Clinical Neurophysiology 125 (2014) 675–684

Contents lists available at ScienceDirect

Clinical Neurophysiology

journal homepage: www.elsevier .com/locate /c l inph

Electroencephalographic activity of preterm infants is increasedby Family Nurture Intervention: A randomized controlled trialin the NICU

1388-2457/$36.00 � 2013 International Federation of Clinical Neurophysiology. Published by Elsevier Ireland Ltd. All rights reserved.http://dx.doi.org/10.1016/j.clinph.2013.08.021

⇑ Corresponding author. Address: New York State Psychiatric Institute, 1051Riverside Drive, Unit 40, New York, NY 10032, USA. Tel.: +1 212 543 5101; fax: +1212 253 4234.

E-mail address: mgw13@columbia.edu (M.G. Welch).1 These authors contributed equally to the work.2 The FNI Trial Group members: Ladan Afifi, Howard F. Andrews, David A. Bateman,

Amy Bechar, Beatrice S. Beebe, Susan A. Brunelli, Kathryn E. Carnazza, Christine Y.Chang, Patricia A. Farrell, Ewelina S. Fiedor, Amie A. Hane, Qais Karim, ShannaKofman, Yasmine A. Koukaz, John M. Lorenz, Mary T. McKiernan, David P. Merle,Richard A. Polin, Sorana Sopterian.

Martha G. Welch a,b,c,d,⇑,1, Michael M. Myers a,b,d,1, Philip G. Grieve b,e, Joseph R. Isler b, William P. Fifer a,b,d,Rakesh Sahni b, Myron A. Hofer a,d, Judy Austin f, Robert J. Ludwig a, Raymond I. Stark b, the FNI Trial Group2

a Department of Psychiatry, Columbia University College of Physicians & Surgeons, New York, NY 10032, USAb Department of Pediatrics, Columbia University College of Physicians & Surgeons, New York, NY 10032, USAc Department of Pathology & Cell Biology, Columbia University College of Physicians & Surgeons, New York, NY 10032, USAd Department of Developmental Neuroscience, New York State Psychiatric Institute, New York, NY 10032, USAe Department of Biomedical Engineering, Columbia University, New York, NY 10027, USAf Mailman School of Public Health, Columbia University, New York, NY 10032, USA

a r t i c l e i n f o h i g h l i g h t s

Article history:Accepted 17 August 2013Available online 17 October 2013

Keywords:Maternal separationMother–infant interventionEEGPremature birthBrain maturation

� Family Nurture Intervention (FNI) is designed to facilitate affective connections between prematurelyborn infants and their mothers.

� At term age, prematurely born infants with FNI had robust increases in frontal brain activity.� Promoting early nurturing interactions may improve neurobehavioral outcomes of preterm infants.

a b s t r a c t

Objective: To assess the impact of Family Nurture Intervention (FNI) on electroencephalogram (EEG)activity in preterm infants (26–34 weeks gestation).Methods: Two groups were tested in a single, level IV neonatal intensive care unit (NICU; standard care orstandard care plus FNI) using a randomized controlled trial design. The intervention consists of sessionsdesigned to achieve mutual calm and promote communication of affect between infants and their moth-ers throughout the NICU stay. EEG recordings were obtained from 134 infants during sleep at �35 and�40 weeks postmenstrual age (PMA). Regional brain activity (power) was computed for 10 frequencybands between 1 and 48 Hz in each of 125 electrodes.Results: Near to term age, compared to standard care infants, FNI infants showed robust increases in EEGpower in the frontal polar region at frequencies 10 to 48 Hz (20% to 36% with p-values <0.0004). Effectswere significant in both quiet and active sleep, regardless of gender, singleton-twin status, gestationalage (26–30 or 30–35 weeks) or birth weight (<1500 or >1500 g).Conclusion: FNI leads to increased frontal brain activity during sleep, which other investigators find pre-dictive of better neurobehavioral outcomes.Significance: FNI may be a practicable means of improving outcomes in preterm infants.� 2013 International Federation of Clinical Neurophysiology. Published by Elsevier Ireland Ltd. All rights

reserved.

1. Introduction

Preterm birth at <37 weeks gestational age (GA) in the UnitedStates has an incidence of 12% (Hamilton et al., 2013). The problemaddressed by this current study is that accompanying earlypreterm birth is a period of prolonged physical and emotionalseparation of mothers and infants during neonatal care withincreased risk for adverse outcomes during subsequent growthand development. This vulnerability manifests as neurobehavioraldisorders that include attention deficits (Johnson and Marlow,

Fig. 1. FNI Timeline. (A) Scent Cloth exchange began immediately after enrollment.(B) Sustained Touch and Vocal Soothing began in the isolette as soon as infant wasclinically stable and continued throughout stay. (C) Wrapped and Skin to SkinHolding took place as soon as it was clinically appropriate for the infant to come outof the incubator. (D) Family sessions occurred mainly near the end of the NICU stay.

676 M.G. Welch et al. / Clinical Neurophysiology 125 (2014) 675–684

2011), executive dysfunction (Baron et al., 2012; Peterson et al.,2000), depression and psychotic disorders (Nosarti et al., 2012),and autism spectrum disorder (Pinto-Martin et al., 2011). Numer-ous preclinical and clinical studies over the course of the past fiftyyears have confirmed that early mother infant separation leads to awide array of adverse physiological and behavioral consequencesthat persist throughout life (Coe et al., 1988; Hofer, 1996; Sanchezet al., 2001). However, other preclinical studies indicate thatincreasing mother/infant interactions, such as buffered stress re-sponses, can have positive effects on outcome (Meaney et al.,1996). Over the past two decades, specific interventions directedat premature infant care in Neonatal Intensive Care Units (NICU),such as Newborn Developmental Care and Assessment Program(NIDCAP) (Als et al., 1994, 2012), skin-to-skin care (Conde-Agudeloet al., 2011, Feldman et al., 2002) and massage therapy (Field et al.,2006, 2010, Vickers et al., 2004), have shown significant effects onoutcomes. However, these interventions are not universally acceptedand further studies are necessary to document efficacy in preterm in-fants (Conde-Agudelo et al., 2011, Ohlsson and Jacobs, 2013).

This research study tests a new approach for improving develop-ment of preterm infants, Family Nurture Intervention (FNI). The dis-tinguishing feature of FNI is that the intervention staff facilitatescommunication of affect and physiologic co-regulation betweenmother and infant. The intervention starts early and continues overthe full course of hospitalization, thereby minimizing the effects ofmother–infant separation in the Neonatal Intensive Care Unit(NICU). The intervention includes mother and infant scent cloth ex-change, sustained touch and vocal soothing in the isolette with eyecontact, wrapped or skin-to-skin holding and family-based supportsessions (Welch et al., 2012) (see Methods below).

