Association between skin-to-skin contact and cognition and behaviour at 5 years in children born at 24–31 weeks’ gestation: a secondary analysis of a national population-based prospective cohort study

Summary

Background

Long-term effects of Skin-to-Skin Contact (SSC) on neurodevelopmental outcome for preterm children are still debated. We aimed to evaluate associations between SSC, cognition and behavior at 5 years among children born at 24–31 weeks of gestation.

Methods

In this secondary analysis of the French Etude Epidémiologique sur les Petits Ages Gestationnels (EPIPAGE-2), a nationwide prospective population-based cohort study conducted in all neonatal units across 25 French regions from Mar 28, 2011, to Dec 31, 2011, we examined preterm-born children divided into two groups based on exposure to skin-to-skin contact (SSC) during the first week of life. Outcomes at 5 years included full-scale intelligence quotient (FSIQ; Wechsler Preschool and Primary Scale of Intelligence, 4th edition) and behavioural difficulties (Strengths and Difficulties Questionnaire, SDQ), using a contemporaneous group of term-born children from the ELFE cohort as a reference. The ELFE cohort enrolled participants during four inclusion periods in 2011: Apr 01–04, 2011; Jun 27–Jul 04, 2011; Sep 27–Oct 04, 2011; and Nov 28–Dec 05, 2011. Included children were born at 24–31 weeks of gestation, remained in the same neonatal unit during the first week of life, were alive at 7 days, had no severe cerebral malformations or decisions to limit/withhold care, and had data on SSC exposure. Results are given after multiple imputation using propensity score methods with inverse probability of treatment weighting approach.

Findings

Of 3669 live-born preterm children, 2726 children were eligible and 2666 included with SSC information during the first week of life available; 2561 survived to 5 years (1328 boys (51.8%). Among survivors, 1581 received SSC during the first week of whom 987 had full follow-up assessment; 980 were not exposed of whom 558 had full follow-up assessment. SSC exposure increased with gestational age (21.7% at 24 weeks to 78.0% at 31 weeks), with variability across neonatal units (range 5%–100%). Compared with non-exposed children, exposed children had higher FSIQ scores (mean difference +2.3, 95% CI [+0.3 to 4.3], p = 0.024) and were more likely to score ≥−1 standard deviation (odds ratio 1.33, 95% CI [1.04–1.70], p = 0.023). There was no significant difference in SDQ scores between groups (mean difference −0.3, 95% CI [−1.1 to 0.5], p = 0.42 and score ≥90th percentile odds ratio 0.92 [0.64; 1.33], p = 0.66).

Interpretation

While the duration of SSC the first week of life and SSC beyond the first week were not accounted for, our findings suggest that SSC during the first week of life may be associated with an increased likelihood of higher FSIQ at 5 years. These results point to a potential neuroprotective role of SSC during this early sensitive postnatal period and highlight important areas for improvement. Variability in SCC implementation across units deserves attention and further research is needed to explore dose–response relationships and timing effects.

Funding

The National Research Agency (via the French Equipex Program of Investments in the Future); the French Institute of Public Health Research/Institute for Public Health and its partners at the French Health Ministry; the National Institute of Cancer; the National Institute of Health and Medical Research; the National Solidarity Fund for Autonomy; and the PremUP Foundation.

Research in context.

Evidence before this study

The long-term neurodevelopmental effects of Skin-to-Skin Contact (SSC), particularly beyond infancy, remain largely unknown. We searched PubMed for articles published from database inception to Dec 31, 2024 with an age filter (preschool child 2–5 years and child 6–12 years) using the terms “skin-to-skin”, “kangaroo”, “cognitive”, “behaviour”, “neurodevelopmental outcome”. Of the 64 studies identified, 46 involved preterm-born children; among these we identified 2 studies focusing on targeted preterm outcomes: one in childhood and one in adulthood, based on small groups of children or children born from 1993 to 1998, with attrition bias. Long-term effects of SSC on neurodevelopmental outcome for extremely and very preterm children have never been evaluated in recent large cohort studies. The aim of the study was to evaluate the associations between SSC initiated during the first week of life and cognitive and behavioural outcomes at 5 years of age, among children born between 24 and 31 completed weeks of gestation, within a large population-based cohort of very preterm children.

