Neonatal systemic inflammation and the risk of low scores on measures of reading and mathematics achievement at age 10 years among children born extremely preterm

Abstract

Background

Difficulties with reading and math occur more commonly among children born extremely preterm than among children born at term. Reasons for this are unclear.

Methods

We measured the concentrations of 27 inflammatory-related and neurotrophic/angiogenic proteins (angio-neurotrophic proteins) in multiple blood specimens collected a week apart during the first postnatal month from 660 children born before the 28^th week of gestation who at age 10 years had an IQ ≥ 70 and a Wechsler Individual Achievement Test 3^rd edition (WIAT-III) assessment. We identified four groups of children, those who had a Z-score ≤ −1 on the Word Reading assessment only, on the Numerical Operations assessment only, on both of these assessments, and on neither, which served as the referent group. We then modeled the risk of each learning limitation associated with a top quartile concentration of each protein, and with high and lower concentrations of multiple proteins.

Results

The protein profile of low reading scores was confined to the third and fourth postnatal weeks when increased risks were associated with high concentrations of IL-8 and ICAM-1 in the presence of low concentrations of angio-neurotrophic proteins. The profile of low math scores was very similar, except it did not include ICAM-1. In contrast, the profile of low scores on both assessments was present in each of the first four postnatal weeks. The increased risks associated with high concentrations of TNF-α in the first two weeks and of IL-8 and ICAM-1 in the next two weeks were modulated down by high concentrations of angio-neurotrophic proteins.

Conclusions

High concentrations of angio-neurotrophic proteins appear to reduce/moderate the risk of each learning limitation associated with systemic inflammation. The three categories of limitations have protein profiles with some similarities, and yet some differences, too.

1. Introduction

Compared to children born at term, those born very preterm are at increased risk of reading(1) and related limitations,(2) and math limitations.(3–7) These children are more likely than others to have structural and functional impairments of the brain.(8–12) We reasoned that if very preterm children who had systemic inflammation during the first postnatal month are more likely than others to have structural and/or functional abnormalities of the brain,(13–16) and children who have structural and/or functional abnormalities of the brain are more likely than others to have learning limitations,(8–12) then children who had systemic inflammation during the first postnatal month might be at increased risk of learning problems.

The “paucity of protectors” hypothesis offers one explanation for why extremely preterm newborns are at increased risk of limitations of brain development, and of brain damage. According to this hypothesis the mother and/or placenta provide the fetus with proteins needed for brain maturation, and birth before the newborn can synthesize adequate amounts of these proteins deprives the brain of needed brain-maturation enhancers.(17) These proteins with neurotrophic properties, which are often growth factors, are now recognized as having angiogenic properties,(18–20) and capable of reducing the risk of brain damage and/or promoting repair.(21, 22) Some of these proteins are viewed as mainly neurotrophic, while others are primarily considered angiogenic.(23, 24) Regardless, these proteins almost invariably have both neurotrophic and angiogenic properties, (24–32) prompting the names “angioneurins”(33–39) and “angioglioneurins.”(40) To emphasize the trophic properties of these proteins, we prefer the name “angio-neurotrophic protein.”

Data from the ELGAN (Extremely Low Gestational Age Newborn) Study provided an opportunity to explore this possibility because of the availability of the concentrations of 27 proteins with inflammation-related and/or neurotrophic/angiogenic properties in blood specimens obtained during the first postnatal month from children born before the 28^th week of gestation and assessments of their educational achievements at age 10 years with the Wechsler Individual Achievement Test, Third Edition (WIAT-III).

