Association Between Early Administration of High-Dose Erythropoietin in Preterm Infants and Brain MRI Abnormality at Term-Equivalent Age

Importance  Premature infants are at risk of developing encephalopathy of prematurity, which is associated with long-term neurodevelopmental delay. Erythropoietin was shown to be neuroprotective in experimental and retrospective clinical studies.

Objective  To determine if there is an association between early high-dose recombinant human erythropoietin treatment in preterm infants and biomarkers of encephalopathy of prematurity on magnetic resonance imaging (MRI) at term-equivalent age.

Design, Setting, and Participants  A total of 495 infants were included in a randomized, double-blind, placebo-controlled study conducted in Switzerland between 2005 and 2012. In a nonrandomized subset of 165 infants (n=77 erythropoietin; n=88 placebo), brain abnormalities were evaluated on MRI acquired at term-equivalent age.

Interventions  Participants were randomly assigned to receive recombinant human erythropoietin (3000 IU/kg; n=256) or placebo (n=239) intravenously before 3 hours, at 12 to 18 hours, and at 36 to 42 hours after birth.

Main Outcomes and Measures  The primary outcome of the trial, neurodevelopment at 24 months, has not yet been assessed. The secondary outcome, white matter disease of the preterm infant, was semiquantitatively assessed from MRI at term-equivalent age based on an established scoring method. The resulting white matter injury and gray matter injury scores were categorized as normal or abnormal according to thresholds established in the literature by correlation with neurodevelopmental outcome.

Results  At term-equivalent age, compared with untreated controls, fewer infants treated with recombinant human erythropoietin had abnormal scores for white matter injury (22% [17/77] vs 36% [32/88]; adjusted risk ratio [RR], 0.58; 95% CI, 0.35-0.96), white matter signal intensity (3% [2/77] vs 11% [10/88]; adjusted RR, 0.20; 95% CI, 0.05-0.90), periventricular white matter loss (18% [14/77] vs 33% [29/88]; adjusted RR, 0.53; 95% CI, 0.30-0.92), and gray matter injury (7% [5/77] vs 19% [17/88]; adjusted RR, 0.34; 95% CI, 0.13-0.89).

Conclusions and Relevance  In an analysis of secondary outcomes of a randomized clinical trial of preterm infants, high-dose erythropoietin treatment within 42 hours after birth was associated with a reduced risk of brain injury on MRI. These findings require assessment in a randomized trial designed primarily to assess this outcome as well as investigation of the association with neurodevelopmental outcomes.

Trial Registration  clinicaltrials.gov Identifier: NCT00413946

Survival of premature infants has improved over the past decades, but at the expense of an increase in the number of infants affected by long-term developmental disabilities.¹ Recent studies confirm the major neuropathological substrate to be encephalopathy of prematurity, characterized by white matter lesions, white matter loss, and abnormalities in cortical development.^2,3 Magnetic resonance imaging (MRI) allows the characterization of specific features of encephalopathy of prematurity, including structural changes of brain white matter and gray matter.^3,4

Pharmacological interventions to improve neurological outcome in preterm infants are limited.⁵ Nevertheless, experimental data have indicated potential targets for neuroprotection.⁶ Erythropoietin and its receptor are expressed in the central nervous system, where erythropoietin has been shown to exert neuroprotection in animal models.^7,8

Because erythropoietin has been routinely used in neonatology to treat anemia of prematurity, a few observational or retrospective studies have reported beneficial effects of the hematopoietic treatment on long-term neurodevelopmental outcome.^9-11 Moreover, a recent study reported improved cognitive scores at 18 to 22 months in preterm infants treated with low doses of erythropoietin or darbepoetin in the period up to 35 gestational weeks.¹² In terms of safety, a prospective, open-label, dose escalation trial demonstrated that high-dose recombinant erythropoietin given at 24-hour intervals in the first 3 days of life was well tolerated by premature infants¹³ and no adverse events were found.¹⁴

The current trial investigates the effect of high-dose recombinant human erythropoietin specifically for neuroprotection in premature infants. This report presents an intermediate outcome of the trial, white matter disease,¹⁵ evaluated as white matter injury (WMI) and gray matter injury (GMI) on MRI at term-equivalent age.