FNI was based both on a method used to treat emotional,behavioral and developmental disorders in older children (Welchet al., 2006) and on basic research in animal models that revealedsensorimotor, olfactory and nurture-based activities within nor-mal mother–infant relationships (Hofer, 1975, 1994, 1996; Myerset al., 1989a,b). In recent laboratory studies, central neural mech-anisms have been found for these effects, involving epigeneticregulation of gene expression patterns in specific brain regions,from infancy to adulthood, as a result of variations in earlymother-infant interactions (Meaney et al., 1996; Meaney, 2001;Jensen and Champagne, 2012; Zhang et al., 2013; Caldji et al.,2011). FNI can be safely and feasibly implemented in a level IVNICU without any negative impact on length of stay (Welchet al., 2013).

Quantitative characterization of features of the EEG unique toearly development provides state-dependent indices for trackingchanges in specific neurophysiologic mechanisms critical for nor-mal development of cortical function (Grieve et al., 2005, 2007,2008; Isler et al., 2005; Myers et al., 1997, 1993, 2012; Sahniet al., 1995, 2005). The impact of prior NICU interventions on braindevelopment has been assessed using electroencephalogram (EEG)measures of complexity and coherence of brain activity duringsleep (Als et al., 2012; Kaffashi et al., 2012; Ludington-Hoe et al.,2006). Prior studies have found that preterm infants suffer fromfunctional abnormalities mediated by the frontal cortex, includingattention, emotion regulation, language and cognition (Baron et al.,2012; Johnson and Marlow, 2011; Nosarti et al., 2012; Petersonet al., 2000; Pinto-Martin et al., 2011). Other studies have foundthat activity (power) at higher frequencies measured during sleepin preterm and term infants is predictive of Bayley neurobehavioralscores later in development (Scher et al., 1996). This current studyobtained high-density, 124-lead, EEG recordings as a novel meansto assess changes in brain activity from a NICU intervention. Thecurrent findings show near term age that the FNI group showshighly significant increases in EEG power in high frequency bandsin the frontal cortex.

2. Methods

2.1. FNI trial design

Data for this study were obtained in a single-center, parallelgroup, randomized controlled trial (RCT) in the level IV NICU atMorgan Stanley Children’s Hospital of New York at Columbia Uni-versity Medical Center. ClinicalTrials.gov has registered this trial(#NCT01439269). Columbia University Medical Center Institu-tional Review Board approved all recruitment, consent, and studyprocedures. The target enrollment for this study was 150 infants.The original protocol restricted the gestational ages of the infantsto 26 to 32 weeks 6 days. After eight months of enrollment, we ob-tained approval to increase the age to 34 weeks 6 days in order toincrease the rate of enrollment, but the target enrollment of 150infants was not changed. Excluded were mothers that did notunderstand and speak English, or had a history of drug addictionor severe mental illness. Entry into the study required that familieshave at least one adult other than the mother in the home. Ex-cluded infants were those with a birth weight below the third per-centile for gestational age or significant congenital defects. Shownin the consort chart (Supplementary Figure S1) are the number ofeligible infants and consents obtained.

2.2. FNI activities and procedures

Following consent, mothers assigned to the FNI group met withNurture Specialists who worked with the mothers and their fami-lies throughout the study to facilitate all aspects of the interven-tion. Mothers of infants assigned to the SC condition received thecare that is standard for infants admitted to the NICU. The timelineof the intervention and the sequence of mother-infant interactionsare depicted in Fig. 1.

The first FNI activities took place with the infant confined to iso-lette care. As soon as possible after birth, the mothers started reci-procal scent cloth exchanges. Two small cotton scent cloths weregiven to the mothers, one to wear in their bra and the other toplace under the head of their infant. Mothers were encouraged torepeat this cycle at each visit to the NICU. As infants became morestable, Nurture Specialists facilitated FNI mothers in making affec-tive contact with their infants through the ports of the isolette,using firm and sustained touch and ‘containment’, speaking emo-tionally to their infants in their native language and when possible,making eye contact. When infants were medically able to leave theincubator, Nurture Specialists helped FNI mothers to engage inskin-to-skin, or non-skin-to-skin holding to continue vocalsoothing and eye to eye contact. Mothers were facilitated in engag-ing regularly in these activities for a minimum of 1 h at a time.When family members were available, the Nurture Specialist

Fig. 2. Schematic of high density EEG net. The 124 electrodes in the EEG net aredistributed within ten standard brain regions for the regional analyses of the study.Note 10 midline leads (open circles) were excluded from statistical analyses byregion.

M.G. Welch et al. / Clinical Neurophysiology 125 (2014) 675–684 677

engaged them in sessions that discussed the importance of the FNIactivities between mother and infant, and offered support to thefamilies with ongoing problems and with their needs arising fromthe arrival of their new infant.

2.3. EEG acquisition and sleep state coding

EEG recording used 128-electrode data acquisition systems(EGI, Inc., Eugene, Oregon); however, only 124 leads are used in in-fants. Data were obtained between 11 am and 4 pm within�30 min after a normally scheduled feeding. A study session of�90 min duration was required to acquire �60 min of EEG data.The details of EEG acquisition are described in prior work (Grieveet al., 2008) Following consultation with the manufacturer of theEEG system, we used special methods to maintain electrode/scalpcontact, including covering the electrode net with plastic wrap todiminish evaporation of the saline in the electrode sponges andcovering the net and plastic wrap with an elastic material (Surgi-last). Application of the net and these extra measures usually tookabout 5 min. These procedures generally allowed impedances to bekept below 50 kilo ohms per manufacturer’s recommendation. TheEEG was sampled at 1000 samples/s using a vertex reference. Thevoltage from each lead was band-pass filtered from 0.1 to 400 Hzand then digitized with 16 bits per sample at the rate of 1000 sam-ples per second. Then, in software, data were re-referenced to theaverage of the 124 lead EEG montage and processed to obtain mea-sures of power (lV2) at specific frequencies for each electrode. Thenegative of the average reference approximates the voltage at thevertex lead as referenced to a point at infinity, thus producingthe 125th lead.

Research assistants performed visual coding of quiet and activesleep, indeterminate, awake and crying states once each minute.This coding was done continuously throughout the EEG study withbehavioral criteria previously shown to be appropriate for preterminfants (Stefanski et al., 1984).

2.4. EEG signal artifact rejection

Multiple steps were taken to diminish inclusion of EEG datacontaminated by movement-related or other sources of non-corti-cal electrical activity. The standard deviation of voltage was com-puted for each 30-s epoch of each lead. Individual lead data wereexcluded if the standard deviation of the lead exceeded 50 lV.Epochs were excluded if more than 20 of the 125 leads exceededthe standard deviation criterion and entire studies were excludedif more than 80% of their epochs were excluded.