Added value of this study

This study, based on prospectively collected data from more than 2500 children born extremely and very preterm in 2011 in France, using a contemporaneous group of term-born children as a reference, reveals substantial variability in exposure to SSC during the first week of life across neonatal units and gestational age groups. Exposure to SSC the first week of life increased with gestational age. SSC exposure was associated with improved cognitive outcomes at 5 years of age, as measured by the Wechsler Preschool and Primary Scale of Intelligence (WPPSI), 4th edition. Among exposed vs non-exposed children, full-scale intelligence quotient mean score was higher and score ≥−1 standard deviation was observed more frequently. No significant differences in behavioural difficulties, assessed using the Strengths and Difficulties Questionnaire (SDQ), were observed between SSC-exposed and non-exposed groups. These findings highlight important areas for improvement and support the potential neuroprotective effects of SSC during this sensitive early period.

Implications of all the available evidence

Improving outcomes for children born extremely and very preterm remains a significant challenge, as no substantial gains in cognitive outcomes have been observed in recent years. SSC, widely recognized and promoted as a safe intervention to enhance short-term outcomes in this high-risk population, offers a promising opportunity to help reduce the developmental gap compared to term-born peers. Although progress has been made since the 2011 data, considerable variability in practice persists, highlighting ongoing disparities and indicating that there is still room for improvement during this sensitive early postnatal period. To ensure equitable access—particularly given the well-established benefits of SSC and the expressed desire of most parents—it is essential to continue efforts to standardise and promote this low-cost intervention across all care settings. Future studies are warranted to better understand the dose–response relationship and optimal timing of SSC, not only for mothers but also for fathers of extremely and very preterm children.

Introduction

Skin-to-skin contact (SSC) immediately after birth is a proven, life-saving intervention, particularly in low- and middle-income countries, where it significantly reduces neonatal mortality among preterm children.¹ Endorsed by the World Health Organization (WHO) and supported by extensive research, SSC has been shown to promote physiological stability, enhance breastfeeding outcomes, and strengthen parent-infant bonding.^2,3 In high-income countries, SSC is considered both safe and feasible for extremely preterm (24–27 weeks) and very preterm (28–31 weeks) children.⁴ Nevertheless, despite these well-documented benefits, the implementation of SSC continues to vary widely between and within countries.⁵ In 2022, a Swedish Neonatal Quality Register study indicate that the existing evidence for SSC is poorly translated into clinical practice for extremely and very preterm children.⁴ The long-term neurodevelopmental effects of SSC, particularly beyond infancy, remain largely unknown.

After the neonatal stay, short term neurodevelopmental impacts of SSC alone, independently of other infant and family centered care interventions, are controversial with no effect on development at 12 months.⁶ However, the first hours after birth are considered a sensitive early period during which biobehavioural interactions between parents and infants activate specific neuroendocrine systems that influence gene expression, brain development, and the formation of parent-infant attachment.⁷ Limiting the toxic stress associated with maternal-neonate separation, and promoting optimal sensory environment for preterm-born neonates, SSC seems neuroprotective in the long term.^7,8 Prolonged and continuous SSC at least 18 h/day in the Kangaroo Mother Care program positively influenced the premature brain networks and synaptic efficacy up to adolescence,⁹ and was associated with neurodevelopmental benefits until adulthood.¹⁰ Late but repeated daily SSC has also been shown to improve child physiologic organization and cognitive control at 10 years of age in children born very and moderately preterm.⁷ Early and frequent SSC in extremely preterm children was associated with early cognitive and communication performance at 12 months.¹¹ However, all these studies were based on small groups of children, or children born from 1993 to 1998, with attrition bias, and long-term effects of SSC on neurodevelopmental outcome for extremely and very preterm children have never been evaluated in recent large cohort studies.

