Browse the corpus

Homocystinuria caused by cystathionine beta synthase (CBS) deficiency is a rare autosomal recessive disorder of sulfur amino acid metabolism. The disease manifests in childhood by vascular, neurologic, and connective tissue abnormalities,1,2 although increasing numbers of mildly affected adults with only thromboembolism are reported.3-6 Various approaches have been used to examine the population frequency of homocystinuria, but the true incidence of the disease is unknown. Newborn screening, based on finding individuals with elevated blood methionine concentrations, has apparently been detected almost exclusively patients with severe pyridoxine nonresponsive form of disease7 and with a worldwide incidence of 1:65 000-1:900 000.8 However, estimates calculated from the heterozygote frequency for the most common mutation c.833T>C (p.I278T) in newborns suggested an incidence of at least one order of magnitude higher.9-11 In agreement with these findings Refsum et al12 reported even higher population frequency of another CBS mutation c.1105 C>T, which was present in 0.8% of CBS chromosomes among unselected Norwegian newborns. The calculated incidence of homocystinuria caused by 6 mutations including the c.1105 C>T may be thus unusually high with 1 patient with CBS deficiency expected in each of 6400 Norwegian newborns.12

ncy of another CBS mutation c.1105 C>T, which was present in 0.8% of CBS chromosomes among unselected Norwegian newborns. The calculated incidence of homocystinuria caused by 6 mutations including the c.1105 C>T may be thus unusually high with 1 patient with CBS deficiency expected in each of 6400 Norwegian newborns.12 The c.1105 C>T mutation is localized in a CpG dinucleotide in exon 10 and leads to replacement of arginine in position 369 of the CBS polypeptide by cysteine (p.R369C). The pathogenicity of this mutation is unclear. The c.1105 C>T was originally described in 3 unrelated pyridoxine-responsive patients of Norwegian,14 Dutch,13 and Anglo-Celtic4 origin, which strongly supports the notion of its pathogenicity (Table I). In contrast, Kim et al14 showed in a yeast system that expression of the p.R369C variant ensures normal to above-normal growth of CBS-deficient strain, suggesting none or at most a mild effect of p.R369C on the enzyme activity. High frequency of the c.1105 C>T allele and its putative pathogenicity may have important epidemiologic consequences, such as for neonatal screening of homocystinuria. Our study was therefore aimed at (1) determining the frequency of this variant in another Caucasian population and (2) examining the pathogenicity of the p.R369C mutant by expression studies in prokaryotic and eukaryotic systems.

ve important epidemiologic consequences, such as for neonatal screening of homocystinuria. Our study was therefore aimed at (1) determining the frequency of this variant in another Caucasian population and (2) examining the pathogenicity of the p.R369C mutant by expression studies in prokaryotic and eukaryotic systems. Methods Samples We used 300 anonymous peripheral blood spots from Prague and 300 archived umbilical cord blood samples from Brno representing Bohemian and Moravian region, respectively, from a study that was described previously.10 Genomic DNA was isolated with the QIAamp DNA Mini Kit (Qiagen, Valencia, California). This study was approved by the Ethics Committee of Charles University in Prague—First Faculty of Medicine. Genotyping To detect the c.1105C>T alleles, we used polymerase chain reaction (PCR)-RFLP technique to amplify exon 10 from genomic DNA using specific primers—forward: 5′-CAgTgCCCACCCCAgCTCATTA-3′ and reverse: 5′-ggCCTCCTCCCCTCCCAgTTCT-3′ and Klentaq polymerase (GeneAge Technologies, Praha, Czech Republic).

Methods Samples We used 300 anonymous peripheral blood spots from Prague and 300 archived umbilical cord blood samples from Brno representing Bohemian and Moravian region, respectively, from a study that was described previously.10 Genomic DNA was isolated with the QIAamp DNA Mini Kit (Qiagen, Valencia, California). This study was approved by the Ethics Committee of Charles University in Prague—First Faculty of Medicine. Genotyping To detect the c.1105C>T alleles, we used polymerase chain reaction (PCR)-RFLP technique to amplify exon 10 from genomic DNA using specific primers—forward: 5′-CAgTgCCCACCCCAgCTCATTA-3′ and reverse: 5′-ggCCTCCTCCCCTCCCAgTTCT-3′ and Klentaq polymerase (GeneAge Technologies, Praha, Czech Republic). Kim et al14 used HaeII for PCR-RFLP analysis of c.1105 C>T (p.R369C); however, the loss of restriction site in their assay may be also caused by the mutation c.1106 G>A (p.R369H). To explore the possibility that either c.1105 C>T or c.1106 G>A was present in Czech newborns, the PCR products were first digested with HhaI to detect these 2 mutant alleles. The HhaI-positive samples were further digested with Cac8I and DraIII to check for the presence of c.1105 C>T and c.1106 G>A mutation, respectively. The latter mutation was not detected in the studied cohort.

was present in Czech newborns, the PCR products were first digested with HhaI to detect these 2 mutant alleles. The HhaI-positive samples were further digested with Cac8I and DraIII to check for the presence of c.1105 C>T and c.1106 G>A mutation, respectively. The latter mutation was not detected in the studied cohort. Expression Constructs Both the wild-type CBS and the mutant c.1105C>T constructs were derived from pHCS3 expression plasmid.15 The mutation was introduced into the wild-type expression plasmid with GeneTailor Site–directed mutagenesis kit (Invitrogen, Carlsbad, California) on the basis of manufacturer procedure. The sequences of mutagenic primers for c.1105 C>T were as follow—sense: 5′-gCTgCgTggTCATTCTgCCCgACTCagTgC-3′ and antisense: 5′-gCAgAATgACCACgCAgCACTggCCCTCCT-3′. For the purpose of eukaryotic expression both the wild-type and c.1105 C>T containing CBS were cloned into pTRE2hyg vector (Clontech Laboratories, Mountain View, California). The presence of c.1105 C>T and the absence of additional mutations in the entire coding CBS sequence of the pHCS3 and pTRE2hyg constructs were verified by dideoxy sequencing with ALF sequencer (Amersham Pharmacia Biotech, Piscataway, New Jersey).

into pTRE2hyg vector (Clontech Laboratories, Mountain View, California). The presence of c.1105 C>T and the absence of additional mutations in the entire coding CBS sequence of the pHCS3 and pTRE2hyg constructs were verified by dideoxy sequencing with ALF sequencer (Amersham Pharmacia Biotech, Piscataway, New Jersey). Prokaryotic Expression Both the wild-type and p.R369C CBS enzymes were expressed in DH5alpha cells (Gibco BRL, Carlsbad, California). The cells were grown at 37° C or 18° C, respectively, in Super Optimal Broth (SOB) media supplemented with ampicillin 100 μg/mL. After the OD600 of cultures reached ∼0.5, the cells were induced with isopropyl β-D-1-thiogalactopyranoside (final concentration 0.33 mmol/L) and grown for an additional 3 hours or overnight at 37° C or 18° C, respectively. The pKK vector without CBS insert was used as a negative control. Lysates from Escherichia coli cell pellets were prepared by sonication in Tris-Cl buffer with addition of Protease Inhibitor Cocktail suitable for use with bacterial cell extracts (Sigma Aldrich, St. Louis, Missouri).15 Concentration of protein in lysates was determined as described by Lowry et al16 with bovine serum albumin used as a standard. All expressions were done in triplicate, and the numbers shown in this article represent means.

or Cocktail suitable for use with bacterial cell extracts (Sigma Aldrich, St. Louis, Missouri).15 Concentration of protein in lysates was determined as described by Lowry et al16 with bovine serum albumin used as a standard. All expressions were done in triplicate, and the numbers shown in this article represent means. Eukaryotic Expression in CHO-K1 Cells Because the CHO-K1 cells do not express endogenous CBS at significant levels, they are a suitable system for studying CBS mutants. Two 75-cm2 flasks containing Chinese hamster ovary–derived cells that express the reverse tetracycline-controlled transactivator (BD Biosciences, San Jose, California) were transiently transfected with pTRE-2hyg plasmid bearing either the wild-type or mutant human CBS sequences with Lipofectamine-2000 (Invitrogen, Carlsbad, California). The CBS expression was induced 4 hours after transfection with doxycycline (final concentration 200 ng/mL). About 20 hours after transfection the cells were harvested mechanically and pooled, cell extracts were prepared by osmotic lysis with phosphate buffer 30 mmol/L with PEE-W1 detergent,17 and protein concentration was determined as described above. All expressions were done in triplicate, and means are shown in the article. For kinetic studies large-scale expression was carried out, and higher variability in the amounts of mutant tetramers in lysates was observed (data not shown).

th PEE-W1 detergent,17 and protein concentration was determined as described above. All expressions were done in triplicate, and means are shown in the article. For kinetic studies large-scale expression was carried out, and higher variability in the amounts of mutant tetramers in lysates was observed (data not shown). CBS Activity and Kinetics Measurement The CBS activity was assayed at homocysteine 10 mmol/L and serine 10 mmol/L by the previously published procedure15 with incubation at 37° C for 1 hour for E. coli lysates and 4 hours for CHO lysates with the following modification: mixture of unlabeled and 14C-labeled serine was replaced by 2H-labeled serine 10 mmol/L (Cambridge Isotope Laboratories, Andover, Massachusetts) and the amount of 2H-labeled cystathionine produced was determined by LC-MS-MS with a commercially available kit for amino acid analysis (EZ:faast; Phenomenex, Torrance, California) (Krijt et al, unpublished data). The kinetic studies were performed in CHO lysates (approximately 30 to 40 μg of total cellular protein in 50 μL reaction) incubated for 30 minutes at 37° C. The above-described assay procedure was used maintaining one substrate at saturating concentration (40 mmol/L and 20 mmol/L of homocysteine and 2H-labeled serine, respectively) while using the concentration of the other substrate of 0.31, 0.62, 1.25, 2.5, 5, 7.5, 10, 12.5, 15, 20, 30, and 40 mmol/L. The Km and Vmax were calculated with KaleidaGraph software (Synergy Software, Reading, Pennsylvania).

ng concentration (40 mmol/L and 20 mmol/L of homocysteine and 2H-labeled serine, respectively) while using the concentration of the other substrate of 0.31, 0.62, 1.25, 2.5, 5, 7.5, 10, 12.5, 15, 20, 30, and 40 mmol/L. The Km and Vmax were calculated with KaleidaGraph software (Synergy Software, Reading, Pennsylvania). Western Blot Analysis For Western blot analysis, cell lysates containing total protein 10 μg were electrophoresed on 4% to 15% gradient native polyacrylamide gels (BioRad precast gels) with Laemmli buffer system with or without SDS for SDS-PAGE and native-PAGE gel, respectively. The separated proteins were transferred onto polyvinylidene difluoride membrane (Immobilon-P; Millipore, Billerica, Massachusetts) with semi-dry blotting transfer technique. After transfer and overnight blocking of nonspecific binding sites with 5% non-fat dry milk, the membrane was incubated with immunopurified anti-CBS antibody diluted 1:5000 in 3% bovine serum albumin in 1 × phosphate-buffered saline solution (PBS) for 1 hour. After a series of washes the membrane was subsequently incubated with secondary anti-rabbit IgG antibody conjugated with HRP for 30 minutes (Pierce, Rockford, Illinois); the secondary antibody was diluted 1:30 000 in 5% non-fat dry milk dissolved in 1 × PBS/0.2% Tween 20. After a second series of washes the signal was visualized using the West Pico Super Signal system (Pierce Biotechnology, Rockford, Illinois) followed by bioimaging system ChemiGenius-Q (Syngene Inc, Frederick, Maryland) with cooled CCD camera. The amount of tetramers and high-molecular weight aggregates were quantified with Gene Tools software (Syngene Inc).

al was visualized using the West Pico Super Signal system (Pierce Biotechnology, Rockford, Illinois) followed by bioimaging system ChemiGenius-Q (Syngene Inc, Frederick, Maryland) with cooled CCD camera. The amount of tetramers and high-molecular weight aggregates were quantified with Gene Tools software (Syngene Inc). Statistical Methods Expected birth prevalence was calculated on the basis of Hardy-Weinberg equilibrium with a model published previously10 with added frequency of the c.1105C>T alleles. Confidence intervals of estimates were calculated with S-plus (Tibco Software, Inc, Palo Alto, California) and R software.

al was visualized using the West Pico Super Signal system (Pierce Biotechnology, Rockford, Illinois) followed by bioimaging system ChemiGenius-Q (Syngene Inc, Frederick, Maryland) with cooled CCD camera. The amount of tetramers and high-molecular weight aggregates were quantified with Gene Tools software (Syngene Inc). Statistical Methods Expected birth prevalence was calculated on the basis of Hardy-Weinberg equilibrium with a model published previously10 with added frequency of the c.1105C>T alleles. Confidence intervals of estimates were calculated with S-plus (Tibco Software, Inc, Palo Alto, California) and R software. Results Frequency of the c.1105C>T Mutation and Expected Birth Prevalence of Homocystinuria We detected six c.1105C>T heterozygotes among 600 unselected Czech newborns; 1 individual was present in the Bohemian cohort and 5 in the Moravian cohort, respectively. The observed allelic frequency of c.1105C>T is 0.005 (95%CI 0.0018-0.011), which is similar to that observed in Norway.12 Assuming Hardy-Weinberg equilibrium the expected birth prevalence of homozygotes for this mutation is 1:40 000 (95%CI 1:8000-1:295 000). We have previously estimated incidence of homocystinuria in the Czech Republic with a composite model.10 In the model we have combined direct detection of 2 mutant alleles in unselected newborn blood samples with an inferred population frequency of 8 additional mutant alleles calculated from their number in patients who were ascertained by biochemical screening during a 20-year period. After adding the data on c.1105C>T from this study into the above-mentioned composite model, the expected birth prevalence of homocystinuria in the Czech Republic increased to 1:15 500 newborns (95% CI 1:7500-1:38 000).

from their number in patients who were ascertained by biochemical screening during a 20-year period. After adding the data on c.1105C>T from this study into the above-mentioned composite model, the expected birth prevalence of homocystinuria in the Czech Republic increased to 1:15 500 newborns (95% CI 1:7500-1:38 000). In Silico Analyses of the p.R369C Mutant Arginine 369 is located in the beta strand 11 at a dimer-dimer interface of the CBS enzyme.18 In their 3-dimensional model of full-length CBS, Yamanishi et al19 predicted that this mutation abolishes an ionic interaction with D506 residue located in the carboxyterminal regulatory CBS domain, suggesting possible destabilization of tertiary structure. Analysis with Clustal-W and a series of CBS orthologs obtained from NCBI database showed that arginine 369 is highly conserved in mammals, and in other organisms, this positively charged amino acid may be replaced by positively charged lysine or uncharged tyrosine, valine, leucine, methionine or asparagine residues but not by cysteine (data not shown). This imperfect conservation of arginine in position 369 of the CBS polypeptide during evolution suggests that this amino acid residue is not absolutely critical for cystathionine beta-synthase function. Taken together these computer-simulated data suggest that the effect of this mutation on enzyme function may be rather mild.

perfect conservation of arginine in position 369 of the CBS polypeptide during evolution suggests that this amino acid residue is not absolutely critical for cystathionine beta-synthase function. Taken together these computer-simulated data suggest that the effect of this mutation on enzyme function may be rather mild. Properties of p.R369C Mutant Expressed in E. coli The p.R369C mutant expressed at 37° C was present in bacterial extracts in amounts somewhat lower than the wild-type enzyme (66 % of CBS signal relative to the wild-type upon analysis by denaturing SDS-PAGE/western blotting; data not shown). Most of these mutant enzyme molecules misfolded, as demonstrated by the presence of only 8% of correctly folded tetrameric fraction of the p.R369C relative to the wild-type CBS enzyme when non-denaturing conditions for PAGE/Western blotting were used and data were normalized for CBS abundance (Figure). The relative activity—that is, specific activity normalized for the abundance of tetramers—of the p.R369C mutant was decreased to 34% of the wild-type (Table II).

relative to the wild-type CBS enzyme when non-denaturing conditions for PAGE/Western blotting were used and data were normalized for CBS abundance (Figure). The relative activity—that is, specific activity normalized for the abundance of tetramers—of the p.R369C mutant was decreased to 34% of the wild-type (Table II). It was shown previously20 that lower expression temperature may facilitate the attainment of correct tertiary and quaternary structure of misfolding-prone mutants. Indeed, folding and assembly of mutant p.R369C polypeptide chains expressed at 18° C improved substantially as the bacterial extracts contained 23% of tetramer compared with wild-type and the relative activity simultaneously increased to 67% (Table II). To examine whether the p.R369C mutation may alter the response to the natural allosteric activator of CBS—S-adenosylmethionine (SAM)—we measured the catalytic activity of the mutant expressed at 18° C in the absence and presence of 1 mmol/L SAM. The activity of the mutant increased 4.4 times in the presence of SAM, which is similar to the increase of 5.7 for the wild-type enzyme; these data demonstrate that the mutant enzyme retains its normal allosteric activation by S-adenosylmethionine.

mutant expressed at 18° C in the absence and presence of 1 mmol/L SAM. The activity of the mutant increased 4.4 times in the presence of SAM, which is similar to the increase of 5.7 for the wild-type enzyme; these data demonstrate that the mutant enzyme retains its normal allosteric activation by S-adenosylmethionine. The above data show that the p.R369C CBS mutant is intrinsically prone to misfolding with simultaneous loss of activity, and that this property may be reversed by folding permissive conditions. Moreover, the expression studies demonstrated that even the correctly folded and assembled tetramers exhibit only 34% to 67% relative activity per tetramer amount compared with the wild-type enzyme. It was shown previously that despite their limitations the prokaryotic expression systems permit examination of intrinsic properties of mutant enzymes such as cytosolic phenylalanine hydroxylase, and expression in mammalian systems mimics more faithfully the in vivo situation in human beings.21 Therefore we subsequently tested the properties of the mutant enzyme in a mammalian system.

tic expression systems permit examination of intrinsic properties of mutant enzymes such as cytosolic phenylalanine hydroxylase, and expression in mammalian systems mimics more faithfully the in vivo situation in human beings.21 Therefore we subsequently tested the properties of the mutant enzyme in a mammalian system. Properties of p.R39C Mutant Expressed in CHO-K1 Cells The p.R369C mutant expressed in mammalian cells was present in amounts similar to the wild-type enzyme (105% of the wild-type enzyme when assessed by Western blotting under denaturing conditions, data not shown). Surprisingly, the amounts of correctly folded tetramers and higher order oligomers were slightly increased to 129% compared with wild-type enzyme after correction for the CBS abundance. However, the relative activity of these correctly folded mutant oligomers was decreased to 54% of the wild-type tetramers (Table II). Taken together the net effect of the p.R369C substitution on quaternary structure and activity appears to be moderate in mammalian cells.

e enzyme after correction for the CBS abundance. However, the relative activity of these correctly folded mutant oligomers was decreased to 54% of the wild-type tetramers (Table II). Taken together the net effect of the p.R369C substitution on quaternary structure and activity appears to be moderate in mammalian cells. All experiments showed that despite varying extent of correct folding in either the mammalian or prokaryotic system the relative activity of the p.R369C mutant is similarly decreased to about half of the wild-type enzyme. Such behavior of the p.R369C mutant is expected to result from altered kinetic properties. Therefore we determined the Michaelis constant Km for both substrates and we measured the Vmax of reaction in crude CHO cell extracts. The p.R369C mutant exhibited Km for serine and homocysteine similar to Km of the wild-type CBS (Table II). In contrast the Vmax of the mutant was reduced to 29% of the wild-type enzyme and to 55% when corrected for the abundance of tetramers in samples that were used for kinetic studies. Considering the limitations of analyses in crude extracts all these data suggest that the p.R369C mutant does not have lower affinity for its 2 substrates but that the velocity of the reaction is approximately halved.

nd to 55% when corrected for the abundance of tetramers in samples that were used for kinetic studies. Considering the limitations of analyses in crude extracts all these data suggest that the p.R369C mutant does not have lower affinity for its 2 substrates but that the velocity of the reaction is approximately halved. Discussion The surprisingly high frequency of the c.1105C>T allele among Norwegian newborns suggests that this CBS mutation may be the most common cause of homocystinuria in North Europeans.12 However, functional studies indicated that it may only be a frequent neutral polymorphism, although the authors discussed a possibility that the yeast may not model this mutation properly.14 Because detailed functional studies and data on frequency in other populations have been lacking, we aimed in our study at exploring the properties of p.R369C mutant enzyme and at determining the prevalence of the c.1105C>T allele in another European population.

ty that the yeast may not model this mutation properly.14 Because detailed functional studies and data on frequency in other populations have been lacking, we aimed in our study at exploring the properties of p.R369C mutant enzyme and at determining the prevalence of the c.1105C>T allele in another European population. From clinical standpoint the question whether the c.1105C>T mutation causes homocystinuria or whether it is only a functionally neutral genetic variant is of key importance. Three lines of evidence support the hypothesis that it is indeed pathogenic. First, this mutation was found at 1 parental allele in 2 independent patients with CBS deficiency (Table I; patients 1 and 2), demonstrating that the p.R369C mutant enzyme produced from the c.1105C>T allele has impaired catalytic activity because it is not able to functionally complement the mutant protein synthesized from the other parental allele. Second, the expression studies presented in this article strongly suggest that the mutant enzyme is intrinsically prone to misfolding but more folding permissive conditions of mammalian cells enable formation of normal amounts of tetramers; in human beings carrying this mutation the tendency to misfold may only manifest under conditions that disfavor folding. Third, regardless of the expression system, the relative activity of p.R369C mutant normalized for tetramer abundance is decreased to about one half of the wild-type enzyme. We have shown that this low activity is not caused by altered binding of substrates but rather by a decreased velocity of the enzymatic reaction.

Third, regardless of the expression system, the relative activity of p.R369C mutant normalized for tetramer abundance is decreased to about one half of the wild-type enzyme. We have shown that this low activity is not caused by altered binding of substrates but rather by a decreased velocity of the enzymatic reaction. The properties of mutant enzyme expressed in cells, however, cannot unequivocally answer the question whether the p.R369C mutation is pathogenic in vivo. The evidence has to come from clinical observations; however, there is a scarcity of reports on CBS-deficient patients carrying the c.1105 C>T variant. This discrepancy between the numbers of expected and reported homozygotes/compound heterozygotes for c.1105 C>T may be due to several reasons, for example, (1) patients may die antenatally, (2) patients may have a phenotypically different disease, which does not resemble homocystinuria (such an example is the mutation p.V377I in mevalonate kinase gene22), (3) the disease manifests in an unrecognized mild adult-onset form, or (4) patients do not have any clinical symptoms at all. The late onset of symptoms, lack of mental retardation, and of connective tissue involvement, and pyridoxine responsiveness in the known patients carrying the p.R369C mutation favor the latter 2 hypotheses of a mild to null phenotype, which possibly escapes diagnosis. It is puzzling that such individuals are not being reported among patients with moderate to severe hyperhomocysteinemia who had been ascertained during a thrombophilia workout. In summary, it is conceivable that in human beings the p.R369C mutation is not functionally neutral in vivo although severity and spectrum of clinical consequences is at present unknown; only future reports on phenotypic features in homozygotes and additional compound heterozygotes carrying this mutation will resolve this issue.

, it is conceivable that in human beings the p.R369C mutation is not functionally neutral in vivo although severity and spectrum of clinical consequences is at present unknown; only future reports on phenotypic features in homozygotes and additional compound heterozygotes carrying this mutation will resolve this issue. The second aim of our study was to determine the frequency of c.1105 C>T in another European population. The 0.5% frequency of c.1105 C>T alleles in predominantly Slavic population of the Czech Republic is similar to 0.8% frequency determined by Refsum et al12 in Norwegian newborns or 0.5% shown for 200 North American adult control subjects by Kim et al.14 Although data from other European populations are lacking, similar frequencies in unrelated Norwegians, Czechs, and North Americans suggest that this variant allele may be of ancient origin and that it may be common in populations of European descent. This high population frequency of mutant CBS alleles may have important consequences for newborn screening. The expected frequency of homocystinuria because of 6 mutations in Norway12 and 11 mutations in the Czech Republic (this study) are similarly high, being 1:6400 and 1:15 500, respectively. Provided that p.R369C is indeed pathogenic, the expected frequency of homocystinuria in 2 European populations is similar to the frequency of other common inborn errors of metabolism such as phenylketonuria or MCAD deficiency.7 Such frequency and encouraging efficacy of therapeutic interventions in early detected patients with CBS deficiency (summarized in reference 2) rank thus CBS deficiency among suitable candidates for an efficient newborn screening.

of other common inborn errors of metabolism such as phenylketonuria or MCAD deficiency.7 Such frequency and encouraging efficacy of therapeutic interventions in early detected patients with CBS deficiency (summarized in reference 2) rank thus CBS deficiency among suitable candidates for an efficient newborn screening. Unfortunately, the existing newborn screening programs for CBS deficiency are based on detecting moderately to grossly elevated methionine concentrations by various techniques, including tandem mass spectrometry. It has been shown that patients may not be ascertained if higher cut-off methionine levels are used.23 Especially the pyridoxine-responsive forms, which are present in at least 50% of patients with homocystinuria,1 seem to escape diagnosis by newborn screening programs.7 This is demonstrated by only a single patient with pyridoxine-responsive homocystinuria detected among 36 Irish and U.S. patients that were ascertained in the newborn period.8,23 Total homocysteine measurement in blood samples24 or dried blood spots25 may be a more reliable approach to detecting neonates with homocystinuria. Unfortunately, total homocysteine is not routinely analyzed in newborn screening programs because of the necessity of releasing homocysteine from its bond to plasma proteins in an additional pre-derivatization reduction step. Because cystathionine is decreased and methionine normal to increased in CBS deficiency, the addition of a cystathionine assay with calculation of cystathionine-to-methionine ratio may be an approach for increasing the number of detected presymptomatic newborn patients with homocystinuria. The molecular epidemiology data reported previously12,14,26 and in this study challenge our perception of homocystinuria as a rare disease and should stimulate a discussion on how to increase the efficacy of newborn screening for homocystinuria in populations of European origin.

ic newborn patients with homocystinuria. The molecular epidemiology data reported previously12,14,26 and in this study challenge our perception of homocystinuria as a rare disease and should stimulate a discussion on how to increase the efficacy of newborn screening for homocystinuria in populations of European origin. The authors would like to express their gratitude to Petra Melenovská, MSc, Jana Kopecká, MSc, and Alena Dutá for technical help, Markéta Pavlíková, MSc, for calculating the confidence intervals, and J. P. Kraus, PhD, for providing the anti-CBS antibodies. Supported by the Wellcome Trust International Senior Research Fellowship in Biomedical Science in Central Europe (reg. No 070255/Z/03/Z). Institutional support was provided by the Research Project of the Ministry of Education of the Czech Republic (reg. No MSM0021620806). The authors declare no conflict of interests.

The authors would like to express their gratitude to Petra Melenovská, MSc, Jana Kopecká, MSc, and Alena Dutá for technical help, Markéta Pavlíková, MSc, for calculating the confidence intervals, and J. P. Kraus, PhD, for providing the anti-CBS antibodies. Supported by the Wellcome Trust International Senior Research Fellowship in Biomedical Science in Central Europe (reg. No 070255/Z/03/Z). Institutional support was provided by the Research Project of the Ministry of Education of the Czech Republic (reg. No MSM0021620806). The authors declare no conflict of interests. Figure Quaternary structure of p.R369C mutant under different expression conditions. The p.R369C mutant was expressed in both E. coli and CHO cell-based systems, crude extracts were electrophoresed in gradient polyacrylamide gels under non-denaturing conditions followed by Western blotting (see Methods for details). A representative gel is shown, the sharply demarcated fractions (CBS)4, (CBS)8, or (CBS)12 contain 4, 8, or 12 CBS subunits, respectively; the smear in the high molecular weight fraction contains misfolded CBS. The individual lanes contain either the p.R369C mutant (M), wild-type CBS (WT), or blank (B, i.e. either E. coli transformed with empty vector or CHO cells transfected with empty vector pTRE2hyg, respectively). The line “Average amount of tetramers/oligomers” shows the relative amounts of the sum of correctly folded mutant tetramers and oligomers relative to the wild-type CBS enzyme, data are means of 3 independent expressions with SD in parentheses.

y vector or CHO cells transfected with empty vector pTRE2hyg, respectively). The line “Average amount of tetramers/oligomers” shows the relative amounts of the sum of correctly folded mutant tetramers and oligomers relative to the wild-type CBS enzyme, data are means of 3 independent expressions with SD in parentheses. Table I Known patients with a CBS deficiency carrying the c.1105C>T (p.R369C) mutation Ancestry Genotype⁎ Disease severity Vitamin B6† Comments Reference Norwegian-sib pair Allele 1: p.R369C Allele 2: p.I278T variable among siblings + A sibpair diagnosed at 33 and 27 years of age, respectively. The older sister was symptom free, the younger brother had severe psychiatric disease. Plasma tHcy concentrations in these 2 patients were 245 and 130 μmol/L before, and 8 and 18 μmol/L after pyridoxine administration (40 mg/d), respectively. 14 Anglo-Celtic Allele 1: p.R369C Allele 2: c.533del18

variable among siblings + A sibpair diagnosed at 33 and 27 years of age, respectively. The older sister was symptom free, the younger brother had severe psychiatric disease. Plasma tHcy concentrations in these 2 patients were 245 and 130 μmol/L before, and 8 and 18 μmol/L after pyridoxine administration (40 mg/d), respectively. 14 Anglo-Celtic Allele 1: p.R369C Allele 2: c.533del18 mild + The patient manifested by pulmonary embolus at the age of 23 years after the birth of her son. She was ascertained by family screening at 24 years of age since her son, who inherited from her the 533del18 mutant allele, was diagnosed with CBS deficiency. 4 Dutch Both alleles: [p.R369C; p.R491C] severe + Severely affected patient with ectopic lenses, other connective tissue abnormalities, thromboembolism and psychosis but without mental retardation, age of diagnosis is not given. The relative contribution of the 2 mutations linked in cis to the severe inactivation of CBS enzyme has not been determined. 13 ⁎ Genotypes of both alleles are shown with description of either cDNA nucleotide change (signified by “c.”) or of the predicted amino acid substitution (signified by “p.”). † This column shows pyridoxine responsiveness; in references 4 and 11 the vitamin B6 responsiveness is defined as decrease of plasma tHcy below 50 or 60 μmol/L after pyridoxine treatment, respectively. Table II Catalytic activity and kinetic properties of the p.R369C mutant

mild + The patient manifested by pulmonary embolus at the age of 23 years after the birth of her son. She was ascertained by family screening at 24 years of age since her son, who inherited from her the 533del18 mutant allele, was diagnosed with CBS deficiency. 4 Dutch Both alleles: [p.R369C; p.R491C] severe + Severely affected patient with ectopic lenses, other connective tissue abnormalities, thromboembolism and psychosis but without mental retardation, age of diagnosis is not given. The relative contribution of the 2 mutations linked in cis to the severe inactivation of CBS enzyme has not been determined. 13 ⁎ Genotypes of both alleles are shown with description of either cDNA nucleotide change (signified by “c.”) or of the predicted amino acid substitution (signified by “p.”). † This column shows pyridoxine responsiveness; in references 4 and 11 the vitamin B6 responsiveness is defined as decrease of plasma tHcy below 50 or 60 μmol/L after pyridoxine treatment, respectively. Table II Catalytic activity and kinetic properties of the p.R369C mutant Wild-type CBS p.R369C mutant Specific activity in crude extracts (nmol cystathionine/mg total cellular protein/hour)⁎ E. coli 37° C 142.7 ± 27.2 2.8 ± 2.2 E. coli 18° C 85.5 ± 27.0 13.1 ± 0.4 CHO-K1 37° C 111.1 ± 39.7 71.2 ± 27.3 Relative activity in crude extracts (specific activity corrected for tetramer abundance, in % of wild-type enzyme) E. coli 37° C 100% 34% ± 26% E. coli 18° C 100% 67% ± 8% CHO-K1 37° C 100% 54% ± 24% Kinetic properties of wild-type and p.R369C mutant enzyme in crude extracts after expression in CHO-K1 cells† Km serine (mmol/L) 10.1 8.1 Km homocysteine (mmol/L) 5.8 7.5 Vmax (nmol/h/mg total cellular protein/) 1715 500 Normalized Vmax (nmol/h/arbitrary units of CBS tetramer) 100% 55% ⁎ Numbers shown are means of 3 independent expression experiments with SD.

9C mutant enzyme in crude extracts after expression in CHO-K1 cells† Km serine (mmol/L) 10.1 8.1 Km homocysteine (mmol/L) 5.8 7.5 Vmax (nmol/h/mg total cellular protein/) 1715 500 Normalized Vmax (nmol/h/arbitrary units of CBS tetramer) 100% 55% ⁎ Numbers shown are means of 3 independent expression experiments with SD. † Numbers shown are means of 2 independent experiments (with the exception of Km for homocysteine, which was determined only once).

Our results, together with other findings, suggest that excluding children with ADHD from services and interventions on the basis of the presence of mild ID is clinically unwarranted, given that children with ADHD and ID do not seem to differ from those without ID in terms of ADHD subtype and number of ADHD symptoms. They are more likely to have CD, however. It also appears that they differ from children with ID alone, suggesting that ID does not drive the link to conduct problems. Thus, services that deal with ADHD should be well placed to manage ADHD in children with mild ID; however, they will need access to the types of social and clinical interventions that will also help manage associated conduct problems. Appendix Table III Clinical features of children with ADHD + ID (clinical sample) and those with ID only (ALSPAC sample) after matching the samples based on IQ Variables ADHD + ID (n = 58) ID-only (n = 58) Statistics n (%) Mean (SD) n (%) Mean (SD) OR 95% CI P Age, years 10.6 (2.9) 10.8 (0.1) 0.96 0.80-1.15 .64 Male sex 50 (86.2) 31 (53.4) 0.18 0.07-0.46 2.5E-04 ODD symptom count 4.1 (2.3) 0.2 (0.7) 5.08 2.63-9.79 1.2E-06∗ CD symptom count (out of 7) 1.1 (1.5) 0.0 (0.3) 11.41 2.82-46.13 6.4E-04∗ ODD diagnosis 28 (48.3) 2 (3.4) 32.09 6.26-164.56 3.2E-05∗ CD diagnosis 16 (16.7) 0 (0.0) NA NA NA† ∗ A adjusted for the covariates: child's age at time of assessment, sex, and IQ. † Because there were no CD diagnoses in the ID-only group, statistical calculation was not applicable.

Variables ADHD + ID (n = 58) ID-only (n = 58) Statistics n (%) Mean (SD) n (%) Mean (SD) OR 95% CI P Age, years 10.6 (2.9) 10.8 (0.1) 0.96 0.80-1.15 .64 Male sex 50 (86.2) 31 (53.4) 0.18 0.07-0.46 2.5E-04 ODD symptom count 4.1 (2.3) 0.2 (0.7) 5.08 2.63-9.79 1.2E-06∗ CD symptom count (out of 7) 1.1 (1.5) 0.0 (0.3) 11.41 2.82-46.13 6.4E-04∗ ODD diagnosis 28 (48.3) 2 (3.4) 32.09 6.26-164.56 3.2E-05∗ CD diagnosis 16 (16.7) 0 (0.0) NA NA NA† ∗ A adjusted for the covariates: child's age at time of assessment, sex, and IQ. † Because there were no CD diagnoses in the ID-only group, statistical calculation was not applicable. Table IV Clinical features of children with ADHD + ID compared with those with IQ 70-84 and those with IQ ≥85

Attention deficit hyperactivity disorder (ADHD) is a disabling condition, affecting 1.4%-6% of children.1 Little is known of the clinical presentation and etiology of ADHD in children with intellectual disability (ID), because those with lower cognitive ability (IQ scores <70) are often excluded from studies of ADHD,2 despite evidence that ADHD is more common in children with ID, and that the risk increases with increasing severity of ID.3 It has been suggested that ADHD does not occur in children with ID, and that any inappropriate behavior in children with ID is secondary to “mental impairment.”4 That view is not supported by current evidence, however. Studies have shown that ADHD occurs more commonly in these children but may be underdiagnosed owing to such issues as “diagnostic overshadowing,” the tendency of clinicians to overlook additional psychiatric diagnoses after a diagnosis of ID is made, or “masking,” in which the clinical characteristics of a mental disorder are masked by a cognitive, language, or speech deficit.5 A population-based study estimating the prevalence of psychiatric diagnoses in children with ID identified hyperkinetic disorder as the most common psychiatric disorder.6 Studies of children with mild and borderline ID have identified ADHD in 8%-39% of cases.7-9 A crucial clinical issue is whether or not the clinical pattern of comorbidity in this group is the same as that seen in children with ADHD but without ID. This is important in determining the level and type of services and clinical care required for this subgroup.

derline ID have identified ADHD in 8%-39% of cases.7-9 A crucial clinical issue is whether or not the clinical pattern of comorbidity in this group is the same as that seen in children with ADHD but without ID. This is important in determining the level and type of services and clinical care required for this subgroup. In the present study, we compared the rates of comorbid problems and ADHD symptom levels in 2 groups of children with ADHD, 1 group with ID (ADHD + ID group) and the other group without ID (ADHD-only group). Consistent with previous studies of ADHD, we defined ID is an IQ test score <70. We hypothesized that the ADHD profiles in the 2 groups (ADHD + ID [IQ <70] vs ADHD-only [IQ ≥70]) would be highly similar in terms of symptoms, rates of subtypes, and patterns of comorbid problems (ie, oppositional behaviors, conduct disorder [CD], anxiety, and depression).

ADHD, we defined ID is an IQ test score <70. We hypothesized that the ADHD profiles in the 2 groups (ADHD + ID [IQ <70] vs ADHD-only [IQ ≥70]) would be highly similar in terms of symptoms, rates of subtypes, and patterns of comorbid problems (ie, oppositional behaviors, conduct disorder [CD], anxiety, and depression). Methods Participants were recruited from more than 30 child and adolescent mental health services or community pediatric outpatient clinics in Wales, England, and Scotland for a genetic study of ADHD. Given this study's focus on evaluating for the presence of nonsyndromal ID in children with ADHD, International Statistical Classification of Diseases and Related Health Problems, 10th revision10 and Diagnostic and Statistical Manual of Mental Disorders, 4th edition (DSM-IV)11 exclusion criteria were used. Children with a known diagnosis of schizophrenia, autism spectrum disorder (ASD), bipolar disorder, Tourette syndrome, epilepsy, brain damage, or any other neurologic or genetic disorder were excluded. Information on these conditions was derived from a questionnaire completed by the referring clinician, diagnostic interview information obtained from parents, and quality control of genetic data performed as part of the genetic study. Children with IQ <50 were also excluded, because the study focused on mild ID, and the assessment measures have not yet been validated in individuals with severe ID.

e referring clinician, diagnostic interview information obtained from parents, and quality control of genetic data performed as part of the genetic study. Children with IQ <50 were also excluded, because the study focused on mild ID, and the assessment measures have not yet been validated in individuals with severe ID. A total of 971 children met the inclusion criteria and had sufficient data for analysis. All of these children met the DSM-IV11 or Diagnostic and Statistical Manual of Mental Disorders, 3rd edition revised (DSM-III-R)12 criteria for a diagnosis of ADHD, which was confirmed through research diagnostic interviews.13 The children ranged in age from 5 to 17 years (mean age, 10.1 ± 2.8 years), and included 148 females (15.2%). The study received ethical approval from the North West England and Wales Multicentre Research Ethics Committees. For all subjects, written informed consent was obtained from parents and assent/consent from children.

ildren ranged in age from 5 to 17 years (mean age, 10.1 ± 2.8 years), and included 148 females (15.2%). The study received ethical approval from the North West England and Wales Multicentre Research Ethics Committees. For all subjects, written informed consent was obtained from parents and assent/consent from children. Cognitive ability was assessed using the Wechsler Intelligence Scale for Children versions III (n = 381) and IV (n = 590)14,15 to obtain an estimate of full-scale IQ (using all required subtests). Two versions of this assessment tool were used because version IV was released during the study period. The assessment was performed by trained psychologists. In children who had recently undergone IQ assessment in school, that score was used to determine ID status. In accordance with International Statistical Classification of Diseases and Related Health Problems, 10th revision and DSM-IV criteria, children with an IQ score of 50-69 were considered to have mild mental retardation/ID and classified in the ADHD + ID group. Children with an IQ score ≥70 were classified in the ADHD-only group.

nce with International Statistical Classification of Diseases and Related Health Problems, 10th revision and DSM-IV criteria, children with an IQ score of 50-69 were considered to have mild mental retardation/ID and classified in the ADHD + ID group. Children with an IQ score ≥70 were classified in the ADHD-only group. ADHD symptoms, impairment, and diagnoses were confirmed using the Child and Adolescent Psychiatry Assessment (CAPA),13 a research diagnostic interview with parents. Interviews were performed by trained psychologists supervised weekly by a child psychiatrist. Interrater reliability for ADHD was perfect (κ = 1.0). Information on ADHD symptoms and school impairments was obtained using the Child ADHD Teacher Telephone Interview,16 the DuPaul teacher rating scale,17 or the Conners teacher rating scale.18 A diagnosis of ADHD required that the child have symptoms meeting DSM-IV or DSM-III-R criteria, substantial impairment from symptoms at home, and pervasive symptoms and impairment in the school setting.

e Child ADHD Teacher Telephone Interview,16 the DuPaul teacher rating scale,17 or the Conners teacher rating scale.18 A diagnosis of ADHD required that the child have symptoms meeting DSM-IV or DSM-III-R criteria, substantial impairment from symptoms at home, and pervasive symptoms and impairment in the school setting. The CAPA was also used to assess current symptoms, impairment, and DSM-IV diagnoses of comorbid oppositional defiant disorder (ODD), CD, anxiety disorders (ie, generalized anxiety disorder, social anxiety, and separation anxiety), depression, and mania. Comorbid symptoms were also assessed using the child version of the CAPA19 for children aged ≥12 years. Comorbid anxiety or depression symptoms were endorsed if reported by the parent or child. Owing to the scarcity of anxiety and depression diagnoses in the sample, only symptoms of these disorders could be analyzed. Interrater reliability for parent-rated CD symptoms was very good (intraclass correlation, 0.98).

years. Comorbid anxiety or depression symptoms were endorsed if reported by the parent or child. Owing to the scarcity of anxiety and depression diagnoses in the sample, only symptoms of these disorders could be analyzed. Interrater reliability for parent-rated CD symptoms was very good (intraclass correlation, 0.98). Avon Longitudinal Study of Parents and Children To compare clinical variables found to be associated in the primary analysis in the ADHD + ID and ADHD-only groups, we turned to the Avon Longitudinal Study of Parents and Children (ALSPAC), a large, well-characterized longitudinal dataset. Details of the study methodology are available elsewhere.20 Ethical approval for all aspects of the study was obtained from the ALSPAC Law and Ethics Committee and the local Research Ethics Committees. Parents provided written consent and the children provided assent at each assessment. IQ had been assessed at age 8 years using the Wechsler Intelligence Scale for Children version III.14 Children who scored between 50 and 69 on the IQ test and had no diagnosis of ADHD or ASD were included in our analysis. A total of 74 children (1.2% of the ALSPAC sample with complete data on these measures) met these criteria. Data on ADHD, ASD, ODD, and CD symptoms and diagnoses were collected from participants at age 128 months, using the parent and teacher Development and Well-Being Assessment.21 Complete clinical data were available for 58 children, who constituted the ID-only group. These children were 10-11 years old at the time of clinical assessment (mean, 10.8 ± 0.1 years), and 27 were female (46.6%).

rom participants at age 128 months, using the parent and teacher Development and Well-Being Assessment.21 Complete clinical data were available for 58 children, who constituted the ID-only group. These children were 10-11 years old at the time of clinical assessment (mean, 10.8 ± 0.1 years), and 27 were female (46.6%). Statistical Analyses The ADHD clinical sample was divided into those with ID (ADHD + ID; n = 97) and those without ID (ADHD-only; n = 874). The 2 groups were compared on each of the clinical factors identified. All descriptive statistics are presented as raw scores for ease of interpretation. Where a variable was nonnormally distributed, the scores were naturally logarithmically transformed, and analyses were run on transformed scores. Clinical predictor variables were used to predict binary outcomes (ADHD + ID or ADHD-only) using regression analyses. All analyses included child's age at the time of assessment as a covariate. Sex was not included as a covariate, because it was not associated with the presence or absence of ID. Clinical variables were assessed both categorically and continuously, whenever relevant. All analyses were performed using SPSS version 16 (IBM, Armonk, New York). To take into account multiple testing, Bonferroni correction for the number of variables tested was used, α was set at P = .003 (0.05/15) for the 15 tests performed.

variables were assessed both categorically and continuously, whenever relevant. All analyses were performed using SPSS version 16 (IBM, Armonk, New York). To take into account multiple testing, Bonferroni correction for the number of variables tested was used, α was set at P = .003 (0.05/15) for the 15 tests performed. Based on our results, a hypothesis-driven comparison of the ADHD + ID and ID-only samples was performed for rates of diagnoses and symptom counts for ODD and CD, adjusted for the covariates age, sex, and IQ. CD items available in both datasets were summed and used to generate the CD diagnoses (lying, fighting, breaking curfew, stealing, truancy, running away from home, and bullying). Results Sample Description: Clinical ADHD Sample At the time of assessment, 74.3% of the children (n = 721) met the criteria for DSM-IV ADHD Combined type, 6.0% (n = 58) met the criteria for DSM-IV ADHD Inattentive type, 9.6% (n = 93) met the criteria for DSM-IV ADHD Hyperactive-Impulsive type, and the remaining 10.2% (n = 99) met the criteria for DSM-III-R ADHD. The rates of comorbid disorders were 45.2% (n = 435) for DSM-IV ODD, 17.6% (n = 171) for DSM-IV CD, 6.1% (n = 59) for any anxiety disorder, and 1.1% (n = 11) for any depressive disorder. The mean IQ test scores were 61.8 ± 5.4 for the ADHD + ID group and 87.8 ± 11.4 for the ADHD-only group; the range of scores was 50-69 for the ADHD + ID group and 70-139 for the ADHD-only group. The scores were normally distributed for the sample as a whole.

iety disorder, and 1.1% (n = 11) for any depressive disorder. The mean IQ test scores were 61.8 ± 5.4 for the ADHD + ID group and 87.8 ± 11.4 for the ADHD-only group; the range of scores was 50-69 for the ADHD + ID group and 70-139 for the ADHD-only group. The scores were normally distributed for the sample as a whole. Sample Description: ID-Only ALSPAC Sample The mean IQ score of the ID-only group was 64.8 ± 4.5. At the 128-month assessment, 2 of the children in the ID-only group met the criteria for DSM-IV ODD (3.4%), and none met the criteria for DSM-IV CD. Analysis of Clinical Features in Children with ADHD The ADHD + ID group was older than the ADHD-only group, but the 2 groups did not differ in terms of sex distribution. Table I presents the descriptive statistics and results of regression analyses with age included as a covariate. Although a trend for the ADHD + ID group to be more likely to have the DSM-IV Combined ADHD subtype was seen, this result did not withstand correction for multiple testing. Otherwise, the 2 groups of children were similar in terms of ADHD subtypes and Inattentive, Hyperactive-Impulsive, and total ADHD symptoms. The 2 groups also had similar rates of ODD diagnoses and of anxiety and depression symptoms. There was a trend for children with ADHD + ID to have on more ODD symptoms on average. The ADHD + ID group had more symptoms of CD and were more likely to have a diagnosis of CD; these associations remained after multiple testing was taken into consideration.

tes of ODD diagnoses and of anxiety and depression symptoms. There was a trend for children with ADHD + ID to have on more ODD symptoms on average. The ADHD + ID group had more symptoms of CD and were more likely to have a diagnosis of CD; these associations remained after multiple testing was taken into consideration. Analysis of Clinical Features in Children with ID-Only Results of the comparison of ADHD + ID and ID-only groups (clinical sample vs population sample) are presented in Table II. The 2 groups differed significantly in terms of sex (more boys in the ADHD + ID group), but not in age at assessment. Although the range of IQ scores was similar in the 2 groups, the ADHD + ID group had significantly lower scores (OR, 0.88; 95% CI, 0.82-0.95; P = .001). After adjusting for sex, age, and IQ score, the children in the ADHD + ID group were significantly more likely to have a diagnosis of ODD, and had significantly more symptoms of ODD and CD. Statistical assessment of the between-group difference in the rate of CD diagnoses was not possible, given the rate of 0 in the ID-only group.

ing for sex, age, and IQ score, the children in the ADHD + ID group were significantly more likely to have a diagnosis of ODD, and had significantly more symptoms of ODD and CD. Statistical assessment of the between-group difference in the rate of CD diagnoses was not possible, given the rate of 0 in the ID-only group. Discussion In our clinical sample of children with ADHD, those with and without mild ID (IQ score 50-69) exhibited similar patterns of ADHD subtypes, total number of ADHD symptoms, and comorbidity with ODD, anxiety, and depression. The children in the ADHD + ID group did have higher rates of CD symptoms and diagnoses, however. To explore these results further and test whether these differences were related to an increase in behavioral problems in children with ID in general, we compared our ADHD + ID group with a population-based sample of children with ID but without ADHD. We found significantly higher rates of ODD diagnoses and ODD and CD symptoms in the ADHD + ID group, suggesting that the combination of ADHD and ID results in increases in the rates of comorbid CD and ODD beyond the rates of these disorders in individuals with ID only or ADHD only. The mean IQ score was lower in the ADHD + ID group compared with the ID-only group. This difference in IQ may be related to selective attrition in the ALSPAC sample,22 clinical ascertainment effects, or the effects of ADHD on IQ test performance. Regardless, when we matched the groups on IQ scores by selecting a subsample of children with ADHD + ID (n = 58, to match the number of children in the ID-only group), we obtained the same pattern of results (Table III; available at www.jpeds.com).

l ascertainment effects, or the effects of ADHD on IQ test performance. Regardless, when we matched the groups on IQ scores by selecting a subsample of children with ADHD + ID (n = 58, to match the number of children in the ID-only group), we obtained the same pattern of results (Table III; available at www.jpeds.com). ADHD is one of the most common forms of psychopathology in children with ID.7 Generally, children with ID are neglected in the medical field, perhaps in part because of their poor ability to communicate and social disadvantages.2 They are frequently excluded from clinical, etiologic, and treatment studies,2 and thus few studies to date have examined the clinical presentation, etiology, patterns of service use, and treatment of ADHD in children with ID. Moreover, there are few standardized norms or guidelines in the classification systems for identifying “normal” or “usual” amounts of inattention, overactivity and impulsivity in persons with ID.

date have examined the clinical presentation, etiology, patterns of service use, and treatment of ADHD in children with ID. Moreover, there are few standardized norms or guidelines in the classification systems for identifying “normal” or “usual” amounts of inattention, overactivity and impulsivity in persons with ID. Studies with clinic-referred children and community and population samples have shown consistently higher rates of ADHD in children with ID and higher rates of ID in children with ADHD.23-25 There has been less work on whether the clinical features and levels of ADHD symptomatology differ between children with ID and those without ID. One study found equivalent levels of ADHD symptoms in preschool-age children with ADHD and normal IQ and those with ADHD and ID.26 A population study assessed overactivity, inattention, and impulsivity symptoms using the Strengths and Difficulties Questionnaire in children with IQ <70 and those with IQ ≥70 and found no differences between the 2 groups.24 A recent longitudinal study compared children with ADHD with IQ <85 (indicating ID or borderline IQ) and those with IQ ≥85 and found similar inattentive and hyperactive-impulsive ADHD symptom trajectories over a 3-year period.27 An important difference noted in that sample was that the children with lower IQ tended to meet the diagnostic criteria for ADHD at an earlier age and to have more diagnostic stability than the children with higher IQ, indicating a more severely impairing form of ADHD. A study that addressed this question from the other direction found that although the children with developmental delay (IQ <85) had a higher rate of ADHD than the typically developing children (IQ ≥85), the pattern of ADHD subtypes and levels of inattentive and hyperactive-impulsive symptoms were similar in the 2 groups.9

udy that addressed this question from the other direction found that although the children with developmental delay (IQ <85) had a higher rate of ADHD than the typically developing children (IQ ≥85), the pattern of ADHD subtypes and levels of inattentive and hyperactive-impulsive symptoms were similar in the 2 groups.9 The few studies examining whether rates of comorbid psychiatric disorders in ADHD differ in children with and without ID have shown mixed results. Some studies have found higher rates of comorbid impairments in social skills, conduct problems, aggression, and noncompliance in children with ADHD and ID compared with children with ADHD and normal IQ.26,28 In contrast, in community samples, the profiles of comorbidity with emotional and conduct problems in children with ADHD symptoms did not differ according to the presence of mild ID (IQ <70)24 or using a cutoff of IQ <85.29 A third study found higher rates of noncompliance, anxiety, depression, and social problems in those with ADHD and ID compared with those with ID alone.30

dity with emotional and conduct problems in children with ADHD symptoms did not differ according to the presence of mild ID (IQ <70)24 or using a cutoff of IQ <85.29 A third study found higher rates of noncompliance, anxiety, depression, and social problems in those with ADHD and ID compared with those with ID alone.30 A limitation of previous research is that questionnaires generally have been used to assess comorbidity rather than the more in-depth standardized diagnostic assessment methods used in the present study. Another strength of the present study is the relatively large sample size of the ADHD group compared with previous studies. The IQ cutpoint used to delineate comparison groups in the literature varies, such that in some studies, children with borderline ID (IQ 70-85) are included in the group of children with mild ID. Although the present study focused on comparing children based on an IQ cutpoint of 70, reanalysis of the data after dividing the children into 3 groups (mild ID: IQ 50-69 [n = 97], borderline ID: IQ 70-84 [n = 380], and typically developing: IQ ≥85 [n = 494]) showed a similar pattern of results (Table IV; available at www.jpeds.com).

sent study focused on comparing children based on an IQ cutpoint of 70, reanalysis of the data after dividing the children into 3 groups (mild ID: IQ 50-69 [n = 97], borderline ID: IQ 70-84 [n = 380], and typically developing: IQ ≥85 [n = 494]) showed a similar pattern of results (Table IV; available at www.jpeds.com). An important limitation of the present study is that the presence of ID was defined based primarily on IQ score, because no measure of adaptive functioning was available. Another limitation is that disharmonic IQ profiles were not considered, and thus children with significant performance IQ and verbal IQ discrepancies might not necessarily be considered to have ID in clinical practice. However, given that IQ alone is often the basis for exclusion criteria in ADHD research, the implications of our findings are valid in this context. Samples are also likely to be heterogeneous because of ascertainment differences in referred samples. Our sample comprised referred cases, and thus the higher rate of CD in our cohort may be related to the fact that these children were identified with ADHD and referred. Because this subgroup of children with ADHD + ID are excluded from virtually all clinical, etiologic, and treatment studies, further work is needed to verify our findings. A further limitation of this study is that we were unable to take into account the variable effects of medication timing and dosage. However, in those children receiving a stimulant medication, whether or not they took the medication on the day of testing had no effect on whether they were classified as ADHD + ID or ADHD-only.

r limitation of this study is that we were unable to take into account the variable effects of medication timing and dosage. However, in those children receiving a stimulant medication, whether or not they took the medication on the day of testing had no effect on whether they were classified as ADHD + ID or ADHD-only. A related question that is beyond the scope of the present study is the extent to which the presence of ID in children with ADHD indexes a different etiology. Previous work has shown that children with ADHD + ID are significantly more likely to have large, rare structural deletions or duplications of DNA, called copy number variants (CNVs).31 Importantly however, CNVs are also associated with ADHD without ID.32 In addition, the presence of such CNVs does not appear to index a distinct pattern of etiologic correlates, in the form of various prenatal and perinatal risk factors, or a distinct clinical profile in children with ADHD with and without ID.33 Thus, whether the etiology and risk correlates of ADHD are substantially different in affected children with and without ID remains unclear, and further work is needed to explore this question.

rm of various prenatal and perinatal risk factors, or a distinct clinical profile in children with ADHD with and without ID.33 Thus, whether the etiology and risk correlates of ADHD are substantially different in affected children with and without ID remains unclear, and further work is needed to explore this question. Our results, together with other findings, suggest that excluding children with ADHD from services and interventions on the basis of the presence of mild ID is clinically unwarranted, given that children with ADHD and ID do not seem to differ from those without ID in terms of ADHD subtype and number of ADHD symptoms. They are more likely to have CD, however. It also appears that they differ from children with ID alone, suggesting that ID does not drive the link to conduct problems. Thus, services that deal with ADHD should be well placed to manage ADHD in children with mild ID; however, they will need access to the types of social and clinical interventions that will also help manage associated conduct problems. Appendix Table III Clinical features of children with ADHD + ID (clinical sample) and those with ID only (ALSPAC sample) after matching the samples based on IQ

Variables ADHD + ID (n = 58) ID-only (n = 58) Statistics n (%) Mean (SD) n (%) Mean (SD) OR 95% CI P Age, years 10.6 (2.9) 10.8 (0.1) 0.96 0.80-1.15 .64 Male sex 50 (86.2) 31 (53.4) 0.18 0.07-0.46 2.5E-04 ODD symptom count 4.1 (2.3) 0.2 (0.7) 5.08 2.63-9.79 1.2E-06∗ CD symptom count (out of 7) 1.1 (1.5) 0.0 (0.3) 11.41 2.82-46.13 6.4E-04∗ ODD diagnosis 28 (48.3) 2 (3.4) 32.09 6.26-164.56 3.2E-05∗ CD diagnosis 16 (16.7) 0 (0.0) NA NA NA† ∗ A adjusted for the covariates: child's age at time of assessment, sex, and IQ. † Because there were no CD diagnoses in the ID-only group, statistical calculation was not applicable. Table IV Clinical features of children with ADHD + ID compared with those with IQ 70-84 and those with IQ ≥85 Variables Group 2: IQ 70-84 (n = 380) Statistics∗ (group 2 compared with ADHD + ID) Group 3: IQ ≥85 (n = 494) Statistics∗ (group 3 compared with ADHD + ID) n (%) Mean (SD) OR 95% CI P† n (%) Mean (SD) OR 95% CI P† Age, years 10.2 (2.8) 0.87 0.80-0.94 .0004 9.8 (2.7) 0.82 0.76-0.88 4.5E−07 Male sex 317 (83.4) 1.28 0.68-2.44 .45 422 (85.4) 1.10 0.58-2.08 .76 DSM-IV Combined ADHD diagnosis 277 (73.5) 0.54 0.30-0.96 .04 354 (72.0) 0.49 0.28-0.87 .01 DSM-IV Inattentive ADHD diagnosis 21 (5.5) 1.44 0.52-3.99 .49 30 (6.1) 1.79 0.66-4.87 .25 DSM-IV Hyperactive-Impulsive ADHD diagnosis 41 (10.8) 1.86 0.76-4.54 .17 46 (9.3) 1.55 0.63-3.76 .34 DSM-III-R ADHD diagnosis only 34 (9.0) 1.78 0.67-4.69 .25 55 (11.2) 2.23 0.86-5.77 .10 DSM-IV ODD diagnosis 160 (42.6) 1.19 0.75-1.90 .45 239 (48.8) 1.51 0.95-2.37 .08 DSM-IV CD diagnosis 82 (21.6) 0.55 0.34-0.89 .02 54 (10.9) 0.25 0.15-0.41 7.8E−08 ADHD symptoms: inattentive 7.3 (1.7) 0.79 0.46-1.36 .40‡ 7.3 (1.7) 0.80 0.47-1.36 .41‡ ADHD symptoms: hyperactive-impulsive 7.8 (1.3) 0.81 0.45-1.45 .48‡ 7.7 (1.6) 0.66 0.37-1.18 .16‡ ADHD symptoms: total 15.2 (2.4) 0.94 0.85-1.04 .21 15.1 (2.4) 0.92 0.83-1.01 .09 DSM-IV ODD symptom count 3.9 (2.4) 0.95 0.86-1.05 .33 3.7 (2.3) 0.89 0.81-0.99 .02 DSM-IV CD symptom count 1.3 (1.76) 0.53 0.37-0.75 .0003‡ 0.9 (1.3) 0.34 0.24-0.48 1.5E−09‡ DSM-IV anxiety symptoms 1.1 (1.9) 0.84 0.58-1.20 .33‡ 1.1 (2.0) 0.82 0.57-1.18 .29‡ DSM-IV depression symptoms 1.2 (1.4) 0.72 0.47-1.11 .13‡ 1.3 (1.3) 0.86 0.57-1.31 .49‡ ∗ All clinical analyses are adjusted for the covariate child's age.

symptom count 1.3 (1.76) 0.53 0.37-0.75 .0003‡ 0.9 (1.3) 0.34 0.24-0.48 1.5E−09‡ DSM-IV anxiety symptoms 1.1 (1.9) 0.84 0.58-1.20 .33‡ 1.1 (2.0) 0.82 0.57-1.18 .29‡ DSM-IV depression symptoms 1.2 (1.4) 0.72 0.47-1.11 .13‡ 1.3 (1.3) 0.86 0.57-1.31 .49‡ ∗ All clinical analyses are adjusted for the covariate child's age. † Critical P value corrected for multiple testing: P < .003. ‡ Transformed. We thank the families, pediatricians, and Child and Adolescent Mental Health Service clinicians who supported this project. We also thank the field team members for clinical sample collection and Michael O'Donovan, Michael Owen, Peter Holmans, Lindsey Kent, and Sharifah Syed for assistance with the dataset collected for the genetic study. We also thank all of the families who took part in the ALSPAC study, the midwives for their help in recruiting these participants, and the entire ALSPAC team, including interviewers, computer and laboratory technicians, clerical workers, research scientists, volunteers, managers, receptionists, and nurses. Clinical ADHD sample was funded by the Baily Thomas Charitable Trust, Action Medical Research, and the Wellcome Trust. The ALSPAC sample was funded by the UK Medical Research Council, the Wellcome Trust (092731), and the University of Bristol. The authors declare no conflicts of interest. Table I Clinical features of the ADHD + ID and ADHD-only groups

Clinical ADHD sample was funded by the Baily Thomas Charitable Trust, Action Medical Research, and the Wellcome Trust. The ALSPAC sample was funded by the UK Medical Research Council, the Wellcome Trust (092731), and the University of Bristol. The authors declare no conflicts of interest. Table I Clinical features of the ADHD + ID and ADHD-only groups Variable ADHD + ID (n = 97) ADHD only (n = 874) Statistics∗ n (%) Mean (SD) n (%) Mean (SD) OR 95% CI P† Age, y 11.4 (3.0) 10.0 (2.7) 1.19 1.11-1.28 4.6E-06 Male sex 84 (86.6) 739 (84.6) 0.85 0.46-1.56 .60 DSM-IV Combined ADHD diagnosis 79 (82.3) 631 (72.6) 1.96 1.12-3.40 .02 DSM-IV Inattentive ADHD diagnosis 5 (5.2) 51 (5.8) 0.62 0.23-1.62 .32 DSM-IV Hyperactive-Impulsive ADHD diagnosis 6 (6.2) 87 (10.0) 0.59 0.25-1.41 .23 DSM-III-R ADHD diagnosis only 5 (5.2) 89 (10.2) 0.49 0.19-1.26 .14 DSM-IV ODD diagnosis 36 (37.5) 399 (46.1) 0.74 0.48-1.14 .17 DSM-IV CD diagnosis 35 (36.1) 136 (15.6) 2.69 1.69-4.28 2.8E-05 ADHD symptoms: inattentive 7.5 (1.5) 7.3 (1.7) 1.25 0.75-2.09 .39‡ ADHD symptoms: hyperactive-impulsive 7.8 (1.4) 7.8 (1.5) 1.38 0.79-2.40 .25‡ ADHD symptoms: total 15.4 (2.1) 15.1 (2.4) 1.08 0.98-1.18 .12 DSM-IV ODD symptom count 4.1 (2.3) 3.8 (2.4) 1.09 0.99-1.19 .08 DSM-IV CD symptom count 2.1 (2.2) 1.0 (1.5) 2.38 1.71-3.32 2.5E-07‡ DSM-IV anxiety symptoms 1.2 (2.0) 1.1 (1.9) 1.21 0.86-1.70 .28‡ DSM-IV depression symptoms 1.8 (1.8) 1.3 (1.3) 1.26 0.84-1.89 .26‡ ∗ All clinical analyses were adjusted for the covariate child's age. † Critical P value corrected for multiple testing: P < .003. ‡ Transformed.

Variable ADHD + ID (n = 97) ADHD only (n = 874) Statistics∗ n (%) Mean (SD) n (%) Mean (SD) OR 95% CI P† Age, y 11.4 (3.0) 10.0 (2.7) 1.19 1.11-1.28 4.6E-06 Male sex 84 (86.6) 739 (84.6) 0.85 0.46-1.56 .60 DSM-IV Combined ADHD diagnosis 79 (82.3) 631 (72.6) 1.96 1.12-3.40 .02 DSM-IV Inattentive ADHD diagnosis 5 (5.2) 51 (5.8) 0.62 0.23-1.62 .32 DSM-IV Hyperactive-Impulsive ADHD diagnosis 6 (6.2) 87 (10.0) 0.59 0.25-1.41 .23 DSM-III-R ADHD diagnosis only 5 (5.2) 89 (10.2) 0.49 0.19-1.26 .14 DSM-IV ODD diagnosis 36 (37.5) 399 (46.1) 0.74 0.48-1.14 .17 DSM-IV CD diagnosis 35 (36.1) 136 (15.6) 2.69 1.69-4.28 2.8E-05 ADHD symptoms: inattentive 7.5 (1.5) 7.3 (1.7) 1.25 0.75-2.09 .39‡ ADHD symptoms: hyperactive-impulsive 7.8 (1.4) 7.8 (1.5) 1.38 0.79-2.40 .25‡ ADHD symptoms: total 15.4 (2.1) 15.1 (2.4) 1.08 0.98-1.18 .12 DSM-IV ODD symptom count 4.1 (2.3) 3.8 (2.4) 1.09 0.99-1.19 .08 DSM-IV CD symptom count 2.1 (2.2) 1.0 (1.5) 2.38 1.71-3.32 2.5E-07‡ DSM-IV anxiety symptoms 1.2 (2.0) 1.1 (1.9) 1.21 0.86-1.70 .28‡ DSM-IV depression symptoms 1.8 (1.8) 1.3 (1.3) 1.26 0.84-1.89 .26‡ ∗ All clinical analyses were adjusted for the covariate child's age. † Critical P value corrected for multiple testing: P < .003. ‡ Transformed. Table II Clinical features of children with ADHD + ID (clinical sample) and those with ID only (ALSPAC sample)

Variable ADHD + ID (n = 97) ADHD only (n = 874) Statistics∗ n (%) Mean (SD) n (%) Mean (SD) OR 95% CI P† Age, y 11.4 (3.0) 10.0 (2.7) 1.19 1.11-1.28 4.6E-06 Male sex 84 (86.6) 739 (84.6) 0.85 0.46-1.56 .60 DSM-IV Combined ADHD diagnosis 79 (82.3) 631 (72.6) 1.96 1.12-3.40 .02 DSM-IV Inattentive ADHD diagnosis 5 (5.2) 51 (5.8) 0.62 0.23-1.62 .32 DSM-IV Hyperactive-Impulsive ADHD diagnosis 6 (6.2) 87 (10.0) 0.59 0.25-1.41 .23 DSM-III-R ADHD diagnosis only 5 (5.2) 89 (10.2) 0.49 0.19-1.26 .14 DSM-IV ODD diagnosis 36 (37.5) 399 (46.1) 0.74 0.48-1.14 .17 DSM-IV CD diagnosis 35 (36.1) 136 (15.6) 2.69 1.69-4.28 2.8E-05 ADHD symptoms: inattentive 7.5 (1.5) 7.3 (1.7) 1.25 0.75-2.09 .39‡ ADHD symptoms: hyperactive-impulsive 7.8 (1.4) 7.8 (1.5) 1.38 0.79-2.40 .25‡ ADHD symptoms: total 15.4 (2.1) 15.1 (2.4) 1.08 0.98-1.18 .12 DSM-IV ODD symptom count 4.1 (2.3) 3.8 (2.4) 1.09 0.99-1.19 .08 DSM-IV CD symptom count 2.1 (2.2) 1.0 (1.5) 2.38 1.71-3.32 2.5E-07‡ DSM-IV anxiety symptoms 1.2 (2.0) 1.1 (1.9) 1.21 0.86-1.70 .28‡ DSM-IV depression symptoms 1.8 (1.8) 1.3 (1.3) 1.26 0.84-1.89 .26‡ ∗ All clinical analyses were adjusted for the covariate child's age. † Critical P value corrected for multiple testing: P < .003. ‡ Transformed. Table II Clinical features of children with ADHD + ID (clinical sample) and those with ID only (ALSPAC sample) Variables ADHD + ID (n = 97) ID-only (n = 58) Statistics n (%) Mean (SD) n (%) Mean (SD) OR 95% CI P Age, y 11.4 (3.0) 10.8 (0.1) 1.10 0.96-1.27 .16 Male sex 84 (86.6) 31 (53.4) 0.18 0.08-0.39 1.4E-05 ODD symptom count 4.1 (2.3) 0.2 (0.7) 5.54 2.86-10.75 4.0E-07∗ CD symptom count (out of 7) 1.2 (1.4) 0.0 (0.3) 13.66 3.25-57.42 3.6E-04∗ ODD diagnosis 47 (49.0) 2 (3.4) 30.99 6.38-150.39 2.0E-05∗ CD diagnosis 16 (16.7) 0 (0.0) NA NA NA† NA, not applicable.

96-1.27 .16 Male sex 84 (86.6) 31 (53.4) 0.18 0.08-0.39 1.4E-05 ODD symptom count 4.1 (2.3) 0.2 (0.7) 5.54 2.86-10.75 4.0E-07∗ CD symptom count (out of 7) 1.2 (1.4) 0.0 (0.3) 13.66 3.25-57.42 3.6E-04∗ ODD diagnosis 47 (49.0) 2 (3.4) 30.99 6.38-150.39 2.0E-05∗ CD diagnosis 16 (16.7) 0 (0.0) NA NA NA† NA, not applicable. ∗ All clinical analyses adjusted for the covariates child's age at time of assessment, sex, and IQ. † Because there were no CD diagnoses in the ID-only group, statistical calculation was not applicable.

There is a need to have effective lifestyle modifications that target the growing group of obese children with dyslipidemia. The beneficial health effects of plant-based diets in adults are known. Studies have suggested that a low-fat, vegan diet (no animal products) may promote weight loss, lower body mass index (BMI), and improve lipoprotein profiles and insulin sensitivity and possibly prevent CVD.6–10 Those who follow a vegetarian diet (no animal products except for dairy and/or eggs) typically have lower cholesterol levels and a lower risk for coronary heart disease than non-vegetarians.11–13 Additionally, vegetarian diets have been shown to not only prevent but reverse heart disease in adults.15, 16 Whether the benefits of a plant-based (only plants and whole grains, limited avocado and nuts) no added fat diet (PB) extend to children is not known. We therefore conducted a four-week randomized trial comparing a PB with the American Heart Association diet (AHA)17 in children ages 9–18 with BMI >95% and total cholesterol >169mg/dL and one of their parents. Similar to the PB, the AHA encourages fruits, vegetables, whole grains and low sodium intake but permits non-whole grains, low-fat dairy, selected plant oils, and lean meat and fish in moderation.

eart Association diet (AHA)17 in children ages 9–18 with BMI >95% and total cholesterol >169mg/dL and one of their parents. Similar to the PB, the AHA encourages fruits, vegetables, whole grains and low sodium intake but permits non-whole grains, low-fat dairy, selected plant oils, and lean meat and fish in moderation. The aim of this study was to determine if a PB and/or AHA significantly change anthropometric measurements and/or biomarkers of inflammation and CVD risk after a 4-week intervention in obese, hypercholesterolemic children age 9–18 years old and one of their parents. Our hypothesis was that both groups would show improvement in the studied outcomes and the improvements might be greater for the PB than AHA. METHODS This was a prospective randomized 4-week trial (from April 20, 2013, to May 18, 2013) of either a PB or AHA. It was approved by our institutional review board. We enrolled 30 children seen in a large Midwestern hospital system’s predominantly middle class pediatric practices between the ages of 9–18 years with a last recorded BMI greater than the 95th percentile for age and sex and most recent total cholesterol greater than 169 mg/dL. A parent or guardian also participated in the study and was assigned to follow the same diet that was given to his/her child to help with dietary compliance. Pregnant women were excluded from the study.

corded BMI greater than the 95th percentile for age and sex and most recent total cholesterol greater than 169 mg/dL. A parent or guardian also participated in the study and was assigned to follow the same diet that was given to his/her child to help with dietary compliance. Pregnant women were excluded from the study. A computerized search of Cleveland Clinic medical records identified 1,278 potential participants (Figure 1; available at www.jpeds.com). Eligible patients were invited by letter to participate in the study. Those interested contacted the principal investigator (PI) and were enrolled on a first come, first served basis. Informed consent was obtained from the participants who were eighteen years of age and older. Participants younger than eighteen years provided assent with parent/guardian approval. There was a time gap between the last recorded measurements, obtaining informed consent, and the start of the study. During this time gap before the start of the study 6 previously obese (BMI >95%) children had become overweight (BMI 85%–95%), and one hypercholesterolemic (>169 mg/dL) child’s cholesterol had decreased to 169 mg/dL. Each child and parent pair received a fifty dollar stipend for each of the four weeks of the study.

dy. During this time gap before the start of the study 6 previously obese (BMI >95%) children had become overweight (BMI 85%–95%), and one hypercholesterolemic (>169 mg/dL) child’s cholesterol had decreased to 169 mg/dL. Each child and parent pair received a fifty dollar stipend for each of the four weeks of the study. Participants assigned to the PB were instructed to avoid all animal products and added fat, and to limit intake of nuts and avocado.15 The AHA group was allowed 30% of calories from total fat, 7% of calories from saturated fat, less than 300 mg cholesterol and less than 1,500 mg of sodium daily.17 All participants received standardized teaching at the time of consent to learn how to record a 24-hour dietary history. Participants completed two 3-day dietary histories consisting of two weekdays and one weekend day; one before the start of the study and one during the study. During the study, participants attended a total of 4 weekly 2-hour classes specific to their assigned diet consisting of one hour of nutrition education and one hour of cooking lessons with recipes provided.

ries consisting of two weekdays and one weekend day; one before the start of the study and one during the study. During the study, participants attended a total of 4 weekly 2-hour classes specific to their assigned diet consisting of one hour of nutrition education and one hour of cooking lessons with recipes provided. Classes were led by acknowledged study collaborators. Weeks one and two focused on reading labels, where to buy food, food preparation, and how to stay on the assigned diets when eating away from home. Weeks three and four reviewed healthy food choices, the effects of diet on health, discussions of what worked and what did not work for the study participants. At the fifth and final study session, after all laboratory samples and measurements were obtained, participants had the option to attend an introductory class on the diet they were not assigned.

thy food choices, the effects of diet on health, discussions of what worked and what did not work for the study participants. At the fifth and final study session, after all laboratory samples and measurements were obtained, participants had the option to attend an introductory class on the diet they were not assigned. At the start of the 4-week trial, fasting blood samples for biomarkers of inflammation, and CVD risk were obtained. The biomarkers of inflammation were myeloperoxidase [MPO] and high sensitivity C-reactive protein (hsCRP)— which are biomarkers for inflammation and cardiovascular risk in prepubescent obese children and adults,18, 19 as well as IL-6, ALT, and AST. The CVD risk biomarkers were total cholesterol, triglycerides, high-density lipoprotein cholesterol (HDL-C), and low-density lipoprotein cholesterol (LDL-C), and included biomarkers for diabetes; HgbA1c, fasting glucose and insulin. Laboratory analysis included total cholesterol, triglycerides, low-density lipoprotein cholesterol (LDL-C), high-density lipoprotein cholesterol (HDL-C) by standard enzymatic methodology, hemoglobin A1c (HgbA1c) (in percentages %) by turbidimetric inhibition immunoassay, insulin by chemiluminescence immunoassay, high sensitivity C-reactive protein (hs-CRP) by immunoturbidometric assay and fasting plasma glucose by glucose hexokinase method. All analyses were performed in the Preventive Research Laboratory and Lab Diagnostic Core, Cleveland Clinic.

At the start of the 4-week trial, fasting blood samples for biomarkers of inflammation, and CVD risk were obtained. The biomarkers of inflammation were myeloperoxidase [MPO] and high sensitivity C-reactive protein (hsCRP)— which are biomarkers for inflammation and cardiovascular risk in prepubescent obese children and adults,18, 19 as well as IL-6, ALT, and AST. The CVD risk biomarkers were total cholesterol, triglycerides, high-density lipoprotein cholesterol (HDL-C), and low-density lipoprotein cholesterol (LDL-C), and included biomarkers for diabetes; HgbA1c, fasting glucose and insulin. Laboratory analysis included total cholesterol, triglycerides, low-density lipoprotein cholesterol (LDL-C), high-density lipoprotein cholesterol (HDL-C) by standard enzymatic methodology, hemoglobin A1c (HgbA1c) (in percentages %) by turbidimetric inhibition immunoassay, insulin by chemiluminescence immunoassay, high sensitivity C-reactive protein (hs-CRP) by immunoturbidometric assay and fasting plasma glucose by glucose hexokinase method. All analyses were performed in the Preventive Research Laboratory and Lab Diagnostic Core, Cleveland Clinic. Measurements (height, weight, mid-arm circumference, waist circumference, and blood pressure) were also obtained at the start of the trial. BMI was calculated by dividing weight in kilograms by height in meters squared. Measurements of the physical activity of the children and adolescents were self-reported using the Physical Activity Questionnaire.20 The PAQ consists of 9 questions which ask subjects to rate their physical activity for the previous 7 days, at different times of day and days of week, and how often they engaged in specific activities. All items are presented on a 5-point scale where 1 is low activity and 5 is high activity; the overall PAQ score is a mean of the 9 questions. All measurements were repeated at the completion of the study for comparison with baseline. Race/ethnicity was self-reported to help determine the comparability of the study groups. At the conclusion of the four-week trial, all participants completed a validated Food Acceptability Questionnaire,21 which subjectively rated the ease of following their assigned diets and their general like or dislike of the diet.

ace/ethnicity was self-reported to help determine the comparability of the study groups. At the conclusion of the four-week trial, all participants completed a validated Food Acceptability Questionnaire,21 which subjectively rated the ease of following their assigned diets and their general like or dislike of the diet. The sample size of 15 adults and children per group was calculated to substantially exceed, even with a 20% drop-out rate, the 6–7 patients per group required to provide a power of 90% at a significance level of 0.05 to detect the within-group changes from baseline in total cholesterol described previously (mean ± standard deviation decrease of 60±26 mg/dl)14 versus a null hypothesis mean decrease of 25±26mg/dl. We did not power our study to demonstrate statistically significant differences between two effective dietary interventions. Families were randomized to the two study groups in a 1:1 ratio in blocks of four families, stratified by the child’s age group (age strata 9–13 years vs. 14–18 years). The randomization was performed by Ms. Worley using an SAS computer program between the end of enrollment and the first weekly session. Demographics, comorbidities, and body measurements were collected in a REDCap database,22 using double data entry. Laboratory values were provided in an Excel sheet. Diet journals were entered into and analyzed using Nutrition Data System for Research (NDSR) software.

