Analyses of this population-based cohort of extremely preterm born infants did not show statistical evidence that either growth between birth and discharge, or growth between birth and the age of two years were associated with 2-year neurodevelopmental outcomes.
The association between postnatal growth and neurodevelopment in preterm infants has been studied repeatedly over the last decades [10; 28]. However, considerable heterogeneity exists in the analyzed patient collective, sample size, primary outcomes, assessment measures and adjustment for confounding variables. For example, included patients were categorized as small for gestational age versus adequate for gestational age and growth was categorized according to quartiles of the normative values of the patient collective . In both cases, analyses are less precise than analyzing growth parameters as continuous variables as performed in the present study. Furthermore, several studies measured growth as weight gain in gram/kg/d, which is not as precise as analyses of z-scores. Other studies did not adjust for confounding variables, had small sample size or did not focus on extremely preterm infants as our study. Thus, only a small number of studies analyzed the association between postnatal growth and neurodevelopment in extremely preterm infants precisely in a large collective.
Overall, the majority of publications imply that an association between postnatal growth and neurodevelopment exists, but some important limitations are present in most of them. In particular only few studies adjusted development for socio-economic status, which is an important predictor of neurodevelopment .
The French EPIPAGE study documented increased risk of cognitive impairment and inattention-hyperactivity at the age of five years in preterm infants < 32 weeks who were born small for gestational age. Furthermore, impaired postnatal growth of patients with appropriate for gestational age birthweight was associated with cerebral palsy and school difficulties . In contrast to our study, patients of higher gestational age were included and growth was assessed only at the age of six months. Moreover, eight outcome parameters were presented and no correction for multiple testing was applied.
Similar to our study Belfort et al. assessed the relationship between growth and neurocognitive development in a large American collective of ELGAN . While presenting multiple analyses, lower weight gain was not associated with higher or lower risk of low MDI, low PDI, cerebral palsy or microcephaly. Significant associations with neurodevelopment were only detected in subsets of infants (weight z-score <-2 at the age of 12 month, not considering growth).
Based on the above-mentioned literature, it is difficult to draw strong conclusions about the impact of postnatal growth on neurodevelopment in the extreme preterm population. In light of the present findings it seems that postnatal growth is not a predictor of neurodevelopment as described previously.
While z-scores for all four growth parameters declined between birth and hospital discharge, length and body weight showed a catch-up growth at FU2. Only head circumference showed a z-score below − 0.4 at FU2. Since patients with catch-up growth might have a better outcome than patients without catch-up growth, the growth rate might be an explanation for our results.
It is important to state that this study did not include intrauterine growth as a risk factor, although extensive studies have documented that intrauterine growth restriction is associated with impaired development [29–32].
Over the last decades, a major aim in neonatology was to improve growth of preterm infants to enable optimal development. However, results of our study question if focusing on weight gain improves neurodevelopment of extremely preterm infants. Moreover, a large number of studies have documented a positive association between postnatal weight gain and adiposity, insulin resistance as well as increased blood pressure [28; 33; 34]. Therefore, potential positive effects of increased weight gain have to be balanced against existing risks. In fact, current nutritional strategies do not aim to only improve growth, but to optimize the quantity and quality of the intake, to increase breast milk consumption, while trying to reproduce body composition resembling that of term infants as much as possible.
Important questions about optimal growth remain unanswered. In particular, nutritional status in preterm infants is very complex and measuring growth in grams and centimeters describes growth only quantitatively, but not qualitatively. Assessment of lean body mass or supply of micronutrients might give additional information about optimal growth.
The following weaknesses limit the generalizability of the present findings. First, the retrospective design of the study implies a reporting bias. Since patients who were assessed at FU2 were slightly sicker (see supl. Table 1) than patients who were not assessed, a bias seems possible.
Second, the study includes patients born 2006–2012 with FU2 not later than 2014, which limits generalizability of our results. Despite advances in neonatal care several studies show that neither mortality, nor the rate of neurodevelopment has improved substantially over the last decades. In fact, a recent meta-analysis concluded that no definite trend of improved neurodevelopment at school age for neurosensory, cognitive, academic achievement, motor or executive function exists[35; 36]. However, some studies suggest that the spectrum of NDI shifted towards less severe CP and less severe sensory impairment. Nevertheless, we speculate that more recent data would provide results comparable to our study.
Third, multiple testing required a Bonferroni correction. While maintaining the overall type I error at 5%, this also had the consequence of increasing the type II error, such that we could have missed a few significant results. In particular, the adjusted association between delta2 weight z-score and NDI was potentially clinically impactful, reaching a p-value of 0.0069. This would have been significant if we had chosen to investigate less associations. Similar remarks hold for the adjusted association between delta2 weight, respectively delta 2 BMI with PDI. It will be interesting to see whether such potential associations can be confirmed in future studies.
Fourth, postnatal development was assessed with three different tools. While the composite outcome NDI included the results of all three developmental assessment methods, we included a secondary analysis considering only BSID-II data focusing on the largest subgroup of the cohort studied.
The strengths of the present study include the large size of the cohort, the utilization of standardized neurodevelopmental measures, and the prospective nature of the dataset regarding a geographically defined population.