The Etude éPIdémiologique sur les Petits Ages GEstationnels 2 (EPIPAGE-2), a national population-based prospective cohort study, with data collected in the neonatal period on SSC and follow-up at 5^½ years,¹² provides a unique opportunity to study the long-term effects of SSC on neurodevelopmental outcomes among children born extremely and very preterm. The aim of the study was to evaluate associations between SSC started during the first week of life, cognition and behaviour at 5^1/2 years among children born between 24 and 31 completed gestational weeks included in the EPIPAGE-2 follow-up study.

Methods

Study population

Recruitment took place in all neonatal units in 25 French regions (21 of the 22 metropolitan regions and all overseas regions) from Mar 28, 2011, to Dec 31, 2011, during an eight-month period for children born at 24–26 weeks, and a six-month period for those born at 27–31 weeks. Eligible children were those born at 24–31 weeks, hospitalized in the same neonatal unit the first week of life, alive at 7 days after birth, without severe cerebral malformation or decision to limit or withhold care during the first week of life, and with information available regarding SSC during the first week of life. Triplets and quadruplets were excluded. Children were followed up from Sep 1, 2016, to Dec 31, 2017.

Skin-to-skin contact (SSC)

Preterm-born children were divided into 2 groups, exposed and non-exposed, based on whether SSC was performed during the first week of life or not. Information about SSC during the first week of life was recorded at day 7 (yes/no), and if yes, the date of the first SSC was reported. Duration of SSC and the date of SSC if it was performed after the first week of life were not recorded in both groups. All data were prospectively collected during neonatal hospitalization.

Outcomes

The two main outcome measures at 5^½ years were the level of intelligence and behavioural difficulties, using a contemporaneous group of term-born children from the ELFE cohort study¹² as a reference. The ELFE cohort enrolled participants during four inclusion periods in 2011: Apr 01–04, 2011; Jun 27–Jul 04, 2011; Sep 27–Oct 04, 2011; and Nov 28–Dec 05, 2011.¹³

Cognitive domain—we used the French version of the Wechsler Preschool and Primary Scale of Intelligence, fourth edition,¹⁴ which measures cognitive ability in five domains (verbal comprehension, visuospatial indices, fluid reasoning, working memory, processing speed), resulting in a composite full-scale-intelligence-quotient (FSIQ) reflecting the overall level of intelligence. Results are reported as mean and standard deviations (SD), and score ≥−1 SD, using cut-off points according to the mean and standard deviations of the reference sample mentioned above.¹²

Behaviour—we used the Strengths and Difficulties Questionnaire (SDQ),¹⁵ a self-administered parental questionnaire designed to screen for symptoms of hyperactivity/inattention, emotional, conduct, and peer problems. A total score is calculated by summing the scores of each domain, with higher scores indicating more difficulties. Children with a total score greater than or equal to the 90th centile of a reference sample¹² are usually considered as having a substantially raised probability of formally diagnosed mental health problems.¹⁶