2.5. EEG power analysis, outlier rejection and statistical analyses

After dropping all defined artifacts, data were available from atleast one EEG study for each of 134 infants (63 SC, 71 FNI). Fromthese data, EEG power was computed using 1 s FFTs for each ofthe 125 leads. Average power for each of 10 frequency bands (1–3 , 4–6, 7–9, 10–12, 13–15, 16–18, 19–21, 22–24, 25–36 and 37–48 Hz) was computed for 30-s epochs throughout the session.

To define outliers in the power data, regression analyses of logpower versus PMA at the time of the recording were conducted foreach lead, frequency band and sleep state (2500 total analyses; 125leads � 10 bands � 2 states). Outliers from the regression line weredefined when residuals with Studentized value exceeded 1.98 (i.e.p � 0.05 with df � 300). These values were set to missing. Finally,muscle activity can contribute to outlier values of power in a leadat nearly all frequencies. As a percentage of EEG power in a band,these outliers are larger in high frequency bands (Shackmanet al., 2009). Accordingly, to limit the contribution of muscle activ-ity, if data from a lead were set to missing because of excessive

power in the highest frequency band (37–48 Hz), then data forall frequency bands for that lead were set to missing.

With elimination of outliers and acceptance of a minimum of3 min of EEG data within each sleep state to increase the stabilityof measures of EEG power, the median power for all epochs of ac-tive sleep and quiet sleep for each lead and frequency band wasdetermined. This allowed the inclusion of 293 early and near termstudies (129 SC, 164 FNI) for AS analyses and 290 early and lateterm studies (127 SC, 163 FNI) in the QS analyses. In both AS andQS, there was a trend for a greater fraction of SC studies to be ex-cluded as compared with the fraction of FNI studies that were ex-cluded. However, in neither state were these group differences inexclusions significantly different (AS: X2 = 2.52, p = 0.11; QS:X2 = 3.04, p = 0.08).

EEG studies were divided into 2 age groups; preterm age (33.8–36.9 weeks PMA) and a near term age (37.2–44.4 weeks PMA). 80%of the preterm age studies and 25% of the near term age studieswere conducted in the NICU. The remaining studies were done inthe follow-up clinic. These percentages did not differ betweenthe study groups. For overall descriptive purposes, the first phaseof the analyses performed t-tests to characterize differences be-tween FNI and SC for each lead, for each frequency band, for eachsleep state, and for each of the two study ages.

To determine whether group differences were statistically sig-nificant, ANOVAs of log EEG power for each study age groupwere conducted for each of the five regions (frontal polar, fron-tal, temporal, parietal, occipital) (Fig. 2) for each of the 10 fre-quency bands and for each sleep state. These analyses had onebetween-subjects factor (FNI vs. SC) and one within-subjects fac-tor (left hemisphere, right hemisphere). Gestational age at birthand PMA at the time of the EEG studies were included ascovariates.

When a given subject had more than one study within an agerange, results were averaged over the multiple studies within thatrange. Approximately one third of the infants had multiple studieswithin the early range and another third within the near to termage range. The distribution of multiple versus single studies wasnot significantly different between the two groups.

678 M.G. Welch et al. / Clinical Neurophysiology 125 (2014) 675–684

Within each age group 100 tests were conducted (10 frequencybands � 5 regions � 2 sleep states), producing p-values that re-flected the main effects of treatment. To control for multiple test-ing, we employed the False Discovery Rate procedure (Benjaminiand Hochberg, 1995) with a threshold set at 10%. We performedadditional analyses on subsets of infants with EEG studies whohad EEG studies at both ages (33 SC, 43 FNI). These were groupby repeated measures (age) ANOVAs and were limited to left fron-tal polar power in the 19–21 Hz band.

3. Results

3.1. Demographics and clinical characteristics

There were no significant differences between SC and FNIgroups with regard to parent demographics, maternal or infantbirth or clinical conditions (Supplementary Tables S1–S3).

3.2. EEG study characteristics

One hundred thirty-four infants (63 SC, 71 FNI) from 105 (51 SC,54 FNI) pregnancies including singletons and twins provided EEGdata for these analyses (Supplementary Table S4). The recordingswere �1 h in duration in both study groups and in both age groups.In the early EEG studies, FNI infants had fewer epochs identified asoutliers in both active sleep (AS) and quiet sleep (QS) states thandid SC infants.

3.3. Electrode by electrode differences in EEG power at the two ageranges

To evaluate the effect of FNI on EEG power, we conducted t-tests for each electrode, within each sleep state and PMA rangeand for each of 10 frequency bands. Electrode maps of the p-valuesfrom these t-tests are shown in Fig. 3 for quiet sleep and in Fig. 4for active sleep. Included in the maps are data for each of 10 fre-quency bands across the spectral range from 1 to 48 Hz for both

Fig. 3. EEG power during quiet sleep. In quiet sleep FNI had effects on EEG power at themost apparent in frontal regions. Shown are electrode maps of the p-values for the diffespectral frequency range from 1 to 48 Hz and PMA range for each of 10 frequency band

the early and near term studies. Results obtained at the near toterm age range indicate there were highly significant effects ofthe intervention (red to dark brown electrodes, lower sets of elec-trodes in Figs 3 and 4). In both sleep states, the intervention led toregional increases in power, particularly in the most frontal elec-trodes at frequencies above 10 Hz. At the early study age range,there were no robust differences in power between the two groups.

3.4. Regional effects of FNI on EEG power

To test whether the individual electrode differences in Figs 3and 4 were regionally significant, we conducted a series of ANOVAsin which the main effect of FNI vs. SC was determined across tenregions (Fig. 2), ten frequency bands, two sleep states, and bothage ranges. Data from left and right hemispheres were includedas a within-subjects factor in these analyses and gestational ageat birth and PMA at the time of EEG study were covariates.

ANOVAS found no significant intervention group by hemisphereinteractions (Supplementary Tables S5–S8). Thus, results only fo-cus on the main effect of intervention group. Identification of thesubset of p-values that passed the 10% False Discovery Rate thresh-old revealed that near to term age, power above 10 Hz in the leadsover the frontal polar cortex was significantly affected by the inter-vention. As shown in Table 1, there was increased power in this re-gion in FNI infants that ranged from 19% to 36% greater than that inSC infants. The effects of FNI appeared to be more pronounced inQS, although robust changes were found in both sleep states. Atfrequencies above 10 Hz, seven of the QS and six of the AS ANOVAtests passed the 10% FDR threshold.

3.5. FNI increased EEG power in both the first and second half of studyperiod

To test whether the effects of FNI were consistent over thecourse of the trial, we divided the data into the first half and secondhalf of the study period for enrollment in the RCT. We focused onthe data obtained at near to term age in QS. Although multiple

older age range (after �6 weeks of FNI) but not the early range. These effects wererences in power between FNI and SC infants in t-tests for each electrode, across thes across all 30 s epochs that were behaviorally coded as quiet sleep.