2. Methods

2.1 Participants

The ELGAN study is a multi-center prospective, observational study of the risk of structural and functional neurologic disorders in extremely preterm infants.(41) A total of 1506 infants born before the 28^th week of gestation were enrolled during the years 2002–2004 and 1198 survived to age 10 years. At age 10 years, 966 of these were recruited for an age-appropriate assessment of cognition, executive function, behaviors, and achievement, 889 (92%) returned for follow up, and 874 were administered the neurocognitive tests (Table 1). Because children who had early systemic inflammation are at heightened risk of cognitive impairment,(42) and children who are cognitively impaired do not do well on academic achievement tests, we wanted to restrict our search for any relationship between early systemic inflammation and learning limitations not attributed to global limitation to children who were not cognitively impaired. Thus, the sample for the analyses presented here is restricted to the 666 children who had a DAS-II verbal IQ ≥ 70 and non-verbal IQ ≥ 70. Enrollment and consent procedures for this follow up study were approved by the institutional review boards of all participating institutions.

Table 1.

Sample description

	Yes
Enrolled	1506
Survived to age 10 years	1198
Recruited for assessment at age 10 years	966
Returned for an assessment at age 10 years	889
Proteins measured in blood collected on 2 separate days	857
DAS-II verbal IQ ≥ 70 and non-verbal IQ ≥ 70	666
Word Reading and Numberical Operations Z-scores available	660
Word Reading Z-score ≤ −1 + Numerical Operations Z-score > −1	48
Numerical Operations Z-score ≤ −1 + Word Reading Z-score > −1	113
Word Reading Z-score ≤ −1 + Numerical Operations ≤ −1	69
Word Reading Z-score > −1 + Numerical Operations > −1	430

2.2 Procedures

All families who participated in the previous follow up were contacted by mail and then by phone to invite them to participate in the 10-year follow up. Lost to follow-up families were searched for on state vaccination registries, and other openly-available websites. Facebook was also used where approved by the local institution’s IRB.

Families willing to participate were scheduled for one visit during which all of the tests reported here were administered in 3 to 4 hours, including breaks. The assessments were selected to provide the most comprehensive information about neurocognitive and academic function in one testing session. While the child was tested, the parent or primary caregiver completed questionnaires regarding the child’s educational, medical, and neurological status and behavior, and completed the Kaufman Brief Intelligence Test – 2 (KBIT-2) nonverbal subscale.(43)

2.2.1 General cognitive ability

General cognitive ability (or IQ) was assessed with the School-Age Differential Ability Scales–II (DAS-II) Verbal and Nonverbal Reasoning scales.(44) We required that children have scores of 70 or higher on both scales to be included in our sample for these analyses.

2.2.2 Academic Function

The Wechsler Individual Achievement Test-III (WIAT-III) provides grade- and age-adjusted standard scores for the Word Reading and Numeric Operations subtests.(45) We defined each learning limitation as a Z-score ≤ −1 (i.e., below the 16^th centile, which is equivalent to a score ≤ 85) on a grade-based WIAT-III achievement test.(46) Thus, we identified four mutually-exclusive groups, reading limitation only (Word Reading Z-score ≤ −1, Numerical Operations Z-score > −1), math limitation only (Numerical Operations Z-score ≤ −1, Word Reading Z-score > −1), both reading and math limitations (Word Reading Z-score ≤ −1, Numerical Operations Z-score ≤ −1), and neither limitation (Word Reading Z-score > −1, Numerical Operations Z-score > −1) (Table 1).

Because the reading and math limitations occur more commonly than would be expected if they were independent of each other,(9, 47–51) we considered it prudent to view children with the combination as possibly having protein profiles that differed from those of children with either isolated limitation. Thus, we have three outcomes of interest, isolated reading limitation (i.e., not accompanied by a math limitation), isolated math limitation (i.e., not accompanied by a reading limitation), and the combination of reading and math limitations. Children with these entities were compared to children who had no reading or math limitation.

2.2.3 Systemic inflammation ascertainment procedures

Drops of blood were collected on filter paper on the first postnatal day (range: 1–3 days), the 7^th postnatal day (range: 5–8 days), the 14^th postnatal day (range: 12–15 days), the 21^st postnatal day (range: 19–23 days), and the 28^th postnatal day (range: 26–29). All blood was from the remainder of specimens obtained for clinical indications. Dried blood spots were stored at −70°C in sealed bags with a desiccant until processed. Details about the elution of proteins from the blood spots are provided elsewhere.(52)

Because some spots are thicker than others, we have normalized our measurements to mg of total protein in the eluent. The total protein concentration in each eluted sample was determined by BCA assay (Thermo Scientific, Rockford, IL) using a multi-label Victor 2 counter (Perkin Elmer, Boston, MA) and the measurements of each analyte normalized to mg total protein.