Statistical analysis

In the overall cohort, to ensure representativeness, percentages were weighted to consider the differences in the recruitment times for the children born at 24–26 weeks and those born at 27–31 weeks and associations between the exposure and outcomes were adjusted for gestational age. To control for the non-random assignment of children in exposure groups and adjust for confounding, we conducted a propensity score analysis with inverse probability of treatment weighting (IPWT) approach.¹⁷ The propensity score was defined as the children's probability of being exposed to SSC, based on maternal, pregnancy-related, neonatal and unit characteristics, estimated with a logistic regression model. The following variables were included in the model, maternal and pregnancy characteristics: maternal age (<25, 25–35, or > 35 years), primiparity (yes, no), mother country of birth (France, foreign-born), maternal education (less than high school, high school, 1–2 years of graduate studies, ≥3 years of graduate studies), parents' socio-economic status (based on parental occupation, classified into five categories), living in couple relationship (yes, no), cause of preterm birth categorized in five classes: (preterm labour, preterm premature rupture of membranes, hypertensive disorders, isolated foetal growth restriction, and other causes including triplet or quadruplet births and undefined), antenatal corticosteroids administration (yes, no), tocolysis use (yes, no), mode of delivery (vaginal, caesarean), maternal anaesthesia (general anaesthesia: yes, no) and multiple pregnancy (yes, no); neonatal characteristics at birth: gestational age (in completed weeks), sex, small-for-gestational age (defined as a birth weight <10th centile for gestational age and sex based on French intrauterine growth curves¹⁸), Apgar score <7 at 5 min, surfactant administration (yes, no), hemodynamic treatment during the first 72 h of life (yes, no), initiation of lactation the first week of life (yes, no); unit characteristics: neonatal unit volume of activity (number of preterm infants born <32 weeks admitted in 2011, categorized in four groups) and implementation of a developmental care program (Newborn Individualised Developmental Care and Assessment Program (NIDCAP)¹⁹) (yes, no). Average treatment effect (ATE) weights were used to estimate the treatment effects in the whole population.¹⁷ Balance statistics were graphically assessed by convergence of the balance measures after iterations of the model and by standardised difference. A standardised difference of 10% or more is generally considered meaningful. The reported measures of association are odds ratio (OR) and risk difference for categorical variables and mean difference for continuous variables with 95% confidence intervals (CIs). Generalised estimating equation (GEE) regression analysis with an exchangeable correlation structure and robust standard error was used to account for the non-independence of outcomes within neonatal units. Missing baseline and outcomes variables were handled by multiple imputations by chained equations with the R package ‘mice’ using variables that were included in the propensity score and outcomes. A large number of predictors was included to ensure the “missing at random” assumption during imputation and for the reliability of the imputed results. Missing data for baseline characteristics ranged from 0% to 9.0%, exceeding 5.0% only for data on living in a couple relationship and maternal educational. Missing data for 5^½ years FSIQ and behavioural scores were 38% and 40% respectively. We generated fifty independent imputed datasets with thirty iterations each. Estimates were pooled according to Rubin's rule.²⁰ Propensity score analysis was performed for each imputed dataset to estimate the exposition effect. Subgroup analyses were performed by gestational age group 24–27 and 28–31 weeks, with interaction p-values to test whether the effect differs significantly between subgroups.

We performed several sensitivity analyses. Firstly, to address potential bias from extreme weights in our IPTW-based propensity score analysis, we performed an overlap weight cohort analysis, assigning (1–propensity score) to exposed children and the propensity score itself to non-exposed children. This approach prevents extreme weights and outliers from dominating results.²¹ Secondly, to control in a different way for the non-random assignment of children in exposure groups and adjust for confounding, we also performed a matching with replacement by gestational age in weeks, and propensity score within a calliper of 0.2 standard deviations of the logit of the propensity score. Thirdly, in the overall cohort, we used logistic regression models (complete cases and multiple imputation) to assess the relationship between SSC and outcomes, adjusting for gestational age, sex, birth weight centile, parent's socio-economic status, maternal education, and country of birth. Fourthly, we used an instrumental variable approach because propensity score cannot completely remove hidden biases from unmeasured confounders.²² NICU preference for SSC during the first week of life served as the instrument, using a two-stage residual inclusion approach, adjusting for gestational age, sex, birth weight centile, parent's socio-economic status, maternal education, and country of birth (eMethods1 in Supplement).²³

All tests were two sided. p Values < 0.05 were considered significant. All analyses were performed with SAS software (V.9.4) and R software (4.0.2).