Fig. 4. EEG power during active sleep. In active sleep FNI had effects on EEG power at the older age range (after �6 weeks of FNI) but not the early range. These effects weremost apparent in frontal regions. Shown are electrode maps of the p-value for the differences in power between FNI and SC infants in t-tests for each electrode, across thespectral frequency range from 1 to 48 Hz and PMA range for each of 10 frequency bands across all 30 s epochs that were behaviorally coded as active sleep.

Table 1Results from analyses of Frontal Polar EEG power from the near to term studies (PMA37–44.4 weeks). For each sleep state (active and quiet) analyses were repeatedmeasures ANOVAs of log EEG power for each of 10 frequency bands. These analyseshad one between subjects factor (Group: FNI vs. SC) and one within subjects factor(hemisphere). The p-values shown are for the main effect by Group. Note the numberof subjects across frequency bands and states was not constant because outliers wereset to missing for individual cells in these analyses. There were no significantinteractions between treatment group and hemisphere. In both quiet and active sleepin all frequency bands, the power was greater in FNI than in SC infants. Presented inbold are the p-values which passed a 10% False Discovery Rate threshold based on pvalues from 200 tests (5 regions � 10 frequency bands � 2 ages � 2 states).

Frequencyband (Hz)

Quiet sleep Active sleep

FNI vs.SC (%)

p (Group Ns) FNI vs.SC (%)

p (Group Ns)

1–3 +8 0.26438 (44/51) +11 0.19612 (44/50)4–6 +16 0.07060 (44/53) +5 0.38996 (44/50)7–9 +19 0.03022 (44/52) +12 0.10840 (44/50)10–12 +21 0.00364 (44/53) +20 0.00360 (44/51)13–15 +21 0.00501 (44/50) +20 0.00459 (44/49)16–18 +30 0.00018 (44/52) +20 0.00351 (44/50)19–21 +32 0.00003 (44/52) +19 0.00375 (44/49)22–24 +32 0.00006 (44/52) +20 0.00525 (44/49)25–36 +36 0.00007 (44/53) +23 0.00307 (44/51)37–48 +31 0.00121 (44/53) +20 0.01474 (44/51)

M.G. Welch et al. / Clinical Neurophysiology 125 (2014) 675–684 679

frequency bands showed highly significant differences, we focusedon the band that had the most robust results (19–21 Hz). Inspec-tion of the electrode by electrode results in Figs. 3 and 4 suggestedsomewhat more robust results on the left side. Thus, we conductedthese post hoc tests using data from the left frontal polar region. AnANOVA for effects of group (FNI vs. SC) and study period halfshowed there was a highly significant difference between groupswith FNI greater than SC (F(1, 92 = 20.06, p < 0.00003), but no sig-nificant group by study half interaction (p = 0.99). Post-hoc t-testsshowed a significant effect of group in both halves of the interven-tion (p < 0.007 (SC/FNI, n = 24/26) and p < 0.001 (n = 21/27),respectively).

3.6. FNI increased EEG power in both males and females

A gender by group comparison determined whether the effectsof FNI treatment on EEG power were similar in males and females.The ANOVA showed a robust main effect of group with FNI greaterthan SC (F(1, 92) = 20.17, p < 0.00003), but no significant interac-tion between group and sex (p = 0.60). The mean power for malesand females in each group is shown (Fig. 5A). Post-hoc t-testsshowed a significant effect of FNI in both males and females(p < 0.001(SC/FNI, n = 20/28) and p < 0.007 (n = 25/25),respectively).

3.7. FNI increased EEG power in both singletons and twins

We also conducted an ANOVA to determine if the effects of FNIwere similar in singletons and twins (Fig. 5B). There was a highlysignificant effect of group with FNI greater than SC (F(1,92) = 21.19, p < 0.00002) but there was also a significant interac-tion between twin/singleton status and group (F(1, 92) = 6.18,p < 0.02). Post-hoc tests showed that the effects of FNI were signif-icant in both singletons and twins (p < 0.03 (SC/FNI, n = 27/29) andp < 0.01 (n = 18/24), respectively).

3.8. FNI increased EEG power independent of age or weight at birth

Next, we divided the infants into two groups based on gesta-tional age at birth. There were 28 infants born 630 weeks (12 SC,16 FNI) and 70 infants born >30 weeks (33 SC, 37 FNI). There wasa significant main effect of age group with FNI greater than SC(F(1, 94) = 14.43, p < 0.0003) but no significant interaction betweenintervention group and birth age group (p = 0.48). The mean powerfor each birth age group is shown in Fig. 5C. Post-hoc t-testsshowed there was significantly greater power in the FNI group inboth age groups (630 weeks PMA, p < 0.03;>30 weeks PMA,p < 0.001). The effect of the intervention in the left frontal polar re-gion at 19–21 Hz was also significant for infants born <1500 g(p < 0.005(SC/FNI, n = 18/32)) as well as infants born P1500 g(p < 0.001(n = 27/21)) with greater power in the FNI group.

Fig. 5. Effects of FNI on EEG power are found in multiple subsets of infants. EEG power (mean ± SE) in electrodes located over the left frontal cortex in the frequency bandfrom 19 to 21 Hz for subgroups of SC and FNI infants. Results are presented for comparisons of subjects grouped by (A) gender, (B) birth status, twin vs. singleton, (C)gestational age (GA) at birth, (D) Pre- or post-discharge EEG obtained near to term age.

680 M.G. Welch et al. / Clinical Neurophysiology 125 (2014) 675–684

3.9. FNI increased EEG power independent of testing before or afterdischarge

The majority of near to term EEG studies were conducted post-discharge in our follow-up clinic (34 SC, 34 FNI) while a few wereobtained in the NICU prior to discharge (10 SC, 12 FNI). Analysesshowed a main effect of group with FNI greater than SC (F(1,86) = 14.54, p < 0.0003) but no significant interaction betweengroup and discharge status (p = 0.56) (Fig. 5D). Moreover, therewas a significant effect of FNI in both of these subsets (tested inNICU p < 0.02; tested in follow-up clinic p < 0.001).