The sample size of 15 adults and children per group was calculated to substantially exceed, even with a 20% drop-out rate, the 6–7 patients per group required to provide a power of 90% at a significance level of 0.05 to detect the within-group changes from baseline in total cholesterol described previously (mean ± standard deviation decrease of 60±26 mg/dl)14 versus a null hypothesis mean decrease of 25±26mg/dl. We did not power our study to demonstrate statistically significant differences between two effective dietary interventions. Families were randomized to the two study groups in a 1:1 ratio in blocks of four families, stratified by the child’s age group (age strata 9–13 years vs. 14–18 years). The randomization was performed by Ms. Worley using an SAS computer program between the end of enrollment and the first weekly session. Demographics, comorbidities, and body measurements were collected in a REDCap database,22 using double data entry. Laboratory values were provided in an Excel sheet. Diet journals were entered into and analyzed using Nutrition Data System for Research (NDSR) software. STATISTICAL ANALYSES Mean daily nutrients were computed for each subject within the pre-study and during-study periods. The BMI of the children was converted to age- and sex- adjusted percentiles and their corresponding z-scores; statistical analysis was performed on the z- scores. Parent and child subjects were analyzed separately because their outcomes were likely to be correlated, given genetic and environmental similarities. For the primary analysis, within- group changes from baseline to week 4 were computed and their means estimated with 95% confidence intervals; log-transformations of baseline and week 4 values of variables were performed as needed. For the secondary analysis, the PB and AHA groups were compared at the end of the trial, adjusting for baseline values, using analysis of covariance (ANCOVA) models. Where needed to meet model assumptions, both the baseline and week 4 values of variables were log-transformed. Study groups were compared on responses to each question on the food acceptability questionnaire using Fisher’s exact tests and Cochran-Armitage trend tests. Sample sizes for individual variables reflect missing data. All analyses were performed on a complete- case basis. All tests were two-tailed and performed at a significance level of 0.05. SAS 9.2 software (SAS Institute, Cary, NC) was used for all analyses and R 3.0.0 (The R Foundation for Statistical Computing) was used for plots.

for individual variables reflect missing data. All analyses were performed on a complete- case basis. All tests were two-tailed and performed at a significance level of 0.05. SAS 9.2 software (SAS Institute, Cary, NC) was used for all analyses and R 3.0.0 (The R Foundation for Statistical Computing) was used for plots. RESULTS Sixteen families were randomized to the PB and 14 families to the AHA. Two families, both in the PB group were lost to follow-up. One discontinued after the first week, and the other after the third week. Both families were excluded from the analysis because no end of study data was available. The final study cohort consisted of 28 families, 14 in each group (Figure 1). There were no significant between group differences in baseline demographic, nutrient, and clinical outcomes (Tables I–III; Tables II and III available at www.jpeds.com).

excluded from the analysis because no end of study data was available. The final study cohort consisted of 28 families, 14 in each group (Figure 1). There were no significant between group differences in baseline demographic, nutrient, and clinical outcomes (Tables I–III; Tables II and III available at www.jpeds.com). The total energy intake and the intake of almost all measured nutrients significantly decreased in children and adults in both groups, and dietary fiber intake significantly increased only in PB diet group (both children and adults) based on dietary histories completed during the study compared with those completed at baseline (Figure 2 and Table II). When comparing the PB and AHA groups during the study, children and adults in the PB group had a significantly lower intake of total protein, animal protein, cholesterol, total saturated fat, vitamin D, vitamin B12, percent of calories from fat and percent of calories from saturated fat. Children and adults in the PB group also had a significantly higher intake of total carbohydrates and dietary fiber than children in the AHA group. During the study, children and adults of both groups significantly reduced and increased intakes of the same nutrients, except for a trans-fat decrease only in adults on PB. Within-study energy intake and total fat intake were not significantly different between the two study groups, in children or adults.

e AHA group. During the study, children and adults of both groups significantly reduced and increased intakes of the same nutrients, except for a trans-fat decrease only in adults on PB. Within-study energy intake and total fat intake were not significantly different between the two study groups, in children or adults. Two goals for the children on PB were to consume no animal products and add no fat. During the study the mean (standard deviation) daily reported animal protein intake decreased from 42.32 (13.21) g to 2.24 (4.45) g (P <0.001) and the % of calories from fat and saturated fat was 18.04% (8.56%) and 3.59% (2.17%) respectively. The goals for the AHA children’s group were to consume <30% of total calories from fat, <7% of calories from saturated fat, <1,500 mg sodium, and <300 mg cholesterol. The respective mean (standard deviation) reported values during the study were 25.38% (6.12%), 7.59% (2.38%), 1,699 (897.71) mg and 144 (105.57) mg. Adults in both groups reported similar changes (Table II). These results suggest good but not perfect compliance with the assigned diets.

sodium, and <300 mg cholesterol. The respective mean (standard deviation) reported values during the study were 25.38% (6.12%), 7.59% (2.38%), 1,699 (897.71) mg and 144 (105.57) mg. Adults in both groups reported similar changes (Table II). These results suggest good but not perfect compliance with the assigned diets. On The Food Acceptability Questionnaire18 using a seven-point response scale both children and parents in the PB group reported more difficulty purchasing the necessary food for their diet than the children and parents in the AHA group. Median difficulty ratings were 3 (“Slightly difficult”) in the PB group and 5 (“Fairly easy”) in the AHA group for both parent and child subject. Mean PB vs. AHA ratings were 3.7 vs. 5.1 for children and 3.5 vs. 5.1 for parents. There were no other statistically significant differences between the groups on how well they liked these foods, liked the taste, appearance appeal, how boring, ease of preparation, ease of maintaining diet at restaurants, effort to stay on diet, effect on cost of food purchases, satisfaction felt after meals, and overall satisfaction.

er statistically significant differences between the groups on how well they liked these foods, liked the taste, appearance appeal, how boring, ease of preparation, ease of maintaining diet at restaurants, effort to stay on diet, effect on cost of food purchases, satisfaction felt after meals, and overall satisfaction. In the group of children on PB, there were statistically significant (P<0.05) mean decreases in nine measures: BMI Z-score (−0.14), systolic blood pressure (−6.43 mm Hg), weight (−3.05 kg), mid-arm circumference (−2.02 cm), total cholesterol (−22.5 mg/dL), LDL (−13.14 mg/dL), hsCRP (−2.09 mg/L), MPO (−75.34 pmol/L), and insulin (−5.42 uU/ml). In the AHA children group, there were statistically significant mean decreases in 5 measures: weight (−1.55 kg), waist circumference (−2.96 cm), mid-arm circumference (−1.14 cm), high density lipoprotein (−2.93 mg/dL), and MPO (−69.23 pmol/L). Both groups had statistically significant increases in Hgb A1c (PB +0.17%, AHA +0.21%) (Figure 3 and Table III). In the group of adults on PB, there were statistically significant (P<0.05) mean decreases in eight measures: BMI (−1.29kg/m2), systolic blood pressure (−7.96mm Hg), weight (−3.64kg), mid-arm circumference (−1.32cm), total cholesterol (−33.79mg/dL), HDL (−8.14mg/dL), LDL (−27.0mg/dL), and Hgb A1C (−0.16%). In the AHA adult group, there were statistically significant mean decreases in three measures: BMI (−0.73kg/m2), weight (−2.01kg), and HDL (−4.93mg/dL). The only statistically significant mean increase was AST (+4.43 U/L) in the AHA group (Figure 3 and Table III).

L (−8.14mg/dL), LDL (−27.0mg/dL), and Hgb A1C (−0.16%). In the AHA adult group, there were statistically significant mean decreases in three measures: BMI (−0.73kg/m2), weight (−2.01kg), and HDL (−4.93mg/dL). The only statistically significant mean increase was AST (+4.43 U/L) in the AHA group (Figure 3 and Table III). The only statistically significant differences between the PB and AHA groups after the intervention were that the children in the PB group had significantly lower week 4 BMI Z-scores and hsCRP levels. Parents in the PB group had significantly lower total cholesterol, LDL, and Hgb A1C than parents in the AHA group. The only significant change favoring AHA was a 1% difference in children’s waist circumference. The primary analysis of our study was of evaluable cases.

The only statistically significant differences between the PB and AHA groups after the intervention were that the children in the PB group had significantly lower week 4 BMI Z-scores and hsCRP levels. Parents in the PB group had significantly lower total cholesterol, LDL, and Hgb A1C than parents in the AHA group. The only significant change favoring AHA was a 1% difference in children’s waist circumference. The primary analysis of our study was of evaluable cases. In this study of PB in children we were most interested in determining if adhering to a PB would improve cardiovascular risk. A secondary intent-to-treat analysis was performed including the two child/parent pairs on PB who failed to complete the study. For those pairs, we assumed that there was no change from baseline in outcomes where baseline measures were available, and for measures with unavailable baseline data (insulin, IL-6, and myeloperoxidase), we assigned the median baseline value of the measure, computed separately for all parent and all child subjects, to both the baseline and end of study values. There were no differences between the evaluable-case analysis and intent-to-treat analysis in the statistical significance of within- group changes in outcome. The only differences between the evaluable-case analysis and intent- to-treat analysis in the statistical significance of the PB and AHA group comparisons were that children in the PB group no longer had significantly lower week 4 hsCRP or vitamin D intake, and parents in the PB no longer had significantly lower total protein or percent of calories from fat (data not shown).

and intent- to-treat analysis in the statistical significance of the PB and AHA group comparisons were that children in the PB group no longer had significantly lower week 4 hsCRP or vitamin D intake, and parents in the PB no longer had significantly lower total protein or percent of calories from fat (data not shown). DISCUSSION We believe our inability to demonstrate more than a few significant differences between our intervention groups, versus many significant differences from baseline values, most likely reflects the fact our study was powered to detect changes from baseline values in the PB group with 4 weeks intervention.14

and intent- to-treat analysis in the statistical significance of the PB and AHA group comparisons were that children in the PB group no longer had significantly lower week 4 hsCRP or vitamin D intake, and parents in the PB no longer had significantly lower total protein or percent of calories from fat (data not shown). DISCUSSION We believe our inability to demonstrate more than a few significant differences between our intervention groups, versus many significant differences from baseline values, most likely reflects the fact our study was powered to detect changes from baseline values in the PB group with 4 weeks intervention.14 Statistically significant differences that occurred between baseline and week 4 that might be of possible concern for increased cardiovascular risk include the elevation of Hgb A1c in both the plant-based and AHA children groups. However, the only statistically significant change in HOMA-IR in our study was in the PB children’s group with a mean ratio (95% CI) of −1.25 (−2.01, −0.99), P-value 0.004, which suggests significantly decreased insulin resistance. In the PB children’s group FIRI (fasting immunoreactive insulin) and in the adult PB group HgbA1c decreased significantly. All these values suggest, if anything, a decreased risk of diabetes in the PB group. Other studies have documented beneficial effects of PB on diabetes.6, 7, 24 The statistically significant decreases of HDL in both the adult and children AHA group and the adult PB group are also of some concern. The decrease in serum HDL cholesterol after 4 weeks was most likely associated with early weight loss.25 Vegan diets have been previously reported to be associated with decreased HDL, but vegan diets are also associated with a decreased risk of heart disease.8,15, 26 However, we are unaware of similar reports of lowering HDL on the AHA diet. There is heterogeneity described in the composition and function of HDL, and further characterization of HDL after exposure to these diets, might prove helpful in establishing the significance of these findings of decreased HDL.27, 28 The statistically significant elevation of AST in the adults on the AHA diet is of unknown significance. Mild, transient increases in ALT and AST values have been observed in weight loss studies, but usually returned to below baseline levels after substantial weight loss.29, 30

of these findings of decreased HDL.27, 28 The statistically significant elevation of AST in the adults on the AHA diet is of unknown significance. Mild, transient increases in ALT and AST values have been observed in weight loss studies, but usually returned to below baseline levels after substantial weight loss.29, 30 Statistically significant reported differences in the PB and AHA diet in adults and children of decreased cholesterol, fat and calorie intake and increased fiber on the PB diet most likely were of benefit to the obese hypercholesterolemic children studied.9, 24 Previous reports demonstrate that dietary records, especially in individuals with higher BMI, commonly underestimate intake by nearly 50%, so the actual intake is difficult to know for certain. The reported decrease of protein intake, especially in the context of probable under-reporting of intake, is not of concern as PB diets have been shown to provide adequate protein and most other nutrients in adults and children.31–33 The decreased intake of Vitamins B12 and D reported in the adult and child PB groups found in the current study have been noted previously. Vitamin B 12 definitely and Vitamin D probably warrant supplements for those on long-term PB diets. Intake of key nutrients is generally adequate in a balanced vegan diet, but it is still essential to monitor consumption of protein, n-3 fatty acids, iron, zinc, iodine, and calcium in long-term vegans.31–33 Dietary intake of n-3 long-chain polyunsaturated fatty acids (PUFAs) eicosapentaenoic acid and docosahexaenoic acid, is low in vegans compared with omnivores.34 Therefore, n-3 fatty acids, particularly in a no added fat vegan diet, should be especially carefully monitored and may also require supplementation with algae-derived n-3 PUFAs. The only significant described problem in our middle class study population for diet acceptance was the difficulty purchasing food. Cost may be a barrier to a PB in lower socioeconomic populations. In another study the only identified barrier to adherence was the effort required.35 If the PB diet is to achieve ever increasing adaptation, barriers to easy, affordable access to plant-based, no added fat foods will need to be reduced.

purchasing food. Cost may be a barrier to a PB in lower socioeconomic populations. In another study the only identified barrier to adherence was the effort required.35 If the PB diet is to achieve ever increasing adaptation, barriers to easy, affordable access to plant-based, no added fat foods will need to be reduced. The major limitations of our study are that this was a small study conducted for a short period of time in a select group of middle class patients with less than completely reliable measures of compliance and with no direct health outcomes measures. Also, the AHA is considered a standard of care and was used as a comparison group—there was no placebo- controlled group. There is also concern that long-term compliance with the PB could be problematic. This is especially true given the difficulties expressed by parents and children in finding food to purchase for the diet in our study, and in the effort required to maintain a PB in a previous study.21 However, there are studies describing good acceptability and compliance with a PB. 15,16,35,36

could be problematic. This is especially true given the difficulties expressed by parents and children in finding food to purchase for the diet in our study, and in the effort required to maintain a PB in a previous study.21 However, there are studies describing good acceptability and compliance with a PB. 15,16,35,36 Plant-based diets are generally recognized as safe for children and adolescents as long as the intake of key nutrients is monitored and appropriate supplements are provided. The results of our study suggest that the documented benefits of PB in adults, including, but not limited to, decreased overweight and obesity and decreased cardiovascular risk, most likely would be seen in children. These benefits, especially given the known onset of CVD in childhood, could improve the lifetime health of those populations who choose to eat a PB beginning in childhood. Supported by National Center for Advancing Translational Sciences of the National Institutes of Health (UL1TR000439) and Research Program Committee and Pediatric Research Fund Grants from the Cleveland Clinic.

Plant-based diets are generally recognized as safe for children and adolescents as long as the intake of key nutrients is monitored and appropriate supplements are provided. The results of our study suggest that the documented benefits of PB in adults, including, but not limited to, decreased overweight and obesity and decreased cardiovascular risk, most likely would be seen in children. These benefits, especially given the known onset of CVD in childhood, could improve the lifetime health of those populations who choose to eat a PB beginning in childhood. Supported by National Center for Advancing Translational Sciences of the National Institutes of Health (UL1TR000439) and Research Program Committee and Pediatric Research Fund Grants from the Cleveland Clinic. Alicia Thomas, MS, RD, LD (Case Western Reserve University), taught the American Heart Association Diet and contributed to the study design. Jane Esselstyn, RN, is an author and teacher who taught the Plant-Based No Added Fat Diet. Stacy Payne, BS, Megan Villarreal, BA, and Kay Stelmach, RN, helped coordinate the study and with members of their Cleveland Clinic Clinical Research Unit aided in data collection. Alan Pratt, MT (ASCP) (Preventive Research Laboratory and Lab Diagnostic Core, Cleveland Clinic), managed laboratory analysis. Amy Moore, BA (Department of Scientific Publications, Cleveland Clinic), provided medical editing. Lorelei Woody, MLIS (Cleveland Clinic Alumni Library), helped with manuscript preparation. Vaishal Shah, MD, MPH (Cleveland Clinic Children’s), did data entry and collection, and Stephanie Bernstein, BSDH, also helped with data entry. Sana Mansoor, MD, Alex Golden, MD, Asif Padiyath, MBBS, Kari Gali, MSN, RN, CPNP (all from Cleveland Clinic Children’s), performed data collection. Francette Malone provided secretarial support, and June Cassano, RN, BSN, MBA (both from Cleveland Clinic Children’s), helped with budget preparation. John Kirwan, PhD (Department of Pathobiology Cleveland Clinic and Department of Nutrition Case Western Reserve University, funded by NIH RO1 AG12834 and DK089547, and UL1 RR024989), facilitated the participation of Tammie Kong, Adam Weier, and Alicia Thomas and reviewed the initial study design.

ith budget preparation. John Kirwan, PhD (Department of Pathobiology Cleveland Clinic and Department of Nutrition Case Western Reserve University, funded by NIH RO1 AG12834 and DK089547, and UL1 RR024989), facilitated the participation of Tammie Kong, Adam Weier, and Alicia Thomas and reviewed the initial study design. No reprints are available The authors declare no conflicts of interest. Registered with ClinicalTrials.gov: NCT01817491 This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final citable form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain. LIST OF ABBREVIATIONS AND ACRONYMS CVDCardiovascular disease PBPlant-based (only plants and whole grains, limited avocado and nuts) no added fat diet AHAAmerican Heart Association Diet (also encourages fruits, vegetables, whole grains and low sodium intake but permits non-whole grains, low-fat dairy, selected plant oils, and lean meat and fish in moderation.) HOMA-IRHomeostasis Model Assessment –Insulin Resistance=FIRI (fasting immunoreactive insulin mU/l)×FPG (fasting plasma glucose mmol/l)/22.5 Figure 1 Consort Diagram Figure 2 Nutrient Outcomes Within Groups SFA-Saturated Fatty Acids, Trans-Trans Fats Figure 3 Clinical Outcomes Within Groups SBP-Systolic Blood Pressure, DBP-Diastolic Blood Pressure

HOMA-IRHomeostasis Model Assessment –Insulin Resistance=FIRI (fasting immunoreactive insulin mU/l)×FPG (fasting plasma glucose mmol/l)/22.5 Figure 1 Consort Diagram Figure 2 Nutrient Outcomes Within Groups SFA-Saturated Fatty Acids, Trans-Trans Fats Figure 3 Clinical Outcomes Within Groups SBP-Systolic Blood Pressure, DBP-Diastolic Blood Pressure Table 1 Demographics Factor Plant-based (N=14) AHA (N=14) Child: male 5(36) 5(36) Child: Age, years 15.0(9.0,18.0) 15.0(9.0,18.0) Child: race • None given 1(7) 0(0) • White 10(71) 10(71) • Black 2(14) 4(29) • Asian 1(7) 0(0) Child: Hispanic ethnicity* 0(0) 1(9) Child: BMI at baseline • Overweight 4(29) 2(14) • Obese 10(71) 12(86) Child: Blood pressure at baseline • Normal 8(57) 8(57) • Prehypertension 5(36) 4(29) • Hypertension 1(7) 2(14) Child: High cholesterol at baseline (>169) 14(100) 13(93) Parent: male 5(36) 4(29) Parent: Age (years) 46.5(37.0,61.0) 46.0(35.0,51.0) Parent: race • White 10(71) 9(64) • Black 3(21) 4(29) • Asian 1(7) 0(0) • More than one race 0(0) 1(7) Parent: Hispanic ethnicity* 0(0) 1(10) Parent: BMI at baseline • Normal weight 1(7) 3(21) • Overweight 4(29) 3(21) • Obese 9(64) 8(57) Parent: Blood pressure at baseline • Normal 7(50) 3(21) • At risk 3(21) 8(57) • High 4(29) 3(21) Parent: High cholesterol at baseline (>199) 8(57) 8(57) * Data not available for all subjects. Missing values: Child: Hispanic ethnicity = 6, Parent: Hispanic ethnicity = 8. Values presented as Means with SDs indicated in parenthesis, Median [P25, P75], Median (min, max) or N (column %).

Factor Plant-based (N=14) AHA (N=14) Child: male 5(36) 5(36) Child: Age, years 15.0(9.0,18.0) 15.0(9.0,18.0) Child: race • None given 1(7) 0(0) • White 10(71) 10(71) • Black 2(14) 4(29) • Asian 1(7) 0(0) Child: Hispanic ethnicity* 0(0) 1(9) Child: BMI at baseline • Overweight 4(29) 2(14) • Obese 10(71) 12(86) Child: Blood pressure at baseline • Normal 8(57) 8(57) • Prehypertension 5(36) 4(29) • Hypertension 1(7) 2(14) Child: High cholesterol at baseline (>169) 14(100) 13(93) Parent: male 5(36) 4(29) Parent: Age (years) 46.5(37.0,61.0) 46.0(35.0,51.0) Parent: race • White 10(71) 9(64) • Black 3(21) 4(29) • Asian 1(7) 0(0) • More than one race 0(0) 1(7) Parent: Hispanic ethnicity* 0(0) 1(10) Parent: BMI at baseline • Normal weight 1(7) 3(21) • Overweight 4(29) 3(21) • Obese 9(64) 8(57) Parent: Blood pressure at baseline • Normal 7(50) 3(21) • At risk 3(21) 8(57) • High 4(29) 3(21) Parent: High cholesterol at baseline (>199) 8(57) 8(57) * Data not available for all subjects. Missing values: Child: Hispanic ethnicity = 6, Parent: Hispanic ethnicity = 8. Values presented as Means with SDs indicated in parenthesis, Median [P25, P75], Median (min, max) or N (column %). Table 2 Nutrients Compared Between Plant-based and American Heart Association Diets

Factor Plant-based (N=14) AHA (N=14) Child: male 5(36) 5(36) Child: Age, years 15.0(9.0,18.0) 15.0(9.0,18.0) Child: race • None given 1(7) 0(0) • White 10(71) 10(71) • Black 2(14) 4(29) • Asian 1(7) 0(0) Child: Hispanic ethnicity* 0(0) 1(9) Child: BMI at baseline • Overweight 4(29) 2(14) • Obese 10(71) 12(86) Child: Blood pressure at baseline • Normal 8(57) 8(57) • Prehypertension 5(36) 4(29) • Hypertension 1(7) 2(14) Child: High cholesterol at baseline (>169) 14(100) 13(93) Parent: male 5(36) 4(29) Parent: Age (years) 46.5(37.0,61.0) 46.0(35.0,51.0) Parent: race • White 10(71) 9(64) • Black 3(21) 4(29) • Asian 1(7) 0(0) • More than one race 0(0) 1(7) Parent: Hispanic ethnicity* 0(0) 1(10) Parent: BMI at baseline • Normal weight 1(7) 3(21) • Overweight 4(29) 3(21) • Obese 9(64) 8(57) Parent: Blood pressure at baseline • Normal 7(50) 3(21) • At risk 3(21) 8(57) • High 4(29) 3(21) Parent: High cholesterol at baseline (>199) 8(57) 8(57) * Data not available for all subjects. Missing values: Child: Hispanic ethnicity = 6, Parent: Hispanic ethnicity = 8. Values presented as Means with SDs indicated in parenthesis, Median [P25, P75], Median (min, max) or N (column %). Table 2 Nutrients Compared Between Plant-based and American Heart Association Diets CHILDREN Energy (kcal) 1686.32 (650.59) 1208.02 (288.96) −478.30 (496.93) 1638.16 (484.17) 1115.90 (462.71) −522.26 (289.68) 1.14 (0.93, 1.38) 0.19 Outcome Plant-Based (N=14) Mean (SD) AHA (N=14) Mean (SD) PB - AHA Adjusted Mean Difference (95% CI) at Week 4 PB/AHA Adjusted Mean Ratio (95% CI) at Week 4 P- value Baseline During study Change Baseline During study Change Total Fat (g) 65.72 (30.97) 26.60 (17.90) −39.12 (26.57) 69.66 (20.34) 33.39 (18.31) −36.27 (12.93) 0.82 (0.57, 1.19) 0.29 Total Carbohydrate (g) 217.35 (97.86) 215.83 (58.11) −1.52 (74.58) 198.09 (73.71) 152.39 (73.06) −45.70 (46.51) 53.18 (15.48, 90.88) 0.008 Total Protein (g) 63.61 (17.76) 39.61 (6.92) −24.00 (16.98) 60.06 (20.81) 56.38 (17.83) −3.68 (20.89) −17.69 (−27.70, −7.68) 0.001 Animal Protein (g) 42.32 (13.21) 2.24 (4.45) −40.08 (14.35) 38.61 (17.38) 38.25 (15.48) −0.36 (21.99) −36.19 (−45.28, −27.10) <0.001 Cholesterol (mg) 181.23 (81.81) 12.93 (30.63) −168.30 (79.95) 169.90 (84.68) 143.70 (105.57) −26.20 (104.27) −134.32 (−192.59, −76.05) <0.001 Total Saturated Fatty Acids (SFA) (g) 21.97 (8.78) 5.46 (4.52) −16.51 (8.51) 22.41 (11.32) 10.20 (6.75) −12.22 (7.32) 0.49 (0.33, 0.75) 0.002 Total Trans-Fatty Acids (TRANS) (g) 2.11 (1.22) 0.80 (1.03) −1.31 (1.63) 2.69 (1.24) 1.05 (0.54) −1.64 (1.28) 0.51 (0.24, 1.08) 0.078 Omega-3 Fatty Acids (g) 1.33 (1.16) 0.62 (0.51) −0.71 (1.17) 1.57 (0.83) 0.87 (0.41) −0.70 (0.81) −0.22 (−0.58, 0.13) 0.21 Total Dietary Fiber (g) 13.68 (4.27) 29.01 (7.91) 15.32 (6.30) 12.92 (5.15) 13.58 (7.65) 0.66 (3.71) 2.22 (1.79, 2.75) <0.001 Vitamin D (calciferol) (mcg) 4.08 (1.88) 1.47 (1.44) −2.61 (1.65) 3.78 (4.21) 2.45 (1.66) −1.33 (4.38) 0.53 (0.31, 0.92) 0.026 Vitamin B-12 (cobalamin) (mcg) 4.75 (3.47) 2.03 (2.55) −2.72 (3.33) 3.08 (1.69) 2.96 (1.40) −0.12 (1.74) 0.45 (0.25, 0.82) 0.011 Calcium (mg) 806.89 (327.11) 395.80 (84.63) −411.09 (297.61) 611.89 (315.44) 493.82 (249.68) −118.07 (309.23) 0.78 (0.60, 1.02) 0.069 Iron (mg) 16.32 (10.20) 15.39 (6.58) −0.92 (11.38) 11.72 (4.88) 10.33 (5.13) −1.39 (3.97) 4.12 (−0.57, 8.82) 0.082 Sodium (mg) 2841.92 (962.40) 1697.75 (401.15) −1144.17 (1160.52) 2842.88 (886.84) 1699.27 (897.71) −1143.61 (473.64) −1.20 (−499.3

) 611.89 (315.44) 493.82 (249.68) −118.07 (309.23) 0.78 (0.60, 1.02) 0.069 Iron (mg) 16.32 (10.20) 15.39 (6.58) −0.92 (11.38) 11.72 (4.88) 10.33 (5.13) −1.39 (3.97) 4.12 (−0.57, 8.82) 0.082 Sodium (mg) 2841.92 (962.40) 1697.75 (401.15) −1144.17 (1160.52) 2842.88 (886.84) 1699.27 (897.71) −1143.61 (473.64) −1.20 (−499.3 3, 496.93) 0.99 % Calories from Fat 33.48 (6.80) 18.04 (8.56) −15.44 (8.24) 36.99 (6.67) 25.38 (6.12) −11.61 (8.50) −6.15 (−11.98, −0.32) 0.039 % Calories from SFA 11.78 (4.15) 3.59 (2.17) −8.19 (4.39) 11.36 (2.47) 7.59 (2.38) −3.77 (1.80) 0.43 (0.31, 0.59) <0.001 PARENTS Energy (kcal) 1964.39 (950.71) 1235.70 (335.76) −728.70 (839.79) 1660.18 (422.06) 1142.41 (368.92) −517.77 (460.19) 1.07 (0.86, 1.32) 0.54 Total Fat (g) 84.93 (50.55) 25.84 (14.22) −59.09 (50.54) 70.90 (26.34) 33.93 (20.18) −36.96 (29.81) 0.75 (0.49, 1.13) 0.16 Total Carbohydrate (g) 218.33 (101.44) 222.89 (67.31) 4.56 (81.29) 190.82 (47.77) 158.45 (54.82) −32.38 (48.99) 52.30 (11.62, 92.97) 0.014 Total Protein (g) 83.12 (36.04) 41.33 (11.46) −41.79 (34.43) 69.31 (18.91) 57.32 (13.69) −11.99 (19.44) −17.71 (−27.61, −7.80) 0.001 Animal Protein (g) 54.22 (26.26) 1.30 (2.27) −52.93 (25.89) 46.43 (16.55) 35.40 (13.93) −11.03 (12.93) −35.41 (−42.90, −27.93) <0.001 Cholesterol (mg) 241.61 (161.29) 7.93 (14.13) −233.69 (155.19) 212.75 (89.96) 119.64 (67.33) −93.11 (117.02) −112.16 (−151.00, −73.33) <0.001 Total Saturated Fatty Acids (SFA) (g) 28.14 (17.45) 5.03 (4.21) −23.11 (16.49) 22.60 (9.85) 10.12 (5.62) −12.48 (11.48) 0.45 (0.29, 0.72) 0.002 Total Trans-Fatty Acids (TRANS) (g) 3.15 (2.21) 0.47 (0.28) −2.68 (2.24) 3.07 (2.57) 1.04 (0.49) −2.03 (2.62) 0.35 (0.18, 0.69) 0.004 Omega-3 Fatty Acids (g) 1.62 (0.62) 0.94 (0.67) −0.69 (0.96) 1.63 (0.68) 0.84 (0.72) −0.79 (0.69) 0.10 (−0.44, 0.64) 0.71 Total Dietary Fiber (g) 20.11 (9.59) 31.99 (8.24) 11.88 (11.32) 17.52 (7.53) 16.42 (6.50) −1.10 (7.22) 1.99 (1.55, 2.55) <0.001 Vitamin D (calciferol) (mcg) 2.67 (1.82) 0.96 (1.39) −1.72 (2.35) 3.71 (3.32) 2.87 (2.64) −0.84 (3.98) 0.45 (0.25, 0.80) 0.008 Vitamin B-12 (cobalamin) (mcg) 3.97 (2.35) 1.01 (1.72) −2.97 (2.54) 3.37 (1.56) 3.38 (3.04) 0.02 (3.35) 0.32 (0.18, 0.58) <0.001 Calcium (mg) 780.10 (333.35) 387.65 (170.62) −392.45 (350.28) 572.29 (123.52) 524.45 (245.01) −47.84 (160.68) 0.71 (0.49, 1.01) 0.059 Iron (mg) 14.76 (6.30) 14.94 (13.51) 0.19 (13.37) 11.54 (3.01) 11.19 (4.12) −0.35 (3.21) 1.10 (0.77, 1.57) 0.58 Sodium (mg) 3515.11 (1899.46) 1708.01 (464.17) −1807.11 (1748.47) 3235.59 (1109.71) 175

780.10 (333.35) 387.65 (170.62) −392.45 (350.28) 572.29 (123.52) 524.45 (245.01) −47.84 (160.68) 0.71 (0.49, 1.01) 0.059 Iron (mg) 14.76 (6.30) 14.94 (13.51) 0.19 (13.37) 11.54 (3.01) 11.19 (4.12) −0.35 (3.21) 1.10 (0.77, 1.57) 0.58 Sodium (mg) 3515.11 (1899.46) 1708.01 (464.17) −1807.11 (1748.47) 3235.59 (1109.71) 175 4.93 (787.72) −1480.66 (1241.75) −78.00 (−574.67, 418.67) 0.75 % Calories from Fat 35.73 (7.37) 17.40 (6.81) −18.33 (7.69) 36.58 (6.90) 24.44 (7.32) −12.14 (8.71) −6.77 (−12.07, −1.46) 0.015 % Calories from SFA 11.69 (4.22) 3.31 (1.73) −8.38 (3.78) 11.66 (2.91) 7.30 (2.16) −4.36 (4.01) −3.99 (−5.53, −2.45) <0.001 Table 3 Clinical Outcomes

66 (1241.75) −78.00 (−574.67, 418.67) 0.75 % Calories from Fat 35.73 (7.37) 17.40 (6.81) −18.33 (7.69) 36.58 (6.90) 24.44 (7.32) −12.14 (8.71) −6.77 (−12.07, −1.46) 0.015 % Calories from SFA 11.69 (4.22) 3.31 (1.73) −8.38 (3.78) 11.66 (2.91) 7.30 (2.16) −4.36 (4.01) −3.99 (−5.53, −2.45) <0.001 Table 3 Clinical Outcomes Outcome Plant-Based (N=14) Mean (SD) AHA (N=14) Mean (SD) PB - AHA Adjusted Mean Difference (95% CI) at Week 4 PB/AHA Adjusted Mean Ratio (95% CI) at Week 4 P-value Baseline Week 4 Change Baseline Week 4 Change CHILDREN BMI percentile 96.36 (2.63) 95.24 (2.89) −1.12 (1.69) 98.01 (1.81) 97.93 (1.79) −0.08 (0.42) −1.17 (−2.20, −0.14) 0.028 BMI Z score 1.90 (0.38) 1.77 (0.39) −0.14 (0.16) 2.19 (0.38) 2.17 (0.37) −0.03 (0.07) −0.13 (−0.24, −0.03) 0.017 Systolic BP (mm Hg) 116.25 (14.10) 109.82 (9.67) −6.43 (8.86) 111.50 (20.16) 106.36 (9.87) −5.14 (17.64) 1.87 (−4.41, 8.15) 0.55 Diastolic BP (mm Hg) 68.79 (9.61) 66.18 (8.75) −2.61 (10.25) 69.79 (11.56) 65.43 (7.43) −4.36 (14.67) 1.01 (0.92, 1.11) 0.83 Weight (kg) 79.62 (23.43) 76.57 (23.38) −3.05 (3.27) 87.71 (19.85) 86.16 (19.01) −1.55 (1.83) −1.71 (−3.80, 0.39) 0.11 Waist circ (cm) 97.11 (12.05) 95.58 (12.22) −1.53 (3.01) 104.89 (11.99) 101.93 (12.53) −2.96 (4.04) 1.32 (−1.66, 4.30) 0.37 Midarm circ (cm) 33.30 (5.02) 31.29 (4.25) −2.02 (2.17) 35.06 (3.64) 33.91 (3.37) −1.14 (1.63) −1.25 (−2.61, 0.11) 0.070 PAQ 2.44 (0.63) 2.66 (0.83) 0.22 (0.53) 2.36 (0.74) 2.20 (0.63) −0.16 (0.54) 0.39 (−0.02, 0.80) 0.060 CHOL (mg/dL) 213.79 (41.97) 191.29 (70.34) −22.50 (36.31) 194.07 (25.69) 177.57 (25.53) −16.50 (28.19) −10.34 (−36.48, 15.80) 0.42 TRIG (mg/dL) 139.93 (139.03) 114.43 (89.77) −25.50 (95.26) 119.29 (63.31) 106.14 (52.58) −13.14 (36.19) 1.01 (0.77, 1.32) 0.94 HDL (mg/dL) 55.43 (19.10) 49.50 (16.11) −5.93 (9.37) 43.36 (6.96) 40.43 (6.17) −2.93 (3.71) 0.17 (−5.07, 5.40) 0.95 LDL (mg/dL) 132.21 (29.04) 119.07 (54.22) −13.14 (35.54) 126.86 (19.06) 115.86 (18.45) −11.00 (23.31) 0.95 (0.80, 1.13) 0.54 GLU (mg/dL) 98.86 (45.38) 99.79 (48.89) 0.93 (6.01) 98.57 (49.50) 97.93 (46.81) −0.64 (5.02) 1.01 (0.96, 1.05) 0.81 hsCRP (mg/L) 5.11 (8.33) 3.03 (4.64) −2.09 (7.10) 4.09 (3.76) 6.87 (8.63) 2.78 (8.76) 0.46 (0.21, 1.00) 0.049 ALT (U/L) 18.50 (6.64) 19.29 (14.14) 0.79 (14.35) 25.36 (27.45) 24.21 (28.33) −1.14 (3.46) 1.00 (0.76, 1.32) 0.98 AST (U/L) 18.43 (3.55) 21.21 (8.40) 2.79 (8.96) 22.50 (13.95) 22.50 (17.77) 0.00 (4.69) 1.13 (0.92, 1.40) 0.23 IL-6 (pg/ml) 1.72 (1.10) 1.55 (1.34) −0.17 (0.89) 2.33 (2.50) 2.14 (1.97) −0.

0.21, 1.00) 0.049 ALT (U/L) 18.50 (6.64) 19.29 (14.14) 0.79 (14.35) 25.36 (27.45) 24.21 (28.33) −1.14 (3.46) 1.00 (0.76, 1.32) 0.98 AST (U/L) 18.43 (3.55) 21.21 (8.40) 2.79 (8.96) 22.50 (13.95) 22.50 (17.77) 0.00 (4.69) 1.13 (0.92, 1.40) 0.23 IL-6 (pg/ml) 1.72 (1.10) 1.55 (1.34) −0.17 (0.89) 2.33 (2.50) 2.14 (1.97) −0. 19 (3.14) 0.26 (0.05, 1.37) 0.11 MPO (pmol/L) 287.54 (147.35) 212.20 (138.87) −75.34 (37.34) 303.79 (89.92) 234.56 (87.95) −69.23 (72.92) 0.95 (0.79, 1.13) 0.52 HgbA1c 5.77 (1.60) 5.94 (1.63) 0.17 (0.17) 6.14 (1.82) 6.34 (1.82) 0.21 (0.12) 0.99 (0.97, 1.01) 0.51 Insulin (uU/ml) 17.33 (14.12) 11.91 (9.40) −5.42 (5.68) 23.19 (12.57) 26.36 (26.25) 3.16 (25.62) 0.70 (0.43, 1.12) 0.13 PARENTS BMI 33.26 (7.69) 31.98 (7.57) −1.29 (1.14) 37.08 (12.66) 36.35 (12.20) −0.73 (0.92) −0.69 (−1.47, 0.09) 0.082 Systolic BP (mm Hg) 125.18 (16.98) 117.21 (17.18) −7.96 (12.28) 121.64 (15.07) 118.50 (11.90) −3.14 (14.42) 0.97 (0.89, 1.05) 0.39 Diastolic BP (mm Hg) 78.25 (12.06) 74.79 (8.75) −3.46 (11.06) 79.29 (10.34) 72.64 (8.58) −6.64 (12.19) 1.03 (0.94, 1.13) 0.46 Weight (kg) 93.33 (27.18) 89.70 (26.28) −3.64 (3.41) 100.79 (32.16) 98.77 (30.73) −2.01 (2.66) −1.95 (−4.16, 0.26) 0.081 Waist circ (cm) 104.41 (15.05) 102.47 (18.20) −1.94 (5.80) 112.11 (24.55) 111.62 (25.13) −0.49 (5.73) −1.14 (−5.75, 3.48) 0.62 Midarm circ (cm) 35.52 (6.61) 34.20 (6.19) −1.32 (2.06) 35.89 (7.67) 36.24 (8.61) 0.35 (3.16) −1.68 (−3.80, 0.44) 0.11 CHOL (mg/dL) 210.43 (51.63) 176.64 (40.80) −33.79 (30.84) 214.43 (45.28) 207.29 (55.66) −7.14 (24.93) −27.29 (−48.68, −5.90) 0.014 TRIG (mg/dL) 130.07 (80.18) 136.29 (89.70) 6.21 (41.71) 117.21 (53.57) 134.07 (75.74) 16.86 (36.86) 0.95 (0.76, 1.19) 0.67 HDL (mg/dL) 56.14 (17.24) 48.00 (15.60) −8.14 (5.40) 59.57 (13.47) 54.64 (15.70) −4.93 (5.61) 0.94 (0.87, 1.02) 0.16 LDL (mg/dL) 128.36 (42.00) 101.36 (36.06) −27.00 (26.72) 131.36 (43.67) 125.86 (50.02) −5.50 (20.89) −21.92 (−40.37, −3.46) 0.022 GLU (mg/dL) 101.50 (31.28) 106.43 (53.88) 4.93 (24.65) 100.64 (22.69) 95.21 (14.36) −5.43 (12.40) 1.06 (0.97, 1.16) 0.20 hsCRP (mg/L) 3.60 (4.16) 3.36 (3.52) −0.24 (1.36) 5.23 (6.20) 5.44 (5.15) 0.21 (1.90) 0.68 (0.44, 1.07) 0.091 ALT (U/L) 21.14 (8.73) 22.00 (13.69) 0.86 (7.01) 20.64 (7.43) 25.21 (13.73) 4.57 (9.69) 0.85 (0.67, 1.08) 0.17 AST (U/L) 18.93 (4.41) 19.07 (5.59) 0.14 (5.70) 18.14 (5.04) 22.57 (7.71) 4.43 (6.73) 0.83 (0.68, 1.03) 0.084 IL-6 (pg/ml) 7.86 (21.09) 8.02 (20.83) 0.16 (0.90) 2.13 (1.53) 1.94 (1.06) −0.19 (0.85) 1.14 (0.83, 1.57) 0.40 MPO (pmol/L

22.00 (13.69) 0.86 (7.01) 20.64 (7.43) 25.21 (13.73) 4.57 (9.69) 0.85 (0.67, 1.08) 0.17 AST (U/L) 18.93 (4.41) 19.07 (5.59) 0.14 (5.70) 18.14 (5.04) 22.57 (7.71) 4.43 (6.73) 0.83 (0.68, 1.03) 0.084 IL-6 (pg/ml) 7.86 (21.09) 8.02 (20.83) 0.16 (0.90) 2.13 (1.53) 1.94 (1.06) −0.19 (0.85) 1.14 (0.83, 1.57) 0.40 MPO (pmol/L ) 297.59 (145.78) 314.50 (236.87) 16.91 (110.66) 256.14 (70.52) 257.91 (87.62) 1.78 (45.93) 0.93 (0.80, 1.10) 0.39 HgbA1c 5.89 (0.97) 5.73 (0.88) −0.16 (0.16) 5.81 (0.51) 5.94 (0.79) 0.14 (0.49) 0.96 (0.92, 1.00) 0.031 Insulin (uU/ml) 13.46 (8.25) 10.34 (6.67) −3.11 (8.56) 16.33 (12.33) 13.18 (7.96) −3.15 (5.47) 0.87 (0.67, 1.13) 0.27

uring different eras of neonatal care.2 Second, we hypothesized that mathematic attainment9,16 may be susceptible to country specific schooling and that outcomes may, thus, differ between cohorts; that is, prediction from one cohort to another may be less accurate compared with predictions of basic cognitive abilities. Methods Two prospective geographically defined birth cohorts were included, the BLS and the EPICure study. Descriptive characteristics of the BLS and EPICure study participants are in Table I.

e been born in the intermediate socioeconomic groups, and have a birthweight of ≥2500 grams (Table I; available at www.jpeds.com). The mean IQ of those subjects with full growth data was 98.9 points (95% CI, 98.4-99.4), compared with 96.0 (95% CI, 95.2-96.7) among those who did not have full anthropometric assessments. About 1 of every 3 mothers had ≤4 years of schooling, and most of the sample belonged to low-income families at birth (1 minimum wage in 1982 corresponded with US$50.00). The prevalence of low birthweight (birthweight < 2500 grams) was 7.2% and 14.2% of the studied subjects had a birthweight that was <10th percentile according to gestational age and sex. Concerning nutritional status in childhood, 12.8% of the subjects had a height-for-age z-score of ≤−2 SD at 2 years of age. The mean IQ at 30 years of age was 98 points and the average number of years of schooling was 11.4 (Table II; available at www.jpeds.com).

See related article, p 1417 Around 15 million babies worldwide (∼10% of all births) are born preterm (<37 weeks gestational age [GA]) each year. Changes in reproduction patterns and improved neonatal medicine have led to increased numbers of moderately (32-33 weeks GA) and late preterm (34-36 weeks GA) births and increased survival rates of those born very preterm (<32 weeks GA). Despite improved neonatal care, prematurity remains the leading cause of infant mortality and long-term morbidity today,1 and the high prevalence of cognitive problems (>20%) in preterm populations has not changed over the last 2 decades.2

) births and increased survival rates of those born very preterm (<32 weeks GA). Despite improved neonatal care, prematurity remains the leading cause of infant mortality and long-term morbidity today,1 and the high prevalence of cognitive problems (>20%) in preterm populations has not changed over the last 2 decades.2 Studies suggest that delivery at any gestation other than full-term may confer an insult to brain development3 rendering survivors at risk for adverse cognitive and educational outcomes, particularly in mathematics.4-6 It remains controversial whether the dose response effect of GA on early mathematical abilities is linear6 or curvilinear.7 Emerging evidence from different cohorts demonstrate a significant impact of GA at birth on basic cognitive abilities (eg, IQ, mathematic processing)8,9 and mathematic attainment,4,6,10 but there is uncertainty about its specific nature and magnitude. The relationship of GA with cognitive and educational outcomes may be affected by differences in neonatal care across cohorts or eras of care, particularly across the 1980s and 1990s, with increased survival following advances in surfactant treatment, ventilation techniques, or nutrition.1,2,11,12 Furthermore, cognitive abilities and attainment may be affected by socioeconomic status (SES) and early education.13,14

eonatal care across cohorts or eras of care, particularly across the 1980s and 1990s, with increased survival following advances in surfactant treatment, ventilation techniques, or nutrition.1,2,11,12 Furthermore, cognitive abilities and attainment may be affected by socioeconomic status (SES) and early education.13,14 We investigated the association of GA with cognitive ability (IQ), basic mathematic processing, and mathematic attainment assessed during second grade of elementary school (8 years of age) in the Bavarian Longitudinal Study (BLS) cohort born 1985/1986 in the South of Germany at 27-41 weeks GA. We then used the regression functions identified in the BLS sample to predict IQ, basic mathematic processing, and mathematic attainment assessed at second grade in the United Kingdom (UK) (6 years) and 11 years of age using the same tests in the EPICure national cohort of extremely preterm (EP) children born in 1995 in the whole of the UK and Ireland at 23-25 weeks GA. We, first, hypothesized that the effects of GA on IQ and basic mathematic processing8,15 are universal; that is, similar deficits would be found across cohorts assessed in different countries and during different eras of neonatal care.2 Second, we hypothesized that mathematic attainment9,16 may be susceptible to country specific schooling and that outcomes may, thus, differ between cohorts; that is, prediction from one cohort to another may be less accurate compared with predictions of basic cognitive abilities.

uring different eras of neonatal care.2 Second, we hypothesized that mathematic attainment9,16 may be susceptible to country specific schooling and that outcomes may, thus, differ between cohorts; that is, prediction from one cohort to another may be less accurate compared with predictions of basic cognitive abilities. Methods Two prospective geographically defined birth cohorts were included, the BLS and the EPICure study. Descriptive characteristics of the BLS and EPICure study participants are in Table I. BLS Cohort The enrollment procedures have been described in detail elsewhere.17-19 A total of 7505 infants (10.6% of all live births) who were born between January 1985 and March 1986 in Southern Bavaria, Germany, and required admission to a children's hospital within the first 10 days of life were invited to participate in this study (index children). In addition, 916 term-born infants who received normal postnatal care were identified in the same hospitals. Ethical approval was obtained from the Ethics Committee of the University of Munich Children's Hospital and the Bavarian Health Council (Landesärztekammer). Analyses for this study use follow-up data at 8 years. At this age, we assessed 336 very preterm survivors and a sample of 1169 children born >31 weeks GA stratified by child sex, family SES, and degree of neonatal risk. Of these, 156 children could not complete the full battery of tests and were excluded. Data from 20 EP children (<27 weeks GA) were excluded as the number was too small to allow for appropriate statistical estimates. Finally, 40 children born post-term (>41 weeks GA) were excluded given the established association with adverse developmental outcomes.20 The final BLS sample for this study thus comprised 1289 children born between 27 and 41 weeks GA. All tests were standardized according to 584 children born full term (39-41 weeks) within the sample (298 receiving normal postnatal care and 286 index full-term children) who were followed to 8 years.

opmental outcomes.20 The final BLS sample for this study thus comprised 1289 children born between 27 and 41 weeks GA. All tests were standardized according to 584 children born full term (39-41 weeks) within the sample (298 receiving normal postnatal care and 286 index full-term children) who were followed to 8 years. EPICure The EPICure study included EP infants who were born before 26+0 weeks GA in the UK and Ireland from March through December 1995. The sampling of the study population has been described previously.10,21 Ethics approval was granted by the Trent Multicenter Research Ethics Committee. In total, 241 and 219 survivors were followed to age 6 and 11 years, respectively. Children with severe physical disability who could not complete the tests were excluded (n = 48), leaving 171 EP children. Cognitive abilities and mathematics attainment were assessed at 6 years and mathematic processing at 11 years. All tests were standardized according to full-term control children (37-41 weeks gestation) from the same classes in mainstream schools at 6 (n = 160) and 11 years of age (n = 153).22,23 Measures In both studies, GA (completed weeks) was calculated from maternal reports of the last menstrual period and serial ultrasounds during pregnancy.23,24 In both studies, psychologists assessed cognitive abilities using the Kaufman-Assessment Battery for Children (K-ABC).25,26 This yielded a mental processing composite (MPC) score indicating general cognitive ability (IQ).

d from maternal reports of the last menstrual period and serial ultrasounds during pregnancy.23,24 In both studies, psychologists assessed cognitive abilities using the Kaufman-Assessment Battery for Children (K-ABC).25,26 This yielded a mental processing composite (MPC) score indicating general cognitive ability (IQ). Children in both studies were administered a Mathematics Estimation Test9,27 at age 8 (BLS) and 11 (EPICure) years, respectively. Tasks were presented to children in book form with 12 items assessing the estimation of dot array and number line magnitude, as well as judgments of approximate length and distance (Figure 1; available at www.jpeds.com). Item responses were scored for accuracy and summarized into a total score. Test scores were standardized based on term controls in each study separately (standardized control mean 100; SD 15). In both studies, the age-appropriate K-ABC arithmetic subtest (separate from the MPC) assessed children's attainment in mathematics.25,26 At the time of the K-ABC assessment, children in both cohorts had received, on average, 2 years of formal school education. For the purpose of comparison between the 2 cohorts, MPC and K-ABC mathematics scores were standardized according to the full-term control children in each study separately (standardized control mean 100; SD 15). Children who could not be assessed because of severe cognitive disability were assigned a score of 39 for IQ and mathematics attainment (ie, 1 point below the minimal possible assessment score).

ere standardized according to the full-term control children in each study separately (standardized control mean 100; SD 15). Children who could not be assessed because of severe cognitive disability were assigned a score of 39 for IQ and mathematics attainment (ie, 1 point below the minimal possible assessment score). Analyses were controlled for family SES, child sex, and small for GA (SGA) birth. Infants were classified as SGA if they weighed less than the sex specific 10th percentile for their GA according to the national German standard weight charts (1985-1986)28 and based on the UK child growth foundation charts in EPICure.29 Family SES was classified into 3 categories corresponding to high, medium, and low using parental education and occupation.30,31 Statistical Analyses Missing data was imputed with full information maximum likelihood estimates. BLS In order to identify the best fitting model for GA effects on outcomes in the BLS cohort, piecewise linear regressions were fitted for IQ, mathematic processing, and mathematic attainment using the STATA v 12 nonlinear functions (Stata Corp, College Station, Texas) and Taylor change-point analysis tools. This method was selected based on previous findings of a nonlinear effect of GA on cognitive and mathematic outcomes.7,32 Piecewise regressions were used to identify the week of GA at which test performance differed significantly above and below a change point. Final models were adjusted for family SES, child sex, and SGA.

method was selected based on previous findings of a nonlinear effect of GA on cognitive and mathematic outcomes.7,32 Piecewise regressions were used to identify the week of GA at which test performance differed significantly above and below a change point. Final models were adjusted for family SES, child sex, and SGA. EPICure Accuracy of predicted IQ, mathematic processing, and mathematic attainment scores for EP children was evaluated by inserting their observed scores into the piecewise regressions fitted to the BLS sample. The 50% and 75% prediction intervals (ie, predicted scores ± 0.67 (Z0.52) and 1.04 (Z0.252), root mean square errors [RMSEs]) were then calculated. RMSEs indicate the SDs of the residuals (ie, the difference between observed and predicted scores). The precision of these predictions was examined by the range of EPICure observed scores (ie, 25th-75th percentiles) that fell within these 50% (1 RMSE) and 75% (2 RMSEs) prediction intervals.33 A prediction was assumed to be very precise if the 25th-75th percentiles of observed scores were covered within the 50% prediction interval. All models were controlled for family SES, child sex, and SGA birth.

ores (ie, 25th-75th percentiles) that fell within these 50% (1 RMSE) and 75% (2 RMSEs) prediction intervals.33 A prediction was assumed to be very precise if the 25th-75th percentiles of observed scores were covered within the 50% prediction interval. All models were controlled for family SES, child sex, and SGA birth. Results The Effect of Birth at 27-41 Weeks GA on IQ, Mathematic Processing, and Mathematic Attainment in the BLS Cohort Piecewise regressions showed that GA exerted differential effects on IQ and mathematics attainment below vs above 34 weeks (95% CI: 31 weeks, 37 weeks; and 32 weeks, 36 weeks, respectively) and on basic mathematic processing below vs above 36 weeks (95% CI: 34 weeks, 38 weeks). Table II and Figure 2 show that after controlling for SES, sex, and SGA, children's IQ and mathematics attainment scores decreased by 2.34 points (95% CI: −2.99, −1.70) and 2.76 points (95% CI: −3.40, −2.11) with each lower week of GA below 34 weeks, respectively. There were no significant differences among children born at 34-41 weeks GA for both outcomes. Basic mathematic processing scores decreased by 1.77 points (95% CI: −2.20, −1.34) with each week of GA below 36 weeks, and there was no significant effect of GA for children born at 36-41 weeks. In addition to GA, low SES had strong negative effects on outcomes. For example, the effect of low SES on IQ was equivalent to that of 5 weeks of GA below the change point (34 weeks). On average, SGA birth had negative effects on all outcomes across the gestation spectrum and girls had worse mathematics attainment and basic mathematic processing scores than boys (Table II).

n outcomes. For example, the effect of low SES on IQ was equivalent to that of 5 weeks of GA below the change point (34 weeks). On average, SGA birth had negative effects on all outcomes across the gestation spectrum and girls had worse mathematics attainment and basic mathematic processing scores than boys (Table II). Predicted Performance of EPICure Children According to BLS Regression Functions Accuracy of predicted IQ, mathematic processing, and mathematic attainment scores for EP children was evaluated by inserting their observed scores into the piecewise regressions fitted to the BLS sample. Figure 3 shows distributions of EPICure Study children's observed scores (box plots) vs their predicted scores (lines) with 50% and 75% prediction intervals based on the BLS data. Both observed IQ and basic mathematic processing scores between 25th and 75th percentiles were mostly covered within the 50% prediction interval (Figures 3 and 4; Figure 4 available at www.jpeds.com), showing observed and predicted scores by GA in both BLS and EPICure children. Thus, consistent with hypothesis 1, BLS children's scores (27-41 weeks GA) allowed accurate prediction of IQ and basic mathematic processing scores of children born at 23-25 weeks GA in another country one decade later. In contrast, the top one-half of the observed EPICure mathematics attainment scores were only within the range of the 75% prediction interval; they deviated more than 1 RMSE from the predicted scores. Thus, EPICure children had higher mathematics attainment scores than was predicted from BLS data.

Predicted Performance of EPICure Children According to BLS Regression Functions Accuracy of predicted IQ, mathematic processing, and mathematic attainment scores for EP children was evaluated by inserting their observed scores into the piecewise regressions fitted to the BLS sample. Figure 3 shows distributions of EPICure Study children's observed scores (box plots) vs their predicted scores (lines) with 50% and 75% prediction intervals based on the BLS data. Both observed IQ and basic mathematic processing scores between 25th and 75th percentiles were mostly covered within the 50% prediction interval (Figures 3 and 4; Figure 4 available at www.jpeds.com), showing observed and predicted scores by GA in both BLS and EPICure children. Thus, consistent with hypothesis 1, BLS children's scores (27-41 weeks GA) allowed accurate prediction of IQ and basic mathematic processing scores of children born at 23-25 weeks GA in another country one decade later. In contrast, the top one-half of the observed EPICure mathematics attainment scores were only within the range of the 75% prediction interval; they deviated more than 1 RMSE from the predicted scores. Thus, EPICure children had higher mathematics attainment scores than was predicted from BLS data. Discussion This study investigated the effect of birth across the whole gestation spectrum on cognitive abilities and mathematics attainment in middle childhood. The relationships between GA and outcomes were best fitted using piecewise regressions. These indicated deficits in IQ and mathematics attainment for children born <34 weeks GA and in basic mathematic processing abilities for children born <36 weeks GA. In addition, regression functions that were identified in the BLS sample accurately predicted EP children's IQ and basic mathematic processing scores in the EPICure study; however EP children's mathematic attainment in the EPICure study was better than predicted by performance of the BLS cohort.

as <10th percentile according to gestational age and sex. Concerning nutritional status in childhood, 12.8% of the subjects had a height-for-age z-score of ≤−2 SD at 2 years of age. The mean IQ at 30 years of age was 98 points and the average number of years of schooling was 11.4 (Table II; available at www.jpeds.com). With respect to the confounding variables, socioeconomic status was associated inversely with the prevalence of low birthweight, small-for-gestational age, and low weight and height for age z-score at 2 and 4 years of age. The proportion of children whose conditional length at 2 years of age variable was ≥1 z-score was positively associated with socioeconomic status, whereas conditional height at 4 years of age and conditional weight variables were independent of socioeconomic variables (Table III; available at www.jpeds.com). IQ, years of schooling, and income at 30 years of age were positively associated with socioeconomic status at birth and childhood. Maternal smoking during the pregnancy was associated with lower IQ, years of schooling, and income at 30 years of age (Table IV; available at www.jpeds.com).

dren born <36 weeks GA. In addition, regression functions that were identified in the BLS sample accurately predicted EP children's IQ and basic mathematic processing scores in the EPICure study; however EP children's mathematic attainment in the EPICure study was better than predicted by performance of the BLS cohort. First, these results provide evidence for a universal effect of GA at birth on long-term cognitive and basic mathematic processing abilities. Using data obtained from children born at 27-41 weeks gestation in Germany in 1985/1986, we successfully predicted IQ and mathematic processing scores in EP children born at 23-25 weeks gestation in the UK and Ireland in 1995. Given the particular decade that elapsed between recruitment of the 2 cohorts, EPICure children received pioneering new treatments such as surfactant administration that highly increased survival of EP children. Despite this, there was no equivalent improvement in their basic cognitive abilities. This is consistent with recent findings that compared 2 EP cohorts born in 1995 and 2006 that showed increased survival of EP infants but no improvement in neurodevelopmental outcomes.2 The precision of these predictions across different populations, decades, and health care systems indicates that underlying neurodevelopmental,32 rather than childhood environmental factors, may explain adverse effects of preterm birth on basic cognitive abilities. Accordingly, recent neuro-imaging studies have indicated changes in brain structure, function, and connectivity in relation to gestation at birth.34-36 Future studies that include neuro-imaging across cohorts may provide more direct evidence of similarly altered brain development. Our findings suggest that despite significant improvements in neonatal intensive care, there is considerable temporal and cross-national consistency in long-term cognitive abilities, at least in high income countries such as Germany and the UK. Increased survival, thus, provides no indication of improved cognitive function among children born moderately or very preterm.

ements in neonatal intensive care, there is considerable temporal and cross-national consistency in long-term cognitive abilities, at least in high income countries such as Germany and the UK. Increased survival, thus, provides no indication of improved cognitive function among children born moderately or very preterm. Second, EPICure study children had higher mathematics attainment scores than were predicted by BLS data. Mathematics attainment was measured when children in both cohorts had received, on average, 2 years of formal schooling. There are, however, some important differences between the study populations' educational contexts. In the UK, children must enter compulsory schooling by 5 years of age, which is usually preceded by a year in reception class. Moreover, most children with special educational needs are admitted to mainstream school and receive extra and often individual help within class. Given the UKs inclusive education policies, only children with severe disabilities are admitted to special schools. Within the EPICure Study, only 6 (3.5%) EP children in the dataset used for analysis were in special schools. In contrast, in Germany, children had formal school entry assessments by community pediatricians that were used to stream children before entering elementary school. Those who passed the school entry tests entered elementary school in September after their 6th birthdays. Those who failed the school entry examination were either delayed for one year, (ie, entered school a year later at age 7 years) (96 children [7.4%] of the whole BLS sample [27-41 weeks GA]; 77 out of 319 children born <34 weeks GA [24.1%]) or were directly streamed into special schooling (73 children [5.7%] of the whole BLS sample; 44 out of 319 children born <34 weeks GA [13.8%], respectively). Thus German preterm children who often have mathematics achievement problems9 are less likely to receive support at mainstream school level. These discrepancies between the UK's and Germany's education systems may explain why EPICure children were doing much better in mathematics attainment than was expected.

ctively). Thus German preterm children who often have mathematics achievement problems9 are less likely to receive support at mainstream school level. These discrepancies between the UK's and Germany's education systems may explain why EPICure children were doing much better in mathematics attainment than was expected. In order to evaluate to what extent streaming into special schooling and delayed school entry may have accounted for German children's mathematic underachievement, we repeated our analyses only including BLS children who had entered mainstream school at the age appropriate time; EPICure children's observed mathematic attainment scores were now much lower than predicted (Figure 5; available at www.jpeds.com). This strongly suggests that special help within mainstream school may help children to attain mathematics abilities beyond their general cognitive abilities.37 The late preschool and early school years may represent a sensitive time for acquiring mathematical skills and preterm children may be highly sensitive to teacher or parent interventions at this time.8,38

hin mainstream school may help children to attain mathematics abilities beyond their general cognitive abilities.37 The late preschool and early school years may represent a sensitive time for acquiring mathematical skills and preterm children may be highly sensitive to teacher or parent interventions at this time.8,38 Over and above the significant effect of GA, SES strongly predicted children's cognitive and mathematics abilities. The effect of growing up in a low SES family was equivalent to that of 2 (for basic mathematic processing) to 5 (for IQ) weeks decrease in GA for children born with a GA below the respective change points. Thus, our study confirms previously shown powerful influences of the social environment on preterm and full-term individual's long-term cognitive and educational outcomes.39 Further analyses indicated that there was no interaction effect between SES and GA, but rather an additive detrimental effect of low SES on preterm children's cognitive and educational outcomes, as has been shown before.17 Thus, low SES has similar adverse effects on cognitive and mathematics abilities for children born across the whole gestation spectrum.

was no interaction effect between SES and GA, but rather an additive detrimental effect of low SES on preterm children's cognitive and educational outcomes, as has been shown before.17 Thus, low SES has similar adverse effects on cognitive and mathematics abilities for children born across the whole gestation spectrum. This study evaluated the long-term effects of preterm birth across the whole gestation spectrum and cross validated findings in another cohort while controlling for key confounders. In both studies, children's abilities were assessed with the same standardized tests allowing for direct comparison across cohorts. To control for country specific impacts or the Flynn effect40 of increasing standardized test scores over time, the performance of preterm children was standardized according to term-born children recruited at birth in the BLS and classroom controls in the EPICure study.18 To estimate the association of gestation with outcome measures, those who were admitted to a children's hospital (index children) or had normal postnatal care between 39 and 41 weeks were combined in the BLS. These 2 full-term groups did not differ in basic mathematic processing and mathematics attainment scores but slightly in IQ (99.1 and 101.6; mean difference: −2.5 [95% CI: −4.6, −0.4]). Thus, overall IQ may have been slightly underestimated in the full-term range in the analysis but weighting did not alter prediction results. In both samples, K-ABC assessments were administered when children were at the same point in their school careers (ie, age 8 years in the BLS and age 6 years in EPICure, thus, they all had 2 years of school experience). The Mathematics Estimation Test was, however, administered when BLS children were 8 years and EPICure children were 11 years old. Although this timing difference may be seen as a caveat, the ability to nevertheless predict EPICure children's scores on the basis of an assessment done in another country at a different age and time point in children's school careers further strengthens the validity of our findings. Although our results suggest a universal effect of GA on childhood outcomes our findings are solely based on 2 European studies in high income countries which, despite some differences, may also present a number of similarities. In order to confirm the universal effect described here future studies should cross-validate our findings using non-European preterm samples.

GA on childhood outcomes our findings are solely based on 2 European studies in high income countries which, despite some differences, may also present a number of similarities. In order to confirm the universal effect described here future studies should cross-validate our findings using non-European preterm samples. With regard to SGA classifications, the BLS and EPICure cohorts both used nationally appropriate growth chart samples. However, the generally small number of EP children born SGA in the EPICure study, a study of children at the limits of survival at the time, may be due to growth approximations in the charts at the time rather than actual growth chart data. Finally, IQ and math processing and attainment were assessed between 6 and 11 years. Future studies may include longer follow-up. However, both IQ and math tests have been shown to be highly predictive of outcomes in adulthood and even old age.41,42

ations in the charts at the time rather than actual growth chart data. Finally, IQ and math processing and attainment were assessed between 6 and 11 years. Future studies may include longer follow-up. However, both IQ and math tests have been shown to be highly predictive of outcomes in adulthood and even old age.41,42 The ability to predict long-term outcomes in general cognitive abilities and basic mathematic processing from one national cohort to another, both over time and with different neonatal services and social and education systems, suggests that neurodevelopmental rather than childhood environmental factors explain the long-term effects of gestation at birth. The finding that EPICure children had higher mathematic attainment scores than predicted suggests that national differences in elementary education may have substantial effects on preterm children's educational attainment chances despite similar general cognitive functioning. This may warrant further research including randomized controlled trials of tailored education interventions for preterm children. Appendix Figure 1 Mathematics Estimation Test example items. Figure 4 Observed and predicted mean change of outcomes according to GA at birth in the EPICure Study (23-25 weeks GA) and BLS (27-41 weeks GA). Blue vertical lines: 95% CIs of observed means (circles: BLS and squares: EPICure); X: GA change points; black solid horizontal lines: predicted means below the GA change point; dashed horizontal lines: predicted means above the GA change point.

A at birth in the EPICure Study (23-25 weeks GA) and BLS (27-41 weeks GA). Blue vertical lines: 95% CIs of observed means (circles: BLS and squares: EPICure); X: GA change points; black solid horizontal lines: predicted means below the GA change point; dashed horizontal lines: predicted means above the GA change point. Figure 5 Observed and predicted mean change of mathematic attainment scores only for children in mainstream schools according to GA at birth in the BLS (N = 1094; 27-41 weeks GA) and in the EPICure Study (N = 165; 23-25 weeks GA). Blue vertical lines: 95% CIs of observed means (circles: BLS and squares: EPICure); X: GA change points; black solid horizontal lines: predicted means below the GA change point; dashed horizontal lines: predicted means above the GA change point. Supported by the Nuffield Foundation (EDU/40442) and the German Research Foundation (JA 1913). The contents are solely the responsibility of the authors and do not necessarily represent the official views of funding bodies. The authors declare no conflicts of interest. Figure 2 Observed and predicted mean change in outcomes according to GA at birth in the BLS (Germany; 27-41 weeks GA). Grey vertical lines: 95% CIs of observed means (circles); X: GA change points; black solid horizontal lines: predicted means below the GA change point; dashed horizontal lines: predicted means above the GA change point.

and predicted mean change in outcomes according to GA at birth in the BLS (Germany; 27-41 weeks GA). Grey vertical lines: 95% CIs of observed means (circles); X: GA change points; black solid horizontal lines: predicted means below the GA change point; dashed horizontal lines: predicted means above the GA change point. Figure 3 EPICure Study observed score distributions (box plots∗) with predicted mean scores (solid lines) and 50% (dashed lines) and 75% (dotted lines) prediction intervals based on the BLS cohort. Precision of prediction was examined by calculating the percentiles of box plots within prediction intervals and show that observed IQ and basic mathematic processing scores were mostly covered within the 50% prediction interval. ∗The bottom and top of each box are the 25th and 75th percentiles of observed scores, respectively. The line in the middle is the 50th percentile and hollow circles are observed mean scores. Table I Descriptive characteristics of BLS and EPICure children included in analyses

Figure 3 EPICure Study observed score distributions (box plots∗) with predicted mean scores (solid lines) and 50% (dashed lines) and 75% (dotted lines) prediction intervals based on the BLS cohort. Precision of prediction was examined by calculating the percentiles of box plots within prediction intervals and show that observed IQ and basic mathematic processing scores were mostly covered within the 50% prediction interval. ∗The bottom and top of each box are the 25th and 75th percentiles of observed scores, respectively. The line in the middle is the 50th percentile and hollow circles are observed mean scores. Table I Descriptive characteristics of BLS and EPICure children included in analyses BLS children (N = 1289) EPICure children (N = 171) IQ 96.97 (16.70) 78.63 (16.62) Basic mathematic processing 97.63 (15.58) 83.99 (16.12) Mathematic attainment 96.87 (16.82) 81.84 (19.84) GA 36.52 (3.94) 24.53 (0.66) Sex (boys) 655 (50.81%) 74 (43.27%) Age 8.34 (0.23) 6.28 (0.46) SES High 386 (29.95%) 67 (45.58%) Medium 485 (37.63%) 34 (23.13%) Low 418 (32.43%) 46 (31.29%) SGA 325 (25.21%) 14 (8.19%) Data are presented as mean (SD) for numerical variables or numbers (percentages [%]) for categorical variables. Please note that EPICure children's basic mathematic processing abilities (Mathematics Estimation Test) were assessed at 11 years of age (mean = 10.91 [SD = 0.37]). Table II The association of GA with cognitive and mathematic performance in the BLS cohort (N = 1289) adjusted for child sex, family SES, and SGA status

BLS children (N = 1289) EPICure children (N = 171) IQ 96.97 (16.70) 78.63 (16.62) Basic mathematic processing 97.63 (15.58) 83.99 (16.12) Mathematic attainment 96.87 (16.82) 81.84 (19.84) GA 36.52 (3.94) 24.53 (0.66) Sex (boys) 655 (50.81%) 74 (43.27%) Age 8.34 (0.23) 6.28 (0.46) SES High 386 (29.95%) 67 (45.58%) Medium 485 (37.63%) 34 (23.13%) Low 418 (32.43%) 46 (31.29%) SGA 325 (25.21%) 14 (8.19%) Data are presented as mean (SD) for numerical variables or numbers (percentages [%]) for categorical variables. Please note that EPICure children's basic mathematic processing abilities (Mathematics Estimation Test) were assessed at 11 years of age (mean = 10.91 [SD = 0.37]). Table II The association of GA with cognitive and mathematic performance in the BLS cohort (N = 1289) adjusted for child sex, family SES, and SGA status Regression coefficient β (95% CI) IQ Basic mathematic processing Mathematic attainment GA change point 34 36 34 <GA change point∗ −2.34 (−2.99, −1.70) −1.77 (−2.20, −1.34) −2.76 (−3.40, −2.11) >GA change point∗ 0.16 (−1.45, 1.78) −0.33 (−1.68, 1.01) 0.16 (−1.46, 1.78) Females −1.24 (−2.89, 0.42) −2.83 (−4.45, −1.21) −5.01 (−6.67, −3.36) High SES 1 1 1 Medium SES −6.57 (−8.61, −4.54) −2.97 (−4.96, −0.98) −6.41 (−8.45, −4.37) Low SES −11.85 (−13.95, −9.75) −4.96 (−7.02, −2.90) −9.28 (−11.39, −7.17) SGA −5.48 (−7.44, −3.53) −2.79 (−4.71, −0.88) −5.29 (−7.24, −3.32) ∗ Coefficient β for each GA relative to GA change point.

Cardiovascular disease (CVD) is the leading cause of death globally, and the population incidence of CVD and related metabolic disorders is higher in low- and middle-income countries (LMICs) than in the rest of the world.1, 2 Mortality because of premature CVD is increasing in South Asian countries such as India.1, 3, 4 Growth patterns in early life are important predictors of adult CVD risk factors.4, 5 Lower weight at birth6, 7 and/or during infancy8 and higher weight or BMI during childhood or adolescence9, 10, 11, 12 are associated with a higher risk of adult hypertension, type 2 diabetes mellitus (T2DM), and CVD. These relationships may be mediated by effects on body composition; weight and BMI at birth and during infancy positively predict adult lean mass more strongly than adult fat mass, and during late childhood and adolescence, they predict fat mass more strongly.13, 14, 15 Few studies have examined associations of height (linear) growth in early life with later CVD risk.

body composition; weight and BMI at birth and during infancy positively predict adult lean mass more strongly than adult fat mass, and during late childhood and adolescence, they predict fat mass more strongly.13, 14, 15 Few studies have examined associations of height (linear) growth in early life with later CVD risk. Understanding when in childhood growth relates to later CVD risk may guide the timing of interventions to prevent disease. Identifying specific ages when linear growth and weight (soft tissue) gain predict later outcomes is complicated by the fact that serial measurements of height or weight within an individual are strongly positively correlated, and height and weight are correlated with each other. Two-way conditional growth measures, which are adjusted for prior size, and mutually adjust weight and height for each other, have been developed to overcome these limitations.14 We have used 2-way conditional growth analysis to study independent relationships of linear growth and weight gain, during defined periods of infancy, childhood, and adolescence, with adult CVD risk markers using data from the Vellore birth cohort in India.

ch other, have been developed to overcome these limitations.14 We have used 2-way conditional growth analysis to study independent relationships of linear growth and weight gain, during defined periods of infancy, childhood, and adolescence, with adult CVD risk markers using data from the Vellore birth cohort in India. Methods The Vellore Birth Cohort includes individuals born within defined areas of Vellore town and adjoining rural villages in Tamil Nadu, India during 1969-1973.9 The current analysis used data from the 2218 cohort members for whom birth measurements were available and who were followed up as young adults during 1998-2002. Height and weight were measured prospectively by trained research staff, using standardized methods, at birth, during infancy (up to 3 months of age), childhood (6-8 years of age), adolescence (10-15 years of age), and adulthood. Children had up to 3 measurements in the first 3 months of age, up to 2 measurements between 6 and 8 years of age, and up to 5 measurements between 10 and 15 years of age. The study was approved by the institutional ethics committee, and all study participants provided written informed consent.

of age), and adulthood. Children had up to 3 measurements in the first 3 months of age, up to 2 measurements between 6 and 8 years of age, and up to 5 measurements between 10 and 15 years of age. The study was approved by the institutional ethics committee, and all study participants provided written informed consent. Adult follow-up took place at a median (IQR) age of 28.1 years (27.4, 28.8).9 Measurements included weight to the nearest 0.1 kg; height to the nearest 1 mm, measured using a Harpenden portable stadiometer (Holtain Ltd, Crymych, Dyfed, Wales); and waist circumference (WC) measured to the nearest 1 mm, midway between the costal margin and iliac crest in expiration. Body mass index (BMI) was calculated using the formula weight (kg)/length or height (m) squared. Information was collected on place of residence (both current and at birth), attained education level, current tobacco and alcohol use, physical activity, and socioeconomic status. Education was recorded in 7 groups from no schooling to a professional qualification. Participants were defined as current tobacco users or nonusers. Frequency and quantity of consumption of beer, wine, and spirits were converted into units of alcohol per week. A score was derived as a summary estimate of daily physical activity as described previously.10 A 6-point scale ranging from “almost entirely sedentary” to “heavy physical work” was used to classify work-related activity. In addition, the scoring included time spent in domestic and leisure activities and daily mode of transport (walking, cycling). Time periods for each activity were multiplied by metabolic constants derived from published tables of the relative energy expenditure of each task, and summed to create the final physical activity score. Socioeconomic status was assessed by recording possession of up to 15 household items.11 Details of anthropometry and CVD risk factors measurements are described elsewhere.12

constants derived from published tables of the relative energy expenditure of each task, and summed to create the final physical activity score. Socioeconomic status was assessed by recording possession of up to 15 household items.11 Details of anthropometry and CVD risk factors measurements are described elsewhere.12 Data Analyses Analysis Sample In selecting ages for the growth analysis, we aimed to include (in addition to birth and adulthood) infancy, childhood, and adolescence. The exact ages selected were based upon availability of data. No infant data were collected before March 1, 1971 or in December 1973, and therefore, we excluded births at these times (n = 326), leaving 1892. We also excluded 7 cohort members without birth length and 7 without adult height and weight, leaving 1878. Among the 1878, infant measurements were available for 1613 (median [IQR] 2.9 months [2.0, 3.0]), childhood measurements for 1680 (6.4 years of age [5.9, 6.9]), and adolescent measurements for 1108 (14.9 years of age [14.4, 19.4]) (Figure 1; available at www.jpeds.com). All measurements were converted into within-cohort age- and sex-specific z scores [(subject mean-cohort mean)/cohort SD]. Exact values at age 3 months, and age 6.5 and 15 years were then obtained by interpolation of the z scores, using the nearest measurements to that age and within 2 months for the infant value and 2 years for the childhood and adolescent values, and back-transformation to the units of measurement.

hort mean)/cohort SD]. Exact values at age 3 months, and age 6.5 and 15 years were then obtained by interpolation of the z scores, using the nearest measurements to that age and within 2 months for the infant value and 2 years for the childhood and adolescent values, and back-transformation to the units of measurement. We examined the associations of weight and height z scores separately, at each age, with each CVD risk marker, using linear regression, first adjusted for adult age alone (model 1) and then by additional adjustment for adult body size (BMI and height, model 2). Model 1 (“forward-looking” approach) addresses the question: What is the net association of size at each age with the adult outcome? Model 2 (“backward-looking” approach) addresses the question: Given that this person has achieved a particular adult BMI and height, is there any remaining effect on the outcome of size at earlier ages, or do the earlier measurements have all their effect through their contribution to adult size? Both models were adjusted for year of birth and sociodemographic variables. We used interaction tests to examine whether associations between body size and cardiovascular risk markers differed between the sexes, and because there were more statistically significant interactions than expected by chance, all analyses were stratified by sex.

usted for year of birth and sociodemographic variables. We used interaction tests to examine whether associations between body size and cardiovascular risk markers differed between the sexes, and because there were more statistically significant interactions than expected by chance, all analyses were stratified by sex. We constructed sex-specific and height- and weight-specific conditional variables, which are standardized residuals derived from regressing size z scores at each age on prior size measures.16 Conditional height is current height accounting for all prior height and weight measures. Conditional relative weight gain is current weight accounting for current height and all prior weight and height measures. For example, adult conditional relative weight was derived by regressing adult weight on adult height, and weight and height or length at age 15 and 6.5 years, 3 months, and birth.

weight measures. Conditional relative weight gain is current weight accounting for current height and all prior weight and height measures. For example, adult conditional relative weight was derived by regressing adult weight on adult height, and weight and height or length at age 15 and 6.5 years, 3 months, and birth. Conditional relative weight and height gain variables represent children's deviation from expected size based on their own prior measures and on the growth of the other children in the cohort, and can be interpreted as representing greater or less than expected soft tissue gain and linear growth respectively. For example, a child with a positive conditional relative weight at 6.5 years of age is heavier than expected given his/her current height and prior size and, thus, had a faster rate of soft tissue gain from age 3 months to 6.5 years. Again, we created “forward-looking” models adjusted only for adult age (model 1), and “backward-looking” models further adjusted for adult BMI and height (model 2). We included 907 participants who had data at all selected ages for conditional analysis. Analyses were undertaken using SPSS v 22 (SPSS Inc, Chicago, Illinois) and Stata v 13.1 (StataCorp, College Station, Texas).

or adult age (model 1), and “backward-looking” models further adjusted for adult BMI and height (model 2). We included 907 participants who had data at all selected ages for conditional analysis. Analyses were undertaken using SPSS v 22 (SPSS Inc, Chicago, Illinois) and Stata v 13.1 (StataCorp, College Station, Texas). Results When compared with the remainder of the original cohort of 10 691 live singleton births, the 1878 men and women included in this analysis were less likely to have been born to a primiparous mother (studied 11.7%; not studied 19.8%; P < .001). There were differences in the education level of the head of the household at the time of their birth, but these were small (illiterate: 12.2% and 12.7%; attended school 5th standard: 36.5% and 32.4%; 6-11 standard: 47.1% and 45.8%; college graduate: 5.4% and 4.1%; P < .001). There were no significant differences in parental weight and height. Subjects studied were similar at birth (2795  vs 2815 g; P = .19) and heavier at 3 months of age (4858 vs 4706 g; P < .001), but there were no significant differences in measurements during childhood. Subjects studied during adolescence were lighter than the subjects not studied though the differences were small.

ight. Subjects studied were similar at birth (2795  vs 2815 g; P = .19) and heavier at 3 months of age (4858 vs 4706 g; P < .001), but there were no significant differences in measurements during childhood. Subjects studied during adolescence were lighter than the subjects not studied though the differences were small. Early life and adult characteristics of the study sample are shown in Table I. Compared with an international growth reference, they were small at all ages in childhood; mean World Health Organization z scores for length, weight, and BMI at birth were -0.72, -1.17, and -1.28 respectively. Equivalent data for 3 months of age were -1.34, -2.00, and -1.70; for 6.5 years of age -2.27, -2.35, and -1.19 and for 15 years of age -2.11, -2.89, and -1.95. At 28 years of age, the prevalence of underweight (BMI <18.5 kg/m2) was 28%, overweight (BMI 25-30 kg/m2) 11%, and obesity (BMI >30 kg/m2) 2%. Despite a low mean BMI, 14% of men and 17% of women had central obesity defined by the International Diabetes Federation as a waist circumference (WC) >90 cm in men and >80 cm in women.17 The prevalence of hypertension, impaired glucose tolerance, and T2DM was high at 9.4%, 16.6%, and 2.8% respectively. Estimates for diastolic blood pressure (DBP) were similar to those for systolic blood pressure (SBP), and are not shown.

etes Federation as a waist circumference (WC) >90 cm in men and >80 cm in women.17 The prevalence of hypertension, impaired glucose tolerance, and T2DM was high at 9.4%, 16.6%, and 2.8% respectively. Estimates for diastolic blood pressure (DBP) were similar to those for systolic blood pressure (SBP), and are not shown. Unadjusted for Adult Size As expected, length or height at all ages was positively correlated with adult height; correlations strengthened with increasing age of the earlier measurement, from r = 0.28 in men and 0.25 in women for birth length to r = 0.66 in men and 0.79 in women for 15-year height. Similarly BMI at all earlier ages correlated positively with adult BMI (r = 0.13 in men and 0.07 in women for BMI at birth, and 0.52 in both sexes for 15-year BMI).

increasing age of the earlier measurement, from r = 0.28 in men and 0.25 in women for birth length to r = 0.66 in men and 0.79 in women for 15-year height. Similarly BMI at all earlier ages correlated positively with adult BMI (r = 0.13 in men and 0.07 in women for BMI at birth, and 0.52 in both sexes for 15-year BMI). Taller adult height was associated with higher WC, BP, and homeostatic model assessment-insulin resistance (HOMA-IR) in both sexes, but with lower 120-minute glucose and cholesterol concentrations in women (Table II). Higher adult BMI was associated with higher WC, BP, glucose, cholesterol, and triglyceride concentrations, and HOMA-IR, and lower adult high-density lipoprotein cholesterol concentration in both sexes (Table III). Longer birth length and higher BMI at birth were associated with lower 120-minute glucose concentration among women. Height and BMI at all ages were strongly positively related to adult WC (Table II, Table III). Taller height at 6.5 and 15 years of age was associated with higher BP in both sexes. Taller height at 15 years of age was also associated with higher cholesterol and triglyceride concentrations in men and higher HOMA-IR in women. Higher 15-year BMI was associated with higher BP in both sexes, and with higher cholesterol and triglyceride concentrations in men.

as associated with higher BP in both sexes. Taller height at 15 years of age was also associated with higher cholesterol and triglyceride concentrations in men and higher HOMA-IR in women. Higher 15-year BMI was associated with higher BP in both sexes, and with higher cholesterol and triglyceride concentrations in men. Adjusted for Adult Size The positive associations between earlier height and BMI and adult CVD risk markers were attenuated, and most were no longer statistically significant, after adjusting for adult height and BMI (Table II, Table III). Exceptions were BP and cholesterol, which remained positively related to 15-year height. In addition, some inverse associations became apparent. Longer birth length was associated with lower BP, HOMA-IR, and triglyceride concentrations in women (Table II). Higher BMI at birth was associated with lower 120-minute glucose concentration in both sexes (Table III). Higher BMI at 3 months of age was associated with lower WC in women. Higher BMI at 6.5 years of age was associated with lower 120-minute glucose in both sexes, and lower WC, cholesterol, and triglyceride concentrations in men. Taller height at 6.5 years of age was associated with lower triglyceride and higher adult high-density lipoprotein cholesterol concentration in men (Table II). Higher BMI at 15 years of age was associated with lower WC and 120-minute glucose concentration in both sexes, HOMA-IR in men, and cholesterol and triglycerides in women.

ght at 6.5 years of age was associated with lower triglyceride and higher adult high-density lipoprotein cholesterol concentration in men (Table II). Higher BMI at 15 years of age was associated with lower WC and 120-minute glucose concentration in both sexes, HOMA-IR in men, and cholesterol and triglycerides in women. Longitudinal Analysis—Unadjusted for Adult Size The 907 men and women included in the longitudinal analysis were shorter at birth by 0.48 cm than the 971 not included (P < .001) but did not differ significantly in birth weight or height and weight at 3 months, 6.5, 15, or 28 years of age, any of the cardiometabolic risk markers, or rural v urban residence at birth or currently. The results are presented in Figure 2. Greater linear growth at all ages was associated with higher adult WC in both sexes. Greater linear growth between birth and 3 months of age was associated with higher cholesterol and triglyceride concentrations in men, and higher BP in women. Greater linear growth between 3 months and 6.5 years of age was associated with higher BP and HOMA-IR in men, and DBP in women. Greater linear growth from 6.5 to 15 years of age was associated with higher BP in both sexes, higher HOMA-IR in women, and higher total cholesterol and triglycerides in men. Linear growth between 15 and 28 years of age was unrelated to risk factors, and there was little increase in height between these ages (mean 16 cm in men and 5 cm in women) (Table IV; available at www.jpeds.com).

th higher BP in both sexes, higher HOMA-IR in women, and higher total cholesterol and triglycerides in men. Linear growth between 15 and 28 years of age was unrelated to risk factors, and there was little increase in height between these ages (mean 16 cm in men and 5 cm in women) (Table IV; available at www.jpeds.com). The strongest finding was that greater conditional relative weight gain between 15 and 28 years of age was associated with a more adverse CVD risk profile in both sexes. Greater conditional weight gain at all ages was associated with higher adult WC. Conditional relative weight gain between birth and 3 months of age was unrelated to the risk factors except for WC. Greater conditional relative weight gain between 3 months and 6.5 years of age was associated with higher SBP in both sexes. Greater conditional relative weight gain from 6.5 to 15 years of age was associated with higher triglyceride concentrations in both sexes, higher SBP and DBP in women, and higher total cholesterol in men. The positive associations of conditional relative weight gain between 15 and 28 years of age with risk factors were considerably larger in magnitude than those for conditional relative weight gain before the age of 15 years (Table V; available at www.jpeds.com).