2.3 Proteins measured

The Genital Tract Biology Laboratory at the Brigham and Women’s Hospital in Boston Massachusetts eluted all blood spots as previously described and measured all proteins reported here. The laboratory used the Meso Scale Discovery to measure: C-Reactive Protein (CRP), SAA, Myloeperoxidase (MPO) Interleukin-1 β (IL-1β), IL-6, IL-6 Receptor (IL-6R), Tumor Necrosis Factor-α (TNF-α), TNF Receptor-1 (TNFR-1), TNFR-2, IL-8 (CXCL8), Regulated upon Activation, Normal T-cell Expressed, and Secreted (RANTES; CCL5), Intercellular Adhesion Molecule-1 (ICAM-1; CD54), Vascular Cell Adhesion Molecule-1 VCAM-1; CD106), Vascular Endothelial Growth Factor (VEGF), VEGF Receptor-1 (VEGFR-1, also known as sFLT-1), VEGFR-2 (KDR), Insulin-Like Growth Factor-1 (IGF) Binding Protein-1 (IGFBP-1), thyroid stimulating hormone (TSH), Metalloproteinase (MMP)-9, and erythropoietin (EPO).

A multiplex immunobead assay manufactured by R&D Systems (Minneapolis, MN) and a MAGPIX Luminex reader (R&D Systems) was used to measure angiopoietin-1 (Ang-1), Ang-2, placenta growth factor (PIGF), Neurotrophin-4 (NT-4), Brain Derived Neurotrophic Factor (BDNF), and basic Fibroblast Growth Factor (bFGF). ELISA (R&D Systems) was used to measure IGF-1.

One set of measurements was made in 2009–2010, and the second in 2015, and, while the distributions of each were similar, they were not identical. In addition, the protein concentrations varied with gestational age, and with the postnatal day of collection.(53, 54) Consequently, we stratify our sample into 30 groups defined by two measurement sets (2009–2010, 2015), three gestational age categories (23–24, 25–26, 27 weeks), and five postnatal days of blood collection (1, 7, 14, 21 and 28). Because we were interested in the contribution of high concentrations, and the concentrations of most proteins did not follow a normal distribution, the distribution of each protein’s concentration was dichotomized into the highest quartile and the lower three quartiles among children in each of the 30 groups.

2.4 Data analyses

Because we did not evaluate achievement scores in a full-term comparison group, we relied on comparisons to the WIAT-III normative sample.(45) We calculated Z-scores to allow for the differences in school grade at the time of the assessment. In our sample, the distribution of grade-adjusted scores on the WIAT-III Word Reading and Numerical Operations subtests had a larger number of children at the low end of the distribution than expected, prompting us to focus on those who had scores one or more standard deviations below the expected mean (Z-score ≤ −1).

We tested the null hypotheses that among children who have both DAS-II Verbal and Non-verbal IQ scores ≥ 70, elevated protein concentrations (defined as in the top quartile) are not associated with risk of three learning limitations, reading only (Word Reading Z-scores ≤ −1 with Numerical Operations Z-scores > −1), math only (Numerical Operations Z-scores ≤ −1 with Word Reading Z-scores > −1), and both reading and math limitations (Z-scores ≤ −1 on both Word Reading and Numerical Operations assessments). Thus, in essence, we tested two hypotheses for each dysfunction. One is that elevated concentrations of inflammation-related proteins are associated with increased risk, while the other is that elevated concentrations of angio-neurotrophic proteins are associated with reduced risk.

1. Individual days: single proteins

We evaluated the hypothesis that children who had a top quartile concentration of each protein on each day were not at greater risk of a reading limitation, and separately of a mathematics limitation, than children who had a lower concentration of each protein on each of five days during the first postnatal month (Appendix Tables Reading-A, Math-A, and Reading-and-Math-A, which are summarized in Table 2).

