjects. No RCT of that magnitude has examined postdischarge nutrition. Unfortunately, of the trials cited here, 50% had total trial sizes at the end of the study period of less than 50 subjects and only 3 had more than 100 subjects.

A solution to having multiple underpowered studies is meta-analysis. This was feasible when analyzing growth data (above), because there is some agreement that body weight, length and OFC are a standard outcome package. But various neurodevelopmental tests were performed in these trials, including the Neonatal Behavioral Assessment Scale, the Fagan Test of Infant Intelligence, visual grating acuity and contrast sensitivity, the Griffith Mental Development Scale (GMDS), Knobloch Passamenick and Sherrard’s test (KPS), the Bayley Scales of Infant Development (BSID), Wechsler Intelligence Scale for Children (WISC), motor functioning, and performance in school English and mathematics examinations. To amalgamate these disparate tests in meta-analyses would not just reflect proverbial comparison of apples and oranges but also of pears, bananas, grapefruits, cherries, etc.

Given the diversity of tests and testing periods, the major problem of underpowering when the RCTs are considered individually, and difficulty in statistically combining RCTs in a meta-analysis, our expectation should be that it would be difficult to detect neurodevelopmental benefits at all even if they exist. In statistical terms, the potential for type II error is great. The hypothesis that enhanced postdischarge nutrition affects neurodevelopment favorably would be expected to be unprovable in most individual studies. This is not just because of underpowering but because some of these varied studies might include some intervention that was not optimal, for instance, a low protein to energy ratio diet. However, there might be occasional studies in which a favorable effect was large enough to be detected. Given this background, the findings are summarized below.

The postdischarge fortification trial of O’Connor et al assigned preterm infants for 3 months to a control breastfed group or a group in which 50% of the mother’s milk was fortified, which promoted a 1-cm increase in OFC at 1 year CA. (27) Visual development was used as a surrogate for neurodevelopment at 4 and 6 months’ CA. Visual acuity was significantly higher in the fortified group at both ages with a large effect size (1.2 SD at 6 months). Trends of higher contrast sensitivity were noted with fortification. The BSID at 1 year CA showed that the fortified group had a 9-point higher psychomotor development index, which did not reach significance in this small study. (46)

Two of the 11 formula trials showed a significant positive effect on development of using a nutrient-enriched formula after discharge. Teller et al (25) reported that the RCT by Werkman and Carlson showed an advantage of PTF compared with TF in 104 preterm infants using the Fagan Test of Infant Intelligence. (25)(45) Marini et al, using the GMDS at 2 years, showed a significant benefit for the PTF group over the TF group because, despite the small sample, the effect size was large: an 8-point advantage. (43) Teller et al (25) reported a trend of higher GMDS scores in the 2 RCTs by Agosti et al (16)(44) and of higher KPS and BSID scores at 9 and 18 months, respectively, in the RCT by Lucas et al. (1) An 8-year follow-up by Ruys et al of the trial by Amesz et al did not detect cognitive or motor changes in those fed PDF versus TF but the sample size was small. (47)(48) A trial by Lucas et al (1) did not show significant differences between PDF versus TF in performance in English and mathematics examinations at 11 and 16 years of age in a novel linkage study using UK educational databases. (49) However, although the RCT by Cooke et al of PTF versus TF showed no difference in infant developmental scores between groups using the BSID, a recent 10-year follow-up by Embleton et al used the more sensitive WISC (abbreviated form). (15)(50) The original study had 3 groups: 2 groups compared PTF with TF until 6 months’ CA and a third crossover group switched from PTF to TF at term equivalent. (15) The 3 groups had 37, 37, and 18 subjects, respectively, and examined full-scale IQ and 4 subtests. One subtest score, Processing Speed Index (PSI) was 10 points higher in the crossover group than the TF group (P=.04) with a trend toward a 5-point higher PSI score in the PTF versus TF group.