Ethics approval

The EPIPAGE-2 study was approved by the National Data Protection Authority (Commission Nationale de l’Informatique et des Libertés, CNIL n°911009), the Consultative Committee on the Treatment of Information on Personal Health Data for Research Purposes (CCTIRS: Comité Consultatif sur le Traitement de l'Information en matière de Recherche, approval granted November 18, 2010; reference number 10.626), and the Committee for the Protection of People Participating in Biomedical Research (CPP: Comité de Protection des Personnes, approval granted March 18, 2011, reference CPP SC-2873).¹² This study followed the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) reporting guideline for cohort studies. The ELFE study was approved by the National Data Protection Authority (Commission Nationale de l’Informatique et des Libertés, CNIL n°910504), the Consultative Committee on the Treatment of Information on Personal Health Data for Research Purposes (CCTIRS: Comité Consultatif sur le Traitement de l'Information en matière de Recherche; reference number 2011X716AU), and the Committee for the Protection of People Participating in Biomedical Research (CPP: Comité de Protection des Personnes, reference CPP n°IDFIX-11-024).¹³ Recruitment and data collection in both cohorts occurred only after families had received information and provided written informed consent to participate in the study.

Role of the funding source

The funder of the study had no role in study design, data collection, data analysis, data interpretation, or writing of the report. AM, LM and VP have access to and verify the underlying study data, and take full responsibility for the integrity of the data, the accuracy of the data analysis, and for the decision to submit for publication. They are the guarantors.

Results

Study population

Among 2726 children born at 24–31 weeks in the hospital where he/she was hospitalized during the first week of life, information about SSC started during the first week of life was available for 2666 children (2561 survivors at 5^½ years); 1607 (61.7%) were exposed to SSC the first week of life (1581 (63.0%) survivors at 5^½ years) and 1059 (38.3%) non-exposed or exposed after the first week of life (980 (37.0%) survivors at 5^½ years) (Fig. 1). Rates of SSC across neonatal units varied from 4.8% to 100.0% (eFigure S1 in Supplement) without any correlation between SSC and patient volume in each neonatal unit. In the overall cohort, children in the exposed vs non-exposed group were more likely to have mothers born in France, primiparous, with higher educational level and socio-economic status, were more often singleton, with higher gestational age, with higher initiation of lactation, were more likely admitted in a unit with lower volume activity and with NIDCAP implementation, and were less likely to be small-for-gestational age, to have a low Apgar score, to receive surfactant and hemodynamic treatment. Additional characteristics of exposed vs non-exposed children in the overall cohort are shown in Table 1. Exposure to SSC increased with gestational age, from 21.7% (10/46) for children born at 24 weeks to 78.0% (488/626) for children born at 31 weeks (eTable 1 in Supplement). Characteristics of children participating and non-participating at inclusion and at 5^½ years are detailed in eTable 2 in Supplement. At 5½ years, participating children were more likely to have older mothers, to be born in France to primiparous mothers, to come from families with higher educational and socio-economic status, and to have been more exposed to SSC.

Fig. 1.

Flow chart of the study population. Legend: Abbreviations: D0, day of birth; D7, day 7 after birth. ∗ included parental questionnaire and an evaluation by a psychologist.

Table 1.

Baseline characteristics of mothers, pregnancies, neonates, and units according to exposure to skin-to-skin, N = 2561.