Table 2.

Statistically-significant odds ratios for Z-scores ≤ −1 at age 10 years on the reading (R) assessment only, the mathematics (M) assessment only, or the co-occurrence of low scores on both (B), associated with a concentration in the top quartile of the protein listed on the left, among children whose DAS-II verbal and nonverbal IQs were ≥ 70. This table summarizes Appendix Tables A (individual days) and B (epochs) for these entities. Upper-case regular typeface indicates statistically significantly increased odds ratios, while lower-case italic typeface indicates statistically significantly reduced odds ratios (p < 0.05). Adjustment has been made for birth weight Z-score < −1 and mother’s K-BIT Z-score < −1.

	Protein in highest quartile on: individual day					Two days of epoch
	Day 1	Day 7	Day 14	Day 21	Day 28	Early	Late
Proteins traditionally viewed as inflammation-associated
CRP			M, B	M, B
SAA			M
MPO					M		R, M
IL-1β			B
IL-6					R
IL-6R	R
TNF-α			M
TNF-R1			M
TNF-R2			M	M, B		M
IL-8				M	M		M
RANTES					b		b
ICAM-1	m		B	B			B
VCAM-1
Proteins traditionally viewed as neurotrophic and/or angiogenic
TSH
EPO						B
NT-4		b
BDNF			m		b
bFGF	m
IGF-1
IGFBP-1	M	R		B
VEGF		m
VEGF-R1
VEGF-R2			M
PIGF
Ang-1		b		b
Ang-2

2. Early epoch: single proteins

We evaluated the hypotheses that children who had a top quartile concentration of each protein on at least two days during the first two postnatal weeks (when specimens were obtained on days 1, 7, and 14) were not at greater risk of a reading limitation, and separately of a mathematics limitation, than children who had lower concentrations (Appendix Tables Reading-B, Math-B, and Reading-and-Math-B, which are also summarized in Table 2).

3. Late epoch: single proteins

Similarly, we compared children who had top quartile concentrations of each protein on both days 21 and 28 (the late epoch) to their peers with lower concentrations of each protein (Appendix Tables Reading-B, Math-B, and Reading-and-Math-B, which are also summarized in Table 2).

4. Early epoch: Combinations of inflammation-related and angio-neurotrophic proteins

We evaluated the hypotheses that children who had top quartile concentrations of one or both of two proteins (one an inflammation-related protein and the other an angio-neurotrophic protein) on at least two days during the first two postnatal weeks (i.e., the early epoch) were not at higher or lower risk of each learning limitation than children who had lower concentrations of both proteins (Appendix Tables Reading-C, Math-C, and Reading-and-Math-C, and summarized in Table 3).

Table 3.

This is a summary of all 3 Appendix Tables C (which evaluate the risk of a Z-score ≤ −1 on the WIAT-III academic achievement assessment in light of top quartile concentrations of two proteins at a time for the early epoch). An inflammation-related protein is identified at the top of each set of columns, and the angio-neurotrophic proteins are identified on the left. A “V” in the cell indicates that a top quartile concentration of the angio-neurotrophic protein (listed on the left) appears to modulate the increased risk of the learning limitation (identified at the top of each column) associated with a top quartile concentration of the inflammation-related protein that heads each set of columns.

	TNF-α			IL-8			ICAM-1
	Read	Math	Both	Read	Math	Both	Read	Math	Both
IL-6R			V
MMP-9			V
RANTES
EPO
NT-4
BDNF
bFGF			V					V
IGF-1
VEGF			V				V	V
VEGF-R1		V	V					V
VEGF-R2			V
PIGF			V					V
Ang-1								V
Ang-2								V

5. Late epoch: Combinations of inflammation-related and angio-neurotrophic proteins

Similarly, we evaluated the hypotheses that children who had top quartile concentrations of one or both of two proteins (one an inflammation-related protein and the other an angio-neurotrophic protein) on two days a week apart during the third and fourth postnatal weeks (i.e., the late epoch) were not at higher or lower risk of each learning limitation than children who had lower concentrations of both proteins (Appendix Tables Reading-D, Math-D, and Reading-and-Math-D, and summarized in Table 4).