Thus, despite the major issue of underpowering of trials in this field, and no meta-analyses, 7 of 13 individual postdischarge supplementation trials showed at least some indication of higher neurodevelopmental performance in those preterm infants fed enriched diets, which in 4 trials reached significance for the test or subtest used. No study showed an adverse effect. We cannot assume all interventions were optimal for promoting neurodevelopment, but we note that the trial by Marini et al, in which cognitive benefits were marked, the enriched-formula group had a high protein to energy ratio. (43) Hopefully, if more patients from these trials are followed, some standardization in outcome testing will facilitate meta-analyses.

RCTs and meta-analyses such as the ones described are used as the gold standard for underpinning clinical care. However, new postdischarge RCTs may seldom be initiated due to ethical constraints in randomly assigning infants to potentially suboptimal diets. Yet the paucity of evidence from meta-analyses of neurodevelopmental outcomes limits the interpretation of findings from these underpowered individual trials. Such limitations of RCT evidence are not new to neonatology: for instance, there is agreement that the mother’s own milk should be used for preterm infants, partly because of its association with enhanced neurodevelopment; however, not a single RCT has established this because randomization would be unethical. Similarly, postdischarge nutrition practice must also rely on evidence other than RCTs.

The best supporting observational evidence is the suggested association between early somatic or head growth, including growth in the postdischarge period, and better developmental scores over the long term. Ong et al reviewed 18 studies on weight gain and 15 on head growth in infants born preterm, which generally showed a positive relationship with neurodevelopment; studies have also shown that head size, shown here to be increased with postdischarge nutritional supplementation, relates to current and later cognitive performance. (8)(37)(38) Nevertheless, observational data must be examined critically for potential confounding because many factors may influence both growth and development in the infant population, independently resulting in spurious correlation. The relationship between growth and development requires statistical adjustment for confounding factors. Thus, in the 10-year follow-up study by Embleton et al, an important observation was that infant weight and head circumference gains were significantly related to full-scale IQ at 10 years, despite adjusting for several potential confounders. (50)

In summary, evidence from RCTs about the neurodevelopmental impact of postdischarge nutrition in preterm infants is unsatisfactory. However, with 7 of 13 trials showing at least some leaning toward higher developmental scores in the enriched postdischarge diet group, the evidence is consistent with the hypothesis that nutrient supplementation after discharge could favor neurodevelopment, given major underpowering of previous studies. However, observational data link faster somatic and head growth with neurodevelopmental benefit, and cognitive performance has been linked to early head size. Thus, the totality of evidence is somewhat more compelling. In any case, lack of equipoise could preclude future postdischarge RCTs, and reliance on non-RCT data is then inevitable, as with breastfeeding.

RCTs discussed here generally do not report adverse metabolic or tolerance issues with nutrient-enriched diets that would contraindicate their use in healthy growing former preterm infants. However, a potentially important safety matter deserves attention—the possibility that promoting rapid early growth could increase cardiovascular risk. The trade-off between promoting growth to enhance neurodevelopment and avoiding rapid growth to reduce cardiovascular risk is now a familiar concept. The postnatal growth acceleration hypothesis of Singhal and Lucas is that rapid postnatal growth increases the risk of cardiovascular disease. (9) This unifying hypothesis was rooted in evidence from programming studies in animals, epidemiologic studies in humans, and randomized trials in preterm and SGA infants. Thus, nutritional RCTs in preterm infants with 16-year follow-up showed early nutritional influences on key components of the metabolic syndrome, consistent with the growth acceleration hypothesis. (51)(52)(53)(54) RCTs in SGA term infants showed that feeding a nutrient-enriched diet to promote early growth increased later issues of blood pressure and body fat. (55) Despite many studies, there is some controversy in this area. (8)