This study reports on a subset of patients enrolled in a randomized, double-blind, placebo-controlled trial testing high-dose recombinant human erythropoietin for the neuroprotection of premature infants (Figure 1). The study was conducted in Switzerland by the Swiss EPO Neuroprotection Trial Group between 2005 and 2012, in 3 university hospitals (Basel, Geneva, and Zurich) and 2 district hospitals (Aarau and Chur) and included all eligible infants born from 26 weeks to 31 weeks and 6 days of gestation. Exclusion criteria were presence of a genetically defined syndrome, severe congenital malformation adversely affecting life expectancy, or abnormality known to affect neurodevelopment. The primary outcome of the trial, neurodevelopmental outcome at 2 years of age, has not yet been reported. This study reports an intermediate outcome, white matter disease, evaluated as WMI and GMI on MRI at term-equivalent age. The study was conducted at 2 of the 5 hospitals; 3 hospitals did not have MRI scanners. The Swiss drug surveillance unit and the local ethical committees approved the protocol (Supplement 1), and written informed consent was obtained from parents or guardians.

Randomization lists for block randomization with variable block length were created by an independent statistician for each participating center and provided only to the pharmacy involved in production of the medication. Study medication was randomly assigned to each patient number in a 1:1 allocation, using a computer-based random-number generator. Erythropoietin or an equivalent volume of normal saline placebo was administered intravenously before 3 hours of age, at 12 to 18 hours, and at 36 to 42 hours after birth. A single dose consisted of 25 μg (3000 IU) of recombinant human erythropoietin per kilogram of body weight dissolved in 1 mL of sterile water. The treatment protocol was established based on a previously published safety trial.¹⁴

Magnetic resonance imaging was performed at term-equivalent age without sedation. Noise was attenuated with earplugs (attenuation: 24 dB; Earsoft; Aearo) and Minimuffs (attenuation: 7 dB; Natus). T1-weighted and T2-weighted high-resolution images were acquired on two 3T magnetic resonance systems: an HDxt MRI scanner (GE Medical Systems) in Zurich and a TrioTim system (Siemens Medical Solutions) in Geneva. Acquisition parameters were set to obtain compatible images in terms of quality and contrast (eAppendix in Supplement 2).

Brain MRI at term-equivalent age was used to evaluate the presence and degree of white matter disease, including GMI and WMI, and punctate white matter lesions. White matter and gray matter abnormalities in brain MRI were scored based on the system of Woodward et al.¹⁶ The WMI score was obtained by adding the subscores of white matter signal abnormality (the so-called diffuse excessive high signal intensity), periventricular white matter volume loss, presence of cystic abnormalities, ventricular dilatation, and thinning of corpus callosum (Figure 2). The GMI score was obtained by adding the subscores of cortical abnormalities, quality of gyral maturation, and size of subarachnoid space. In the work of Woodward et al,¹⁶ each of these subscores was evaluated according to a 3-point scale, with 1 = normal, 2 = mild abnormality, and 3 = moderate to severe abnormality. Recent studies demonstrated that mild diffuse excessive high signal intensity on MRI at term-equivalent age is a developmental phenomenon not related to negative neurodevelopmental outcome^17,18; thus, it was considered normal and scored as 1 in this analysis (instead of 2 in the original score). The original scoring of Woodward et al¹⁶ was used for all the other subscores. Thus, the range of possible scores for the WMI score was from 5 to 15 and for the GMI score was from 3 to 9.

No minimal clinically important differences were defined in the work of Woodward et al.^16,19 In the literature, these scores are usually interpreted using a threshold established by correlation with neurodevelopmental outcome at 2,¹⁶ 4, and 6 years of age.¹⁹ According to Woodward et al, we categorized the WMI score as normal (≤6) or abnormal (>6). In the work of Woodward et al, the GMI score was considered normal at 5 or lower. However, neurodevelopmental outcome scores were quite low in the infants who were considered normal.¹⁶ Therefore, we considered the GMI score to be normal at 4 or lower and abnormal at higher than 4. All WMI and GMI subscores were considered abnormal when greater than 1.