Table 4.

This is a summary of all 3 Appendix Tables D (which evaluate the risk of a Z-score ≤ −1 on the WIAT-III academic achievement assessment in light of top quartile concentrations of two proteins at a time for the late epoch). The inflammation-related protein is identified at the top of each set of columns, and the angio-neurotrophic proteins are identified on the left. A “V” in the cell indicates that a top quartile concentration of the angio-neurotrophic proteins (listed on the left) appeared to modulate the increased risk of the learning limitation (identified at the top of each column) associated with a top quartile concentration of the inflammation-related protein that heads each set of columns. The small ‘v’s in other cells indicate the same pattern of modulated risk, but with too few children to achieve statistical significance.

	TNF-α			IL-8			ICAM-1
	Read	Math	Both	Read	Math	Both	Read	Math	Both
IL-6R				v					V
MMP-9				v			v		V
RANTES				v	V
EPO				v	V				V
NT-4				v	V	V	v		V
BDNF				v					V
bFGF				v
IGF-1							v		V
VEGF				v			v
VEGF-R1				v			v
VEGF-R2					V
PIGF				v	V		v		V
Ang-1									V
Ang-2		V				V	v		V

We created multinomial logistic regression models that allowed us to calculate odds ratios and 95% confidence intervals of the risk of each LL among children who had a top quartile concentration of each protein to the risk among children who had a concentration below the 75^th centile. In separate multinomial logistic regression models, we evaluated the risk of each LL associated with top quartile concentrations of two proteins simultaneously relative to the risk among children who had lower concentrations of both proteins. The indicator that an angio-neurotrophic protein appears to have protective effects is a statistically significant increased risk of the LL among children whose blood had a top quartile concentration of the inflammation-related protein and a lower concentration of the angio-neurotroph, while the risk was not increased among children whose blood had a top quartile concentration of the inflammation-related protein and a top quartile concentration of the angio-neurotroph.

In the larger sample, children who had top quartile concentrations of inflammation-related proteins were no more likely than their peers who had lower concentrations to have a mother who had limited educational achievement, a low KBIT-2, or was eligible for government-provided medical care insurance (Medicaid). Thus, confounding by social class is not an issue. Similarly, other potential confounders such as the child’s sex were also not associated with systemic inflammation. On the other hand, low birth weight Z-score, indicative of fetal growth restriction, is associated in this sample with elevated concentrations of some proteins,(55, 56) and with low Word Reading scores.(57) Thus, low birth weight Z-score is a potential confounder. Low maternal education achievement (defined as a KBIT-2 < 85) is not associated with protein concentration,(56) and is therefore not a confounder. We have adjusted for this variable, however, because we consider it the most likely effect modifier.

3. Results

The details of individual learning limitations are provided in the appendix along with the relevant tables. Here, we present an overview and identify commonalities and differences among the three learning limitation groups (reading only, math only, both reading and math).

3.1 Sample (Table 1)

Of the 660 children for whom 2 blood specimens were available, who had both DAS-II verbal and non-verbal IQ scores of IQ ≥ 70, and had a WIAT-III assessment, 48 had a Z-score ≤ −1 on the Word Reading component only, 113 had a Z-score ≤ −1 on the Numerical Operations component only, and 69 had a Z-score ≤ −1 on both the Word Reading and Numerical Operations assessments.

3.2 Individual proteins on individual days and on 2 days of each epoch (Table 2)

Of all the cells in Table 2, only the one for top quartile concentrations of MPO on both days of the late epoch is identified with increased risks of both the reading only and math only limitations. Increased risks of the math only and the reading+math (identified in the table with a B for both) limitations are associated with top quartile concentrations of CRP on days 14 and 21, and TNF-R2 on day 21. Increased risks of math only limitations are associated with top quartile concentrations of TNF-R2 on days 14 and 21 and during the early epoch, and IL-8 on days 21 and 28 and during the late epoch. Increased risks of reading+math limitations are associated with top quartile concentrations of ICAM-1 on days 14 and 21 and during the late epoch.