However, the question is whether the postdischarge period is a critical one for adverse programming of the metabolic syndrome. The RCTs in preterm and SGA term infants (cited above) used an enriched diet to promote growth from close to birth onwards, and evidence suggests that the early weeks and months after birth are critical. However, when postdischarge nutrition commences, infants are generally several months old. A follow-up study of our postdischarge intervention trial at 5 to 7 years found that PDF did not increase later body fat or blood pressure, raising the hypothesis that after around 3 months postnatally sensitivity to early growth promotion in terms of cardiovascular risk may have diminished or disappeared. (56) An 8-year follow-up by Ruys et al of a postdischarge RCT of PDF versus TF initially conducted by Amesz et al, though relatively small, also failed to show that PDF led to an increase in any of numerous later metabolic and clinical factors that form part of different definitions of the metabolic syndrome. (47)(57) Thus, RCT evidence so far has not indicated that use of an enriched formula in the postdischarge period promoted cardiovascular risk in infants born preterm. In any case, there is a view that improved neurodevelopment would take precedence over cardiovascular risk. (11)

• A major limitation in the preterm infant postdischarge nutrition field is the incomplete evidence from RCTs that explore diet and neurodevelopment; underpowering of studies and difficulty in combining RCTs incorporating disparate outcomes are key obstacles.

• RCT evidence on the value of providing fortification to breastfed postdischarge preterm infants is scarce and conflicting. This is problematic since breastfeeding is the favored practice.

• The RCTs are principally focused on healthy, stable postdischarge infants born preterm. There are few trial data relating to care of the more high-risk infants who may have residual respiratory or gut disease or are tube fed and cannot regulate volume intake.

• The intervention periods of postdischarge RCTs range between 1 and 12 months but inadequate data define which is optimal.

• Maximum birthweight for trial eligibility has ranged from 1,500 to 1,850 g so, currently, the common use of nutrient-enriched postdischarge diets in infants born above 1,850 g has not been formally studied.

• Given few recent RCTs of postdischarge nutrition, whether modern improvements in predischarge nutrition might reduce the need for postdischarge nutrition interventions has received insufficient formal study; however, observational data suggest that even infants currently receiving care commonly exhibit extrauterine growth restriction at discharge. (58)(59)(60)

1. Evidence from RCTs shows that feeding enriched diets to preterm infants after discharge increases body weight usually including good quality lean tissue accretion, increases length, and has a significant effect on OFC, a measure of brain volume.

2. Evidence from underpowered RCTs for favorable dietary effects on cognitive development is limited; however, coupled with observational studies linking faster growth with developmental benefits, it seems prudent to ensure good postdischarge growth.

3. An increase in growth rate specifically in the postdischarge period has not so far been shown in RCTs to increase later cardiovascular risk in those born preterm.

4. An understanding of the physiology of the infant’s feeding volume adjustment in relation to energy intake has led to recognition of the value of postdischarge diets with a high protein to energy ratio in former preterm infants.

5. Varying response to nutritional intervention means that some infants will inevitably require special attention. However, although some units restrict nutritional intervention to infants with residual clinical problems or those considered high risk for poor growth infants, the evidence shows that significant growth promotion can be achieved in healthy, low-risk infants after discharge.

6. PDF or nutritional supplementation of human milk are used to address postdischarge undernutrition of preterm infants as evidenced by potentially remediable restriction of somatic and head growth. However, current evidence does not define a rigid intervention period for enhanced postdischarge nutrition, and practice varies between countries. One common practice in the United States is to use PDF or human milk fortification routinely in infants weighing less than 1,500 g at birth for 12 weeks and reassess growth. Some infants may require supplementation for substantially longer to achieve catch-up growth.

We are most grateful to Dr Steven Abrams and Professor Atul Singhal for reviewing our manuscript and to Professor Nicholas Embleton for his helpful advice.

BMC

bone mineral content

BSID

Bayley Scales of Infant Development

CA

corrected age

GMDS

Griffith Mental Development Scale

ICV

intracranial volume

KPS

Knobloch Passamenick and Sherrard test

OFC

occipitofrontal head circumference

PDF

postdischarge formula

PTF

preterm formula

RCT

randomized controlled trial

SGA

small for gestational age

TF

term formula

WISC

Wechsler Intelligence Scale for Children