The presence of punctate white matter lesions was also assessed, as these lesions are known to be associated with alterations in later neurodevelopment.^17,20 Images were categorized as having no punctate lesions (score of 0), 6 or fewer (score of 1), or more than 6 punctate lesions (score of 2), as described by De Bruïne et al.¹⁷

A single investigator (R.H.L.) scored all scans, and 20 scans were rerated by a second experienced investigator (C.H.). Agreement between investigators was good for WMI (intraclass correlation coefficient, 0.75; 95% CI, 0.44-0.90) and moderate for GMI (κ = 0.51; 95% CI, 0.23-0.79). The main investigator rerated 20 scans within an interval of several months and intrarater agreement was excellent for WMI (intraclass correlation coefficient, 0.95; 95% CI, 0.86-0.98) and good for GMI (κ = 0.75; 95% CI, 0.49-1). Raters were blind to infants’ perinatal history and treatment group.

The main objective of the present study was to compare WMI, GMI, and punctate white matter lesion scores between the randomized recombinant human erythropoietin and placebo groups. Because sample sizes were moderate, we first verified if there were substantial differences between the groups in terms of clinical characteristics and neonatal morbidities.

Semiquantitative WMI and GMI scores were first compared between groups using a linear regression model. Then, to facilitate clinical interpretation, these 2 scores, their composing subscores, and the punctate white matter lesion score were categorized as normal/abnormal according to the clinically relevant cutoffs presented above and were compared between groups using the χ² test or the Fisher exact test as appropriate.

Because birth weight is a well-known confounder and was different between the 2 groups, we used multivariable regression models (linear regression for continuous scores; log-binomial regression when categorized as normal/abnormal) to adjust for the treatment effect.

A post hoc sensitivity analysis was performed by excluding infants whose treatment adherence with the protocol could not be verified or who received a small dose of recombinant human erythropoietin at 1 month of life or later to treat anemia.

Because we did not test the general null hypothesis that all null hypotheses are true simultaneously, we did not correct for multiple comparisons.

We used R software version 2.15.2 (http://www.R-project.org) for all analyses. Statistical significance was assessed at the 2-sided P<.05 level for all analyses.

Of the 495 infants randomized in the trial (n=256 to recombinant human erythropoietin; n=239 to placebo), 34 were excluded because of subsequent discovery of exclusion criteria (such as genetic syndromes) or nonadherence to inclusion criteria (due to errors in the reporting of gestational age). Because only 2 of 5 hospitals had MRI scanners available, and some parents refused the MRI, of the 461 infants remaining in the trial, 175 had an MRI, of which the scans of 10 infants were of insufficient quality for the analyses (Figure 1). Therefore, the analysis was performed on 165 infants with good-quality MRI scans (n=77 in the recombinant human erythropoietin group and n=88 in the placebo group).

At the time of the analysis, only deaths at term-equivalent age were registered for the 461 eligible infants in the trial; there was no statistically significant difference in deaths between the recombinant human erythropoietin group (12/236 [5.1%]) and the placebo group (12/225 [5.3%]; P>.99). Moreover, owing to the trial design, no deaths at term-equivalent age were observed for the 175 infants who had an MRI.

No statistically significant differences in birth weight, gestational age, or clinical characteristics were found between the 175 infants with MRI at term-equivalent age and the 286 infants included in the trial who did not have an MRI (eTable 1 in Supplement 2). There were differences between the 165 infants with MRIs included and the 10 infants excluded for low image quality, with those excluded having a lower gestational age and higher incidence of bronchopulmonary dysplasia and retinopathy of prematurity (eTable 2 in Supplement 2).

Table 1 summarizes the clinical data of the 165 infants analyzed in this study. Birth weight was slightly higher in infants who received recombinant human erythropoietin than in controls. Continuous positive airway pressure, postnatal steroids, and required number of days of mechanical ventilation were similar between groups. Other perinatal morbidities such as sepsis and retinopathy of prematurity were similar between groups. Magnetic resonance imaging was performed at a median gestational age of 40 weeks and 5 days (interquartile range, 39 weeks and 5 days to 42 weeks and 3 days; range, 35 weeks and 5 days to 44 weeks and 6 days), with no significant differences between treatment groups. Two enrolled infants born at 25 weeks and 6 days of gestation and 2 born at 32 weeks of gestation (1 day younger and 1 day older than the inclusion age range, respectively) were included in this analysis because the clinical maturity of these infants was judged to correspond to the inclusion age range (26 weeks to 31 weeks and 6 days of gestation) and to increase statistical power, given that MRI data are difficult to obtain in this population.