ICAM-1 is the only inflammation-related protein whose top quartile concentration is associated with reduced risk of a learning limitation (math only on day 1). In contrast, high concentrations of multiple angio-neurotrophic proteins were associated with reduced risk of math only limitations (including BDNF on day 14, bFGF on day 1, and VEGF on day 7) and both reading+math limitations (including RANTES on day 28 and during the late epoch, NT-4 on day 7, BDNF on day 28, and Ang-1 on days 7 and 21). We created Table 2 to help us identify the inflammation-related and AN-trophic proteins that might be most suitable for examining in light of each other. Because we decided that Tables 3 and 4 should have only 3 inflammation-related proteins these tables do not include CRP, TNF-R2, SAA, MPO, IL-1β, IL-6, TNF-R1, TNF-R2, and VCAM-1. We did retain rows for IL-6R, RANTES, and MMP-9 because each has what appear to be inflammation-modulating capabilities in the ELGAN Study population.

3.3 Sets of proteins during the early epoch (Table 3)

The increased risk of both the math limitation associated with top quartile concentrations of TNF-α was modulated by top quartile concentrations of VEGF-R1 only. The increased risk of reading+math limitations together (but not of either individually) associated with top quartile concentrations of TNF-α was modulated by top quartile concentrations of IL-6R, MMP-9, bFGF, VEGF, VEGF-R1, VEGF-R2, and PIGF.

3.4 Sets of proteins during the late epoch (Table 4)

The increased risk of math only limitations associated with top quartile concentrations of TNF-α was modulated by top quartile concentrations of Ang-2, while the increased risk associated with top quartile concentrations of IL-8 was modulated by top quartile concentrations of RANTES, EPO, NT-4, VEGF-R2, and PIGF. The increased risk of both reading+math limitations together (but not of either individually) associated with top quartile concentrations of ICAM-1 was moderated by top quartile concentrations of IL-6R, MMP-9, EPO, NT-4, BDNF, IGF-1, PIGF, Ang-1, and Ang-2.

Because of the small number of children who had the reading limitation only, the modulation of increased risks associated with top quartile concentrations of IL-8 and ICAM-1 by angio-neurotrophic proteins never achieved statistical significance. In Table 4, we identify with a lower case “v” the many risk ratios of 2.0 or greater in Appendix Table 3b cells characterized by top quartile concentrations of IL-8 or ICAM-1 and lower concentrations of the angio-neurotrophic protein.

4. Discussion

4.1 Our strategy

Because systemic inflammation appears to increase the risk of perinatal brain damage,(13–16) and angio-neurotrophic proteins have the capacity to minimize brain damage,(21, 58–64) we evaluated the contribution of high (which we have operationally-defined as top-quartile) concentrations of inflammation-related proteins in light of the presence or absence of high concentrations of angio-neurotrophic proteins. In essence, our ability to identify increased risk associated with an inflammation-related protein is expected to be diminished in the setting of relative abundance of an angio-neurotrophic protein. Similarly, an angio-neurotrophic protein cannot be so identified in a setting that lacks the potential for increased risk.

4.2 What we found

Our main findings are that in the extremely preterm newborn, high blood concentrations of inflammation-related proteins are associated with increased risk of mathematics limitations and to a lesser extent, reading limitations, and that angio-neurotrophic proteins appear to modulate the increased risks of learning limitations potentially attributable to systemic inflammation. Some of our findings suggest that reading and math limitations have distinguishable protein profiles. For example, angio-neurotrophic proteins appears able to modulate the increased risks of reading limitations and combined reading and math limitations associated with high concentrations of IL-8 and ICAM-1 during the late epoch, while (with the single exception of Ang-2) isolated math limitations do not share this profile. Many of these same angio-neurotrophic proteins (i.e., VEGF, VEGF-R1, VEGF-R2, Ang-1, and Ang-2) appear to modulate the risks of the combined reading and math limitations associated with high concentrations of TNF-α during the early epoch.