3.10. FNI alters developmental trajectories of EEG power

Next, we examined developmental trajectories of EEG power byanalyzing changes in power with age (controlling for gestationalage at birth and PMA at testing). Repeated measures ANOVAs wereconducted using data from subjects with measurements both earlyat 33.8–36.9 PMA and near-term at 37.2–44.4 PMA (43 FNI, 33 SC).The developmental trajectories for EEG power in the 19–21 Hzband during QS for the left and right sides of five brain regionsare shown in Fig. 6. Results showed that left frontal polar power in-creased in FNI infants with PMA, whereas SC infants showed nosignificant change with PMA (Fig. 6A). This pattern of change withage resulted in a significant age by group interaction (p < 0.0001). Asignificant age by group interaction was also found for right frontalpolar power (p < 0.001). Even though the differences in frontalpower at term age were not significant, the change from the earlyPMA studies to the near to term age studies was greater in FNI in-fants in both hemispheres (interaction term left frontal p < 0.0005,right frontal p < 0.008; Fig. 6B). Significant age by group interac-tions were also found for the left temporal (p < 0.03; Fig. 6C), aswell as for the left and right parietal regions (p < 0.008, p < 0.005,respectively; Fig. 6D), and for the right occipital region (p < 0.01;Fig. 6E). Post-hoc tests showed that power increased significantly(p < 0.02 to p < 0.00001) in FNI in all but the left temporal region.

In contrast, power increased in SC infants only in the left occipitalregion (p < 0.05). Moreover, in SC, EEG power in the left frontal re-gion decreased with age (p < 0.02).

4. Discussion

Preterm infants are a growing segment of the population. Theseinfants are taken from their uterine environment before fetaldevelopment is complete and are exposed to an artificial environ-ment. Advances in technology have made it possible to reducemortality at early gestational ages. However, adverse outcomesin preterm infants remain high.

There are several reports of positive impact of non-medicalinterventions on outcomes following neonatal intensive care (Alset al., 1994, 2012, Conde-Agudelo et al., 2011, Feldman et al.,2002, 2006, 2010, Vickers et al., 2004), although comprehensiveassessments of these interventions suggest that they have limitedusefulness in resolving problems of long-term outcome (Ohlssonand Jacobs, 2013, Symington and Pinelli, 2006). The Newborn Indi-vidualized Developmental Care and Assessment Program (NIDCAP)focuses on providing care of NICU infants tailored to their observedneurobehavioral characteristics (Als et al., 1994). NIDCAP results inimproved medical and neurobehavioral outcomes (Als et al., 2004;Buehler et al., 1995; McAnulty et al., 2009). These effects are corre-lated with MRI- and EEG-based evidence of increased connectivitybetween frontal and occipital brain regions following NIDCAPintervention (Als et al., 2004). Encouraging mothers to engage inskin-to-skin care of their preterm infants has positive effects upongrowth and clinical course in the NICU (Conde-Agudelo et al.,2011), as well as longer term neurobehavioral outcomes (Feldmanet al., 2002). Skin-to-skin care has also been reported to increaseEEG-based measures of complexity, interpreted as enhanced mat-uration of brain function (Kaffashi et al., 2012; Scher et al., 2009).Massage intervention in preterm infants has been shown to in-crease growth rates and reduce length of stay in the NICU (Fieldet al., 2010; Field et al., 2006, Vickers et al., 2004). Two reports

Fig. 6. Developmental trajectories of EEG power are altered by FNI. EEG power (loge 19–21 Hz (lV2/Hz); mean ± SE) for left and right sides of the (A) frontal polar, (B) frontal,(C) temporal, (D) parietal and (E) occipital brain regions recorded during quiet sleep for SC (circles/dashed line) and FNI (Square/ solid line) infants measured between 34 and<37 weeks PMA vs. P37 to 44 weeks PMA. The p-values are from ANOVAs and reflect the significance level of the interaction between age and intervention group controllingfor gestational age at birth and PMA at testing.

M.G. Welch et al. / Clinical Neurophysiology 125 (2014) 675–684 681

from a small trial (N = 10/group) found that massage therapy accel-erated developmental decreases in inter-burst intervals in EEGactivity, shortened latencies in visual evoked responses, andcaused a transient enhancement of visual acuity at three monthsof age (Guzzetta et al., 2009, 2011). These studies also showed thatmassage therapy led to higher levels of EEG power during activesleep.

Although FNI incorporates some aspects of the above inter-ventions, it differs in central focus. FNI specifically facilitatesthe pairing of mother’s communication of emotion with ex-change of scent, comforting touch, vocal soothing, holding andthe experience of mutual calming. This pairing of affect commu-nication and sensory interaction may underlie another mecha-nism that could explain the effects of FNI on brain function;

namely, learning. The neural substrates for learning exist veryearly in development and have been demonstrated in early mo-tor learning paradigms in rat fetuses (Robinson et al., 2008).Other studies have demonstrated that the human fetus learnsto recognize the mother’s voice (DeCasper and Fifer, 1980; Fiferand Moon, 1994; Moon and Fifer, 2000) and scent (Marlier et al.,1997; Schaal et al., 2000). New learning from olfactory stimuli(Sullivan et al., 1986a,b) guides behavior during the early post-natal period, with temperature and touch being key determi-nants of affiliative learning (Kojima and Alberts, 2011).Learning in infants can occur during a wide range of circum-stances, such as feeding (Delaunay-El Allam et al., 2006) andeven during sleep (Fifer et al., 2010). Based on this body of earlylearning studies and on the design of our intervention, it seems

682 M.G. Welch et al. / Clinical Neurophysiology 125 (2014) 675–684

very likely that some form of learning may underlie the effectson brain development of infants following FNI.

Indeed, our results show that near to term age infants who re-ceive facilitated FNI exhibit robust differences in brain activity(EEG power), when compared to infants receiving standard care.The increases in EEG power in FNI infants ranged from 19% to36% in frequencies from 10 to 48 Hz. The greatest percent change(+36%) occurred during quiet sleep in the low gamma frequencyrange (25–36 Hz). These findings are particularly convincing be-cause they were obtained in a large randomized controlled trialthat employed fine grained high density EEG records with repeatedmeasurements over early development, and held true regardless ofgender, singleton or twin status, gestational age at birth, or birthweight. Further, they held for infants enrolled in the first or secondhalf of the study, as well as for infants studied near to term agewhile still in the NICU or after discharge.

FNI exerts its most highly significant effects in the frontal polarregions at beta and low gamma frequencies. Results from otherstudies have shown that increased high frequency power predictsbetter developmental outcomes at later ages. Results from ourprior study in term infants with cardiac anomalies showed a posi-tive correlation between left frontal beta power, during sleep, andBayley scores at 18 months (Williams et al., 2012). This is consis-tent with prior work in both term and preterm infants (Scheret al., 1996; Tarullo et al., 2012). In older awake infants, frontalgamma power, thought to be related to level of arousal and atten-tion, was correlated with cognitive abilities (Benasich et al., 2008).Resting frontal EEG gamma power measured in awake infants at16 months of age predicted language capabilities at 3 years ofage (Gou et al., 2011). These studies support the hypothesis thatthe increases in high frequency power we demonstrated in pre-term infants following FNI presage improved neurobehavioral out-comes at later stages of development.