The cumulative distribution of total WMI scores within the recombinant human erythropoietin and placebo groups is presented in Figure 3. Quantitative investigation via linear regression of the WMI score depending on treatment group showed a statistically significant association with improved total WMI scores for recombinant human erythropoietin–treated infants (mean score, 5.91 [SD, 1.33] for recombinant human erythropoietin and 6.43 [SD, 1.76] for placebo; mean difference, −0.52; 95% CI, −1.01 to −0.04; P = .04), and the association was maintained when adjusting for birth weight (mean difference, −0.55; 95% CI, −1.04 to −0.06; P = .03). When categorizing the total WMI score as normal or abnormal, the difference between the recombinant human erythropoietin group (22%) and placebo group (36%) was not statistically significant (difference, 14.3%; 95% CI, 0.4%-27.6%; P = .07) but became statistically significant when adjusted for birth weight in log-binomial regression analysis (adjusted risk ratio [RR], 0.58; 95% CI, 0.35-0.96; P = .03) (Table 2).

Frequencies of abnormality for all WMI subscores are presented in Table 2. Concerning the presence of white matter signal abnormalities, the difference between the placebo group and the recombinant human erythropoietin group was not statistically significant (11% vs 3%; P = .06). However, the difference became significant when adjusting for birth weight in log-binomial regression (adjusted RR, 0.20; 95% CI, 0.05-0.90; P = .04). Significantly more infants in the placebo group (33%) showed periventricular white matter loss compared with the recombinant human erythropoietin group (18%; P = .048), and the difference remained statistically significant when adjusting for birth weight in log-binomial regression (adjusted RR, 0.53; 95% CI, 0.30-0.92; P = .02).

A similar number of infants in both groups had punctate white matter lesions (16% in the placebo group vs 14% in the recombinant human erythropoietin group; P = .97). No infants in either group had more than 6 punctate lesions.

Figure 3 presents the cumulative distribution of total GMI scores within the recombinant human erythropoietin and placebo groups. Linear regression analysis of the GMI score showed no statistically significant association with recombinant human erythropoietin treatment (mean score, 3.64 [SD, 0.70] for recombinant human erythropoietin and 3.85 [SD, 0.81] for placebo; mean difference, −0.21; 95% CI, −0.45-0.03; P = .08), and the same was true when adjusted for birth weight (mean difference, −0.19; 95% CI, −0.43-0.05; P = .13). However, recombinant human erythropoietin treatment was associated with significantly fewer abnormal total GMI scores in the recombinant human erythropoietin group (7%) compared with the placebo group (19%) (difference, 12%; 95% CI, 2%-22%; P = .03) (Table 2). The association remained statistically significant when adjusting for birth weight in log-binomial regression analysis (adjusted RR, 0.34; 95% CI, 0.13-0.89; P = .03).

When excluding from multivariable analyses 25 infants whose treatment adherence to the protocol could not be verified and/or who were identified as having received a small dose of recombinant human erythropoietin to treat anemia (eTable 3 in Supplement 2), the associations, assessed by the RR adjusted for birth weight, between recombinant human erythropoietin treatment and abnormal total WMI score (RR, 0.59; 95% CI, 0.35-1.01; P = .05) and abnormal total GMI score (RR, 0.34; 95% CI, 0.13-0.89; P = .03) were similar, even though the WMI score was no longer significant. However, this lack of significance can reasonably be attributed to the decreased statistical power.

In this analysis of an intermediate outcome of a randomized, placebo-controlled trial on the role of high-dose recombinant human erythropoietin for neuroprotection of premature infants, the extent of brain abnormalities was measured on MRI at term-equivalent age. Recombinant human erythropoietin treatment within 42 hours after birth was associated with a reduced risk of WMI and GMI, white matter signal abnormalities, and periventricular white matter volume loss, all imaging biomarkers of the recently described encephalopathy of prematurity.³

Encephalopathy of prematurity is an ensemble of disturbances in the preterm brain resulting both from brain injury and from subsequent failure of normal development. As a neuroprotective agent, erythropoietin potentially addresses both processes. On one hand, it prevents acute injury by mechanisms such as the inhibition of glutamate release, modulation of intracellular calcium metabolism, induction of neuronal antiapoptotic factors, reduction of inflammation, decrease in nitric oxide–mediated injury, and direct antioxidant effects.⁶ On the other hand, erythropoietin was shown to influence developmental mechanisms by promoting the proliferation and differentiation of preoligodendrocytes and by stimulating growth factors.^6,8

The major neuropathologic feature of encephalopathy of prematurity is cerebral WMI, including foci of necrosis, axonal damage, reactive astrocytosis, and microglial activation.^2,21 On in vivo MRI, these neuropathologic features translate to punctate high signal intensities on T1-weighted images, diffuse high signal intensities on T2-weighted images, or cysts, as measured by the clinical scoring system used in the current trial.¹⁶ Erythropoietin treatment in the trial was associated with a reduction of white matter signal abnormality scores, indicating a reduction of the main histopathology of white matter disease in preterm infants.