BP and DBP in women, and higher total cholesterol in men. The positive associations of conditional relative weight gain between 15 and 28 years of age with risk factors were considerably larger in magnitude than those for conditional relative weight gain before the age of 15 years (Table V; available at www.jpeds.com). Adjusted for Adult Size All the above positive associations were attenuated, and most became nonsignificant after adjusting for adult size. However, greater linear growth between birth and 3 months of age remained associated with higher cholesterol and triglyceride concentrations in men, and higher DBP in women; greater linear growth between 3 months and 6.5 years of age with higher DBP in men; and greater linear growth between 6.5 and 15 years of age with higher BP and total cholesterol and triglyceride concentration in men, and DBP in women. Discussion Overall, our study shows that greater height and BMI after birth and through to adolescence are associated with higher CVD risk factors in adulthood, and these associations are largely attributable to adult height and BMI. The significant differences between males and females were mainly due to differences in the specific ages at which earlier size or growth were related to the outcomes.

hrough to adolescence are associated with higher CVD risk factors in adulthood, and these associations are largely attributable to adult height and BMI. The significant differences between males and females were mainly due to differences in the specific ages at which earlier size or growth were related to the outcomes. Strengths of the study include the use of prospectively recorded growth data and the use of conditional growth analysis, which enabled us to examine independent effects of weight and height growth during specific age periods. Conditional variables are uncorrelated, and expressing them as z scores allows direct comparison of coefficients within regression models. This offers advantages over other representations of growth, and provides more information than weight gain alone. However, we acknowledge that conditional relative weight gain cannot differentiate between lean and fat gain. Very few birth cohorts have longitudinal measurements of lean and fat tissue through childhood, but recent data from a younger South Indian cohort has shown that the adipose component of childhood weight gain contributes most strongly to associations of weight gain with CVD risk markers.18 A limitation of the study was cohort attrition, mainly because of deaths and out-migrations; of the original 10 691 live singleton births, we studied 1878 (18%). Those studied differed from the original cohort in a variety of ways (for example infant mortality was higher among lower birth weight individuals). In a within-cohort analysis like this, these differences would not of themselves introduce bias, but would do so only if the associations between early size/growth and cardiometabolic outcomes differed between those studied and not studied. The conditional method requires data at all ages of interest, which led to the further loss of 971 (51.7%) participants from the longitudinal analysis. The remaining 48% of participants did not differ significantly in any early life characteristics, except that they were shorter by 0.48 cm at birth.

ed and not studied. The conditional method requires data at all ages of interest, which led to the further loss of 971 (51.7%) participants from the longitudinal analysis. The remaining 48% of participants did not differ significantly in any early life characteristics, except that they were shorter by 0.48 cm at birth. Positive associations of BMI and conditional relative weight gain from midchildhood onward with adult risk markers are in agreement with recent studies from LMICs14 and developed countries10, 14 showing that greater BMI gain during childhood and adolescence is associated with an increased risk of adult hypertension, T2DM, metabolic syndrome, and coronary heart disease. Accelerated weight gain, or upward crossing of weight percentiles, at this stage of the life-course is, thus, associated with higher adult risk in all populations studied,19 even among children who are relatively light and thin in absolute terms, like the Indian children in our study. Weight gain at this stage of life appears to lead to greater fat than lean gain.13, 20 Little is being done in any country to measure children in a way that could pick up this harmful growth pattern. If children are monitored at all with future CVD risk in mind, action tends to be limited to the treatment of established obesity, and ignores upward change within the normal BMI range (“children becoming obese relative to themselves”).

ountry to measure children in a way that could pick up this harmful growth pattern. If children are monitored at all with future CVD risk in mind, action tends to be limited to the treatment of established obesity, and ignores upward change within the normal BMI range (“children becoming obese relative to themselves”). Pediatricians in LMICs tend to promote weight gain during infancy because it is known to increase survival and benefit neurodevelopment. This practice has been questioned recently because of data from developed countries showing that rapid infant weight gain is associated with an increased risk of adult obesity and insulin resistance.21 However, the Hertfordshire and Finland cohorts showed that greater weight or BMI gain in infancy (under 2 years of age) was associated with a lower risk of coronary heart disease and T2DM22, 23 suggesting that infancy (like fetal life) is a period of plasticity during which good nutrition improves the development of metabolically active tissues, and build greater lean body mass. Similar concepts have led to the focus on promoting nutrition during “the first 1000 days” (from conception to the end of the second postnatal year) to promote long-term health. We found that weight gain below the age of 3 months was associated with a higher adult WC but not with other adult risk markers. Overall, we conclude that promoting either linear growth or infant weight gain in LMICs is unlikely to increase later CVD risk.

end of the second postnatal year) to promote long-term health. We found that weight gain below the age of 3 months was associated with a higher adult WC but not with other adult risk markers. Overall, we conclude that promoting either linear growth or infant weight gain in LMICs is unlikely to increase later CVD risk. Taller height and faster linear growth during infancy, childhood, and adolescence were associated with higher WC. Positive associations between height or height gain and BP, insulin resistance, cholesterol, and triglycerides were less consistent and mostly explained by greater adult size, although the associations of BP with linear growth 6.5-15 years of age, and of cholesterol with linear growth 0-3 months of age and 6.5-15 years of age in men remained significant after adjustment for adult size. An association of WC with height is not surprising, given that WC is a measure of frame size as well as of central adiposity. An association between BP and height in children is well described and possibly represents a physiological adaptation to perfuse a longer arterial system.24 Taller people have a lower risk of CVD, despite higher BP,25 so this adaptation may not carry adverse implications for health. Data from the Helsinki cohort has related greater height gain between birth and 7 years of age, to an increased risk of adult hypertension and coronary heart disease.26, 27, 28 Conversely, recent data from the 1946 United Kingdom birth cohort showed that shorter height in early childhood was associated with higher adult carotid intima media thickness.29 The association between childhood height growth and insulin resistance may reflect reverse causality; insulin has growth promoting properties, and higher insulin concentrations in childhood may increase height growth. Reasons for the positive association between infant length growth and adult cholesterol and triglyceride concentrations, however, are unclear. Taller childhood height can reflect an accelerated “tempo” of growth, resulting in earlier maturation and puberty. In western populations, earlier puberty has been associated with increased adult CVD risk factors, for reasons that are not clear.30, 31 Our cohort lacks data on puberty, however, earlier maturation is generally followed by a reduced final height, because of premature epiphyseal fusion, and so this explanation does not fit well with our data, in which WC, BP, and HOMA-IR were positively associated with final height.

for reasons that are not clear.30, 31 Our cohort lacks data on puberty, however, earlier maturation is generally followed by a reduced final height, because of premature epiphyseal fusion, and so this explanation does not fit well with our data, in which WC, BP, and HOMA-IR were positively associated with final height. Future studies examining linear growth in early life in relation to later CVD risk would be useful. Longer birth length was associated with lower glucose and cholesterol in women, and after adjusting for adult size, with lower HOMA-IR and triglycerides. Higher BMI at birth was associated with lower glucose concentration in women, and in men after adjusting for adult size, similar to previous reports that showed inverse associations between birth weight and T2DM.7 Shorter birth length was associated with a higher risk of later T2DM in another Indian birth cohort.32 An association between smaller birth size and higher adult BP, found in many birth cohorts,33 was not seen in our study, as in other Indian studies.34 Associations of small size at birth with later CVD risk markers tend to be weaker in LMICs than in high income settings,21 which may reflect less within-cohort heterogeneity in birth weight, because of a lack of newborns in the upper range of birth weight, and/or lower adult BMI in LMICs; in western birth cohorts, the highest prevalence of T2DM is among men and women who were small at birth and became obese adults.

high income settings,21 which may reflect less within-cohort heterogeneity in birth weight, because of a lack of newborns in the upper range of birth weight, and/or lower adult BMI in LMICs; in western birth cohorts, the highest prevalence of T2DM is among men and women who were small at birth and became obese adults. Monitoring childhood weight and height, and active intervention to prevent or reverse upward crossing of BMI percentiles may reduce later CVD risk. Individuals who had greater linear growth during childhood and/or became taller as adults had higher adult WC, BP, insulin resistance, and cholesterol concentrations. It is not clear whether these associations reflect an increased risk of future CVD,35, 36 and associations between linear growth and adult CVD risk need further investigation. Infant weight gain was positively related to adult WC, but unrelated to BP or the biochemical CVD risk markers, suggesting that the common clinical practice in LMICs of promoting infant weight gain to enhance survival and neurodevelopment is unlikely to have either adverse or beneficial implications for future CVD risk. Appendix Figure 1 Cohort flow chart. *Number of participants used for cross-sectional analysis. Figure 1Table IV Associations of conditional height from birth to 28 years of age with adult height, BMI, and cardiovascular risk factors, including “forward-looking (model 1)” and “backward-looking (model 2)” analyses

Monitoring childhood weight and height, and active intervention to prevent or reverse upward crossing of BMI percentiles may reduce later CVD risk. Individuals who had greater linear growth during childhood and/or became taller as adults had higher adult WC, BP, insulin resistance, and cholesterol concentrations. It is not clear whether these associations reflect an increased risk of future CVD,35, 36 and associations between linear growth and adult CVD risk need further investigation. Infant weight gain was positively related to adult WC, but unrelated to BP or the biochemical CVD risk markers, suggesting that the common clinical practice in LMICs of promoting infant weight gain to enhance survival and neurodevelopment is unlikely to have either adverse or beneficial implications for future CVD risk. Appendix Figure 1 Cohort flow chart. *Number of participants used for cross-sectional analysis. Figure 1Table IV Associations of conditional height from birth to 28 years of age with adult height, BMI, and cardiovascular risk factors, including “forward-looking (model 1)” and “backward-looking (model 2)” analyses Table IV Men Women Model 1 Model 2 Model 1 Model 2 β* 95% CI P value β* 95% CI P value β* 95% CI P value β* 95% CI P value Adult height (cm) Birth 0.30 0.23,0.37 <.001 - - - 0.31 0.25,0.38 <.001 - - - 0-3mo 0.26 0.20,0.32 <.001 - - - 0.20 0.15,0.25 <.001 - - - 3 mo-6.5 y 0.45 0.39,0.50 <.001 - - - 0.45 0.39,0.50 <.001 - - - 6.5-15 y 0.34 0.28,0.40 <.001 - - - 0.40 0.35,0.45 <.001 - - - 15-28 y BMI(kg/m2) Birth 0.10 0.02,0.19 <.001 - - - 0.15 0.07,0.24 .001 - - - 0-3 mo 0.00 -0.07,0.08 .9 - - - 0.06 -0.01,0.13 .1 - - - 3 mo-6.5 y 0.20 0.13,0.28 <.001 - - - 0.07 -0.005,0.14 .06 - - - 6.5-15 y 0.12 0.04,0.19 .003 - - - 0.10 0.03,0.17 .005 - - - 15-28 y WC(cm) Birth 2.47 2.00,2.93 <.001 0.19 -0.26,0.64 .4 2.01 1.59,2.42 <.001 0.14 -0.31,0.59 .5 0-3 mo 1.06 0.67,1.45 <.001 0.24 -0.14,0.61 .2 0.87 0.53,1.21 <.001 -0.34 -0.70,0.01 .06 3 mo-6.5 y 3.28 2.89,3.66 <.001 0.12 -0.31,0.55 .5 2.03 1.67,2.39 <.001 0.48 0.03,0.92 .04 6.5-15 y 2.60 2.20,3.00 <.001 0.30 -0.11,0.70 .1 1.56 1.22,1.90 <.001 -0.19 -0.60,0.23 .3 15-28 y 1.67 1.28,2.07 <.001 - - - 0.6 0.26,0.95 <.001 - - - SBP(mm Hg) Birth 1.04 -0.08,2.15 .07 -0.05 -1.26,1.17 .9 0.45 -0.74,1.64 .4 -0.34 -1.67,0.98 .6 0-3 mo 0.57 -0.37,1.50 .2 0.19 -0.83,1.21 .7 1.21 0.23,2.19 .02 0.69 -0.36,1.73 .2 3 mo-6.5 y 1.83 0.90,2.75 <.001 0.34 -0.84,1.52 .5 0.91 -0.12,1.94 .08 0.19 -1.11,1.49 .7 6.5-15 y 2.46 1.50,3.42 <.001 1.37 0.27,2.48 .02 1.84 0.87,2.82 <.001 1.06 -0.15,2.27 .09 15-28 y 0.73 -0.21,1.68 .1 - - - 0.38 -0.61,1.37 .4 - - - DBP(mm Hg) Birth 1.42 0.52,2.31 <.001 1.00 0.03,1.97 .04 0.23 -0.68,1.15 .6 -0.36 -1.38,0.66 .4 0-3 mo 0.41 -0.34,1.16 .2 0.49 -0.31,1.31 .2 1.34 0.64,2.15 .002 1.01 0.20,1.82 .01 3 mo-6.5 y 1.41 0.75,2.39 <.001 0.95 0.02,1.89 .04 0.87 0.07,1.66 .03 0.51 -0.48,1.52 .3 6.5-15 y 1.95 1.18,2.72 <.001 1.59 0.71,2.47 <.001 1.65 0.90,2.39 <.001 1.19 0.25,2.12 .01 15-28 y -0.31 -1.06,0.45 .4 - - - -0.15 -0.91,0.60 .6 - - - Log glucose 120 min(mmol/L) Birth 0.04 0.01,0.07 .02 0.03 -0.01,0.06 .1 -0.04 -0.07,-0.01 .02 -0.05 -0.08,-0.01 .01 0-3 mo 0.004 -0.02,0.03 .7 0.01 -0.02,0.04 .5 0.01 -0.02,0.03 .4 0.004 -0.02,0.03 .7 3 mo-6.5 y 0.004 -0.02,0.03 .7 -0.01 -0.04,0.02 .6 -0.01 -0.04,0.02 .4 -0.01 -0.

.45 .4 - - - -0.15 -0.91,0.60 .6 - - - Log glucose 120 min(mmol/L) Birth 0.04 0.01,0.07 .02 0.03 -0.01,0.06 .1 -0.04 -0.07,-0.01 .02 -0.05 -0.08,-0.01 .01 0-3 mo 0.004 -0.02,0.03 .7 0.01 -0.02,0.04 .5 0.01 -0.02,0.03 .4 0.004 -0.02,0.03 .7 3 mo-6.5 y 0.004 -0.02,0.03 .7 -0.01 -0.04,0.02 .6 -0.01 -0.04,0.02 .4 -0.01 -0. 05,0.02 .4 6.5-15 y 0.02 -0.003,0.05 .08 0.02 -0.01,0.05 .2 -0.004 -0.03,0.02 .7 -0.01 -0.04,0.02 .5 15-28 y -0.01 -0.04,0.02 .4 - - - -0.01 -0.03,0.02 .5 - - - HOMA-IR Birth 0.11 -0.01,0.24 .08 0.07 -0.06,0.21 .2 -0.16 -0.32,0.001 .05 -0.25 -0.43,-0.07 .01 0-3 mo 0.04 -0.06,0.15 .4 0.03 -0.08,0.15 .5 0.08 -0.06,0.21 .2 0.02 -0.12,0.17 .7 3 mo-6.5 y 0.13 0.02,0.23 .02 0.07 -0.06,0.21 .2 0.05 -0.09,0.19 .4 -0.02 -0.19,0.16 .8 6.5-15 y 0.08 -0.02,0.19 .1 0.05 -0.08,0.17 .4 0.16 0.03,0.29 .02 0.08 -0.08,0.24 .3 15-28 y 0.02 -0.09,0.12 .7 - - - 0.02 -0.11,0.16 .7 - - - Cholesterol(mmol/L) Birth 0.04 -0.06,0.13 .4 0.01 -0.10,0.11 .8 -0.03 -0.12,0.06 .5 -0.05 -0.15,0.05 .3 0-3 mo 0.09 0.01,0.17 .03 0.1 0.01,0.19 .03 0.03 -0.05,0.10 .4 0.01 -0.06,0.09 .7 3 mo-6.5 y 0.04 -0.05,0.12 .3 0.00 -0.10,0.10 .9 0.05 -0.03,0.12 .2 0.06 -0.04,0.15 .2 6.5-15 y 0.13 0.04,0.21 <.001 0.11 0.01,0.20 .03 0.01 -0.06,0.09 .7 0.01 -0.08,0.15 .8 15-28 y -0.03 -0.11,0.05 .4 - - - -0.06 -0.13,0.02 .1 - - - Log triglycerides(mmol/L) Birth 0.02 -0.04,0.07 .5 -0.01 -0.07,0.05 .7 -0.01 -0.06,0.04 .6 -0.05 -0.11,0.01 .08 0-3 mo 0.05 0.01,0.10 .02 0.05 0.01,0.10 .03 0.002 -0.04,0.04 .9 -0.02 -0.07,0.02 .3 3 mo-6.5 y 0.02 -0.02,0.07 .3 -0.01 -0.07,0.04 .6 0.02 -0.03,0.06 .4 -0.009 -0.06,0.05 .7 6.5-15 y 0.10 0.05,0.14 <.001 0.07 0.02,0.12 .01 0.03 -0.01,0.07 .1 0.002 -0.05,0.05 .9 15-28 y -0.01 -0.05,0.03 .6 - - - 0.00 -0.04,0.04 .9 - - - HDL cholesterol(mmol/L) Birth -0.01 -0.03,0.02 .6 0.004 -0.02,0.03 .7 -0.02 -0.05,0.01 .1 -0.01 -0.04,0.02 .4 0-3 mo -0.001 -0.02,0.02 .9 0.005 -0.02,0.03 .6 0.02 0.000,0.05 .05 0.03 0.00,0.05 .02 3 mo-6.5 y -0.001 -0.02,0.02 .9 0.01 -0.01,0.04 .30 -0.005 -0.03,0.02 .7 0.004 -0.03,0.03 .8 6.5-15 y 0.001 -0.02,0.02 .1 0.01 -0.01,0.04 .3 -0.002 -0.02,0.02 .8 0.01 -0.02,0.03 .6 15-28 y -0.01 -0.03,0.01 .2 - - - -0.01 -0.03,0.02 .5 - - - β* represents unit change in outcome variable with unit change in exposure.

.02 3 mo-6.5 y -0.001 -0.02,0.02 .9 0.01 -0.01,0.04 .30 -0.005 -0.03,0.02 .7 0.004 -0.03,0.03 .8 6.5-15 y 0.001 -0.02,0.02 .1 0.01 -0.01,0.04 .3 -0.002 -0.02,0.02 .8 0.01 -0.02,0.03 .6 15-28 y -0.01 -0.03,0.01 .2 - - - -0.01 -0.03,0.02 .5 - - - β* represents unit change in outcome variable with unit change in exposure. For each risk factor,all 10 conditional variables were included together in regression models. Model 1 was adjusted for adult age; model 2 was additionally adjusted for adult size(BMI and height). All models were adjusted for sociodemographic variables(rural or urban residence at birth,rural or urban residence currently,attained education level,adult household possessions score,smoking history,physical activity,and alcohol consumption). Table V Associations of conditional relative weight gain from birth to 28 years of age with adult height, BMI, and cardiovascular risk factors, including “forward-looking (model 1)” and “backward-looking (model 2)” analyses

For each risk factor,all 10 conditional variables were included together in regression models. Model 1 was adjusted for adult age; model 2 was additionally adjusted for adult size(BMI and height). All models were adjusted for sociodemographic variables(rural or urban residence at birth,rural or urban residence currently,attained education level,adult household possessions score,smoking history,physical activity,and alcohol consumption). Table V Associations of conditional relative weight gain from birth to 28 years of age with adult height, BMI, and cardiovascular risk factors, including “forward-looking (model 1)” and “backward-looking (model 2)” analyses Table V Men Women Model 1 Model 2 Model 1 Model 2 β* 95% CI P value β* 95% CI P value β* 95% CI P value β* 95% CI P value Adult height (cm) Birth 0.09 0.03,0.15 .003 - - - 0.11 0.06,0.17 <.001 - - - z weight 0-3 mo 0.10 0.04,0.16 .001 - - - 0.16 0.11,0.21 <.001 - - - z weight 3 mo-6.5 y 0.00 -0.06,0.06 .1 - - - -0.02 -0.08,0.03 .4 - - - z weight 6.5-15 y -0.24 -0.29,-0.18 <.001 - - - -0.28 -0.33,-0.23 <.001 - - - BMI(kg/m2) Birth 0.18 0.11,0.26 <.001 - - - 0.12 0.04,0.19 .002 - - - 0-3mo 0.07 -0.004,0.15 .07 - - - 0.04 -0.03,0.11 .3 - - - 3 mo-6.5 y 0.14 0.06,0.22 <.001 - - - 0.20 0.13,0.27 <.001 - - - 6.5-15 y 0.40 0.33,0.48 <.001 - - - 0.46 0.39,0.54 <.001 - - - WC(cm) Birth 1.85 1.46,2.25 <.001 -0.23 -0.59,0.14 .2 1.46 1.11,1.82 <.001 0.24 -0.12,0.59 .1 0-3 mo 0.67 0.26,1.09 <.001 -0.09 -0.46,0.29 .6 0.07 -0.27,0.41 .6 -0.2 -0.55,0.15 .2 3 mo-6.5 y 0.78 0.37,1.19 <.001 -0.65 -1.02,-0.28 <.001 1.49 1.15,1.83 <.001 -0.02 -0.36,0.32 .9 6.5-15 y 3.04 2.65 3.43 <.001 -0.46 -0.87,-0.04 .03 2.97 2.63,3.32 <.001 -0.82 -1.24,-0.40 <.001 15-28 y 7.67 7.25,8.08 <.001 - - - 6.89 6.53,7.25 <.001 - - - SBP(mm Hg) Birth 0.78 -0.16,1.73 .1 -0.22 -1.21,0.77 .6 0.89 -0.12,1.90 .08 0.39 -0.65,1.43 .4 0-3 mo 0.28 -0.72,1.28 .5 -0.06 -1.08,0.96 .9 0.78 -0.20,1.76 .1 0.63 -0.39,1.65 .2 3 mo-6.5 y 1.06 0.07,2.04 .04 0.37 -0.64,1.37 .4 1.30 0.33,2.27 .01 0.74 -0.26,1.74 .1 6.5-15 y 0.11 -0.83,1.05 .8 -1.63 -2.77,-0.50 <.001 0.98 -0.01,1.97 .05 -0.38 -1.63,0.87 .5 15-28 y 4.03 3.03,5.03 <.001 - - - 2.79 1.75,3.83 <.001 - - - DBP(mm Hg) Birth 0.60 -0.16,1.36 .1 -0.07 -0.86,0.71 .8 0.50 -0.27,1.28 .2 0.05 -0.74,0.86 .8 0-3 mo 0.61 -0.18,1.41 .1 0.47 -0.33,1.29 .2 0.44 -0.30,1.20 .2 0.47 -0.31,1.25 .2 3 mo-6.5 y 0.24 -0.55,1.03 .5 -0.30 -1.10,0.49 .4 0.66 -0.07,1.41 .07 0.003 -0.76,0.77 .9 6.5-15 y 0.26 -0.49,1.02 .4 -1.49 -2.39,-0.59 .001 1.09 0.33,1.85 .005 -0.76 -1.72,0.20 .1 15-28 y 3.08 2.27,3.88 <.001 - - - 3.09 2.29,3.89 <.001 - - - Log glucose 120 min(mmol/L) Birth -0.02 -0.04,0.01 .2 -0.03 -0.06,-0.01 .02 -0.02 -0.04,0.01 .1 -0.03 -0.05,0.00 .06 0-3 mo -0.01 -0.03,0.02 .6 -0.01 -0.04,0.02 .4 -0.01 -0.03,0.02 .5 -0.01 -0.03,0.02 .6 3 mo-6.5 y 0.00 -0.03,0.03 .8 -0.01 -0.04,0.01

.20 .1 15-28 y 3.08 2.27,3.88 <.001 - - - 3.09 2.29,3.89 <.001 - - - Log glucose 120 min(mmol/L) Birth -0.02 -0.04,0.01 .2 -0.03 -0.06,-0.01 .02 -0.02 -0.04,0.01 .1 -0.03 -0.05,0.00 .06 0-3 mo -0.01 -0.03,0.02 .6 -0.01 -0.04,0.02 .4 -0.01 -0.03,0.02 .5 -0.01 -0.03,0.02 .6 3 mo-6.5 y 0.00 -0.03,0.03 .8 -0.01 -0.04,0.01 .2 0.00 -0.03,0.02 .8 -0.02 -0.04,0.01 .2 6.5-15 y 0.02 -0.01,0.04 .2 -0.03 -0.06,0.00 .09 0.01 -0.01,0.04 .3 -0.03 -0.06,0.00 .09 15-28 y 0.07 0.04,0.10 <.001 - - - 0.06 0.03,0.08 <.001 - - - HOMA-IR Birth 0.06 -0.05,0.16 .2 0.02 -0.09,0.13 .7 0.08 -0.05,0.22 .2 0.03 -0.11,0.17 .6 0-3 mo 0.04 -0.07,0.15 .4 0.03 -0.09,0.14 .6 0.06 -0.08,0.19 .4 0.05 -0.09,0.18 .5 3 mo-6.5 y -0.04 -0.15,0.07 .4 -0.07 -0.18,0.04 .2 0.15 0.02,0.28 .07 0.08 -0.06,0.21 .2 6.5-15 y -0.11 -0.21,0.000 .05 -0.19 -0.31,-0.06 <.001 0.00 -0.13,0.14 .9 -0.18 -0.35,-0.01 .04 15-28 y 0.16 0.05,0.27 .01 - - - 0.31 0.17,0.45 <.001 - - - Cholesterol(mmol/L) Birth 0.08 -0.01,0.16 .0 0.03 -0.06,0.11 .5 0.03 -0.05,0.10 .5 0.00 -0.08,0.08 .9 0-3 mo 0.00 -0.09,0.08 .9 -0.01 -0.10,0.08 .8 -0.05 -0.12,0.03 .2 -0.03 -0.11,0.05 .4 3 mo-6.5 y -0.03 -0.12,0.06 .4 -0.07 -0.16,0.02 .1 0.04 -0.03,0.11 .2 -0.01 -0.09,0.06 .7 6.5-15 y 0.10 0.02,0.19 .01 -0.03 -0.13,0.07 .5 0.04 -0.04,0.11 .3 -0.12 -0.22,-0.03 .01 15-28 y 0.24 0.15,0.33 <.001 - - - 0.21 0.13,0.29 <.001 - - - Log triglycerides (mmol/L) Birth 0.04 -0.01,0.08 .0 0.001 -0.05,0.05 .9 -0.02 -0.06,0.02 .3 -0.05 -0.09,-0.002 .04 0-3 mo 0.02 -0.03,0.07 .3 0.01 -0.03,0.06 .5 0.000 -0.04,0.04 .9 -0.002 -0.05,0.04 .9 3 mo-6.5 y -0.03 -0.07,0.02 .2 -0.06 -0.10,-0.01 .02 0.03 -0.02,0.07 .2 -0.01 -0.05,0.03 .5 6.5-15 y 0.07 0.02,0.11 <.001 -0.03 -0.08,0.02 .2 0.05 0.004,0.09 .03 -0.06 -0.11,-0.01 .03 15-28 y 0.18 0.13,0.23 <.001 - - - 0.16 0.12,0.20 <.001 - - - HDL cholesterol (mmol/L) Birth 0.01 -0.01,0.03 .3 0.02 -0.003,0.04 .1 0.003 -0.02,0.03 .8 0.01 -0.02,0.03 .5 0-3 mo -0.01 -0.03,0.01 .4 -0.01 -0.03,0.02 .6 -0.003 -0.03,0.02 .8 0.000 -0.02,0.02 .9 3 mo-6.5 y -0.002 -0.02,0.02 .8 0.003 -0.02,0.02 .8 -0.01 -0.03,0.01 .4 -0.01 -0.03,0.02 .6 6.5-15 y -0.01 -0.03,0.01 .1 -0.01 -0.03,0.02 .5 -0.02 -0.04,0.01 .1 -0.01 -0.04,0.02 .5 15-28 y -0.02 -0.04,0.000 .05 - - - -0.02 -0.04,0.01 .1 - - - β* represents unit change in outcome variable with unit change in exposure.

9 3 mo-6.5 y -0.002 -0.02,0.02 .8 0.003 -0.02,0.02 .8 -0.01 -0.03,0.01 .4 -0.01 -0.03,0.02 .6 6.5-15 y -0.01 -0.03,0.01 .1 -0.01 -0.03,0.02 .5 -0.02 -0.04,0.01 .1 -0.01 -0.04,0.02 .5 15-28 y -0.02 -0.04,0.000 .05 - - - -0.02 -0.04,0.01 .1 - - - β* represents unit change in outcome variable with unit change in exposure. For each risk factor,all 10 conditional variables were included together in regression models. Model 1 was adjusted for adult age; model 2 was additionally adjusted for adult size(BMI and height). All models were adjusted for sociodemographic variables(rural or urban residence at birth,rural or urban residence currently,attained education level,adult household possessions score,smoking history,physical activity,and alcohol consumption). We thank the subjects who participated, the medical officers, and the field and laboratory staff for their kind cooperation. Funded by the Medical Research Council (United Kingdom) (MRC_G0400519, MRC_MC_UP_A620_1016, and MRC_MC_UU_12011/3), the Department for International Development, the British Heart Foundation (BHF_RG/98001 and BHF_CS/15/4/31493), and Society for Nutrition, Education and Health Action (India). S.V. is funded by the Throne-Holst Foundation, Stockholm, Sweden. The authors declare no conflicts of interest.

0_1016, and MRC_MC_UU_12011/3), the Department for International Development, the British Heart Foundation (BHF_RG/98001 and BHF_CS/15/4/31493), and Society for Nutrition, Education and Health Action (India). S.V. is funded by the Throne-Holst Foundation, Stockholm, Sweden. The authors declare no conflicts of interest. Figure 2 A-G, Associations of conditional height and conditional relative weight gain from birth to 28 years of age with adult height, BMI, and cardiovascular risk factors, including “forward-looking (unadjusted for adult size)” and “backward-looking (adjusted for adult size)” analyses. Estimates represented as unit increase B, in outcome measurement along with 95% CI with unit increase in exposure variable. Gray squares represent men, and black circles represent women. Figure 2Table I Characteristics of the study participants

Figure 2 A-G, Associations of conditional height and conditional relative weight gain from birth to 28 years of age with adult height, BMI, and cardiovascular risk factors, including “forward-looking (unadjusted for adult size)” and “backward-looking (adjusted for adult size)” analyses. Estimates represented as unit increase B, in outcome measurement along with 95% CI with unit increase in exposure variable. Gray squares represent men, and black circles represent women. Figure 2Table I Characteristics of the study participants Table I n Males n Females Height (cm) Birth 981 48.4(2.7) 897 47.9(2.8) 3 mo 834 58.4(3.1) 779 57.4(3.0) 6.5 y 874 107.0(5.4) 806 106.3(5.5) 15 y 558 150.5(9.0) 550 148.8(6.1) Adult 981 166.4(6.6) 897 153.9(5.9) Weight(kg) Birth 981 2.84(0.49) 897 2.75(0.49) 3 mo 834 4.99(0.82) 779 4.68(0.72) 6.5 y 874 16.0(2.0) 806 15.8(2.0) 15 y 558 35.7(6.7) 550 37.3(6.0) Adult 981 57.2(11.2) 897 49.1(10.4) BMI(kg/m2) Birth 981 12.1(1.8) 897 12.0(1.8) 3 mo 834 14.7(2.1) 779 14.2(1.9) 6.5 y 874 13.9(1.4) 806 14.0(1.4) 15 y 558 15.6(1.5) 550 16.8(2.0) Adult 981 20.6(3.4) 897 20.7(4.0) Adult variables Age (y) 981 28.0(1.0) 897 28.2(0.9) WC (cm) 981 77.3(10.5) 897 70.8(9.4) Hip circumference(cm) 981 86.8(7.3) 897 88.7(8.7) SBP (mm Hg) 981 112.2(11.2) 897 101.4(10.8) DBP (mm Hg) 981 72.7(9.3) 897 72.4(8.6) Total cholesterol (mmol/L) 981 4.0(0.9) 897 3.9(0.8) HDL cholesterol (mmol/L) 970 0.98(0.24) 893 1.09(0.24) LDL cholesterol (mmol/L) 968 2.5(0.7) 893 2.4(0.7) Triglyceride (mmol/L)* 980 1.0(0.8,1.5) 897 0.8(0.6,1.1) Fasting glucose (mg/dL)* 981 98(92,104) 897 96(90,102) 2-h glucose(mg/dL)* 981 110(94,131) 897 115(99,134) Fasting insulin (mIU/mL)* 981 5.7(2.6,9.7) 897 3.2(1.2,6.2) 2-h insulin (mIU/mL)* 981 21.7(11.9,39.1) 897 20.8(10.7,37.3) HOMA-IR* 981 1.35(0.60,2.41) 897 0.77(0.28,1.49) Urban at birth(%) 981 29.7 897 26.1 Urban as adult(%) 981 45.7 897 41.8 Education: ≤ primary (%) 981 18 897 36.7 Education: middle and high (%) 981 57.1 897 48.9 Education: > high (%) 981 24.9 897 14.3 Material possessions score 981 5(3.8) 897 5(2.7) Current smoker (%) 981 42.9 897 2.7 Alcohol consumption: any (%) 981 54 897 0.8 Physical activity score 981 1407(832,1890) 897 1854(1427,2409) HDL,high-density lipoprotein; LDL,low-density lipoprotein.

nd high (%) 981 57.1 897 48.9 Education: > high (%) 981 24.9 897 14.3 Material possessions score 981 5(3.8) 897 5(2.7) Current smoker (%) 981 42.9 897 2.7 Alcohol consumption: any (%) 981 54 897 0.8 Physical activity score 981 1407(832,1890) 897 1854(1427,2409) HDL,high-density lipoprotein; LDL,low-density lipoprotein. “High” education refers to higher secondary school,diplomas,other graduates,and higher degrees. * For skewed variables,data are presented as median and IQR. Table II Cross-sectional analysis of z score height from birth to adulthood and adult cardiovascular risk outcomes

nd high (%) 981 57.1 897 48.9 Education: > high (%) 981 24.9 897 14.3 Material possessions score 981 5(3.8) 897 5(2.7) Current smoker (%) 981 42.9 897 2.7 Alcohol consumption: any (%) 981 54 897 0.8 Physical activity score 981 1407(832,1890) 897 1854(1427,2409) HDL,high-density lipoprotein; LDL,low-density lipoprotein. “High” education refers to higher secondary school,diplomas,other graduates,and higher degrees. * For skewed variables,data are presented as median and IQR. Table II Cross-sectional analysis of z score height from birth to adulthood and adult cardiovascular risk outcomes Table II Men Women Model 1 Model 2 Model 1 Model 2 β (95% CI) P value β (95% CI) P value β (95% CI) P value β (95% CI) P value WC(cm) Birth 1.27(0.60,1.94) <.001 -0.13(-0.40,0.15) .4 0.99(0.37,1.62) .002 -0.14(-0.42,0.14) .3 3 mo 0.99(0.32,1.66) .004 0.04(-0.24,0.31) .8 0.90(0.30,1.49) .003 -0.43(-0.69,-0.16) .002 6.5 y 2.57(1.94,3.21) <.001 0.01(-0.31,0.34) .9 1.69(1.09,2.29) <.001 0.14(-0.16,0.45) .4 15 y 3.51(2.71,4.31) <.001 0.35(-0.09,0.80) .1 3.10(2.33,3.88) <.001 -0.21(-0.74,0.32) .4 28 y 2.32(1.70,2.94) <.001 - 1.63(1.04,2.21) <.001 - SBP (mm Hg) Birth 0.55(-0.21,1.30) .2 -0.36(-1.10,0.37) .3 -0.37(-1.15,-0.41) .4 -1.03(-1.80,-0.27) .008 3 mo 0.52(-0.22,1.26) .2 -0.27(-1.00,0.46) .5 0.59(-0.16,1.33) .1 -0.11(-0.86,0.65) .8 6.5 y 1.65(0.90,2.40) <.001 -0.16(-1.03,0.70) .7 1.10(0.33,1.89) .006 0.10(-0.76,0.97) .8 15 y 2.89(1.95,3.82) <.001 1.53(0.29,2.76) .02 2.55(1.54,3.57) <.001 1.28(-0.20,2.76) .09 28 y 2.10(1.40,2.80) <.001 - 1.21(0.48,1.95) .001 - Log glucose 120-min(mmol/L) Birth 0.01(-0.01,0.03) .3 0.01(-0.01,0.03) .5 -0.02(-0.04,-0.00) .04 -0.2(-0.04,0.00) .06 3 mo 0.01(-0.01,0.03) .3 0.01(-0.01,0.03) .2 0.01(-0.01,0.02) .6 0.01(-0.01,0.03) .3 6.5 y -0.00(-0.02,0.02) .8 -0.01(-0.04,0.01) .3 -0.02(-0.03,0.00) .1 -0.01(-0.03,0.01) .4 15 y 0.02(-0.01,0.04) .2 0.01(-0.03,0.04) .7 -0.02(-0.05,0.00) .08 -0.02(-0.06,0.02) .3 28 y -0.01(-0.03,0.01) .3 - -0.03(-0.04,-0.01) .003 - Log HOMA-IR Birth 0.07(-0.01,0.15) .1 0.04(-0.05,0.12) .4 -0.05(-0.15,0.05) .4 -0.11(-0.21,0.01) .03 3 mo -0.05(-0.15,0.05) .4 0.03(-0.06,0.12) .5 0.08(-0.02,0.17) .1 0.01(-0.09,0.11) .8 6.5 y 0.08(-0.02,0.17) .1 -0.02(-0.12,0.08) .7 0.07(-0.04,0.17) .2 -0.03(-0.15,0.08) .6 15 y 0.07(-0.04,0.17) .2 -0.02(-0.15,0.12) .8 0.16(0.02,0.29) .03 -0.02(-0.22,0.18) .8 28 y 0.08(0.00,0.16) .05 - 0.12(0.02,0.21) .02 - Cholesterol(mmol/L) Birth -0.01(-0.07,0.05) .7 -0.04(-0.10,0.02) .2 -0.02(-0.07,-0.04) .6 -0.02(-0.08,0.04) .5 3 mo 0.03(-0.03,0.09) .3 0.03(-0.03,0.09) .4 0.01(-0.05,0.06) .8 -0.00(-0.06,0.06) 1.0 6.5 y 0.05(-0.01,0.11) .1 -0.01(-0.08,0.06) .7 -0.11(-0.06,0.06) .9 0.01(-0.06,0.07) .8 15 y 0.13(0.05,0.20) .002 0.11(0.00,0.21) .04 -0.01(-0.09,0.07) .8 -0.02(-0.13,0.09) .7 28 y 0.

02) .2 -0.02(-0.07,-0.04) .6 -0.02(-0.08,0.04) .5 3 mo 0.03(-0.03,0.09) .3 0.03(-0.03,0.09) .4 0.01(-0.05,0.06) .8 -0.00(-0.06,0.06) 1.0 6.5 y 0.05(-0.01,0.11) .1 -0.01(-0.08,0.06) .7 -0.11(-0.06,0.06) .9 0.01(-0.06,0.07) .8 15 y 0.13(0.05,0.20) .002 0.11(0.00,0.21) .04 -0.01(-0.09,0.07) .8 -0.02(-0.13,0.09) .7 28 y 0. 00(-0.05,0.06) .9 - -0.06(-0.11,-0.00) .05 - Log triglycerides(mmol/L) Birth -0.01(-0.04,0.03) .8 -0.03(-0.06,0.01) .1 -0.02(-0.06,0.01) .2 -0.04(-0.07,-0.01) .02 3 mo 0.02(-0.02,0.05) .3 0.01(-0.02,0.05) .6 0.00(-0.03,0.03) .9 -0.02(-0.05,0.02) .3 6.5 y 0.01(-0.04,0.03) .8 -0.07(-0.11,-0.03) .001 0.00(-0.03,0.03) 1.0 -0.01(-0.05,0.02) .5 15 y 0.06(0.01,0.10) .01 0.03(-0.03,0.09) .3 0.03(-0.02,0.07) .2 -0.01(-0.08,0.05) .7 28 y 0.02(-0.02,0.05) .3 - -0.01(-0.04,0.02) .6 - HDL cholesterol(mmol/L) Birth -0.00(-0.02,0.01) .8 0.00(-0.01,0.02) .7 -0.02(-0.03,0.00) .1 -0.01(-0.03,0.01) .3 3 mo 0.01(-0.01,0.02) .6 0.01(-0.01,0.03) .3 0.00(-0.02,0.02) .8 0.01(-0.01,0.03) .3 6.5 y 0.01(-0.01,0.03) .2 0.03(0.01,0.05) .009 -0.00(-0.02,0.01) .7 0.01(-0.01,0.03) .5 15 y 0.01(-0.01,0.03) .3 0.03(-0.00,0.05) .06 -0.01(-0.03,0.02) .5 0.02(-0.02,0.05) .3 28 y -0.01(-0.03,0.1) .2 - -0.01(-0.03,0.00) .1 - Separate regression analyses were performed for height at each age.

.02) .8 0.01(-0.01,0.03) .3 6.5 y 0.01(-0.01,0.03) .2 0.03(0.01,0.05) .009 -0.00(-0.02,0.01) .7 0.01(-0.01,0.03) .5 15 y 0.01(-0.01,0.03) .3 0.03(-0.00,0.05) .06 -0.01(-0.03,0.02) .5 0.02(-0.02,0.05) .3 28 y -0.01(-0.03,0.1) .2 - -0.01(-0.03,0.00) .1 - Separate regression analyses were performed for height at each age. β represents unit change in outcome variable with unit change in z height. Number of individuals at each time point were: (males): 981 at birth, 834 at 3 months, 874 at 6.5 years, 558 at 15 years, and 981 at 28 years; and (females): 897 at birth, 779 at 3 months, 806 at 6.5 years, 550 at 15 years, and 897 at 28 years. Model 1 was adjusted for adult age, Model 2 was additionally adjusted for adult size (BMI and height). Other covariates included in both models were age, rural/urban residence (at birth and currently), education, household material possessions, smoking, alcohol consumption, and physical activity. Table III Cross-sectional analysis of z score BMI from birth to adulthood and adult cardiovascular risk outcomes

β represents unit change in outcome variable with unit change in z height. Number of individuals at each time point were: (males): 981 at birth, 834 at 3 months, 874 at 6.5 years, 558 at 15 years, and 981 at 28 years; and (females): 897 at birth, 779 at 3 months, 806 at 6.5 years, 550 at 15 years, and 897 at 28 years. Model 1 was adjusted for adult age, Model 2 was additionally adjusted for adult size (BMI and height). Other covariates included in both models were age, rural/urban residence (at birth and currently), education, household material possessions, smoking, alcohol consumption, and physical activity. Table III Cross-sectional analysis of z score BMI from birth to adulthood and adult cardiovascular risk outcomes Table III Men Women Model 1 Model 2 Model 1 Model 2 β (95% CI) P value β (95% CI) P value β (95% CI) P value β (95% CI) P value WC(cm) Birth 0.95(0.27,1.63) .007 -0.21(-0.48,0.05) .1 0.67(0.06,1.29) .03 0.04(-0.22,0.31) .8 3 mo 1.22(0.53,1.91) .001 -0.18(-0.45,0.09) .2 0.89(0.23,1.55) .009 -0.08(-0.37,0.20) .6 6.5 y 0.66(0.03,1.29) .04 -0.46(-0.72,-0.20) <.001 0.99(0.39,1.60) .001 -0.01(-0.27,0.26) 1.0 15 y 4.14(3.66,5.17) <.001 -0.57(-0.94,-0.20) .003 3.49(2.82,4.16) <.001 -0.70(-1.07,-0.33) <.001 28 y 9.10(8.79,9.41) <.001 - 8.22(7.91,8.53) <.001 - SBP(mm Hg) Birth -0.02(-0.78,0.75) 1.0 -0.54(-1.26,0.18) .1 0.42(-0.35,1.19) .3 0.14(-0.59,0.87) .7 3 mo 0.27(-0.50,1.03) .5 -0.47(-1.19,0.26) .2 0.68(-0.14,1.51) .1 0.23(-0.57,1.03) .6 6.5 y 0.30(-0.43,1.02) .4 -0.04(-0.73,0.65) .9 0.20(-0.58,0.99) .6 -0.15(-0.91,0.61) .7 15 y 1.60(0.65,2.54) .001 -0.62(-1.66,0.43) .2 1.74(0.81,2.66) <.001 0.17(-0.89,1.22) .8 28 y 3.91(3.22,4.60) <.001 - 3.59(2.84,4.33) <.001 - Log glucose 120-min(mmol/L) Birth -0.01(-0.03,0.01) .2 -0.02(-0.04,-0.00) .03 -0.02(-0.04,0.00) .05 -0.02(-0.04,-0.00) .02 3 mo -0.00(-0.02,0.02) .9 -0.01(-0.03,0.01) .4 -0.01(-0.03,0.01) .3 -0.01(-0.03,0.01) .4 6.5 y -0.01(-0.03,0.01) .3 -0.02(-0.04,0.01) .01 -0.02(-0.03,0.00) .1 -0.03(-0.04,-0.01) .007 15 y 0.00(-0.02,0.03) .9 -0.04(-0.07,-0.01) .003 0.00(-0.02,0.02) 1 -0.03(-0.06,-0.01) .01 28 y 0.07(0.05,0.09) <.001 - 0.05(0.03,0.06) <.001 - HOMA-IR Birth 0.06(-0.03,0.14) .2 0.03(-0.05,0.12) .4 0.05(-0.04,0.15) .3 0.03(-0.07,0.13) .6 3 mo 0.03(-0.05,0.12) .5 -0.00(-0.09,0.08) 1.0 -0.01(-0.12,0.10) .8 -0.06(-0.16,0.05) .3 6.5 y -0.03(-0.11,0.05) .5 -0.05(-0.12,0.03) .3 0.08(-0.02,0.18) .1 0.05(-0.05,0.15) .4 15 y -0.03(-0.13,0.07) .6 -0.12(-0.23,-0.01) .04 0.09(-0.03,0.21) .1 -0.11(-0.25,0.03) .1 28 y 0.18(0.10,0.26) <.001 - 0.29(0.19,0.38) <.001 - Cholesterol(mmol/L) Birth -0.00(-0.06,0.06) 1.0 -0.03(-0.09,0.03) .3 0.02(-0.04,0.08) .5 0.01(-0.05,0.06) .8 3 mo -0.00(-0.07,0.06) .9 -0.04(-0.10,0.02) .2 -0.00(-0.06,0.06) 1.0 -0.02(-0.08,0.04) .6 6.5 y -0.01(-0.07,0.04) .6 -0.06(-0.11,0.00) .05 0.01(-0.05,0.06) .9 -0.03(-0.09,0.02) .3 15 y 0.13(0.06,0.21) .001 -0.02(-0.10,0.07) .7 0.03(-0.04,0.10) .4 -0.10(-

1.0 -0.03(-0.09,0.03) .3 0.02(-0.04,0.08) .5 0.01(-0.05,0.06) .8 3 mo -0.00(-0.07,0.06) .9 -0.04(-0.10,0.02) .2 -0.00(-0.06,0.06) 1.0 -0.02(-0.08,0.04) .6 6.5 y -0.01(-0.07,0.04) .6 -0.06(-0.11,0.00) .05 0.01(-0.05,0.06) .9 -0.03(-0.09,0.02) .3 15 y 0.13(0.06,0.21) .001 -0.02(-0.10,0.07) .7 0.03(-0.04,0.10) .4 -0.10(- 0.18,-0.02) .01 28 y 0.28(0.23,0.34) <.001 - 0.21(0.15,0.27) <.001 - Log triglycerides(mmol/L) Birth -0.00(-0.04,0.03) .9 -0.03(-0.06,0.01) .1 -0.00(-0.03,0.03) .8 -0.02(-0.05,0.02) .3 3 mo 0.02(-0.02,0.05) .4 -0.01(-0.05,0.02) .5 0.00(-0.03,0.04) 1.0 -0.02(-0.05,0.02) .4 6.5 y -0.01(-0.05,0.02) .5 -0.04(-0.07,-0.01) .02 0.02(-0.01,0.05) .3 -0.01(-0.04,0.02) .6 15 y 0.07(0.03,0.12) .001 -0.04(-0.09,0.01) .1 0.03(-0.01,0.07) .1 -0.08(-0.12,-0.03) .001 28 y 0.20(0.16,0.23) <.001 - 0.17(0.13,0.20) <.001 - HDL cholesterol(mmol/L) Birth 0.00(-0.02,0.02) .8 0.01(-0.01,0.02) .5 0.00(-0.01,0.02) .7 0.01(-0.01,0.02) .4 3 mo -0.00(-0.02,0.02) .8 0.00(-0.02,0.02) .8 -0.01(-0.03,0.01) .3 -0.01(-0.02,0.01) .6 6.5 y 0.00(-0.01,0.02) .6 0.01(-0.01,0.02) .3 -0.01(-0.03,0.01) .2 -0.01(-0.03,0.01) .3 15 y -0.01(0.03,0.01) .5 0.01(-0.01,0.04) .3 -0.02(-0.04,0.01) .1 0.00(-0.02,0.03) .8 28 y -0.02(-0.04,-0.01) <.001 - -0.04(-0.05,-0.02) <.001 - Separate regression analyses were performed for BMI at each age.

3 -0.01(-0.02,0.01) .6 6.5 y 0.00(-0.01,0.02) .6 0.01(-0.01,0.02) .3 -0.01(-0.03,0.01) .2 -0.01(-0.03,0.01) .3 15 y -0.01(0.03,0.01) .5 0.01(-0.01,0.04) .3 -0.02(-0.04,0.01) .1 0.00(-0.02,0.03) .8 28 y -0.02(-0.04,-0.01) <.001 - -0.04(-0.05,-0.02) <.001 - Separate regression analyses were performed for BMI at each age. β represents unit change in outcome variable with unit change in z BMI. Total number of individuals at each time point were: (males) 981 at birth, 834 at 3 months, 874 at 6.5 years, 558 at 15 years, and 981 at 28 years; and (females): 897 at birth, 779 at 3 months, 806 at 6.5 years, 550 at 15 years, and 897 at 28 years. Model 1 was adjusted for adult age, Model 2 was additionally adjusted for adult size (BMI and height). Other covariates included in both models were age, rural/urban residence (at birth and currently), education, household material possessions, smoking, alcohol consumption, and physical activity.

Physical activity is now well established as important to both the current and future health of children and adolescents.1 Higher levels of physical activity in childhood are associated with favorable metabolic and cardiovascular disease risk profiles,2 increased well-being, and better cognitive and motor development.3, 4 Studies using accelerometers have been conducted mainly in well-nourished children.5, 6, 7, 8, 9 However, little is known about physical activity and their correlates among young children with acute malnutrition, a condition which is likely to affect health and developmental outcomes.10 We are aware of 1 study using accelerometers in Ethiopian children admitted to hospital with severe acute malnutrition (SAM),11 and studies using questionnaire or direct observation methods for children with moderate wasting in India and Mozambique.12, 13 No studies investigating physical activity and correlates using accelerometry are available in children with moderate acute malnutrition (MAM), defined as weight-for-height z score (WHZ) between -3 and -2 (World Health Organization [WHO] 2006),14 and/or a midupper-arm circumference (MUAC) between 115 and 125 mm.15 We aimed to assess the level of accelerometer-based physical activity among 6- to 23-month-old children with MAM in Burkina Faso and to identify clinical, biochemical, anthropometric, and sociodemographic correlates of physical activity.

Studies using accelerometers have been conducted mainly in well-nourished children.5, 6, 7, 8, 9 However, little is known about physical activity and their correlates among young children with acute malnutrition, a condition which is likely to affect health and developmental outcomes.10 We are aware of 1 study using accelerometers in Ethiopian children admitted to hospital with severe acute malnutrition (SAM),11 and studies using questionnaire or direct observation methods for children with moderate wasting in India and Mozambique.12, 13 No studies investigating physical activity and correlates using accelerometry are available in children with moderate acute malnutrition (MAM), defined as weight-for-height z score (WHZ) between -3 and -2 (World Health Organization [WHO] 2006),14 and/or a midupper-arm circumference (MUAC) between 115 and 125 mm.15 We aimed to assess the level of accelerometer-based physical activity among 6- to 23-month-old children with MAM in Burkina Faso and to identify clinical, biochemical, anthropometric, and sociodemographic correlates of physical activity. Methods This is a cross-sectional analysis of baseline data from the TreatFOOD study (Controlled-Trials.com: ISRCTN42569496) among 1609 children with MAM. The activity measures were registered as a secondary outcome. The study was conducted in the Province du Passoré, Burkina Faso, at 5 local health centers (Gomponsom, Latoden, Bagaré, Bokin, and Samba) and a nongovernmental organization (Alliance for International Medical Action, Dakar, Senegal). Children were screened by community health workers using MUAC tapes or by designated screening teams with the use of both MUAC and WHZ. Furthermore, children could be referred to a study site from a health center or present at a site on a caretaker's initiative. The final assessment of study inclusion eligibility was performed at site. Children were enrolled if a diagnosis of MAM was confirmed, defined as WHZ between -3 and -2 (WHO 2006)14 and/or MUAC between 115 and 125 mm.15 In the study site, WHZ was determined using WHO field tables, but anthropometry was later recalculated before analysis. Children were not included if treated for SAM or hospitalized within the past 2 months, were participating in a nutritional program, required hospitalization, or had severe disability.

5 and 125 mm.15 In the study site, WHZ was determined using WHO field tables, but anthropometry was later recalculated before analysis. Children were not included if treated for SAM or hospitalized within the past 2 months, were participating in a nutritional program, required hospitalization, or had severe disability. The study protocol was approved by the Ethics Committee for Health Research in Burkina Faso (2012-8-059), and consultative approval was obtained from the Danish National Committee on Biomedical Research Ethics (1208204). Consent was obtained verbally and in writing (signature or fingerprints) from caretakers of the children before inclusion. The study was carried out in accordance with the declaration of Helsinki and international ethical guidelines for biomedical research involving human subjects, published by the Council for International Organizations of Medical Sciences. Medical treatment was provided according to an adapted version of the Integrated Management of Childhood Illness guidelines.16

the declaration of Helsinki and international ethical guidelines for biomedical research involving human subjects, published by the Council for International Organizations of Medical Sciences. Medical treatment was provided according to an adapted version of the Integrated Management of Childhood Illness guidelines.16 Sociodemographic, Clinical, Biochemical, and Anthropometric Data Collection At enrollment, a nurse conducted a clinical examination and collected data using structured questionnaires for sociodemographic variables (number of people in the household, house ownership, fuel used in cooking, type of employment, child birth day) and breastfeeding status (breastfed or not on the day of enrollment). Fever was defined as axillary temperature ≥37.5°C. Upper and lower respiratory tract infections were diagnosed by experienced pediatric nurses based on an adapted version of the Integrated Management of Childhood Illness.16 The morbidity data presented were collected at enrollment when initiating the activity measurement, and not repeated during the measurement period. Venous blood (2.5 mL) was collected to carry out rapid antigen test for Plasmodium falciparum malaria (SD Bioline, Malaria antigen P.f.), and to determine hemoglobin level (HB 301; HemoCue, Ängelholm, Sweden); anemia was defined as <11 g/dL. Serum was separated and stored at -20°C. C-reactive protein (CRP) and α1-acid glycoprotein (AGP) were determined using a simple sandwich enzyme-linked immunosorbent assay.17 We defined CRP ≥10 mg/L and AGP ≥1 g/L as abnormal, indicating systemic inflammation. Weight (model 881; Seca, Hamburg, Germany) and length (wooden length board) were measured to the nearest 100 g and 1 mm, respectively. MUAC was measured to nearest 1 mm at the midpoint between the olecranon and the acromion process using a standard measuring tape. All measurements were done in duplicate. The anthropometry measurements were done by trained staff and equipment was checked daily. Standardization sessions were carried out prior to the start of the trial to ensure precision and accuracy of measurements. During the trial, anthropometry staff were closely supervised by the anthropometry team leader and the site supervisor. Movement ability of the children was defined as not able to crawl/walk, able to crawl, or able to walk as assessed by measurement staff based on observation using an adapted version of the Malawi Developmental Assessment Tool.18

try staff were closely supervised by the anthropometry team leader and the site supervisor. Movement ability of the children was defined as not able to crawl/walk, able to crawl, or able to walk as assessed by measurement staff based on observation using an adapted version of the Malawi Developmental Assessment Tool.18 Physical Activity Measures and Questionnaire Data Physical activity was measured objectively using a triaxial accelerometer (ActiGraph GT3X+; ActiGraph, Pensacola, Florida). The accelerometer was attached to an elastic belt placed on the skin at the right side of the hip and worn for 6 consecutive days (6 × 24 hours). Caretakers were instructed to only let enrolled children wear the device and to make sure that the accelerometer was placed on the right hip during the monitoring period. Monitors could be removed during bathing. We used data recorded by the device starting 7 hours after leaving the clinic and ending 7 hours before returning to the clinic to avoid recording unusual activity caused by the need to attend the clinical appointments. After monitor removal, the caregiver was interviewed using a structured physical activity questionnaire including perception and acceptability of the device, episodes of device removal, and whether children were carried and if so how many times per day (coded as never, 1-2 times per day, 3-6 times per day, more than one-half of the day, or all the day).

r was interviewed using a structured physical activity questionnaire including perception and acceptability of the device, episodes of device removal, and whether children were carried and if so how many times per day (coded as never, 1-2 times per day, 3-6 times per day, more than one-half of the day, or all the day). Data Analyses The recorded activity data were uploaded from the monitors using the Actilife 6 Software (ActiGraph). Raw accelerometer data were collected at a rate of 100 Hz. Data were integrated to 10-second epochs to permit detection of short bouts of activity.6, 19 Each axis (x, y, and z) was converted to counts per min (but still in 10-second resolution), following which vector magnitude was calculated as the square root of sum of the three squared count values. We included data from all times of the measurement period including night (and other sleep) time in the analysis, except the 7 hours in the beginning and end of the file (see above) and periods marked as nonwear. We defined nonwear time as continuous runs of zero activity ≥90 minutes. Days with <8 hours valid wear data and participants with <1 valid day of recording were excluded from the present analyses. We calculated total physical activity as mean vector magnitude over valid days (counts per minute, cpm).

marked as nonwear. We defined nonwear time as continuous runs of zero activity ≥90 minutes. Days with <8 hours valid wear data and participants with <1 valid day of recording were excluded from the present analyses. We calculated total physical activity as mean vector magnitude over valid days (counts per minute, cpm). All statistical analyses were performed using Stata v 12 (StataCorp, College Station, Texas). Anthropometric WHZ and height-for-age z score (HAZ) were calculated using the package “zscore06” in Stata. Variables were tested for normality by histograms and Shapiro-Wilk tests. Means ± SD were calculated for normally distributed variables and median (IQR) for non-normally distributed variables. To determine associations between activity and covariates, we first built unadjusted models comparing volume of physical activity between groups of morbidity, biochemistry, and anthropometry. Second, we adjusted for age and sex (model 1), and finally for all covariates including age, sex, paternal and maternal profession, season of measurement, breastfeeding, number of children under 5 years of age in household, carrying status, and movement ability (model 2). The Figure is based on random effect mixed models with age and sex as fixed effects and child as a random effect.

covariates including age, sex, paternal and maternal profession, season of measurement, breastfeeding, number of children under 5 years of age in household, carrying status, and movement ability (model 2). The Figure is based on random effect mixed models with age and sex as fixed effects and child as a random effect. Results A total of 1609 eligible children, predominantly Mossi, were enrolled in the study from September 2013 to August 2014, after consent from caretakers. Of the 1609 enrolled, 29% of children were enrolled based on MUAC only, 50% based on WHZ and MUAC, and 21% based on WHZ only, as previously reported.20 Among these, 1544 (96%) children had baseline physical activity data and were included in the analysis. The median (IQR) age was 11.3 (8.2; 16) months. More than one-half the children were girls, and almost all children were breastfed (Table I). The majority of children were from families with fuel for cooking based on “coal/wood/straw,” and from families who were owners of their own house. The mean (±SD) MUAC, WHZ, and HAZ were 123 (±4) mm, -2.2 (±0.5), and -1.7 (±1.1), respectively. As previously reported,21 comorbidities were common (Table II). The 65 (4%) children who were excluded from analyses did not differ from those included with respect to age, proportion of girls, proportion of breastfeeding or prevalence of fever, positive malaria test, diarrhea, cough, or raised levels of CRP or AGP (data not shown).

reported,21 comorbidities were common (Table II). The 65 (4%) children who were excluded from analyses did not differ from those included with respect to age, proportion of girls, proportion of breastfeeding or prevalence of fever, positive malaria test, diarrhea, cough, or raised levels of CRP or AGP (data not shown). Level and Associations of Physical Activity Of the 1544 children with physical activity data, 1498 (97%) completed 6 consecutive days of recording with a daily median wear time of 24 hours. At the first day of enrollment (7 hours excluded), the 25th, 10th, 5th, and 1st percentiles of wear time were 16.7, 11.4, 11.4, and 10.6 hours, respectively. The mean (±SD) total physical activity was 707 (±180) cpm, with age being inversely associated (Table I). Compared with children below 12 months of age, those aged 12-17 months and 18-23 months had 34 (95% CI, 13; 54) and 121 (95% CI, 97; 145) cpm lower activity, respectively. Judging from the diurnal pattern of activity, waking hours began on average between 6 a.m. and 7 a.m., from which time activity increased up to 9 a.m., then declined to reach a local nadir at around 2 p.m. and increased again until reaching a peak at 7 p.m. and decreased thereafter (Figure). The highest accumulation of activity occurred between 6 p.m. and 7 p.m. During daytime hours, younger children were more active than older children. In unadjusted models, children who were not able to crawl/walk had 51 (95% CI, 29; 72) cpm lower activity and 38 (95% CI, 14; 62) cpm higher activity compared with those classified as able to crawl or walk, respectively. There was no difference between boys and girls but children of farming parents had higher activity levels. Ethnic group and socioeconomic status based on fuel for cooking and house ownership were not associated with activity (P > .20, data not shown). Breastfed children were 239 (95% CI, 202; 276) cpm more active than those not breastfed. Neither anthropometric indicators nor admission criteria were associated with physical activity (P > .20, data not shown).

s based on fuel for cooking and house ownership were not associated with activity (P > .20, data not shown). Breastfed children were 239 (95% CI, 202; 276) cpm more active than those not breastfed. Neither anthropometric indicators nor admission criteria were associated with physical activity (P > .20, data not shown). Multivariable analysis, including all variables from Table I as covariates, did not change the difference between age groups; children younger than <12 months of age had higher adjusted activity than children aged 12-17 months (β 51, 95% CI, 26; 75) and 18-23 months (β 106, 95% CI, 71; 141). The adjustment only marginally affected the role of breastfeeding (β 199, 95% CI, 160; 239) but increased the impact of ability to crawl (β 71, 95% CI, 48; 93) and walk (β 71, 95% CI, 37; 106). Also father's profession and season for measurement were unaffected by adjustment. On the contrary, the impact on activity by mother's profession or carrying status seemed to be confounded and no longer significant after adjustment. In unadjusted analyses, most measures of clinically assessed morbidity (except diarrhea), anemia, and inflammation (assessed using acute phase proteins) were negative correlates of physical activity (Table II). In the adjusted models, signs of respiratory tract infection (upper and lower respiratory tract infections, cough) were no longer associated with activity. However, fever, malaria, anemia, and inflammation remained strongly associated with lower volume of activity, with estimates only marginally changed compared with unadjusted models.

odels, signs of respiratory tract infection (upper and lower respiratory tract infections, cough) were no longer associated with activity. However, fever, malaria, anemia, and inflammation remained strongly associated with lower volume of activity, with estimates only marginally changed compared with unadjusted models. Finally, none of the anthropometric measures, WHZ, HAZ, and MUAC, were associated with physical activity; HAZ (β -5, 95% CI, -13; 3), WHZ (β 9, 95% CI, -9; 28), and MUAC (β 1, 95% CI, -2; 3). None of the selection criteria was associated with physical activity; WHZ only vs WHZ and MUAC (β 4, 95% CI, -19; 26) or MUAC only vs WHZ and MUAC (β 6, 95% CI, -20; 33). Discussion Among this large group of young (6-23 months of age) children from Burkina Faso with MAM, we found activity to be correlated with measures of sociodemographic status and morbidity.

Finally, none of the anthropometric measures, WHZ, HAZ, and MUAC, were associated with physical activity; HAZ (β -5, 95% CI, -13; 3), WHZ (β 9, 95% CI, -9; 28), and MUAC (β 1, 95% CI, -2; 3). None of the selection criteria was associated with physical activity; WHZ only vs WHZ and MUAC (β 4, 95% CI, -19; 26) or MUAC only vs WHZ and MUAC (β 6, 95% CI, -20; 33). Discussion Among this large group of young (6-23 months of age) children from Burkina Faso with MAM, we found activity to be correlated with measures of sociodemographic status and morbidity. Few studies from low-income countries are available using similar equipment, but most of these used a different study design in that they did not collect data for the full 24 hours of the day, and there were also differences with respect to accelerometer data reduction methods. This makes direct comparison somewhat difficult but do allow relative comparisons of the within-population associations with covariates. A single-center study from Ethiopia in a small group of children with SAM used an identical approach to the one used in the present study.11 Compared with the Ethiopian study, we found a 5-fold higher level of physical activity in children with MAM from Burkina Faso (707 vs 141.5 cpm), suggesting that the degree of malnutrition is a likely determinant of movement in this age group. We did not find any difference in activity between boys and girls among children with MAM in our study, possibly reflecting that activity of young healthy children may not differ by sex. The lack of difference in physical activity between boys and girls is consistent with studies from Belgium among 20-month-old children,6 from Australia among 19-month-old children,22 from Sweden among 2-year-old children,8 and from The Netherlands.9 With respect to the association with motor development milestones, ability to crawl or walk were associated with higher activity levels compared with children who are less developed; remarkably, this effect was observed independently of age and how much the child was being carried.

children,8 and from The Netherlands.9 With respect to the association with motor development milestones, ability to crawl or walk were associated with higher activity levels compared with children who are less developed; remarkably, this effect was observed independently of age and how much the child was being carried. Diurnal patterns in activity showed peaks of activity in the morning and afternoon. This could represent times where children were more engaged in unstructured play. The decrease of activity observed during midday is likely due to feeding and subsequent napping, although we have no observational data to confirm this. These diurnal patterns are, however, consistent with studies among 36-month-old children from New Zealand23 and Australia.24 Although poor nutritional status is considered to have a negative effect on activity,25, 26 we did not see any association between the anthropometric measures and physical activity, possibly because all children included in this cohort had anthropometric measures within the narrow MAM range. Breastfed children, who were younger, seemed to be more active. Although this could have been influenced by the children being carried, both the age and the breastfeeding effects remained significant after adjustment for how much the child was carried, suggesting that breastfeeding may play an important role in the nutritional support for activity of children with MAM.

more active. Although this could have been influenced by the children being carried, both the age and the breastfeeding effects remained significant after adjustment for how much the child was carried, suggesting that breastfeeding may play an important role in the nutritional support for activity of children with MAM. The higher activity seen among children from farming families may have been because they spent more time in the field either playing or in field activities. They could also have been carried while the mother works in field, which could have influenced the registered movement (passive activity); indeed, this effect was no longer significant when other covariates, including carrying status, were being considered. Children enrolled during the rainy season were more active, which was also reported in a study from Zanzibar.25 Because farming activity is linked to the rainy season, this could account for the greater physical activity among children measured during this season. We found a negative association between infection at enrollment and physical activity. Our results were consistent with the study in Zanzibar, where malaria was found to be negatively associated with children's activity, likely because of inflammation, lethargy, and poor appetite.25 Children compensate for lack of dietary energy by decreasing energy expenditure through reduced physical activity.27 Infection and inflammation can lead to a reduction in body mass, which may reduce capacity to perform work or movement.28

ildren's activity, likely because of inflammation, lethargy, and poor appetite.25 Children compensate for lack of dietary energy by decreasing energy expenditure through reduced physical activity.27 Infection and inflammation can lead to a reduction in body mass, which may reduce capacity to perform work or movement.28 Higher hemoglobin was associated with greater physical activity, as has been seen also for children in Mexico and Indonesia.29, 30 Anemia may be related to iron deficiency or to inflammation. Irrespective of the underlying cause, anemia leads to lower oxygen-carrying capacity or reduced cellular oxidative capacity resulting in low energy production associated with low activity levels. It is notable that anemia remained significantly associated with activity in the multivariable analysis, which included infection indicators, suggesting these other mechanisms also may be important. Anemia may reduce children's endurance as has been found in adolescents and school children.31

with low activity levels. It is notable that anemia remained significantly associated with activity in the multivariable analysis, which included infection indicators, suggesting these other mechanisms also may be important. Anemia may reduce children's endurance as has been found in adolescents and school children.31 The strengths of this study include its large sample size, the use of an objective measure of activity with high time resolution, and high compliance covering all 24 hours of the day. Also, this is the first study to investigate physical activity and correlates among young children with MAM using accelerometers. The limitations of the study include a lack of age-matched control data from well-nourished young children from Burkina Faso. Also, most of the children were breastfed and may to some extent have been carried by caregivers and the lack of synchronous activity registration of the caregiver and/or logs did not allow detailed distinguishing passive movement of the child caused by carrying from actual physical activity of the child. Another potential limitation is the lack of a sleep log that would have enabled better comparison with other studies that include only daytime activity. Physical activity declines with age and is associated with infection and inflammation status in children with MAM. However, because younger children are more likely to be carried, future studies should use both accelerometers and activity logs to improve assessment and aid the distinction between passive and active movement.

al activity declines with age and is associated with infection and inflammation status in children with MAM. However, because younger children are more likely to be carried, future studies should use both accelerometers and activity logs to improve assessment and aid the distinction between passive and active movement. We are grateful to the study participants, their families, and the staff of the Alliance for International Medical Action for their valuable contribution to this study. We thank the Ministry of Health in Burkina Faso, the health and village authorities in Province du Passoré, and the staff at the health centers and for their support to this study. Funded by the Danish International Development Assistance (DANIDA; 09-097 LIFE [to C.W.]), the UK Medical Research Council (MC_UU_12015/3 [to S.B.]), Médecins Sans Frontières (Denmark and Norway), World Food Program (received support from the US Agency for International Development's Office of Food for Peace), Alliance for International Medical Action, the European Union, Action Contre la Faim, and Arvid Nilsson Foundation. The authors declare no conflicts of interest. Figure Diurnal patterns in physical activity association with age group. Data represent age- and sex-adjusted means of total physical activity based on random effect mixed model. cpm, count per minute. FigureTable I Background characteristics and physical activity among 1544 children with MAM

Funded by the Danish International Development Assistance (DANIDA; 09-097 LIFE [to C.W.]), the UK Medical Research Council (MC_UU_12015/3 [to S.B.]), Médecins Sans Frontières (Denmark and Norway), World Food Program (received support from the US Agency for International Development's Office of Food for Peace), Alliance for International Medical Action, the European Union, Action Contre la Faim, and Arvid Nilsson Foundation. The authors declare no conflicts of interest. Figure Diurnal patterns in physical activity association with age group. Data represent age- and sex-adjusted means of total physical activity based on random effect mixed model. cpm, count per minute. FigureTable I Background characteristics and physical activity among 1544 children with MAM Table I Total physical activity (cpm) n (%) Mean physical activity (±SD) Unadjusted β (95% CI)* P value Adjusted β (95% CI)† P value Age (mo), median (IQR) 11.3(8.2;16) 6-11 837(54.2) 738(173) Ref ref 12-17 441(28.6) 705(176) −34(−54;−13) .001 −51(−75;−26) <.001 18-23 266(17.2) 617(179) −121(−145;−97) <.001 −106(−141;−71) <.001 Sex Girls 844(54.7) 704(172) Ref ref Boys 700(45.3) 711(191) 7(−11;25) .451 10(−7;26) .273 Mother's profession Farmer 1457(94.4) 710(181) Ref ref Others 87(5.6) 667(165) −43(−82;−4) .032 5(−39;50) .818 Father's profession Farmer 1411(91.4) 711(181) Ref ref Others 133(8.6) 667(160) −45(−77;−13) .006 −41(−77;−4) .031 Number of children <5-y-old in household 1-2 783(50.8) 705(177) Ref ref 3-4 553(35.8) 716(183) 11(−9;31) .295 7(−12;25) .486 ≥5 207(13.4) 692(182) −13(−41;14) .348 −19(−45;7) .151 Season of measurement Dry season 1015(65.7) 687(173) Ref ref Rainy season 529(34.3) 745(188) 58(39;77) <.001 52(34;70) <.001 Breastfeeding Breastfed 1456(94.4) 720(171) Ref ref Not breastfed 86(5.6) 481(180) −239(−276;−202) <.001 −199(−239;−160) <.001 Carrying status Never carried 16(1.1) 575(174) ref ref 1-2 times per d 407(27.1) 682(181) 107(17;196) .020 34(−50;117) .427 3-6 times per d 980(65.4) 717(778) 141(53;230) .002 50(−33;133) .234 More than one-half of the d 91(6.1) 739(187) 164(68;259) .001 83(−6;172) .068 All the d 5(0.3) 644(185) 68(−112;249) .456 12(−155;178) .891 Movement ability Not able to crawl/walk 403(26) 694(168) ref ref Crawling 726(47) 744(181) 51(29;72) <.001 71(48;93) <.001 Walking 415(27) 656(175) −38(−62;−14) <.001 71(37;106) <.001 β, regression coefficient (difference in physical activity volume between groups); ref, reference group.