Based on studies in older infants, one might have expected tofind effects of FNI on the early development of EEG power left/rightasymmetry in our subjects in the 7–9 Hz frequency range. Thesestudies showed asymmetry in EEG power in frontal areas, particu-larly at frequencies between 6 and 9 Hz in association with mater-nal depression (Dawson et al., 1999a,b), responses to fearfulstimuli (Diaz and Bell, 2012), orphanage rearing (McLaughlinet al., 2011), and quality of maternal care (Hane and Fox, 2006;Hane et al., 2010). In contrast to all of the prior published asymme-try work, our data was obtained during sleep. In our study, we didnot find significant left/right group differences in power, or anygroup by hemisphere interactions in the 7–9 Hz band, in post-menstrual age group, sleep state, or brain region. It is possible thatleft/right differences will emerge at later ages, but we did not re-cord EEG beyond the near to term age range or in the awake state.

Although we found that main developmental effects of FNI werelocalized to frontal polar power, we also found significant group-by-age interactions in multiple brain regions. These results suggestthat there are widespread changes in developmental trajectories ofbrain activity that could be mediated by FNI. Since this study waslimited to a specific age range during sleep, it is also conceivablethat increases in EEG power might be seen in the FNI group inother brain regions and/or at earlier ages under other test condi-tions, such as during the awake state or following stimuli thatevoke EEG responses.

It is important to note that the effects of FNI on infant brainactivation occurred with an apparently small ‘‘dose’’ of the inter-vention. Scent cloth exchange was made approximately 2–3 timesper week, and Vocal Soothing and Sustained Touch with Eye Con-tact sessions took place approximately three times per week. Skinto Skin Holding sessions took place an average of two times perweek for at least one hour each time. In total, Nurture Specialist-facilitated intervention sessions occurred on average approxi-

mately 4 h per week. This relatively small amount of time spentin nurturing activities, however, may not be an accurate indicatorof the intervention effect. In other words, dose and effect may notbe directly correlated. Behavioral and emotional effects might havebeen sustained in both mother and infant beyond each session.Sensory stimulation and associated brain activation during earlydevelopment can elicit cascades of hormone and neurochemicalevents that can shape infant brain development (Johnston et al.,2001; Kuhn and Schanberg, 1998; Park and Poo, 2013). Thus, inaddition to learning mechanisms, we speculate that the increasedbrain activity we see following FNI may serve to shape activity-dependent aspects of brain development and plasticity. In replica-tion studies we plan to adapt EEG-based measures (Clapp et al.,2012) to track changes in brain plasticity as both markers andmechanisms for the effects of FNI.

Our ongoing clinical research will assess the proximal physio-logical and behavioral effects of paired nurturing activities, and ex-plore underlying developmental mechanisms. In follow-up of thisstudy cohort extending to two years of age and longer, we willdetermine the effects of FNI on cognition, language, attention,and emotion regulation.

5. Conclusions

Family Nurture Intervention is a novel intervention strategy forthe neonatal intensive care unit that pairs communication of affectand interactions designed to calm both mother and infant. Com-pared to the SC group, the FNI group showed robust increases inhigh frequency EEG brain power in the frontal polar region at nearto term age. These effects were independent of gestational age orweight at birth, gender, twin status, or discharge status when as-sessed. Such region and frequency-specific increases have beenshown by others to correlate with better neurodevelopment at la-ter ages. Results from this study support the importance of earlymother–infant nurture and of facilitating family nurture in theNICU and suggest that FNI may provide an effective and practicableapproach to improving outcomes in preterm infants.

6. Financial disclosures

Funding for this project was provided by the Einhorn FamilyCharitable Trust and the Fleur Fairman Family to the BrainGut Ini-tiative at Columbia University Medical Center. Funding in part wasalso provided by Columbia University’s CTSA from NCRR/NIH(UL1RR024156). Author MMM received salary support from theNew York State Office of Mental Health Research Scientist position.The authors have no competing financial interests.

Acknowledgments

We would like to thank the NICU staff at the Morgan StanleyChildren’s Hospital of New York as well as the participating fami-lies for their invaluable assistance to our program of research.We are grateful for the support of the Columbia University IrvingInstitute for Clinical and Translational Research, where the bulkof EEG recordings were performed. We also want to acknowledgeand thank our Performance and Safety Monitoring Board for theirthoughtful contributions to the conduct of this study (Bruce Levin,Ph.D., Michael Weitzman, M.D. and Deborah E. Campbell, M.D.).

Appendix A. Supplementary data

Supplementary data associated with this article can be found, inthe online version, at http://dx.doi.org/10.1016/j.clinph.2013.08.021.

M.G. Welch et al. / Clinical Neurophysiology 125 (2014) 675–684 683

References

Als H, Duffy FH, McAnulty G, Butler SC, Lightbody L, Kosta S, et al. NIDCAP improvesbrain function and structure in preterm infants with severe intrauterine growthrestriction. J Perinatol 2012;32:797–803.

Als H, Duffy FH, McAnulty GB, Rivkin MJ, Vajapeyam S, Mulkern RV, et al. Earlyexperience alters brain function and structure. Pediatrics 2004;113:846–57.

Als H, Lawhon G, Duffy FH, McAnulty GB, Gibes-Grossman R, Blickman JG.Individualized developmental care for the very low-birth-weight preterminfant. Medical and neurofunctional effects. JAMA 1994;272:853–8.

Baron IS, Kerns KA, Muller U, Ahronovich MD, Litman FR. Executive functions inextremely low birth weight and late-preterm preschoolers: effects on workingmemory and response inhibition. Child Neuropsychol 2012;18:586–99.

Benasich AA, Gou Z, Choudhury N, Harris KD. Early cognitive and language skills arelinked to resting frontal gamma power across the first 3 years. Behav Brain Res2008;195:215–22.

Benjamini Y, Hochberg Y. Controlling the false discovery rate: a practical andpowerful approach to multiple testing. J R Stat Soc Ser B 1995;57:289–300.

Buehler DM, Als H, Duffy FH, McAnulty GB, Liederman J. Effectiveness ofindividualized developmental care for low-risk preterm infants: behavioraland electrophysiologic evidence. Pediatrics 1995;96:923–32.

Caldji C, Hellstrom IC, Zhang TY, Diorio J, Meaney MJ. Environmental regulation ofthe neural epigenome. FEBS Lett 2011;585(13):2049–58.

Clapp WC, Hamm JP, Kirk IJ, Teyler TJ. Translating long-term potentiation fromanimals to humans: a novel method for noninvasive assessment of corticalplasticity. Biol Psychiatry 2012;71:496–502.

Coe CL, Rosenberg LT, Levine S. Effect of maternal separation on the complementsystem and antibody responses in infant primates. Int J Neurosci1988;40:289–302.