4.3 What others found. How our findings are similar or different

We are not aware of any report that assessed the contribution of systemic inflammation in the extremely preterm newborn to the occurrence of reading and math limitations at school age. However, children who survived sepsis-associated encephalopathy perform less well after than they did previously.(65) Prolonged neonatal medical treatment, especially with mechanical ventilation, which appears to increase the risk of systemic inflammation,(66) has been associated with specific impairments in mathematic tasks.(67) Also, studies that provoked inflammation in adult rodents,(68–70) and reports about adults with systemic inflammation,(71, 72) support the view that concurrent or very recent inflammation impairs learning.(73)

4.4 Why focus on top-quartile concentrations

In this sample, the concentrations of most proteins do not follow a normal distribution.(53) This prompted us to avoid calculating means and standard deviations. Rather, we evaluated the risks of indicators of brain damage among four quartiles of protein concentrations.(74, 75) Rarely did the risks increase with increasing protein concentration quartile. Rather, the highest risks tended to be found among children whose protein concentrations were in the top quartile.

4.5 Why focus on top-quartile concentrations on multiple days

Systemic inflammation on two occasions a week apart during the first two postnatal weeks was associated in this sample with ventriculomegaly when the newborn was in the intensive care nursery,(74) and at age 2 years with low Bayley Scales of Infant Development-II,(76) an attention problem,(77) cerebral palsy,(78) and microcephaly.(75) When reporting these findings we were not sure if the systemic inflammation was intermittent or sustained.(13) We now have more confidence that some of the systemic inflammation is sustained.(56)

The main drivers of postnatal systemic inflammation in this sample are fetal growth restriction,(79) mechanical ventilation,(66) bacteremia,(80) and necrotizing enterocolitis.(81) Perhaps efforts to reduce the duration of mechanical ventilation and minimize presumed barotrauma will result in reduced brain damage.(82)

4.6 Categorization of outcome

The dimensional view of reading abilities holds that no sharp break (or discontinuity) separates those with a reading limitation and their peers who are better readers.(83) The same can be said for mathematics abilities.

Epidemiologists, fully aware that measures of a dysfunction represent a continuum, prefer to establish a cut-off for their efforts to identify the antecedents associated with disease entities defined by dichotomizing a continuous measure of function/dysfunction (e.g., hypertension, diabetes mellitus, glaucoma).(84–86) In keeping with this approach, we dichotomized Z-scores at ≤ −1 on the WIAT-III Word Reading and the Numerical Operations subtests.

4.7 Co-occurrence of reading and math limitations (comorbidity)

Reading and math limitations tend to occur together more commonly than can be explained by chance.(47, 48) Multiple explanations have been offered for the relatively high rate of the co-occurrence of mathematics and verbal/spelling/reading limitations, with most current explanations invoking shared circuits/networks,(9, 49–51)

We report here that all three learning limitations we examined are characterized by increased risks associated with systemic inflammation, and reduction of these risks when angio-neurotrophic proteins are abundant. The prominence of late epoch high concentrations of ICAM-1 with reading limitations, but not with math limitations raise the possibility that these two disorders differ in the inflammation pathways that influence their risk of occurrence.

4.8 Surrogates and developmental regulation

We view each protein as a surrogate for all proteins in its class. For example, although we found inflammatory signals for TNF-α, IL-8, and ICAM-1, we prefer not to focus on these specific proteins, but rather on broader inflammatory processes.(87, 88) Similarly, we view each angio-neurotrophic protein as a surrogate for many other proteins that possess neurotrophic and angiogenic characteristics.

The concept that each protein is only a surrogate for others is based, in part, on our limited understanding of developmental regulation. Not all inflammation-related proteins have the same pattern of developmental regulation,(89, 90) nor do all angio-neurotrophic proteins have the same developmental regulation schedule.(91) In general, however, the more “mature” the newborn (i.e., the further along the developmental regulation schedule) the more s/he should be able to resolve inflammation, and promote brain growth, repair, and well-being.(92)

4.9 Limitations and strengths of this study

Perhaps our main weakness is the relatively small number of children with a reading only limitation. Because of this, odds ratios in the 3 to 8 range were not statistically significant.