112;249) .456 12(−155;178) .891 Movement ability Not able to crawl/walk 403(26) 694(168) ref ref Crawling 726(47) 744(181) 51(29;72) <.001 71(48;93) <.001 Walking 415(27) 656(175) −38(−62;−14) <.001 71(37;106) <.001 β, regression coefficient (difference in physical activity volume between groups); ref, reference group. Sample sizes: Breastfeeding (n = 1542), number of children in household (n = 1543). * Unadjusted linear regression; unadjusted comparisons of volume of physical activity. † Linear regression-based estimates of total PA mutually adjusted for age, sex, mothers' profession, father's profession, number of children <5-y-old in household, season of measurement, breastfeeding, carrying status, and movement ability. Table II Associations of volume of physical activity with morbidity, biochemistry among 1544 children with MAM

† Linear regression-based estimates of total PA mutually adjusted for age, sex, mothers' profession, father's profession, number of children <5-y-old in household, season of measurement, breastfeeding, carrying status, and movement ability. Table II Associations of volume of physical activity with morbidity, biochemistry among 1544 children with MAM Table II Volume of physical activity (cpm) Adjusted linear models Model 1 Model 2 N (%) Mean (SD) Δ physical activity (95%CI) P value β (95% CI) P value β (95% CI) P value Clinical examination, presence of: Upper respiratory tract infection 228(14.8) 688(184) −23(−48;2) .074 −17(−42;8) .173 −11(−35;15) .354 Lower respiratory tract infection 361(23.4) 690(189) −22(−44;−1) .039 −21(−41;−0.1) .049 −15(−35;5) .148 Fever 273(17.7) 678(196) −36(−59;−12) .003 −40(−63;−17) .001 −49(−70;−27) <.001 Cough 413(26.8) 685(667) −30(−50;−9) .004 −27(−46;−7) .008 −16(−36;3) .096 Diarrhea 90(5.8) 704(190) −3(−42;35) .868 −7(−44;31) .729 −2(−38;34) .902 Malaria 615(40) 677(185) −51(−69;−33) <.001 −46(−63;−27) <.001 −44(−61;−27) <.001 Biochemical data* CRP ≥10 (mg/L); (ref: < 10 mg/L) 361(24) 669(189) −50(−71;−29) <.001 −47(−68;−26) <.001 −51(−71;−31) <.001 AGP ≥1 (g/L); (ref: < 1 g/L) 986(65.7) 679(181) −79(−98;−61) <.001 −69(−87;−50) <.001 −60(−78;−42) <.001 Hb <11 (g/dL); (ref: Hb ≥11 g/dL) 1087(70.4) 693(184) −49(−68;−29) <.001 −44(63;25) <.001 −37(−56;−19) <.001 Δ physical activity, difference in physical activity between group and reference; Hb, hemoglobin.

1) <.001 AGP ≥1 (g/L); (ref: < 1 g/L) 986(65.7) 679(181) −79(−98;−61) <.001 −69(−87;−50) <.001 −60(−78;−42) <.001 Hb <11 (g/dL); (ref: Hb ≥11 g/dL) 1087(70.4) 693(184) −49(−68;−29) <.001 −44(63;25) <.001 −37(−56;−19) <.001 Δ physical activity, difference in physical activity between group and reference; Hb, hemoglobin. Data are based on linear regression models: Δ physical activity (95% CI) is difference in volume of physical activity compared with reference group, model 1 (adjusted for age and sex), model 2 (adjusted for age, sex, mother's and father's profession, season of measurement, breastfeeding, number of children <5-y-old in household, carrying status, and movement ability). Sample sizes: Upper respiratory tract infection (n = 1542), lower respiratory tract infection (n = 1543), cough (n = 1541), malaria (n = 1536), and CRP and AGP (n = 1502). * CRP ≥10 mg/L and AGP ≥1 g/L defined as abnormal indicating systemic inflammation. Anemia defined as hemoglobin <11 g/dL.

The first 1000 days of life, starting from conception until around the child's second birthday, increasingly are recognized as essential for child growth, with inadequate growth often indicating serious and potentially irreversible consequences.1, 2, 3, 4, 5 Childhood undernutrition is estimated to contribute to 45% of all the deaths of children younger than 5 years globally6; however, early anthropometric deficits also are associated with long-term consequences for health and educational attainment, extending into adulthood and even into the next generation.3, 4, 7, 8, 9 Thus, the first 1000 days have been suggested to be critical for the prevention of malnutrition.

than 5 years globally6; however, early anthropometric deficits also are associated with long-term consequences for health and educational attainment, extending into adulthood and even into the next generation.3, 4, 7, 8, 9 Thus, the first 1000 days have been suggested to be critical for the prevention of malnutrition. Any measure of inadequate attained growth used for identifying children at risk of adverse events has the inherent limitation that the child already is stunted or wasted to a varying degree, impeding possibilities for prevention and impacts of nutritional interventions. Longitudinal growth measures such as weight velocity or weight gain have a theoretical advantage as they present a picture of the current growth trend, whereas attained growth is a cumulative measure of an altered growth rate that leads to a recognizable malnourished state.10, 11 Few studies have estimated the extent to which measures of longitudinal growth early in life can predict future nutritional status. Although weight at 12 months predicted stunting at 36 months equally well as weight gain from 3 to 6 months in children living the Republic of Congo,12 the detection at an earlier age with weight gain could be advantageous. Iannotti et al13 found that weight gain during the first month of life predicted attained weight and length at 1 year of age, but they did not compare it with attained growth measures. In a study in Peru, no advantage of weight gain assessment to predict underweight at 24 months of age was found compared with attained weight assessment.11 Length gain was not found predictive of wasting or stunting at later ages in Peru and Guatemala.11, 14 These studies all had different approaches to define weight and length gain.

eru, no advantage of weight gain assessment to predict underweight at 24 months of age was found compared with attained weight assessment.11 Length gain was not found predictive of wasting or stunting at later ages in Peru and Guatemala.11, 14 These studies all had different approaches to define weight and length gain. The World Health Organization (WHO) published growth velocity standards in 2009,15 offering the opportunity to score weight and length gain according to age and sex. Two studies have used the WHO growth velocity standards to assess the relationship with future nutritional status, but one focused on the association with obesity and did not compare the predictive ability of weight velocity with other growth measures,16 and the other studied children with cystic fibrosis in the US.17 Studying growth velocities could help to identify critical time windows for prevention or early interventions of undernutrition.7, 8, 18 We therefore aimed to estimate the abilities of weight and length velocity z scores in infancy (according to the WHO Child Growth Standards) to predict stunting, wasting, and underweight at the age of 2 years and compare them with those of the attained growth measures weight-for-age z score (WAZ), length-for-age z score (LAZ), and weight-for-length z score (WLZ).

of weight and length velocity z scores in infancy (according to the WHO Child Growth Standards) to predict stunting, wasting, and underweight at the age of 2 years and compare them with those of the attained growth measures weight-for-age z score (WAZ), length-for-age z score (LAZ), and weight-for-length z score (WLZ). Methods The Etiology, Risk Factors, and Interactions of Enteric Infections and Malnutrition and the Consequences for Child Health and Development study (MAL-ED) was conducted in 8 countries (Bangladesh, Brazil, India, Nepal, Pakistan, Peru, South Africa, and Tanzania). For this analysis, data from the Nepal site were used. The study in Nepal was carried out in the Bhaktapur municipality, located 15 km east of the capital Kathmandu and at about 1400 m above sea level. Bhaktapur had a population of about 78 000 people in 2010.19 Hinduism and Buddhism are the predominant religions practiced in this municipality, and community members are primarily distinguished by the traditional caste system. Tourism and agriculture are the main sources of livelihoods. The climate is humid subtropical, with a hot and wet monsoon season from May to September and a cool and dry season from October to March. A pilot study in 2010 of 100 households with children 24-36 months of age showed that although socioeconomic indicators compared favorably with national averages, 40% of children were stunted.19

s humid subtropical, with a hot and wet monsoon season from May to September and a cool and dry season from October to March. A pilot study in 2010 of 100 households with children 24-36 months of age showed that although socioeconomic indicators compared favorably with national averages, 40% of children were stunted.19 The MAL-ED study is a prospective cohort study. During enrollment from June 2010 to February 2012, 668 deliveries were recorded, with 97% occurring at the hospital. Deliveries outside the hospital were registered by fieldworkers surveying the households. Households with recent deliveries were selected randomly on a weekly basis. The number for children selected each week was based on a prestudy census, which informed the expected birth rate, and the target sample size defined for all 8 sites of the MAL-ED study (ie, to arrive at >200 children enrolled during a period of 2 years).20

th recent deliveries were selected randomly on a weekly basis. The number for children selected each week was based on a prestudy census, which informed the expected birth rate, and the target sample size defined for all 8 sites of the MAL-ED study (ie, to arrive at >200 children enrolled during a period of 2 years).20 With this weekly number, 275 children were selected, and all caretakers of were informed about the MAL-ED study. If informed consent was given, households were screened for enrollment. Participants were excluded if the family had plans to move out of the catchment area for >30 consecutive days during the first 6 months of follow-up; the mother was <16 years of age; the mother had another child already enrolled in the MAL-ED study; the child was not a singleton (ie, twins, triplets); the child's guardian failed to provide signed informed consent; weight at birth or enrollment was <1500 g; or the infant had any of the following indications of serious disease: hospitalization for something other than a typical healthy birth; severe or chronic condition diagnosed by a medical doctor (eg, neonatal disorder; renal, liver, lung, and/or heart disease; congenital conditions); or enteropathies diagnosed by a medical doctor. In total, 240 children were enrolled. Ethical approval for the study was obtained from the Nepal Health Research Council and the Walter Reed Institute of Research (Silver Spring, Maryland). All caretakers of the participating children provided informed consent. This subanalysis was approved by the Central Board of the MAL-ED study.

children were enrolled. Ethical approval for the study was obtained from the Nepal Health Research Council and the Walter Reed Institute of Research (Silver Spring, Maryland). All caretakers of the participating children provided informed consent. This subanalysis was approved by the Central Board of the MAL-ED study. At enrollment (within 17 days after delivery), well-trained fieldworkers interviewed caretakers on the child's date of birth, birth weight (available for 97% of the children), breastfeeding status, and sociodemographic characteristics of the household and took anthropometric measurements using standardized techniques (length, weight, and head circumference). Thereafter, monthly anthropometric measurements were taken until the age of 2 years, resulting in 24 anthropometric measurements for each child. Length was measured with a standard length board (ShorrBoard; Weigh and Measure, LLC, Olney, Maryland), weight with an infant scale (seca, Chino, California), and head circumference with a nonstretch synthetic tape (seca). Each month a supervisor duplicated 10% of the measurements within 24 hours. The interobserver technical error of measurement for these repeated measurements was 0.343 for height and 0.070 for weight.

, weight with an infant scale (seca, Chino, California), and head circumference with a nonstretch synthetic tape (seca). Each month a supervisor duplicated 10% of the measurements within 24 hours. The interobserver technical error of measurement for these repeated measurements was 0.343 for height and 0.070 for weight. Data Management and Statistical Analyses If concern or suspicion was articulated during measurements, raw values were plotted on growth curves. In case of implausible discrepancies to the previous values, measurements were redone immediately. All data were double-entered into a local database, and discrepancies and completeness were checked by the site data entry supervisor. If necessary, remeasurements were taken within the shortest time possible, generally within 2 days. Data were sent to and stored at the Data Coordinating Center at Fogarty International Center (Bethesda, Maryland), which did an external quality control and marked values that exceeded plausible ranges within subsequent measurements (increments >1.5 kg for weight, >3.5 cm for length, and >2 cm for head circumference) for review by the study site. The Data Coordinating Center made Web-based issue logs available to the local teams to enable prompt corrections. In addition, monthly reports provided the sites with feedback on data quality.

easurements (increments >1.5 kg for weight, >3.5 cm for length, and >2 cm for head circumference) for review by the study site. The Data Coordinating Center made Web-based issue logs available to the local teams to enable prompt corrections. In addition, monthly reports provided the sites with feedback on data quality. Data were analyzed with Stata (version 13; StataCorp LP, College Station, Texas). We calculated WAZ, WLZ, LAZ, weight velocity z score (WVZ), and length velocity z score (LVZ) according to the WHO Child Growth Standards.15, 21 We defined wasting, stunting, and underweight as z score ≤−2 for WLZ, LAZ, and WAZ, respectively.

were analyzed with Stata (version 13; StataCorp LP, College Station, Texas). We calculated WAZ, WLZ, LAZ, weight velocity z score (WVZ), and length velocity z score (LVZ) according to the WHO Child Growth Standards.15, 21 We defined wasting, stunting, and underweight as z score ≤−2 for WLZ, LAZ, and WAZ, respectively. For the description of the sample, we report percentages, means with SDs or medians with IQRs as appropriate. For each anthropometric index, we built a separate simple logistic or linear regression model, depending on the format of the outcome, ie, WAZ, WLZ, or LAZ at 2 years of age as continuous variable (linear) or as dichotomous variable with a cut-off at −2 z scores (logistic). The predictor variables, all tested in the regression models one at a time, were the individual growth velocity z scores, for 3- and 6-month increments at the ages 0-3, 0-6, 3-6, 6-9, 6-12, and 9-12 months as well as measures of attained growth at the ages 0, 3, 6, and 12 months. Because weight at birth was lacking for 7 children (3%), we imputed values for birth weight for those by regressing birth weight from the earliest weight measurements. Length was not measured at birth and therefore length within 17 days was used as proxy for birth-length for all 240 children. For all other target ages, we allowed for a deviation of ±3 days, eg, between 2.9 and 3.1 months at the 3-month visit.

for those by regressing birth weight from the earliest weight measurements. Length was not measured at birth and therefore length within 17 days was used as proxy for birth-length for all 240 children. For all other target ages, we allowed for a deviation of ±3 days, eg, between 2.9 and 3.1 months at the 3-month visit. For linear regression models, the R-square is reported in addition to the regression coefficients. For logistic regression models, receiver operating characteristic curves depict the balance between sensitivity and specificity at different threshold levels. ORs and areas under the receiver operating characteristic curves (AUROC) are reported. Results In total, 240 children were enrolled into the study, of which 130 (54%) were male and 233 (97%) delivered at a health facility. Characteristics of the study sample are summarized in Table I. The majority of the mothers (90%) initiated breastfeeding within the first 24 hours after childbirth. Introduction of solid foods was on average at 3 months, although supplementary liquids were given earlier. On an average, exclusive breastfeeding lasted 1 month (IQR 0.6-3.2 months) and total breastfeeding duration 24 months (IQR 23-26 months). A toilet with a flush to a piped sewer system was available in 94% of the households, although 46% of those shared facilities with up to 10 other households. The median monthly household income was approximately 12 000 Nepali rupees (IQR 8000-20 000), corresponding to about 144 US$ (IQR 95-240).

ths (IQR 23-26 months). A toilet with a flush to a piped sewer system was available in 94% of the households, although 46% of those shared facilities with up to 10 other households. The median monthly household income was approximately 12 000 Nepali rupees (IQR 8000-20 000), corresponding to about 144 US$ (IQR 95-240). At the age of 2 years, 4% of the children were classified as wasted (WLZ ≤ −2), 13% as underweight (WAZ ≤ −2), and 21% as stunted (LAZ ≤ −2). Figure 1 displays the proportion of children who were wasted, underweight, and stunted according to age and correspondingly for low weight and length velocity z scores (≤−2) in Figure 2 (available at www.jpeds.com). Mean weight and length velocity z scores for the whole study sample were above the standard mean in the first 3 months but declined with age until about 5 and 13 months, respectively, and improved marginally thereafter. Indicators for mean attained growth (WAZ, WLZ, and LAZ) were low already at birth, improved slightly until about 5 months of age, and deteriorated continuously after that age.

ove the standard mean in the first 3 months but declined with age until about 5 and 13 months, respectively, and improved marginally thereafter. Indicators for mean attained growth (WAZ, WLZ, and LAZ) were low already at birth, improved slightly until about 5 months of age, and deteriorated continuously after that age. When we compared children who were underweight or stunted at 2 years of age with those who were not, it showed that differences in mean z scores for WAZ and LAZ were apparent already at birth and remained throughout the study period up to 2 years of age (Figure 3, B and D). For those underweight at 2 years of age, mean weight velocity z scores werelower for the periods starting during the first 6 months of life (Figure 3, A). For those stunted at 2 years of age, length velocity z scores were lower throughout the first 2 years of life. However, there were some periods where there was a substantial overlap of the 95% CIs of the growth velocity estimates (Figure 2, C).

r for the periods starting during the first 6 months of life (Figure 3, A). For those stunted at 2 years of age, length velocity z scores were lower throughout the first 2 years of life. However, there were some periods where there was a substantial overlap of the 95% CIs of the growth velocity estimates (Figure 2, C). Linear and logistic regression models showed that most periods of the different anthropometric indicators were significant predictors for growth at the age of 2 years. A general trend could be observed, ie, indicators of attained growth during the first year predicted attained growth at 2 years of age better than velocity z scores, Table II, Table III, Table IV (Tables III and IV available at www.jpeds.com). LAZ at 12 months could explain 75% of the variation in LAZ at 24 months, whereas LVZ from 6 to 12 months only explained 24%. For WLZ at 2 years, the R2 of WLZ at 6 months was 0.50 and for WVZ between 0 and 6 months 0.28. Also, more variation of WAZ at 2 years was explained by an indicator of attained growth (WAZ at 12 months, R2 = 0.66), than by a velocity z score (WVZ 0-6 months, R2 = 0.44). The value for AUROC ranged between 91 and 95 for WAZ, LAZ, and WLZ at 12 months and was somewhat lower for weight and length velocity at different time periods (70-84). No difference between girls and boys could be observed. The trend that the older the children, the better the ability to predict nutritional status at 2 years of age, as seen in indicators of attained growth, did not appear in indicators for growth velocity.

weight and length velocity at different time periods (70-84). No difference between girls and boys could be observed. The trend that the older the children, the better the ability to predict nutritional status at 2 years of age, as seen in indicators of attained growth, did not appear in indicators for growth velocity. Discussion In this study, indicators of attained growth during the first year of life predicted stunting, wasting, and underweight at age 2 years better than velocity z scores. WAZ, LAZ, and WLZ at age 12 months had excellent AUROC (91-95) to predict the value of the same indicator at age 24 months. Maximum AUROC values for weight and length velocity in different growth periods were somewhat lower (70-84). In agreement with our study results, Simondon et al12 found weight measured at one time point (12 months) to be most predictive of stunting at age 1-5 years. Nevertheless, in their study, weight velocity from 3 to 6 months had equally high sensitivity and specificity values. They used predicted quarterly weight gains as velocity measure, which does not take measurement error and transient weight losses into account. Temporary weight loss, eg, weight loss caused by disease with catch-up growth during recovery, is well described in literature. Advantages of predictions by modeled growth velocity become clear, although the disadvantage for practical settings, where only raw measurement values are used, needs to be emphasized.

ccount. Temporary weight loss, eg, weight loss caused by disease with catch-up growth during recovery, is well described in literature. Advantages of predictions by modeled growth velocity become clear, although the disadvantage for practical settings, where only raw measurement values are used, needs to be emphasized. In children with cystic fibrosis in the US, attained growth measures (WAZ and LAZ) at age 4 months predicted low WAZ and LAZ (<10th percentile) at 24 months better than WVZ and LVZ at different age periods17 when the WHO Child Growth Standards were used. The authors argue that one reason why attained growth indictors performed better might have been that growth velocity z scores were more sensitive to the therapy of cystic fibrosis. Although velocity z scores increased after the introduction of therapeutic measures, they were still insufficient for most children. They would have needed to be positive for a sufficient amount of time to counterbalance completely the growth deficit seen in attained growth measures. Similar reasons could have interfered with the predicting ability of weight and length velocity z scores in our study, because children found sick were referred to health services and treated. This explanation is strengthened by the findings of a study with children from all 8 sites of the MAL-ED study, in which rates of growth defined by a linear piecewise model were found to be greater after periods with high enteropathogen detections. Still, values for indicators of attained growth were decreasing with age.22

explanation is strengthened by the findings of a study with children from all 8 sites of the MAL-ED study, in which rates of growth defined by a linear piecewise model were found to be greater after periods with high enteropathogen detections. Still, values for indicators of attained growth were decreasing with age.22 In the study of Ruel et al,14 anthropometric indicators were ranked in the same order for their ability to predict stunting at age 3 years as they were in our study, ie, LAZ performed best followed by WAZ, WVZ, LVZ, and WLZ. Attained growth indicators performed better in children aged 6 months compared with 3 months. For children with cystic fibrosis in the US, early attained weight and length (at 4 months) was more predictive than at later ages (6, 12, 18 months), but in their study sample, more children were classified as undernourished in early ages, because the underlying cause (cystic fibrosis) was treated. In our study, the proportion of malnourished children increased with increasing age. This continuous deterioration in nutritional status typically is seen in low-income countries,2 leading to better predictions of future nutritional status with increasing age as seen in our analysis.

ause (cystic fibrosis) was treated. In our study, the proportion of malnourished children increased with increasing age. This continuous deterioration in nutritional status typically is seen in low-income countries,2 leading to better predictions of future nutritional status with increasing age as seen in our analysis. Our hypothesis, that velocity z scores would perform better to predict future growth, was based on the theoretical idea that low growth rates would accumulate and in the end lead to a detectable low nutritional status. This was described in a study in Guatemala,11 where mean weight-for-age of the study sample was not below standard mean before 5 months, although weight gains already were lower much earlier. Even though growth velocity z scores during several time periods significantly predicted growth at age 2 years in our study, they performed worse than attained growth measures. In earlier analyses, however, we have shown that growth velocities were better than attained growth to predict child death within 3 months,23 supported by other studies where growth velocity was lower in the time period just before death whereas no association was shown with attained indices.24, 25 We also found that velocity z scores could depict changes in growth according to the well-known seasonal cycle of food availability in an area heavily depending on subsistence farming, which was not apparent in attained growth.26 This might point to the important advantage of growth velocities over attained growth measures of being able to capture current risk factors, thus representing the current risk profile and better predict short term health consequences.

avily depending on subsistence farming, which was not apparent in attained growth.26 This might point to the important advantage of growth velocities over attained growth measures of being able to capture current risk factors, thus representing the current risk profile and better predict short term health consequences. Differences in mean z scores for weight-for-age and length-for-age already were apparent at birth, emphasizing the importance of intrauterine life and other prenatal factors for optimal growth development. Differences remained throughout the study period, with persistently lower weight and length velocity z scores in those children that were malnourished at age 2 years further augmenting the difference in attained growth. This gives an indication that infancy is still an important period to avoid or hamper critical growth deficits at later ages.

ughout the study period, with persistently lower weight and length velocity z scores in those children that were malnourished at age 2 years further augmenting the difference in attained growth. This gives an indication that infancy is still an important period to avoid or hamper critical growth deficits at later ages. The study has several strengths, including being community-based, with random selection of the children, and only a few children lost to follow-up (5% at the end of the study), reducing the possibility of selection bias. Most children were measured within accurate 1-month intervals with a thorough validation procedure, allowing for a very strict definition of target ages (±3 days) for this analysis. Compared with this, the study by Heltshe et al17 allowed for ±9 days' deviation; however, length was not measured at birth, and length measurements within the first 17 days were used as proxy for birth length. Therefore, length velocity z scores in the periods 0-3 and 0-6 months are artificially low as the result of less time to grow in these 3- and 6-months periods, which could have influenced the predictive abilities in these age periods. Nevertheless, additional analyses, in which we estimated birth length from regression lines based on the 2 subsequent length values, did not change the results substantially (data not shown). Low goodness of fit of the models of LVZ also in later age periods (assessed by R2 and AUROC values) supports the robustness of our findings.

additional analyses, in which we estimated birth length from regression lines based on the 2 subsequent length values, did not change the results substantially (data not shown). Low goodness of fit of the models of LVZ also in later age periods (assessed by R2 and AUROC values) supports the robustness of our findings. Our results show that measuring growth at one time point in infancy seems to be sufficient to distinguish between those at high risk of becoming malnourished and those at lower risk. For low-income settings with high prevalence of malnutrition, where resources are often scarce, simplicity of growth monitoring is likely to encourage health personnel to actually do it. The same conclusion is given by Piwoz et al,11 where weight gain was the best predictor for weight-for-age at 12 months, but because of the favored simplicity, the authors advised using attained weight for monitoring programs. Malnutrition, however, remains an enormous problem in low-income countries with serious consequences and efforts need to be put into optimizing detection and treatment of it. We would like to point out the value of assessing growth cross-sectionally and longitudinally, both reflecting different aspects of growth. The decision on which of the methods to use needs to be evaluated carefully, taking into account the purpose and the resources available. Despite the possible drawbacks of growth velocities concerning their practicality at present,23, 27 because their greater sensitivity to capture influencing factors,26 their potential to predict short-term consequences,23 and their strength to reflect the dynamics of growth rather than status, we think that they could be a valuable tool for research in the field of malnutrition that merits further study.

esent,23, 27 because their greater sensitivity to capture influencing factors,26 their potential to predict short-term consequences,23 and their strength to reflect the dynamics of growth rather than status, we think that they could be a valuable tool for research in the field of malnutrition that merits further study. Appendix Figure 2 Proportion of 240 children enrolled in the MAL-ED study, Nepal, having a low weight or length velocity z score (<−2) according to age. All velocity z scores use a 3-month increment and are plotted at the beginning of the growth period, eg, a velocity z score plotted at 3 months is the velocity for the period from 3 to 6 months. Figure 2Table III Early attained growth and growth velocity z scores on WLZ or wasting (WLZ < −2) at age 24 months

Appendix Figure 2 Proportion of 240 children enrolled in the MAL-ED study, Nepal, having a low weight or length velocity z score (<−2) according to age. All velocity z scores use a 3-month increment and are plotted at the beginning of the growth period, eg, a velocity z score plotted at 3 months is the velocity for the period from 3 to 6 months. Figure 2Table III Early attained growth and growth velocity z scores on WLZ or wasting (WLZ < −2) at age 24 months Table III n Linear regression Logistic regression Coefficient (95% CI) R2 OR (95% CI) AUROC WVZ, mo 0-3  227 0.40(0.30,0.50) 0.22 0.51(0.28,0.93) 0.70 0-6  221 0.45(0.36,0.55) 0.28 0.50(0.26,0.94) 0.69 3-6  160 0.32(0.21,0.43) 0.17 0.84(0.40,1.76) 0.59 6-9  183 0.19(0.06,0.32) 0.05 0.91(0.41,2.02) 0.57 6-12  181 0.21(0.07,0.35) 0.05 0.96(0.40,2.34) 0.52 9-12  218 0.10(−0.03,0.23) 0.01 0.61(0.32,1.15) 0.60 LVZ,mo 0-3  186 −0.08(−0.19,0.03) 0.01 1.18(0.66,2.13) 0.57 0-6  182 −0.01(−0.13,0.11) 0.00 0.89(0.41,1.94) 0.53 3-6  160 0.07(−0.04,0.19) 0.01 1.15(0.56,2.37) 0.51 6-9  183 0.04(−0.08,0.17) 0.00 1.33(0.63,2.81) 0.59 6-12  181 0.05(−0.08,0.18) 0.00 2.03(0.83,4.95) 0.66 9-12  218 0.03(−0.09,0.14) 0.00 1.44(0.77,2.70) 0.65 WAZ,mo 0  227 0.36(0.23,0.48) 0.12 0.36(0.18,0.73) 0.71 3  188 0.54(0.43,0.65) 0.34 0.38(0.20,0.73) 0.78 6  188 0.64(0.54,0.73) 0.49 0.28(0.12,0.66) 0.77 12  223 0.58(0.49,0.67) 0.42 0.33(0.18,0.62) 0.82 LAZ,mo 0  227 0.17(0.06,0.29) 0.04 0.57(0.31,1.04) 0.65 3  188 0.16(0.03,0.30) 0.03 0.54(0.26,1.12) 0.63 6  188 0.26(0.11,0.41) 0.06 0.61(0.26,1.44) 0.60 12  223 0.18(0.06,0.31) 0.04 0.77(0.38,1.54) 0.53 WLZ,mo 0  224 0.25(0.14,0.35) 0.08 0.68(0.37,1.24) 0.63 3  188 0.46(0.36,0.56) 0.31 0.47(0.25,0.89) 0.74 6  188 0.60(0.52,0.69) 0.50 0.22(0.08,0.58) 0.84 12  223 0.60(0.52,0.68) 0.49 0.26(0.12,0.55) 0.91 Table IV Early attained growth and growth velocity z scores on WAZ or underweight (WAZ < −2) at age 24 months

) 0.53 WLZ,mo 0  224 0.25(0.14,0.35) 0.08 0.68(0.37,1.24) 0.63 3  188 0.46(0.36,0.56) 0.31 0.47(0.25,0.89) 0.74 6  188 0.60(0.52,0.69) 0.50 0.22(0.08,0.58) 0.84 12  223 0.60(0.52,0.68) 0.49 0.26(0.12,0.55) 0.91 Table IV Early attained growth and growth velocity z scores on WAZ or underweight (WAZ < −2) at age 24 months Table IV n Linear regression Logistic regression Coefficient (95% CI) R2 OR (95% CI) AUROC WVZ, mo 0-3  227 0.47 (0.38, 0.56) 0.30 0.37 (0.24, 0.57) 0.78 0-6  221 0.55 (0.47, 0.64) 0.44 0.22 (0.13, 0.38) 0.84 3-6  160 0.38 (0.28, 0.49) 0.25 0.38 (0.22, 0.66) 0.81 6-9  183 0.28 (0.15, 0.40) 0.09 0.52 (0.34, 0.81) 0.73 6-12  181 0.38 (0.24, 0.51) 0.15 0.41 (0.24, 0.68) 0.75 9-12  218 0.18 (0.05, 0.31) 0.03 0.68 (0.45, 1.02) 0.59 LVZ, mo 0-3  186 0.03 (−0.08, 0.14) 0.00 0.81 (0.56, 1.19) 0.57 0-6  182 0.12 (−0.00, 0.23) 0.02 0.52 (0.32, 0.84) 0.68 3-6  160 0.12 (0.01, 0.23) 0.03 0.75 (0.51, 1.11) 0.62 6-9  183 0.18 (0.06, 0.31) 0.05 0.64 (0.41, 0.98) 0.62 6-12  181 0.27 (0.15, 0.40) 0.10 0.63 (0.40, 0.98) 0.63 9-12  218 0.15 (0.03, 0.27) 0.03 0.85 (0.57, 1.26) 0.54 WAZ, mo 0  227 0.40 (0.28, 0.52) 0.16 0.32 (0.20, 0.53) 0.76 3  188 0.59 (0.49, 0.69) 0.42 0.23 (0.13, 0.40) 0.87 6  188 0.72 (0.64, 0.80) 0.63 0.07 (0.03, 0.18) 0.93 12  223 0.72 (0.65, 0.79) 0.66 0.06 (0.02, 0.15) 0.95 LAZ, mo 0  227 0.34 (0.23, 0.44) 0.15 0.39 (0.26, 0.59) 0.77 3  188 0.41 (0.28, 0.53) 0.19 0.20 (0.10, 0.38) 0.84 6  188 0.63 (0.51, 0.76) 0.35 0.06 (0.02, 0.17) 0.91 12  223 0.60 (0.50, 0.70) 0.39 0.15 (0.08, 0.28) 0.86 WLZ, mo 0  224 0.11 (−0.00, 0.22) 0.02 0.68 (0.47, 0.98) 0.59 3  188 0.31 (0.21, 0.42) 0.15 0.65 (0.44, 0.98) 0.63 6  188 0.51 (0.41, 0.61) 0.35 0.37 (0.22, 0.62) 0.76 12  223 0.56 (0.47, 0.64) 0.44 0.21 (0.12, 0.37) 0.86

.63 (0.51, 0.76) 0.35 0.06 (0.02, 0.17) 0.91 12  223 0.60 (0.50, 0.70) 0.39 0.15 (0.08, 0.28) 0.86 WLZ, mo 0  224 0.11 (−0.00, 0.22) 0.02 0.68 (0.47, 0.98) 0.59 3  188 0.31 (0.21, 0.42) 0.15 0.65 (0.44, 0.98) 0.63 6  188 0.51 (0.41, 0.61) 0.35 0.37 (0.22, 0.62) 0.76 12  223 0.56 (0.47, 0.64) 0.44 0.21 (0.12, 0.37) 0.86 We thank the staff, parents, and children of the MAL-ED study Bhaktapur site for their contributions. The Etiology, Risk Factors, and Interactions of Enteric Infections and Malnutrition and the Consequences for Child Health and Development Project (MAL-ED) was supported by the Bill & Melinda Gates Foundation (OPP47075). The authors declare no conflicts of interest. Figure 1 Proportion of 240 children enrolled in the MAL-ED study, Nepal, being stunted, wasted, or underweight according to age. Stunting is defined as LAZ < −2, wasting as WLZ < −2, and underweight as WAZ < −2, according to the WHO Child Growth Standards. Figure 1Figure 3 Mean z scores for different anthropometric indices with 95% CI over time from age 0 to 24 months according to A and B, underweight at 24 months or C and D, stunting at 24 months (yes = black circles; no = gray triangles) of 240 children enrolled in the MAL-ED study, Nepal. Underweight and stunting is defined as WAZ and LAZ < −2, respectively. All velocity z scores use a 3-month increment and are plotted at the beginning of the growth period, eg, a velocity z score plotted at 3 months is the velocity for the period from 3 to 6 months.

) of 240 children enrolled in the MAL-ED study, Nepal. Underweight and stunting is defined as WAZ and LAZ < −2, respectively. All velocity z scores use a 3-month increment and are plotted at the beginning of the growth period, eg, a velocity z score plotted at 3 months is the velocity for the period from 3 to 6 months. Figure 3Table I Selected characteristics of the study sample of 240 children aged 0-24 months living in Bhaktapur municipality, Nepal, enrolled into the MAL-ED study, 2010-2012 Table I n Values Male sex, % 240 54 Education, father Ever gone to school, % 104 95 Median duration of education, y (IQR) 99 9 (6-10) Education, mother Ever gone to school, % 236 94 Median duration of education, y (IQR) 221 10 (6-10) Median household incomes, median (IQR)* 236 12 (8-20) Electricity available, % 236 100 Access to flush toilet, % 236 94 Owning a television, % 236 94 Owning a computer, % 236 25 Owning a refrigerator, % 236 25 WAZ, mean (SD) 0-6 mo 2088 −0.52 (1.00) 7-12 mo 1353 −0.52 (0.99) 13-24 mo 2653 −0.82 (0.93) LAZ, mean (SD) 0-6 mo 1855 −0.57 (0.98) 7-12 mo 1354 −0.77 (0.93) 13-24 mo 2653 −1.20 (0.93) WLZ, mean (SD) 0-6 mo 1851 −0.12 (1.11) 7-12 mo 1353 −0.15 (1.01) 13-24 mo 2653 −0.33 (0.91) * Nepali rupees per month in thousands, corresponding to about 144 US$ (IQR 95-240). Table II Early attained growth and growth velocity z scores on LAZ or stunting (LAZ < −2) at 24 months

Table I n Values Male sex, % 240 54 Education, father Ever gone to school, % 104 95 Median duration of education, y (IQR) 99 9 (6-10) Education, mother Ever gone to school, % 236 94 Median duration of education, y (IQR) 221 10 (6-10) Median household incomes, median (IQR)* 236 12 (8-20) Electricity available, % 236 100 Access to flush toilet, % 236 94 Owning a television, % 236 94 Owning a computer, % 236 25 Owning a refrigerator, % 236 25 WAZ, mean (SD) 0-6 mo 2088 −0.52 (1.00) 7-12 mo 1353 −0.52 (0.99) 13-24 mo 2653 −0.82 (0.93) LAZ, mean (SD) 0-6 mo 1855 −0.57 (0.98) 7-12 mo 1354 −0.77 (0.93) 13-24 mo 2653 −1.20 (0.93) WLZ, mean (SD) 0-6 mo 1851 −0.12 (1.11) 7-12 mo 1353 −0.15 (1.01) 13-24 mo 2653 −0.33 (0.91) * Nepali rupees per month in thousands, corresponding to about 144 US$ (IQR 95-240). Table II Early attained growth and growth velocity z scores on LAZ or stunting (LAZ < −2) at 24 months Table II n Linear regression Logistic regression Coefficient (95% CI) R2 OR (95% CI) AUROC WVZ, mo 0-3  227 0.33(0.23,0.44) 0.14 0.46(0.32,0.65) 0.71 0-6  221 0.42(0.32,0.52) 0.23 0.37(0.25,0.54) 0.75 3-6  160 0.28(0.17,0.39) 0.13 0.58(0.39,0.85) 0.72 6-9  183 0.26(0.13,0.39) 0.08 0.56(0.39,0.82) 0.68 6-12  181 0.41(0.28,0.55) 0.17 0.43(0.27,0.68) 0.73 9-12  218 0.20(0.06,0.33) 0.04 0.84(0.59,1.18) 0.55 LVZ,mo 0-3  186 0.16(0.05,0.27) 0.04 0.74(0.53,1.02) 0.61 0-6  182 0.22(0.10,0.35) 0.07 0.65(0.43,0.91) 0.65 3-6  160 0.13(0.01,0.24) 0.03 0.80(0.58,1.10) 0.59 6-9  183 0.28(0.16,0.40) 0.11 0.47(0.32,0.70) 0.70 6-12  181 0.44(0.33,0.56) 0.24 0.29(0.18,0.47) 0.79 9-12  218 0.24(0.13,0.36) 0.07 0.54(0.37,0.77) 0.67 WAZ,mo 0  227 0.27(0.14,0.41) 0.07 0.47(0.32,0.69) 0.68 3  188 0.39(0.26,0.51) 0.17 0.37(0.24,0.56) 0.74 6  188 0.48(0.37,0.59) 0.27 0.27(0.16,0.45) 0.78 12  223 0.56(0.47,0.66) 0.37 0.25(0.16,0.40) 0.81 LAZ,mo 0  227 0.40(0.29,0.51) 0.20 0.37(0.26,0.54) 0.76 3  188 0.54(0.43,0.65) 0.32 0.22(0.13,0.39) 0.82 6  188 0.82(0.72,0.92) 0.57 0.08(0.04,0.18) 0.90 12  223 0.87(0.80,0.93) 0.75 0.02(0.01,0.07) 0.95 WLZ,mo 0  224 −0.13(−0.24,−0.01) 0.02 1.16(0.84,1.60) 0.57 3  188 −0.02(−0.14,0.10) 0.00 0.92(0.67,1.27) 0.50 6  188 0.13(0.01,0.26) 0.02 0.73(0.51,1.04) 0.58 12  223 0.23(0.11,0.34) 0.07 0.67(0.49,0.92) 0.61

In low- and middle-income countries, rapid weight gain in early childhood has clear short-term benefits, including the reduction of morbidity and mortality owing to infectious diseases.1 In contrast, the evidence on the long-term consequences of rapid weight gain in childhood are not as clearcut. Several studies, mostly from high-income countries, reported that rapid weight gain in childhood is associated with a greater risk of obesity2, 3, 4, 5 and hypertriglyceridemic waist phenotype.6 The associations of weight gain in childhood and metabolic cardiovascular risk factors in low- and middle-income countries seem to be different. In a pooled analysis of data from 5 birth cohorts, weight gain in the first 2 years of life was associated with higher fat-free mass in adulthood, whereas weight gain after the first 2 years of life was associated with higher fat and fat-free mass.7 Norris et al8 used data from the same 5 cohorts and found that impaired fasting glucose and type 2 diabetes were not associated with weight gain in the first 4 years of life, whereas weight gain from 48 months increased the risk. These and other studies9, 10, 11 stressed that the long-term consequences of weight gain in childhood depend on timing, with negative consequences being particularly associated with late weight gain rather than gain in the first 1000 days.

he first 4 years of life, whereas weight gain from 48 months increased the risk. These and other studies9, 10, 11 stressed that the long-term consequences of weight gain in childhood depend on timing, with negative consequences being particularly associated with late weight gain rather than gain in the first 1000 days. With respect to human capital, different indicators of malnutrition (wasting, underweight, and stunting) in early childhood are associated negatively with performance on cognitive tests and years of schooling completed later in childhood and adolescence.12, 13, 14, 15, 16, 17 Furthermore, children who recovered from stunting performed slightly better on cognition tests than those who remained stunted, but less well than those who did not experience stunting.18, 19 Additionally, Hoddinott et al20 reported that the height-for-age z-score at 2 years of age was positively associated with years of schooling, performance on cognitive testing, and socioeconomic status in adulthood.

on tests than those who remained stunted, but less well than those who did not experience stunting.18, 19 Additionally, Hoddinott et al20 reported that the height-for-age z-score at 2 years of age was positively associated with years of schooling, performance on cognitive testing, and socioeconomic status in adulthood. With respect to growth in childhood, school achievement and IQ are associated more strongly with early than with later growth in childhood.21, 22, 23, 24, 25, 26 In contrast, Krishnaveni et al27 did not observe an association between linear growth in childhood and early adolescence and IQ in adolescence. Most of these studies relied on weight gain as the exposure variable, and until recently there was no attempt to disentangle the consequences of weight gain from those of linear growth. It is important to disentangle the effect of weight from that of linear growth because they may have different consequences. In 2013, in a pooled analysis of 5 cohort studies from low- and middle-income countries (including data from the 23-year follow-up visit to the 1982 Pelotas cohort), Adair et al9 reported that linear growth in the first 2 years of life was associated more strongly with years of schooling than weight gain, suggesting therefore that nutritional interventions in childhood should be focused in promoting linear growth instead of weight gain. Moreover, growth monitoring programs should also incorporate length/height measurements.

the first 2 years of life was associated more strongly with years of schooling than weight gain, suggesting therefore that nutritional interventions in childhood should be focused in promoting linear growth instead of weight gain. Moreover, growth monitoring programs should also incorporate length/height measurements. The present study aimed to evaluate how birthweight, nutritional status, linear growth, and relative weight gain in childhood are associated with performance in intelligence tests, years of schooling, and monthly income at 30 years of age.

the first 2 years of life was associated more strongly with years of schooling than weight gain, suggesting therefore that nutritional interventions in childhood should be focused in promoting linear growth instead of weight gain. Moreover, growth monitoring programs should also incorporate length/height measurements. The present study aimed to evaluate how birthweight, nutritional status, linear growth, and relative weight gain in childhood are associated with performance in intelligence tests, years of schooling, and monthly income at 30 years of age. Methods In 1982, the maternity hospitals in Pelotas, a southern Brazilian city, were visited daily and all in-hospital births were identified. The 5914 liveborns whose families lived in the urban area of city were examined and their mothers interviewed. In 1984 and 1986, all households located in urban areas of the city were visited in search of cohort members; 5161 and 4979 individuals were evaluated in 1984 and 1986, respectively. From June 2012 to February 2013, cohort members were invited to visit the research clinic to be interviewed and examined. We interviewed 3701 subjects, which added to the 325 known to have died, for a follow-up rate of 68.1%. Concerning the losses to follow-up, we were unable to locate 1055 participants from the original cohort, 467 were living far from Pelotas, 86 refused to take part in this follow-up, and 280 did not attend the clinic despite repeated invitations. Further details on the study methodology have been published elsewhere.28, 29 The Ethical Review Board of the Faculty of Medicine of the Federal University of Pelotas approved the study, and written informed consent was obtained from all participants.

-up, and 280 did not attend the clinic despite repeated invitations. Further details on the study methodology have been published elsewhere.28, 29 The Ethical Review Board of the Faculty of Medicine of the Federal University of Pelotas approved the study, and written informed consent was obtained from all participants. Birthweight was assessed by the hospital staff using pediatric scales that were calibrated weekly by the research team. Gestational age estimate was based on the mother's recall of the date of her last menstrual period. Preterm birth was defined by a gestational age of <37 weeks. In the 1984 and 1986 visits, children were weighed using calibrated scales and their length and height were assessed with portable stadiometers. Weight and height z-scores, according to age and sex, were estimated using the 2006 World Health Organization growth standards.30 Birthweight for gestational age z-scores were calculated using the Williams reference population.31 Outcomes Performance in intelligence tests was evaluated in the 2012-2013 visit, at a mean of 30.2 years of age. Four psychologists who were unaware of participant intrauterine growth and nutritional status in childhood administered the Wechsler Adult Intelligence Scale, third version, which has been validated for the Brazilian population.32 The following subtests were used: arithmetic, digit symbol, similarities, and picture completion. Subjects were asked about the highest grade completed successfully at school, as well as their income in the previous month in Brazilian reais.

Outcomes Performance in intelligence tests was evaluated in the 2012-2013 visit, at a mean of 30.2 years of age. Four psychologists who were unaware of participant intrauterine growth and nutritional status in childhood administered the Wechsler Adult Intelligence Scale, third version, which has been validated for the Brazilian population.32 The following subtests were used: arithmetic, digit symbol, similarities, and picture completion. Subjects were asked about the highest grade completed successfully at school, as well as their income in the previous month in Brazilian reais. Conditional Growth Conditional growth modeling was used in the analyses on the effect of weight and height gain.33 Conditional variables were obtained by regressing current size (weight or length/height) on birthweight and earlier measures of weight and length/height, and standardized residuals were derived. Conditional variables express how a child deviates from its expected height or weight, based on its previous measures and the growth of the studied population. At each time point, the conditional variable represents growth during a time interval, and a positive value represents a weight gain or linear growth faster than predicted in that period. For example, conditional relative weight gain at 2 years of age represents the relative weight from birth to 2 years of age. The conditional variable at 4 years of age represents height or relative weight gain from 2 to 4 years of age. To estimate conditional height, current length or height was regressed on previous weight and length. Therefore, conditional length at 2 years of age was estimated by regressing length-for-age z-scores at 2 years of age on birthweight. In contrast, conditional relative weight was estimated from length/height at that age and previous measures of length/height and weight. Therefore, conditional relative weight at 2 years of age was derived by regressing weight at 2 years of age on birthweight and length at 2 years of age.

2 years of age on birthweight. In contrast, conditional relative weight was estimated from length/height at that age and previous measures of length/height and weight. Therefore, conditional relative weight at 2 years of age was derived by regressing weight at 2 years of age on birthweight and length at 2 years of age. Confounding Variables Family income at birth was defined as total income earned by the family members in the month before the interview. Maternal and paternal years of schooling at birth was defined as years of schooling successfully completed. Household assets index in childhood was defined as based on the ownership of household goods and estimated using factor analysis.34 Maternal skin color was rated by the interviewer during the perinatal study. Maternal smoking during pregnancy was defined as those mothers with a history of smoking in the pregnancy being considered smokers. Breastfeeding duration was based on information on breastfeeding duration was collected in 1984 and 1986, and we used the information closest to the age of weaning to minimize recall bias.

Maternal smoking during pregnancy was defined as those mothers with a history of smoking in the pregnancy being considered smokers. Breastfeeding duration was based on information on breastfeeding duration was collected in 1984 and 1986, and we used the information closest to the age of weaning to minimize recall bias. Data Analyses ANOVA was used to compare means and multiple linear regression to obtain estimates that were adjusted for the following confounders: family income at birth, maternal years of schooling at birth, paternal years of schooling in childhood, household assets index, maternal skin color, and maternal smoking during pregnancy. Estimates on the effect of nutritional status in childhood and conditional growth were further adjusted to breastfeeding duration. Furthermore, for conditional length gain from birth to 2 years of age, estimates were also adjusted for birthweight according to gestational age z-score, whereas for conditional relative weight gain from birth to 2 years of age, analyses also controlled for conditional length gain from 0 to 2 years of age, for conditional length gain from 2 to 4 years of age, birthweight and conditional variables from 0 to 2 years of age were also controlled for, and for relative weight gain from 2 to 4 years of age, length gain from 2 to 4 years of age was also included in the model. Statistical comparisons between groups were based on tests of heterogeneity and linear trend in the case of ordinal variables, and the one with the lower P value was presented. The suest command was used to compare regression coefficients. We tested for interaction between birthweight for gestational age z-scores and conditional growth variables.

een groups were based on tests of heterogeneity and linear trend in the case of ordinal variables, and the one with the lower P value was presented. The suest command was used to compare regression coefficients. We tested for interaction between birthweight for gestational age z-scores and conditional growth variables. Results In the 2012-2013 visit, 3701 subjects were interviewed; added to the 325 known to have died, this represents a follow-up rate of 68.1%. Information on IQ at 30 years of age was available for 3611 subjects (61.1% of the original cohort), whereas complete data on IQ and nutritional status at birth and 2 and 4 years of age, were available for 2477 individuals (41.9% of the original cohort). Subjects with complete data were more likely to be female, to have been born in the intermediate socioeconomic groups, and have a birthweight of ≥2500 grams (Table I; available at www.jpeds.com). The mean IQ of those subjects with full growth data was 98.9 points (95% CI, 98.4-99.4), compared with 96.0 (95% CI, 95.2-96.7) among those who did not have full anthropometric assessments.

I; available at www.jpeds.com). IQ, years of schooling, and income at 30 years of age were positively associated with socioeconomic status at birth and childhood. Maternal smoking during the pregnancy was associated with lower IQ, years of schooling, and income at 30 years of age (Table IV; available at www.jpeds.com). In unadjusted analyses, birthweight was positively associated with IQ, years of schooling, and income at 30 years of age. Controlling for confounding variables reduced the strength of the association, but subjects whose birthweight was ≥3500 grams still presented an higher IQ (2.93 points; 95% CI, 1.47-4.39) and income at 30 years of age (R$ 315; 95% CI, 91-538) than low birthweight subjects. The crude and adjusted results also show significant linear trends for years of schooling and income. Birthweight for gestational age was also positively associated with IQ and years of schooling, but not with income. In contrast, preterm birth was not associated with the outcomes (Table V). Table VI shows that attained weight and length/height for age z-scores at 2 and 4 years of age were associated positively with the 3 outcomes. With the exception of the association between length for age at 2 years of age and family income in the adjusted model, all other associations were significant. Adjustment for confounding attenuated the magnitude of the associations, and in general weight was associated more strongly with the outcomes than length or height.

the exception of the association between length for age at 2 years of age and family income in the adjusted model, all other associations were significant. Adjustment for confounding attenuated the magnitude of the associations, and in general weight was associated more strongly with the outcomes than length or height. The associations with the conditional variables are shown in Table VII. Conditional length at age 2 years of age, a proxy for linear growth from birth to this age, was positively associated with IQ, years of schooling, and income at 30 years of age. The associations were stronger for linear growth than for relative weight at 2 years of age, and the latter association with income was not significant after adjustment. Neither conditional height nor weight at 4 years of age were associated with any of the outcomes in the adjusted analyses. Table VIII shows the interactions between size at birth and the conditional measures. Only 1 of the 4 interactions was significant (P = .04). Conditional length at 4 years of age was associated positively with IQ at 30 years of age only among children whose birthweight for gestational age z-score was in the lowest tertile.

e VIII shows the interactions between size at birth and the conditional measures. Only 1 of the 4 interactions was significant (P = .04). Conditional length at 4 years of age was associated positively with IQ at 30 years of age only among children whose birthweight for gestational age z-score was in the lowest tertile. Discussion In this population-based prospective birth cohort, we observed that attained size (birthweight, weight for age and length, or height for age) are associated with IQ, years of schooling, and income at 30 years of age after controlling for several confounding variables. Associations tended to be stronger for attained weight than for length or height, and were present at both 2 and 4 years of age. In contrast, the picture emerging from conditional analyses is more refined. Linear growth was associated more strongly with the 3 outcomes than relative weight gain, for which there was no evidence of benefit after the age of 2 years of age. Associations with IQ and years of schooling tended to be stronger than those with income. Nevertheless, income was 20% higher in the group whose conditional length was ≥1 SD above what was expected, compared with those ≥1 SD below the expected. Weaker associations with income were expected because of the longer causal chain that presumably links early growth to adult earnings. We also sought interactions between size at birth and postnatal growth. Linear growth from 2 to 4 years of age was related to increased IQ among children with lower birthweights according to gestational age, suggesting potential benefits of late catch-up among those with intrauterine growth restriction, with positive consequences on human capital.

etween size at birth and postnatal growth. Linear growth from 2 to 4 years of age was related to increased IQ among children with lower birthweights according to gestational age, suggesting potential benefits of late catch-up among those with intrauterine growth restriction, with positive consequences on human capital. Our analyses report upon findings based on 2 types of anthropometric variables. Attained size (as weight for age or length/height for age) indicates cumulative growth or weight gain from conception to the age of measurement. Conditional analyses, in contrast, allow investigation of the effect of growth in different time periods. Conditional variables at 2 years of age indicate growth from birth to this age, and those at 4 years of age reflect changes from 2-4 years of age. Also, attained weight for age reflects both linear growth and relative weight gain above and beyond what would have been expected from linear growth alone. These distinctions explain why the 2 sets of analyses show apparently contradictory findings.

ge, and those at 4 years of age reflect changes from 2-4 years of age. Also, attained weight for age reflects both linear growth and relative weight gain above and beyond what would have been expected from linear growth alone. These distinctions explain why the 2 sets of analyses show apparently contradictory findings. With respect to study limitations, subjects included in the analyses were more likely to be female and to belong to the intermediate family income groups when compared with the full cohort. The lower representation of children born with low birthweight and in the poorest income group is associated with their higher mortality in infancy. In contrast, birthweight and nutritional status in childhood were not associated with follow-up rate at the 2012-2013 visit.29 Furthermore, the regression coefficients for nutritional status in childhood was similar among those included and not included in the conditional growth analyses (P = .31). Therefore, selection bias is unlikely.

weight and nutritional status in childhood were not associated with follow-up rate at the 2012-2013 visit.29 Furthermore, the regression coefficients for nutritional status in childhood was similar among those included and not included in the conditional growth analyses (P = .31). Therefore, selection bias is unlikely. Residual confounding should be considered, because socioeconomic conditions are associated with the exposures and the outcomes.35, 36 When we controlled for several confounders measured soon after birth or in childhood in the analyses, which are positively associated with growth and with adult IQ and income,37 as expected, the regression coefficients were attenuated. The strongest argument against residual confounding is that the associations between conditional measures and the outcomes were specific—that is, they were higher for linear growth than for relative weight, and higher at 2 years of age than at 4 years of age. Residual confounding cannot account for such specificity in the results. Moreover, the observed association should not be considered as being owing to residual confounding by maternal intelligence and home stimulation. We controlled for several socioeconomic variables and breastfeeding, which are highly correlated with these unmeasured confounders.37 Therefore, a noncausal pathway has been closed.

r, the observed association should not be considered as being owing to residual confounding by maternal intelligence and home stimulation. We controlled for several socioeconomic variables and breastfeeding, which are highly correlated with these unmeasured confounders.37 Therefore, a noncausal pathway has been closed. Anthropometric assessments were carried out by trained interviewers and most of the confounding variables were measured in the perinatal study or in childhood, with a short recall. These measures, therefore, decreased measurement error and residual confounding. Furthermore, we used a standardized test to assess IQ. Similar to our findings, most studies evaluating the association of growth at different moments in the life cycle with school performance or cognition reported positive associations with early growth, whereas late growth showed no or a weak association.21, 22, 23, 24, 25, 26 These findings are biologically plausible. Brain growth spurt occurs between the last trimester of pregnancy and at about 3-4 years of age, and growth velocity is higher in the first months of life.38 Indeed, most studies on the effect of birthweight and weight gain in childhood found that school achievement or IQ was associated more strongly with growth in childhood than intrauterine growth.21, 22, 23, 24, 39

er of pregnancy and at about 3-4 years of age, and growth velocity is higher in the first months of life.38 Indeed, most studies on the effect of birthweight and weight gain in childhood found that school achievement or IQ was associated more strongly with growth in childhood than intrauterine growth.21, 22, 23, 24, 39 This study on adult intelligence attempted to disentangle the effect of early linear growth from that of weight gain. Weight gain is a combination of linear growth and changes in soft tissue, and conditional relative weight gain represents the weight gain that is not owing to linear growth. We observed that linear growth in early childhood has positive long-term consequences on human capital, not only by improving IQ, but also increasing years of schooling and earning ability. In contrast, relative weight gain had a small impact on IQ and no effect on income. The regression coefficients for linear growth and relative weight gain at a given age may be compared, because both conditional growth variables are expressed as SD scores. In a pooled analysis of 5 cohorts from low- and middle-income countries (including data from the 23-year follow-up visit of the 1982 Pelotas cohort), Adair et al9 reported that birthweight, conditional length, and relative conditional weight at 2 years of age were associated inversely with the odds of failing to complete high school, and that these associations were strongest for conditional length gain. Our results are consistent with these earlier findings and, by showing a link with intelligence, the plausibility of the association with years of schooling is strengthened. As mentioned, brain growth velocity is higher in the first months of life,38 which would explain the specific association of early linear growth with IQ, years of schooling, and income.

arlier findings and, by showing a link with intelligence, the plausibility of the association with years of schooling is strengthened. As mentioned, brain growth velocity is higher in the first months of life,38 which would explain the specific association of early linear growth with IQ, years of schooling, and income. Our results support the relevance of monitoring and promoting linear growth and shows that additional weight gain, after taking into account linear growth, has a small impact on human capital. Therefore, nutrition intervention programs should aim at promoting linear growth instead of weight gain, as well as intrauterine growth. And these interventions should focus in the first 1000 days of life, because they will have long-term consequences on human capital, by increasing IQ, years of schooling, and earning ability. Appendix Table I Baseline characteristics of the cohort members according to inclusion in the conditional growth analyses

Our results support the relevance of monitoring and promoting linear growth and shows that additional weight gain, after taking into account linear growth, has a small impact on human capital. Therefore, nutrition intervention programs should aim at promoting linear growth instead of weight gain, as well as intrauterine growth. And these interventions should focus in the first 1000 days of life, because they will have long-term consequences on human capital, by increasing IQ, years of schooling, and earning ability. Appendix Table I Baseline characteristics of the cohort members according to inclusion in the conditional growth analyses Table I Included in the conditional growth analyses Yes No P value N % N % Sex Male 1199 48.4 1839 53.5 <.001 Female 1278 51.6 1598 46.5 Family income at birth(minimum wage) <.001 ≤1 414 16.8 874 25.6 1.1-3 1222 49.4 1567 45.9 3.1-6 532 21.5 559 16.4 6.1-10 165 6.7 217 6.4 >10 137 5.6 198 5.8 Maternal years of schooling <.001 0-4 704 28.4 1256 36.6 5-8 1086 43.9 1368 39.9 9-11 290 11.7 364 10.6 ≥12 395 16.0 444 12.9 Birthweight (g) <.001 <2500 147 5.9 387 11.3 2500-2999 579 23.4 814 23.7 3000-3499 948 38.3 1272 37.1 ≥3500 803 32.4 959 27.9 Total* 2477 3437 * For some variables, the number of subjects might not sum to the subjects included or not in the conditional growth analysis because of missing information. Table II Characteristics of study sample

Table I Included in the conditional growth analyses Yes No P value N % N % Sex Male 1199 48.4 1839 53.5 <.001 Female 1278 51.6 1598 46.5 Family income at birth(minimum wage) <.001 ≤1 414 16.8 874 25.6 1.1-3 1222 49.4 1567 45.9 3.1-6 532 21.5 559 16.4 6.1-10 165 6.7 217 6.4 >10 137 5.6 198 5.8 Maternal years of schooling <.001 0-4 704 28.4 1256 36.6 5-8 1086 43.9 1368 39.9 9-11 290 11.7 364 10.6 ≥12 395 16.0 444 12.9 Birthweight (g) <.001 <2500 147 5.9 387 11.3 2500-2999 579 23.4 814 23.7 3000-3499 948 38.3 1272 37.1 ≥3500 803 32.4 959 27.9 Total* 2477 3437 * For some variables, the number of subjects might not sum to the subjects included or not in the conditional growth analysis because of missing information. Table II Characteristics of study sample Table II Mean (SD) n % Variables measured at birth Monthly family income (minimum wage) ≤1 706 19.6 1.1-3 1772 49.3 3.1-6 707 19.7 6.1-10 218 6.1 >10 191 5.3 Maternal education (y) 0-4 1154 32.0 5-8 1557 43.2 9-11 394 10.9 ≥12 501 13.9 Maternal skin color White 2969 82.2 Non-white 641 17.8 Birthweight (g) <2500 259 7.2 2500-2999 862 23.9 3000-3499 1356 37.5 ≥3500 1133 31.4 Gestational age (wk) ≤36 164 5.6 37-38 645 22.1 ≥39 2106 72.3 Birthweight for gestational age (z-score) <−1.28 SD 413 14.2 −1.28-0 SD 1296 44.4 >0 SD 1205 41.4 Maternal smoking during pregnancy Yes 1262 35.0 No 2342 65.0 Variables measured in childhood Household assets index score (points) 0.34 (−0.55 to 0.68)* Paternal education (y) 0-4 863 27.8 5-8 1415 45.6 9-11 405 13.1 ≥12 419 13.5 Breastfeeding duration (mo) <1 736 21.1 1-2.9 895 25.6 3-5.9 808 23.1 ≥6 1054 30.2 Height-for-age z-score at 2 years ≤−2 424 12.8 −1.99 to −1 821 24.8 −0.99 to 1 1818 55.1 >1 242 7.3 Weight-for-age z-score at 2 years ≤−2 88 2.7 −1.99 to −1 356 10.8 −0.99 to 1 2225 67.3 >1 637 19.3 Height-for-age z-score at 4 years ≤−2 338 10.4 −1.99 to −1 842 26.0 −0.99 to 1 1842 56.9 >1 216 6.7 Weight-for-age z-score at 4 years ≤−2 65 2.0 −1.99 to −1 431 13.3 −0.99 to 1 2252 69.6 >1 488 15.1 Variables measured at 30 years IQ 98.0 (12.6) Educational attainment (y) 11.4 (4.1) Monthly income (R$) 1000 (530-1890)* * Median and interquartile range.

≤−2 338 10.4 −1.99 to −1 842 26.0 −0.99 to 1 1842 56.9 >1 216 6.7 Weight-for-age z-score at 4 years ≤−2 65 2.0 −1.99 to −1 431 13.3 −0.99 to 1 2252 69.6 >1 488 15.1 Variables measured at 30 years IQ 98.0 (12.6) Educational attainment (y) 11.4 (4.1) Monthly income (R$) 1000 (530-1890)* * Median and interquartile range. Table III Prevalence of birth conditions, nutritional status in childhood and rapid growth in childhood according to socioeconomic, demographic, and maternal smoking variables

≤−2 338 10.4 −1.99 to −1 842 26.0 −0.99 to 1 1842 56.9 >1 216 6.7 Weight-for-age z-score at 4 years ≤−2 65 2.0 −1.99 to −1 431 13.3 −0.99 to 1 2252 69.6 >1 488 15.1 Variables measured at 30 years IQ 98.0 (12.6) Educational attainment (y) 11.4 (4.1) Monthly income (R$) 1000 (530-1890)* * Median and interquartile range. Table III Prevalence of birth conditions, nutritional status in childhood and rapid growth in childhood according to socioeconomic, demographic, and maternal smoking variables Table III Low birthweight (%) Preterm birth (%) Small-for-gestational age (%) Weight-for-age z-score at 2 years ≤−2 (%) Height-for-age z-score at 2 years ≤−2 (%) Weight-for-age z-score at 4 years ≤−2 (%) Height-for-age z-score at 4 years ≤−2 (%) Conditional length z-score at 2 years—≥1 (%) Conditional weight z-score at 2 years—≥1 (%) Conditional height z-score at 4 years—≥1 (%) Conditional weight z-score at 4 years—≥1 (%) Family income at birth(in minimum wages) P < .001 P = .9 P < .00 P < .001 P < .001 P < .00 P < .00 P < .001 P = .57 P = .42 P = .21 ≤1 11.8 56.1 118.8 6.6 23.9 14.8 122.0 5.6 16.2 14.3 16.0 1.1-3 6.2 5.4 15.4 2.5 13.2 1.8 9.5 13.1 13.9 14.0 12.5 3.1-6 6.1 5.9 10.9 0.5 6.2 0.5 5.3 22.4 16.8 16.6 15.8 6.1-10 6.0 6.2 8.3 0.5 4.6 1.0 5.7 23.2 15.9 14.6 13.4 ≥10 4.7 4.8 9.0 0.6 2.9 0.6 1.8 27.0 19.7 19.0 13.1 Maternal years of schooling at birth P = .03 P = .7 P = .00 P < .001 P < .001 P < .00 P < .00 P < .001 P = .48 P = .17 P = .43 0-4 7.8 55.3 515.5 5.1 20.7 14.6 117.0 9.8 15.8 12.4 14.5 5-8 8.0 6.1 15.7 2.3 12.3 1.0 9.6 13.3 14.9 15.7 14.7 9-11 4.3 4.7 11.0 0.3 3.6 0.6 4.0 20.1 15.2 16.3 11.8 ≥12 5.6 5.6 10.0 0.2 3.7 0.4 3.1 26.6 16.0 16.2 12.7 Maternal skin color P = .06 P = .7 P = .61 P < .001 P < .001 P < .00 P < .00 P < .001 P = .22 P = .18 P = .88 White 6.8 05.7 14.0 2.2 11.0 11.9 19.2 16.2 15.4 14.5 13.8 Non-white 8.9 5.3 14.9 4.8 21.5 2.4 16.0 10.4 15.1 17.1 14.9 Household assets index (quintiles) P = .02 P = .3 P < .00 P < .001 P < .001 P < .00 P < .00 P < .001 P = .41 P = .36 P = .22 First 10.0 16.7 118.9 6.6 25.7 14.4 122.6 6.0 14.6 12.2 13.7 Second 7.9 5.4 17.6 2.8 17.0 3.2 11.9 11.2 14.4 14.4 12.6 Third 6.6 5.5 13.3 1.7 8.2 1.2 7.5 15.0 16.7 15.2 12.5 Fourth 4.1 3.8 10.5 1.3 7.0 0.7 4.7 23.0 17.5 17.1 16.7 Fifth 7.0 7.3 9.4 1.0 6.0 0.8 4.7 25.3 13.1 16.6 17.5 Maternal smoking in the pregnancy P < .001 P = .6 P < .00 P < .001 P < .001 P < .00 P < .00 P = .01 P = .004 P = .75 P = .91 No 5.8 75.5 111.1 2.6 10.5 11.9 18.8 16.0 14.4 15.1 14.0 Yes 9.8 5.9 20.4 2.8 17.1 2.3 13.5 13.8 17.5 14.6 13.9 Table IV IQ, years of schooling, and income at 30 years of age, according to socioeconomic, demographic, and maternal smoking variables in the p

.001 P < .001 P < .00 P < .00 P = .01 P = .004 P = .75 P = .91 No 5.8 75.5 111.1 2.6 10.5 11.9 18.8 16.0 14.4 15.1 14.0 Yes 9.8 5.9 20.4 2.8 17.1 2.3 13.5 13.8 17.5 14.6 13.9 Table IV IQ, years of schooling, and income at 30 years of age, according to socioeconomic, demographic, and maternal smoking variables in the p regnancy

.001 P < .001 P < .00 P < .00 P = .01 P = .004 P = .75 P = .91 No 5.8 75.5 111.1 2.6 10.5 11.9 18.8 16.0 14.4 15.1 14.0 Yes 9.8 5.9 20.4 2.8 17.1 2.3 13.5 13.8 17.5 14.6 13.9 Table IV IQ, years of schooling, and income at 30 years of age, according to socioeconomic, demographic, and maternal smoking variables in the p regnancy Table IV IQ (points), mean (95% CI) Years of schooling, mean (95% CI) Monthly income (R$), mean (95% CI) Family income at birth (minimum wage) P < .001 P < .001 P < .001 ≤1 91.5(90.7-92.4) 8.9(8.6-9.1) 940(855-1024) 1.1-3 96.6(96.0-97.1) 10.7(10.6-10.9) 1255(1191-1320) 3.1-6 102.0(101.2-102.9) 13.1(12.8-13.4) 1894(1750-2038) 6.1-10 106.7(105.2-108.2) 14.5(14.1-14.9) 2583(2260-2907) ≥10 110.4(108.9-112.0) 15.8(15.3-16.2) 3208(2787-3628) Maternal years of schooling at birth P < .001 P < .001 P < .001 0-4 92.2(91.5-92.9) 9.2(9.0-9.4) 997(931-1062) 5-8 97.5(97.0-98.1) 11.1(10.9-11.3) 1356(1281-1430) 9-11 103.2(102.1-104.3) 13.2(12.9-13.5) 1870(1679-2063) ≥12 108.6(107.7-109.5) 15.3(15.1-15.6) 2846(2614-3078) Maternal skin color P < .001 P < .001 P < .001 White 99.2(98.8-99.7) 11.6(11.5-11.8) 1616(1548-1683) Non-white 92.1(91.2-93.0) 9.9(9.6-10.2) 977(906-1048) Household assets index(quintiles) P < .001 P < .001 P < .001 First 90.0(89.1-91.0) 8.4(8.1-8.7) 868(797-938) Second 95.2(94.3-96.0) 10.3(10.0-10.6) 1219(1117-1320) Third 100.4(99.7-101.0) 12.1(11.9-12.4) 1564(1466-1662) Fourth 101.3(100.0-102.6) 12.7(12.3-13.1) 1762(1556-1967) Fifth 104.2(103.1-105.2) 13.7(13.3-14.0) 2321(2100-2542) Maternal smoking during pregnancy P < .001 P < .001 P = .001 No 99.1(98.6-99.6) 11.7(11.5-11.8) 1573(1499-1648) Yes 95.9(95.2-96.6) 10.7(10.5-10.9) 1371(1282-1460)

.7-101.0) 12.1(11.9-12.4) 1564(1466-1662) Fourth 101.3(100.0-102.6) 12.7(12.3-13.1) 1762(1556-1967) Fifth 104.2(103.1-105.2) 13.7(13.3-14.0) 2321(2100-2542) Maternal smoking during pregnancy P < .001 P < .001 P = .001 No 99.1(98.6-99.6) 11.7(11.5-11.8) 1573(1499-1648) Yes 95.9(95.2-96.6) 10.7(10.5-10.9) 1371(1282-1460) Funded by Wellcome Trust (086974/Z/08/Z), International Development Research Center (Canada), CNPq, FAPERGS, and the Brazilian Ministry of Health. The authors declare no conflicts of interest. Table V IQ, years of schooling, and income at 30 years, according to birth conditions

.7-101.0) 12.1(11.9-12.4) 1564(1466-1662) Fourth 101.3(100.0-102.6) 12.7(12.3-13.1) 1762(1556-1967) Fifth 104.2(103.1-105.2) 13.7(13.3-14.0) 2321(2100-2542) Maternal smoking during pregnancy P < .001 P < .001 P = .001 No 99.1(98.6-99.6) 11.7(11.5-11.8) 1573(1499-1648) Yes 95.9(95.2-96.6) 10.7(10.5-10.9) 1371(1282-1460) Funded by Wellcome Trust (086974/Z/08/Z), International Development Research Center (Canada), CNPq, FAPERGS, and the Brazilian Ministry of Health. The authors declare no conflicts of interest. Table V IQ, years of schooling, and income at 30 years, according to birth conditions Table V IQ (points) Years of schooling Monthly income (R$) Mean (95% CI) Adjusted regression coefficient (95% CI) Mean (95% CI) Adjusted regression coefficient (95% CI) Mean (95% CI) Adjusted regression coefficient (95% CI) Birthweight (g) P < .001* P < .001* P < .001* P = .003* P < .001* P < .001* <2500 94.5(92.8to96.1) Reference(0) 10.5(10.0to11.0) Reference(0) 1190(1009to1370) Reference(0) 2500-2999 95.7(94.8to96.5) 1.02(−0.47to2.51) 10.6(10.4to10.9) 0.04(−0.43to0.52) 1266(1161to1370) 69(−159to296) 3000-3499 98.5(97.8to99.1) 2.22(0.79to3.65) 11.4(11.2to11.7) 0.29(−0.17to0.75) 1538(1444to1632) 193(−26to412) ≥3500 99.9(99.2to100.6) 2.93(1.47to4.39) 11.9(11.6to12.1) 0.48(0.01to0.95) 1709(1597to1821) 315(91to538) Preterm birth P = .29 P = .29 P = .21 P = .18 P = .84 P = .85 No 98.8(98.4to99.3) Reference(0) 11.6(11.5to11.8) Reference(0) 1592(1523to1660) Reference(0) Yes 97.8(95.8to99.8) −0.86(−2.56to0.84) 11.2(10.6to11.8) −0.36(−0.90to0.17) 1562(1263to1861) −12(−283to259) Birthweight for gestational age(z-score) P < .001* P = .001* P < .001* P < .001* P < .001* P = .10* <−1.28 95.7(94.5to96.9) Reference(0) 10.5(10.2to10.9) Reference(0) 1311(1159to1464) Reference(0) −1.28-0 98.1(97.4to98.8) 1.26(0.06to2.46) 11.3(11.1to11.5) 0.35(−0.02to0.72) 1558(1461to1655) 138(−53to328) >0 100.6(99.9to101.3) 2.33(1.10to3.55) 12.3(12.0to12.5) 0.82(0.44to1.21) 1721(1611to1831) 179(−16to374) Adjusted for family income at birth, parental years of schooling, household asset index score, maternal skin color, and maternal smoking during pregnancy.

(11.1to11.5) 0.35(−0.02to0.72) 1558(1461to1655) 138(−53to328) >0 100.6(99.9to101.3) 2.33(1.10to3.55) 12.3(12.0to12.5) 0.82(0.44to1.21) 1721(1611to1831) 179(−16to374) Adjusted for family income at birth, parental years of schooling, household asset index score, maternal skin color, and maternal smoking during pregnancy. * P-value for linear trend. Table VI IQ, years of schooling, and income at 30 years, according to nutritional status in childhood

(11.1to11.5) 0.35(−0.02to0.72) 1558(1461to1655) 138(−53to328) >0 100.6(99.9to101.3) 2.33(1.10to3.55) 12.3(12.0to12.5) 0.82(0.44to1.21) 1721(1611to1831) 179(−16to374) Adjusted for family income at birth, parental years of schooling, household asset index score, maternal skin color, and maternal smoking during pregnancy. * P-value for linear trend. Table VI IQ, years of schooling, and income at 30 years, according to nutritional status in childhood Table VI IQ (points) Years of schooling Monthly income (R$) Mean (95% CI) Adjusted regression coefficient (95% CI) Mean (95% CI) Adjusted regression coefficient (95% CI) Mean (95% CI) Adjusted regression coefficient (95% CI) Weight-for-age z-score at 2 years P < .001* P < .001* P < .001* P < .001* P < .001* P = .003* ≤−2 88.1(85.7to90.5) Reference(0) 7.6(6.8to8.4) Reference(0) 624(500to748) Reference(0) −1.99to−1 92.3(91.1to93.6) 2.33(−0.14to4.80) 9.5(9.1to9.9) 1.28(0.51to2.05) 1055(925to1185) 226(−147to599) −0.99to1 98.0(97.5to98.6) 4.29(1.99to6.59) 11.3(11.2to11.5) 1.92(1.20to2.64) 1482(1410to1554) 258(−89to605) >1 102.8(101.9to103.7) 5.70(3.21to8.18) 13.0(12.7to13.3) 2.50(1.72to3.27) 1929(1767to2090) 349(−26to724) Height-for-age z-score at 2 years P < .001* P < .001* P < .001* P < .001* P < .001* P = .67† ≤−2 90.6(89.4to91.8) Reference(0) 8.8(8.4to9.2) Reference(0) 1022(900to1143) Reference(0) −1.99to−1 95.4(94.6to96.3) 2.22(0.96to3.48) 10.4(10.1to10.7) 0.80(0.40to1.20) 1258(1158to1357) −41(−235to153) −0.99to1 100.0(99.5to100.6) 3.83(2.62to5.05) 12.0(11.8to12.2) 1.43(1.04to1.81) 1624(1537to1710) −14(−202to173) >1 105.2(103.9to106.5) 4.77(2.93to6.62) 13.8(13.4to14.2) 1.80(1.22to2.38) 2208(1928to2489) 112(−173to397) Weight-for-age z-score at 4 years P < .001* P < .001* P < .001* P < .001* P < .001* P = .005* ≤−2 86.1(83.2to89.0) Reference(0) 7.7(6.8to8.6) Reference(0) 697(526to868) Reference(0) −1.99to−1 93.3(92.1to94.4) 4.53(1.70to7.36) 9.8(9.5to10.2) 1.33(0.44to2.21) 977(867to1087) 9(−407to425) −0.99to1 98.0(97.5to98.5) 5.91(3.20to8.63) 11.3(11.2to11.5) 1.68(0.83to2.53) 1506(1435to1578) 213(−186to613) >1 103.6(102.6to104.6) 7.97(5.07to10.87) 13.3(12.9to13.6) 2.43(1.52to3.34) 1968(1783to2153) 328(−100to757) Height-for-age z-score at 4 years P < .001* P < .001* P < .001* P < .001* P < .001* P = .04* ≤−2 90.7(89.3to92.0) Reference(0) 8.8(8.4to9.3) Reference(0) 866(756to975) Reference(0) −1.99to−1 94.8(94.0to95.7) 1.85(0.45to3.25) 10.2(9.9to10.5) 0.67(0.23to1.10) 1259(1158to1362) 152(−53to358) −0.99to1 100.0(99.4to100.5) 3.71(2.35to5.07) 12.1(11.9to12.3) 1.33(0.91to1.76) 1638(1553to1722) 192(−7to391) >1 105.0(103.6to106.4) 5.33(3.32to7.33) 13.7(13.2to14.1) 1.74(1

Reference(0) 866(756to975) Reference(0) −1.99to−1 94.8(94.0to95.7) 1.85(0.45to3.25) 10.2(9.9to10.5) 0.67(0.23to1.10) 1259(1158to1362) 152(−53to358) −0.99to1 100.0(99.4to100.5) 3.71(2.35to5.07) 12.1(11.9to12.3) 1.33(0.91to1.76) 1638(1553to1722) 192(−7to391) >1 105.0(103.6to106.4) 5.33(3.32to7.33) 13.7(13.2to14.1) 1.74(1 .12to2.37) 2122(1839to2405) 316(19to613) Adjusted for family income at birth, parental years of schooling, household assets index, maternal skin color, maternal smoking during pregnancy, breastfeeding duration, and birthweight. * P value for linear trend. † P value for heterogeneity. Table VII IQ, years of schooling, and income at 30 years, according to conditional growth in childhood

.12to2.37) 2122(1839to2405) 316(19to613) Adjusted for family income at birth, parental years of schooling, household assets index, maternal skin color, maternal smoking during pregnancy, breastfeeding duration, and birthweight. * P value for linear trend. † P value for heterogeneity. Table VII IQ, years of schooling, and income at 30 years, according to conditional growth in childhood Table VII Regression coefficient (95% CI) IQ (points) Years of schooling Monthly income (R$) Crude Adjusted Crude Adjusted Crude Adjusted Conditional lengthat 2 years (z-score) P < .001* P < .001* P < .001* P < .001* P < .001* P = .06† ≤−1 Reference(0) Reference(0) Reference(0) Reference(0) Reference(0) Reference(0) −0.99to−0.01 4.36(2.90to5.83) 2.11(0.79to3.42) 1.50(1.03to1.97) 0.73(0.32to1.13) 336(115to558) 106(−105to317) 0.00to0.99 7.95(6.48to9.41) 3.98(2.64to5.32) 2.68(2.22to3.15) 1.30(0.89to1.72) 458(236to680) 33(−183to249) ≥1 10.52(8.79to12.24) 4.41(2.79to6.02) 3.80(3.25to4.35) 1.62(1.12to2.12) 961(700to1221) 299(39to558) Conditional relative weightat 2 years(z-score) P = .02* P = .04* P = .004† P = .04† P = .02* P = .05* ≤−1 Reference(0) Reference(0) Reference(0) Reference(0) Reference(0) Reference(0) −0.99to−0.01 1.08(−0.37to2.53) 0.26(−1.03to1.55) 0.57(0.11to1.03) 0.26(−0.14to0.66) 61(−159to280) −38(−245to170) 0.00to0.99 1.30(−0.17to2.77) 0.60(−0.71to1.91) 0.84(0.37to1.30) 0.54(0.14to0.95) 184(−39to407) 91(−120to301) ≥1 2.09(0.38to3.81) 1.49(−0.04to3.01) 0.78(0.24to1.33) 0.50(0.03to0.97) 258(−1to518) 170(−76to416) Conditional heightat 4 years(z-score) P < .001† P = .51† P < .001† P = .56† P < .001† P = .15† ≤−1 Reference(0) Reference(0) Reference(0) Reference(0) Reference(0) Reference(0) −0.99to−0.01 2.05(0.58to3.52) 0.63(−0.68to1.95) 0.68(0.22to1.15) 0.14(−0.27to0.54) 367(145to589) 214(3to425) 0.00to0.99 3.03(1.55to4.51) 1.03(−0.31to2.36) 0.98(0.51to1.45) 0.22(−0.19to0.63) 461(237to686) 243(29to457) ≥1 2.50(0.76to4.24) 0.74(−0.83to2.31) 0.71(0.15to1.26) −0.03(−0.51to0.45) 376(112to639) 179(−73to431) Conditional relative weight4 years(z-score) P = .07* P = .15† P = .09* P = .24* P = .15† P = .17† ≤−1 Reference(0) Reference(0) Reference(0) Reference(0) Reference(0) Reference(0) −0.99to−0.01 −0.30(−1.81to1.20) 0.05(−1.30to1.39) −0.24(−0.72to0.24) −0.10(−0.51to0.32) −15(−243to214) 34(−182to250) 0.00to0.99 −0.14(−1.69to1.41) 0.17(−1.21to1.55) −0.24(−0.74to0.25) −0.11(−0.54to0.31) 151(−84to386) 187(−35to409) ≥1 −1.99(−3.80to−0.17) −1.33(−2.95to0.29) −0.57(−1.15to0.01) −0.32(−0.82to0.18) −65(−341to210) 23(−238to284) Adjusted for family income at birth, parent

0.24) −0.10(−0.51to0.32) −15(−243to214) 34(−182to250) 0.00to0.99 −0.14(−1.69to1.41) 0.17(−1.21to1.55) −0.24(−0.74to0.25) −0.11(−0.54to0.31) 151(−84to386) 187(−35to409) ≥1 −1.99(−3.80to−0.17) −1.33(−2.95to0.29) −0.57(−1.15to0.01) −0.32(−0.82to0.18) −65(−341to210) 23(−238to284) Adjusted for family income at birth, parent al years of schooling, household index score, maternal skin color, maternal smoking during pregnancy, birthweight and breastfeeding duration. * Test for linear trend. † Test for heterogeneity. Table VIII Conditional growth analyses of IQ at 30 years according to birthweight for gestational age z-score* Table VIII Adjusted regression coefficient (95% CI) P-value for interaction Birthweight for gestational age z-score First tertile Second tertile Third tertile Conditional length 2 years 1.51(0.74to2.28) 1.02(0.24to1.80) 2.35(1.54to3.15) .44 Conditional relative weight 2 years 0.26(−0.45to0.98) 0.33(−0.42to1.07) 0.88(0.15to1.62) .20 Conditional height 4 years 0.98(0.25to1.71) 0.31(−0.45to1.07) −0.24(−0.97to0.50) .04 Conditional relative weight 4 years −0.34(−1.08to0.39) 0.00(−0.72to0.73) −0.86(−1.58to−0.14) .30 Adjusted for family income at birth, parental years of schooling, household index score, maternal skin color, maternal smoking during pregnancy, and breastfeeding duration. * Results are expressed as change in IQ(points) associated with one z-score of the conditional variable.