	Overall cohorta			IPTW cohortb
	Exposed to skin to skin D0-D7 (n = 1581)	Non-exposed or exposed after D7 (n = 980)	SDc	Exposed to skin to skin D0-D7 (n = 1581)	Non-exposed or exposed after D7 (n = 980)	SDc
	n/N (%)	n/N (%)		%	%
Maternal characteristics at birth
Maternal age
<25 years	291/1581 (18.3)	185/980 (19.0)	1.7	18.9	19.6	1.9
25–34 years	938/1581 (59.4)	571/980 (58.0)	2.9	58.8	57.8	2.1
>35 years	352/1581 (22.3)	224/980 (23.0)	1.8	22.3	22.5	0.7
Primiparous	890/1561 (57.0)	516/974 (52.4)	9.5	54.8	54.1	1.3
Birth in France	1247/1564 (79.8)	693/970 (71.3)	19.9	76.2	75.4	1.7
Level of education
Less than high school	417/1494 (28.0)	316/901 (35.4)	15.8	32.5	32.3	0.5
High school	328/1494 (21.9)	193/901 (21.4)	1.6	21.8	22.9	2.7
1–2 years of graduate studies	330/1494 (22.1)	190/901 (20.9)	3.3	21.2	21.2	0.3
3+ years of graduate studies	419/1494 (27.9)	202/901 (22.4)	12.7	24.5	23.6	2.0
Parents' socio-economic statusd
Executive	371/1519 (24.3)	176/926 (18.9)	13.7	21.7	21.2	1.1
Intermediate	312/1519 (20.7)	189/926 (20.3)	1.6	19.4	18.9	1.3
Administration	409/1519 (26.9)	258/926 (28.2)	2.8	27.5	27.2	0.8
Service, trade	208/1519 (13.7)	131/926 (13.9)	1.3	14.6	15.7	3.3
Manual worker, unemployed	219/1519 (14.5)	172/926 (18.6)	12.0	16.8	17	0.5
Living in a couple relationship	1380/1511 (91.4)	810/903 (89.9)	4.7	90.4	90.1	0.9
Maternal obstetrical characteristics
Cause of preterm birth
Preterm labour	601/1581 (37.6)	358/980 (35.7)	4.0	37.7	35.8	3.9
Preterm premature rupture of membranes	416/1581 (26.1)	260/980 (25.7)	1.0	26.2	26.3	0.2
Hypertensive disorders or Placental abruption group	377/1581 (24.2)	242/980 (25.6)	3.2	23.6	24.3	1.6
Isolated foetal growth restriction	106/1581 (6.8)	67/980 (7.4)	2.0	6.5	6.9	1.5
Other	81/1581 (5.2)	53/980 (5.6)	1.9	5.9	6.7	3.5
Antenatal corticosteroids	1390/1559 (89.2)	838/964 (87)	6.6	88	87.9	0.5
Tocolysis	895/1569 (56.7)	561/976 (56.4)	0.8	57.1	56.1	2.1
Caesarean delivery	1050/1575 (67.0)	680/975 (71.5)	9.9	67.8	69.3	3.2
General maternal anaesthesia	213/1564 (13.5)	141/955 (14.8)	3.9	14.7	15	1.0
Multiple	442/1581 (28.0)	346/980 (35.7)	16.5	30.7	31.4	1.5
Neonatal characteristics
Gestational age (weeks), mean (sd)	29.4 (1.6)	28.2 (2.0)	66.9	28.8 (2.0)	28.8 (2.0)	1.2
24	10/1581 (0.5)	36/980 (2.9)	19.1	1.8	1.8	0.4
25	32/1581 (1.5)	105/980 (8.6)	32.6	5.1	5.4	1.4
26	96/1581 (4.6)	143/980 (11.7)	26.1	10	9.8	0.6
27	123/1581 (8.0)	140/980 (15.4)	23.4	9.9	10.1	0.7
28	188/1581 (12.2)	146/980 (16.1)	11.3	12.8	12.8	0.1
29	263/1581 (17.0)	121/980 (13.3)	10.3	15.1	14.8	0.8
30	381/1581 (24.7)	151/980 (16.6)	19.9	21	21.4	1.0
31	488/1581 (31.6)	138/980 (15.2)	39.4	24.4	23.8	1.4
Assigned male at birth	818/1581 (51.7)	510/980 (52.4)	1.4	52.2	53.4	2.4
Small-for-gestational agee	559/1581 (35.7)	385/980 (40.5)	10.0	36.1	38.2	4.4
Apgar Score <7 at 5 min of life	228/1526 (14.7)	175/923 (18.3)	10.1	16.7	16.8	0.4
Surfactant	830/1563 (52.2)	736/978 (73.6)	45.8	61.5	63.0	3.4
Hemodynamic treatment during the first 72 h of life	207/1557 (13.0)	233/974 (23.2)	26.9	17.5	17.8	0.3
Initiation of lactation the first week of life	1049/1504 (69.9)	595/924 (64.1)	12.4	66.7	66.1	1.2
Units' characteristics
Neonatal unit volume activityf
<55	611/1581 (38.8)	298/980 (30.7)	17.1	35.7	35	1.6
[55–70]	328/1581 (20.7)	229/980 (23.4)	6.5	22.2	22.2	0.2
[70–90]	302/1581 (19.1)	135/980 (13.6)	14.7	17.3	17.9	1.4
≥90	340/1581 (21.4)	318/980 (32.2)	24.6	24.7	24.9	0.4
Neonatal Individualised Developmental Care and Assessment Program (NIDCAP)	410/1581 (25.8)	131/980 (13.1)	32.5	21.6	20.2	3.6