We are also limited by the interrelatedness of early and late systemic inflammation, which limits our ability to tease apart the contributions of each to the occurrence of each learning achievement limitation. We are also limited by the relatively small number of proteins measured. Inflammation is a broad and complex phenomenon,(93, 94) and we have assessed only a very small part of it.

We relied on blood specimens obtained for clinical indications. As their physiology became more stable, infants were less likely than their sicker peers to have blood drawn on days 14, 21, and 28. Consequently, selection bias probably occurred to some extent as the number of children for whom we had specimens from both of the late-epoch days tended to be about 2/3 of the number for whom we had specimens from two of the three early-epoch days.

Not only did the clinical indications for specimens increase the potential for bias, but it also diminished power. Only 305 children provided specimens for both days 21 and 28 (i.e., the late-epoch). This can be expected to provide 20 children in the cell for top-quartile concentrations of both proteins (305 × .25 × .25). With only 8% of children having the combined reading and math limitation, only 1 or 2 of the 20 children in this cell would have this learning limitation. Among the consequences of a small number of subjects are very wide confidence intervals, occasional empty cells, and of course, very limited power to detect a difference between the odds ratio associated with high concentrations of both the inflammation-related protein and the angio-neurotrophic protein (left upper cell) and the odds ratio associated with high concentrations of the inflammation-related and low concentrations of the angio-neurotrophic protein (left lower cell).

This study has sufficient power to identify an odds ratio of 1.9 for the math limitation in the individual analyses and 2.0 in the two-protein analyses. Other strengths are the selection of infants based on gestational age, not birth weight,(95) prospective collection of all data, modest attrition, and finally, protein data of high quality,(19,20) and high content validity.(53, 96–98)

5. Conclusions

In this cohort of children born extremely preterm and at high-risk group of learning problems, high concentrations of inflammation-related proteins during the first postnatal month are associated with increased risks of reading and math limitations, which appear to be modulated by high concentrations of angio-neurotrophic proteins.

Highlights.

• Extremely preterm children are at increased risk of early systemic inflammation

• Those who have early systemic inflammation are at high risk of learning limitations

• Abundance of neurotrophins modulates this heightened risk

Abbreviations

Ang-1

Angiopoietin-1

Ang-2

Angiopoietin-2

BDNF

Brain-Derived Neurotrophic Factor

bFGF

basic Fibroblast Growth Factor

CRP

C-Reactive Protein

DAS-II

Differential Ability Scales–II

EPO

Erythropoietin

ICAM-1

Intercellular Adhesion Molecule-1

IGF-1

Insulin-like growth factor-1

IGFBP-1

Insulin-like growth factor binding protein-1

IL-1β

Interleukin-1 β

IL-6

Interleukin-6

IL-6R

Interleukin-6 Receptor

IL-8

Interleukin-8

KBIT-2

Kaufman Brief Intelligence Test – 2

LL

Learning limitation

MMP-9

Matrix Metalloproteinase-9

MPO

Myeloperoxidase

NT-4

Neurotrophin-4

PIGF

Placenta Growth Factor

RANTES

Regulated upon Activation, Normal T-cell Expressed, and Secreted

SAA

Serum Amyloid A

TNF-R1

Tumor Necrosis Factor-α Receptor-1

TNF-R2

Tumor Necrosis Factor-α Receptor-2

TNF-α

Tumor Necrosis Factor-α

TSH

Thyroid-Stimulating Hormone

VCAM-1

Vascular Cell Adhesion Molecule-1

VEGF

Vascular endothelial growth factor

VEGF-R1

Vascular endothelial growth factor Receptor-1

VEGF-R2

Vascular endothelial growth factor Receptor-2

WIAT-III

Wechsler Individual Achievement Test-III