The prevalence of childhood overweight and obesity is increasing in almost every region of the world.1 At the same time, several low- and middle-income countries continue to confront high rates of undernutrition and impaired child growth, and are thus simultaneously facing a dual burden of malnutrition among children. In low-resource settings in particular, where 156 million children under 5 years of age experience chronic undernutrition,2 the widespread prevalence of undernutrition and impaired growth in early life may have profound impacts on long-term health outcomes. Children who suffer from undernutrition in early life are more likely to have reduced adult stature, impaired cognitive development, lower earning potential, and if they are women, also a higher risk of birth complications.3 Furthermore, the developmental origins of health and disease paradigm proposes that perturbations in the homeostasis of a developing fetus or infant result in long-term changes affecting that individual's risk of future diseases.4 Several studies have documented an increased risk of cardiometabolic risk factors, such as high blood pressure, dyslipidemia, and insulin resistance, among individuals who were born at a low birth weight5, 6, 7, 8 and among those with a low body mass index-for-age z-score (BMIZ) in early childhood.6, 9, 10 The relative importance of perinatal versus infant growth in long-term health outcomes, however, is unknown. Furthermore, much of the evidence on the relationship between early life growth and adult health outcomes comes from wealthy nations,5, 9, 11 despite the growing evidence that the relative influence on growth in early life on later-life health outcomes may be population specific.12 Specifically, there is limited research from sub-Saharan Africa, where the risk factors for poor growth as well as chronic disease outcomes are likely quite different from those in developed nations.

evidence that the relative influence on growth in early life on later-life health outcomes may be population specific.12 Specifically, there is limited research from sub-Saharan Africa, where the risk factors for poor growth as well as chronic disease outcomes are likely quite different from those in developed nations. Recent data suggest that environmental enteric dysfunction (EED), a subclinical condition of the small intestine characterized by villous atrophy, crypt hyperplasia, increased intestinal permeability, inflammatory cell infiltrate, and malabsorption, is associated with growth failure and stunting.13, 14, 15 Some have also hypothesized that EED may increase the risk for cardiometabolic diseases in later life, including insulin resistance and hypertension.16 Although the gold standard for assessing EED is small bowel biopsy, the invasive and expensive nature of these procedures has led researchers to pursue biomarkers that are more suitable for widespread use in community-based settings. Our group has assessed previously the use of anti-flagellin and anti-lipopolysaccharide (LPS) IgA and IgG as biomarkers of EED. We found that anti-LPS and anti-flagellin IgA and IgG concentrations increased over the first year of life in Tanzanian infants, that the concentrations in Tanzanian infants were significantly higher than in healthy controls in Boston, and that elevated anti-LPS and anti-flagellin IgA and IgG concentrations were associated with an increased risk of underweight in infancy.17 In our current study, we assess whether these biomarkers continue to be associated with growth and health outcomes in midchildhood.

ntly higher than in healthy controls in Boston, and that elevated anti-LPS and anti-flagellin IgA and IgG concentrations were associated with an increased risk of underweight in infancy.17 In our current study, we assess whether these biomarkers continue to be associated with growth and health outcomes in midchildhood. Given the limited research on how nutritional status and EED in early infancy relate to long-term growth and cardiometabolic risk factors in children in low-resource settings, we collected data from a cohort of children in Dar es Salaam, Tanzania, to assess the relative importance of infant growth and nutritional status at age 6 weeks of age and change from 6 to 52 weeks of age, as well as biomarkers of EED in infancy, on health and growth outcomes in midchildhood.

in low-resource settings, we collected data from a cohort of children in Dar es Salaam, Tanzania, to assess the relative importance of infant growth and nutritional status at age 6 weeks of age and change from 6 to 52 weeks of age, as well as biomarkers of EED in infancy, on health and growth outcomes in midchildhood. Methods The study sample included children born in Dar es Salaam, Tanzania, who participated in 1 of 2 randomized controlled trials of multiple micronutrient supplementation to infants. The first trial (ClinicalTrials.gov: NCT00197730) randomized 2387 infants born to HIV-infected mothers to either daily administration of multiple micronutrients (vitamins B complex, C, and E) or placebo at 6 weeks of age.18 Randomization of infants occurred between August 2004 and November 2007; follow-up ended in May 2008. The micronutrient supplements did not show an effect on mortality, morbidity, or child growth.18, 19 The second trial (ClinicalTrials.gov: NCT00421668) was implemented with a 2 × 2 factorial design assessing the effect of zinc, zinc plus multivitamins (the same combination of vitamins B complex, C, and E as described), multivitamins alone, or placebo among 2400 infants born to HIV-negative women.20 The second trial found that zinc supplementation reduced the risk of acute respiratory and diarrheal infections,20 but that neither supplement alone nor in combination had an effect on rates of stunting, wasting, or underweight.21 The 2 studies were designed to allow for pooled analyses—they were conducted in overlapping clinics with similar staff, they used identical inclusion/exclusion criteria (other than maternal HIV status), and they collected the same sociodemographic and clinical data on all mothers and children. In both trials, infants were randomized at 6 weeks of age, and mothers were asked to bring the children to the clinic for follow-up visits every 4 weeks after randomization. At each monthly follow-up visit, a trained study nurse measured child anthropometry using standard techniques.22 Weight was measured on a digital infant balance scale with 10-g precision (Tanita, Arlington Heights, Illinois) and length with 1-mm precision using a rigid length board with an adjustable foot piece.

ation. At each monthly follow-up visit, a trained study nurse measured child anthropometry using standard techniques.22 Weight was measured on a digital infant balance scale with 10-g precision (Tanita, Arlington Heights, Illinois) and length with 1-mm precision using a rigid length board with an adjustable foot piece. For the current study, we identified children who participated in the 2 original trials who met the following criteria: children with complete physical descriptions of their home addresses on file, who had anthropometric data at 6 weeks, who had participated in their original trial through 15 months of age, and who were available for contact during the follow-up study recruitment period of June to August 2014. From the 2387 children in the first trial, a list of all children who fulfilled these criteria was generated and simple random sampling was used to select children for follow-up. From the 2400 children in the second trial, we selected from 269 children who had participated in an enteric disease substudy because these children had provided blood specimens at both 6 weeks and 6 months of age.17 Further inclusion criteria in the substudy was that children had length-for-age z-score (LAZ) > −2 at 6 weeks.

e 2400 children in the second trial, we selected from 269 children who had participated in an enteric disease substudy because these children had provided blood specimens at both 6 weeks and 6 months of age.17 Further inclusion criteria in the substudy was that children had length-for-age z-score (LAZ) > −2 at 6 weeks. Laboratory Assessments Microtiter plates were coated with purified Escherichia coli flagellin (100 ng/well) or purified Escherichia coli LPS (2 mg/well). Serum samples from study participants were diluted 1:200 and applied to wells coated with flagellin or LPS. After incubation and washing, the wells were incubated with anti-human IgA (KPL, Milford, Massachusetts) or IgG (GE Healthcare, Little Chalfont, United Kingdom) coupled to a horseradish peroxidase. The quantification of total immunoglobulins was performed with the use of the colorimetric peroxidase substrate tetramethylbenzidine, and absorbance (optical density) was read at 450 nm with the use of an enzyme-linked immunosorbent assay plate reader. Data are reported as optical density-corrected data by subtracting background concentrations, which were determined from the readings in samples that lacked serum.

idase substrate tetramethylbenzidine, and absorbance (optical density) was read at 450 nm with the use of an enzyme-linked immunosorbent assay plate reader. Data are reported as optical density-corrected data by subtracting background concentrations, which were determined from the readings in samples that lacked serum. Follow-Up Assessments For all children who participated in the follow-up study, trained study nurses measured their weight, height, and blood pressure when the children were between 4 and 9 years of age. Diastolic, systolic, and mean arterial blood pressure in mm Hg were measured with a DINAMAP DPC120X-EN (GE Medical Systems Information Technologies Inc, Milwaukee, Wisconsin). Blood pressure measurements were assessed 5 times by a single observer and then averaged. Body weight was measured with an electronic digital scale accurate to 0.1 kg (Tanita), and standing height was measured with a stadiometer to the nearest 0.1 cm. Ethical approval for both parent trials was granted by the Harvard T.H. Chan School of Public Health Human Subjects Committee, the Muhimbili University of Health and Allied Sciences Committee of Research and Publications, the Tanzanian Institute for Medical Research, and the Tanzanian Food and Drug Authority; the follow-up study was approved by the Harvard T.H. Chan School of Public Health Human Subjects Committee and the Muhimbili University of Health and Allied Sciences Committee of Research and Publications. All mothers provided written informed consent to enroll themselves and their children in the original and follow-up studies.

-up study was approved by the Harvard T.H. Chan School of Public Health Human Subjects Committee and the Muhimbili University of Health and Allied Sciences Committee of Research and Publications. All mothers provided written informed consent to enroll themselves and their children in the original and follow-up studies. Statistical Analyses Details on data collection and management from the 2 trials have been published previously.18, 19, 20, 21 In brief, data were double entered by using Microsoft Access software and converted to SAS datasets (version 9.4; SAS Institute, Cary, North Carolina) for analysis. For growth analyses, we calculated age- and sex-specific z-scores for 3 anthropometric indices during infancy: weight-for-length z-score (WLZ), LAZ, and weight-for-age z-score (WAZ) using the 2006 World Health Organization (WHO) growth standards.23 In accordance with WHO recommendations, we set all extreme LAZ (<−6 or >6), WLZ (<−5 or >5), and WAZ (<−6 or >5) values to missing.24 For follow-up anthropometric data, we calculated age- and sex-standardized height-for-age z-score (HAZ), WAZ, and BMIZ using the 2006 WHO growth standards for children younger than 5 years of age and the WHO 2007 Growth Reference for children aged 5 years and older.25 In accordance with WHO guidelines, stunting was defined as HAZ <−2, and overweight as BMIZ >1, and obese as BMIZ >2.

ndardized height-for-age z-score (HAZ), WAZ, and BMIZ using the 2006 WHO growth standards for children younger than 5 years of age and the WHO 2007 Growth Reference for children aged 5 years and older.25 In accordance with WHO guidelines, stunting was defined as HAZ <−2, and overweight as BMIZ >1, and obese as BMIZ >2. Descriptive statistics, including means with standard deviations and frequencies with percentages, were used to summarize sociodemographic information on maternal, child, and household characteristics as well as the growth characteristics of children who participated in the follow-up study. Our outcomes of interest were HAZ, BMIZ, and WAZ; systolic and diastolic blood pressures; and overweight, obesity, and stunting in midchildhood. Our exposures of interest were baseline anthropometric indicators (6-week LAZ, WAZ, and WLZ) and change in each anthropometric indicator from 6 to 52 weeks of age, as well as EED biomarkers (anti-flagellin and anti-LPS IgG and IgA) at 6 weeks and 6 months of age. For each continuous outcome of interest, we first conducted univariate linear regression models with each anthropometric and EED biomarker as predictors of interest. Confounders for multivariate models were selected based on a review of the literature and included which trial the child originally participated in, the trial treatment arm, and the child's sex for all models; maternal height and maternal education for all models for anthropometric outcomes; and child's gestational age at birth as well as age at midchildhood follow-up for all blood pressure models. Multivariate models assessing the effect of anthropometry in infancy included each anthropometric indicator at 6 weeks of age as well as the change in the same indicator from 6 to 52 weeks of age to assess relative importance of the 2 time periods for each indicator. The assumption of linearity was assessed using plots of residuals versus predicted values, and normality based on normal quantile plots.

nthropometric indicator at 6 weeks of age as well as the change in the same indicator from 6 to 52 weeks of age to assess relative importance of the 2 time periods for each indicator. The assumption of linearity was assessed using plots of residuals versus predicted values, and normality based on normal quantile plots. For our primary outcomes (anthropometry and blood pressure in midchildhood), data from 113 subjects and 8 covariates in the multivariate linear regression models, provided 80% power to detect a minimal increase of 7% in the R2 value for a predictor of interest at significance level of 0.05. In models where biomarkers of EED were predictors of interest, the number of subjects was reduced to 66 in our analysis, which provided 80% power to detect a minimal increase of 12% in R2 value at a significance level of 0.05. We also conducted univariate and multivariate log-binomial regression models to assess the relationship between each of our predictors of interest and overweight, obesity, and stunting in midchildhood. Given the limited power to assess these outcomes, multivariate models for binary anthropometric outcomes only adjust for which trial the child participated in, the child's sex, and the corresponding anthropometric indicator at baseline or change from 6 to 52 weeks of age. Multivariate models for stunting in midchildhood also adjust for maternal height. All analyses were conducted in SAS version 9.4 (SAS Institute).

tcomes only adjust for which trial the child participated in, the child's sex, and the corresponding anthropometric indicator at baseline or change from 6 to 52 weeks of age. Multivariate models for stunting in midchildhood also adjust for maternal height. All analyses were conducted in SAS version 9.4 (SAS Institute). Results Research staff attempted to contact 327 children and successfully reconsented and re-enrolled 113 children, 47 from the first trial (children born to HIV-infected mothers, none of whom had tested positive for HIV) and 66 from the second trial (children born to HIV-uninfected mothers). All 66 children from the second trial had blood samples drawn at 6 weeks and 6 months of age. There were no differences in sociodemographic characteristics or early infant growth indicators when comparing children who participated in the follow-up study and the original trials with the exceptional of gestational age in weeks, which was slightly longer in children in the follow-up study compared with the parent trials (40.0 ± 2.3 vs 39.4 ± 2.7). Mean LAZ, WLZ, and WAZ were already below 0 at 6 weeks of age, and the means continued to decline through age 12 months (Table I). The age range for children at follow-up was 4.6-9.8 years (6.8 ± 1.6 years). Children born to HIV-infected mothers (participants in trial 1) were older at follow-up than children born to HIV-uninfected mothers (participants in trial 2). The prevalence of stunting, thinness, overweight, and obesity at follow-up were 4.4%, 5.3%, 8.0%, and 3.5%, respectively.Table I Characteristics of the 113 children who participated in midchildhood (4-9 years of age) assessments

older at follow-up than children born to HIV-uninfected mothers (participants in trial 2). The prevalence of stunting, thinness, overweight, and obesity at follow-up were 4.4%, 5.3%, 8.0%, and 3.5%, respectively.Table I Characteristics of the 113 children who participated in midchildhood (4-9 years of age) assessments Table IMaternal characteristics Age (y) 28.1 ± 5.0* Height (cm) 157.2 ± 6.0 HIV infected, n (%) 47 (41.6) Prior pregnancies, n (%) 0 31 (27.9) 1-4 74 (66.7) ≥5 6 (5.4) Years of formal education 7.8 ± 2.7 Employment, n (%) Housewife without income 61 (55.0) Housewife with income 39 (35.1) Other 11 (9.9) Household characteristics Household possessions†, n (%) 0 31 (27.9) 1-2 31 (27.9) ≥3 49 (44.1) Child characteristics at baseline (5-7 wk) Age at randomization (wk) 5.8 ± 0.3 Male sex, n (%) 51 (45.1) Birthweight (kg) 3.16 ± 0.46 Gestational age at birth (wk) 40.0 ± 2.3 LAZ −0.18 ± 1.03 WAZ −0.27 ± 0.88 WLZ −0.12 ± 1.10 Flagellin IgG (adjusted optical density)‡ 0.43 ± 0.18 Flagellin IgA (adjusted optical density)‡ 0.30 ± 0.23 LPS IgG (adjusted optical density)‡ 0.72 ± 0.38 LPS IgA (adjusted optical density)‡ 0.41 ± 0.31 Child characteristics during 12-month follow-up in original trials Change in LAZ from baseline to 12 months −0.52 ± 1.08 Change in WAZ from baseline to 12 months −0.35 ± 1.06 Change in WLZ from baseline to 12 months −0.49 ± 1.52 Child characteristics at follow-up in midchildhood (4-9 y) Age (y) 6.8 ± 1.6 HAZ −0.47 ± 1.1 WAZ −0.54 ± 1.0 BMIZ −0.40 ± 1.0 Stunting (HAZ <−2) 5 (4.4%) Overweight (BMIZ >1) 9 (8.0%) Obesity (BMIZ >2) 4 (3.5%) Thinness (BMIZ <−2) 6 (5.3%) Systolic blood pressure (mm Hg) 91.8 ± 9.9 Diastolic blood pressure (mm Hg) 54.1 ± 6.5 Mean arterial blood pressure (mm Hg) 64.8 ± 7.7 * Mean ± SD for all continuous variables; number and frequency (%) for all categorical variables.

) Overweight (BMIZ >1) 9 (8.0%) Obesity (BMIZ >2) 4 (3.5%) Thinness (BMIZ <−2) 6 (5.3%) Systolic blood pressure (mm Hg) 91.8 ± 9.9 Diastolic blood pressure (mm Hg) 54.1 ± 6.5 Mean arterial blood pressure (mm Hg) 64.8 ± 7.7 * Mean ± SD for all continuous variables; number and frequency (%) for all categorical variables. † From a list that includes sofa, television, refrigerator and fan. ‡ Only available in children in trial 2 (n = 66). In multivariate models predicting HAZ in midchildhood, LAZ at age 6 weeks and change in LAZ, WLZ, and WAZ from 6 to 52 weeks of age were all correlated positively with HAZ in midchildhood (Table II). Of note, in the model containing 6-week LAZ and change in LAZ from 6 to 52 weeks of age, both variables were significant, independent predictors of midchildhood HAZ (Beta 0.47 [95% CI 0.25-0.68; P < .001] and 0.40 [95% CI 0.20-0.61; P < .001], respectively). None of the EED biomarkers at 6 weeks or 6 months of age were associated significantly with HAZ or any of the anthropometric indicators in midchildhood. We also did not find that treatment in either trial was associated with HAZ in midchildhood or any of the anthropometric indicators.Table II Linear regression models for HAZ, BMIZ, and WAZ in midchildhood (4-9 years of age)

were associated significantly with HAZ or any of the anthropometric indicators in midchildhood. We also did not find that treatment in either trial was associated with HAZ in midchildhood or any of the anthropometric indicators.Table II Linear regression models for HAZ, BMIZ, and WAZ in midchildhood (4-9 years of age) Table II HAZ BMIZ WAZ Univariate models Multivariate models* Univariate models Multivariate models* Univariate models Multivariate models* Baseline (6-week) measurements LAZ† Beta 0.33 0.47 0.04 0.16 0.25 0.42 95% CI 0.14 to 0.53 0.25 to 0.68 0.14 to 0.23 0.06 to 0.39 0.06 to 0.43 0.22 to 0.62 P value .001 <.001 .65 .15 .009 <.001 WAZ† Beta 0.06 0.12 0.35 0.53 0.28 0.46 95% CI 0.19 to 0.30 0.13 to 0.37 0.14 to 0.56 0.30 to 0.76 0.06 to 0.50 0.24 to 0.67 P value .65 .33 .002 <.001 .01 <.001 WLZ† Beta −0.34 −0.05 0.28 0.64 −0.02 0.43 95% CI 0.52 to −0.16 0.29 to 0.19 0.11 to 0.45 0.42 to 0.87 −0.20 to 0.16 0.21 to 0.65 P value <.001 .11 .001 <.001 .82 <.001 Flagellin IgG (adjusted optical density)‡ Beta −1.32 −1.36 0.36 0.46 −0.57 −0.52 95% CI −3.12 to 0.48 −3.11 to 0.38 −1.15 to 1.87 −1.10 to 2.02 −2.15 to 1.00 −2.09 to 1.04 P value .15 .12 .64 .56 .47 .50 Flagellin IgA (adjusted optical density)‡ Beta −0.02 −0.82 0.24 −0.15 0.19 −0.58 95% CI −1.44 to 1.40 −2.26 to 0.62 −0.93 to 1.42 −1.42 to 1.13 −1.04 to 1.41 −1.86 to 0.69 P value .98 .26 .68 .82 .76 .36 LPS IgG (adjusted optical density)‡ Beta 0.01 −0.44 −0.11 −0.19 −0.08 −0.40 95% CI −0.84 to 0.86 −1.28 to 0.41 −0.81 to 0.60 −0.94 to 0.55 −0.81 to 0.66 −1.14 to 0.34 P value .99 .31 .77 .60 .84 .28 LPS IgA (adjusted optical density)‡ Beta −0.61 −0.93 −0.06 −0.08 −0.41 −0.61 95% CI −1.62 to 0.41 −1.91 to 0.05 −0.91 to 0.79 −0.97 to 0.80 −1.29 to 0.47 −1.48 to 0.27 P value .24 .06 .90 .86 .35 .17 Change in anthropometry from 6 to 52 weeks LAZ† Beta 0.19 0.40 0.13 0.19 0.21 0.40 95% CI −0.01 to 0.38 0.20 to 0.61 −0.05 to 0.30 −0.02 to 0.40 0.03 to 0.39 0.21 to 0.59 P value .06 <.001 .16 .08 .02 <.001 WAZ† Beta 0.28 0.35 0.21 0.32 0.33 0.45 95% CI 0.09 to 0.48 0.14 to 0.56 0.03 to 0.39 0.13 to 0.51 0.16 to 0.51 0.27 to 0.63 P value .004 .001 .02 <.001 <.001 <.001 WLZ† Beta 0.33 0.29 0.07 0.38 0.27 0.45 95% CI 0.21 to 0.46 0.11 to 0.47 −0.06 to 0.20 0.21 to 0.54 0.15 to 0.38 0.29 to 0.62 P value <.001 .002 .28 <.001 <.001 <.001 * Multivariate models for HAZ, BMIZ, and WAZ in mid-childhood adjust for which trial the child originally participated in, treatment arm, maternal height (cm), maternal education (0, 1-7 or

45 95% CI 0.21 to 0.46 0.11 to 0.47 −0.06 to 0.20 0.21 to 0.54 0.15 to 0.38 0.29 to 0.62 P value <.001 .002 .28 <.001 <.001 <.001 * Multivariate models for HAZ, BMIZ, and WAZ in mid-childhood adjust for which trial the child originally participated in, treatment arm, maternal height (cm), maternal education (0, 1-7 or ≥8 years), and child's sex and age at follow-up. Multivariate models for BMIZ adjust for which trial the child originally participated in, treatment arm, maternal height (cm), maternal education (0, 1-7 or ≥8 years), and child's sex and age at follow-up. Multivariate models for WAZ adjust for which trial the child originally participated in, treatment arm, maternal height (cm), maternal education (categories), and child's sex and age at follow-up. † Multivariate models with anthropometric indicators as independent variables also include the corresponding anthropometric indicator at 6 weeks or change from 6 to 52 weeks of age, for example, the model reporting the beta parameter for LAZ at 6 weeks adjusts for change in LAZ from 6 to 52 weeks of age, and vice versa. ‡ n = 66 (children from second trial only).

† Multivariate models with anthropometric indicators as independent variables also include the corresponding anthropometric indicator at 6 weeks or change from 6 to 52 weeks of age, for example, the model reporting the beta parameter for LAZ at 6 weeks adjusts for change in LAZ from 6 to 52 weeks of age, and vice versa. ‡ n = 66 (children from second trial only). In the multivariate model adjusting for sociodemographic characteristics, WLZ at 6 weeks of age and change in WLZ from 6 to 52 weeks of age were associated independently with BMIZ in midchildhood (Beta 0.64 [95% CI 0.42-0.87; P < .001] and 0.38 [95% CI 0.21-0.54; P < .001], respectively). Similarly, WAZ at 6 weeks of age and change from 6 to 52 weeks of age were also associated independently with an increase in BMIZ in midchildhood, although the effect estimates were slightly smaller. In multivariate models, 6-week LAZ was not associated significantly with BMIZ in midchildhood, whereas the change in LAZ from 6 to 52 weeks of age was marginally, significantly correlated with BMIZ at follow-up. In multivariate models for WAZ in midchildhood, baseline LAZ and change in LAZ from 6 to 52 weeks of age were correlated independently positively with WAZ, as were baseline WAZ and change in WAZ, and baseline WLZ and change in WLZ.

om 6 to 52 weeks of age was marginally, significantly correlated with BMIZ at follow-up. In multivariate models for WAZ in midchildhood, baseline LAZ and change in LAZ from 6 to 52 weeks of age were correlated independently positively with WAZ, as were baseline WAZ and change in WAZ, and baseline WLZ and change in WLZ. None of the anthropometric indicators at baseline, nor changes from 6 to 52 weeks of age, were significant predictors of systolic or diastolic blood pressures at follow-up in the univariate or multivariate models (Table III). Each additional unit in BMIZ in midchildhood, however, was associated with an increase in systolic blood pressure of 2.21  mm Hg (95% CI 0.40-4.02; P = .02). We also found that flagellin IgA at 6 weeks was a significant predictor of systolic blood pressure in midchildhood in models adjusted for sociodemographic characteristics (Figure). Flagellin IgA at 6 weeks of age was also associated linearly with diastolic blood pressure in midchildhood in multivariate models. To enhance interpretability, we compared quartiles of flagellin IgA and found that children in the highest quartile of flagellin IgA at 6 weeks of age had a mean systolic blood pressure that was 10.3 mm Hg higher (95% CI 2.49-18.11) than infants in the lowest quartile of flagellin IgA at 6 weeks of age in the multivariate models. In additional analyses, we included baseline values and changes in pediatric anthropometry in multivariate models assessing flagellin IgA at 6 weeks of age as a predictor of systolic and diastolic blood pressure in midchildhood; flagellin IgA at 6 weeks of age remained a significant predictor of both blood pressure measures, even after adjusting for infant anthropometry. None of the EED biomarkers at 6 months of age were significant predictors of either blood pressure variable in multivariate models.Figure Flagellin IgA at 6 weeks and systolic and diastolic blood pressure in midchildhood (4-9 years of age).

of both blood pressure measures, even after adjusting for infant anthropometry. None of the EED biomarkers at 6 months of age were significant predictors of either blood pressure variable in multivariate models.Figure Flagellin IgA at 6 weeks and systolic and diastolic blood pressure in midchildhood (4-9 years of age). FigureTable III Linear regression models for systolic and diastolic blood pressure in midchildhood (4-9 years of age)

of both blood pressure measures, even after adjusting for infant anthropometry. None of the EED biomarkers at 6 months of age were significant predictors of either blood pressure variable in multivariate models.Figure Flagellin IgA at 6 weeks and systolic and diastolic blood pressure in midchildhood (4-9 years of age). FigureTable III Linear regression models for systolic and diastolic blood pressure in midchildhood (4-9 years of age) Table III Systolic blood pressure Diastolic blood pressure Univariate models Multivariate models* Univariate models Multivariate models* Baseline (6-week) measurements LAZ† Beta 0.24 0.44 0.43 0.60 95% CI −1.57 to 2.05 −1.73 to 2.61 −0.75 to 1.62 −0.84 to 2.05 P value .79 .69 .47 .41 WAZ† Beta 0.64 1.52 0.50 0.94 95% CI −1.50 to 2.79 −0.87 to 3.91 −0.90 to 1.89 −0.65 to 2.52 P value .55 .21 .48 .24 WLZ† Beta 0.26 2.04 −0.07 0.85 95% CI −1.46 to 1.97 −0.44 to 4.52 −1.19 to 1.04 −0.80 to 2.50 P value .77 .11 .90 .31 Flagellin IgG (adjusted optical density)‡ Beta −0.97 4.58 3.60 7.36 95% CI −15.23 to 13.30 −12.11 to 21.27 −6.17 to 13.36 −4.30 to 19.01 P value .89 .58 .46 .21 Flagellin IgA (adjusted optical density)‡ Beta 17.67 17.81 12.08 13.95 95% CI 7.51 to 27.82 6.50 to 29.1 5.10 to 19.07 6.14 to 21.76 P value <.001 .003 .001 .001 LPS IgG (adjusted optical density)‡ Beta −2.23 −0.58 1.19 2.29 95% CI −8.84 to 4.38 −8.19 to 7.02 −3.36 to 5.74 −3.05 to 7.63 P value .50 .88 .60 .39 LPS IgA (adjusted optical density)‡ Beta −2.21 −1.71 −1.34 −1.39 95% CI −10.19 to 5.78 −10.67 to 7.25 −6.83 to 4.15 −7.72 to 4.94 P value .58 .70 .63 .66 Change in anthropometry from 6 to 52 weeks LAZ† Beta 0.58 0.80 −0.04 0.40 95% CI −1.15 to 2.31 −1.41 to 3.01 −1.18 to 1.10 −01.07 to 1.87 P value .50 .47 .95 .59 WAZ† Beta 0.62 1.10 0.17 0.55 95% CI −1.16 to 2.39 −0.92 to 3.13 −0.99 to 1.33 −0.79 to 1.89 P value .49 .28 .78 .24 WLZ† Beta 0.39 1.25 0.28 0.70 95% CI −0.97 to 1.76 −0.48 to 2.97 −0.52 to 1.09 −0.45 to 1.85 P value .57 .16 .49 .23 Anthropometry in midchildhood (4-9 years) HAZ Beta −0.21 −0.06 −0.51 −0.39 95% CI −1.86 to 1.44 −1.84 to 1.71 −1.59 to 0.58 −1.57 to 0.79 P value .80 .94 .36 .51 BMIZ Beta 1.54 2.21 0.42 0.85 95% CI −0.26 to 3.35 0.40 to 4.02 −0.78 to 1.62 −0.39 to 2.08 P value .09 .02 .49 .18 * Multivariate models for systolic and diastolic blood pressure control for which trial the child participated in, treatment arm, child's sex, gestational age at birth, and age at mid-childhood follow-up.

BMIZ Beta 1.54 2.21 0.42 0.85 95% CI −0.26 to 3.35 0.40 to 4.02 −0.78 to 1.62 −0.39 to 2.08 P value .09 .02 .49 .18 * Multivariate models for systolic and diastolic blood pressure control for which trial the child participated in, treatment arm, child's sex, gestational age at birth, and age at mid-childhood follow-up. Multivariate models diastolic blood pressure adjust for which trial the child originally participated in, whether the child was randomized to multivitamin supplementation, maternal height (cm), child's gestational age at birth, and child's age at follow-up visit. † Multivariate models with anthropometric indicators as predictors also include the corresponding anthropometric indicator at 6 weeks of age or change from 6 to 52 weeks of age, for example, the model reporting the beta parameter for LAZ at 6 weeks adjusts for change in LAZ from 6 to 52 weeks of age, and vice versa. ‡ n = 66 (children from second trial only).

† Multivariate models with anthropometric indicators as predictors also include the corresponding anthropometric indicator at 6 weeks of age or change from 6 to 52 weeks of age, for example, the model reporting the beta parameter for LAZ at 6 weeks adjusts for change in LAZ from 6 to 52 weeks of age, and vice versa. ‡ n = 66 (children from second trial only). In multivariate models for binary outcomes in midchildhood, we found that both baseline WLZ and change in WLZ from 6 to 52 weeks of age, as well as baseline WAZ and change in WAZ, were associated independently and positively with the risk of being overweight in midchildhood (Table IV; available at www.jpeds.com). Of note, each unit increase in baseline WLZ was associated with a 2.5-fold increase in risk of being overweight in midchildhood (aRR 2.47; 95% CI 1.04-5.90) after adjusting for change in WLZ from 6 to 52 weeks of age, child's sex, and original trial, whereas each unit increase in WLZ from 6 to 52 weeks of age was associated with an increased risk of 2.45 (95% CI 1.40-4.30) in the same model. Change in LAZ from 6 to 52 weeks of age was also associated with an increased risk of overweight in multivariate models. In multivariate models for obesity in midchildhood, baseline WAZ and change in all 3 anthropometric indicators from 6 to 52 weeks of age were associated significantly with an increased risk of obesity. In multivariate models for stunting in midchildhood, baseline LAZ and change in LAZ from 6 to 52 weeks of age were associated independently and negatively with risk of stunting (aRR 0.23 [95% CI 0.08-0.68; P = .008] and 0.31 [95% CI 0.13-0.77; P = .01], respectively). Change in WAZ and WLZ from 6 to 52 weeks of age were also associated marginally and significantly with a decreased risk of stunting in midchildhood.

weeks of age were associated independently and negatively with risk of stunting (aRR 0.23 [95% CI 0.08-0.68; P = .008] and 0.31 [95% CI 0.13-0.77; P = .01], respectively). Change in WAZ and WLZ from 6 to 52 weeks of age were also associated marginally and significantly with a decreased risk of stunting in midchildhood. Discussion In this longitudinal study of 113 Tanzanian children, we found that LAZ at age 6 weeks of age and change in LAZ in the first year of life were both independent, positive predictors of HAZ and independent negative predictors of stunting in midchildhood. We also found that the 6-week WAZ and WLZ, as well as changes in WAZ and WAZ in the first year of life, were associated independently and positively with BMIZ in midchildhood, and associated negatively with risk of overweight and obesity. Although linear growth in infancy was also associated with BMIZ, overweight, and obesity, this relationship was less robust than that with weight-related variables; similarly, weight-related indicators in infancy were less strongly correlated with height variables in midchildhood. BMIZ in midchildhood was associated in turn significantly and positively with systolic blood pressure, and serum anti-flagellin IgA in early infancy was also a significant predictor of midchildhood blood pressure.

weight-related indicators in infancy were less strongly correlated with height variables in midchildhood. BMIZ in midchildhood was associated in turn significantly and positively with systolic blood pressure, and serum anti-flagellin IgA in early infancy was also a significant predictor of midchildhood blood pressure. Cohort studies in high-income countries have also found that birthweight and weight gain in early life are positively correlated with BMI, overweight, and obesity later in life26, 27; however, our study is one of few to longitudinally follow-up with infants from a low-income country over many years,28, 29, 30, 31, 32 particularly in sub-Saharan Africa.28 Our findings show that in the resource-limited environment of Dar es Salaam, where poor sanitation, high burdens of infectious disease, and suboptimal infant feeding practices contribute to high levels of undernutrition in infancy,18, 19, 20, 21 weight gain in early life continues to track throughout childhood and ultimately correlates with BMIZ and risk of overweight and obesity in midchildhood.

ere poor sanitation, high burdens of infectious disease, and suboptimal infant feeding practices contribute to high levels of undernutrition in infancy,18, 19, 20, 21 weight gain in early life continues to track throughout childhood and ultimately correlates with BMIZ and risk of overweight and obesity in midchildhood. Stunting in early life increases the risk of impaired cognitive development, adult short stature, lower earning potential, and birth complications in women.3 We found that LAZ at 6 weeks of age and change in LAZ from 6 to 52 weeks of age were associated independently with height in midchildhood, and that they were also correlated independently and inversely with stunting in midchildhood. Our findings thus emphasize the importance of nutritional status in both the perinatal and infant periods to improve long-term nutrition and health outcomes in children. Other studies have also found that early life linear growth predicts attained height later in childhood or adulthood29, 30, 31, 32, 33, 34, 35; however, less clear in the existing literature is the relationship between early life linear growth and later BMI. A recent study pooling data from 5 cohorts in low-income and middle-income countries found that, although change in height in the first 2 years of life was associated with a minor increase in risk of being overweight in adulthood, increases in weight accounting for height in the first 2 years of life was a much stronger predictor of overweight in adulthood.28 Our findings support the hypothesis that early life gains in weight, particularly weight relative to height, has negative consequences for increasing BMI and risk of overweight and obesity. We also found that linear growth in infancy was associated with an increased risk of overweight and obesity in midchildhood, though this association was less robust than that of weight-related indicators and midchildhood BMIZ, overweight, and obesity. Further research on how interventions can promote healthy linear growth without resulting in excess weight gain in early infancy will be essential in addressing the global dual burden of disease.

ough this association was less robust than that of weight-related indicators and midchildhood BMIZ, overweight, and obesity. Further research on how interventions can promote healthy linear growth without resulting in excess weight gain in early infancy will be essential in addressing the global dual burden of disease. This study identified a strong relationship between flagellin antibodies in early infancy and blood pressure in midchildhood. The relationship between morbidity, particularly diarrheal disease in infancy and malnutrition has been well-documented1, 36; however, researchers have only recently begun to explore the important role of EED in malnutrition, highlighting that EED may lead to both micronutrient and macronutrient malabsorption.37, 38, 39 Currently, diagnosing EED remains challenging. Small bowel biopsies have been considered the gold standard for the assessment of the mucosal structure; however, given the invasive and expensive nature of these procedures, research is underway to determine biomarkers that may be more appropriate for widespread use in community-based studies.40, 41 The bacterial protein flagellin mediates bacteria motility and is present in the gut lumen, even in a healthy state; however, flagellin has very limited access to cross the epithelium and reach the mucosal immune system.

ermine biomarkers that may be more appropriate for widespread use in community-based studies.40, 41 The bacterial protein flagellin mediates bacteria motility and is present in the gut lumen, even in a healthy state; however, flagellin has very limited access to cross the epithelium and reach the mucosal immune system. However, a decrease in gut barrier function or an increase in levels of bacteria that produce flagellin might activate the adaptive immune responses to these molecules, resulting in the generation of anti-flagellin and anti-LPS antibodies, as has been shown to occur in patients with short bowel syndrome42 and Crohn's disease.43 Our group has previously shown that anti-flagellin IgA and IgG concentrations increased over the first year of life in Tanzanian infants, that the concentrations in Tanzanian infants were much higher than in healthy controls in Boston, and that elevated anti-flagellin IgA and IgG concentrations were associated with an increased risk of underweight in infancy.17 Our current study also highlights that, in addition to its role in malnutrition in infancy, EED, as measured by antibodies to bacterial components, may also play an important role in long-term chronic disease risk,16 particularly when one considers the research documenting the likelihood of blood pressure tracking from childhood through adulthood.44 Of note, in our sensitivity analyses, we added infant anthropometry to the multivariate models linking 6-week flagellin IgA and midchildhood systolic and diastolic blood pressure, and found that this did not reduce substantially the magnitude of the relationship between flagellin IgA in infancy and midchildhood blood pressure. Our findings, thus, support the hypothesis that EED in infancy may affect long-term cardiometabolic outcomes through pathways external to growth, such systemic inflammation and epigenetic changes in immune function.16

ly the magnitude of the relationship between flagellin IgA in infancy and midchildhood blood pressure. Our findings, thus, support the hypothesis that EED in infancy may affect long-term cardiometabolic outcomes through pathways external to growth, such systemic inflammation and epigenetic changes in immune function.16 Our study has several limitations. We do not have measurements in children between infancy and the midchildhood follow-up, and we only have biomarkers of EED at 6 weeks and 6 months of age, so we are limited in our ability to compare critical periods of growth or to evaluate the potential impact of EED on growth at other ages. It is also worth noting that we only measured a limited set of cardiometabolic risk factors that did not include insulin sensitivity or other vascular outcomes, such as vascular reactivity. Our follow-up study sample may also suffer from selection bias. In particular, the least well-off children, that is, those who became ill, died, or moved since the trial ended, are likely to be excluded from our follow-up sample. However, given that developmental programming is strongest when environmental insults are most severe, if our study had included less healthy children (who likely would have experienced greater growth faltering and EED in infancy), we would have expected to see even stronger associations between infant and midchildhood health indicators. By contrast, our study is also limited by the low prevalence of overweight, obesity, and stunting in this peri-urban African setting and, thus, the limited power for analyses of binary outcomes. Our estimates of overweight and obesity in this sample were, however, comparable with other studies in Tanzanian children.45, 46 Although overweight and obesity are relatively rare in Tanzanian children, a recent study found that 24.1% and 19.2% of adults in Kinondoni, 1 of 3 municipalities in Dar es Salaam, are overweight and obese, respectively.47 We showed that weight gain in early life tracks into midchildhood, and evidence from other contexts indicates that weight gain in midchildhood tracks into adulthood.48 As sub-Saharan Africa continues to undergo the nutrition transition, it will be increasingly important to monitor weight gain in infancy and as children age.

owed that weight gain in early life tracks into midchildhood, and evidence from other contexts indicates that weight gain in midchildhood tracks into adulthood.48 As sub-Saharan Africa continues to undergo the nutrition transition, it will be increasingly important to monitor weight gain in infancy and as children age. Our study, and others, longitudinally tracked pediatric growth from infancy into midchildhood in a sub-Saharan African context.49, 50 Given the rapid pace of the nutrition transition in sub-Saharan Africa and the dual burden of preventing both pediatric undernutrition but also chronic disease later in life,1 our study helps to elucidate the relationship between standard growth indicators in early infancy and health outcomes in later life, as well as the possible role of EED in child health. As the global health community continues to invest in the healthy growth and development of children in sub-Saharan Africa, an expanded emphasis beyond short-term outcomes such a survival and treatment of malnutrition and infectious disease will be essential to promote long-term health and prevent obesity and cardiometabolic diseases. Appendix Table IV Relative risks of overweight, obesity, and stunting in midchildhood (4-9 years)

Our study, and others, longitudinally tracked pediatric growth from infancy into midchildhood in a sub-Saharan African context.49, 50 Given the rapid pace of the nutrition transition in sub-Saharan Africa and the dual burden of preventing both pediatric undernutrition but also chronic disease later in life,1 our study helps to elucidate the relationship between standard growth indicators in early infancy and health outcomes in later life, as well as the possible role of EED in child health. As the global health community continues to invest in the healthy growth and development of children in sub-Saharan Africa, an expanded emphasis beyond short-term outcomes such a survival and treatment of malnutrition and infectious disease will be essential to promote long-term health and prevent obesity and cardiometabolic diseases. Appendix Table IV Relative risks of overweight, obesity, and stunting in midchildhood (4-9 years) Table IV Overweight (BMIZ > 1; N = 113, overweight n = 9) Obesity (BMIZ > 2; N = 113, obese n = 4) Stunting (HAZ < −2; N = 113, stunted n = 6) Univariate models Multivariate models* Univariate models Multivariate models* Univariate models Multivariate models*,† Baseline (6-week) measurements LAZ Risk ratio 1.24 1.94 1.16 2.46 0.25 0.23 95% CI 0.69-2.22 0.96-3.93 0.46-2.90 0.74-8.20 0.11-0.56 0.08-0.68 P value .48 .07 .76 .14 <.001 .008 WAZ Risk ratio 1.46 2.16 1.25 3.33 0.54 0.26 95% CI 0.70-3.01 1.15-4.05 0.41-3.85 1.41-7.87 0.26-1.12 0.04-1.61 P value .31 .02 .69 .006 .10 .15 WLZ Risk ratio 1.10 2.47 1.03 3.85 2.38 0.48 95% CI 0.63-1.92 1.04-5.90 0.43-2.45 0.74-20.17 0.98-5.78 0.11-2.04 P value .74 .04 .95 .11 .06 .32 Change in anthropometry from 6 to 52 weeks LAZ Risk ratio 1.60 2.37 2.03 4.03 0.68 0.31 95% CI 0.99-2.58 1.21-4.63 1.04-3.96 1.32-12.33 0.34-1.36 0.13-0.77 P value .053 .01 .04 .01 .28 .01 WAZ Risk ratio 1.67 3.13 2.24 7.54 0.32 0.22 95% CI 1.18-2.36 1.50-6.52 1.40-3.58 3.12-18.22 0.12-0.86 0.04-1.13 P value .004 .002 <.001 <.001 .02 .07 WLZ Risk ratio 1.52 2.45 2.10 4.66 0.40 0.27 95% CI 1.04-2.23 1.40-4.30 1.19-3.70 1.59-13.67 0.23-0.70 0.06-1.20 P value .03 .002 .01 .005 .02 .08 * Multivariate models adjust for which trial the child originally participated in and child's sex as well as the corresponding anthropometric indicator at 6 weeks or change from 6 to 52 weeks of age, for example, the model reporting the beta parameter for LAZ at 6 weeks of age adjusts for change in LAZ from 6 to 52 weeks of age, and vice versa.

st for which trial the child originally participated in and child's sex as well as the corresponding anthropometric indicator at 6 weeks or change from 6 to 52 weeks of age, for example, the model reporting the beta parameter for LAZ at 6 weeks of age adjusts for change in LAZ from 6 to 52 weeks of age, and vice versa. † Multivariate model for stunting in midchildhood also adjusts for maternal height (cm). We thank the mothers and children, and field teams, including physicians, nurses, midwives, supervisors, laboratory staff, and the administrative staff, who made these studies possible; and Muhimbili National Hospital and Muhimbili University of Health and Allied Sciences in Dar es Salaam for their institutional support. We also thank Ellen Hertzmark for her expert advice. Supported by the National Institutes of Health (R01 HD043688-01, R01 HD048969-01, K24HD058795, K24 DK104676, and 2P30 DK040561), Bill and Melinda Gates Foundation (OPP1066203), and the Feed the Future Food Innovation Lab for Nutrition which is funded by the United States Agency for International Development (USAID; AID-OAA-L-10-00006). C.D. serves as an Editorial Board member for The Journal of Pediatrics. The other authors declare no conflicts of interest.

trients (n = 568)† Plasma zinc, µg/dL 54.2 (14.2) 52.2 (16.6) 55.2 (13.8) 54.4 (13.0) 55.1 (13.5) Zinc deficiency, n (%) 432 (75.4) 110 (74.3) 101 (72.1) 108 (74.5) 113 (80.7) Ferritin, µg/L 31.0 (32.6) 30.3 (32.7) 27.7 (26.6) 27.1 (21.6) 27.2 (30.5) Storage ID, n (%) 150 (26.1) 36 (24.3) 39 (27.9) 37 (25.3) 38 (27.1) sTfR, mg/L 9.4 (7.2) 9.1 (7.6) 9.6 (6.0) 9.3 (6.2) 9.4 (7.6) Functional ID, n (%) 379 (66.0) 91 (61.5) 97 (69.3) 99 (67.8) 92 (65.7) RBP, µmol/L 1.3 (0.4) 1.1 (0.4) 1.1 (0.4) 1.3 (0.3) 1.2 (0.3) VAD, n (%) 9 (1.6) 3 (2.0) 4 (2.7) 1 (0.7) 1 (0.7) Inflammation (n = 568)† AGP, g/L 0.61 (0.42) 0.64 (0.44) 0.63 (0.47) 0.60 (0.41) 0.60 (0.41) AGP>1 g/L, n (%) 119 (20.7) 34 (23.0) 31 (22.1) 26 (17.8) 28 (20.0) CRP, mg/L 0.47 (1.60) 0.57 (1.55) 0.41 (1.19) 0.43 (1.47) 0.46 (2.31) CRP >5 mg/L, n (%) 67 (11.7) 18 (12.2) 13 (9.3) 16 (11.0) 20 (14.3) BMI, body mass index; HFIAS, Household Food Insecurity and Access Score. Zinc deficiency defined as zinc <65 µg/dL, Storage ID as ferritin <12 µg/L, Functional ID as sTfR >8.3 mg/L, and vitamin A deficiency defined as RBP <0.81 µmol/L. * Values represent mean ± SD for continuous variable or n (%) for dichotomized variables. † Values represent median (IQR) for continuous variable or n (%) for dichotomized variables.

Growth faltering and anemia are common in low- and middle-income countries, likely due in part to the coexistence of multiple micronutrient deficiencies, including zinc and iron deficiencies.1 Zinc is involved in DNA and RNA metabolism2 and hence modulates cell replication, differentiation, and growth. As a result, zinc deficiency is associated with a wide spectrum of adverse health events, including linear growth restriction.1 Several meta-analyses have concluded that preventive zinc supplementation has a small, positive impact on linear growth.3, 4, 5, 6 A common finding of these meta-analyses was the presence of a high degree of heterogeneity across the studies, possibly related to setting, age, and other baseline characteristics of study participants. In addition, the evidence suggests that the effects of zinc on growth outcomes differed by whether zinc was delivered alone or together with other micronutrients, particularly iron.4, 5 Thus, additional information is needed regarding the specific contexts in which preventive zinc supplementation is most beneficial.

s. In addition, the evidence suggests that the effects of zinc on growth outcomes differed by whether zinc was delivered alone or together with other micronutrients, particularly iron.4, 5 Thus, additional information is needed regarding the specific contexts in which preventive zinc supplementation is most beneficial. Often, zinc deficiency does not occur in isolation. Supplementation with a multiple micronutrient powder offers a low-cost option to prevent multiple micronutrient deficiencies simultaneously. The World Health Organization (WHO) and United Nations Children's Fund currently promote a micronutrient powder formulation containing 12.5 mg of iron, 4.1-5 mg of zinc, and 13 other micronutrients.7 Despite the proven efficacy of micronutrient powder to prevent iron deficiency and anemia,8 the available evidence suggests a lack of effect on growth outcomes.8, 9 In addition, there are unresolved safety concerns regarding a daily iron dose of 12.5 mg.10, 11 To address these concerns, we used a new micronutrient powder formulation in the present study that contained a lower amount of iron (6 mg/d) and a greater amount of zinc (10 mg/d) than current formulations, along with standard quantities of the 13 other micronutrients.

rns regarding a daily iron dose of 12.5 mg.10, 11 To address these concerns, we used a new micronutrient powder formulation in the present study that contained a lower amount of iron (6 mg/d) and a greater amount of zinc (10 mg/d) than current formulations, along with standard quantities of the 13 other micronutrients. Therapeutic zinc supplementation (20 mg per day for 10-14 days) during episodes of diarrhea is recommended by the WHO and the United Nations Children's Fund along with oral rehydration salts (ORS) to reduce disease severity.12 In some settings, therapeutic zinc supplementation for diarrhea has been associated with a reduced incidence of diarrhea and enhanced growth during the 2-3 months after treatment initiation.13 However, evidence from other studies indicates that these benefits may accrue only during the period of treatment.14 Furthermore, because this strategy requires appropriate recognition of diarrhea and motivation to seek treatment, as well as access to a healthcare facility or pharmacy to obtain the supplements, coverage of therapeutic zinc supplementation programs is often low, typically reaching <30% of children in need.15, 16 Moreover, children have access to therapeutic zinc supplementation only after an episode of diarrhea occurs, so zinc prevention of the first episode would not be possible. Thus, the long-term health benefit of therapeutic zinc supplementation programs is uncertain, and more evidence is needed to compare the relative benefits of preventive vs therapeutic zinc supplementation.

mentation only after an episode of diarrhea occurs, so zinc prevention of the first episode would not be possible. Thus, the long-term health benefit of therapeutic zinc supplementation programs is uncertain, and more evidence is needed to compare the relative benefits of preventive vs therapeutic zinc supplementation. The goal of the present study was to compare the effects of preventive zinc supplementation, provided either as a single micronutrient or as part of new micronutrient powder formulation containing more zinc (10 mg) and less iron (6 mg) than current formulations and therapeutic zinc supplementation, on physical growth, anemia, and biomarkers of zinc and iron status in rural Laotian children. Methods Ethical Approval This trial and the consent procedure were approved by the National Ethics Committee for Health Research, Ministry of Health, Lao PDR, and the institutional review board of the University of California, Davis, California. The trial is registered as the Lao Zinc Study, NCT02428647 (https://clinicaltrials.gov).

l Approval This trial and the consent procedure were approved by the National Ethics Committee for Health Research, Ministry of Health, Lao PDR, and the institutional review board of the University of California, Davis, California. The trial is registered as the Lao Zinc Study, NCT02428647 (https://clinicaltrials.gov). Study Design and Participants In this double-blind, placebo controlled trial, children 6-23 months at enrollment were randomized individually to 1 of 4 interventions and followed for ~36 weeks to assess responses in physical growth, anemia, and micronutrient status.17 The study was implemented from September 2015 through April 2017 in rural communities in Khammouane Province, central Lao PDR. The province was selected because of a high prevalence of stunting among children <5 years,18 the likely high prevalence of zinc deficiency, and the lack of current programs designed to reduce the risk of micronutrient deficiencies or treat diarrhea with therapeutic zinc supplements. A 2015 pilot survey (n = 111) conducted in the province found that >60% of children 6-23 months of age were zinc deficient, based on plasma zinc concentrations <65 µg/dL. The study area (~5300 km2) included 300 rural villages from 5 districts (Nongbok, Xebangfai, Mahaxay, Xaibuathong, and Yommalat). All villages in these districts were invited for enrollment, except in Nongbok. Because the prevalence of stunting in Nongbok was lower than in the other districts, only villages belonging to the catchment area of health centers with a mean stunting prevalence ≥25% were invited to participate.

ibuathong, and Yommalat). All villages in these districts were invited for enrollment, except in Nongbok. Because the prevalence of stunting in Nongbok was lower than in the other districts, only villages belonging to the catchment area of health centers with a mean stunting prevalence ≥25% were invited to participate. Sample Size Considerations To detect an effect size of 0.2 in the comparison of length and weight between any 2 groups, with 90% power and 5% type 1 error rate, a total sample size of 710 children per group was required. An effect size of 0.2 corresponds to a length change of ~1.0 cm, and a weight change of ~0.26 kg, and this effect size was based on the results of previous meta-analyses, which reported mean effect sizes for linear growth and weight gain ranging from 0.12 to 0.35.5 Allowing for 15% attrition, a total sample of 835 children (rounded up to 850) per group was determined, and a total of 3400 children were targeted for enrollment. To detect an effect size of 0.5 in the comparison of biochemical indicators (plasma zinc, ferritin, soluble transferrin receptor [sTfR], retinol binding protein [RBP], C-reactive protein [CRP] and alpha-1-glycoacid protein [AGP], and hemoglobin) between any 2 of the 4 groups, with 90% power and 5% type 1 error rate, a total sample size of 560 children, 115 (or 140 assuming 20% attrition) per group, was required. An effect size of 0.5 for the biochemical indicators corresponds to a difference of 0.5 g/L in hemoglobin, for example. The effect size of 0.5 was based on the results of a recent meta-analysis,5 which reported an overall effect size of 0.6 (CI 0.44-0.77), regarding the effect of zinc supplementation on serum zinc concentrations. To ensure that we would have 140 children per group with baseline and endline biomarker data, blood collection was attempted for the first 760 children enrolled, assuming a combined attrition rate, from losses to follow-up and blood draw failures, of ~30%.

he effect of zinc supplementation on serum zinc concentrations. To ensure that we would have 140 children per group with baseline and endline biomarker data, blood collection was attempted for the first 760 children enrolled, assuming a combined attrition rate, from losses to follow-up and blood draw failures, of ~30%. Inclusion and Exclusion Criteria Children were considered eligible to participate in the study if they were 6-23 months of age at enrollment, their families intended to stay in the study area for the duration of the study, were willing to accept home visits, and at least 1 of the caregivers (mother, father, legal guardian) provided written informed consent. Children were excluded from participation if they demonstrated any of the following health conditions: severe anemia (hemoglobin <70 g/L), severe wasting (defined as weight-for-height z score <–3 with respect to WHO 2006 standards),19 bipedal edema, severe illness warranting hospital referral, congenital abnormalities that may interfere with growth, chronic medical conditions requiring frequent medical attention, known HIV infection of the index child or the child's mother, ongoing use of micronutrient supplements, or current participation in another research study.

vere illness warranting hospital referral, congenital abnormalities that may interfere with growth, chronic medical conditions requiring frequent medical attention, known HIV infection of the index child or the child's mother, ongoing use of micronutrient supplements, or current participation in another research study. Randomization A statistician at University of California Davis randomly assigned the study ID numbers to the 4 study arms, using a block randomization scheme with block lengths of 4 or 8. In the event that multiple siblings in the target age range resided in the same household, only the youngest was enrolled. In the case of twins, both twins were assigned to the same group and received all study-related interventions and follow-up, but only one was selected randomly for inclusion in the data analyses.

. In the event that multiple siblings in the target age range resided in the same household, only the youngest was enrolled. In the case of twins, both twins were assigned to the same group and received all study-related interventions and follow-up, but only one was selected randomly for inclusion in the data analyses. Study Interventions and Follow-Up Children were individually randomized into 1 of 4 groups (Table I; available at www.jpeds.com): (1) the preventive zinc supplementation group, who received a daily preventive zinc supplement tablet containing 7 mg of zinc and placebo therapeutic tablets for diarrhea; (2) the micronutrient powder group, who received a daily preventive micronutrient powder containing 10 mg of zinc, 6 mg of iron and 13 other micronutrients and placebo therapeutic tablets for diarrhea; (3) the therapeutic zinc supplementation group, who received a daily placebo preventive supplement tablet and therapeutic zinc tablets containing 20 mg for 10 days for diarrhea treatment; or (4) the placebo control group, who received daily placebo preventive powder and placebo therapeutic tablets for diarrhea.

arrhea; (3) the therapeutic zinc supplementation group, who received a daily placebo preventive supplement tablet and therapeutic zinc tablets containing 20 mg for 10 days for diarrhea treatment; or (4) the placebo control group, who received daily placebo preventive powder and placebo therapeutic tablets for diarrhea. The preventive and therapeutic zinc and placebo tablets were produced by Nutriset SAS (Malaunay, France). The preventive tablets were distributed in blister packages of 8 tablets and the therapeutic tablets in blister packages containing 10 tablets per strip. The micronutrient powder and placebo powder sachets were produced by DSM Fortitech Asia Pacific (Banting, Malaysia). Standard quality control procedures were used by the manufactures to confirm the nutrient concentrations. The supplements were prelabeled (by the manufacturer) with 4 different numerical codes. In the field, the 4 numerical codes were assigned specific colors (one color per intervention group) to ensure correct delivery to children in the respective study groups. Thus, each color represented one type of preventive intervention (to be taken daily) and the corresponding therapeutic tablets for diarrhea management. In addition, caretakers for each child, irrespective of the study group, were given low-osmolarity ORS sachets for diarrhea management. Supplements and ORS were replenished during weekly home visits, as needed. Caregivers were instructed to begin diarrhea treatment with the therapeutic zinc (or placebo) tablets whenever a child had 3 or more liquid stools per day and to continue the treatment for 10 days, until the blister pack was empty. In addition, caregivers also were instructed to give ORS on any day that the child had 3 or more liquid stools per day.

egin diarrhea treatment with the therapeutic zinc (or placebo) tablets whenever a child had 3 or more liquid stools per day and to continue the treatment for 10 days, until the blister pack was empty. In addition, caregivers also were instructed to give ORS on any day that the child had 3 or more liquid stools per day. At enrollment, caregivers were instructed on how to administer the study products to their child, and field workers repeated these instructions every month during one of the weekly home visits. For both the preventive and therapeutic dispersible zinc and placebo tablets, caregivers were instructed to dissolve 1 tablet with clean water or breastmilk and feed the resulting suspension to the child at least 30 minutes before or after a meal. Caregivers were instructed to mix the entire contents of a micronutrient powder or placebo powder sachet into a small amount of semisolid food that the child could easily consume. Each child was visited weekly for 36 weeks (to replenish supplements and to assess health outcomes), unless lost to follow-up. During the weekly visits, caregivers were interviewed regarding consumption of the intervention products. In addition, used and unused blister packages and sachets were collected to assess adherence.

hild was visited weekly for 36 weeks (to replenish supplements and to assess health outcomes), unless lost to follow-up. During the weekly visits, caregivers were interviewed regarding consumption of the intervention products. In addition, used and unused blister packages and sachets were collected to assess adherence. Study Procedures Pre-Enrollment Orientation and Screening On the day of enrollment, caregivers attended an information session, first in groups and subsequently on individual basis, during which details of the trial, including the overall goals of the study, the study interventions, duration, voluntary participation, and inclusion and exclusion criteria, were explained. After the information session, caregivers who showed interest to participate in the study were asked to sign (or fingerprint) the consent statement in the presence of an independent witness. Children with written, informed parental consent were subsequently screened for eligibility.

exclusion criteria, were explained. After the information session, caregivers who showed interest to participate in the study were asked to sign (or fingerprint) the consent statement in the presence of an independent witness. Children with written, informed parental consent were subsequently screened for eligibility. Data Collection All data were recorded electronically via a customized CommCare-HQ (Dimagi, Boston, Massachusetts) application deployed on portable Samsung tablets (Samsung Galaxy, Tab 4; Samsung Group, Seoul, South Korea). At baseline, duplicate anthropometric assessments were recorded by trained anthropometric teams, using standardized procedures. Measurements included weight to the nearest 0.02 kg (383 balance; SECA, Hamburg, Germany), recumbent length to the nearest 0.1 cm (416 length board; SECA), and mid-upper arm circumference (MUAC) to the nearest 0.1 cm (Shorr-Tape Measuring Tape; Weigh and Measure, Olney, Maryland). In the event that the duplicate measurements differed by >0.1 kg for weight, or by >0.5 cm for recumbent length and MUAC, a third independent measurement was taken.

cm (416 length board; SECA), and mid-upper arm circumference (MUAC) to the nearest 0.1 cm (Shorr-Tape Measuring Tape; Weigh and Measure, Olney, Maryland). In the event that the duplicate measurements differed by >0.1 kg for weight, or by >0.5 cm for recumbent length and MUAC, a third independent measurement was taken. Means were computed using the 2 measures with the lowest absolute differences. Children who were severely wasted (WLZ <–3) at baseline were excluded from participation and referred to the nearest health center or hospital. The anthropometric assessments were repeated after 18 weeks and at endline (32-40 weeks). At baseline, maternal weights were measured to 0.05-kg precision (SECA 874) and maternal heights to 0.1 cm precision (SECA 213). Anthropometric teams were systematically standardized.20 During a total of 4 standardization sessions, the mean technical error of measurement and coefficient of reliability for length were 0.38 cm and 97%, respectively.20

ere measured to 0.05-kg precision (SECA 874) and maternal heights to 0.1 cm precision (SECA 213). Anthropometric teams were systematically standardized.20 During a total of 4 standardization sessions, the mean technical error of measurement and coefficient of reliability for length were 0.38 cm and 97%, respectively.20 Anemia status, based on capillary hemoglobin concentrations, was assessed for all children at baseline and endline using a Hemocue Hb 301 System (Hemocue AB, Angelholm, Sweden). The Hemocue devices was checked weekly against a commercial quality control sample (Eurotrol, Inc, Elizabethtown, Kentucky). For the purpose of evaluating biomarkers of nutritional status and health, ~7 mL samples of venous blood were collected by trained nurses from a subsample of 760 children, using evacuated, trace element-free, 7.5 mL of lithium–heparin tubes (Sarstedt AG & Co, Numbrecht, Germany). The heparinized samples were maintained at 4-8°C in portable cooler boxes until transported to field laboratories for plasma processing within ~3 hours. In the field laboratory, the blood samples were centrifuged (PowerSpin Centrifuge Model LX C856; United Products & Instruments, Inc, Dayton, New Jersey) at 1097g (3100 rpm) for 10 minutes and the plasma was aliquoted into clear or amber microcentrifuge tubes (0.2-1.5 mL per tube), depending on the preplanned analyses. Plasma samples intended for nutritional biomarker assays were stored in the field laboratories at –20°C and later shipped on dry ice to a permanent laboratory at the University of California, Davis, California, where the samples were transferred to another –20°C freezer.

per tube), depending on the preplanned analyses. Plasma samples intended for nutritional biomarker assays were stored in the field laboratories at –20°C and later shipped on dry ice to a permanent laboratory at the University of California, Davis, California, where the samples were transferred to another –20°C freezer. Laboratory Analyses Plasma zinc was analyzed by inductively coupled plasma optical emission spectrophotometry (5100 ICP-OES SVDV, Agilent, Santa Clara, California) at the Children's Hospital of Oakland Research Institute (Oakland, California). Details of the plasma zinc assay by inductively coupled plasma optical emission spectrophotometry have been described previously.21 To summarize, after overnight digestion in trace element grade 70% nitric acid at 60°C, the plasma–nitric acid mixture was diluted to a final concentration of 5.5% nitric acid and then centrifuged for 10 minutes at 3000g. Plasma samples were analyzed in duplicate in the same run. In the event that concentrations of the duplicates differed by 10%, the analyses were repeated. Each batch of samples was analyzed along with a reference sample (Seronorm Trace Element Serum L-1 and L-2; Accurate Chemical and Scientific Corp, Westbury, New York). Iron status markers (ferritin and sTfR), RBP, and inflammatory markers (CRP and AGP) were measured using a sandwiched enzyme-linked immunosorbent assay technique at the VitMin Lab (Willstaett, Germany).22

rence sample (Seronorm Trace Element Serum L-1 and L-2; Accurate Chemical and Scientific Corp, Westbury, New York). Iron status markers (ferritin and sTfR), RBP, and inflammatory markers (CRP and AGP) were measured using a sandwiched enzyme-linked immunosorbent assay technique at the VitMin Lab (Willstaett, Germany).22 Data Entry and Analyses A statistical analysis plan was developed and published before analyses23 and was strictly followed to minimize bias. The groups' identities were revealed only after the data analyses were completed and the study investigators reached consensus on the interpretation of results. All analyses were performed with STATA statistical software, release 13 (StataCorp, Austin, Texas) and SAS version 9.4 (SAS Institute, Cary, North Carolina).

. The groups' identities were revealed only after the data analyses were completed and the study investigators reached consensus on the interpretation of results. All analyses were performed with STATA statistical software, release 13 (StataCorp, Austin, Texas) and SAS version 9.4 (SAS Institute, Cary, North Carolina). Definitions Both absolute and standardized anthropometric measures were used in defining growth outcomes,24 and z scores were calculated based on the WHO Child Growth Standards. We evaluated intervention effects on the following set of primary outcomes: (1) length and length-for-age z scores (LAZ); (2) weight and weight-for-age z-scores (WAZ); (3) weight-for-length z scores (WLZ); (4) MUAC, (5) stunting (LAZ <–2 z scores), underweight (WAZ <–2 z scores), and wasting (WLZ <–2 z-scores); and (6) mean hemoglobin and anemia (hemoglobin <110 g/L). Zinc deficiency was defined as plasma zinc <65 µg/dL, storage iron deficiency as low ferritin (<12 µg/L), and iron deficiency anemia as anemia with low ferritin.25 Elevated sTfR, indicative of functional iron deficiency, was defined as sTfR>8.3 mg/L.25 Because the RBP assays produced concentrations ~13% greater than the NIST standards, vitamin A deficiency was defined as RBP <0.81 µmol/L instead of the commonly used cut-off of 0.7 µmol/L.26 Infant and young children feeding practices were assessed based on caregiver recall using structured survey questions as suggested by WHO.27 Food security was assessed using the Household Food Insecurity Access Scale.28 Indices of socioeconomic status, hygiene and sanitation, and water quality were developed using principal component analyses of available baseline household-level indicators, including land ownership, number of livestock and motorized vehicles, source of drinking water, availability and type of toilet, and income and education level of household head, among others.29

tation, and water quality were developed using principal component analyses of available baseline household-level indicators, including land ownership, number of livestock and motorized vehicles, source of drinking water, availability and type of toilet, and income and education level of household head, among others.29 Statistical Analyses All analyses were done on an intention-to-treat basis among children with available data.30 In all analyses, the intervention group was considered the primary exposure variable. Models first assessed a global difference in treatment effect using a likelihood ratio test, and post-hoc pairwise differences were assessed subsequently in the event of a statistically significant global difference (global P value < .05). In all cases of statistically significant pair-wise comparisons, multiple hypothesis testing adjustments were made to determine the sensitivity of estimates.

ood ratio test, and post-hoc pairwise differences were assessed subsequently in the event of a statistically significant global difference (global P value < .05). In all cases of statistically significant pair-wise comparisons, multiple hypothesis testing adjustments were made to determine the sensitivity of estimates. For anthropometric and anemia outcomes, ANCOVA or modified Poisson regression methods were used to model continuous and binary outcomes, respectively. For each outcome, treatment effect parameters were estimated using minimally adjusted models that controlled for baseline measurement of the respective outcome, age at enrollment, sex, and district as the only covariates. In the assessment of treatment effects on the nutritional biomarkers (both continuous and dichotomous), separate models were constructed using either measured concentrations, or the inflammation-adjusted concentration calculated by adapting procedures recommended by the Biomarkers Reflecting Inflammation and Nutritional Determinant of Anemia project.31, 32, 33 Specifically, adjustment factors reflecting the changes in nutritional biomarkers during inflammation were determined by pooling together the samples collected at baseline from all 4 groups. The adjustment factors (ie, regression coefficient for CRP and/or AGP) were subsequently applied to both baseline and endline samples. In adapting the approach of the Biomarkers Reflecting Inflammation and Nutritional Determinant of Anemia project, we used the 10th percentile of the CRP and AGP concentrations in this population. In addition, adjustment factors for CRP and/or AGP were applied only if the biomarker was significantly associated with CRP or AGP (P < .05 for the regression coefficient). In all models, ferritin, zinc, sTfR, and RBP were log-transformed. Potential effect modification by baseline variables was explored by incorporating interaction terms in the statistical models and further investigated if marginally significant (P < .1). The full list of the specific covariates and effect modifiers is available in the published analysis plan.23

BP were log-transformed. Potential effect modification by baseline variables was explored by incorporating interaction terms in the statistical models and further investigated if marginally significant (P < .1). The full list of the specific covariates and effect modifiers is available in the published analysis plan.23 Results Of the 3830 children screened, 3433 were eligible for enrollment (Figure 1; available at www.jpeds.com). This included 3407 individually randomized children and 26 children from a twin pair (who were excluded at the analytic stage to ensure that only one child per household remained in the data set for analysis). Attrition during the 36 weeks of follow-up was 10% in the therapeutic zinc group, 13% in preventive zinc and control groups, and 17% in the micronutrient powder group (P = .01). The children lost to follow-up were statistically similar to those who completed the study with respect to baseline age, maternal variables, and anemia (Table II; available at www.jpeds.com). Similarly, baseline anthropometric indicators were comparable between those who completed the study vs those who dropped out, except for MUAC, which was slightly, but significantly, lower (P = .04) in the children who completed the study (13.8 ± 1.0 cm) compared with those lost to follow-up (13.9 ± 1.1 cm).

jpeds.com). Similarly, baseline anthropometric indicators were comparable between those who completed the study vs those who dropped out, except for MUAC, which was slightly, but significantly, lower (P = .04) in the children who completed the study (13.8 ± 1.0 cm) compared with those lost to follow-up (13.9 ± 1.1 cm). The average age at baseline was 14.3 ± 5.0 months (Table III). The prevalences of stunting and anemia were 40% and 55%, respectively. Approximately 73% of children were breastfed in the previous 4 weeks, and only 14% met the WHO definition of adequate dietary diversity. Overall reported adherence to the preventive supplement was 91%, ranging from 89% in the micronutrient powder group to 92% in the preventive zinc and therapeutic zinc groups (P < .01; Table IV [available at www.jpeds.com]). This amounted to a daily supplemental zinc intake of ~6.5 mg for the preventive zinc and ~9.0 mg for the micronutrient powder group over the duration of follow-up. Children received diarrhea treatment for an average of 4 days per 100 child days. Children in the therapeutic zinc group received the equivalent of 0.8 mg/d zinc over the course of the study (Table IV).Table III Group-wise comparison of baseline characteristics in children included in the analyses

uration of follow-up. Children received diarrhea treatment for an average of 4 days per 100 child days. Children in the therapeutic zinc group received the equivalent of 0.8 mg/d zinc over the course of the study (Table IV).Table III Group-wise comparison of baseline characteristics in children included in the analyses Table IIICharacteristics All (n = 2943) Preventive zinc (n = 738) Micronutrient powder (n = 701) Therapeutic zinc (n = 764) Control (n = 740) Child characteristics* Age, mo 14.3 ± 5.0 14.1 ± 5.1 14.3 ± 5.0 14.5 ± 5.2 14.1 ± 5.1 Male, n (%) 1503 (51.1) 370 (50.1) 356 (50.8) 392 (51.3) 385 (52.0) Length, cm 72.5 ± 5.5 72.1 ± 5.5 72.5 ± 5.4 72.6 ± 5.7 72.5 ± 5.6 Weight, kg 8.3 ± 1.3 8.2 ± 1.3 8.3 ± 1.3 8.3 ± 1.3 8.2 ± 1.3 MUAC, cm 13.8 ± 1.0 13.7 ± 1.0 13.8 ± 1.0 13.8 ± 1.0 13.8 ± 1.0 LAZ −1.75 ± 1.08 −1.78 ± 1.03 −1.76 ± 1.07 −1.78 ± 1.11 −1.67 ± 1.09 WAZ −1.44 ± 1.01 −1.48 ± 1.00 −1.46 ± 1.02 −1.43 ± 1.03 −1.37 ± 1.00 WLZ −0.71 ± 0.96 −0.74 ± 0.97 −0.73 ± 0.96 −0.69 ± 0.98 −0.67 ± 0.94 Stunting, n (%) 1167 (39.7) 303 (41.1) 288 (41.1) 300 (39.3) 276 (37.3) Underweight, n (%) 808 (27.5) 209 (28.3) 212 (30.2) 211 (27.6) 176 (23.8) Wasting, n (%) 243 (8.3) 72 (9.8) 58 (8.3) 58 (7.6) 55 (7.4) Hemoglobin, g/L 107.0 ± 1.1 107.6 ± 1.1 107.5 ± 1.1 107.8 ± 1.1 108.0 ± 1.0 Anemia, n (%) 1614 (54.8) 406 (55.0) 396 (56.5) 415 (54.3) 397 (53.7) Complementary feeding Breastfeeding, n (%) 2107 (72.6) 529 (73.9) 551 (76.1) 516 (70.4) 511 (70.2) Adequate dietary diversity, n (%) 413 (14.2) 107 (14.9) 91 (12.5) 108 (14.7) 107 (14.7) Iron-rich foods, n (%) 1948 (67.0) 478 (66.8) 478 (65.8) 484 (65.9) 508 (69.6) Household/maternal* District, n (%) Xebangfai 740 (25.1) 183 (24.8) 184 (26.1) 190 (24.9) 184 (24.9) Nongbok 449 (15.3) 112 (15.2) 104 (14.8) 119 (15.6) 114 (15.4) Mahaxay 734 (25.0) 181 (24.5) 178 (25.4) 188 (24.6) 187 (25.3) Xaibuathong 585 (19.9) 152 (20.6) 132 (18.8) 156 (20.5) 145 (19.6) Yommalat 434 (14.7) 110 (14.9) 104 (14.8) 110 (14.4) 110 (14.9) HFIAS 3.2 ± 5.0 3.2 ± 5.0 3.2 ± 5.0 3.2 ± 5.0 3.1 ± 4.0 Maternal age, y 26.8 ± 5.9 26.5 ± 5.7 26.8 ± 5.9 26.7 ± 6.0 27.0 ± 6.0 Maternal BMI, kg/m2 21.4 ± 2.9 21.4 ± 3.0 21.5 ± 2.7 21.4 ± 2.9 21.4 ± 3.0 Micronutrients (n = 568)† Plasma zinc, µg/dL 54.2 (14.2) 52.2 (16.6) 55.2 (13.8) 54.4 (13.0) 55.1 (13.5) Zinc deficiency, n (%) 432 (75.4) 110 (74.3) 101 (72.1) 108 (74.5) 113 (80.7) Ferritin, µg/L 31.0 (32.6) 30.3 (32.7) 27.7 (26.6) 27.1 (21.6) 27.2 (30.5) Storage ID, n (%) 150 (26.1) 36 (24.3) 39 (27.9) 37 (25.3) 38 (27.1)

Zinc deficiency defined as zinc <65 µg/dL, Storage ID as ferritin <12 µg/L, Functional ID as sTfR >8.3 mg/L, and vitamin A deficiency defined as RBP <0.81 µmol/L. * Values represent mean ± SD for continuous variable or n (%) for dichotomized variables. † Values represent median (IQR) for continuous variable or n (%) for dichotomized variables. At endline, the adjusted mean length (79.1 ± 4.9 cm), mean weight (9.6 ± 1.3 kg), and mean MUAC (14.0 ± 1.0 cm) did not differ by study group (Table V). Likewise, the mean LAZ (–1.94 ± 1.0), WAZ (–1.52 ± 1.0), and WLZ (–0.72 ± 0.93) were similar across the 4 groups. These standardized anthropometric scores declined in all groups relative to their corresponding baseline values. There was no treatment effect on stunting (44%-50% across the groups; P = .37) or underweight (26%-30% across the groups; P = .45). Wasting prevalence differed significantly by study group (P = .02) and was significantly lower in the preventive zinc group (4.7%; 95% CI 3.7%-7.0%) compared with the therapeutic zinc group (7.6%; 95% CI 5.5%-9.2%). There were no other significant pair-wise differences in wasting prevalence. A secondary analyses of anthropometry outcomes assessed at midline (~18 months) did not find any treatment effects (data not shown).Table V Effects of 32-40 weeks of supplementation with daily preventive zinc supplements, daily micronutrient powder, or therapeutic zinc supplements for diarrhea on final absolute and standardized anthropometric indices and biochemical outcomes among rural Laotian children

t find any treatment effects (data not shown).Table V Effects of 32-40 weeks of supplementation with daily preventive zinc supplements, daily micronutrient powder, or therapeutic zinc supplements for diarrhea on final absolute and standardized anthropometric indices and biochemical outcomes among rural Laotian children Table VOutcome Preventive zinc Micronutrient powder Therapeutic zinc Control P value* Anthropometry n 739 701 763 740 – Length, cm 79.0 ± 4.8 79.2 ± 4.8 79.2 ± 5.0 79.3 ± 4.9 .41 Weight, kg 9.52 ± 1.25 9.58 ± 1.32 9.63 ± 1.41 9.63 ± 1.32 .92 MUAC, cm 14.0 ± 0.9 14.0 ± 1.0 14.0 ± 1.0 14.0 ± 0.9 .71 LAZ −1.93 ± 0.97 −1.94 ± 1.00 −1.95 ± 1.02 −1.93 ± 1.00 .58 WAZ −1.51 ± 0.88 −1.51 ± 0.93 −1.53 ± 0.97 −1.52 ± 0.90 .76 WLZ −0.72 ± 0.79 −0.72 ± 0.85 −0.73 ± 0.89 −0.74 ± 0.82 .92 Stunting, n (%) 355 (47.0) 337 (47.0) 374 (49.6) 317 (44.4) .37 Underweight, n (%) 217 (28.9) 197 (26.1) 219 (28.5) 203 (30.4) .45 Wasting, n (%) 38 (4.7)a 39 (6.2)ab 57 (7.6)b 35 (4.7)ab .02 Anemia n 722 689 751 726 — Hemoglobin, g/L 110.4 ± 0.4 111.6 ± 0.4 110.6 ± 0.4 110.6 ± 0.4 .09 Anemia, n (%) 324 (44.9) 274 (39.2) 315 (42.4) 313 (43.2) .14 Micronutrient status Inflammation, unadjusted n 145 138 145 140 Plasma zinc, µg/dL 63.4 (61.0, 65.7)c 59.7 (57.6, 62.1)a 54.4 (52.4, 56.4)b 53.6 (51.6, 55.6)b <.001 Zinc deficiency, % 87 (60.0)a 92 (67.1)a 122 (84.7)b 119 (84.5)b <.001 Ferritin, µg/L 28.1 (25.2, 31.0)b 37.2 (33.3, 41.1)a 26.0 (23.3, 28.6)b 26.7 (23.8, 29.3)b <.001 Storage ID (%) 20 (14.2)b 6 (4.8)a 28 (17.6)b 22 (15.4)b .02 sTfR, mg/L 10.2 (9.8, 10.7)b 9.5 (9.1, 9.9)a 10.3 (9.9, 10.7)b 10.4 (10.0, 10.9)b .02 Functional ID (%) 92 (63.9) 78 (58.2) 91 (62.4) 92 (65.9) .49 RBP, µmol/L 1.19 (1.14, 1.23) 1.22 (1.18, 1.27) 1.23 (1.19, 1.28) 1.20 (1.16, 1.25) .49 VAD 11 (7.2) 12 (9.0) 11 (8.0) 7 (4.8) .54 Inflammation-adjusted† n 145 138 145 140 Zinc, µg/dL 66.4 (64.0, 68.9)c 62.2 (59.8, 64.5)a 56.8 (54.7, 58.9)b 56.0 (53.9, 58.1)b <.001 Zinc deficiency (%) 74 (50.7)a 81 (59.1)a 114 (79.2)b 111 (78.6)b <.001 Ferritin, µg/L 18.9 (17.2, 20.7)b 26.0 (23.5, 28.5)a 18.0 (16.4, 19.7)b 18.4 (16.7, 20.1)b <.001 Storage ID (%) 33 (17.0)b 16 (5.8)a 42 (21.1)b 40 (20.3)b <.01 sTfR, mg/L — — — — — Functional ID (%) — — — — — RBP, µmol/L 1.33 (1.28, 1.37) 1.33 (1.29, 1.38) 1.35 (1.31, 1.40) 1.33 (1.29, 1.38) .58 VAD‡ — — — — — Inflammation n 145 138 145 140 CRP, mg/L 0.68 ± 0.08 0.47 ± 0.06 0.58 ± 0.07 0.57 ± 0.08 .17 AGP, g/L 0.69 ± 0.03 0.65 ± 0.03 0.64 ± 0.03 0.65 ± 0.03 .59 ID, iron deficiency; VAD,

— — — — Functional ID (%) — — — — — RBP, µmol/L 1.33 (1.28, 1.37) 1.33 (1.29, 1.38) 1.35 (1.31, 1.40) 1.33 (1.29, 1.38) .58 VAD‡ — — — — — Inflammation n 145 138 145 140 CRP, mg/L 0.68 ± 0.08 0.47 ± 0.06 0.58 ± 0.07 0.57 ± 0.08 .17 AGP, g/L 0.69 ± 0.03 0.65 ± 0.03 0.64 ± 0.03 0.65 ± 0.03 .59 ID, iron deficiency; VAD, vitamin A deficiency. Values represent mean ± SD for continuous anthropometric values; geometric mean (95% CI) for continuous nutritional biomarkers, frequency (marginal prevalence) for all dichotomous variables and geometric ± standard error for CRP and AGP. * Values on the same row with different superscript are significantly different (P < .05) after adjustment for age, sex, and district baseline status. Zinc deficiency defined as zinc < 65 µg/dL, Storage ID as ferritin <12 µg/L, Functional ID as sTfR >8.3 mg/L, and VAD defined as RBP <0.81 µmol/L. † In adjusting for inflammation, adjustment factors were estimated by pooling together baseline data across the 4 groups and subsequently using procedures recommended by the Biomarkers Reflecting Inflammation and Nutritional Determinant of Anemia project to estimate CRP and AGP coefficients. There was no significant association between baseline sTfR and CRP or AGP. Hence, endline sTfR was not adjusted for inflammation. ‡ Prevalence too low to model treatment effects.