Abbreviations: D, days; IPTW, Inverse Probability of Treatment Weighting; SD, standardised difference; CPAP, continuous positive airway pressure.

^aDenominators vary according to the number of missing data for each variable. Percent are weighted to consider the differences in survey design between gestational age groups, proportions are not exactly number of events/numbers in groups due to the weighting.

^bInverse Probability of Treatment Weighting, after multiple imputation for handling missing data.

^cStandardized difference between the 2 exposition groups, reported as percentage.

^dDefined as the highest occupational status between occupations of the mother and the father, or mother only if living alone.

^eSmall-for- gestational age was defined as birth weight less than the 10th percentile for gestational age and sex based on French intrauterine “EPOPé” growth curves (Ego 2016).

^fNumber of neonates born before 32 WG admitted in 2011, obtained from the national hospital discharge database.

Propensity score-inverse probability of treatment weighting analysis

Propensity scores and weights were calculated for 2561 children in the overall cohort. Distributions of propensity scores are summarized in eFigure S2 in Supplement. In the IPWT cohort, standardised difference between the exposed (n = 1581) and non-exposed (n = 980) children were less than 5% in all recorded baseline variables (eg, mean (SD) gestational age of 28.2 weeks (2.0) in both groups) (Table 1).

After multiple imputation, higher FSIQ was observed in exposed vs non-exposed children (mean difference +2.3, 95% CI [0.3; 4.3]), and an FSIQ score ≥−1 SD was observed more frequently in exposed vs non-exposed children respectively (OR, 1.33; 95% CI, 1.04–1.70; absolute risk increase in events [ie, the likelihood of having an FSIQ score ≥−1 SD at 5^½ years] per 100 children, 7.1 [95% CI, 1.0–13.2]). SDQ scores were not significantly different between groups of children exposed and non-exposed to SSC (Table 2; Fig. 2).

Table 2.

Associations between skin-to-skin contact the first week of life and 5^½ years outcomes.

	Overall cohorta		IPTW cohort
	Exposed to skin to skin D0-D7 (n = 1581)	Non-exposed or exposed after D7 (n = 980)	OR or MD (95% CI)b Exposed vs Non-exposed or exposed after D7	p-value
	n/N (%)	n/N (%)
Full scale intelligence quotient (FSIQ)
Mean (SD)	95.6 (15.7)	90.9 (16.3)	2.3 (0.3; 4.3)	0.024
≥−1SDc	970/1581 (61.5)	499/980 (50.8)	1.33 (1.04; 1.70)	0.023
Behavioural difficulties, SDQ score
Mean (SD)	10.7 (5.9)	11.2 (5.9)	−0.3 (−1.1; 0.5)	0.42
≥90th percentilec	174/1581 (10.9)	119/980 (12.2)	0.92 (0.64; 1.33)	0.66

Multivariable analyses after multiple imputation, N = 2561.

Abbreviations: IPTW, Inverse Probability of Treatment Weighting; MD, mean differences; D, days; FSIQ, Full scale intelligence quotient, measured by the Wechsler Preschool and Primary Scale of Intelligence-fourth edition (Wechsler, 2014); SD, Standard deviation; SDQ, Strengths and Difficulties Questionnaire (Goodman, 1997).