Periventricular white matter loss is another feature of encephalopathy of prematurity,³ which has been reproduced by early hypoxia-ischemia and infection-inflammation in the immature sheep model, both assessed by ex vivo high-field MRI.^22,23 Infants treated with erythropoietin in the present study showed a significant association with less periventricular white matter loss, reflecting erythropoietin’s role in protection from cell destruction.⁸ Moreover, both endogenous erythropoietin release from astrocytes posthypoxia²⁴ and exogenous erythropoietin protect the oligodendrocyte precursors and the immature oligodendrocytes from oxidative stress, as well as induce proliferation of oligodendrocyte precursors from the subventricular zone.^25-27 Thus, our in vivo data showing an association between erythropoietin administration and the preservation of periventricular white matter volume supports the neuroprotective effect of erythropoietin on the axonal/premyelinating oligodendrocyte unit.

Encephalopathy of prematurity also affects subplate neurons and late-migrating γ-aminobutyric acid–ergic interneurons,²⁸ both important for cortical development and for the guidance of axonal circuits.^29,30 Erythropoietin has been recently shown to exert neuroprotection on subplate neurons, thus positively influencing cortical development.³¹ Loss of cortical neurons, in particular pyramidal neurons of layer V, has also been described in preterm infants.³² These neurons send axons through the corticospinal tracts, and injury to these axons in the white matter through retrograde degeneration can cause neuronal loss and cortical signal abnormalities, assessed as GMI by MRI.³³ Erythropoietin protects peripheral axons from degeneration³⁴ and therefore may preserve axonal integrity in the white matter as well as protect from secondary neuronal loss and enhance gyral maturation. In the current trial, erythropoietin administration was significantly associated with a lower risk of abnormal GMI scores, mainly due to better gyral maturation scores and less subarachnoid space, indicating better brain growth, which is consistent with the experimental findings.

Encephalopathy of prematurity has been recognized to contribute to the long-term neurodevelopmental deficits in prematurely born children, which affect not only their early development but also their school achievements and their social integration.^35-37

Early imaging biomarkers of encephalopathy of prematurity, such as the ones used in this study, are increasingly recognized to be useful for physicians. Cystic periventricular leukomalacia and MRI-defined delay in myelination of the posterior limb of the internal capsule are well known to correlate with later motor outcome and cerebral palsy, whereas noncystic WMI correlates with cognitive outcome.^37,38 In the study by Woodward et al,¹⁶ moderate to severe abnormal WMI scores (10-15) were predictive of cognitive delay (odds ratio [OR], 3.6), motor delay (OR, 10.3), and cerebral palsy (OR, 9.6), indicating the clinical relevance of these scores for later outcome. Seventy-two percent of preterm infants in that study had abnormal WMI scores and neurodevelopmental scores at 2 years below the normal range. In the current study, 22% of the recombinant human erythropoietin group had abnormal WMI scores vs 36% of the placebo group, which represents a 14% risk reduction for erythropoietin-treated infants. Furthermore, in the study by Woodward et al,¹⁶ abnormal GMI scores were associated with severe cognitive delay or cerebral palsy (OR, 2-3). In a more recent study, punctate white matter lesions and ventricular dilatation or white matter loss predicted poor neurodevelopmental outcome at 2 years, whereas mild diffuse high signal intensities alone did not.¹⁷

In later childhood, children with abnormal white matter appearances on MRI at term-equivalent age had a higher incidence of cerebral palsy, lower verbal performance, and full-scale IQs and were more likely to require special assistance at school, underlining the role of MRI at term-equivalent age as a biomarker of future neurodevelopmental outcome.³⁹