† In adjusting for inflammation, adjustment factors were estimated by pooling together baseline data across the 4 groups and subsequently using procedures recommended by the Biomarkers Reflecting Inflammation and Nutritional Determinant of Anemia project to estimate CRP and AGP coefficients. There was no significant association between baseline sTfR and CRP or AGP. Hence, endline sTfR was not adjusted for inflammation. ‡ Prevalence too low to model treatment effects. At endline, mean inflammation-adjusted plasma zinc concentrations in the preventive zinc (66.4 µg/dL [95% CI 64.0-68.9]) and micronutrient powder (62.2 µg/dL [95% CI 59.8-64.5]) groups were significantly greater than in the therapeutic zinc (56.8 µg/dL [95% CI 54.7-58.9]) and control (56.0 µg/dL [95% CI 53.9-58.1]) groups (Table V, P < .001), resulting in a significantly lower prevalence of zinc deficiency (P < .001) in preventive zinc (60%) and micronutrient powder (67%) groups compared with the therapeutic zinc (85%) and control (85%) groups. Ferritin concentration in the micronutrient powder group (26.0 mg/L [95% CI 23.5-28.5]) was greater (P < .001) than the other 3 groups (~18 mg/L, Table V), translating into ~44-55% reduction in iron deficiency. The micronutrient powder resulted in lower sTfR concentrations, with no impact on functional iron deficiency (Table V). We found no treatment effects on RBP, CRP, or AGP.

0 mg/L [95% CI 23.5-28.5]) was greater (P < .001) than the other 3 groups (~18 mg/L, Table V), translating into ~44-55% reduction in iron deficiency. The micronutrient powder resulted in lower sTfR concentrations, with no impact on functional iron deficiency (Table V). We found no treatment effects on RBP, CRP, or AGP. There was an overall marginal effect of the study group on hemoglobin concentration (P = .09) but not anemia prevalence (P = .14) (Table V). Specifically, the final hemoglobin concentration in the micronutrient powder group (111.6 ± 0.4 g/L) was greater compared with the preventive zinc (110.4 ± 0.4 g/L), therapeutic zinc (110.6 ± 0.4 g/L), and control (110.6 g/L ± 0.4) groups. We observed an interaction-by-baseline hemoglobin status (P = .06), such that among children who were anemic at baseline, the micronutrient powder was associated with a trend to a reduction in anemia of 6-9 percentage points (global P = .06) compared with the other 3 groups (Figure 2), whereas there was no impact among nonanemic children. The micronutrient powder did not prevent the development of anemia among previously nonanemic children (Table VI; available at www.jpeds.com). The prevalence of anemia among previously nonanemic children was 24%-28% (Table VI) and did not differ by group allocation (P = .62).Figure 2 Effects of 32-40 weeks of supplementation with daily preventive zinc supplements, micronutrient powder, or therapeutic zinc supplements for diarrhea on anemia prevalence among rural Laotian children, stratified by baseline hemoglobin concentrations. *Models adjusted for age, sex, district, and baseline hemoglobin. No effects of micronutrient powder on anemia in previously nonanemic children (baseline Hb ≥110; P > .05 for all pairwise comparisons); in previously anemic children (baseline Hb <110), the micronutrient powder reduced the prevalence of anemia by 9 percentage points (vs preventive zinc; P = .008), by 7 percentage points (vs therapeutic zinc; P = .041) and by 6 percentage points (vs control, P = .08). MNP, micronutrient powder; PZ, preventive zinc; TZ, therapeutic zinc.

mic children (baseline Hb <110), the micronutrient powder reduced the prevalence of anemia by 9 percentage points (vs preventive zinc; P = .008), by 7 percentage points (vs therapeutic zinc; P = .041) and by 6 percentage points (vs control, P = .08). MNP, micronutrient powder; PZ, preventive zinc; TZ, therapeutic zinc. Figure 2 Complementary feeding practices did not differ over the course of the study. Based on data pooled from 9 monthly caregiver dietary recalls, the percentage of times children met the WHO recommended criteria for minimum dietary diversity (~29%, P = .72) and minimum meal frequency (~61%, P = .72), and consumption of animal source foods (~90%, P = .18) and iron-rich foods (~84%, P = .44) did not differ by group (Table VII; available at www.jpeds.com). Discussion The results of this community-based, randomized controlled trial indicate that daily preventive zinc supplements providing 7 mg zinc/d and daily micronutrient powder sachets containing 10 mg zinc, despite improving plasma zinc concentrations, had no impact on linear growth or weight gain among young children residing in rural health districts of central Lao PDR. Therapeutic zinc supplementation given for the treatment of diarrhea had no effect on zinc status nor linear growth. In addition, the low-iron, high-zinc micronutrient powder resulted in a significant positive effect on iron status and an overall marginal effect on hemoglobin. Moreover, the micronutrient powder had a marginally significant positive effect on hemoglobin and anemia in children who were anemic at baseline.

inear growth. In addition, the low-iron, high-zinc micronutrient powder resulted in a significant positive effect on iron status and an overall marginal effect on hemoglobin. Moreover, the micronutrient powder had a marginally significant positive effect on hemoglobin and anemia in children who were anemic at baseline. The lack of impact of daily preventive zinc supplementation on physical growth in the present study differs from the results of 4 systematic reviews and meta-analyses on preventive zinc supplementation, all of which found a small positive effect on linear growth and weight gain, albeit with significant heterogeneity of responses across trials.3, 4, 5, 6 The failure of the high-zinc, low iron micronutrient powder to increase linear growth is consistent with previous meta-analyses of micronutrient powder, which did not find an effect of micronutrient powder on growth.8 The lack of any growth impact of therapeutic zinc is not surprising, given the low frequency of therapeutic supplementation and correspondingly small amount of supplemental zinc consumed over the course of the trial by the children in this study group. Our findings regarding therapeutic zinc and growth are consistent with previous studies, which found no impact of short-term zinc supplementation of acutely ill children on their subsequent growth14, 34 and highlights the need for additional interventions beyond the period of zinc treatment.

ildren in this study group. Our findings regarding therapeutic zinc and growth are consistent with previous studies, which found no impact of short-term zinc supplementation of acutely ill children on their subsequent growth14, 34 and highlights the need for additional interventions beyond the period of zinc treatment. Possible reasons for the lack of any growth effect in the present study may be that the study participants were not sufficiently growth restricted or zinc deficient to be able to respond to supplementation, the dose or duration of supplementation was insufficient, possibly because of impaired intestinal absorption, or the children did not adhere to the supplementation protocol. As ~40% of children were stunted and ~75% had low plasma zinc concentration, the children should have been able to respond to supplemental zinc if zinc deficiency was the only factor restricting their growth. Moreover, we found no evidence of effect modification by baseline LAZ, suggesting that even the more severely growth-restricted children did not respond to zinc supplementation. However, the prevalence of chronic inflammation in this population (21% with elevated AGP concentrations) may have had constraining effects on growth. The dose and duration of preventive zinc tablet were consistent with the range of dosing regimens used in earlier studies that did find a growth response. Thus, the dosing regimen does not seem to explain the lack of observed growth impact. Some studies have shown that the absorption of zinc from micronutrient powder is low35 and that the efficacy of zinc may be compromised when zinc is coadministered with iron.36 The dose of 10 mg of zinc used in the micronutrient powder was based on consideration of both the amount of zinc absorbed when mixed with food (<50%)35 and the dose of zinc in micronutrient powder previously shown to increase plasma or serum zinc concentration.37, 38 The fact that both the micronutrient powder and preventive zinc significantly increased plasma zinc concentration relative to the control group provides evidence that the supplemental zinc was being consumed and absorbed, and the most plausible explanation for the lack of growth effect is that growth faltering in this study population may be driven by other factors that are not responsive to zinc supplementation. Additional research is needed to fully understand the etiology of growth faltering in this population.

consumed and absorbed, and the most plausible explanation for the lack of growth effect is that growth faltering in this study population may be driven by other factors that are not responsive to zinc supplementation. Additional research is needed to fully understand the etiology of growth faltering in this population. The statistically significant effect of preventive zinc on wasting is of small magnitude and may represent a spurious finding in view of the lack of any independent effects of the supplements on length and weight gain.

consumed and absorbed, and the most plausible explanation for the lack of growth effect is that growth faltering in this study population may be driven by other factors that are not responsive to zinc supplementation. Additional research is needed to fully understand the etiology of growth faltering in this population. The statistically significant effect of preventive zinc on wasting is of small magnitude and may represent a spurious finding in view of the lack of any independent effects of the supplements on length and weight gain. In settings with a high burden of infections, there is interest in finding the minimum effective iron dose necessary for reducing anemia, without increasing the risk for adverse health outcomes.10, 39 Proposed strategies for achieving this include daily supplementation with 10-12.5 mg for just 3 consecutive months a year and intermittent iron supplementation a few days a week instead of daily.40 The limited evidence on these strategies suggest an overall efficacy (compared with placebo controls) in reducing anemia, albeit with a lower effect size compared with delivering 12.5 mg iron daily.41, 42, 43 In the current study, we intentionally reduced the amount of iron in the micronutrient powder in an attempt to avoid possible adverse effects of iron supplementation. Our results are consistent with this pattern of lower efficacy compared with previous studies of daily supplementation with greater amounts of iron for 2-12 months.8, 44 Although the present study found a marginal impact on anemia prevalence, this effect was lower than in previous micronutrient powder trials,8 possibly due to the lower iron content (6 mg iron /day) in the micronutrient powder provided in the present study. Even in children who were anemic, in whom the micronutrient powder increased hemoglobin concentration and reduced the prevalence of anemia, the relative reduction in anemia prevalence was <16%, compared with ~30% found in trials delivering 12.5 mg of iron.8

ay) in the micronutrient powder provided in the present study. Even in children who were anemic, in whom the micronutrient powder increased hemoglobin concentration and reduced the prevalence of anemia, the relative reduction in anemia prevalence was <16%, compared with ~30% found in trials delivering 12.5 mg of iron.8 In neighboring Savannakhet (south of the study area), an earlier study by Kounnavong et al reported that in children 6-52 months, daily or twice-weekly micronutrient powder containing 12.5 mg of iron was associated with a greater reduction in anemia (32-35 percentage points decrease) compared with children receiving no micronutrient powder (only 10 percentage points decrease) during 24 weeks of follow-up.45 The magnitude of effect reported by Kounnavong et al was greater than observed in our study population (6-9 percentage point reduction), despite a lower reported compliance (73%; defined as consumption on at least 5 days in a week) and a shorter duration of follow-up (up to 24 weeks) in that study.45 This difference in effect may suggest that a greater iron-dose micronutrient powder, even if given intermittently, may be more efficacious than the lower daily dose used in this study. Perhaps more importantly, the marginal response in anemia despite a substantial relative reduction in iron deficiency suggests the presence of other cause of anemia in this population. This view also is supported by the fact that the micronutrient powder did not prevent the incidence of new anemia cases among children who were not anemic at baseline. In Cambodian children, Wieringa et al concluded that hemoglobinopathies, and not iron deficiency, may be more relevant to the burden of anemia.46 Additional evidence is needed to determine the cause-specific attribution fractions for anemia with respect to iron deficiency, hemoglobinopathies, and infections,46, 47 to determine optimal anemia control strategies.

uded that hemoglobinopathies, and not iron deficiency, may be more relevant to the burden of anemia.46 Additional evidence is needed to determine the cause-specific attribution fractions for anemia with respect to iron deficiency, hemoglobinopathies, and infections,46, 47 to determine optimal anemia control strategies. Several strengths of this study include its implementation in a zinc-deficient and thus potentially responsive population, the individual randomization protocol, the frequency of follow-up visits, and the masking procedures used to ensure continued blinding throughout the field implementation and data analyses. In addition, the drop-out rate of ~13% was lower than expected, ensuring a sufficient final sample size for addressing the proposed questions. As indicated previously, the baseline characteristics of the children lost to follow-up were comparable with those who remained in the study. Moreover, among the children lost to follow-up, there were no differences in age, weight, length, and hemoglobin between the groups at baseline.

ddressing the proposed questions. As indicated previously, the baseline characteristics of the children lost to follow-up were comparable with those who remained in the study. Moreover, among the children lost to follow-up, there were no differences in age, weight, length, and hemoglobin between the groups at baseline. A potential weakness of this trial is the nature of the comparison groups. The micronutrient powder and control preventive interventions were formulated in powder forms, whereas the preventive zinc and therapeutic zinc were formulated in tablet forms, potentially undermining the blinding procedure. However, because there were 2 powder groups, and 2 tablet groups, it was impossible to identify the exact intervention allocation for any particular group. A related problem is that micronutrient powder is designed to be taken along with food, whereas the preventive zinc is recommended to be consumed between meals. Thus, it is possible that the children's dietary patterns were affected differently, possibly confounding treatment effects. However, analyses of the dietary data found no difference in meal frequency, dietary diversity, consumption of iron-rich foods, and animal source foods (Table VII) across the groups over the course of the follow-up period, and adjusting for these variables did not change the estimated impact on growth and anemia outcomes.

ever, analyses of the dietary data found no difference in meal frequency, dietary diversity, consumption of iron-rich foods, and animal source foods (Table VII) across the groups over the course of the follow-up period, and adjusting for these variables did not change the estimated impact on growth and anemia outcomes. In conclusion, our data suggest that preventive zinc supplements, provided alone or in combination with other micronutrients at doses of 7-10 mg zinc/d to young children in rural central Lao PDR, improved biomarkers of zinc status but had no impact on physical growth. Also, because the micronutrient powder tended to improve iron status and reduce anemia among initially anemic children, this may be the preferred strategy to deliver zinc and other micronutrients in this population. Additional research is needed to understand the optimal strategy for reducing the high burden of stunting and anemia in this population. Acknowledgments available at www.jpeds.com Appendix Supplement CONSORT 2010 checklist of information to include when reporting a randomised trial.

In conclusion, our data suggest that preventive zinc supplements, provided alone or in combination with other micronutrients at doses of 7-10 mg zinc/d to young children in rural central Lao PDR, improved biomarkers of zinc status but had no impact on physical growth. Also, because the micronutrient powder tended to improve iron status and reduce anemia among initially anemic children, this may be the preferred strategy to deliver zinc and other micronutrients in this population. Additional research is needed to understand the optimal strategy for reducing the high burden of stunting and anemia in this population. Acknowledgments available at www.jpeds.com Appendix Supplement CONSORT 2010 checklist of information to include when reporting a randomised trial. Supplement Figure 1 Consort diagram of study participant progression through the growth and biochemical assessments in the Lao Zinc Study. The micronutrient powder group received 1 micronutrient powder sachet daily (vitamin A [400 µg of retinol equivalents], thiamin [0.5 mg], riboflavin [0.5 mg], niacin [6 mg], vitamin B6 [0.5 mg], folic acid [150 µg of dietary folate equivalents], cyanocobalamin [0.9 µg], ascorbic acid [30 mg], cholecalciferol [5 mg], dl-α-tocopheryl acetate [5 mg of tocopherol equivalents], copper sulfate anhydrous [to provide 0.56 mg copper], potassium iodate [to provide 90 µg iodine], ferrous fumarate [to provide 6 mg of iron], selenium selenite [to provide 17 µg of selenium], zinc gluconate [to provide 10 mg of zinc]) and placebo tablet (1 per day for 10 days) during diarrhea episodes. The preventive zinc group (preventive zinc supplements) received 1 zinc tablet (7 mg/d) daily and placebo tablet (1 per day for 10 days) during diarrhea episodes. The therapeutic zinc group (therapeutic zinc supplements) received 1 placebo tablet daily and a 20-mg zinc tablet (1 per day for 10 days) during diarrhea episodes. Controls received 1 placebo powder sachet daily and placebo tablet (1 per day for 10 days) during diarrhea episodes. All children received ORS during diarrhea episodes, regardless of group allocation.

supplements) received 1 placebo tablet daily and a 20-mg zinc tablet (1 per day for 10 days) during diarrhea episodes. Controls received 1 placebo powder sachet daily and placebo tablet (1 per day for 10 days) during diarrhea episodes. All children received ORS during diarrhea episodes, regardless of group allocation. Figure 1Table I Preventive and therapeutic supplements received by each study group of the Lao Zinc Study Table ISupplements Preventive zinc Micronutrient powder Therapeutic zinc Control Preventive supplement Zinc-containing tablet (7 mg/d) Micronutrient powder (containing 10 mg of zinc, 6 mg of iron, + 13 other micronutrients*) Placebo tablet Placebo powder Therapeutic supplement for diarrhea ORS + placebo tablet ORS + placebo tablet ORS + zinc tablet (20 mg/d) for 10 d during acute diarrhea episodes ORS + placebo tablet * Micronutrient powder contents of 1 daily dose micronutrient powder sachet: vitamin A (400 µg Retinol Equivalents), thiamin (0.5 mg), riboflavin (0.5 mg), niacin (6 mg), vitamin B6 (0.5 mg), folic acid (150 µg of Dietry Folate Equivalents), cyanocobalamin (0.9 µg), ascorbic acid (30 mg), cholecalciferol (5 mg), dl-α-tocopheryl acetate (5 mg of Tocopherol Equivalents), copper sulfate anhydrous (0.56 mg of copper), potassium iodate (90 µg of iodine), ferrous fumarate (6 mg of iron), selenium selenite (17 µg of selenium), and zinc gluconate (10 mg of zinc). Table IV Caregiver report of adherence to study interventions among children participating in the Lao Zinc Study

Table ISupplements Preventive zinc Micronutrient powder Therapeutic zinc Control Preventive supplement Zinc-containing tablet (7 mg/d) Micronutrient powder (containing 10 mg of zinc, 6 mg of iron, + 13 other micronutrients*) Placebo tablet Placebo powder Therapeutic supplement for diarrhea ORS + placebo tablet ORS + placebo tablet ORS + zinc tablet (20 mg/d) for 10 d during acute diarrhea episodes ORS + placebo tablet * Micronutrient powder contents of 1 daily dose micronutrient powder sachet: vitamin A (400 µg Retinol Equivalents), thiamin (0.5 mg), riboflavin (0.5 mg), niacin (6 mg), vitamin B6 (0.5 mg), folic acid (150 µg of Dietry Folate Equivalents), cyanocobalamin (0.9 µg), ascorbic acid (30 mg), cholecalciferol (5 mg), dl-α-tocopheryl acetate (5 mg of Tocopherol Equivalents), copper sulfate anhydrous (0.56 mg of copper), potassium iodate (90 µg of iodine), ferrous fumarate (6 mg of iron), selenium selenite (17 µg of selenium), and zinc gluconate (10 mg of zinc). Table IV Caregiver report of adherence to study interventions among children participating in the Lao Zinc Study Table IVAdherence types* Preventive zinc Micronutrient powder Therapeutic zinc Control P value Observation days 194 603 186 133 198 975 195 804 Days preventive tablet/sachet given, n (%) 179 299 (92.1) 166 009 (89.2) 183 762 (92.4) 178 116 (91.0) <.001 Days therapeutic tablet given, n 7972 7449 8124 7962 .46 Days for which the status of supplement intake was unknown are excluded.

Therapeutic zinc Control P value Observation days 194 603 186 133 198 975 195 804 Days preventive tablet/sachet given, n (%) 179 299 (92.1) 166 009 (89.2) 183 762 (92.4) 178 116 (91.0) <.001 Days therapeutic tablet given, n 7972 7449 8124 7962 .46 Days for which the status of supplement intake was unknown are excluded. Table VI Effects of 32-40 weeks of supplementation with daily preventive zinc supplements, daily micronutrient powder, or therapeutic zinc supplements for diarrhea on endline hemoglobin and anemia in rural Laotian children, stratified by baseline anemia status Table VIBaseline anemia statuses Preventive zinc Micronutrient powder Therapeutic zinc Control P value* Anemic n 395 391 404 388 — Hemoglobin, g/L 106.6 ± 0.5a 108.9 ± 0.5b 106.9 ± 0.5a 107.3 ± 0.5a .005 Anemia at endline, % 59.3 ± 0.02a 50.1 ± 0.02b 57.1 ± 0.02a 56.1 ± 0.02a .06 Not anemic n 341 295 321 332 Hemoglobin, g/L 114.8 ± 0.5 115.3 ± 0.5 114.6 ± 0.5 115.1 ± 0.5 .77 Anemia at endline, % 26.3 ± 0.02 26.2 ± 0.03 24.1 ± 0.02 28.6 ± 0.02 .62 Values on the same row with different superscript are significantly different (P < .05) after adjustment for age, sex, and district baseline status. * Values are mean (hemoglobin) or prevalence (anemia) ± SEM. Adjusted for age, sex, district, and respective baseline values; (%) marginal prevalence at endline. Table VII Complementary feeding practices among children participating in the Laos Zinc Study

Table VIBaseline anemia statuses Preventive zinc Micronutrient powder Therapeutic zinc Control P value* Anemic n 395 391 404 388 — Hemoglobin, g/L 106.6 ± 0.5a 108.9 ± 0.5b 106.9 ± 0.5a 107.3 ± 0.5a .005 Anemia at endline, % 59.3 ± 0.02a 50.1 ± 0.02b 57.1 ± 0.02a 56.1 ± 0.02a .06 Not anemic n 341 295 321 332 Hemoglobin, g/L 114.8 ± 0.5 115.3 ± 0.5 114.6 ± 0.5 115.1 ± 0.5 .77 Anemia at endline, % 26.3 ± 0.02 26.2 ± 0.03 24.1 ± 0.02 28.6 ± 0.02 .62 Values on the same row with different superscript are significantly different (P < .05) after adjustment for age, sex, and district baseline status. * Values are mean (hemoglobin) or prevalence (anemia) ± SEM. Adjusted for age, sex, district, and respective baseline values; (%) marginal prevalence at endline. Table VII Complementary feeding practices among children participating in the Laos Zinc Study Table VII*Indicators Preventive zinc Micronutrient powder Therapeutic zinc Control P value Minimum dietary diversity 29.0 (27.0-31.0) 30.2 (28.2-32.3) 28.6 (26.6-30.6) 29.4 (27.4-31.4) .72 Animal source food 89.5 (88.2-90.7) 91.4 (90.2-92.6) 89.7 (88.4-90.9) 89.5 (88.2-90.6) .18 Iron-rich foods 84.1 (82.7-85.6) 85.4 (84.0-86.9) 83.9 (82.5-85.3) 84.2 (82.8-85.6) .44 Minimum meal frequency 60.8 (58.2-63.3) 61.0 (58.4-63.7) 60.2 (57.6-62.7) 62.4 (59.8-64.9) .72 Complementary feeding practices assessed every 4 weeks throughout the study. * Values represent the percentage of times children met the specified criteria as defined by the WHO30 and the 95% CIs around the point estimate.

Table VII*Indicators Preventive zinc Micronutrient powder Therapeutic zinc Control P value Minimum dietary diversity 29.0 (27.0-31.0) 30.2 (28.2-32.3) 28.6 (26.6-30.6) 29.4 (27.4-31.4) .72 Animal source food 89.5 (88.2-90.7) 91.4 (90.2-92.6) 89.7 (88.4-90.9) 89.5 (88.2-90.6) .18 Iron-rich foods 84.1 (82.7-85.6) 85.4 (84.0-86.9) 83.9 (82.5-85.3) 84.2 (82.8-85.6) .44 Minimum meal frequency 60.8 (58.2-63.3) 61.0 (58.4-63.7) 60.2 (57.6-62.7) 62.4 (59.8-64.9) .72 Complementary feeding practices assessed every 4 weeks throughout the study. * Values represent the percentage of times children met the specified criteria as defined by the WHO30 and the 95% CIs around the point estimate. Table II Comparison of baseline characteristics between children who completed the study and those lost to follow-up Table IICharacteristics Completed the study Lost to follow-up P value n (%) 2988 (87.1) 444 (12.9) — Age 14.2 ± 5.1 14.5 ± 5.0 .28 Length, cm 72.4 ± 5.6 72.7 ± 5.1 .30 Weight, kg 8.28 ± 1.31 8.41 ± 1.27 .06 MUAC, cm 13.8 ± 1.0 13.9 ± 1.1 .04 LAZ −1.76 ± 1.08 −1.75 ± 0.95 .86 WAZ −1.44 ± 1.01 −1.35 ± 0.98 .08 WLZ −0.70 ± 0.96 −0.61 ± 1.00 .06 Stunting, n (%) 1188 (39.8) 166 (37.4) .34 Underweight, n (%) 814 (27.2) 107 (24.1) .16 Wasting, n (%) 241 (8.1) 35 (7.8) .90 Hemoglobin, g/L 108 ± 11 107 ± 11 .18 Anemia, n (%) 1642 (54.9) 261 (58.8) .13 Maternal BMI, kg/m2 21.4 ± 3.1 21.4 ± 3.1 .77 Maternal age, y 28.0 ± 6.0 27.5 ± 5.7 .09 Values represent mean ± SD for continuous variables and frequency (%) for dichotomous variables mean.

07 (24.1) .16 Wasting, n (%) 241 (8.1) 35 (7.8) .90 Hemoglobin, g/L 108 ± 11 107 ± 11 .18 Anemia, n (%) 1642 (54.9) 261 (58.8) .13 Maternal BMI, kg/m2 21.4 ± 3.1 21.4 ± 3.1 .77 Maternal age, y 28.0 ± 6.0 27.5 ± 5.7 .09 Values represent mean ± SD for continuous variables and frequency (%) for dichotomous variables mean. Acknowledgments We thank Dr Souraxay Phrommala and Dr Somphou Sayasone (Lao Tropical and Public Health Instituteh, Vientiane, Lao PDR), and Dr Tholakhan Xaypanya and Dr Sodalai Onenavong (Khammouane Provincial Health Department, Lao PDR) for their project oversight and assistance with community engagement. We thank the members of the Data Safety and Monitoring Board (DSMB) (Niranjan Kissoon, Department of Pediatrics, University of British Columbia; Edward Frongillo, Arnold School of Public Health, University of South Carolina; Reynaldo Martorell, Rollins School of Public Health, Global Health, Emory University; Frank Wieringa, Institut de Recherche pour le Development) for their review and support of the study. We thank Charles Larson of the Canadian Coalition for Global Health Research (Ottawa) for providing expertise on morbidity assessment. We thank the entire Lao Zinc Study team and all of the participating children and their parents, the local communities and health districts of the Khammouane Province, Lao PDR.

e study. We thank Charles Larson of the Canadian Coalition for Global Health Research (Ottawa) for providing expertise on morbidity assessment. We thank the entire Lao Zinc Study team and all of the participating children and their parents, the local communities and health districts of the Khammouane Province, Lao PDR. Supported by the Mathile Institute for the Advancement of Human Nutrition and Nutrition International (formerly the Micronutrient Initiative) (10-1347-UCALIF-07). Other aspects of the Lao Zinc Study were supported by the Bill & Melinda Gates Foundation (OPP1134272). K.B. and the spouse of S.H. work for the Bill & Melinda Gates Foundation, which provided part of the financial support. The sponsors had no involvement in the field implementation, data analyses, and manuscript writing. S.H. worked as a consultant for the Bill & Melinda Gates Foundation. The other authors declare no conflicts of interest. Portions of this study were presented at the American Society for Nutrition Conference, June 12, 2018, Boston, Massachusetts, and at the International Congress on Nutrition, October 17, 2017, Buenos Aires, Argentina.

Warts, hypogammaglobulinemia, infections, and myelokathexis (WHIM) syndrome is a rare immunodeficiency disorder characterized by warts, hypogammaglobulinemia, infections, and abnormal retention of mature neutrophils in the bone marrow (myelokathexis).1,2 This neutropenia, which is associated with a B and T lymphopenia and hypogammaglobulinemia, results in an increased risk for bacterial infections. Patients with WHIM syndrome present major and selective susceptibility to human papillomavirus that may manifest as cutaneous warts and genital dysplasia and cancer. The real prevalence of the disease is unknown but ∼40 cases have been reported.3 Autosomal dominant heterozygous mutations of the gene encoding the CXC chemokine receptor 4 (CXCR4) have been associated with the syndrome4 and lead to a gain of CXCR4 function.5,6 CXCR4 is a G protein–coupled receptor with a unique ligand, CXCL12 (previously named SDF-1).7,8 This signaling axis orchestrates leukocyte trafficking and is involved in the regulation of bone marrow homeostasis, hematopoiesis, and organogenesis.9-11 Notably, besides hematopoietic and central nervous system defects, mice deficient for CXCR4 or CXCL12 exhibit cardiac defects, indicating a role for this axis in ventricular septum formation.12,13

s leukocyte trafficking and is involved in the regulation of bone marrow homeostasis, hematopoiesis, and organogenesis.9-11 Notably, besides hematopoietic and central nervous system defects, mice deficient for CXCR4 or CXCL12 exhibit cardiac defects, indicating a role for this axis in ventricular septum formation.12,13 Tetralogy of Fallot (TOF) is a congenital heart disease characterized by: (1) pulmonary outflow tract obstruction; (2) ventricular septal defect; (3) overriding aortic root; and (4) right ventricular hypertrophy. The first 3 features are the result of abnormalities occurring during embryogenesis, and the fourth one is the consequence of the obstruction to pulmonary blood flow. The etiology is multifactorial. To date, some genetic alterations, the most frequent being represented by microdeletions of chromosome 22, have been associated with TOF but pathogenesis still remains unclear. The incidence is of 3 of every 10 000 live births, which represents approximately 10% of all congenital defects.14-16 Results We report the cases of 3 patients with WHIM syndrome who are affected by TOF (Figure). Taking into consideration the relatively poor prevalence of the two disorders, in the present report we show that patients with WHIM syndrome display an increased risk to develop TOF.

Tetralogy of Fallot (TOF) is a congenital heart disease characterized by: (1) pulmonary outflow tract obstruction; (2) ventricular septal defect; (3) overriding aortic root; and (4) right ventricular hypertrophy. The first 3 features are the result of abnormalities occurring during embryogenesis, and the fourth one is the consequence of the obstruction to pulmonary blood flow. The etiology is multifactorial. To date, some genetic alterations, the most frequent being represented by microdeletions of chromosome 22, have been associated with TOF but pathogenesis still remains unclear. The incidence is of 3 of every 10 000 live births, which represents approximately 10% of all congenital defects.14-16 Results We report the cases of 3 patients with WHIM syndrome who are affected by TOF (Figure). Taking into consideration the relatively poor prevalence of the two disorders, in the present report we show that patients with WHIM syndrome display an increased risk to develop TOF. The first patient is a 19-year-old man with a history of WHIM syndrome identified at age 2.5 years, manifesting as severe neutropenia and recurrent pneumonias, resulting in bronchiectases (Table). There was no evidence of hypogammaglobulinemia and warts. The bone marrow analysis showed myeloid hypercellularity with the presence of mature neutrophils. These hematologic and clinical findings led to the suspicion of WHIM syndrome and thereby to CXCR4 gene sequencing that revealed S338X mutation. The congenital heart disease was suspected at birth and was characterized by the anatomic variant of TOF associated with pulmonary atresia and with anomalies in branch pulmonary arteries. Of note, the patient presented with agenesis of the left-hand fingers with homolateral hypoplasia of the radius. The patient has been maintained on daily subcutaneously administered granulocyte-colony stimulating factor therapy since 2 years of age.

ith pulmonary atresia and with anomalies in branch pulmonary arteries. Of note, the patient presented with agenesis of the left-hand fingers with homolateral hypoplasia of the radius. The patient has been maintained on daily subcutaneously administered granulocyte-colony stimulating factor therapy since 2 years of age. In the second case, the patient had TOF associated with the presence of patent ductus arteriosus documented at birth; the heart disease was surgically corrected at the age of 2 years (Table). This patient is a 15-year-old girl who has had recurrent respiratory infections since early childhood. Severe neutropenia was discovered at age of 2 years, and on that occasion, an analysis of the bone marrow revealed myelokathexis with mature neutrophils presenting morphologic abnormalities consistent with apoptosis. Granulocyte-colony stimulating factor therapy was started at 5 years of age to maintain circulating neutrophil count in the normal range. Because of the occurrence of repeated pneumonia episodes, she has been maintained on antibiotic prophylaxis. In the following years, the observation of hypogammaglobulinemia suggested WHIM syndrome and genetic confirmation of the diagnosis was obtained at age 5 by detection of the R334X mutation in the CXCR4 gene. In the following years, the patient has developed a plantar wart.

has been maintained on antibiotic prophylaxis. In the following years, the observation of hypogammaglobulinemia suggested WHIM syndrome and genetic confirmation of the diagnosis was obtained at age 5 by detection of the R334X mutation in the CXCR4 gene. In the following years, the patient has developed a plantar wart. The third patient is 7 years old and presented shortly after birth with a heart murmur that led to the discovery of a TOF characterized by ventricular septal defect, overriding aorta, and pulmonary infundibular stenosis. The congenital heart disease was surgically corrected in the first months of life. Neutropenia, lymphopenia, and hypogammaglobulinemia were present since infancy, and the analysis of bone marrow showed myelokathexis. Three other family members have had neutropenia and recurrent infections, but none of them had a congenital heart defect. The genetic analysis of CXCR4 revealed the same mutation (S338X) in the 4 subjects. The child underwent treatment with intravenous immunoglobulins, obtaining a satisfactory control of infectious episodes.

other family members have had neutropenia and recurrent infections, but none of them had a congenital heart defect. The genetic analysis of CXCR4 revealed the same mutation (S338X) in the 4 subjects. The child underwent treatment with intravenous immunoglobulins, obtaining a satisfactory control of infectious episodes. Discussion Although TOF is the most common form of cyanotic congenital heart disease, its occurrence in 3 unrelated patients with WHIM syndrome is much higher than expected; the normal occurrence in the general population is 3 of every 10 000 live births. Because of the low number of patients whom we have observed, we cannot draw a correlation between the development of the heart defect with a specific mutation. The detection of the same mutation (S338X) in a family with 4 members affected by WHIM syndrome with only 1 showing TOF rules out the hypothesis that the heterozygous gain-of-function mutation of CXCR4 directly leads to the development of the cardiac defect. Instead, our observations suggest that the WHIM syndrome–associated CXCR4 truncating mutation might increase the risk that this combination of cardiac defects may develop during the formation of the fetal heart. Beyond a role for CXCL12 and CXCR4 in heart, nervous system, and blood vessel development,17 studies show that CXCR7, the recently described second receptor for CXCL12,18,19 has also a role in fetal endothelial biology and heart development—in particular, ventricular septum and heart valve development.20,21 Its germline deletion results in perinatal lethality, and its mutation affects semilunar valve development, contributing to aortic and pulmonary valve stenosis and, in some cases, septal defects.22 Recent findings support the view that CXCR7 modulates CXCR4 function by acting as a scavenger for CXCL1223 and forming heterodimers with CXCR4.20,24,25 In vitro studies on endothelial progenitor cell function show a role for both CXCR4 and CXCR7 in regulating the response of the cells to CXCL12 and, thus, angiogenesis.21 This suggests that interactions between the 2 chemokine receptors may be required for proper valve morphogenesis.

ng heterodimers with CXCR4.20,24,25 In vitro studies on endothelial progenitor cell function show a role for both CXCR4 and CXCR7 in regulating the response of the cells to CXCL12 and, thus, angiogenesis.21 This suggests that interactions between the 2 chemokine receptors may be required for proper valve morphogenesis. Our observations demonstrate that TOF can be a presenting manifestation of WHIM syndrome and that this rare inherited disease should be suspected in children with a congenital heart defect and neutropenia. The recognition of this manifestation of WHIM syndrome might help to prevent the diagnostic delay in the identification of this rare genetic disease. A prompt diagnosis should facilitate the management of leucopenia, which might include in the future CXCR4-targeted therapy as supported by 2 recent studies indicating that the specific CXCR4 antagonist plerixafor may be effective in restoring the cellular blood counts to normal.26,27 We are grateful to Virginia Gulino for critical discussion. Supported by eRare (grant to F.B. and R.B.), Fondazione Cariplo, PRIN 2009, and Telethon Italia (GGP10170A to R.B.). F.B. was supported by AP-HP (Assistance Publique-Hopitaux de Paris) and ANR (Agence Nationale de la Recherche) (ANR-07-MRAR-029) and is a member of the Laboratory of Excellence LERMIT (Laboratory of Excellence in Research on Medication and Innovative Therapeutics) (Investissements d′Avenir). The authors declare no conflicts of interest. Figure Pedigree trees of the three WHIM patients, who are affected by TOF. Table Features of WHIM syndrome in reported patients

Supported by eRare (grant to F.B. and R.B.), Fondazione Cariplo, PRIN 2009, and Telethon Italia (GGP10170A to R.B.). F.B. was supported by AP-HP (Assistance Publique-Hopitaux de Paris) and ANR (Agence Nationale de la Recherche) (ANR-07-MRAR-029) and is a member of the Laboratory of Excellence LERMIT (Laboratory of Excellence in Research on Medication and Innovative Therapeutics) (Investissements d′Avenir). The authors declare no conflicts of interest. Figure Pedigree trees of the three WHIM patients, who are affected by TOF. Table Features of WHIM syndrome in reported patients Feature Patient 1 Patient 2 Patient 3 Mutation (CXCR4) S338X R334X S338X Age at molecular diagnosis 19 y 5 y Birth (neutropenia and affected father) Clinical manifestations at onset TOF TOF TOF; neutropenia Neutropenia (age at onset) + (2.5 y) + (2 y) + (20 d) Myelokathexis + + + Hypogammaglobulinemia (age at onset) − + (5 y) + Not really assessable as early IVIG at the onset Recurrent infections Pneumonia Pneumonia − Warts − +∗ − Pulmonary outflow obstruction Pulmonary valve agenesis Pulmonary infundibular stenosis Pulmonary infundibular stenosis Overriding aorta + + + Ventricular septal defect + + + Right ventricular hypertrophy + − − Age at surgical correction 1 y 2 y 2.5 mo Medical treatment G-CSF; antibiotic prophylaxis G-CSF; antibiotic prophylaxis IVIG G-CSF, Granulocyte-colony stimulating factor; IVIG, Intravenous immunoglobulins. ∗ One episode without relapse.

Marfan syndrome (MFS) is a multi-system connective tissue disorder resulting from mutation in FBN1, the gene encoding fibrillin-1.1 MFS occurs in 1 in 3,000 live births and cardiovascular complications, especially aortopathy, are the leading cause of morbidity and premature mortality.2 Progressive aortic dilation is common with up to 80% of adults having dilation of the aortic root.3 In 1965, Wheat et al demonstrated that the use of reserpine improved survival of patients with aortic dissection.4 Subsequently, Halpern et al demonstrated that β-blocker therapy decreased myocardial contractility in two patients with MFS.5 Since that time, β-blocker therapy began to be used widely in this patient population,6 and remains the first-line therapy for the prevention of aortic complications in MFS.7. However, more recent studies have shown mixed results as to the efficacy of β-blocker therapy in these patients.8 Studies have shown decreased aortic growth rates in MFS patients taking angiotensin converting enzyme inhibitors (ACEI)9 and angiotensin-II receptor blockers.10 We sought to revisit the effects of both ACEI and β-blocker therapy on AGV in patients with MFS.

ts as to the efficacy of β-blocker therapy in these patients.8 Studies have shown decreased aortic growth rates in MFS patients taking angiotensin converting enzyme inhibitors (ACEI)9 and angiotensin-II receptor blockers.10 We sought to revisit the effects of both ACEI and β-blocker therapy on AGV in patients with MFS. Methods We performed a retrospective review of all patients with MFS seen at Arkansas Children’s Hospital between January 1, 1976 and January 1, 2013. Patients with MFS were identified using multiple institutional databases including those from the echocardiography and cardiac catheterization laboratories, the cardiology clinic, all cardiothoracic surgeries, and the Division of Genetics. All available clinical data were reviewed and were recorded. Echocardiograms were performed with the patient in the supine position using commercially available ultrasound machines (Siemens Acuson Sequoia 512 with 10, 7, 5, and 3 MHz probes and Philips iE33 with 12, 8, and 5 MHz probes). Two-dimensional measurements were made in accordance with the recommendations of the American Society of Echocardiography using parasternal long-axis views of the aortic annulus, aortic sinus of Valsalva, sinotubular junction and ascending aorta.11 Measurements were made from inner edge to inner edge during ventricular systole.

sional measurements were made in accordance with the recommendations of the American Society of Echocardiography using parasternal long-axis views of the aortic annulus, aortic sinus of Valsalva, sinotubular junction and ascending aorta.11 Measurements were made from inner edge to inner edge during ventricular systole. The decision to initiate pharmacologic therapy was primarily based on the presence of aortic measurements above the normal range reported by Roman et al12 or accelerated progressive dilation. The selection of a pharmacologic agent and the dose were provider dependent; there were no formal algorithms. After the report from our institution by Yetman et al,9 the use of ACEI as primary therapy at our institution increased. Anthropometric data were used to calculate the body surface area (BSA) at each patient encounter using the Dubois formula.13 A normative control comparison dataset for aortic dimension and growth rate was created by using the calculated BSA of each patient with MFS at each encounter using the formula: aortic root dimension = 24.0(BSA in m2)1/3 + 0.1(Age) – 4.3.14 This normative control dataset was then compared against actual measured aortic dimensions in the patient cohort.

aset for aortic dimension and growth rate was created by using the calculated BSA of each patient with MFS at each encounter using the formula: aortic root dimension = 24.0(BSA in m2)1/3 + 0.1(Age) – 4.3.14 This normative control dataset was then compared against actual measured aortic dimensions in the patient cohort. Statistical Analyses Summary statistics were expressed as frequency and percentage for categorical variables, and as mean ± standard deviation for continuous variables, except for the ages of the treatment groups, which are expressed as mean with first (Q1) and third (Q3) quartiles. To compare aortic growth velocities between medication groups, a mixed model was developed for the aortic dimension as a function of age, medication group (none, β-blocker, ACEI, or normative control), and the interaction between the two. A restricted cubic spline was used for age when fitting the mixed model with regard to the non-linear relationship between aortic dimension and age. A compound symmetry variance matrix was used to take into account the correlated measurements from the same patient. Additional mixed models were fitted for blood pressures and heart rates to assess their differences among three medication groups (none, β-blocker, or ACEI). All the data were analyzed using statistical software SAS 9.4 (SAS Institute Inc., Cary, NC). P-values < 0.05 were considered to indicate statistical significance.

dditional mixed models were fitted for blood pressures and heart rates to assess their differences among three medication groups (none, β-blocker, or ACEI). All the data were analyzed using statistical software SAS 9.4 (SAS Institute Inc., Cary, NC). P-values < 0.05 were considered to indicate statistical significance. Results A total of 67 patients with confirmed MFS were identified (34/67, 51% female). The mean ± SD age at first encounter was 13 ± 10 years, with a mean followup length of 7.6 ± 5.8 years. There were 839 patient encounters with a median number of 10 (range 2–42) encounters per patient. For the patients in the untreated group, the mean age was 9.8 years (Q1: 3.9 years; Q3: 17.2 years), which was lower than either the β-blocker or ACEI groups (p<0.001). For those treated with β-blocker therapy, the mean age was 16.9 years (Q1: 12.1 years; Q3: 22.2 years). β-blockers used included daily atenolol (45.9%), twice daily metoprolol (48.5%), and thrice daily propranolol (5.6%). The mean dose per patient was 0.95 ± 0.63 mg/kg (total daily dose 47.3 ± 23.5mg). The mean systolic (109 ± 16 mmHg) and diastolic (67 ± 8 mmHg) blood pressures were not different from patients who were not treated with medication. The mean heart rate in the β-blocker group (78 ± 19 bpm) was significantly lower compared with those who were untreated (90 ± 24 bpm, p=0.001).

daily dose 47.3 ± 23.5mg). The mean systolic (109 ± 16 mmHg) and diastolic (67 ± 8 mmHg) blood pressures were not different from patients who were not treated with medication. The mean heart rate in the β-blocker group (78 ± 19 bpm) was significantly lower compared with those who were untreated (90 ± 24 bpm, p=0.001). For those treated with ACEI therapy, the mean age was 16.8 years (Q1: 10.7 years; Q3: 24.4 years). ACEIs used included daily lisinopril (12.9%), twice daily enalapril (85.5%), and thrice daily captopril (1.6%). The mean ACEI dose per patient was 0.22 ± 0.1 mg/kg (total daily dose 12.7 ± 6.9 mg). The mean systolic (113 ± 16 vs. 106 ± 20 mmHg, p=0.005) and diastolic (68 ± 10 vs. 64 ± 10 mmHg, p=0.005) blood pressures were significantly higher in the ACEI group compared with the untreated group. The mean heart rate in the ACEI group (83 ± 17 bpm) was significantly lower compared with the untreated group (90 ± 24 bpm, p=0.003). The systolic and diastolic blood pressures were higher in the ACEI group compared with the β-blocker group (p=0.015 and p=0.019, respectively), and there was no difference in heart rate (p=0.696).

eart rate in the ACEI group (83 ± 17 bpm) was significantly lower compared with the untreated group (90 ± 24 bpm, p=0.003). The systolic and diastolic blood pressures were higher in the ACEI group compared with the β-blocker group (p=0.015 and p=0.019, respectively), and there was no difference in heart rate (p=0.696). The aortic size and AGV for the three largest groups (none, ACEI, and β-blocker) were determined (Figure 1) and intergroup comparisons were made, as well as comparisons with the calculated, expected normative control aortic root dimensions (Figure 2). The aortic dimensions were significantly larger in all groups compared with the normative control dataset (p<0.001). At younger ages, the aortic dimensions in the β-blocker group were significantly larger than the ACEI and untreated groups (p<0.01; Figure 1). There was no difference between the ACEI and untreated groups with regard to aortic dimensions at younger ages (p=0.636). The aortic dimensions and growth velocities varied between the four groups. Aortic growth velocity was significantly attenuated in the β-blocker, and nearly approximated that in the normative control dataset. The attenuation of the AGV in the ACEI group was less than that of the β-blocker group compared with those not treated with medication.

d growth velocities varied between the four groups. Aortic growth velocity was significantly attenuated in the β-blocker, and nearly approximated that in the normative control dataset. The attenuation of the AGV in the ACEI group was less than that of the β-blocker group compared with those not treated with medication. A total of 14 patients underwent 17 surgeries. Aortic root replacement was performed in 11/14 patients (79%). The median age at aortic root replacement was 17 years (range 9–35 years). Two patients experienced aortic dissections: one treated with metoprolol experienced acute dissection of the entire thoracic aorta and survived to surgery; the other was treated with enalapril and found to have an asymptomatic dissection of the aortic root. Of the 11 patients who underwent aortic surgery, pharmacologic treatment prior to surgery included: ACEI in 5; β-blocker in 4; and no therapy in 2. One patient in the study group died approximately 2 years after aortic root replacement. Further details of the death could not be obtained.

dissection of the aortic root. Of the 11 patients who underwent aortic surgery, pharmacologic treatment prior to surgery included: ACEI in 5; β-blocker in 4; and no therapy in 2. One patient in the study group died approximately 2 years after aortic root replacement. Further details of the death could not be obtained. Discussion The present study demonstrates a significant attenuation of the AGV in pediatric patients with MFS treated with β-blocker therapy. In fact, the growth velocity was almost identical to the predicted normal growth velocity for the cohort. On average, the patients in the β-blocker group began with a larger aortic dimension compared with the patients in the ACEI and untreated groups, and this difference resolved over time, a finding that may suggest that the early initiation of β-blocker therapy should be considered in patients with MFS. This is a significant finding because aortic wall tension increases as the aortic diameter increases, which results in acceleration of aortic dilation at larger diameters.15 The expected result would then be that the AGV should be faster in the β-blocker group; however, the opposite was found in the β-blocker group indicating a definite mitigating effect on the AGV in patients with MFS.

s the aortic diameter increases, which results in acceleration of aortic dilation at larger diameters.15 The expected result would then be that the AGV should be faster in the β-blocker group; however, the opposite was found in the β-blocker group indicating a definite mitigating effect on the AGV in patients with MFS. Prior investigators have demonstrated decreases in AGV in patients with MFS treated with β-blocker.16–18 Recently, Mueller et al demonstrated a decrease in AGV in patients with MFS who were treated with either β-blockers or angiotensin II receptor blockers.19 However, other studies have suggested that β-blocker therapy does not alter the AGV or clinical outcomes in patients with MFS.20, 21 Gersony et al conducted a meta-analysis of six studies available at the time,22 and concluded that β-blocker therapy was ineffective in patients with MFS.

iotensin II receptor blockers.19 However, other studies have suggested that β-blocker therapy does not alter the AGV or clinical outcomes in patients with MFS.20, 21 Gersony et al conducted a meta-analysis of six studies available at the time,22 and concluded that β-blocker therapy was ineffective in patients with MFS. In the original studies used to conduct their meta-analysis, most of the original authors arrived at different conclusions than Gersony et al. Silverman et al showed that death was twice as common in the untreated versus the β-blocker treated group.23 Those patients treated with β-blocker therapy lived longer (p=0.01). Legget et al reported that aortic complications were more common in the β-blocker treatment group; however, the aortic complication group also had significantly larger aortas at the initiation of the study (p<0.0001).21 Further, in that study patients were dichotomized, without clear delineation as to why, to those who had β-blocker therapy for ≥1 year versus those who had therapy for < 1 year. Similar to the work of Silverman et al, Roman et al reported that in 113 patients with MFS those who had aortic complications had larger aortic size at presentation (p<0.005), were significantly older (p<0.01), and had significantly faster AGV (p<0.05); however, β-blocker therapy use only trended toward being more common in the complication group (86% vs 66%, p=non-significant).15 Those authors stated that the trend toward increased β-blocker use in the group with complications was because they started out with larger aortas, which represents a significant selection bias.15 Salim et al reported that the aorta is dilated at young ages in MFS and showed that the AGV was faster in those who were untreated.24 In that study, the 5 patients treated with β-blocker therapy who underwent surgical intervention had larger aortas at the time of enrollment. In their randomized trial of propranolol therapy in MFS, Shores et al reported in the propranolol group a decreased rate of aortic dilation, fewer deaths, and significantly fewer patients who reached one or more predetermined clinical end-points.17 Finally, Tahernia reported three patients who were treated with β-blocker therapy and three who were not.25 With a mean follow-up of 3.3 years, the β-blocker group had a mean AGV of 0.2 mm/year. The three untreated patients had a mean follow-up of 3 years with a mean AGV of 1.4 mm/year.

etermined clinical end-points.17 Finally, Tahernia reported three patients who were treated with β-blocker therapy and three who were not.25 With a mean follow-up of 3.3 years, the β-blocker group had a mean AGV of 0.2 mm/year. The three untreated patients had a mean follow-up of 3 years with a mean AGV of 1.4 mm/year. In the consideration of studies evaluating AGV, it is imperative to consider the size of the aorta at the time of the initiation of pharmacotherapy. In accordance with the law of Laplace, when wall thickness and blood pressure are stable, wall tension in the aorta increases as the aortic diameter increases,26 resulting in acceleration of aortic dilation at larger diameters.15 To make a truly meaningful comparison between therapeutic groups, either both groups have to begin with the same diameters or the group that is shown to have a decreased growth rate must have begun at a larger diameter, thus showing a diminution of AGV. If the aorta is larger in one group than the other, those patients are beginning the study at a different point in the aortic disease process than the group with the smaller diameter. The present study is not subject to this concern because those patients treated with β-blocker therapy began, on average, with a larger aortic diameter than either the ACEI group or the untreated group and were still demonstrated to have a lower AGV.

the aortic disease process than the group with the smaller diameter. The present study is not subject to this concern because those patients treated with β-blocker therapy began, on average, with a larger aortic diameter than either the ACEI group or the untreated group and were still demonstrated to have a lower AGV. The difference in results between the β-blocker and ACEI groups in the present study seem to support the hypothesis that a decrease in myocardial contractility is the explanation for decreased AGV in patients with MFS treated with β-blockers. In the present study, there were no clinically significant differences in blood pressure or heart rate between the β-blocker and ACEI groups; however, the AGV in the β-blocker group was significantly less than either the ACEI or untreated groups. However, other explanations cannot be ruled out, such as heretofore-unknown pleiotropic effects of β-blockers. Yetman et al demonstrated that there was a decrease in the AGV and improvement in the aortic distensibility in patients with MFS treated with ACEI versus those treated with β-blocker.9 The present study does not corroborate those findings. In the study by Yetman et al, the subject ages and the doses of both medications were similar to those in the present study. The length of follow-up was shorter in that prior study (3.0 ± 0.2 years) compared with the present study (7.6 ± 5.8 years).

ith β-blocker.9 The present study does not corroborate those findings. In the study by Yetman et al, the subject ages and the doses of both medications were similar to those in the present study. The length of follow-up was shorter in that prior study (3.0 ± 0.2 years) compared with the present study (7.6 ± 5.8 years). The present study is limited by its retrospective nature. The lack of a true control group also is a limitation. However, a previously published and validated formula for the determination of predicted aortic size was used in combination with the anthropometric data from the present study cohort to create a predicted “normal” aortic growth curve, which allowed for comparison of the study measurements against a unique dataset specific to the study cohort. The patients in the non-treatment group were younger than those in either the β-blocker or ACEI group. An older population would be expected to have an increased AGV, which if anything would blunt any effect of the medication on slowing AGV. These findings suggest that β-blocker therapy may be more beneficial. Early introduction of β-blocker therapy in patients with MFS should be considered particularly even prior to the demonstration of aortic dilation.

The present study is limited by its retrospective nature. The lack of a true control group also is a limitation. However, a previously published and validated formula for the determination of predicted aortic size was used in combination with the anthropometric data from the present study cohort to create a predicted “normal” aortic growth curve, which allowed for comparison of the study measurements against a unique dataset specific to the study cohort. The patients in the non-treatment group were younger than those in either the β-blocker or ACEI group. An older population would be expected to have an increased AGV, which if anything would blunt any effect of the medication on slowing AGV. These findings suggest that β-blocker therapy may be more beneficial. Early introduction of β-blocker therapy in patients with MFS should be considered particularly even prior to the demonstration of aortic dilation. This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final citable form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain. The authors declare no conflicts of interest.

This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final citable form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain. The authors declare no conflicts of interest. Acknowledgments The authors wish to than Bruce S. Alpert, MD (previously at the University of Tennessee College of Medicine; currently a case reviewer for the health care industry), for his thoughtful review and comments on the manuscript. Supported in part by National Center for Advancing Translational Sciences/National Institutes of Health (1 UL1 RR029884). Figure 1 Aortic growth in in patients with Marfan syndrome: (A) without treatment; (B) treated with beta-blocker; and (C) treated with ACE-inhibitor. (D) Aortic growth in the predicted, normative control dataset for the cohort. Figure 2 Comparison of aortic growth velocities over time in treated and untreated groups and in the predicted, normative control dataset.

Children with eczema and asthma have been reported to have poorer mental well-being than healthy children; childhood asthma has been associated with anxiety, depression, emotional and behavioral problems, and treatment for a mental health problem,1-9 and childhood eczema with increased emotional problems and a higher risk of attention deficit/hyperactivity disorder (ADHD).10-14 Childhood eczema and asthma are also associated with maternal anxiety and depression; mothers who are anxious or depressed during pregnancy, the postpartum period, and beyond are at increased risk of having a child with asthma or wheezing.15-18 There is also evidence of reverse causation; caring for a child with eczema or asthma is associated with higher levels of anxiety, depression, sleep deprivation, and reduced quality of life for the child's parents and family.19-24 Therefore, as children with eczema and asthma are more likely to have a mother with anxiety or depression compared with healthy children, maternal mental health may be an important mediator of the association between eczema and asthma and child mental well-being. A better understanding of the role of maternal mental health could help elucidate the mechanisms that underpin the association between eczema and asthma and child mental well-being, and if causal, could aid the development of interventions to improve the mental and physical health of children affected by these conditions.25

ter understanding of the role of maternal mental health could help elucidate the mechanisms that underpin the association between eczema and asthma and child mental well-being, and if causal, could aid the development of interventions to improve the mental and physical health of children affected by these conditions.25 Few previous studies on the mental well-being of children with eczema or asthma have considered maternal mental health. Additional limitations to the existing literature include the frequent use of small, clinic-based, convenience samples, a lack of inclusion of potentially important confounders, and few studies focusing on younger children.11,26,27 Few studies have considered both eczema and asthma despite the frequent co-occurrence of these conditions. To address these limitations, we used data from a large, longitudinal, population-based birth cohort. We hypothesized that maternal anxiety and depression would mediate the association between rash and wheeze and child mental well-being (Figure; available at www.jpeds.com). Our first aim was to determine whether rash or wheeze symptoms were associated with internalizing (ie, anxious and depressive) or externalizing (ie, oppositional and hyperactive) behaviors at the age of 8 years, and whether duration of symptoms was an important factor. Our second aim was to determine whether these associations remained after adjustment for maternal anxiety and depression.

ociated with internalizing (ie, anxious and depressive) or externalizing (ie, oppositional and hyperactive) behaviors at the age of 8 years, and whether duration of symptoms was an important factor. Our second aim was to determine whether these associations remained after adjustment for maternal anxiety and depression. Methods Subjects were participants in the Avon Longitudinal Study of Parents and Children (ALSPAC). Details on ALSPAC have been published previously,28 and a fully searchable data dictionary is available online (www.bristol.ac.uk/alspac/). In brief, ALSPAC recruited pregnant women with expected dates of delivery between April 1, 1991 and December 31, 1992 who lived in a defined geographic area (Avon, United Kingdom). The children have been studied throughout their lives using maternal or self-report questionnaires and, from the age of 7 years, approximately annual research clinic visits. There were 14 062 live births, 13 988 children were alive at 1 year and over 7000 mothers completed the 8-year questionnaire. The study sample in this paper comprises the 7250 singletons with outcome data at 8 years of age. Ethical approval for the study was obtained from the ALSPAC Ethics and Law Committee and the Local Research Ethics Committees.

13 988 children were alive at 1 year and over 7000 mothers completed the 8-year questionnaire. The study sample in this paper comprises the 7250 singletons with outcome data at 8 years of age. Ethical approval for the study was obtained from the ALSPAC Ethics and Law Committee and the Local Research Ethics Committees. Measures Exposure - Child Rash, Wheeze, and Atopy Child rash and wheeze symptoms were reported in mailed, self-completion questionnaires sent to the mother when her child was aged 6 months, 18 months, 2 years 6 months, 3 years 6 months, 4 years 9 months, 5 years 9 months, 6 years 9 months, and 7 years 7 months (Table I; available at www.jpeds.com). When questionnaires contained more than 1 question about wheeze or rash, a symptom was coded as being present if the mother reported yes to any 1 of the questions. The time periods covered by the questionnaires were categorized as infancy (up to 18 months), preschool (18 months up to 4 years 9 months), and school age (4 years 9 months up to 7 years 7 months). For each time period, symptoms were coded as being present if the mother reported that the child had the symptom in at least 1 questionnaire. For rash and wheeze separately, symptoms were categorized as ‘none’ (no symptoms in any time period); ‘early onset, transient’ (symptoms in infancy and/or preschool only); ‘persistent’ (symptoms in infancy and/or preschool, and at school age); and ‘late onset’ (symptoms at school age only). Atopy status was known for 5004 children in our sample who attended an ALSPAC clinic when aged 7.5 years; atopy was determined by skin prick testing and defined as a positive response (≥2 mm weal) to any one of Dermatophagoides pteronyssinus, grass, or cat allergen with a negative response to diluent solution. As previously reported, this definition identified >95% of subjects with any positive response to a wider panel of allergens.29

ermined by skin prick testing and defined as a positive response (≥2 mm weal) to any one of Dermatophagoides pteronyssinus, grass, or cat allergen with a negative response to diluent solution. As previously reported, this definition identified >95% of subjects with any positive response to a wider panel of allergens.29 Outcome—Child Mental Well-Being Child mental well-being at age 8 years 1 month was maternally reported using the parental version of the Goodman Strengths and Difficulties Questionnaire, a validated behavioral screening tool for children and adolescents.30 The peer and emotion subscales can be summed to give an internalizing problems score (range 0-20), and the hyperactivity and conduct subscales summed to give an externalizing problems score (range 0-20).31 All 4 subscales can be summed to give a total difficulties score (TDS) (range of 0-40); a higher score reflects more difficulties. For each of the TDS, externalizing and internalizing scores, children were classified as having a high score or not (high defined as >90th percentile). The cut-offs for a high score were ≥7 for internalizing, ≥10 for externalizing, and ≥15 for TDS.

score (TDS) (range of 0-40); a higher score reflects more difficulties. For each of the TDS, externalizing and internalizing scores, children were classified as having a high score or not (high defined as >90th percentile). The cut-offs for a high score were ≥7 for internalizing, ≥10 for externalizing, and ≥15 for TDS. Other Variables Maternal anxiety and depression were measured in pregnancy and at the outcome time-point (when child was 8 years old). Depression was measured by the Edinburgh Postnatal Depression Scale. Although this measure was originally designed for use with postnatal women, none of the 10 items is specific to this period and it has been validated for use at other times32; it was chosen as it does not contain somatic items that could confound normal symptoms in pregnancy with depression. Anxiety was measured in pregnancy by the 8 items of the anxiety subscale of the Crown-Crisp Experiential Index33 and 8 years later, by the 20-trait anxiety items of the State-Trait Anxiety Inventory.34 Quartiles of depression and anxiety scores were calculated for use in analyses as they were not normally distributed.

Anxiety was measured in pregnancy by the 8 items of the anxiety subscale of the Crown-Crisp Experiential Index33 and 8 years later, by the 20-trait anxiety items of the State-Trait Anxiety Inventory.34 Quartiles of depression and anxiety scores were calculated for use in analyses as they were not normally distributed. A number of other child, mother, and socioeconomic variables thought to be potential confounders or mediators (based on previous literature or on theoretical grounds) were also included in analyses. Child variables were maternally reported: sex; ethnicity (White, non-White [no further disaggregation was possible due to small numbers]); age at outcome; and child wakes at night (no, once, twice or more). Socioeconomic position (SEP) was reported during pregnancy: highest maternal education (university degree, A level, O level, vocational/none); housing tenure (owned/mortgaged, privately rented, council rented, other); and financial difficulties (quartiles of score with range 0-40, where 0 is no financial difficulties). Maternal smoking (no, yes) was reported during pregnancy and at the outcome time point, and maternal insufficient sleep (no, yes) when the child was aged 7 years.

d/mortgaged, privately rented, council rented, other); and financial difficulties (quartiles of score with range 0-40, where 0 is no financial difficulties). Maternal smoking (no, yes) was reported during pregnancy and at the outcome time point, and maternal insufficient sleep (no, yes) when the child was aged 7 years. Missing Data Multiple imputation using chained equations was used to replace missing exposure and confounder data with predictions based on information observed in the sample. The percentage of missing data was below 10% for most variables (Table II; available at www.jpeds.com); 47% of the children had complete data, with another 40% having between 1 and 4 missing variables. Twenty imputed datasets were created and analyzed using mi estimate commands in Stata 13 (StataCorp, College Station, Texas). Complete case analysis was also performed; results were consistent with those from the imputed data and are available from authors on request. Atopy data were not imputed; a separate complete case analysis was performed for this exposure.

lyzed using mi estimate commands in Stata 13 (StataCorp, College Station, Texas). Complete case analysis was also performed; results were consistent with those from the imputed data and are available from authors on request. Atopy data were not imputed; a separate complete case analysis was performed for this exposure. Statistical Analyses χ2 tests were used to assess whether distributions of the outcome and confounder variables differed by rash or wheeze status. To determine whether maternal mental health could be a mediator in the relationship between child rash or wheeze and child mental well-being, we first used multivariable logistic regression models to test for an association between the exposure and the potential mediator (child rash/wheeze and maternal anxiety/depression) and between the potential mediator and the outcome (maternal anxiety/depression and child mental well-being).35 Multivariable logistic regression models were then used to model the association between child rash and wheeze and the child mental well-being outcomes. Eight multivariable logistic regression models were fitted for each binary outcome (high TDS, high internalizing problems, high externalizing problems). All models adjusted for baseline confounders (child demographics, SEP, mother's age at delivery). Model 1 added rash and wheeze individually to these baseline confounders; model 2 then included both rash and wheeze together. Models 3 and 4 added maternal anxiety and depression, in pregnancy and at the outcome time point, respectively, to the variables in model 2. Model 5 adjusted for maternal anxiety and depression at both time points. Finally, we included maternal smoking (model 6), maternal sleep (model 7), and child sleep (model 8). Interaction terms were fitted to test whether the relationship between rash/wheeze and the child mental well-being outcomes differed by child sex; these were not significant and so models were adjusted for sex but not stratified. We also tested for interactions between rash and wheeze, and between maternal mental health and rash and wheeze; none of these were significant and so are not included in the models presented.

l well-being outcomes differed by child sex; these were not significant and so models were adjusted for sex but not stratified. We also tested for interactions between rash and wheeze, and between maternal mental health and rash and wheeze; none of these were significant and so are not included in the models presented. Results The 7250 children (3681 boys and 3569 girls) had a mean age of 8.2 years at outcome assessment. Compared with the excluded sample (children with no outcome data), study children were more likely to be of higher SEP and to have ever had a rash. The mothers of the study children reported lower anxiety and depression than the mothers of the excluded sample and were less likely to smoke (Table III; available at www.jpeds.com).

ith the excluded sample (children with no outcome data), study children were more likely to be of higher SEP and to have ever had a rash. The mothers of the study children reported lower anxiety and depression than the mothers of the excluded sample and were less likely to smoke (Table III; available at www.jpeds.com). In the study sample, more mothers reported that their child had ever had a rash (71%) than wheeze (46%). The patterning of symptoms by maternal SEP differed for rash and wheeze (Table IV). Mothers whose child had wheeze were more likely to have reported financial difficulties and were less likely to have owned their own home. In contrast, mothers whose child had early onset rash were the most likely to be educated to degree level. Children with a persistent rash were the most likely to have had wheeze, and children with wheeze at school age were the most likely to have had a rash. Children with wheeze were more likely to have a mother who smoked, but there was no association between maternal smoking and child rash status (Table IV). Children with wheeze and rash at school age were the most likely to wake during the night (Table IV). Maternal sleep was correlated with child sleep; of mothers who reported insufficient sleep, 21% said their child woke at least once at night compared with 14% of mothers who reported sufficient sleep. Maternal anxiety and depression scores, in pregnancy and at the outcome time point, were associated with child rash and wheeze and with child externalizing, internalizing, and TDS (Tables IV and V and Figure; Table V available at www.jpeds.com).

once at night compared with 14% of mothers who reported sufficient sleep. Maternal anxiety and depression scores, in pregnancy and at the outcome time point, were associated with child rash and wheeze and with child externalizing, internalizing, and TDS (Tables IV and V and Figure; Table V available at www.jpeds.com). Internalizing and externalizing continuous scores were moderately correlated (r = 0.39). However, most of the children with a high score on 1 measure did not have a high score on the other measure; 482 children had a high externalizing score only, 548 children had a high internalizing score only, and 257 had both a high externalizing and internalizing score. One-fifth of the children (20.4%) who had a skin prick test had a positive reaction, indicating atopy. Children with early onset transient rash (OR 1.61, 95% CI, 1.29-2.01) and persistent rash (3.31, 2.70-4.05) were more likely to be atopic than children who had never had a rash; there was no difference for late onset rash. For wheeze, those with persistent (4.49, 3.74-5.40) and late onset (3.24, 2.38-4.41) wheeze were more likely to be atopic than those with never wheeze, but there was no difference for early onset transient wheeze.

e likely to be atopic than children who had never had a rash; there was no difference for late onset rash. For wheeze, those with persistent (4.49, 3.74-5.40) and late onset (3.24, 2.38-4.41) wheeze were more likely to be atopic than those with never wheeze, but there was no difference for early onset transient wheeze. Children with rash had elevated odds of having a high internalizing, externalizing, and TDS relative to those who had never had a rash. ORs were higher for externalizing problems than internalizing problems, and were higher for those with a rash at school age (whether persistent or late onset) compared with those with early onset transient symptoms (Table VI, model 1). Children with persistent wheeze, and to a lesser extent early onset transient wheeze, had higher odds of internalizing and externalizing problems, and of having a high TDS, than children who had never wheezed. There was no association between late onset wheeze and any of the mental well-being outcomes. Rash and wheeze were associated with the outcomes independent of each other, although including both in the same model did attenuate associations for persistent rash and persistent wheeze for all outcomes (Table VI, model 2).