Conde-Agudelo A, Belizan JM, Diaz-Rossello J. Kangaroo mother care to reducemorbidity and mortality in low birthweight infants. Cochrane Database Syst Rev2011. CD002771.

Dawson G, Frey K, Panagiotides H, Yamada E, Hessl D, Osterling J. Infants ofdepressed mothers exhibit atypical frontal electrical brain activity duringinteractions with mother and with a familiar, nondepressed adult. Child Dev1999a;70:1058–66.

Dawson G, Frey K, Self J, Panagiotides H, Hessl D, Yamada E, et al. Frontal brainelectrical activity in infants of depressed and nondepressed mothers: relation tovariations in infant behavior. Dev Psychopathol 1999b;11:589–605.

DeCasper AJ, Fifer WP. Of human bonding: newborns prefer their mothers’ voices.Science 1980;208:1174–6.

Delaunay-El Allam M, Marlier L, Schaal B. Learning at the breast: preferenceformation for an artificial scent and its attraction against the odor of maternalmilk. Infant Behav Dev 2006;29:308–21.

Diaz A, Bell MA. Frontal EEG asymmetry and fear reactivity in different contexts at10 months. Dev Psychobiol 2012;54:536–45.

Feldman R, Weller A, Sirota L, Eidelman AI. Skin-to-skin contact (Kangaroo care)promotes self-regulation in premature infants: sleep-wake cyclicity, arousalmodulation, and sustained exploration. Dev Psychol 2002;38:194–207.

Field T, Diego M, Hernandez-Reif M. Preterm infant massage therapy research: areview. Infant Behav Dev 2010;33:115–24.

Field T, Diego MA, Hernandez-Reif M, Deeds O, Figuereido B. Moderate versus lightpressure massage therapy leads to greater weight gain in preterm infants.Infant Behav Dev 2006;29:574–8.

Fifer WP, Byrd DL, Kaku M, Eigsti IM, Isler JR, Grose-Fifer J, et al. Newborn infantslearn during sleep. Proc Natl Acad Sci USA 2010;107:10320–3.

Fifer WP, Moon CM. The role of mother’s voice in the organization of brain functionin the newborn. Acta Paediatr 1994;397(Suppl):86–93.

Gou Z, Choudhury N, Benasich AA. Resting frontal gamma power at 16, 24 and36 months predicts individual differences in language and cognition at 4 and5 years. Behav Brain Res 2011;220:263–70.

Grieve PG, Isler JR, Izraelit A, Peterson BS, Fifer WP, Myers MM, et al. EEG functionalconnectivity in term age extremely low birth weight infants. Clin Neurophysiol2008;119:2712–20.

Grieve PG, Myers MM, Stark RI, Housman S, Fifer WP. Topographic localization ofelectrocortical activation in newborn and two- to four-month-old infants inresponse to head-up tilting. Acta Paediatr 2005;94:1756–63.

Grieve PG, Stark RI, Isler JR, Housman SL, Fifer WP, Myers MM. Electrocorticalfunctional connectivity in infancy: response to body tilt. Pediatr Neurol2007;37:91–8.

Guzzetta A, Baldini S, Bancale A, Baroncelli L, Ciucci F, Ghirri P, et al. Massageaccelerates brain development and the maturation of visual function. J Neurosci2009;29:6042–51.

Guzzetta A, D’Acunto MG, Carotenuto M, Berardi N, Bancale A, Biagioni E, et al. Theeffects of preterm infant massage on brain electrical activity. Dev Med ChildNeurol 2011;53(Suppl. 4):46–51.

Hamilton BE, Hoyert DL, Martin JA, Strobino DM, Guyer B. Annual Summary of VitalStatistics: 2010–2011. Pediatrics 2013.

Hane AA, Fox NA. Ordinary variations in maternal caregiving influence humaninfants’ stress reactivity. Psychol Sci 2006;17:550–6.

Hane AA, Henderson HA, Reeb-Sutherland BC, Fox NA. Ordinary variations in humanmaternal caregiving in infancy and biobehavioral development in earlychildhood: a follow-up study. Dev Psychobiol 2010;52:558–67.

Hofer MA. Studies on how early maternal separation produces behavioral change inyoung rats. Psychosom Med 1975;37:245–64.

Hofer MA. Hidden regulators in attachment, separation, and loss. Monogr Soc ResChild Dev 1994;59:192–207.

Hofer MA. On the nature and consequences of early loss. Psychosom Med1996;58:570–81.

Isler JR, Garland M, Stark RI, Grieve PG. Local coherence oscillations in the EEGduring development in the fetal baboon. Clin Neurophysiol 2005;116:2121–8.

Jensen CL, Champagne FA. Epigenetic and neurodevelopmental perspectives onvariation in parenting behavior. Parent Sci Pract 2012;12:202–11.

Johnson S, Marlow N. Preterm birth and childhood psychiatric disorders. Pediatr Res2011;69:8R–11R.

Johnston MV, Nishimura A, Harum K, Pekar J, Blue ME. Sculpting the developingbrain. Adv Pediatr 2001;48:1–38.

Kaffashi F, Scher MS, Ludington-Hoe SM. Loparo KA. An analysis of the kangaroocare intervention using neonatal EEG complexity: A preliminary study. ClinNeurophysiol; 2012.

Kojima S, Alberts JR. Warmth from skin-to-skin contact with mother is essential forthe acquisition of filial huddling preference in preweanling rats. Dev Psychobiol2011;53:813–27.

Kuhn CM, Schanberg SM. Responses to maternal separation: mechanisms andmediators. Int J Dev Neurosci 1998;16:261–70.

Ludington-Hoe SM, Johnson MW, Morgan K, Lewis T, Gutman J, Wilson PD, et al.Neurophysiologic assessment of neonatal sleep organization: preliminaryresults of a randomized, controlled trial of skin contact with preterm infants.Pediatrics 2006;117:e909–23.

Marlier L, Schaal B, Soussignan R. Orientation responses to biological odours in thehuman newborn. Initial pattern and postnatal plasticity. C R Acad Sci III1997;320:999–1005.

McAnulty G, Duffy FH, Butler S, Parad R, Ringer S, Zurakowski D, et al. Individualizeddevelopmental care for a large sample of very preterm infants: health,neurobehaviour and neurophysiology. Acta Paediatr 2009;98:1920–6.

McLaughlin KA, Fox NA, Zeanah CH, Nelson CA. Adverse rearing environments andneural development in children: the development of frontalelectroencephalogram asymmetry. Biol Psychiatry 2011;70:1008–15.

Meaney MJ, Diorio J, Francis D, Widdowson J, LaPlante P, Caldji C, et al. Earlyenvironmental regulation of forebrain glucocorticoid receptor gene expression:implications for adrenocortical responses to stress. Dev Neurosci1996;18:49–72.