The current study has some limitations. First, the report does not include neurodevelopmental outcome data. Second, not all patients enrolled in the trial had access to MRI, since 3 of 5 centers did not have MRI available, and some parents of eligible infants subsequently refused the MRI acquisition. Additionally, 10 of 175 infants were excluded from analyses because of low-quality images. Therefore, and despite an original randomized design, selection bias may exist. These limitations also result in the current study being underpowered to detect treatment effects on rare cerebral lesions, such as cystic periventricular leukomalacia. Third, we did not use a correction for multiple comparisons. However, such methods are relevant when one wants to test the general null hypothesis that all null hypotheses are true simultaneously; such a correction would have increased the likelihood of type II error. Truly important differences would be deemed not significant, which may not be the best approach when studying a vulnerable population. Fourth, the study used a very short dosing regimen (up to 42 hours after birth) in accordance with the experimental data available at the time of study design, therefore not taking into consideration recent experimental data on long-term erythropoietin treatment.⁷ Fifth, even though in the literature moderate to severe abnormal WMI and GMI scores were associated with clinically significant neurodevelopmental impairments, we do not currently know what is the minimum clinically important difference of these scores. However, we estimate that a 12% to 14% risk difference is clinically relevant.

In an analysis of secondary outcomes of a randomized clinical trial of preterm infants in Switzerland, high-dose erythropoietin treatment within 42 hours after birth was associated with a reduced risk of brain WMI and GMI. These findings require assessment in a randomized trial designed primarily to assess this outcome, as well as investigation of the association with neurodevelopmental outcomes.

Corresponding Author: Petra Susan Hüppi, MD, Division of Development and Growth, Department of Pediatrics, University Hospital of Geneva, Rue Willy Donzé 6, 1211 Genève 14, Switzerland (petra.huppi@hcuge.ch).

Author Contributions: Dr Hüppi had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis. Drs Leuchter and Gui contributed equally to this article.

Study concept and design: Hagmann, Martin, Bucher, Hüppi.

Acquisition, analysis, or interpretation of data: Leuchter, Gui, Poncet, Hagmann, Lodygensky, Martin, Koller, Darqué, Hüppi.

Drafting of the manuscript: Leuchter, Darqué.

Critical revision of the manuscript for important intellectual content: Leuchter, Gui, Poncet, Hagmann, Lodygensky, Martin, Koller, Bucher, Hüppi.

Statistical analysis: Leuchter, Gui, Poncet, Hagmann.

Obtained funding: Bucher, Hüppi.

Administrative, technical, or material support: Gui, Lodygensky, Koller, Hüppi.

Study supervision: Hagmann, Martin, Bucher, Hüppi.

Conflict of Interest Disclosures: All authors have completed and submitted the ICMJE Form for Disclosure of Potential Conflicts of Interest. Dr Leuchter reports grants from Nestlé Research and Development. Dr Hüppi reports grants from the European Community Framework 6/7 and Nestlé Research and Development. No other disclosures were reported.

Swiss EPO Neuroprotection Trial Group: The following local investigators and hospitals participated in this study: Aarau: Kinderklinik Kantonsspital Aarau (Georg Zeilinger, MD; Sylviane Pasquier, MD); Basel: Universitätskinderklinik UKBB (Christoph Bührer, MD; René Glanzmann, MD; Sven Schulzke, MD); Chur: Abteilung für Neonatologie, Kantons- und Regionalspital (Brigitte Scharrer, MD; Walter Bär, MD); Geneva: Division of Development and Growth, Department of Pediatrics, Geneva University Hospitals (S. Sizonenko, MD, PhD), Neonatology Unit, Department of Pediatrics, Geneva University Hospitals (Riccardo Pfister, MD, PhD); Geneva Centre d’Imagerie Biomédicale (F. Lazeyras, PhD); Zürich: UniversitätsSpital Zürich, Klinik für Neonatologie (Jean-Claude Fauchère, MD); Zentrum für MR-Forschung (Ruth O’Gorman, PhD; Ianina Scheer, MD; Hadwig Speckbacher, MTRA); Zurich Pharmacy (D. Fetz, B. Christen).

Funding/Support: This trial was sponsored by the Swiss National Science Foundation (grant SNF 3200B0-108176).

Role of the Funders/Sponsors: The sponsor had no role in the design and conduct of the study; collection, management, analysis, and interpretation of the data; preparation, review, or approval of the manuscript; or decision to submit the manuscript for publication.

Additional Contributions: We thank the Plateforme de Recherche Clinique of the Department of Pediatrics, University Hospital, Geneva, for help.