There was no association between late onset wheeze and any of the mental well-being outcomes. Rash and wheeze were associated with the outcomes independent of each other, although including both in the same model did attenuate associations for persistent rash and persistent wheeze for all outcomes (Table VI, model 2). Maternal anxiety and depression when the child was 8 years old, but not during pregnancy, accounted for the association between child rash and internalizing symptoms and partly for the association with externalizing symptoms (Table VI, models 3 and 4). On the other hand, maternal anxiety and depression during pregnancy, rather than in childhood, appeared more relevant in accounting for the association between internalizing and externalizing symptoms and child persistent wheeze. Further adjustment for maternal smoking or maternal sleep did not modify these associations (Table VI, model 6 and 7). Child sleep explained some of the association between the outcomes and persistent wheeze, but not rash (Table VI, model 8). In the subsample of children who had their atopy status determined by skin prick test (n = 5004), there was no association between atopy and the odds of having a high internalizing (OR 0.91, 95% CI, 0.72-1.15), externalizing (0.91, 0.72-1.16), or total difficulties (1.03, 0.82-1.30) score in age-adjusted models. Complete confounder data was available for 3584 of these children; adjustment made little difference to the associations observed (Table VII; available at www.jpeds.com).

gh internalizing (OR 0.91, 95% CI, 0.72-1.15), externalizing (0.91, 0.72-1.16), or total difficulties (1.03, 0.82-1.30) score in age-adjusted models. Complete confounder data was available for 3584 of these children; adjustment made little difference to the associations observed (Table VII; available at www.jpeds.com). Discussion Both rash and wheeze status were related to child mental well-being at the age of 8 years in our large, population-based cohort study; persistent and late onset rash, and persistent wheeze, were associated with both internalizing and externalizing symptoms. These results confirm, in a large prospective study, previous results of increased internalizing5,10,14,26,27 and externalizing11,36 problems in children with eczema and asthma. We have demonstrated the importance of including measures of maternal anxiety and depression, both during and after pregnancy, when examining the relationship between rash and wheeze and child mental well-being. For persistent and late onset rash, we found that maternal anxiety and depression, particularly after the birth of the child rather than during pregnancy, accounted for the association with internalizing symptoms and partly for externalizing symptoms. Worse sleep did not mediate the association. However, maternal anxiety and depression at both time points attenuated the association with persistent wheeze. Worse child sleep also contributed to the association with persistent wheeze.

or the association with internalizing symptoms and partly for externalizing symptoms. Worse sleep did not mediate the association. However, maternal anxiety and depression at both time points attenuated the association with persistent wheeze. Worse child sleep also contributed to the association with persistent wheeze. In our study, children who only had symptoms in infancy or preschool tended to have poorer mental well-being at 8 years compared with those with no symptoms at any age; early onset transient wheeze was associated with more internalizing problems, and early onset transient rash with more externalizing problems. Similarly, a large German cohort study found that children who cleared their eczema by age 2 years tended to have more emotional and conduct problems at age 10 years than children who had never had eczema.14 We found generally similar sized associations for persistent and late onset rash, which contrasts with the German study which found that emotional problems were greater the longer eczema persisted14; age at outcome and categories of symptom duration differed between the studies, which may explain the different findings. We mutually adjusted for rash and wheeze status to account for the fact that these conditions often coexist, and found both symptoms to be independently associated with the outcomes. There is limited evidence from previous literature on the relative impact of eczema and asthma on child mental well-being, and findings are mixed. In the German study, eczema was associated with mental well-being independent of atopic comorbidity but associations for asthma were not reported.14 In 4 other studies that included measures of both eczema and asthma, all of which had ADHD as their outcome, 3 found only eczema was independently associated with ADHD in fully adjusted models,12,13,36 but another study reported that asthma, but not eczema, was associated with ADHD.37

ns for asthma were not reported.14 In 4 other studies that included measures of both eczema and asthma, all of which had ADHD as their outcome, 3 found only eczema was independently associated with ADHD in fully adjusted models,12,13,36 but another study reported that asthma, but not eczema, was associated with ADHD.37 Our mediation model assumes that child rash and wheeze predict maternal anxiety and depression, which in turn co-occur with poor child mental well-being (ideally child mental well-being would have been measured after maternal anxiety and depression in order to truly establish mediation). As maternal anxiety and depression in pregnancy occur before child rash and wheeze develop, they cannot mediate the association; however, child rash and wheeze could be a mechanism through which maternal depression and anxiety in pregnancy predict later child well-being. A true mediating mechanism could involve both behavioral and biological pathways. Caregivers with poor mental health feel less empowered to deal with their child's condition or to manage it effectively, are less knowledgeable about their child's medications, and their children are at increased risk of poor medication adherence, increased hospitalization, and lower asthma-related quality of life.38-42 Therefore, poor mental health may impair a mother's ability to cope with the demands of her child's condition, including administering the appropriate treatment regimes, resulting in a worsening of her child's physical symptoms.43

tion adherence, increased hospitalization, and lower asthma-related quality of life.38-42 Therefore, poor mental health may impair a mother's ability to cope with the demands of her child's condition, including administering the appropriate treatment regimes, resulting in a worsening of her child's physical symptoms.43 Many children with asthma and eczema, and consequently their parents, suffer from disrupted sleep,21,44-46 and previous studies have reported that sleep quality is a key factor in the association between eczema, asthma, and behavioral problems in children.13,44,47,48 In the present study, maternal and child sleep were related, but only child sleep attenuated the association between wheeze and internalizing and externalizing problems. Maternal mental health may, therefore, have direct effects on a child's mental health, as well as indirect effects via poor disease management resulting in a worsening of symptoms, which in turn leads to disrupted sleep and an increase in behavioral difficulties. However, such pathways are likely to be cyclical as poor mental well-being in asthmatic children is associated with increased disease severity over time.49 Furthermore, a child's physical symptoms may affect how the mother parents her child; studies have found perceived child vulnerability50 and negative parenting behaviors51 to mediate the association between parental depression and child internalizing symptoms in children with asthma. This highlights the complexity of the relationships between child physical symptoms and child and maternal mental health.

studies have found perceived child vulnerability50 and negative parenting behaviors51 to mediate the association between parental depression and child internalizing symptoms in children with asthma. This highlights the complexity of the relationships between child physical symptoms and child and maternal mental health. In addition to the potential behavioral pathways described above, there may be biological pathways linking eczema and asthma to mental well-being. Eczema and asthma in children are both associated with stress, which may be mediated by common underlying immunologic mechanisms, such as atopy and systemic inflammatory responses.11 It is thought that systemic inflammatory responses and stress could influence behavior and emotions via effects on the brain.11,52 Maternal mental health could play a role in such a biological pathway, as parental stress and depression have been reported to predict increases in a child's inflammatory profile.41 There was no association between atopy and child mental well-being in our study, suggesting that associations between rash and wheeze and mental well-being were not mediated by allergy; however, we did not have relevant biomarkers of systemic inflammation to explore this pathway further.

child's inflammatory profile.41 There was no association between atopy and child mental well-being in our study, suggesting that associations between rash and wheeze and mental well-being were not mediated by allergy; however, we did not have relevant biomarkers of systemic inflammation to explore this pathway further. Rather than a mediating mechanism, it is possible in this observational analysis that the attenuating effect of maternal mental health is due to confounding. For example, lower SEP is associated with both maternal mental health and more child wheeze. However, we found that the role of maternal mental health remained after adjustment for three measures of SEP. Furthermore, confounding by SEP is not likely given the association between maternal SEP and child rash is in the opposite direction. Mothers with poorer mental health are more likely to smoke, and maternal smoking is associated with both child asthma and eczema, and child mental health.53 However, in this study maternal smoking did not explain any of the excess risk of behavioral problems associated with child rash or wheeze.

the opposite direction. Mothers with poorer mental health are more likely to smoke, and maternal smoking is associated with both child asthma and eczema, and child mental health.53 However, in this study maternal smoking did not explain any of the excess risk of behavioral problems associated with child rash or wheeze. A further potential explanation is reporting bias. A mother's mental health may influence how she perceives and reports both her child's symptoms and behavior.49 This is important as the majority of studies, including ours, use maternal reports. All of our measures, except for atopy, were maternally reported thereby raising the possibility of bias from shared method variance. Finally, associations could be due to reverse causality as maternal mental health may be reduced as a result of the child having poor mental health. We have not been able to test for reverse causality in our study, and it is difficult to disentangle these potential mechanisms in observational studies.

method variance. Finally, associations could be due to reverse causality as maternal mental health may be reduced as a result of the child having poor mental health. We have not been able to test for reverse causality in our study, and it is difficult to disentangle these potential mechanisms in observational studies. If maternal anxiety and depression have a true mediating role, treatment to improve maternal mental health could optimize the well-being of children with asthma and eczema. The idea of ‘treating the parents to heal the child’ has been proposed for children with asthma and eczema, but a better understanding of the role of parental psychological well-being in child health is needed.25 Although there is some limited evidence that interventions, such as cognitive behavioral therapy for anxiety disorders, can improve asthma-related quality of life in both adults54 and children,55 we are unaware of any intervention to date that has examined the impact of improving maternal anxiety and depression on the well-being of children with asthma or eczema.

interventions, such as cognitive behavioral therapy for anxiety disorders, can improve asthma-related quality of life in both adults54 and children,55 we are unaware of any intervention to date that has examined the impact of improving maternal anxiety and depression on the well-being of children with asthma or eczema. Strengths of this current study include the use of a large, longitudinal, population-based cohort, and the inclusion of several potentially important confounders. In addition, we included both rash and wheeze, and demonstrated the importance of accounting for the duration of symptoms in the association of rash and wheeze with later mental health. Our study also has limitations. We had no objective measure of rash and wheeze severity, and not all reported rash and wheeze will be due to eczema or asthma, which may limit comparison with previous studies. However, the wheeze questions are consistent with those used by the International Study of Asthma and Allergies in Childhood and are a reliable way of diagnosing asthma in large epidemiologic studies where direct physician assessment would not be feasible.56 Also, a subset of the children had their skin assessed at an ALSPAC clinic when aged 49 months; observed eczema was associated with maternally reported child rash. It is possible that some children developed symptoms in the 6-month period between the last measure of rash/wheeze status (at age 7 years 7 months) and the outcome measures (at age 8 years) and so may have had their symptom status misclassified; however, such errors would lead to our estimates being conservative. Child sleep quality was measured at age 6 years 9 months and maternal sleep at age 7 years 1 month; these variables may not reflect sleep quality at the outcome time point, particularly if rash or wheeze symptoms changed after sleep quality was measured. Lastly, in this study we have only considered child mental well-being at age 8 years; we have not considered how mental well-being, or its associations with rash and wheeze, develop over time.

lect sleep quality at the outcome time point, particularly if rash or wheeze symptoms changed after sleep quality was measured. Lastly, in this study we have only considered child mental well-being at age 8 years; we have not considered how mental well-being, or its associations with rash and wheeze, develop over time. In conclusion, maternal anxiety and depression explained some of the association between child rash and wheeze and externalizing and internalizing symptoms in childhood. Further research would help determine whether this is because maternal mental health mediates the association, or whether it is due to reporting bias, confounding or reverse causality. A better understanding of the role of maternal mental health could help in the development of interventions to improve the quality of life of affected children and their families. Appendix Figure Theoretical model of role of maternal anxiety and depression in association between child allergic symptoms and child mental well-being. Table I ALSPAC questionnaire items on rash and wheeze

In conclusion, maternal anxiety and depression explained some of the association between child rash and wheeze and externalizing and internalizing symptoms in childhood. Further research would help determine whether this is because maternal mental health mediates the association, or whether it is due to reporting bias, confounding or reverse causality. A better understanding of the role of maternal mental health could help in the development of interventions to improve the quality of life of affected children and their families. Appendix Figure Theoretical model of role of maternal anxiety and depression in association between child allergic symptoms and child mental well-being. Table I ALSPAC questionnaire items on rash and wheeze Child's age when questionnaire sent to mother “Has your child had….” (since birth for 6-month questionnaire, since the previous questionnaire time point for all others) Rash 6 months, 18 months 1. A rash in the joints and creases of his/her body (eg, behind the knees, under the arms)? 2. An itchy, dry, oozing or crusted rash on the face, forearms or shins? 2 years 6 months, 3 years 6 months, 4 year 9 months, 5 years 9 months, 6 years 9 months 1. An itchy, dry skin rash in the joints and creases of his/her body (eg, behind the knees, elbows, under the arms)? 2. An itchy, dry rash on his hands? 3. An itchy, dry rash on his feet? 7 years 7 months 1. A rash? Wheeze 6 months, 18 months, 2 years 6 months, 3 years 6 months, 4 years 9 months, 5 years 9 months, 6 years 9 months 1. Wheezing with whistling on his chest when he/she breathed? 2. Wheezing? 7 years 7 months 1. Wheezing? Table II Description of missing data for each of the exposure and confounder variables

1. A rash? Wheeze 6 months, 18 months, 2 years 6 months, 3 years 6 months, 4 years 9 months, 5 years 9 months, 6 years 9 months 1. Wheezing with whistling on his chest when he/she breathed? 2. Wheezing? 7 years 7 months 1. Wheezing? Table II Description of missing data for each of the exposure and confounder variables n (%) missing (7250 = 100%)∗ Child sex 0 Mother's age at delivery 0 Child ethnic group 295 (4.1) Housing tenure 186 (2.6) Maternal education 184 (2.5) Financial difficulties 363 (5.0) Anxiety in pregnancy 719 (9.9) Depression in pregnancy 624 (8.6) Anxiety when child aged 8 years 536 (7.4) Depression when child aged 8 years 139 (1.9) Maternal smoking in pregnancy 63 (0.87) Maternal smoking when child 8 years 27 (0.37) Maternal sleep 701 (9.7) Child sleep 800 (11.0) Wheeze 6 months 323 (4.5) 18 months 310 (4.3) 2 years 6 months 490 (6.8) 3 years 6 months 484 (6.7) 4 years 9 months 495 (6.8) 5 years 9 months 702 (9.7) 6 years 9 months 740 (10.2) 7 years 7 months 599 (8.3) Rash 6 months 374 (5.2) 18 months 321 (4.4) 2 years 6 months 547 (7.5) 3 years 6 months 480 (6.6) 4 years 9 months 886 (12.2) 5 years 9 months 1126 (15.5) 6 years 9 months 980 (13.5) 7 years 7 months 610 (8.4) ∗ The study sample was defined as singleton children with complete outcome data, n = 7250. Table III Comparison of study population and those with incomplete data

n (%) missing (7250 = 100%)∗ Child sex 0 Mother's age at delivery 0 Child ethnic group 295 (4.1) Housing tenure 186 (2.6) Maternal education 184 (2.5) Financial difficulties 363 (5.0) Anxiety in pregnancy 719 (9.9) Depression in pregnancy 624 (8.6) Anxiety when child aged 8 years 536 (7.4) Depression when child aged 8 years 139 (1.9) Maternal smoking in pregnancy 63 (0.87) Maternal smoking when child 8 years 27 (0.37) Maternal sleep 701 (9.7) Child sleep 800 (11.0) Wheeze 6 months 323 (4.5) 18 months 310 (4.3) 2 years 6 months 490 (6.8) 3 years 6 months 484 (6.7) 4 years 9 months 495 (6.8) 5 years 9 months 702 (9.7) 6 years 9 months 740 (10.2) 7 years 7 months 599 (8.3) Rash 6 months 374 (5.2) 18 months 321 (4.4) 2 years 6 months 547 (7.5) 3 years 6 months 480 (6.6) 4 years 9 months 886 (12.2) 5 years 9 months 1126 (15.5) 6 years 9 months 980 (13.5) 7 years 7 months 610 (8.4) ∗ The study sample was defined as singleton children with complete outcome data, n = 7250. Table III Comparison of study population and those with incomplete data Sample∗ (n = 7250) Complete case† (n = 5513) Excluded sample‡ (N = 6367) Imputed Unimputed N = 7250 Child high internalizing problems score High (%) 11.1 11.1 10.6 / N = 7250 Child high externalizing problems score High (%) 10.2 10.2 9.6 / N = 7250 Child high TDS High (%) 10.5 10.5 9.7 / N = 6934 N = 2246 Child rash§ None (%) 29.0 30.5 30.8 37.4 Early onset transient (%) 29.9 30.9 30.6 34.7 Early onset persistent (%) 33.4 31.3 31.2 22.8 Late onset (%) 7.6 7.3 7.4 5.1 N = 6958 N = 2296 Child wheeze§ None (%) 53.6 54.8 55.4 53.5 Early onset transient (%) 28.0 27.6 27.2 30.5 Early onset persistent (%) 14.1 13.5 13.3 13.0 Late onset (%) 4.4 4.1 4.1 3.0 N = 7250 N = 6367 Child sex Female (%) 49.2 49.2 48.9 47.5 N = 6955 N = 4814 Child ethnic group White (%) 96.3 96.4 96.9 92.9 N = 7250 N = 6367 Mother's age at delivery Mean (years) 29.1 29.1 29.3 26.7 N = 7064 N = 5639 Housing tenure Owned/mortgaged (%) 82.0 82.3 85.2 62.0 Privately rented (%) 5.7 5.6 4.8 9.3 Council rented (%) 9.4 9.2 7.8 24.4 Other (%) 3.0 2.9 2.2 4.3 N = 7066 N = 5036 Maternal education Degree (%) 16.4 16.6 17.1 7.6 A-level (%) 26.1 26.3 27.3 17.3 O-Level (%) 35.3 35.2 35.7 33.7 None/Vocational (%) 22.3 22.0 19.8 41.3 N = 6531 N = 4935 Anxiety in pregnancy Lowest quartile (%) 31.6 31.8 32.8 28.3 Highest quartile (%) 19.1 18.9 18.2 24.8 N = 6626 N = 5034 Depression in pregnancy Lowest quartile (%) 22.0 22.2 22.8 17.2 Highest quartile (%) 24.3 23.9 22.9 33.0 N = 6714 N = 212 Anxiety when child aged 8 years Lowest quartile (%) 25.3 25.6 26.1 25.9 Highest quartile (%) 24.3 23.9 23.0 28.3 N = 7111 N = 253 Depression when child aged 8 years Lowest quartile (%) 30.0 30.0 31.0 29.6 Highest quartile (%) 23.5 23.5 21.9 27.3 N = 7187 N = 5861 Maternal smoking in pregnancy Yes (%) 20.3 20.2 18.7 35.2 N = 7223 N = 5498 N = 294 Maternal smoking when child 8 years Yes (%) 19.4 19.4 18.1 22.8 N = 6549 N = 5165 N = 1514 Maternal sleep Insufficient sleep (%) 38.0 38.0 37.7 39.8 N = 6450 N = 5121 N = 1639 Child wakes at night Never (%) 83.2 83.6 84.3 82.3 Once (%) 14.3 14.2 13.8 14.0 Twice or more (%) 2.4 2.2 1.8 3.8 ∗ The study sample was defined as singleton children with complete outcome data, n = 7250.

2.8 N = 6549 N = 5165 N = 1514 Maternal sleep Insufficient sleep (%) 38.0 38.0 37.7 39.8 N = 6450 N = 5121 N = 1639 Child wakes at night Never (%) 83.2 83.6 84.3 82.3 Once (%) 14.3 14.2 13.8 14.0 Twice or more (%) 2.4 2.2 1.8 3.8 ∗ The study sample was defined as singleton children with complete outcome data, n = 7250. Missing covariates were imputed. Table compares imputed and unimputed variables for the study sample. For the unimputed data, denominators differ by variable and are stated in the Table. The imputed sample was used in analysis. † The complete case population consists of singleton children who have complete outcome, exposure, and key confounder data. This sample was used in a sensitivity analysis. ‡ The 6367 children in the excluded sample are singletons alive at 1 year who do not have outcome data; denominators differ by variable and are stated in the Table. § For the unimputed, complete case, and excluded samples, the overall rash and wheeze variables were derived for children who had rash/wheeze reported at least once in each time period. Table V Association between maternal depression and anxiety and child mental well-being outcomes (high internalizing, externalizing, and total difficulties)

§ For the unimputed, complete case, and excluded samples, the overall rash and wheeze variables were derived for children who had rash/wheeze reported at least once in each time period. Table V Association between maternal depression and anxiety and child mental well-being outcomes (high internalizing, externalizing, and total difficulties) OR (95% CI)∗ High internalizing High externalizing High TDS Pregnancy Maternal depression Low (Q1, reference) Q2 1.28 (0.97-1.68) 1.10 (0.84-1.44) 1.03 (0.76-1.38) Q3 1.73 (1.35-2.21)† 1.41 (1.09-1.83)† 1.66 (1.28-2.15)† High (Q4) 2.58 (2.01-3.32)† 2.05 (1.59-2.64)† 2.69 (2.07-3.48)† Maternal anxiety Low (Q1, reference) Q2 1.42 (1.11-1.81)† 1.03 (0.80-1.32) 1.19 (0.92-1.54) Q3 1.92 (1.54-2.41)† 1.43 (1.14-1.79)† 1.87 (1.48-2.35)† High (Q4) 2.75 (2.19-3.45)† 1.96 (1.56-2.46)† 2.65 (2.11-3.34)† Outcome time point Maternal depression Low (Q1, reference) Q2 1.53 (1.19-1.97)† 1.76 (1.36-2.28)† 1.85 (1.41-2.43)† Q3 2.27 (1.78-2.89)† 2.18 (1.69-2.82)† 2.73 (2.10-3.55)† High (Q4) 4.03 (3.23-5.04)† 4.04 (3.20-5.10)† 5.24 (4.11-6.67)† Maternal anxiety Low (Q1, reference) Q2 1.42 (1.11-1.81)† 1.03 (0.80-1.32) 1.19 (0.92-1.54) Q3 1.92 (1.54-2.41)† 1.43 (1.14-1.79)† 1.87 (1.48-2.35)† High (Q4) 2.75 (2.19-3.45)† 1.96 (1.56-2.46)† 2.65 (2.11-3.34)† ∗ Adjusted for maternal age at birth, financial difficulties, maternal education, housing tenure. † P < .05. Table VII Association between child atopy status and child mental well-being at age 8 years

OR (95% CI)∗ High internalizing High externalizing High TDS Pregnancy Maternal depression Low (Q1, reference) Q2 1.28 (0.97-1.68) 1.10 (0.84-1.44) 1.03 (0.76-1.38) Q3 1.73 (1.35-2.21)† 1.41 (1.09-1.83)† 1.66 (1.28-2.15)† High (Q4) 2.58 (2.01-3.32)† 2.05 (1.59-2.64)† 2.69 (2.07-3.48)† Maternal anxiety Low (Q1, reference) Q2 1.42 (1.11-1.81)† 1.03 (0.80-1.32) 1.19 (0.92-1.54) Q3 1.92 (1.54-2.41)† 1.43 (1.14-1.79)† 1.87 (1.48-2.35)† High (Q4) 2.75 (2.19-3.45)† 1.96 (1.56-2.46)† 2.65 (2.11-3.34)† Outcome time point Maternal depression Low (Q1, reference) Q2 1.53 (1.19-1.97)† 1.76 (1.36-2.28)† 1.85 (1.41-2.43)† Q3 2.27 (1.78-2.89)† 2.18 (1.69-2.82)† 2.73 (2.10-3.55)† High (Q4) 4.03 (3.23-5.04)† 4.04 (3.20-5.10)† 5.24 (4.11-6.67)† Maternal anxiety Low (Q1, reference) Q2 1.42 (1.11-1.81)† 1.03 (0.80-1.32) 1.19 (0.92-1.54) Q3 1.92 (1.54-2.41)† 1.43 (1.14-1.79)† 1.87 (1.48-2.35)† High (Q4) 2.75 (2.19-3.45)† 1.96 (1.56-2.46)† 2.65 (2.11-3.34)† ∗ Adjusted for maternal age at birth, financial difficulties, maternal education, housing tenure. † P < .05. Table VII Association between child atopy status and child mental well-being at age 8 years OR (95% CI) Model 1 + age Model 2 M1 + SEP Model 3 M2 + maternal anxiety and depression in pregnancy Model 4 M2 + maternal anxiety and depression when child 8 years Model 5 M2 + maternal anxiety and depression at both time points Model 6 M5 + maternal smoking Model 7 M5 + maternal sleep Model 8 M5 + child sleep Atopy (reference = no atopy) Complete case N = 5004 N = 4619 N = 4295 N = 4322 N = 4029 N = 4018 N = 3780 N = 3709 High Internalizing score 0.91 (0.72-1.15) 0.89 (0.69-1.13) 0.92 (0.71-1.19) 0.88 (0.68-1.14) 0.89 (0.68-1.17) 0.89 (0.68-1.17) 0.84 (0.63-1.12) 0.81 (0.60-1.08) High Externalizing score 0.91 (0.72-1.16) 0.95 (0.74-1.22) 0.93 (0.72-1.21) 0.88 (0.68-1.15) 0.88 (0.67-1.16) 0.89 (0.67-1.17) 0.86 (0.65-1.14) 0.86 (0.64-1.15) High TDS 1.03 (0.82-1.30) 1.04 (0.81-1.33) 1.07 (0.83-1.38) 0.99 (0.76-1.28) 1.02 (0.78-1.34) 1.02 (0.78-1.35) 0.97 (0.73-1.29) 0.98 (0.74-1.32) All models restricted to those with complete data N = 3584 N = 3584 N = 3584 N = 3584 N = 3584 N = 3584 N = 3584 N = 3584 High internalizing score 0.85 (0.64-1.13) 0.84 (0.63-1.12) 0.84 (0.62-1.12) 0.81 (0.60-1.09) 0.81 (0.60-1.09) 0.81 (0.61-1.10) 0.81 (0.60-1.09) 0.80 (0.59-1.08) High externalizing score 0.85 (0.63-1.13) 0.85 (0.64-1.14) 0.85 (0.64-1.14) 0.85 (0.64-1.14) 0.83 (0.62-1.12) 0.84 (0.62-1.13) 0.84 (0.62-1.13) 0.83 (0.62-1.12) High TDS 0.97 (0.73-1.29) 0.98 (0.74-1.31) 0.98 (0.74-1.31) 0.96 (0.71-1.29) 0.95 (0.71-1.28) 0.96 (0.71-1.29) 0.96 (0.71-1.29) 0.95 (0.71-1.28) Model 1: child sex, age (at skin prick test and at outcome).

(0.64-1.14) 0.85 (0.64-1.14) 0.85 (0.64-1.14) 0.83 (0.62-1.12) 0.84 (0.62-1.13) 0.84 (0.62-1.13) 0.83 (0.62-1.12) High TDS 0.97 (0.73-1.29) 0.98 (0.74-1.31) 0.98 (0.74-1.31) 0.96 (0.71-1.29) 0.95 (0.71-1.28) 0.96 (0.71-1.29) 0.96 (0.71-1.29) 0.95 (0.71-1.28) Model 1: child sex, age (at skin prick test and at outcome). Model 2: Model 1 plus ethnicity, mother's age, maternal education, financial difficulties, housing tenure. Model 3: Model 2 plus maternal anxiety and depression in pregnancy. Model 4: Model 2 plus maternal anxiety and depression when child aged 8 years. Model 5: Model 2 plus maternal anxiety and depression in pregnancy and when child aged 8 years. Model 6: Model 5 plus maternal smoking in pregnancy and when child aged 8 years. Model 7: Model 5 plus maternal sleep. Model 8: Model 5 plus child sleep. We are grateful to all the families who took part in this study, the midwives for their help in recruiting them, and the ALSPAC team, which includes interviewers, computer and laboratory technicians, clerical workers, research scientists, volunteers, managers, receptionists, and nurses. ☆ This is an open access article under the CC BY license (http://creativecommons.org/licenses/by/3.0/). Supported by the UK Medical Research Council and the United Kingdom Wellcome Trust (092731) and the University of Bristol (United Kingdom) core support for ALSPAC. B.G. is funded by a Wellcome Trust fellowship (089979). The authors declare no conflicts of interest. Table IV Summary of child mental well-being outcomes and confounders/mediators, overall and by rash and wheeze status

Supported by the UK Medical Research Council and the United Kingdom Wellcome Trust (092731) and the University of Bristol (United Kingdom) core support for ALSPAC. B.G. is funded by a Wellcome Trust fellowship (089979). The authors declare no conflicts of interest. Table IV Summary of child mental well-being outcomes and confounders/mediators, overall and by rash and wheeze status Child's age when questionnaire posted to mother Overall (n = 7250, 100%) Rash‡ Wheeze‡ None Early onset transient Persistent Late onset None Early onset transient Persistent Late onset Outcome - child mental well-being (SDQ) TDS (% high)∗,† 8 years 1 month 10.5 8.4 10.1 12.2 12.7 8.8 11.2 15.4 10.9 Internalizing problems score (% high)∗,† 8 years 1 month 11.1 9.4 10.5 12.7 12.9 9.4 11.8 15.7 12.2 Externalizing problems score (% high)∗,† 8 years 1 month 10.2 7.9 10.0 12.0 11.6 8.6 10.6 16.1 8.7 Exposures - child rash and wheeze status Wheeze∗ None (%) birth to 7 years 7 months 53.6 62.9 53.1 45.3 56.5 Early onset transient (%) 28.0 25.8 31.0 27.0 28.2 Persistent (%) 14.1 7.8 12.3 22.1 9.8 Late onset (%) 4.4 3.5 3.5 5.6 5.5 Rash† None (%) birth to 7 years 7 months 29.0 34.1 26.8 16.1 23.5 Early onset transient (%) 29.9 29.7 33.2 26.2 24.1 Persistent (%) 33.4 28.2 32.2 52.4 42.8 Late onset (%) 7.6 8.0 7.7 5.3 9.6 Other variables Maternal age at delivery (mean)∗,† 29.1 28.9 28.9 29.2 29.6 29.2 28.9 28.7 29.2 Socioeconomic status Maternal education (% with university degree)∗ Pregnancy 16.4 12.9 17.4 19.0 14.3 17.0 15.4 14.7 19.5 Housing tenure (% with mortgage/owned)† Pregnancy 82.0 81.3 82.4 82.7 80.1 84.9 80.0 75.6 78.9 Financial difficulties (% none)† Pregnancy 40.3 40.7 40.5 40.0 39.5 44.3 35.5 35.3 37.5 Mental health in pregnancy Depression Score (EPDS) (% Q4 - high)∗,† Pregnancy 24.3 23.1 24.7 25.9 20.3 20.6 27.7 31.2 25.6 Anxiety score (% Q4 - high)∗,† Pregnancy 19.1 18.2 18.5 20.8 17.9 15.9 22.6 24.5 19.1 Mental health at outcome time point Depression score (EPDS) (% Q4 - high)∗,† 8 years 1 month 23.5 20.3 23.1 25.8 27.6 21.8 24.7 27.3 24.3 Anxiety score (% Q4 - high)∗,† 8 years 1 month 24.3 21.9 24.1 25.7 28.4 23.1 25.2 27.0 25.2 Maternal smoking Yes in pregnancy (%)† Pregnancy 20.3 20.4 21.8 18.8 21.2 17.2 23.7 25.0 21.9 Yes at outcome time point (%)† 8 years 1 month 19.4 19.1 21.2 18.1 19.2 16.5 22.9 22.9 21.5 Maternal sleep Insufficient sleep (%)∗,† 7 years 1 month 38.0 34.0 38.8 41.0 37.4 34.7 40.3 45.5 40.8 Child sleep Wakes at night (% ≥1 times per night)∗,† 6 years 9 months 16.8 15.0 16.0 18.0 20.9 15.2 16.4 22.1 21.3 EPDS, Edinburgh Postnatal Depres

oint (%)† 8 years 1 month 19.4 19.1 21.2 18.1 19.2 16.5 22.9 22.9 21.5 Maternal sleep Insufficient sleep (%)∗,† 7 years 1 month 38.0 34.0 38.8 41.0 37.4 34.7 40.3 45.5 40.8 Child sleep Wakes at night (% ≥1 times per night)∗,† 6 years 9 months 16.8 15.0 16.0 18.0 20.9 15.2 16.4 22.1 21.3 EPDS, Edinburgh Postnatal Depres sion Scale; SDQ, Strengths and Difficulties Questionnaire. ∗ Difference between rash categories in distribution of variable statistically significant (P < .05). † Difference between wheeze categories in distribution of variable statistically significant (P < .05). ‡ Numbers in each rash and wheeze category not shown as they differ across imputed datasets. Table VI Association between child rash and wheeze status and child mental well-being at age 8 years

∗ Difference between rash categories in distribution of variable statistically significant (P < .05). † Difference between wheeze categories in distribution of variable statistically significant (P < .05). ‡ Numbers in each rash and wheeze category not shown as they differ across imputed datasets. Table VI Association between child rash and wheeze status and child mental well-being at age 8 years OR (95% CI) Model 1 Rash and wheeze in separate models Model 2 Rash and wheeze in same model Model 3 M2 + maternal anxiety and depression in pregnancy Model 4 M2 + maternal anxiety and depression when child 8 years Model 5 M2 + maternal anxiety and depression at both time points Model 6 M5 + maternal smoking Model 7 M5 + maternal sleep Model 8 M5 + child sleep High internalizing symptoms Rash (reference = no rash) Early onset transient 1.14 (0.92-1.42) 1.10 (0.89-1.38) 1.09 (0.88-1.36) 1.04 (0.83-1.31) 1.04 (0.83-1.31) 1.05 (0.83-1.31) 1.03 (0.82-1.30) 1.04 (0.83-1.30) Persistent 1.42 (1.16-1.74)† 1.31 (1.07-1.61)† 1.28 (1.04-1.57)∗ 1.21 (0.98-1.50) 1.21 (0.98-1.50) 1.21 (0.98-1.50) 1.20 (0.97-1.48) 1.20 (0.97-1.49) Late onset 1.46 (1.07-1.99)∗ 1.43 (1.05-1.95)∗ 1.44 (1.06-1.98)∗ 1.28 (0.93-1.77) 1.30 (0.94-1.79) 1.30 (0.94-1.79) 1.30 (0.94-1.80) 1.27 (0.92-1.75) Wheeze (reference = no wheeze) Early onset transient 1.22 (1.02-1.46)∗ 1.21 (1.01-1.44)∗ 1.15 (0.96-1.38) 1.17 (0.97-1.40) 1.15 (0.96-1.38) 1.15 (0.96-1.38) 1.14 (0.95-1.37) 1.15 (0.95-1.38) Persistent 1.67 (1.35-2.06)‡ 1.58 (1.28-1.96)‡ 1.48 (1.19-1.83)‡ 1.54 (1.23-1.92)‡ 1.50 (1.20-1.87)‡ 1.50 (1.20-1.87)‡ 1.47 (1.18-1.84)† 1.45 (1.16-1.81)† Late onset 1.30 (0.89-1.91) 1.25 (0.85-1.84) 1.20 (0.81-1.76) 1.21 (0.82-1.78) 1.20 (0.81-1.77) 1.20 (0.81-1.78) 1.18 (0.80-1.75) 1.16 (0.79-1.72) High Externalizing Symptoms Rash (reference = no rash) Early onset transient 1.33 (1.06-1.67)∗ 1.29 (1.03-1.63)∗ 1.28 (1.02-1.62)∗ 1.22 (0.97-1.54) 1.22 (0.97-1.54) 1.21 (0.96-1.53) 1.21 (0.96-1.53) 1.21 (0.96-1.52) Persistent 1.74 (1.40-2.15)‡ 1.61 (1.29-2.00)‡ 1.58 (1.27-1.97)‡ 1.47 (1.17-1.83)† 1.47 (1.18-1.84)† 1.48 (1.18-1.85)† 1.46 (1.17-1.83)† 1.46 (1.17-1.84)† Late onset 1.62 (1.17-2.25)† 1.60 (1.15-2.22)† 1.62 (1.17-2.25)† 1.43 (1.03-2.01)∗ 1.45 (1.04-2.03)∗ 1.44 (1.03-2.02)∗ 1.45 (1.04-2.03)∗ 1.41 (1.00-1.98)∗ Wheeze (reference = no wheeze) Early onset transient 1.13 (0.93-1.37) 1.10 (0.91-1.34) 1.07 (0.88-1.30) 1.06 (0.87-1.29) 1.05 (0.86-1.29) 1.04 (0.85-1.27) 1.05 (0.86-1.28) 1.05 (0.86-1.28) Persistent 1.74 (1.41-2.15)‡ 1.59 (1.28-1.97)‡ 1.52 (1.22-1.89)‡ 1.55 (1.25-1.94)‡ 1.54 (1.24-1.93)‡ 1.53 (1.23-1.91

1 (1.00-1.98)∗ Wheeze (reference = no wheeze) Early onset transient 1.13 (0.93-1.37) 1.10 (0.91-1.34) 1.07 (0.88-1.30) 1.06 (0.87-1.29) 1.05 (0.86-1.29) 1.04 (0.85-1.27) 1.05 (0.86-1.28) 1.05 (0.86-1.28) Persistent 1.74 (1.41-2.15)‡ 1.59 (1.28-1.97)‡ 1.52 (1.22-1.89)‡ 1.55 (1.25-1.94)‡ 1.54 (1.24-1.93)‡ 1.53 (1.23-1.91 )‡ 1.52 (1.22-1.90)‡ 1.47 (1.18-1.84)† Late onset 0.99 (0.62-1.57) 0.93 (0.59-1.47) 0.90 (0.57-1.44) 0.89 (0.56-1.41) 0.88 (0.56-1.41) 0.87 (0.55-1.39) 0.88 (0.55-1.40) 0.85 (0.54-1.36) High TDS Rash (reference = no rash) Early onset transient 1.25 (0.99-1.59) 1.21 (0.96-1.54) 1.20 (0.95-1.52) 1.14 (0.90-1.45) 1.15 (0.90-1.46) 1.15 (0.90-1.45) 1.13 (0.89-1.44) 1.13 (0.89-1.44) Persistent 1.61 (1.31-1.98)‡ 1.49 (1.21-1.84)‡ 1.46 (1.18-1.81)‡ 1.36 (1.09-1.68)† 1.36 (1.10-1.69)† 1.37 (1.10-1.70)† 1.35 (1.08-1.67)† 1.35 (1.09-1.68)† Late onset 1.68 (1.23-2.29)† 1.65 (1.21-2.25)† 1.69 (1.23-2.31)† 1.47 (1.06-2.03)∗ 1.50 (1.08-2.07)∗ 1.50 (1.08-2.07)∗ 1.50 (1.09-2.08)∗ 1.45 (1.04-2.01)∗ Wheeze (reference = no wheeze) Early onset transient 1.20 (1.00-1.44) 1.17 (0.98-1.41) 1.12 (0.93-1.35) 1.13 (0.93-1.37) 1.11 (0.92-1.35) 1.11 (0.91-1.35) 1.11 (0.91-1.34) 1.11 (0.91-1.35) Persistent 1.66 (1.34-2.04)‡ 1.54 (1.24-1.91)‡ 1.43 (1.15-1.78)‡ 1.49 (1.20-1.87)‡ 1.46 (1.17-1.83)† 1.46 (1.16-1.82)† 1.42 (1.14-1.78)† 1.38 (1.10-1.73)† Late onset 1.24 (0.81-1.90) 1.18 (0.77-1.80) 1.13 (0.73-1.73) 1.13 (0.73-1.74) 1.12 (0.72-1.73) 1.11 (0.72-1.72) 1.10 (0.71-1.70) 1.07 (0.69-1.66) All models adjust for child sex, age, ethnicity, mother's age, maternal education, financial difficulties, housing tenure.

6 (1.16-1.82)† 1.42 (1.14-1.78)† 1.38 (1.10-1.73)† Late onset 1.24 (0.81-1.90) 1.18 (0.77-1.80) 1.13 (0.73-1.73) 1.13 (0.73-1.74) 1.12 (0.72-1.73) 1.11 (0.72-1.72) 1.10 (0.71-1.70) 1.07 (0.69-1.66) All models adjust for child sex, age, ethnicity, mother's age, maternal education, financial difficulties, housing tenure. Model 1: separate model for rash and wheeze status. Model 2: rash status, wheeze status. Model 3: Model 2 plus maternal anxiety and depression in pregnancy. Model 4: Model 2 plus maternal anxiety and depression when child aged 8 years. Model 5: Model 2 plus maternal anxiety and depression in pregnancy and when child aged 8 years. Model 6: Model 5 plus maternal smoking in pregnancy and when child aged 8 years. Model 7: Model 5 plus maternal sleep. Model 8: Model 5 plus child sleep. ∗ P < .05. † P < .01. ‡ P < .001.

Extremely low birth weight (ELBW, birth weight <1000 g) infants depend on parenteral nutrition (PN) for several weeks.1 With ongoing PN, parenteral nutrition associated cholestasis (PNAC) is a well-described pathology2 that may progress to liver failure.3 Soybean oil-based lipid emulsions are rich in proinflammatory ω-6 long chain polyunsaturated fatty acids (LC-PUFAs) and phytosterols4 that are both accused to trigger PNAC.5,6 To overcome these untoward effects alternative oil-based lipid emulsions were developed. Here, the latest generation contains fish oil, which provides ω-3 LC-PUFAs—that are less proinflammatory compared with ω-6 LC-PUFAs—and is devoid of phytosterols.4 Infants with PNAC because of intestinal failure who were treated with a lipid emulsion exclusively based on fish oil showed marked improvements of liver function.7,8 Most recently, a mixed lipid emulsion composed of soybean oil, medium chain triglycerides (MCTs), olive oil, and fish oil (SMOF-LE) has become available, and is licensed for pediatric use in the European Union, but not the US because of insufficient data to date on adequate supply with essential fatty acids (EFA). Because of its mixed nature, SMOF-LE contains less ω-6 LC-PUFA4,9 and phytosterols10 but more ω-3 LC-PUFA and also more alpha-tocopherol4,9 compared with soybean oil-based lipid emulsion. Parenteral nutrition using SMOF-LE may, therefore, prevent PNAC compared with the current standard soybean oil-based lipid emulsion. We therefore hypothesized that PN using SMOF-LE would reduce PNAC in ELBW infants. The current trial was set up to compare SMOF-LE and soybean oil-based lipid emulsion for PN of ELBW infants with PNAC as the primary outcome.

y, therefore, prevent PNAC compared with the current standard soybean oil-based lipid emulsion. We therefore hypothesized that PN using SMOF-LE would reduce PNAC in ELBW infants. The current trial was set up to compare SMOF-LE and soybean oil-based lipid emulsion for PN of ELBW infants with PNAC as the primary outcome. Methods This double-blind randomized controlled trial (RCT) (ClinicalTrials.gov: NCT01585935) was performed at the level IV11 neonatal care unit of the University Children’s Hospital Vienna (Medical University of Vienna, Vienna, Austria). The primary aim was to assess whether a mixed lipid emulsion (SMOF-LE) would reduce PNAC in ELBW infants compared with a soybean oil-based lipid emulsion. Secondary aims were to explore the impact on other morbidities. Recruitment was initiated in June 2012. Eligible participants were ELBW infants admitted before 24 hours of life. Infants with cholestasis (conjugated bilirubin >1.5 mg/dL [25 μmol/L]) before intervention and higher order multiples were not eligible. Infants with conditions associated with cholestasis independent of PN (ie, infection with cytomegalovirus, HIV, hepatitis B or C, rhesus mediated hemolysis, cystic fibrosis, inborn errors of metabolism or primary liver diseases) were not eligible or excluded postrandomization.12 Participants were randomized using permuted blocks (ratio 1:1, block size of 4) and stratified according to sex and birth weight (<750 vs ≥750 g) using a software13 prepared by an independent statistician, who kept the randomization sequence concealed until the end of the study. To account for correlation of twins, the first twin was randomized and the second assigned to the opposite treatment.14 Intervention was started within the first 120 hours of life. Infants received either a mixed lipid emulsion composed of 30% soybean oil, 30% MCT, 25% olive oil, and 15% fish oil (SMOFlipid 20%; Fresenius Kabi, Bad Homburg, Germany; SMOF-LE; ω-6:ω-3 ratio 2.5:1) or a soybean oil-based lipid emulsion (Intralipid 20%; Fresenius Kabi, Bad Homburg, Germany; soybean oil-based lipid emulsion; ω-6:ω-3 ratio 8:1) for PN over 24 hours. Both products have similar side effects and are registered for infants in Europe, but in the US SMOFlipid 20% (Fresenius Kabi, Bad Homburg, Germany) is only registered for adults.

d emulsion (Intralipid 20%; Fresenius Kabi, Bad Homburg, Germany; soybean oil-based lipid emulsion; ω-6:ω-3 ratio 8:1) for PN over 24 hours. Both products have similar side effects and are registered for infants in Europe, but in the US SMOFlipid 20% (Fresenius Kabi, Bad Homburg, Germany) is only registered for adults. Based on a content of 35 mg/mL (range: 28-50) linoleic acid and a requirement of 0.25 g/kg/day15 to prevent EFA deficiency, SMOF-LE theoretically will prevent EFA deficiency at 1.4 g/kg/day (range: 1-1.8). Based on a prevalence of PNAC of 25% at our unit,16 a χ2 test indicated that 100 infants/group were required to detect a relative reduction by 60% (from 25% to 10%) with a 2-sided 5% significance level and a power of 80%. Assuming a dropout rate of 18%, we aimed at recruiting 122 infants/group in a 3-year period.

Based on a content of 35 mg/mL (range: 28-50) linoleic acid and a requirement of 0.25 g/kg/day15 to prevent EFA deficiency, SMOF-LE theoretically will prevent EFA deficiency at 1.4 g/kg/day (range: 1-1.8). Based on a prevalence of PNAC of 25% at our unit,16 a χ2 test indicated that 100 infants/group were required to detect a relative reduction by 60% (from 25% to 10%) with a 2-sided 5% significance level and a power of 80%. Assuming a dropout rate of 18%, we aimed at recruiting 122 infants/group in a 3-year period. Participants, healthcare providers, data collectors, and outcome adjudicators were blinded. The investigational products were visually indistinguishable. Taste or smell was regarded irrelevant because of the closed mode of application; study lipids were stored at room temperature and applied using light protected syringes and infusion lines. A blinding team uninvolved in clinical decisions established the blinding code and masked the glass containers using opaque labels designated “Lipid A” or “Lipid B.” Labels were resistant to detachment, in particular by 70% alcohol used during aseptic preparation. Neonatal nurses who prepared the study lipids for PN were part of the blinding team. Discarded containers were controlled for blinding integrity. The attending physicians prescribed the study lipids together with PN using a computer program (catoPAN; Cato Software Solutions, Becton Dickinson, Vienna, Austria) customized to include the designations “Lipid A” and “Lipid B.” Participants received full PN from birth using soybean oil-based lipid emulsion (1 g/kg/day) and were switched to study lipids after enrollment. Lipids were dosed up to 3 g/kg/day at the discretion of the attending physicians and reduced in relation to enteral nutrition (increased up to 20 mL/kg/day). Serum triglycerides were measured at least weekly. Lipids were halted for 24 hours at triglyceride levels >400 mg/dL (4.5 mmol/L) or down titrated >250 mg/dL (2.8 mmol/L). Parenteral nutrition was stopped at 140-160 mL/kg/day of enteral feeds. Urodeoxycholic acid was administered to patients that developed cholestasis. Parenteral fish oil (Omegaven; Fresenius Kabi, Bad Homburg, Germany) was permitted as rescue therapy (1 g/kg/day) if conjugated bilirubin was >6 mg/dL (100 μmol/L). Infants were followed until their 44th week of postmenstrual age (PMA), discharge, or transfer to another hospital.

to patients that developed cholestasis. Parenteral fish oil (Omegaven; Fresenius Kabi, Bad Homburg, Germany) was permitted as rescue therapy (1 g/kg/day) if conjugated bilirubin was >6 mg/dL (100 μmol/L). Infants were followed until their 44th week of postmenstrual age (PMA), discharge, or transfer to another hospital. Patient data were collected from the electronic charts (ICCA; Phillips Medical Systems, Eindhoven, The Netherlands) and discharge letters. Data on PN were collected from the prescription software (catoPAN; Cato Software Solutions, Becton Dickinson, Vienna, Austria). Demographic characteristics were recorded as shown in Table I. A full course of prenatal steroids was defined as 2 doses of betamethasone administered 24 hours apart. Surfactant (Curosurf; Chiesi, Parma, Italy) was administered to all infants <28 + 0 weeks of gestational age17 or else if deficiency was suspected (>35% oxygen with saturation ranges of 88%-96%). Anthropometry was performed by the attending nurses and z scores calculated using growth curves by Fenton et al.18,19 Small for gestational age was defined as birth weight <10th percentile (z score <–1.28). The primary outcome PNAC, measures of liver function, and neonatal morbidities were recorded as shown in Table II.

A full course of prenatal steroids was defined as 2 doses of betamethasone administered 24 hours apart. Surfactant (Curosurf; Chiesi, Parma, Italy) was administered to all infants <28 + 0 weeks of gestational age17 or else if deficiency was suspected (>35% oxygen with saturation ranges of 88%-96%). Anthropometry was performed by the attending nurses and z scores calculated using growth curves by Fenton et al.18,19 Small for gestational age was defined as birth weight <10th percentile (z score <–1.28). The primary outcome PNAC, measures of liver function, and neonatal morbidities were recorded as shown in Table II. PNAC was defined as conjugated bilirubin levels >1.5 mg/dL (25 μmol/L) at 2 consecutive measurements16 by spectrophotometric quantitation (Vitros Chemistry System; Ortho Clinical Diagnostics, Raritan, New Jersey). Peak levels of liver enzymes (aspartate aminotransferase, alanine aminotransferase, γ-glutamyltransferase [GT], and alkaline phosphatase [AP]) during hospitalization were identified. Blood sampling was performed weekly as long as PN was required and then every 7-14 days. Retinopathy of prematurity (ROP) was screened for by indirect ophthalmoscopy starting at 5 weeks of age. Treatment (laser or intravitreal ranibizumab) was performed at ROP stage ≥3. Culture proven sepsis was detected by blood culture (BacT/Alert Pediatric FAN; BioMerieux, Marcy l’Etoile, France) drawn after birth and before any antibiotic treatment. Intraventricular hemorrhage and cystic periventricular leukomalacia were diagnosed by cerebral ultrasound performed every 7-14 days. Necrotizing enterocolitis (NEC) was diagnosed clinically (Bell’s stage ≥IIa20) or after surgical exploration. Focal intestinal perforation was defined as perforation in an otherwise healthy bowel. All infants received probiotics21 and lactoferrin.22 Bronchopulmonary dysplasia (BPD) was defined as supplementary oxygen after 36 + 0 weeks PMA. Persistent ductus arteriosus was treated if hemodynamically significant (enddiastolic blood flow in the left pulmonary artery >20 cm/s or backward flow in the abdominal aorta)23 using ibuprofen. Pulmonary hypertension was diagnosed by measurement of tricuspid regurgitation or right-to-left/bidirectional shunt across the ductus arteriosus24 and treated using inhaled nitric oxide, sildenafil or both.

olic blood flow in the left pulmonary artery >20 cm/s or backward flow in the abdominal aorta)23 using ibuprofen. Pulmonary hypertension was diagnosed by measurement of tricuspid regurgitation or right-to-left/bidirectional shunt across the ductus arteriosus24 and treated using inhaled nitric oxide, sildenafil or both. Data on study lipids, nutrition and growth were recorded as shown in Table III. Therapy adherence was calculated as the percentage study lipids were correctly provided; >80% was considered highly adherent.25 Enteral feeds were provided every 3 hours; the median volume of a single feed per kg in the first week of life was calculated. For growth analysis (anthropometry with z score difference from birth to discharge), only survivors were analyzed to avoid distortion of measurements by perimortal edema.

red highly adherent.25 Enteral feeds were provided every 3 hours; the median volume of a single feed per kg in the first week of life was calculated. For growth analysis (anthropometry with z score difference from birth to discharge), only survivors were analyzed to avoid distortion of measurements by perimortal edema. Statistical Analyses The χ2 test was used to examine the relationship between the intervention and PNAC. The primary outcome was additionally analyzed by binary logistic regression (Table IV). Possible confounders were specified according to literature (male sex,26 sepsis,27 compound outcome of NEC, focal intestinal perforation and gastrointestinal surgery,28 birth weight,29 total days on PN,3 z score of birth weight,16 and enteral nutrition in the first week of life per kg birth weight16) and tested in univariable logistic regression models for significant influence. Because of the low number of events the confounders included in the final model were restricted to those with univariate P < .01. Furthermore, the covariates of the final model were reduced according to Akaike information criterion to avoid overfitting. Secondary outcome measures were compared between the treatment groups with χ2 test for categorical data and Mann-Whitney U test for continuous data. Analyses were performed by intention to treat (ITT). As sensitivity analysis, calculations were additionally carried out on the per protocol set (treatment adherence ≥80%,25 admission ≥28 days according to the study protocol). Because of opposed allocation of twins, calculations including the mother as random factor were not necessary.14 There was no interim analysis.

). As sensitivity analysis, calculations were additionally carried out on the per protocol set (treatment adherence ≥80%,25 admission ≥28 days according to the study protocol). Because of opposed allocation of twins, calculations including the mother as random factor were not necessary.14 There was no interim analysis. The study was conducted in conformance with the Declaration of Helsinki, International Conference on Harmonisation / Good Clinical Practice guidelines, and the respective European Union directives embedded in the Austrian drug act. Written consent from 1 parent was sufficient due to low risk for participants.30 Patients were insured as legally required. The study was approved by the institution’s ethic committee (EK 2011/1030) and registered at European Clinical Trial Database (EudraCT 2011-005456-33) and clinicaltrials.gov (NCT01585935).

ug act. Written consent from 1 parent was sufficient due to low risk for participants.30 Patients were insured as legally required. The study was approved by the institution’s ethic committee (EK 2011/1030) and registered at European Clinical Trial Database (EudraCT 2011-005456-33) and clinicaltrials.gov (NCT01585935). Results A total of 274 ELBW infants were screened: 223 infants were available for ITT analysis, 206 for per protocol analysis (Figure; available at www.jpeds.com). Three participants did not receive the intervention (death: n = 2; error: n = 1), 3 participants were withdrawn due to conditions associated with cholestasis (mitochondriopathy: n = 1; Gaucher disease type II: n = 1; undefined genetic syndrome: n = 1), and 1 participant was withdrawn due to cholestasis before intervention. Thus, 7 participants (SMOF-LE: n = 4; soybean oil-based lipid emulsion: n = 3) were excluded from ITT analysis after randomization.12 Recruitment ended after the intended number of participants was included (July 2015); follow-up lasted until October 2015. Premature unblinding (accidentally or as a result of a serious adverse event) did not occur. All analyses were prespecified according to the study protocol. Demographic characteristics are shown in Table I. Study participants were born between 23 + 1 and 33 + 4 weeks PMA (median 26 + 0 weeks) and weighed between 390 and 998 g (median 775 g).

Results A total of 274 ELBW infants were screened: 223 infants were available for ITT analysis, 206 for per protocol analysis (Figure; available at www.jpeds.com). Three participants did not receive the intervention (death: n = 2; error: n = 1), 3 participants were withdrawn due to conditions associated with cholestasis (mitochondriopathy: n = 1; Gaucher disease type II: n = 1; undefined genetic syndrome: n = 1), and 1 participant was withdrawn due to cholestasis before intervention. Thus, 7 participants (SMOF-LE: n = 4; soybean oil-based lipid emulsion: n = 3) were excluded from ITT analysis after randomization.12 Recruitment ended after the intended number of participants was included (July 2015); follow-up lasted until October 2015. Premature unblinding (accidentally or as a result of a serious adverse event) did not occur. All analyses were prespecified according to the study protocol. Demographic characteristics are shown in Table I. Study participants were born between 23 + 1 and 33 + 4 weeks PMA (median 26 + 0 weeks) and weighed between 390 and 998 g (median 775 g). The primary outcome PNAC was 36% lower using SMOFLE compared with soybean oil-based lipid emulsion, a relative difference that was statistically not significant (Table II). We found no significant difference in PNAC characteristics, highest conjugated bilirubin, or liver enzymes. Mortality, hospitalization, culture proven sepsis and morbidity (visual, gastrointestinal, neurologic, cardiovascular, and pulmonary) did not differ significantly. Infants receiving fish oil as a rescue therapy did not differ between groups in mortality (SMOF-LE: 0/3 vs soybean oil-based lipid emulsion: 1/4; P = .35) or normalization of PNAC until discharge (SMOFLE: 1/3 vs soybean oil-based lipid emulsion: 0/4; P = .21). A multivariable model including relevant confounders showed no significant difference for the primary outcome (Table IV).

between groups in mortality (SMOF-LE: 0/3 vs soybean oil-based lipid emulsion: 1/4; P = .35) or normalization of PNAC until discharge (SMOFLE: 1/3 vs soybean oil-based lipid emulsion: 0/4; P = .21). A multivariable model including relevant confounders showed no significant difference for the primary outcome (Table IV). Therapy adherence was high and equal between groups (Table III). Both groups received PN for a comparable time with similar amounts of lipids. Triglyceride levels were measured at similar intravenous lipid supply and did not differ significantly, as well as the incidence of hypertriglyceridemia. Feeding volumes in the first week of life and the use of mother’s milk were similar. There was no significant difference in anthropometry from birth to the end of the study. Discussion In this prospective, double-blind randomized trial in ELBW infants a mixed lipid emulsion composed of soybean oil, MCT, olive oil, and fish oil did not significantly reduce the incidence of PNAC compared with a soybean oil-based lipid emulsion. We found no effect on measures such as ROP, BPD, and growth.

Therapy adherence was high and equal between groups (Table III). Both groups received PN for a comparable time with similar amounts of lipids. Triglyceride levels were measured at similar intravenous lipid supply and did not differ significantly, as well as the incidence of hypertriglyceridemia. Feeding volumes in the first week of life and the use of mother’s milk were similar. There was no significant difference in anthropometry from birth to the end of the study. Discussion In this prospective, double-blind randomized trial in ELBW infants a mixed lipid emulsion composed of soybean oil, MCT, olive oil, and fish oil did not significantly reduce the incidence of PNAC compared with a soybean oil-based lipid emulsion. We found no effect on measures such as ROP, BPD, and growth. Lipid emulsions based on soybean oil are the currently recommended product for provision of parenteral lipids in preterm infants,4,31 and the only licensed lipid emulsion for infants in the US.9 Because side effects of PN on liver function were ascribed to the specific properties of soybean oil, research focused on reducing the excess of ω-6 LC-PUFAs and phytosterols in lipid emulsions by admixing alternative oils (MCT and olive) and most recently fish oil providing ω-3 LC-PUFAs to counterbalance proinflammatory ω-6 effects.4 In this regard, SMOFLE is the most recent development and was shown to improve the supply with the crucial ω-3 LC-PUFA docosahexaenoic acid (DHA) while lowering soybean oil exposure by 70% and phytosterol exposure by roughly 50%.10 Also, the higher supply with vitamin E compared with soybean oil-based lipid emulsion may prevent PNAC as recently shown in preterm piglets.32 However, until now clinical trials have not provided enough evidence for improvement of clinical outcomes such as PNAC to justify a change of the current guidelines that recommend soybean oil-based lipid emulsion.

E compared with soybean oil-based lipid emulsion may prevent PNAC as recently shown in preterm piglets.32 However, until now clinical trials have not provided enough evidence for improvement of clinical outcomes such as PNAC to justify a change of the current guidelines that recommend soybean oil-based lipid emulsion. Four out of 6 trials that have tested SMOF-LE vs soybean oil-based lipid emulsion in 310 preterm infants reported on PNAC as secondary outcome,33–36 without a significant effect in a recent meta-analysis.31 However, the overall incidence of PNAC was 5% across studies, therefore, the relatively low number of infants analyzed may not suffice to exclude a clinically meaningful effect. Our trial was the first designed to investigate SMOF-LE for prevention of PNAC as primary outcome. To attain a high baseline incidence, we exclusively recruited the vulnerable group of ELBW infants with an anticipated PNAC incidence of 25% at our unit.16 However, the power to prove our hypothesis was lowered by an observed PNAC incidence of only 15.9% in the current trial, attributable to an accelerated weaning from PN compared with the planning phase (−10 days).16 This shorter time on PN is an important limitation of our study and possibly related to the implementation of probiotics at our unit before the start of this trial in 201021 and their preventive effect against NEC. Reduced fear of NEC possibly encouraged clinicians to establish full enteral feeds faster.

(−10 days).16 This shorter time on PN is an important limitation of our study and possibly related to the implementation of probiotics at our unit before the start of this trial in 201021 and their preventive effect against NEC. Reduced fear of NEC possibly encouraged clinicians to establish full enteral feeds faster. Although our study showed that SMOF-LE did not significantly prevent PNAC in ELBW, our results cannot be generalized to infants with a substantially longer time on PN such as those with intestinal failure. The same applies to infants with established PNAC. Here, a significant decline of conjugated bilirubin was demonstrated in a pilot RCT after switching from soybean oil-based lipid emulsion to SMOF-LE, however, only after adjusting the original analysis and eliminating a statistical outlier.37 In our study, PNAC also resolved more quickly, however, the numbers were not statistically significant (Table II).

ated bilirubin was demonstrated in a pilot RCT after switching from soybean oil-based lipid emulsion to SMOF-LE, however, only after adjusting the original analysis and eliminating a statistical outlier.37 In our study, PNAC also resolved more quickly, however, the numbers were not statistically significant (Table II). Besides protection of liver function, other morbidities associated with prematurity such as BPD38 might have been affected by the intervention with SMOF-LE. So far meta-analyses did not show a significant influence on BPD.39 On the contrary, Collins et al recently reported their trial of fish oil supplementation in preterm infants and surprisingly found a significantly increased risk for BPD.40 In that study, fish oil was used as enteral add-on from birth to 36 weeks PMA providing a total DHA supply even beyond fetal accretion rates.40 This is in contrast to our study using SMOF-LE, where fish oil is provided parenterally and only as long as PN was needed. However, Collins et al also questioned the safety of PN using SMOF-LE. We and other RCTs found no difference in BPD using SMOF-LE compared with soybean oil-based lipid emulsion.33–36 Furthermore, we found no affection of other outcomes related to inflammatory processes such as pulmonary hypertension41 and persistent ductus arteriosus42 or of any other neonatal morbidity and death.

MOF-LE. We and other RCTs found no difference in BPD using SMOF-LE compared with soybean oil-based lipid emulsion.33–36 Furthermore, we found no affection of other outcomes related to inflammatory processes such as pulmonary hypertension41 and persistent ductus arteriosus42 or of any other neonatal morbidity and death. In recent years, the ω-3 LC-PUFA DHA was included into models of ROP as another nonoxygen-regulated factor.43 Supplementation with fish oil to improve DHA enrichment of the retina was suggested to prevent ROP.43 Here, SMOF-LE would not only provide DHA but also higher amounts of vitamin E4 known to prevent ROP.44 Two RCTs studied the effect of SMOF-LE33 or a pure fish oil LE45 on ROP as primary outcome. In the study by Beken et al, SMOF-LE significantly reduced any ROP.33 Pawlik et al showed that co-application of a pure fish oil LE and soybean oil-based lipid emulsion significantly prevented severe ROP.45 A recent systematic review suggested a preventive effect based on RCTs and observational studies.46 In the present study, we did not find any effect against any stage of ROP. Compared with the study by Pawlik et al, we provided less fish oil using SMOF-LE, which may indicate some dose dependency. However, the lack of effect compared with Beken et al (ROP reduced by 80%) who also used SMOF-LE is striking.33 In this context, it seems important that infants in our study were more immature by 4 weeks compared with Beken et al. As ROP typically occurs after 30 weeks PMA, the timing of fish oil supplementation from birth and retinal DHA enrichment in relation to the actual PMA probably matters. In this respect, well-designed pharmacologic studies on dosing and timing of SMOF-LE or pure fish oil LE in relation to the actual PMA are necessary. However, it is discouraging that the trial by Collins et al, who applied enteral fish oil in high amounts, also showed no effect on ROP.40

he actual PMA probably matters. In this respect, well-designed pharmacologic studies on dosing and timing of SMOF-LE or pure fish oil LE in relation to the actual PMA are necessary. However, it is discouraging that the trial by Collins et al, who applied enteral fish oil in high amounts, also showed no effect on ROP.40 Safety and tolerance of SMOF-LE is another important issue inconsistently referred to in literature. Although lipid clearance was even improved in adults,47 serum triglycerides were found significantly higher in preterm infants using SMOF-LE compared with soybean oil-based lipid emulsion, though at doses exceeding current recommendations (>3 g/kg/day).34 In our study, serum triglyceride levels and the incidence of hypertriglyceridemia did not differ between SMOF-LE and soybean oil-based lipid emulsion.

significantly higher in preterm infants using SMOF-LE compared with soybean oil-based lipid emulsion, though at doses exceeding current recommendations (>3 g/kg/day).34 In our study, serum triglyceride levels and the incidence of hypertriglyceridemia did not differ between SMOF-LE and soybean oil-based lipid emulsion. Postnatal growth failure is a frequently observed problem in preterm infants.48 Although SMOF-LE and soybean oil-based lipid emulsion provide similar amounts of energy, improved supply with ω-3 LC-PUFAs might impact on head growth as DHA rapidly accumulates in the fetal brain in late pregnancy.49 Furthermore, a recent meta-analysis of biochemical aspects showed lower levels of arachidonic acid using SMOF-LE50 and raised concerns because of an association of low arachidonic acid levels and growth failure.51 In our study, we found no significant impact of SMOF-LE on anthropometry. Moreover, there is a lack of studies demonstrating the efficacy of SMOF-LE in the prevention of EFA deficiency in preterm infants. As we did not measure EFA in our study, we can only speculate. A supply of less than 0.25 g/kg/day of linoleic acid may cause EFA deficiency in preterm infants.15 It therefore seems prudent to aim at a supply of 2 g/kg/day SMOF-LE if enteral nutrition is low in the first week of life, or likely even at 2.5 g/kg/day in exclusively parenterally nourished ELBW infants, to safely prevent EFA deficiency. We thank Eva Wissmann for excellent support with drug accountability.

Postnatal growth failure is a frequently observed problem in preterm infants.48 Although SMOF-LE and soybean oil-based lipid emulsion provide similar amounts of energy, improved supply with ω-3 LC-PUFAs might impact on head growth as DHA rapidly accumulates in the fetal brain in late pregnancy.49 Furthermore, a recent meta-analysis of biochemical aspects showed lower levels of arachidonic acid using SMOF-LE50 and raised concerns because of an association of low arachidonic acid levels and growth failure.51 In our study, we found no significant impact of SMOF-LE on anthropometry. Moreover, there is a lack of studies demonstrating the efficacy of SMOF-LE in the prevention of EFA deficiency in preterm infants. As we did not measure EFA in our study, we can only speculate. A supply of less than 0.25 g/kg/day of linoleic acid may cause EFA deficiency in preterm infants.15 It therefore seems prudent to aim at a supply of 2 g/kg/day SMOF-LE if enteral nutrition is low in the first week of life, or likely even at 2.5 g/kg/day in exclusively parenterally nourished ELBW infants, to safely prevent EFA deficiency. We thank Eva Wissmann for excellent support with drug accountability. Funded by the Austrian Science Fund (FWF, KLI99-B00). Study lipids were supplied free of charge by Herba Chemosan (Graz, Austria). A.R. received funding from Fresenius Kabi (Graz, Austria) to employ a clinical research nurse The other authors declare no conflicts of interest. BPDBronchopulmonary dysplasia DHADocosahexaenoic acid EFAEssential fatty acid ELBWExtremely low birth weight ITTIntention to treat

Funded by the Austrian Science Fund (FWF, KLI99-B00). Study lipids were supplied free of charge by Herba Chemosan (Graz, Austria). A.R. received funding from Fresenius Kabi (Graz, Austria) to employ a clinical research nurse The other authors declare no conflicts of interest. BPDBronchopulmonary dysplasia DHADocosahexaenoic acid EFAEssential fatty acid ELBWExtremely low birth weight ITTIntention to treat LC-PUFALong chain polyunsaturated fatty acid MCTMedium chain triglyceride NECNecrotizing enterocolitis PMAPostmenstrual age PNParenteral nutrition PNACParenteral nutrition associated cholestasis RCTRandomized controlled trial ROPRetinopathy of prematurity SMOF-LELipid emulsion composed of soybean oil, medium chain triglycerides, olive oil, and fish oil Figure Flow chart showing enrollment with reasons for study exclusion and analysis.

NECNecrotizing enterocolitis PMAPostmenstrual age PNParenteral nutrition PNACParenteral nutrition associated cholestasis RCTRandomized controlled trial ROPRetinopathy of prematurity SMOF-LELipid emulsion composed of soybean oil, medium chain triglycerides, olive oil, and fish oil Figure Flow chart showing enrollment with reasons for study exclusion and analysis. Table I Demographic characteristics Parameter SMOF-LE (n = 110) S-LE (n = 113) Obstetric parameters Multiple pregnancy 30 (27) 36 (32) Cesarean delivery 100 (91) 101 (89) Prenatal steroids (full course) 69 (63) 65 (59) PROM 38 (35) 42 (37) Preeclampsia 16 (15) 16 (14) Neonatal parameters Umbilical artery pH 7.31 [7.25 to 7.36] * 7.30 [7.24 to 7.35] † Apgar—5 min 8 [8 to 9] 8 [8 to 9] ‡ Male sex 64 (58) 73 (65) Surfactant 97 (88) 98 (87) Gestational age (wk+d) 25 + 6 [24 + 6 to 27 + 3] 26 + 2 [25 + 0 to 28 + 0] Birth weight (g) 788 [648 to 891] 760 [610 to 884] Z score −0.4 [−1.1 to −0.2] −0.7 [−1.4 to − 0.2] Birth length (cm) 34 [31 to 35] ‡ 33 [31 to 35] ‡ Z score −0.3 [−1.1 to 0.7] ‡ −0.3 [−1.5 to 0.4] ‡ Birth head circumference (cm) 24 [23 to 25] § 24 [23 to 25] Z score −0.1 [−0.7 to 0.6] § −0.2 [−1.0 to −0.2] Small for gestational age 22 (20) 38 (34) PROM, premature rupture of membranes; S-LE soybean oil-based lipid emulsion. Analysis by ITT. Categorical data are presented as numbers with percentages in round parentheses and were tested using the χ2 test. Continuous data are presented as the median and IQR in squared parentheses and were tested using the Mann-Whitney U Test. * Data of 15 patients missing. † Data of 21 patients missing.

Table I Demographic characteristics Parameter SMOF-LE (n = 110) S-LE (n = 113) Obstetric parameters Multiple pregnancy 30 (27) 36 (32) Cesarean delivery 100 (91) 101 (89) Prenatal steroids (full course) 69 (63) 65 (59) PROM 38 (35) 42 (37) Preeclampsia 16 (15) 16 (14) Neonatal parameters Umbilical artery pH 7.31 [7.25 to 7.36] * 7.30 [7.24 to 7.35] † Apgar—5 min 8 [8 to 9] 8 [8 to 9] ‡ Male sex 64 (58) 73 (65) Surfactant 97 (88) 98 (87) Gestational age (wk+d) 25 + 6 [24 + 6 to 27 + 3] 26 + 2 [25 + 0 to 28 + 0] Birth weight (g) 788 [648 to 891] 760 [610 to 884] Z score −0.4 [−1.1 to −0.2] −0.7 [−1.4 to − 0.2] Birth length (cm) 34 [31 to 35] ‡ 33 [31 to 35] ‡ Z score −0.3 [−1.1 to 0.7] ‡ −0.3 [−1.5 to 0.4] ‡ Birth head circumference (cm) 24 [23 to 25] § 24 [23 to 25] Z score −0.1 [−0.7 to 0.6] § −0.2 [−1.0 to −0.2] Small for gestational age 22 (20) 38 (34) PROM, premature rupture of membranes; S-LE soybean oil-based lipid emulsion. Analysis by ITT. Categorical data are presented as numbers with percentages in round parentheses and were tested using the χ2 test. Continuous data are presented as the median and IQR in squared parentheses and were tested using the Mann-Whitney U Test. * Data of 15 patients missing. † Data of 21 patients missing. ‡ Data of 2 patients missing. § Data of 1 patient missing.

Analysis by ITT. Categorical data are presented as numbers with percentages in round parentheses and were tested using the χ2 test. Continuous data are presented as the median and IQR in squared parentheses and were tested using the Mann-Whitney U Test. * Data of 15 patients missing. † Data of 21 patients missing. ‡ Data of 2 patients missing. § Data of 1 patient missing. Table II Neonatal outcome Outcome SMOF-LE (n = 110) S-LE (n = 113) P Primary outcome and liver function (peak levels throughout the study) PNAC 11 (10) 18 (16) .20 Occurrence (day of life) 23 [11-36] 20 [9-42] .95 Death* 2 (18) 3 (16) .91 Rescue therapy using fish oil* 3 (27) 4 (22) .76 Normalized before discharge* 6 (55) 5 (28) .15 Conjugated bilirubin (mg/dL) 0 [0-0.22] 0 [0-0.38] .67 γ-GT (U/mL) 148 [95-243] † 157 [101-217] ‡ .94 AST (U/mL) 41 [32-67] † 48 [34-80] ‡ .13 ALT (U/mL) 32 [30-41] † 34 [29-45] ‡ .40 AP (U/mL) 518 [396-665] † 494 [376-671] ‡ .70 Neonatal morbidity Death 8 (7) 8 (7) .96 Hospitalization (d) 81 [59-105] 79 [63-97] .69 Retinopathy of prematurity (any) 60 (58) 56 (55) .69 Highest grade (grade 1-5) 1 [0-2] 1 [0-2] .63 Requiring treatment (severe ROP) 9 (8) 10 (9) .79 Sepsis, culture proven 24 (22) 26 (23) .83 Intraventricular hemorrhage III/IV 12 (11) 9 (8) .45 Cystic periventricular leukomalacia 3 (3) 4 (4) .73 NEC ≥IIa 8 (7) 8 (7) .96 Focal intestinal perforation 4 (4) 5 (4) .77 Abdominal surgery 13 (12) 14 (12) .90 Days on mechanical ventilation 6 [0-10] 6 [0-10] .51 Chronic lung disease 19 (17) 21 (19) .80 Steroid treatment 11 (10) 17 (15) .26 PDA requiring treatment 56 (51) 68 (60) .16 Number of ibuprofen cycles 2 [1-3] 2 [1-3] .82 Surgical ligation 5 (5) 6 (5) .80 Pulmonary hypertension 23 (21) 31 (28) .27 iNO/sildenafil treatment 19 (17) 28 (25) .18 ALT, Alanine transaminase; AST, Aspartate transaminase; AP, Alkaline phosphatase; GT, glutamyltransferase.

PDA requiring treatment 56 (51) 68 (60) .16 Number of ibuprofen cycles 2 [1-3] 2 [1-3] .82 Surgical ligation 5 (5) 6 (5) .80 Pulmonary hypertension 23 (21) 31 (28) .27 iNO/sildenafil treatment 19 (17) 28 (25) .18 ALT, Alanine transaminase; AST, Aspartate transaminase; AP, Alkaline phosphatase; GT, glutamyltransferase. Analysis by ITT. Categorical data are presented as numbers with percentages in round parentheses and were tested using the chi-square test. Continuous data are presented as median and interquartile range in squared parentheses and were tested using the Mann-Whitney U test. P values <.05 were considered statistically significant. SI conversion factors: To convert bilirubin to µmol/L multiply values by 17.1, to convert triglycerides to mmol/L multiply values by .011. * Percentage within the subgroup of infants with PNAC. † Data of 2 patient missing. ‡ Data of 1 patients missing.

Analysis by ITT. Categorical data are presented as numbers with percentages in round parentheses and were tested using the chi-square test. Continuous data are presented as median and interquartile range in squared parentheses and were tested using the Mann-Whitney U test. P values <.05 were considered statistically significant. SI conversion factors: To convert bilirubin to µmol/L multiply values by 17.1, to convert triglycerides to mmol/L multiply values by .011. * Percentage within the subgroup of infants with PNAC. † Data of 2 patient missing. ‡ Data of 1 patients missing. Table III Nutrition and growth Outcome SMOF-LE (n = 110) S-LE (n = 113) P Study drug and nutrition Therapy adherence >80% 109 (99) 110 (97) .33 Time on PN (d) 23 [17-37] 24 [17-35] .87 Time on parenteral lipids (d) 20 [15-35] 21 [15-29] .91 Total parenteral lipids (g/kg) 38.2 [26.4-68.5] 39.0 [28.1-55.3] .56 Total study lipids (g/kg) 32.6 [21.3-63.8] 34.4 [23.0-50.8] .57 Highest triglycerides (mg/dL) 208 [142-323] * 207 [147-305] † .85 Parenteral lipids at measurement 2.1 [1.5-2.5]* 2.0 [1.2-2.5]† .30 Hypertriglyceridemia 39 (39) 38 (37) .71 Feeding first wk (mL per feed/kg) 3.3 [2.3-4.1] 3.1 [2.3-4.1] .55 Mothers milk at discharge (any) 85 (75) 80 (69) .29 Anthropometry at discharge ‡ Weight at discharge (g) 2594 [2124 to 3029] 2479 [2175 to 2956] .23 Δ z score (birth to end of study) −0.7 [−1.0 to −0.2] −0.7 [−1.1 to −0.2] .84 Length (cm) 45 [42.5 to 47] 44 [41.5 to 47] .14 Δ z Score (birth to end of study) −1.4 [−2.3 to 0.8] −1.5 [−2.1 to 1.0] .64 Head circumference (cm) 32 [30.6 to 33.5] 32 [30.7 to 33.1] .45 Δ z Score (birth to end of study) −1.0 [−1.7 to −0.3] −1.0 [−1.6 to −0.3] .92 Analysis by ITT. Categorical data are presented as numbers with percentages in round parentheses and were tested using the χ2 test. Continuous data are presented as median and IQR in squared parentheses and were tested using the Mann-Whitney U Test. P values <.05 were considered statistically significant.

to −0.3] .92 Analysis by ITT. Categorical data are presented as numbers with percentages in round parentheses and were tested using the χ2 test. Continuous data are presented as median and IQR in squared parentheses and were tested using the Mann-Whitney U Test. P values <.05 were considered statistically significant. * Data of 11 patients missing. † Data of 10 patients missing. ‡ For analysis of growth, infants who died were excluded (SMOF-LE, n = 104; S-LE, n = 101). Table IV Multivariable analysis on the risk for PNAC ITT (n = 223) Per protocol (n = 206) Parameter aOR 95% CI P aOR 95% CI P SMOF-LE 0.428 0.155-1.187 .10 0.457 0.155-1.347 .16 NEC/FIP/GI surgery 5.481   1.894-15.868   .002 5.016   1.528-16.464   .008 Time on PN (d) 1.051 1.023-1.080 <.001 1.059 1.027-1.092 <.001 Feeding first wk (mL per feed/kg) 0.739 0.487-1.112 .16 0.665 0.413-1.073 .10 FIP, focal intestinal perforation; GI, gastrointestinal. Analysis by ITT and per protocol. Binary logistic regression analysis showing the odds for the effect of SMOF-LE on PNAC compared with S-LE, adjusted for “compound outcome of NEC, FIP, and GI surgery,” time on parenteral nutrition and the median volume per enteral feed per kg birth weight in the first week of life. P values < .05 were considered statistically significant.

Insulin-like growth factor-1 (IGF-1) is an important fetal growth regulator, with IGF-1 levels increasing with gestational age, particularly during the second and third trimesters of pregnancy.1,2 After preterm birth, serum IGF-1 levels decrease rapidly and remain low for the first weeks of life relative to corresponding fetal levels in utero.3,4 Longitudinal studies have reported an association between lower serum IGF-1 levels at birth in extremely preterm infants and an increased risk of retinopathy of prematurity (ROP), bronchopulmonary dysplasia (BPD), neurodevelopmental impairment, and growth impairment.5–9 Preclinical models also support associations between IGF-1 and complications of prematurity. In mice, IGF-1 absence delays normal retinal vascular development,10 and recombinant human (rh)IGF-1 administration reduces risk of oxygen-induced retinopathy.11 Additionally, rhIGF-1 administration in a hyperoxia-induced model of BPD decreases signs of disease in newborn rats.12 Angiogenesis is an important process in both retinal and lung development,10,13 and it may represent a common underlying mechanism affected by low IGF-1 levels in ROP and BPD.10,14 In addition, IGF-1 is neuroprotective in rat pups affected by germinal matrix hemorrhage.15 Together, these data suggest that ROP, BPD, brain injury/neurodevelopmental impairment, and growth restriction could be ameliorated by supplementing postnatal serum IGF-1 to corresponding fetal levels in extremely preterm infants.

addition, IGF-1 is neuroprotective in rat pups affected by germinal matrix hemorrhage.15 Together, these data suggest that ROP, BPD, brain injury/neurodevelopmental impairment, and growth restriction could be ameliorated by supplementing postnatal serum IGF-1 to corresponding fetal levels in extremely preterm infants. We are investigating the use of rhIGF-1 complexed with its binding protein rhIGFBP-3 (rhIGF-1/rhIGFBP-3) to prevent complications of prematurity. Early clinical studies conducted between June 2010 and July 2013 demonstrated feasibility of rhIGF-1/rhIGFPB-3 infusion without safety concerns.16,17 In the current study, we hypothesized that rhIGF-1/rhIGFBP-3 administration by continuous intravenous infusion would decrease the severity of ROP and other complications of prematurity.

studies conducted between June 2010 and July 2013 demonstrated feasibility of rhIGF-1/rhIGFPB-3 infusion without safety concerns.16,17 In the current study, we hypothesized that rhIGF-1/rhIGFBP-3 administration by continuous intravenous infusion would decrease the severity of ROP and other complications of prematurity. Methods This phase 2, multicenter, randomized, standard of care con-current control, assessor-masked study evaluated the efficacy and safety of rhIGF-1/rhIGFBP-3 in decreasing the severity of ROP and other complications of prematurity (ClinicalTrials.gov: NCT01096784). The trial was conducted at 20 clinical sites in Italy, the Netherlands, Poland, Sweden, the United Kingdom, and the US. Study drug was administered from within 24 hours after birth until postmenstrual age (PMA) 296/7 weeks, with follow-up evaluations up to a PMA of 404/7 weeks (Figure 1). All infants’ parents/guardians provided written informed consent. The study was reviewed/approved by relevant institutional review boards/independent ethics committees. Additional details on safety monitoring and interim analyses are provided in the Methods section of Appendix 2 (available at www.jpeds.com). The study adhered to International Conference on Harmonization Good Clinical Practice guidelines and the tenets of the Declaration of Helsinki.

ards/independent ethics committees. Additional details on safety monitoring and interim analyses are provided in the Methods section of Appendix 2 (available at www.jpeds.com). The study adhered to International Conference on Harmonization Good Clinical Practice guidelines and the tenets of the Declaration of Helsinki. Patient Population Infants with gestational age at birth of 230/7–276/7 weeks were eligible for enrollment. Exclusion criteria included monozygotic twins, detectable gross malformation, known/suspected chromosomal abnormality, genetic disorder/syndrome, a persistent blood glucose level of <2.5 mmol/L or >10 mmol/L on the day of birth, anticipated need for administration of rh erythropoietin during treatment, a history of maternal diabetes requiring insulin, and clinically significant neurologic disease (germinal matrix hemorrhage allowed). Randomization and Masking Infants were allocated to rhIGF-1/rhIGFBP-3 or standard neonatal care (controls) in a 1:1. Dizygotic twins were randomized to the same study arm. Randomization was stratified by gestational age (<26, ≥26 weeks) centrally, using the permuted-block randomization approach. Investigators were not masked to treatment assignment, but certain assessments were masked. ROP stage was evaluated by 2 centralized independent pediatric ophthalmologists (and adjudicated by a third) and cranial ultrasound scans by a single central examiner.

centrally, using the permuted-block randomization approach. Investigators were not masked to treatment assignment, but certain assessments were masked. ROP stage was evaluated by 2 centralized independent pediatric ophthalmologists (and adjudicated by a third) and cranial ultrasound scans by a single central examiner. Treatment Regimen rhIGF-1/rhIGFBP-3 (mecasermin rinfabate, 50 μg/mL solution), a 1:1 molar ratio of the noncovalent complex of rhIGF-1 and rhIGFBP-3, was administered via continuous intravenous infusion through a central or peripheral line. Interruptions in the infusion of ≥1 hour were recorded. Standard care was determined based on the individual preterm infant’s condition following local protocols. The Methods section in Appendix 2 summarizes information on permitted and prohibited concomitant medications. Dosing and Target IGF-1 Levels rhIGF-1/rhIGFBP-3 dosing was standardized to 250 μg/kg/24 hours with the intention of maintaining serum IGF-1 levels within 28–109 μg/L, estimated as the normal physiologic intrauterine range based on prior literature.18–20 The dose was decreased to 125 μg/kg/24 hours if the infant’s serum IGF-1 levels exceeded the upper bound for 2 consecutively scheduled samples (plus a confirmatory sample 12 hours after the previous 2 consecutive samples). The Methods section in Appendix 2 provides details on sampling intervals and methods for IGF-1 measurement.

as decreased to 125 μg/kg/24 hours if the infant’s serum IGF-1 levels exceeded the upper bound for 2 consecutively scheduled samples (plus a confirmatory sample 12 hours after the previous 2 consecutive samples). The Methods section in Appendix 2 provides details on sampling intervals and methods for IGF-1 measurement. Outcomes The primary endpoint was maximum severity of ROP stage across all retinal examinations, based on retinal camera (RetCam, Clarity Medical Systems Inc, Pleasanton, California) images of the dilated fundus. ROP assessments were performed every 1–2 weeks between PMA 31 and 40 weeks. ROP was classified according to the International Classification.21 For treatment, the recommendations of the Early Treatment for Retinopathy of Prematurity Cooperative Group were followed.22 The International Classification is based on an ordinal scale with higher numbers indicating a more severe outcome: 0, 1, 2, 3, 3+, 4, and 5.

ified according to the International Classification.21 For treatment, the recommendations of the Early Treatment for Retinopathy of Prematurity Cooperative Group were followed.22 The International Classification is based on an ordinal scale with higher numbers indicating a more severe outcome: 0, 1, 2, 3, 3+, 4, and 5. A prespecified key secondary endpoint was time between day of birth and day of discharge from neonatal care. Other secondary outcome measures included incidence of BPD and intraventricular hemorrhage (IVH) and assessment of growth (weight, length, and head circumference). BPD was assessed by need for oxygen use during the first 28 days after birth and by oxygen challenge testing at PMA of 363/7 weeks.23,24 Definitions of mild, moderate, and severe BPD were based on the National Institute of Child Health and Human Development criteria for preterm infants born before 32 weeks of gestation.23 The presence of cerebral hemorrhage was assessed by cranial ultrasound scanning before study inclusion, at postnatal days 3, 7, 14, and 21 (±1 day), and at PMA 40 weeks (±4 days), and graded between 0 and 4 using the Papile/Bowerman scoring method.25,26 Ultrasound images were graded by a single reader masked to study group. Brain volumetric measurements were performed on magnetic resonance images obtained at PMA 40 weeks and will be reported separately.