Meaney MJ. Maternal care, gene expression, and the transmission of individualdifferences in stress reactivity across generations. Annu Rev Neurosci2001;24:1161–92.

Moon CM, Fifer WP. Evidence of transnatal auditory learning. J Perinatol2000;20:S37–44.

Myers MM, Brunelli SA, Shair HN, Squire JM, Hofer MA. Relationships betweenmaternal behavior of SHR and WKY dams and adult blood pressures of cross-fostered F1 pups. Dev Psychobiol 1989a;22:55–67.

Myers MM, Brunelli SA, Squire JM, Shindeldecker RD, Hofer MA. Maternal behaviorof SHR rats and its relationship to offspring blood pressures. Dev Psychobiol1989b;22:29–53.

Myers MM, Fifer WP, Grose-Fifer J, Sahni R, Stark RI, Schulze KF. A novelquantitative measure of Trace-alternant EEG activity and its association withsleep states of preterm infants. Dev Psychobiol 1997;31:167–74.

Myers MM, Grieve PG, Izraelit A, Fifer WP, Isler JR, Darnall RA, et al. Developmentalprofiles of infant EEG: overlap with transient cortical circuits. Clin Neurophysiol2012;123:1502–11.

Myers MM, Stark RI, Fifer WP, Grieve PG, Haiken J, Leung K, et al. A quantitativemethod for classification of EEG in the fetal baboon. Am J Physiol1993;265:R706–14.

Nosarti C, Reichenberg A, Murray RM, Cnattingius S, Lambe MP, Yin L, et al. Pretermbirth and psychiatric disorders in young adult life. Arch Gen Psychiatry2012;69:E1–8.

Ohlsson A, Jacobs SE. NIDCAP: a systematic review and meta-analyses ofrandomized controlled trials. Pediatrics 2013;131:e881–93.

Park H, Poo MM. Neurotrophin regulation of neural circuit development andfunction. Nat Rev Neurosci 2013;14:7–23.

Peterson BS, Vohr B, Staib LH, Cannistraci CJ, Dolberg A, Schneider KC, et al. Regionalbrain volume abnormalities and long-term cognitive outcome in preterminfants. JAMA 2000;284:1939–47.

Pinto-Martin JA, Levy SE, Feldman JF, Lorenz JM, Paneth N, Whitaker AH. Prevalenceof autism spectrum disorder in adolescents born weighing <2000 g. Pediatrics2011;128:883–91.

Robinson SR, Kleven GA, Brumley MR. Prenatal development of interlimb motorlearning in the rat fetus. Infancy 2008;13:204–28.

Sahni R, Schulze KF, Kashyap S, Ohira-Kist K, Fifer WP, Myers MM. Sleeping positionand electrocortical activity in low birthweight infants. Arch Dis Child FetalNeonatal Ed 2005;90:F311–5.

Sahni R, Schulze KF, Stefanski M, Myers MM, Fifer WP. Methodological issues incoding sleep states in immature infants. Dev Psychobiol 1995;28:85–101.

Sanchez MM, Ladd CO, Plotsky PM. Early adverse experience as a developmentalrisk factor for later psychopathology: evidence from rodent and primatemodels. Dev Psychopathol 2001;13:419–49.

Schaal B, Marlier L, Soussignan R. Human foetuses learn odours from their pregnantmother’s diet. Chem Senses 2000;25:729–37.

Scher MS, Ludington-Hoe S, Kaffashi F, Johnson MW, Holditch-Davis D, Loparo KA.Neurophysiologic assessment of brain maturation after an 8-week trial of skin-to-skin contact on preterm infants. Clin Neurophysiol 2009;120:1812–8.

684 M.G. Welch et al. / Clinical Neurophysiology 125 (2014) 675–684

Scher MS, Steppe DA, Banks DL. Prediction of lower developmental performances ofhealthy neonates by neonatal EEG-sleep measures. Pediatr Neurol1996;14:137–44.

Shackman AJ, McMenamin BW, Slagter HA, Maxwell JS, Greischar LL, Davidson RJ.Electromyogenic artifacts and electroencephalographic inferences. Brain Topogr2009;22:7–12.

Stefanski M, Schulze K, Bateman D, Kairam R, Pedley TA, Masterson J, et al. A scoringsystem for states of sleep and wakefulness in term and preterm infants. PediatrRes 1984;18:58–62.

Sullivan RM, Brake SC, Hofer MA, Williams CL. Huddling and independent feeding ofneonatal rats can be facilitated by a conditioned change in behavioral state. DevPsychobiol 1986a;19:625–35.

Sullivan RM, Hofer MA, Brake SC. Olfactory-guided orientation in neonatal rats isenhanced by a conditioned change in behavioral state. Dev Psychobiol.1986b;19:615–23.

Symington A, Pinelli J. Developmental care for promoting development andpreventing morbidity in preterm infants. Cochrane Database Syst Rev 2006.CD001814.

Tarullo AR, Lee J, Gera S, Condon C, Tamura EK, Grieve PG, et al. Languagedevelopment linked to newborn frontal gamma power. Dev Psychobiol2012;54:769.

Vickers A, Ohlsson A, Lacy JB, Horsley A. Massage for promoting growth anddevelopment of preterm and/or low birth-weight infants. Cochrane DatabaseSyst Rev 2004. CD000390.

Welch MG, Hofer MA, Brunelli SA, Stark RI, Andrews HF, Austin J, et al. Familynurture intervention (FNI): methods and treatment protocol of a randomizedcontrolled trial in the NICU. BMC Pediatr 2012;12:14.

Welch MG, Hofer MA, Stark RI, Andrews HF, Austin J, Glickstein SB, et al.Randomized controlled trial of Family Nurture Intervention in the NICU:Assessments of length of stay, feasibility and safety. BMC Pediatrics2013;13:148. http://dx.doi.org/10.1186/1471-2431-13-148.

Welch MG, Northrup RS, Welch-Horan TB, Ludwig RJ, Austin CL, Jacobson JS.Outcomes of Prolonged Parent-Child Embrace Therapy among 102 childrenwith behavioral disorders. Complement Ther Clin Pract 2006;12:3–12.

Williams IA, Tarullo AR, Grieve PG, Wilpers A, Vignola EF, Myers MM, et al. Fetalcerebrovascular resistance and neonatal EEG predict 18-monthneurodevelopmental outcome in infants with congenital heart disease.Ultrasound Obstet Gynecol 2012;40:304–9.

Zhang TY, Labonte B, Wen XL, Turecki G, Meaney MJ. Epigenetic mechanisms for theearly environmental regulation of hippocampal glucocorticoid receptor geneexpression in rodents and humans. Neuropsychopharmacology2013;38:111–23.