1 (±1 day), and at PMA 40 weeks (±4 days), and graded between 0 and 4 using the Papile/Bowerman scoring method.25,26 Ultrasound images were graded by a single reader masked to study group. Brain volumetric measurements were performed on magnetic resonance images obtained at PMA 40 weeks and will be reported separately. Safety Assessments Adverse events (AEs) and serious AEs (SAEs) were recorded from receipt of informed consent until final study examination/sampling at PMA 40 weeks. Investigator Verbatim Terms describing AEs were coded using the Medical Dictionary for Regulatory Activities (MedDRA; version 16.0) to MedDRA Preferred Terms (Table I [available at www.jpeds.com] for reported Verbatim Terms for commonly observed AEs). The Methods section in Appendix 2 describes other definitions and assessments of AEs.

atim Terms describing AEs were coded using the Medical Dictionary for Regulatory Activities (MedDRA; version 16.0) to MedDRA Preferred Terms (Table I [available at www.jpeds.com] for reported Verbatim Terms for commonly observed AEs). The Methods section in Appendix 2 describes other definitions and assessments of AEs. Statistical Analyses A sample size of 40 infants per treatment group (80 total) was estimated to provide 80% power (significance level, 5%) to demonstrate a statistically significant difference in distribution of ROP maximum severity between groups (primary endpoint). The ROP outcome stages in the current study were classified as 0, 1, 2, 3, and >3, which is the expected possible range of ROP stages that would be encountered in the study, based on analysis of Swedish infants screened for ROP from 2004 to 2008. Based on Swedish registry data,27 the proportion of children with each ROP outcome for the control group is provided in Table II (available at www.jpeds.com). An estimated treatment effect among treated infants is also presented. The estimated treatment effect was calculated based on the following assumptions: For each outcome of the maximum severity of ROP stage, it is assumed that 25% of the children will not benefit from the treatment, 25% will have their maximum severity of ROP stage reduced by 1 level (eg, from 2 to 1), and 50% will have their maximum severity of ROP stage reduced 2 levels (eg, from 3+ to 2). The null hypothesis tested whether the distribution of maximum severity of ROP stage across all retinal examinations was the same for both treatment groups. Assuming a 30% nonevaluable/dropout rate, 120 infants (60 per group) were to be randomly assigned. The study was powered for the primary ROP endpoint only. A sample size of 80 evaluable infants was also estimated to provide adequate power for the key secondary endpoint (time to discharge from neonatal intensive care), but no power calculations were performed for BPD or IVH secondary endpoints. The Methods section in Appendix 2 provides additional details on the statistical evaluation of all endpoints.

infants was also estimated to provide adequate power for the key secondary endpoint (time to discharge from neonatal intensive care), but no power calculations were performed for BPD or IVH secondary endpoints. The Methods section in Appendix 2 provides additional details on the statistical evaluation of all endpoints. The intention-to-treat population included all enrolled infants assigned a randomization number. The full analysis set (FAS) was defined as all randomized infants receiving study drug or standard care. For rhIGF-1/rhIGFBP-3–treated infants, the FAS was the same as the intention-to-treat population because all randomized infants started treatment. The evaluable set (ES) included treated infants in the FAS who had ≥70% of serum IGF-1 levels within the target range (28–109 μg/L) and who received ≥70% of the intended duration of infusion of rhIGF-1/rhIGFBP-3 (overall infusion length excluding interruptions of ≥1 hour). For the standard care group, the FAS was also considered to be the ES. The safety population included randomized infants receiving study drug or standard care for whom ≥1 safety assessment was completed. The pharmacokinetic population included infants receiving study drug who had ≥1 blood samples drawn after administration.

he standard care group, the FAS was also considered to be the ES. The safety population included randomized infants receiving study drug or standard care for whom ≥1 safety assessment was completed. The pharmacokinetic population included infants receiving study drug who had ≥1 blood samples drawn after administration. Results The first infant was enrolled September 19, 2014, and the last infant completed March 30, 2016. Overall, 121 infants were enrolled, of whom 29 did not complete the study (19 of 29 owing to death; Figure 1). Sixty-one infants received rhIGF-1/rhIGFBP-3 and 60 standard of care. Thirty-five of 61 treated infants (57.4%) and 32 of 60 control infants (53.3%) were born before 26 weeks gestational age (Table III). Mean average daily dose of rhIGF-1/rhIGFBP-3 was 248.1 μg/kg/24 hours (range, 131.1–250.0 μg/kg/24 hours); the total duration of exposure was 23.8 days (range, 0.1–45.3 days); the ratio of duration of exposure to expected duration (birth to 296/7 weeks PMA) was 0.86 (range, 0.0–1.0); and the number of infusion interruptions of ≥1 hour was 4.0 per treated infant (range, 0–83 per treated infant). Table IV (available at www.jpeds.com) provides details on exposure by gestational age strata. Among treated infants, 52 of 61 received ≥70% of the expected treatment duration and 28 of 61 had ≥70% of IGF-1 levels within the target range. Overall target exposure (based on duration and IGF-1 level) was achieved for 24 of 61 treated infants (ES). The Results section in Appendix 2 and Figure 2 (available at www.jpeds.com) summarize information on attained serum IGF-1 levels.

treatment duration and 28 of 61 had ≥70% of IGF-1 levels within the target range. Overall target exposure (based on duration and IGF-1 level) was achieved for 24 of 61 treated infants (ES). The Results section in Appendix 2 and Figure 2 (available at www.jpeds.com) summarize information on attained serum IGF-1 levels. Primary Endpoint: ROP Considering the FAS, 25.5% of rhIGF-1/rhIGFBP-3–treated infants developed ROP stage ≥3 vs 18.0% of controls; there was no statistically significant difference between the 2 groups in distribution of maximum severity of ROP (P = .06; Table V). In the ES, similar proportions of treated vs control infants had ROP of stage ≥3 (18.2% vs 18.0%, respectively; P = .24 for severity distribution between groups). A breakdown by gestational age strata for ROP/other endpoints is presented in Table VI, Table VII, and Table VIII (available at www.jpeds.com). The number of infants who received treatment for ROP was similar between treatment groups: standard of care, 7 (all laser therapy); rhIGF-1/rhIGFBP-3, 7 (6 laser therapy, 1 anti-vascular endothelial growth factor only). Post hoc analysis of IGF-1 levels by ROP severity (<3, ≥3) across weeks after birth showed a clear separation for mean IGF-1 profiles between treated and control infants; in the rhIGF-1/rhIGFBP-3 group, an association of higher serum IGF-1 with less severe ROP was observed (Figure 3, A).

evels, but a similar trend toward a decrease in grades 3–4 IVH in the ES vs the FAS was observed. Although it remains unclear why such a marked effect was seen for BPD relative to ROP, these observations support the hypothesis that increasing serum IGF-1 levels can decrease the severity of complications of prematurity. An analysis of the achieved serum IGF-1 levels indicates that attainment of target levels could be further optimized to potentially improve outcomes. An interim pharmacokinetic analysis in the first 10 treated infants supported the appropriateness of the 250 μg/kg/24 hours dose to achieve IGF-1 target levels.20 However, in the treated population, although the proportion of IGF-1 measurements within target was substantially greater than among controls, fewer than one-half of infants (28/61 [45.9%]) achieved ≥70% of IGF-1 measurements within the target range (see the Results section in Appendix 2). It is possible that the slightly lower than anticipated target attainment could relate to technical aspects of drug administration or that intercurrent proinflammatory states could have led to fluctuations in IGF-1 levels.34 Alternatively, these observations may point to a need for further dose optimization.

Primary Endpoint: ROP Considering the FAS, 25.5% of rhIGF-1/rhIGFBP-3–treated infants developed ROP stage ≥3 vs 18.0% of controls; there was no statistically significant difference between the 2 groups in distribution of maximum severity of ROP (P = .06; Table V). In the ES, similar proportions of treated vs control infants had ROP of stage ≥3 (18.2% vs 18.0%, respectively; P = .24 for severity distribution between groups). A breakdown by gestational age strata for ROP/other endpoints is presented in Table VI, Table VII, and Table VIII (available at www.jpeds.com). The number of infants who received treatment for ROP was similar between treatment groups: standard of care, 7 (all laser therapy); rhIGF-1/rhIGFBP-3, 7 (6 laser therapy, 1 anti-vascular endothelial growth factor only). Post hoc analysis of IGF-1 levels by ROP severity (<3, ≥3) across weeks after birth showed a clear separation for mean IGF-1 profiles between treated and control infants; in the rhIGF-1/rhIGFBP-3 group, an association of higher serum IGF-1 with less severe ROP was observed (Figure 3, A). Key Secondary Endpoint: Time to Discharge From Neonatal Care In the overall population (FAS), the median time to discharge was 82 days (range, 55–115 days) in the rhIGF-1/rhIGFBP-3 group and 74 days (range, 51–107 days) among controls. The difference between groups was not statistically significant (P = .37). In the ES, median time to discharge was 74 days in both study arms (P = .43).

overall population (FAS), the median time to discharge was 82 days (range, 55–115 days) in the rhIGF-1/rhIGFBP-3 group and 74 days (range, 51–107 days) among controls. The difference between groups was not statistically significant (P = .37). In the ES, median time to discharge was 74 days in both study arms (P = .43). Secondary Endpoints A statistically significant difference in distribution of BPD severity was observed between groups (FAS, P = .04; ES, P = .02; Table V), with an apparent shift toward milder BPD cases with treatment. In the FAS, 21.3% of infants with BPD assessments (10/47) in the rhIGF-1/rhIGFBP-3 group had severe BPD vs 44.9% of controls (22/49). The difference between treatment groups was more pronounced in the ES, with 4.8% of treated infants (1/21) with severe BPD vs 44.9% of controls. The BPD analyses did not include the 19 all-cause deaths that occurred in the study, including 12 deaths (20%) in the rhIGF-1/rhIGFBP-3 group and 7 deaths (12%) in the standard care group. We conducted post hoc analyses for BPD that included all-cause deaths during the study period. In the additional analyses, deaths were grouped with severe BPD. These analyses showed that, when deaths were included, the trend for a decrease in severe BPD among treated infants remained (37.3% in rhIGF-1/rhIGFBP-3 group vs 51.8% in the standard of care group Table IX [available at www.jpeds.com]).

the study period. In the additional analyses, deaths were grouped with severe BPD. These analyses showed that, when deaths were included, the trend for a decrease in severe BPD among treated infants remained (37.3% in rhIGF-1/rhIGFBP-3 group vs 51.8% in the standard of care group Table IX [available at www.jpeds.com]). In this study, we observed a shift in the distribution of severity of BPD, which was most evident in the difference between groups in the percentage of subjects with severe BPD. If moderate cases were combined with severe cases and deaths, 52.5% of the rhIGF-1/rhIGFBP-3 group (31/59) and 60.7% of the standard care group (34/56) had BPD. When early deaths (within the first 14 days) were excluded, as suggested by Higgins et al in the 2016 National Institute of Child Health and Human Development [NICHD] workshop,28 31.5% of the rhIGF-1/rhIGFBP-3 group (17/54) vs 50.0% of the standard of care group (27/54) had severe BPD. Post hoc analysis of IGF-1 levels by BPD severity (mild, moderate, severe) across weeks after birth showed a clear separation for mean IGF-1 profiles between treated and control infants, with a higher serum IGF-1 associated with less severe BPD in the treated group (Figure 3, B). A smaller proportion of infants in the rhIGF-1/rhIGFBP-3 group had severe IVH (grades 3–4) than among controls (Table V): FAS, 13.1% vs 23.3%, respectively; ES, 8.3% vs 23.3%, respectively (not statistically significant). The numbers of IVH events were too small to explore an exposure-response relationship between IGF-1 levels and IVH grade.

infants in the rhIGF-1/rhIGFBP-3 group had severe IVH (grades 3–4) than among controls (Table V): FAS, 13.1% vs 23.3%, respectively; ES, 8.3% vs 23.3%, respectively (not statistically significant). The numbers of IVH events were too small to explore an exposure-response relationship between IGF-1 levels and IVH grade. rhIGF-1/rhIGFBP-3 treatment did not affect rates of change of length, weight, or head circumference compared with standard of care (Figure 4 [available at www.jpeds.com]). Similar results were seen in both the FAS and ES. Safety All infants had ≥1 treatment-emergent AEs, with the exception of 1 infant in the rhIGF-1/rhIGFBP-3 group. For 13.1% of treated infants (8/61), treatment-emergent AEs were considered possibly related to study drug. The most common events overall (related or unrelated) were patent ductus arteriosus (90.2% of treated infants vs 85.0% of controls) and neonatal anemia (75.4% vs 73.3%, respectively; Table X; available at www.jpeds.com). Additionally, 78.7% of treated infants (48/61) and 61.7% of control infants (37/60) had SAEs.

Figure 3. IGF-1 levels by A, ROP severity* and B, BPD severity† by postnatal week. *Mean (±SE) serum IGF-1 levels and ROP severity (<3, ≥3) in the rhIGF-1/rhIGFBP-3 and standard neonatal care groups by postnatal week. †Mean (±SE) serum IGF-1 levels and BPD severity (mild, moderate, or severe) in the rhIGF-1/rhIGFBP-3 and standard neonatal care groups by postnatal week. Note: If an infant had multiple IGF-1 levels in a day, then IGF-1 level was averaged for the day. Figure 4. A, Average weight, B, length, and C, head circumference by treatment group (FAS). Table I. Verbatim terms reported in relation to common treatment-emergent AEs AE/preferred term Verbatim terms Patent ductus arteriosus “Persistent ductus arteriosus,” “patent ductus arteriosus that requires treatment,” “patent ductus arteriosus that need treatment,” “PDA,” “large PDA,” “large patent ductus arteriosus,” “patent ductus arteriosus, moderately hemodynamically significant,” “patent ductus arteriosus, minor hemodynamically significant,” “PDA of 3.4 mm detected on first ECHO,” “reopening of ductus arteriosus,” “reopening of patent ductus arteriosus,” “big patent ductus arteriosus,” “small patent ductus arteriosus,” “patent ductus arteriosus, hemodynamically significant,” “patent ductus arteriosa without any clinical significance,” “small PDA again,” “homonymic significant patent ductus arteriosus,” “patent ductus arteriosus (PDA),” “small closing patent ductus arteriosus,” “patent ductus arteriosus 2.1 mm,” “patent ductus arteriosus with shunt pulsating flow pattern,” “PDA prior to treatment,” “patent ductus arteriosus from echocardiogram,” “PDA observed on echocardiogram” Anemia neonatal “Anemia,” “anemi,” “anemia requiring transfusion,” “anaemia,” “anaemia of prematurity,” “anemie,” “tired, pale, listless due to anemia,” “anaemia needed transfusion,” “neonatal anaemia,” “anemia of prematurity,” “symptomatic anemia” Neonatal respiratory distress syndrome “Respiratory distress syndrome,” “RDS,” “respiratory distress,” “hyaline membrane disease,” “infant respiratory distress syndrome,” “respiratory distress/hyaline membrane disease” Jaundice neonatal “Jaundice,” “jaundiced,” “neonatal jaundice,” “icterus,” “jaundice requiring phototherapy,” “intermittent jaundice,” “jaundice requiring treatment with phototherapy” Infantile apneic attack “Apneas,” “apneas,” “apnea,” “apnea of prematurity,” “apnea,” “apnoea,” “prematurity apnea,” “apnoea crisis,” “recurrent apnea,” “recurrent apnoeas,” “sudden and severe apnoea,” “severe apnoea,” “apnea episode,” “apnea neonatal,” “crisis of apnea” Neonatal hypotension “Hypotension,” “low blood pressure,” “hypotension (low blood pressure),” “systemic hypotension,” “hypotensive (mean 21),” “intermittent hypotension,” “hypotension (intermittent),” “hypotension, MAP 26” Hyperglycemia “Hyperglycemia,”

drug. The most common events overall (related or unrelated) were patent ductus arteriosus (90.2% of treated infants vs 85.0% of controls) and neonatal anemia (75.4% vs 73.3%, respectively; Table X; available at www.jpeds.com). Additionally, 78.7% of treated infants (48/61) and 61.7% of control infants (37/60) had SAEs. Fatal SAEs were reported in 19.7% of treated infants (12/61) and 11.7% of control infants (7/60); none were considered treatment related (Table XI). The imbalance of deaths was concentrated in the subgroup born at a gestational age of <26 weeks and driven by 1 iatrogenic death owing to a misplaced umbilical catheter (which caused intra-abdominal hemorrhage leading to multiorgan failure), 1 infant with severe respiratory distress with onset before study drug infusion, and 3 more cases of necrotizing enterocolitis (NEC). Overall, NEC AEs (fatal and nonfatal) were balanced between groups (6 infants [9.8%] in the treated group vs 5 infants [8.3%] in the standard care group). Further explorations revealed no apparent differences in modes of delivery, transfusions, sepsis events, antibiotic/probiotic use, or nutrition between infants with NEC who died and those who survived. In addition to NEC (which can be misclassified as a bowel perforation), 1 infant who died in the rhIGF-1/rhIGFBP-3 group had an intestinal perforation and 2 infants who died in the standard care group had intestinal perforations. Overall, including nonfatal cases, there was 1 case of intestinal perforation in the rhIGF-1/rhIGFBP-3 group, compared with 5 in the standard of care group. A full summary of safety data, including severe AEs, and AEs of interest is provided in the Results section in Appendix 2.

care group had intestinal perforations. Overall, including nonfatal cases, there was 1 case of intestinal perforation in the rhIGF-1/rhIGFBP-3 group, compared with 5 in the standard of care group. A full summary of safety data, including severe AEs, and AEs of interest is provided in the Results section in Appendix 2. Discussion This study did not meet the primary endpoint of reducing maximum severity of ROP. However, there was a marked decrease in the proportion of rhIGF-1/rhIGFBP-3–treated infants who developed severe BPD, and a shift in patterns of BPD severity to milder cases; a similar trend was observed for IVH but was not statistically significant. rhIGF-1/rhIGFBP-3 was well-tolerated in this study; there were no safety signals. The overall results of this study support continued evaluation of rhIGF-1/rhIGFBP-3 for the prevention of complications of prematurity in extremely preterm infants.

similar trend was observed for IVH but was not statistically significant. rhIGF-1/rhIGFBP-3 was well-tolerated in this study; there were no safety signals. The overall results of this study support continued evaluation of rhIGF-1/rhIGFBP-3 for the prevention of complications of prematurity in extremely preterm infants. Although a trend toward higher serum IGF-1 in rhIGF-1/rhIGFBP-3–treated infants with no or lower stages of ROP was observed, the reason for lack of overall effect on ROP is not clear. It may be that the dosing regimen of rhIGF-1/rhIGFBP-3 requires further optimization. Alternatively, practice variability in RetCam assessments may have limited the observed treatment effect. A further consideration is the variation between sites in target oxygen saturation measures and compliance with these levels. This consideration is particularly important in light of findings from 5 landmark clinical trials, published shortly before commencement of the current study, which studied the effects of targeting lower (85%–89%) vs higher (91%–95%) oxygen saturation targets on neurodevelopmental impairment and severe ROP in extremely preterm infants.29–31 The US Surfactant, Positive Pressure, and Oxygenation Randomized Trial (SUPPORT) associated a lower oxygenation target of 85%–90%, as compared with a higher target of 91%–95%, with increased mortality and decreased ROP among survivors.29 Combined findings from the Benefits of Oxygen Saturation Targeting II (BOOST II) trials,30 conducted in the UK, Australia, and New Zealand, were largely consistent with the SUPPORT study. The investigators reported increased mortality in the lower oxygen target group vs the higher (among infants with revised oximeter software) and a lower rate of severe ROP in the lower saturation range vs the higher (among infants in the total study population).30 Conversely, the Canadian Oxygen Trial (COT) found no significant difference in mortality or the incidence of severe ROP between the lower and higher oxygen target groups.31 Findings from the SUPPORT and BOOST II trials led to changes in neonatal clinical practice, where the risk of ROP needed to be considered against the risk of increased mortality, with the result that many centers across countries adopted higher oxygen target ranges.

between the lower and higher oxygen target groups.31 Findings from the SUPPORT and BOOST II trials led to changes in neonatal clinical practice, where the risk of ROP needed to be considered against the risk of increased mortality, with the result that many centers across countries adopted higher oxygen target ranges. More recent studies have reported increased rates and severity of ROP among extremely preterm infants after changing from lower to higher oxygen target ranges.32,33 It is possible that higher oxygen target ranges used in centers in the current study may have affected the observed incidences of ROP in the study population. The key secondary endpoint of time to discharge from neonatal care also was not statistically different between groups. However, it is difficult to draw conclusions from these data because assessment of length of stay in intensive care is easily confounded. There also was a lack of effect of rhIGF-1/hIGFBP-3 on growth measures, which could have been confounded by nutritional differences across sites; unfortunately, nutrition data were not recorded consistently, precluding further analysis of a potential effect.

ent of length of stay in intensive care is easily confounded. There also was a lack of effect of rhIGF-1/hIGFBP-3 on growth measures, which could have been confounded by nutritional differences across sites; unfortunately, nutrition data were not recorded consistently, precluding further analysis of a potential effect. By contrast, the shift in severity distribution of BPD was marked and suggests a hypothesis-generating trend with higher serum IGF-1. The rate of IVH events was too low to evaluate correlations with IGF-1 levels, but a similar trend toward a decrease in grades 3–4 IVH in the ES vs the FAS was observed. Although it remains unclear why such a marked effect was seen for BPD relative to ROP, these observations support the hypothesis that increasing serum IGF-1 levels can decrease the severity of complications of prematurity.

pendix 2). It is possible that the slightly lower than anticipated target attainment could relate to technical aspects of drug administration or that intercurrent proinflammatory states could have led to fluctuations in IGF-1 levels.34 Alternatively, these observations may point to a need for further dose optimization. The observed safety profile of rhIGF-1/rhIGFBP-3 was encouraging. There was an imbalance in deaths between the treatment arms concentrated in the subgroup with a gestational age of <26 weeks; however, careful evaluation of the causes of death by an independent data monitoring committee did not raise safety concerns (see the Results section in Appendix 2). With regard to other safety considerations, despite the duration of intravenous infusion, data collected to date across rhIGF-1/rhIGFBP-3 clinical trials have shown no signs of an increase in pathogen-confirmed sepsis. Also in the current study, the proportion of infants with hypoglycemia was similar between groups. Of note, rhIGF-1/rhIGFBP-3 was associated with a lower incidence of hyperglycemia relative to standard care, suggesting possible improvements in glycemic control. A previous phase of the study (B/C) found that rhIGF-1/rhIGFBP-3 was well-tolerated in extremely preterm infants and that the incidence of AEs was similar for infants treated with rhIGF-1/rhIGFBP-3 vs control infants.35 Further, consistent with the current trial, rates of hypoglycemia were similar between groups in the phase B/C study. These outcomes will be of interest for further investigation.

ated in extremely preterm infants and that the incidence of AEs was similar for infants treated with rhIGF-1/rhIGFBP-3 vs control infants.35 Further, consistent with the current trial, rates of hypoglycemia were similar between groups in the phase B/C study. These outcomes will be of interest for further investigation. The limitations of this study included the smaller than anticipated number of infants eligible for inclusion in the ES, albeit an arbitrary a priori definition. Another limitation was the randomization by gestational age centrally and not by study site, which resulted in an imbalance between gestational age strata at high enrolling sites, and may have been an additional confounder. Additionally, there were technical challenges to performing certain assessments (eg, RetCam) and variability of practices across and within sites. The analysis of AEs was limited by some reporting inconsistency, largely owing to a lack of international consensus on defining and classifying AEs in infants based on their severity. A further consideration is that infants who died were not included in the analyses, and this factor may have had an impact on our findings related to the degree of difference in severe BPD between the 2 groups. However, post hoc analyses showed similar, although attenuated, results when all-cause death was included with severe BPD as an outcome. Although the prevalence of ROP, BPD, and IVH was comparable with the observed prevalence in the extremely preterm population,36,37 the sample size was relatively small. The latter limitation obviously precludes the evaluation of a possible confounding effect of mortality on differences in respiratory morbidity between study groups at a gestational age of 36 weeks. A larger investigation that examines a higher dose IGF-I with more standardized and harmonized approaches to safety assessments is needed to clarify this issue, including the evaluation of treatment effects on NEC. Preclinical models indicate that IGF-I supplementation and subsequent enhancement of vascular endothelial growth factor receptor-2 signaling may have a preventive effect on development of NEC.38–42 Higher levels of plasma IGF-1 were also associated with reduced incidence of NEC in a prospective analysis of very low birth weight infants in the NIRTURE study.43 The administration of IGF-1/IGFBP3 would not be expected to increase the risk of NEC.

r-2 signaling may have a preventive effect on development of NEC.38–42 Higher levels of plasma IGF-1 were also associated with reduced incidence of NEC in a prospective analysis of very low birth weight infants in the NIRTURE study.43 The administration of IGF-1/IGFBP3 would not be expected to increase the risk of NEC. Of note, there were additional protocol limitations associated with the methodology used to assess presence of IVH (a single reader of cranial ultrasound examinations, grading only by the Papile/Bowerman method25,26), which may have impacted the frequency of the various grades seen in both treatment groups. Post hoc analyses are ongoing using an adjudicated reader and 2 additional scoring methods44,45 for the assessment of ultrasound scans to evaluate whether the frequency of the various grades varied as a function of the scoring method; details of the comparison (also including magnetic resonance imaging data) will be reported separately.

are ongoing using an adjudicated reader and 2 additional scoring methods44,45 for the assessment of ultrasound scans to evaluate whether the frequency of the various grades varied as a function of the scoring method; details of the comparison (also including magnetic resonance imaging data) will be reported separately. Based on the results of this study and given the clear unmet need for therapies to decrease the overall morbidity burden in extremely preterm infants,46–49 continued investigation of rhIGF-1/rhIGFBP-3 for the prevention of complications of prematurity is planned in a larger clinical trial. The protocol for this study is in development, with the prevention of the onset of chronic lung disease (indicated by reductions in respiratory complications) at a corrected age of 12 months as the primary endpoint, and a decrease in the severity of BPD at a PMA of 36 weeks and IVH severity at a PMA of 40 weeks as separate key secondary endpoints. Of importance, long-term outcomes after short-term exposure to rhIGF-1/rhIGFBP-3 also are being investigated in an extension study (PEDAL; NCT02386839) over 5 years.

he primary endpoint, and a decrease in the severity of BPD at a PMA of 36 weeks and IVH severity at a PMA of 40 weeks as separate key secondary endpoints. Of importance, long-term outcomes after short-term exposure to rhIGF-1/rhIGFBP-3 also are being investigated in an extension study (PEDAL; NCT02386839) over 5 years. Funded by Shire. Shire participated in the design of the study, the collection and analysis of data, and preparation of the clinical study report. D.L. and I.H-P. hold stock/stock options in Premalux AB, and received consulting fees from Shire. B.H. received consulting fees from Premacure AB and Shire. L.R. received consulting fees and research support from Shire. N.M. received consulting fees from Shire, and partial funding from the Department of Health’s National Institute for Health Research Biomedical Research Centre’s funding scheme at University College London Hospitals/University College London. K.B., F.B., J.H., O.M-N., M.vW., and L.S. received consulting fees from Shire. D.D. received consulting fees from Shire, and received consulting fees from Ipsen regarding other indications for IGF-1 therapies. N.B., A.T., M.H., E.J., A.M., and J-K.C. are employees of and own stock/stock options in Shire. M.T.’s university received consulting fees from Shire. A.H. holds stock/stock options in Premalux AB, and received consulting fees from Shire. C.D., A.M., P.R, and C.G. declare no conflicts of interest.

for IGF-1 therapies. N.B., A.T., M.H., E.J., A.M., and J-K.C. are employees of and own stock/stock options in Shire. M.T.’s university received consulting fees from Shire. A.H. holds stock/stock options in Premalux AB, and received consulting fees from Shire. C.D., A.M., P.R, and C.G. declare no conflicts of interest. We thank Rui Tang, an employee of Shire, who provided assistance with additional statistical analysis. We also thank Valérie Boissel, PhD, and Rosalind Bonomally, MSc, of Excel Scientific Solutions, who provided medical writing assistance funded by Shire. The authors thank the infants, their families, and the investigators/contributors who participated in this study (list of participants available at www.jpeds.com [Appendix 1]).

hank Valérie Boissel, PhD, and Rosalind Bonomally, MSc, of Excel Scientific Solutions, who provided medical writing assistance funded by Shire. The authors thank the infants, their families, and the investigators/contributors who participated in this study (list of participants available at www.jpeds.com [Appendix 1]). Appendix 1 Study Investigators and Other Contributors by Clinical Site/Affiliation Clinical sites Study investigators/contributors Lund University, Skane University Hospital, Department of Clinical Sciences Lund, Pediatrics, Lund, Sweden Margareta Gebka, RN; Ann-Cathrine Berg, RN Department of Neonatology, Department of Clinical Science, Intervention and Technology (CLINTEC), Karolinska Institute and Karolinska University Hospital, Stockholm, Sweden MireilleVanpee, MD; DirkWackernagel, MD; Stefan Löfgren, MD Careggi University Hospital of Florence, University of Florence, Florence, Italy Chiara Poggi, MD; Elena Gozzini, MD; Tommaso Bianconi, MD; Saverio Frosini, MD; Simone Pratesi, MD; Iuri Corsini MD; Venturella Vangi, MD; Ada Kura, PhD UCL EGA Institute for Women’s Health, London, UK Rashmi Gandhi, MBBS, MRCPCH; Joanna Lawson, FRCS(Ed), FRCOphth; Gina Burquis, RN The Department of Pediatrics and the Wellcome-Trust MRC Institute of Metabolic Science, University of Cambridge, Cambridge, UK Lynn Thomson, RSCN The Children’s Hospital at the University of Oklahoma Health Sciences Center, Oklahoma City, OK Michael Siatkowski, MD; Birju Shah, MD; Michael McCoy; Michelle Blunt; Kelly Satnes; Kimberly Benjamin; Lindsay DePace East Carolina University, Greenville, NC Devon Kuehn, MD, Sherry L. Moseley, RN; Ryan Moore, MD St Mary’s Hospital, Central Manchester University NHS Foundation Trust, Manchester Academic Health Science Centre and Division of Developmental Biology & Medicine, School of Medical Sciences, University of Manchester, Manchester, UK S. Nedungadi, MRCPCH; S. Biswas, FRCOphth; G. Ciotti, MD; V. Tang, MD; A. Hendrickson; K. Dockery; N. Booth Departments of Pediatrics and Ophthalmology, University of Wisconsin, Madison, WI Michael Struck, MD; De-Ann Pillers, MD, PhD; Yasmin Bradfield, MD; Melanie Schmitt, MD; Erica Riedesel, MD Neonatal Intensive Care Unit, St Peter’s Hospital, Chertsey, Surrey, UK Nicky Holland, RN; Edit Molnar, MD Department of Woman and Child Health, University Hospital A.

lmology, University of Wisconsin, Madison, WI Michael Struck, MD; De-Ann Pillers, MD, PhD; Yasmin Bradfield, MD; Melanie Schmitt, MD; Erica Riedesel, MD Neonatal Intensive Care Unit, St Peter’s Hospital, Chertsey, Surrey, UK Nicky Holland, RN; Edit Molnar, MD Department of Woman and Child Health, University Hospital A. Gemelli, Rome, Italy Costantino Romagnoli, MD; Velia Purcaro, MD; Francesca Serrao, MD; Patrizia Papacci, MD; Mikael GhennetTesfagabir, MD; Rossella Iannotta, MD; Maria Sofia Cori, MD; Antonio Baldascino, MD; Cecilia Zuppi, MD; Giovanni Luca Scaglione, MSc, PhD; Andrea Cocci, MLT; Elisa De Paolis, MSc; Cesare Colosimo, MD; Tommaso Verdolotti, MD; Marino Gentile; Marta Romagnoli VU University Medical Center, Amsterdam, the Netherlands Dana Yumani, MD; Annemieke de Lang Institute of Translational Medicine, University of Liverpool, Liverpool, UK Patrick McGowan, RGN; Karen Harvey, RSCN; Joanne Windrow, RSCN Department for Women’s and Children’s Health, University of Padua, Padua, Italy Paola Lago, MD Department of Pediatrics, University Hospitals Coventry and Warwickshire NHS Trust, Coventry, UK Prakash Satodia, MD University of Mississippi Medical Center, Jackson, MS Abhay Bhatt, MD University of South Alabama Children’s & Women’s Hospital, Mobile, AL Fabien Eyal, MD Neonatal-Perinatal Medicine, Mercy and St. Luke’s Hospitals, St. Louis, MO Erzsebet Jung, MD Department of Neonatology, Poznan University of Medical Sciences, Poznan, Poland Janusz Gadzinowski, MD Additional contributors (not affiliated with study sites) Affiliation Contributor University of Rochester Medical Center, Rochester, NY Matthew D. Gearinger, MD; Mina M. Chung, MD; Henry Wang, MD Department of Pediatric Ophthalmology and Strabismus, Ludwig-Maximilians-University Munchen, Munich, Germany Birgit Lorenz, MD Appendix 2 Methods Safety Monitoring. A clinical study monitor ensured that the investigation was conducted according to protocol design and regulatory requirements through frequent site visits and communications. Infant safety was monitored on a continuous basis until the last infant completed the last scheduled study visit/assessment, and quarterly safety review meetings were held throughout the study. Safety data collected during the trial were reported annually to competent authorities in the form of a Development Safety Update Report. Additionally, an independent data monitoring committee provided an ongoing review and assessment of safety data.

and quarterly safety review meetings were held throughout the study. Safety data collected during the trial were reported annually to competent authorities in the form of a Development Safety Update Report. Additionally, an independent data monitoring committee provided an ongoing review and assessment of safety data. A special data monitoring committee meeting was to be convened if the safety-related stopping rules (ie, if a death occurred that was considered possibly or probably related to the study drug) were met. Interim Analyses. Two interim analyses were planned and conducted. The first was an analysis of dosing/target attainment conducted when 10 treated infants had completed the rhIGF-1/IGFBP-3 dosing phase of the study (reported previously by Chung et al1). In the second, a conditional power analysis was performed on unmasked data when 60 infants had completed the study or withdrew early. The analysis was performed by an external independent statistician to assess the appropriateness of the sample size and assumptions made regarding the distribution of the maximum severity of ROP.

cond, a conditional power analysis was performed on unmasked data when 60 infants had completed the study or withdrew early. The analysis was performed by an external independent statistician to assess the appropriateness of the sample size and assumptions made regarding the distribution of the maximum severity of ROP. Concomitant Medications. Predefined medications (including hydrocortisone, betamethasone, dexamethasone, ibuprofen, dopamine, dobutamine, epinephrine, budesonide, furosemide, nitric oxide, Curosurf/other surfactants, indomethacin, and other medications given to treat AEs) and procedures administered to infants from time of informed consent through to a PMA of 404/7 weeks were regarded as permitted. Treatment with fresh frozen plasma, which is associated with a short-term increase of serum concentrations of IGF-1,2 was permitted and recorded. Anti-vascular endothelial growth factor medication and rh erythropoietin were prohibited. Delivery of parenteral/enteral nutrition (including glucose) was performed according to local guidelines.

frozen plasma, which is associated with a short-term increase of serum concentrations of IGF-1,2 was permitted and recorded. Anti-vascular endothelial growth factor medication and rh erythropoietin were prohibited. Delivery of parenteral/enteral nutrition (including glucose) was performed according to local guidelines. IGF-1 Blood Sampling. Blood samples for IGF-1 measurement were obtained at baseline (predose), 12 hours ± 30 minutes, and 24 hours ± 30 minutes after the infusion started, then every 72 ± 1 hours until a PMA of 296/7 weeks, and then 1 hour ± 30 minutes after the infusion stopped for treated infants. Controls were tested at baseline (day of birth), 12 hours ± 30 minutes, 24 hours ± 30 minutes, and then every 168 ± 1 hours after baseline until a PMA of 296/7 weeks. Additional sampling occurred at PMA 32, 36, and 40 weeks. IGF-1 was determined locally by a validated ELISA (Mediagnost GmbH, Reutlingen, Germany), and in a central laboratory by radioimmunoassay (PPD Laboratories, Richmond, Virginia).3

30 minutes, and then every 168 ± 1 hours after baseline until a PMA of 296/7 weeks. Additional sampling occurred at PMA 32, 36, and 40 weeks. IGF-1 was determined locally by a validated ELISA (Mediagnost GmbH, Reutlingen, Germany), and in a central laboratory by radioimmunoassay (PPD Laboratories, Richmond, Virginia).3 Safety Assessments. Hypoglycemia was defined within the trial as plasma glucose levels of <2.5 mmol/L measured during predefined schedules of blood glucose monitoring (see Blood Glucose Monitoring section), and from blood glucose measurements performed on clinical indication. Hyperglycemia was defined as plasma glucose levels of >10 mmol/L. Echocardiography was interpreted by a pediatric cardiologist to define presence of clinically significant patent ductus arteriosus on postnatal days 2 and 4, and according to clinical judgment after these time points. All infants were assessed for tonsillar hypertrophy by visual examination of their tonsils during a weekly physical examination and at term age (PMA of 40 weeks) by a pediatric specialist.

nically significant patent ductus arteriosus on postnatal days 2 and 4, and according to clinical judgment after these time points. All infants were assessed for tonsillar hypertrophy by visual examination of their tonsils during a weekly physical examination and at term age (PMA of 40 weeks) by a pediatric specialist. Blood Glucose Monitoring. rhIGF-1/rhIGFBP-3 Group. For infants who were fed every 2 hours or who were fed intravenously or through continuous enteral feed, blood glucose measurement was performed every 4 hours during the first 72 hours (days 1–3) after starting infusion with rhIGF-1/rhIGFBP-3. For infants who were fed every 3 hours, blood glucose was measured every 3 hours during the first 72 hours (days 1–3) after starting infusion with rhIGF-1/rhIGFBP-3. During days 4–7, blood glucose was measured every 6 hours for those infants who were fed either every 2 or every 3 hours. After the first 7 days of the infusion, blood glucose was analyzed twice daily until the completion of infusion (PMA of 296/7 weeks). Thereafter, blood glucose was measured weekly until the end of the study. If blood glucose was <3 mmol/L during ongoing rhIGF-1/rhIGFBP-3 infusion and more frequent monitoring was deemed necessary based on the clinical judgment of the investigator, a blood glucose sample was then taken every hour until a blood glucose of ≥3 mmol/L was reached. Clinical sites had specific protocols in place to monitor hypoglycemia.

mmol/L during ongoing rhIGF-1/rhIGFBP-3 infusion and more frequent monitoring was deemed necessary based on the clinical judgment of the investigator, a blood glucose sample was then taken every hour until a blood glucose of ≥3 mmol/L was reached. Clinical sites had specific protocols in place to monitor hypoglycemia. Standard Neonatal Care Group. Blood glucose was analyzed every 6 hours for the first 7 days. Thereafter, it was analyzed twice daily until a PMA of 296/7 weeks, provided that blood was available for sampling. If blood was not available, the sample collection interval was increased. After PMA 296/7 weeks, blood glucose was measured weekly until the end of the study.

analyzed every 6 hours for the first 7 days. Thereafter, it was analyzed twice daily until a PMA of 296/7 weeks, provided that blood was available for sampling. If blood was not available, the sample collection interval was increased. After PMA 296/7 weeks, blood glucose was measured weekly until the end of the study. Statistical Analyses. The primary endpoint between the 2 groups was analyzed using a generalized Cochran–Mantel–Haenszel row means score statistic with modified ridit scores adjusting for gestational age strata. This statistical test generates an estimate of an association between an exposure and an outcome after adjusting for or taking into account confounding factors. The sample size of 80 evaluable infants also was to provide adequate power for the key secondary endpoint of time to discharge from neonatal care, for which the difference between the 2 treatment groups was tested using the stratified version of the Wilcoxon rank-sum test and adjusted for gestational age status. In the event of death, time to discharge was imputed using the median value of all infants with available data within each gestational age strata. The comparison between treatment groups for ordered category of BPD (absent, mild, moderate, severe) at PMA 36 weeks was made using Cochran–Mantel–Haenszel row mean score statistics with modified ridit scores adjusting for gestational age strata. The grade of the hemorrhage (grades 1–4) was summarized descriptively by treatment group and gestational age strata. The difference in the distribution of IVH severity seen between the rhIGF-1/rhIGFBP-3 (FAS or ES) and standard care groups was analyzed in a post hoc analysis using Cochran–Mantel–Haenszel row mean score statistics. Weight, length, and head circumference were analyzed separately, using a linear mixed-model repeated measurement analysis over all postbaseline visits, with change from baseline in each parameter as the outcome variable. The model included treatment time (days), treatment by time interaction, gestational age strata as a fixed effect, infant as a random effect, and baseline value as covariate. Time (days) was calculated relative to the date of the baseline assessment and was used as a continuous covariate.

meter as the outcome variable. The model included treatment time (days), treatment by time interaction, gestational age strata as a fixed effect, infant as a random effect, and baseline value as covariate. Time (days) was calculated relative to the date of the baseline assessment and was used as a continuous covariate. Results Serum IGF-1 Levels. For rhIGF-1/rhIGFBP-3–treated infants, 66.2% of IGF-1 measurements were within the targeted physiological intrauterine range (28–109 μg/L) vs 6.3% for the standard care group (predose data and data after the end of the final infusion were excluded). Very few IGF-1 measurements (1.5%) in treated infants were above the upper bound of the targeted range. Figure 2 shows the mean (SD) serum IGF-1 concentrations over the duration of the study in both the rhIGF-1/rhIGFBP-3 and standard care groups; mean concentrations were within target range during the infusion period for treated infants and below range for the standard care group. The onset of endogenous IGF-1 production was estimated between weeks 30 and 32 (corresponding approximately with cessation of treatment), after which both groups had mean IGF-1 levels within the target range. Additional Safety Results. Severe AEs. In the rhIGF-1/rhIGFBP-3 vs standard care groups, 73.8% of infants (45/61) vs 65.0% of infants (39/60) experienced SAEs, respectively. Only 1 treated infant (1.6%) had 2 SAEs considered possibly related to study drug (1 AE of IVH and 1 separate AE of apnea).

Results Serum IGF-1 Levels. For rhIGF-1/rhIGFBP-3–treated infants, 66.2% of IGF-1 measurements were within the targeted physiological intrauterine range (28–109 μg/L) vs 6.3% for the standard care group (predose data and data after the end of the final infusion were excluded). Very few IGF-1 measurements (1.5%) in treated infants were above the upper bound of the targeted range. Figure 2 shows the mean (SD) serum IGF-1 concentrations over the duration of the study in both the rhIGF-1/rhIGFBP-3 and standard care groups; mean concentrations were within target range during the infusion period for treated infants and below range for the standard care group. The onset of endogenous IGF-1 production was estimated between weeks 30 and 32 (corresponding approximately with cessation of treatment), after which both groups had mean IGF-1 levels within the target range. Additional Safety Results. Severe AEs. In the rhIGF-1/rhIGFBP-3 vs standard care groups, 73.8% of infants (45/61) vs 65.0% of infants (39/60) experienced SAEs, respectively. Only 1 treated infant (1.6%) had 2 SAEs considered possibly related to study drug (1 AE of IVH and 1 separate AE of apnea). Discontinuations Owing to AEs. A total of 20 infants (11 [18%] in the rhIGF-1/rhIGFBP-3 group and 9 [15%] in the standard care group) discontinued the study owing to treatment-emergent AEs, which were fatal SAEs in 90% of infants (18/20). One additional infant in the treated group had an SAE with fatal outcome, but the primary reason for discontinuation was withdrawal of consent.

in the rhIGF-1/rhIGFBP-3 group and 9 [15%] in the standard care group) discontinued the study owing to treatment-emergent AEs, which were fatal SAEs in 90% of infants (18/20). One additional infant in the treated group had an SAE with fatal outcome, but the primary reason for discontinuation was withdrawal of consent. AEs of Interest. The frequency of pathogen-confirmed sepsis was similar between the rhIGF-1/rhIGFBP-3 and standard care groups (65.9% vs 67.4%, respectively). Coagulase-negative staphylococcal sepsis was reported in similar percentages of infants in each group: 36.4% vs 37.2% in the rhIGF-1/rhIGFBP-3 and standard care groups, respectively. Hypoglycemia occurred in a similar proportion in each group: 29.5% of infants (18/61) in the rhIGF-1/rhIGFBP-3 group had AEs of hypoglycemia (considered possibly related to treatment for 4 infants) vs 31.7% (19/60) in the standard care group. Additionally, 39.3% of infants (24/61) in the rhIGF-1/rhIGFBP-3 group had AEs of hyperglycemia vs 48.3% (29/60) in the standard care group (Table VIII). There were no cases of intracranial hypertension or tonsillar hypertrophy in either group. AEAdverse event BPDBronchopulmonary dysplasia ESEvaluable set FASFull analysis set IGF-1Insulin-like growth factor-1 IVHIntraventricular hemorrhage NECNecrotizing enterocolitis PMAPostmenstrual age rhRecombinant human ROPRetinopathy of prematurity rhIGF-1/rhIGFBP-3rhIGF-1 complexed with its binding protein rhIGFBP-3 SAESerious AE

AEs of Interest. The frequency of pathogen-confirmed sepsis was similar between the rhIGF-1/rhIGFBP-3 and standard care groups (65.9% vs 67.4%, respectively). Coagulase-negative staphylococcal sepsis was reported in similar percentages of infants in each group: 36.4% vs 37.2% in the rhIGF-1/rhIGFBP-3 and standard care groups, respectively. Hypoglycemia occurred in a similar proportion in each group: 29.5% of infants (18/61) in the rhIGF-1/rhIGFBP-3 group had AEs of hypoglycemia (considered possibly related to treatment for 4 infants) vs 31.7% (19/60) in the standard care group. Additionally, 39.3% of infants (24/61) in the rhIGF-1/rhIGFBP-3 group had AEs of hyperglycemia vs 48.3% (29/60) in the standard care group (Table VIII). There were no cases of intracranial hypertension or tonsillar hypertrophy in either group. AEAdverse event BPDBronchopulmonary dysplasia ESEvaluable set FASFull analysis set IGF-1Insulin-like growth factor-1 IVHIntraventricular hemorrhage NECNecrotizing enterocolitis PMAPostmenstrual age rhRecombinant human ROPRetinopathy of prematurity rhIGF-1/rhIGFBP-3rhIGF-1 complexed with its binding protein rhIGFBP-3 SAESerious AE Figure 1. A, Study design and B, patient disposition. *Informed consent was obtained before birth or within 24 hours after birth. † One infant had an SAE with a fatal outcome, but the primary reason for discontinuation was withdrawal of consent. ‡ All infants discontinued owing to an SAE with fatal outcome. § Seven of 9 discontinuations were owing to SAEs with fatal outcome.

formed consent was obtained before birth or within 24 hours after birth. † One infant had an SAE with a fatal outcome, but the primary reason for discontinuation was withdrawal of consent. ‡ All infants discontinued owing to an SAE with fatal outcome. § Seven of 9 discontinuations were owing to SAEs with fatal outcome. Figure 2. Mean (SD) serum IGF-1 concentrations over time in infants in the standard neonatal care and rhIGF-1/rhIGFBP-3 groups (n = 121). Figure 3. IGF-1 levels by A, ROP severity* and B, BPD severity† by postnatal week. *Mean (±SE) serum IGF-1 levels and ROP severity (<3, ≥3) in the rhIGF-1/rhIGFBP-3 and standard neonatal care groups by postnatal week. †Mean (±SE) serum IGF-1 levels and BPD severity (mild, moderate, or severe) in the rhIGF-1/rhIGFBP-3 and standard neonatal care groups by postnatal week. Note: If an infant had multiple IGF-1 levels in a day, then IGF-1 level was averaged for the day. Figure 4. A, Average weight, B, length, and C, head circumference by treatment group (FAS). Table I. Verbatim terms reported in relation to common treatment-emergent AEs

vere apnoea,” “apnea episode,” “apnea neonatal,” “crisis of apnea” Neonatal hypotension “Hypotension,” “low blood pressure,” “hypotension (low blood pressure),” “systemic hypotension,” “hypotensive (mean 21),” “intermittent hypotension,” “hypotension (intermittent),” “hypotension, MAP 26” Hyperglycemia “Hyperglycemia,” “hyperglycaemia,” “hyperglucemia,” “hyperglicemia,” “hyperglycemia (intermittent),” “hyperglycemia, blood glucose 26 mmol, insulin infusion commenced,” “hyperglycemia, blood glucose 58 mmol statdose of insulin given,” “hyperglycemia, blood glucose 45 mmol, insulin infusion concentration and dose increased” Neonatal hyponatremia “Hyponatremia,” “hyponatremi,” “low natrium,” “hyponatriemia,” “hyponatraemia,” “mild hyponatremia,” “severe hyponatremia” Sepsis neonatal “Suspected sepsis, blood cultures negative,” “suspected septicemia,” “suspected sepsis,” “sepsis,” “sepsis (blood culture negative),” “sepsis due to Enterococcus faecalis, Staphilococcus aureus, and Staphilococcus haemoliticus,” “late onset neonatal sepsis,” “late onset sepsis,” “suspected sepsis not confirmed (CRP negative),” “suspected sepsis (blood culture and tracheal aspiration were negative, the adverse event was not confirmed,” “suspected sepsis (clinical instability, elevation of CRP but blood and CSF cultures were negative),” “clinical sepsis (desaturations, elevation of CRP but blood culture was negative),” “suspected sepsis (recurrent apneas but blood culture and CRP were negative),” “suspected sepsis (blood culture was negative) for clinical instability (desaturations),” “sepsis (etiology unknown, blood culture was negative),” “clinical sepsis (PCR, blood culture negative),” “possibility to sepsis, no bacteria found in blood culture,” “presumed sepsis (blood cultures negative after 6 days),” “sepsis presumed, no growth on blood culture or nasopharyngeal aspirate,” “presumed sepsis, blood cultures negative,” “sepsis of unknown etiology,” “suspected sepsis, not confirmed,” “suspected sepsis, not confirmed, etiology unknown,” “neonatal septicemia, staphylococcus capitis grown from blood culture,” “possible sepsis, cultures negative,” “suspected sepsis, rising infection parameters, no bacteria detected in cultures,” “early onset sepsis,” “presumed long line sepsis, cultures negative,” suspected sepsis due to severe desaturations,” “suspected sepsis (CSF and blood cultures were negative),” “clinical sepsis, etiology unknown (cultures remained negative),” “sepsis presumed in line with routine ne

s, no bacteria detected in cultures,” “early onset sepsis,” “presumed long line sepsis, cultures negative,” suspected sepsis due to severe desaturations,” “suspected sepsis (CSF and blood cultures were negative),” “clinical sepsis, etiology unknown (cultures remained negative),” “sepsis presumed in line with routine ne onatal care of extreme premature baby, antibiotics stopped when sepsis ruled out,” “sepsis suspected and therefore treated with antibiotics but blood cultures negative after 6 days,” “sepsis suspected, blood cultures showed no growth after 6 days incubation,” “sepsis (no growth from blood cultures after 6 days incubation, no other source of infection suspected),” “sepsis suspected, no sepsis confirmed,” “suspected sepsis (in line with routine preterm neonatal care) (blood cultures negative after 6 days) antibiotics stopped,” “presumed sepsis, no confirmed on blood cultures,” “central line associated blood stream infection from PICC line,” “clinical sepsis (bloody stool), and blood cultures negative” Hypoglycemia neonatal “Hypoglycemia,” “asymptomatic hypoglycaemia,” “hypoglicemia,” “hypoglycaemia,” “hypoglycaemia as peripheral venous line delivering TPN was leaking,” “hypoglycaemia (intermittent),” “hypoglycaemia, asymptomatic,” “hypoglycaemia, 1.8,” “hypoglycaemia, 1.6,” “hypoglycaemia (1.0 mmol/l at 6 pm)” Metabolic acidosis “Metabolic acidosis,” “metabolic acidosis on blood gas” Staphylococcal sepsis “Septicemia: stafyloccus epidermidis,” “staphylococcus epidermis in the blood,” “coagulase negative staphylococci septicaemia,” “sepsis (staphylococcus epidermidis),” “sepsis due to Bacillus Amyloliquefaciens and Staphilococcus Epidermidis,” “sepsis due to staphylococcus epidermidis,” “sepsis by staphylococcus aureus,” “sepsis by oxacillin-resistant staphylococcus epidermis,” “sepsis by Staphilococcus Aureus,” “sepsis by Staphilococcus epidermidis probably originated by a skin lesion,” “sepsis by staphylococcus warneri,” “sepsis by MRSA,” “sepsis from staphylococcus capitis,” “sepsis from staphylococcus epidermidis,” “coagulase-negative staphylococcal sepsis,” “positive blood culture-staphylococcus epidermidis,” “neonatal septicaemia, staphylococcus capitis grown from blood culture,” “positive blood culture, staphylococcus epidermidis,” “mild sepsis (Staphylococcus epidermidis),” “septicemia with coagulase negative staphylococci,” “sepsis staph aureus and KNS,” “staphylococcus sepsis,” “sepsis due to staphylococcus epidermidis and staphylococcus warneri,” “sepsis

capitis grown from blood culture,” “positive blood culture, staphylococcus epidermidis,” “mild sepsis (Staphylococcus epidermidis),” “septicemia with coagulase negative staphylococci,” “sepsis staph aureus and KNS,” “staphylococcus sepsis,” “sepsis due to staphylococcus epidermidis and staphylococcus warneri,” “sepsis from staphylococcus capitis and epidermidis,” “staphylococcus aureus bacteraemia sepsis,” “sepsis by methicillin-resistant staphylococcus aureaus (MRSA),” “clinical sepsis by staphylococcus capitis oxacilline resistant,” “sepsis staphylococcus haemolitycus,” “sepsis from Staphylococcus Haemolitycus,” “sepsis, blood cultures Staphyloccus capitis positive,” “blood cultures showed mixed coagulase negative staphylococci,” “sepsis, confirmed coag neg staph from blood cultures,” “sepsis suspected, confirmed Staphylococcis capitis,” “presumed sepsis, coagulase negative staph isolated from aerobic bottle after 1 day incubation,” “sepsis, Staphylococcus pettenkoferi,” “sepsis confirmed, Staphylococcus aureus and Staphylococcus epidermitis,” “positive blood culture staphylococci epidermidis,” “positive blood culture (gram positive cocci) Staphylococcus Epidermidis,” “sepsis (positive blood culture Staphylococcus epidermidis noted),” “coagulase negative staphylococci septicaemia,” “sepsis from staphylococcus epidermidis catheter associated” Neonatal hypoxia “Increasing need of oxygen,” “oxygen saturation decreased,” “neonatal hypoxia,” “repeated desaturations,” “low saturation,” “poor saturations and blood gases,” “desaturations,” “episodes of desaturations,” “desaturation crisis,” “frequent desaturations,” “persistent need of oxygen in nCPAP with clinical instability (frequent desaturations),” “persistent need of oxygen in CPAP,” “persistent need of oxygen,” “prolonged need of oxygen,” “hypoxia,” “desaturation, need of increased oxygen,” “increased need of oxygen,” “oxygen saturation dips,” “necessity of nCPAP at 28 days of life without oxygen,” “oxygenation problems due to PIE,” “respiratory step up (increase in the oxygen requirement),” “profound desaturation after a feed,” “desaturation following a feed,” “poor oxygen saturation,” “oxygen desaturations,” “desaturation associated with feeding,” “low oxygen saturation (14%)” Neonatal respiratory failure “Respiratory insufficiency,” “respiratory crisis,” “pulmonary insufficiency,” “respiratory failure, ventilator dependent,” “respiratory failure unable to ventilate,” “reintubation for impending respiratory failure,” “res

“desaturation associated with feeding,” “low oxygen saturation (14%)” Neonatal respiratory failure “Respiratory insufficiency,” “respiratory crisis,” “pulmonary insufficiency,” “respiratory failure, ventilator dependent,” “respiratory failure unable to ventilate,” “reintubation for impending respiratory failure,” “res piratory instability,” “respiratory failure unable to ventilate infant,” “respiratory failure” Hypokalemia “Hypokalemia,” “ipokaliemia,” “hypokalaemia,” “hypokalemi,” “hypopotassemia” Bradycardia neonatal “Bradycardia,” “bradycardia episodes,” “bradycardia (mean heart rate 56.3 bpm),” “bradycardia (heart rate <60 bpm)” Hyperbilirubinemia neonatal “Hyperbilirubinemia,” “hyperbilirubinemi,” “hyperbillirubin,” “hyperbillirubinemi,” “hyperbilirubinaemia,” “conjugated hyperbilirubaemia” Pulmonary hypertension “Persistent pulmonary hypertension,” “pulmonary hypertension,” “Persistent pulmonary hypertension of the newborn,” “severe pulmonary hypertension on study echocardiogram” CPAP, continuous positive airway pressure; CRP, C-reactive protein; CSF, cerebrospinal fluid; MRSA, methicillin-resistant Staphylococcus aureus; nCPAP, nasal continuous positive airway pressure; PICC, peripherally inserted central catheter; PIE, pulmonary interstitial emphysema. Table II. Anticipated distribution of the maximum severity of ROP (standard care and rhIGF-1/rhIGFBP-3 groups) used for calculation of sample size*

piratory instability,” “respiratory failure unable to ventilate infant,” “respiratory failure” Hypokalemia “Hypokalemia,” “ipokaliemia,” “hypokalaemia,” “hypokalemi,” “hypopotassemia” Bradycardia neonatal “Bradycardia,” “bradycardia episodes,” “bradycardia (mean heart rate 56.3 bpm),” “bradycardia (heart rate <60 bpm)” Hyperbilirubinemia neonatal “Hyperbilirubinemia,” “hyperbilirubinemi,” “hyperbillirubin,” “hyperbillirubinemi,” “hyperbilirubinaemia,” “conjugated hyperbilirubaemia” Pulmonary hypertension “Persistent pulmonary hypertension,” “pulmonary hypertension,” “Persistent pulmonary hypertension of the newborn,” “severe pulmonary hypertension on study echocardiogram” CPAP, continuous positive airway pressure; CRP, C-reactive protein; CSF, cerebrospinal fluid; MRSA, methicillin-resistant Staphylococcus aureus; nCPAP, nasal continuous positive airway pressure; PICC, peripherally inserted central catheter; PIE, pulmonary interstitial emphysema. Table II. Anticipated distribution of the maximum severity of ROP (standard care and rhIGF-1/rhIGFBP-3 groups) used for calculation of sample size* ROP 0 1 2 3 >3 Total Standard care (%) 26 15 24 18 17 100 rhlGF-1/rhIGFBP-3 (%) 49 19 19 9 4 100 * Standard care group distribution based on registry data (Austeng. Arch Ophthalmol. 2009;127:1315–9); estimated treatment effect in the rhIGF-1/rhIGFBP-3 group based on the following assumptions: For each outcome of the maximum severity of ROP stage, it is assumed that 25% of the children will not benefit from the treatment, 25% will have their maximum severity of ROP stage reduced by 1 level (eg, from 2 to 1), and 50% will have their maximum severity of ROP stage reduced 2 levels (eg, from 3+ to 2).

sumptions: For each outcome of the maximum severity of ROP stage, it is assumed that 25% of the children will not benefit from the treatment, 25% will have their maximum severity of ROP stage reduced by 1 level (eg, from 2 to 1), and 50% will have their maximum severity of ROP stage reduced 2 levels (eg, from 3+ to 2). Table III. Demographic characteristics and maternal/perinatal histories

sumptions: For each outcome of the maximum severity of ROP stage, it is assumed that 25% of the children will not benefit from the treatment, 25% will have their maximum severity of ROP stage reduced by 1 level (eg, from 2 to 1), and 50% will have their maximum severity of ROP stage reduced 2 levels (eg, from 3+ to 2). Table III. Demographic characteristics and maternal/perinatal histories Standard care rhIGF-1/rhIGFBP-3 FAS rhIGF-1/rhIGFBP-3 ES Characteristics (n = 60) (n = 61) (n = 24) Sex, no. (%) Male 39 (65.0) 39 (63.9) 14 (58.3) Female 21 (35.0) 22 (36.1) 10 (41.7) Gestational age group, no. (%) <26 wk 32 (53.3) 35 (57.4) 10 (41.7) ≥26 wk 28 (46.7) 26 (42.6) 14 (58.3) Gestational age Mean, wk 254/7 254/7 257/7 ±SD, d ±10 ±8 ±9 SGA, no. (%) 10 (16.7) 11 (18.0) NA Weight at birth Mean, kg 0.804 0.780 0.847 SD, kg 0.174 0.183 0.192 Race, no. (%) Asian 5 (8.3) 4 (6.6) 1 (4.2) Black or African American 9 (15.0) 5 (8.2) 3 (12.5) White 42 (70.0) 49 (80.3) 19 (79.2) Other 4 (6.6) 3 (4.9) 1 (4.2) Mode of delivery, no. (%) Vaginal 27 (45.0) 25 (41.0) 10 (41.7) Cesarean 33 (55.0) 36 (59.0) 14 (58.3) Maternal infections, no. (%) 14 (23.3) 11 (18.0) 3 (12.5) Clinical chorioamnionitis, no. (%) 6 (10.0) 10 (16.4) 2 (8.3) Maternal antibiotics, no. (%) 38 (63.3) 32 (52.5) 12 (50.0) Antenatal steroids, no. (%) 60 (100.0) 61 (100.0) 24 (100.0) Fertility therapy, no. (%) 9 (15.0) 10 (16.4) 2 (8.3) IVF 7 (11.7) 10 (16.4) 2 (8.3) Ovulation stimulation 2 (3.3) 0 0 Preterm labor, no. (%) 53 (88.3) 50 (82.0) 19 (79.2) Preterm premature rupture of membranes, no. (%) 20 (33.3) 18 (29.5) 8 (33.3) Preeclampsia, no. (%) 5 (8.3) 7 (11.5) 2 (8.3) IVF, in vitro fertilization; NA, not available; SGA, small for gestational age.

VF 7 (11.7) 10 (16.4) 2 (8.3) Ovulation stimulation 2 (3.3) 0 0 Preterm labor, no. (%) 53 (88.3) 50 (82.0) 19 (79.2) Preterm premature rupture of membranes, no. (%) 20 (33.3) 18 (29.5) 8 (33.3) Preeclampsia, no. (%) 5 (8.3) 7 (11.5) 2 (8.3) IVF, in vitro fertilization; NA, not available; SGA, small for gestational age. Table IV. rhIGF-1/rhIGFBP-3 exposure (safety analysis set) rhIGF-1/rhIGFBP-3 Overall population Gestational age <26 wk Gestational age ≥26 wk Variables (n = 61) (n = 35) (n = 26) Total duration of exposure, d* Mean 23.8 27.4 18.9 Range 0.1–45.3 0.1–45.3 1.8–26.7 Ratio of duration of exposure to expected duration, d† Mean 0.86 0.85 0.88 Range 0.0–1.0 0.0–1.0 0.6–1.0 Overall dose, μg/kg‡ Mean 5907.5 6849.8 4639.0 Range 16.0–11,321.9 16.0–11,321.9 460.9–6664.1 Average daily dose, μg/kg/24 hours§ Mean 248.1 250.0 245.4 Range 131.1–250.0 250.0–250.0 131.1–250.0 Interruptions, no. Mean 4.0 5.7 1.6 Range 0–83 0–83 0–9 Length interruptions, h Mean 7.5 10.2 3.9 Range 0.0–52.6 0.0–52.6 0.0–30.8 * Total duration of exposure defined as (study medication end date – study medication start date) – duration of interruptions. † Ratio of duration of exposure to expected duration defined as total duration of exposure / (29 weeks × 7 + 6 days) or the last day in the study–(birth weeks × 7 + day) + 1. ‡ Total dose (μg/kg) defined as the sum of weight-adjusted doses during the exposure. § Average daily dose (μg/kg/24 hours) defined as overall dose/total duration of exposure across the entire study. Table V. Maximum severity of ROP stage, severity of BPD, and IVH by grades (FAS and ES)

† Ratio of duration of exposure to expected duration defined as total duration of exposure / (29 weeks × 7 + 6 days) or the last day in the study–(birth weeks × 7 + day) + 1. ‡ Total dose (μg/kg) defined as the sum of weight-adjusted doses during the exposure. § Average daily dose (μg/kg/24 hours) defined as overall dose/total duration of exposure across the entire study. Table V. Maximum severity of ROP stage, severity of BPD, and IVH by grades (FAS and ES) Standard care rhIGF-1/rhIGFBP-3 (n = 60) FAS (n = 61) ES (n = 24) ROP Infants with ROP examination, no. 50 47 22 Infants with maximum severity of ROP of stage, no. (%) 0 24 (48.0) 14 (29.8) 8 (36.4) 1 4 (8.0) 4 (8.5) 2 (9.1) 2 13 (26.0) 17 (36.2) 8 (36.4) 3 3 (6.0) 6 (12.8) 2 (9.1) 3+ 6 (12.0) 6 (12.8) 2 (9.1) 4 0 0 0 5 0 0 0 ≥3 9 (18.0) 12 (25.5) 4 (18.2) Missing, no.* 10 14 2 P value† .06 .24 BPD Infants with BPD assessment, no. 49 47 21 Severity of BPD, no. (%) No BPD 4 (8.2) 4 (8.5) 2 (9.5) Mild 16 (32.7) 23 (48.9) 13 (61.9) Moderate 5 (10.2) 9 (19.1) 5 (23.8) Severe 22 (44.9) 10 (21.3) 1 (4.8) Unable to determine 2 (4.1) 1 (2.1) 0 P value† .04‡ .02‡ IVH IVH grade, no. (%) 0–1 42 (70.0) 49 (80.3) 20 (83.3) 2 4 (6.7) 4 (6.6) 2 (8.3) 3 9 (15.0) 6 (9.8) 2 (8.3) 4 5 (8.3) 2 (3.3) 0 P value† .14 .18 * The majority (11/12 treated [FAS] and 5/7 control infants) of infants who died in this study had died before the first scheduled ROP assessment at week 31. Other reasons for not being evaluated included withdrawal of consent and difficulties in capturing quality RetCam images. † Cochran–Mantel–Haenszel row mean score test.

Standard care rhIGF-1/rhIGFBP-3 (n = 60) FAS (n = 61) ES (n = 24) ROP Infants with ROP examination, no. 50 47 22 Infants with maximum severity of ROP of stage, no. (%) 0 24 (48.0) 14 (29.8) 8 (36.4) 1 4 (8.0) 4 (8.5) 2 (9.1) 2 13 (26.0) 17 (36.2) 8 (36.4) 3 3 (6.0) 6 (12.8) 2 (9.1) 3+ 6 (12.0) 6 (12.8) 2 (9.1) 4 0 0 0 5 0 0 0 ≥3 9 (18.0) 12 (25.5) 4 (18.2) Missing, no.* 10 14 2 P value† .06 .24 BPD Infants with BPD assessment, no. 49 47 21 Severity of BPD, no. (%) No BPD 4 (8.2) 4 (8.5) 2 (9.5) Mild 16 (32.7) 23 (48.9) 13 (61.9) Moderate 5 (10.2) 9 (19.1) 5 (23.8) Severe 22 (44.9) 10 (21.3) 1 (4.8) Unable to determine 2 (4.1) 1 (2.1) 0 P value† .04‡ .02‡ IVH IVH grade, no. (%) 0–1 42 (70.0) 49 (80.3) 20 (83.3) 2 4 (6.7) 4 (6.6) 2 (8.3) 3 9 (15.0) 6 (9.8) 2 (8.3) 4 5 (8.3) 2 (3.3) 0 P value† .14 .18 * The majority (11/12 treated [FAS] and 5/7 control infants) of infants who died in this study had died before the first scheduled ROP assessment at week 31. Other reasons for not being evaluated included withdrawal of consent and difficulties in capturing quality RetCam images. † Cochran–Mantel–Haenszel row mean score test. ‡ Difference in the distribution of BPD severity seen between the rhIGF-1/rhIGFBP-3 (FAS or ES) and standard care groups is statistically significant. Table VI. Maximum severity of ROP stage across all examinations by a central pediatric ophthalmologist by gestational age strata (FAS and ES)

† Cochran–Mantel–Haenszel row mean score test. ‡ Difference in the distribution of BPD severity seen between the rhIGF-1/rhIGFBP-3 (FAS or ES) and standard care groups is statistically significant. Table VI. Maximum severity of ROP stage across all examinations by a central pediatric ophthalmologist by gestational age strata (FAS and ES) Standard care rhIGF-1/rhIGFBP-3 FAS ES Gestational age groups (n = 32) (n = 35) (n = 10) <26 wk Infants with ROP examination, no. 26 25 9 Infants with maximum severity of ROP of stage, no. (%) 0 8 (30.8) 5 (20.0) 2 (22.2) 1 2 (7.7) 2 (8.0) 0 2 9 (34.6) 7 (28.0) 4 (44.4) 3 2 (7.7) 5 (20.0) 1 (11.1) 3+ 5 (19.2) 6 (24.0) 2 (22.2) 4 0 0 0 5 0 0 0 ≥3 7 (26.9) 11 (44.0) 3 (33.3) Missing 6 10 1 ≥26 wk n = 28 n = 26 n = 14 Infants with ROP examination, no. 24 22 13 Infants with maximum severity of ROP of stage, no. (%) 0 16 (66.7) 9 (40.9) 6 (46.2) 1 2 (8.3) 2 (9.1) 2 (15.4) 2 4 (16.7) 10 (45.5) 4 (30.8) 3 1 (4.2) 1 (4.5) 1 (7.7) 3+ 1 (4.2) 0 0 4 0 0 0 5 0 0 0 ≥3 2 (8.3) 1 (4.5) 1 (7.7) Missing 4 4 1 Table VII. Severity of BPD by gestational age strata (FAS and ES)

, no. 24 22 13 Infants with maximum severity of ROP of stage, no. (%) 0 16 (66.7) 9 (40.9) 6 (46.2) 1 2 (8.3) 2 (9.1) 2 (15.4) 2 4 (16.7) 10 (45.5) 4 (30.8) 3 1 (4.2) 1 (4.5) 1 (7.7) 3+ 1 (4.2) 0 0 4 0 0 0 5 0 0 0 ≥3 2 (8.3) 1 (4.5) 1 (7.7) Missing 4 4 1 Table VII. Severity of BPD by gestational age strata (FAS and ES) Standard care rhIGF-1/ rhIGFBP-3 Gestational age groups (n = 32) FAS (n = 35) ES (n = 10) <26 wk Infants with BPD assessment, no. 25 25 8 Severity of BPD, no. (%) No BPD 1 (4.0) 1 (4.0) 0 Mild 5 (20.0) 10 (40.0) 4 (50.0) Moderate 4 (16.0) 7 (28.0) 3 (37.5) Severe 14 (56.0) 6 (24.0) 1 (12.5) Unable to determine 1 (4.0) 1 (4.0) 0 ≥26 wk n 28 26 14 Infants with BPD assessment, no. 24 22 13 Severity of BPD, no. (%) No BPD 3 (12.5) 3 (13.6) 2 (15.4) Mild 11 (45.8) 13 (59.1) 9 (69.2) Moderate 1 (4.2) 2 (9.1) 2 (15.4) Severe 8 (33.3) 4 (18.2) 0 Unable to determine 1 (4.2) 0 0 Table VIII. Percentage of infants with IVH by grade and gestational age strata (FAS and ES) Standard care rhIGF-1/ rhIGFBP-3 Gestational age groups (n = 32) FAS (n = 35) ES (n = 10) <26 wk IVH grade, no. (%) 0–1 20 (62.5) 27 (77.1) 9 (90.0) 2 3 (9.4) 3 (8.6) 1 (10.0) 3 6 (18.8) 3 (8.6) 0 4 3 (9.4) 2 (5.7) 0 ≥26 wk N 28 26 14 IVH grade, no. (%) 0–1 22 (78.6) 22 (84.6) 11 (78.6) 2 1 (3.6) 1 (3.8) 1 (7.1) 3 3 (10.7) 3 (11.5) 2 (14.3) 4 2 (7.1) 0 0 Table IX. Post hoc analysis of BPD including all-cause death in the severe BPD category

62.5) 27 (77.1) 9 (90.0) 2 3 (9.4) 3 (8.6) 1 (10.0) 3 6 (18.8) 3 (8.6) 0 4 3 (9.4) 2 (5.7) 0 ≥26 wk N 28 26 14 IVH grade, no. (%) 0–1 22 (78.6) 22 (84.6) 11 (78.6) 2 1 (3.6) 1 (3.8) 1 (7.1) 3 3 (10.7) 3 (11.5) 2 (14.3) 4 2 (7.1) 0 0 Table IX. Post hoc analysis of BPD including all-cause death in the severe BPD category Standard care rhIGF-1/rhIGFBP-3 FAS (n = 60) (n = 61) Infants with BPD assessment or death, no.* 56† 59† Severity of BPD No. (%) 95% CI No. (%) 95% CI No BPD/mild 20 (35.7) (24.5–48.8) 27 (45.8) (33.7–58.3) Moderate 5 (8.9) (3.9–19.3) 9 (15.3) (8.2–26.5) Severe/death 29 (51.8) (39.0–64.3) 22 (37.3) (26.1–50.1) Unable to determine 2 (3.6) (1.0–12.1) 1 (1.7) (0.3–9.0) * There were 19 all-cause deaths that occurred during the study, 12 deaths in the rhIGF-1/rhIGFBP-3 group and 7 deaths in the standard care group. † Four infants in the standard care group and 2 infants in the rhIGF-1/rhIGFBP-3 FAS were withdrawn from the study before being assessed for BPD. Table X. Most common treatment-emergent AEs (preferred terms, occurring in ≥20% in any treatment group)

Standard care rhIGF-1/rhIGFBP-3 FAS (n = 60) (n = 61) Infants with BPD assessment or death, no.* 56† 59† Severity of BPD No. (%) 95% CI No. (%) 95% CI No BPD/mild 20 (35.7) (24.5–48.8) 27 (45.8) (33.7–58.3) Moderate 5 (8.9) (3.9–19.3) 9 (15.3) (8.2–26.5) Severe/death 29 (51.8) (39.0–64.3) 22 (37.3) (26.1–50.1) Unable to determine 2 (3.6) (1.0–12.1) 1 (1.7) (0.3–9.0) * There were 19 all-cause deaths that occurred during the study, 12 deaths in the rhIGF-1/rhIGFBP-3 group and 7 deaths in the standard care group. † Four infants in the standard care group and 2 infants in the rhIGF-1/rhIGFBP-3 FAS were withdrawn from the study before being assessed for BPD. Table X. Most common treatment-emergent AEs (preferred terms, occurring in ≥20% in any treatment group) Standard care rhIGF-1/rhIGFBP-3 (n = 60) (n = 61) Infants, Events, Infants, Events, AEs no. (%) no. no. (%) no. Patent ductus arteriosus 51 (85.0) 71 55 (90.2) 80 Anemia neonatal 44 (73.3) 157 46 (75.4) 210 Neonatal respiratory distress syndrome 34 (56.7) 49 29 (47.5) 34 Jaundice neonatal 30 (50.0) 38 28 (45.9) 34 Infantile apneic attack 17 (28.3) 35 27 (44.3) 48 Neonatal hypotension 18 (30.0) 30 25 (41.0) 35 Hyperglycemia 29 (48.3) 58 24 (39.3) 41 Neonatal hyponatremia 22 (36.7) 39 23 (37.7) 43 Sepsis neonatal* 15 (25.0) 30 23 (37.7) 41 Hypoglycemia neonatal 19 (31.7) 28 18 (29.5) 22 Metabolic acidosis 22 (36.7) 46 17 (27.9) 48 Staphylococcal sepsis 19 (31.7) 25 16 (26.2) 24 Neonatal hypoxia† 13 (21.7) 22 14 (23.0) 17 Neonatal respiratory failure‡ 14 (23.3) 18 14 (23.0) 22 Hypokalemia 11 (18.3) 18 14 (23.0) 24 Bradycardia neonatal§ 5 (8.3) 6 13 (21.3) 19 Hyperbilirubinemia neonatal 14 (23.3) 18 12 (19.7) 14 Pulmonary hypertension 12 (20.0) 14 8 (13.1) 8 Events of ROP, BPD, and IVH also were reported as treatment-emergent AEs in some but not all cases; however, they are not included in the table because these are efficacy outcomes.

ardia neonatal§ 5 (8.3) 6 13 (21.3) 19 Hyperbilirubinemia neonatal 14 (23.3) 18 12 (19.7) 14 Pulmonary hypertension 12 (20.0) 14 8 (13.1) 8 Events of ROP, BPD, and IVH also were reported as treatment-emergent AEs in some but not all cases; however, they are not included in the table because these are efficacy outcomes. * Verbatim terms include suspected/presumed sepsis and some microbiologically confirmed cases of sepsis, but do not encompass all cases of sepsis (nonspecific term). See Table I for a list of reported verbatim terms under the preferred term “sepsis neonatal.” † See Table I for verbatim terms reported under the preferred term “neonatal hypoxia.” ‡ See Table I for verbatim terms reported under the preferred term “neonatal respiratory failure.” § Frequency in rhIGF-1/rhIGFBP-3 group at least twice that of the standard care group. Table XI. Fatal treatment-emergent SAEs (preferred terms) Standard care rhIGF-1/rhIGFBP-3 Gestational age <26 wk Gestational age >26 wk Gestational age <26 wk Gestational age ≥26 wk Total deaths, no. 4 3 8 4 Days since birth <7 2 0 3 2 IVH Neonatal respiratory failure Pulmonary hypertension Neonatal respiratory failure Pulmonary hemorrhage Escherichia sepsis* IVH >7 and ≤14 0 0 0 0 >14 and ≤28 2 1 4 1 Renal failure neonatal† Staphylococcal sepsis‡ Sepsis neonatal§ Renal failure neonatal† Citrobacter sepsis Renal failure neonatal† NEC neonatal NEC neonatal >28 0 2 1 1 Neonatal respiratory failure Neonatal respiratory failure¶ NEC neonatal Intestinal obstruction * Early-onset sepsis owing to Escherichia coli.

2 1 4 1 Renal failure neonatal† Staphylococcal sepsis‡ Sepsis neonatal§ Renal failure neonatal† Citrobacter sepsis Renal failure neonatal† NEC neonatal NEC neonatal >28 0 2 1 1 Neonatal respiratory failure Neonatal respiratory failure¶ NEC neonatal Intestinal obstruction * Early-onset sepsis owing to Escherichia coli. † Verbatim Terms under the Preferred Term “renal failure neonatal” were “renal insufficiency” and “acute renal failure.” ‡ Catheter-related sepsis owing to Staphylococcus epidermidis. § Sepsis owing to Enterococcus faecalis, Staphylococcus aureus, and Staphylococcus haemolyticus. ¶ Verbatim terms under the preferred term “neonatal respiratory failure” were “respiratory insufficiency” and “respiratory failure.”