^aWeighted to take into account the differences in survey design between gestational age groups.

^bGeneralised estimation equation regression analysis (GEE) to account for non-independence of outcomes between neonatal units, with inverse probability of treatment weighting.

^cCut-off of the distribution related to a reference group born at term (Pierrat, 2021).

Fig. 2.

Sensitivity analyses of the associations between skin-to-skin contact the first week of life and 5^½ years outcomes, N = 2561. Legend: Abbreviations: IPTW, Inverse Probability of Treatment Weighting; MD, mean differences; OR, Odd ratio; FSIQ, Full scale intelligence quotient, measured by the Wechsler Preschool and Primary Scale of Intelligence-fourth edition (Wechsler, 2014); SD, Standard deviation; SDQ, Strengths and Difficulties Questionnaire (Goodman, 1997). ^aInverse Probability of Treatment Weighting, Generalised Estimation Equation regression analysis (GEE) to account for the non-independence of outcomes between neonatal units, after multiple imputation. ^bExposed children receive a weight of (1- propensity score) and children non-exposed receive a weight of the propensity score itself. Generalised Estimation Equation regression analysis (GEE) to account for non-independence of outcomes between neonatal units, after multiple imputation. ^cGeneralised Estimation Equation regression analysis (GEE) to account for non-independence of outcomes between neonatal units, adjusted for gestational age, sex, birth weight centile, parents' socio-economic status, mother level of education, mother birth in France. ^dMatching with replacement by gestational age in weeks, and propensity score. After multiple imputation. 1515 survivors at 5 years in exposed group are matched to 518 survivors non-exposed. ^eCut-off of the distribution related to a reference group born at term (Pierrat, 2021). ^fTwo-stage residual inclusion estimation approach, Generalised Estimation Equation regression analysis (GEE) to account for non-independence of outcomes between neonatal units, adjusted for gestational age, sex, birth weight centile, parents' socio-economic status, mother level of education, mother birth in France.

Sensitivity and subgroup analyses

In the overlap cohort (N = 2561) or in the matched cohort (1518 survivors at 5^½ years in the exposed group matched to 518 survivors non-exposed to SSC), the association between the exposure and FISQ was consistent with the analysis performed in the IPWT cohort (Fig. 2). In the adjusted overall cohort (after multiple imputation, N = 2561), results were similar to the IPWT cohort (FISQ score ≥−1 SD adjusted OR 1.30, 95% CI (1.04–1.63)) (Fig. 2). In the overall cohort (N = 2561), the instrumental variable (ie NICU preference, (partial F statistic = 43)) was associated with SSC exposition, and not with indicators usually associated with outcomes (eTables 3 and 4 in Supplement). Using this instrumental variable approach, association between the exposure and FISQ was consistent with the IPWT cohort (Fig. 2). Whatever the sensitivity analyses, the association between the exposure and SDQ scores was consistent with the analysis performed in the IPWT cohort (Fig. 2). In the subgroup analysis by gestational age group, SSC during the first week of life in children born at 28–31 weeks was significantly associated with an increase in the likelihood of surviving with higher FSIQ (mean difference +2.9, 95% CI [+0.6 to 5.2] and score ≥−1 SD odd ratio 1.41, 95% CI [1.04–1.90]) at 5^½ years (Fig. 3). Behavioural difficulties (>90th percentile) were significantly different among subgroups (Fig. 3).

Fig. 3.

Subgroup analysis by gestational age group of the associations between skin-to-skin contact the first week of life and 5^½ years outcomes, 24–27 weeks n = 685 and 28–31 weeks n = 1876. Legend: Abbreviations: IPTW, Inverse Probability of Treatment Weighting; OR, Odd ratio; MD, mean differences; D, days; FSIQ, Full scale intelligence quotient, measured by the Wechsler Preschool and Primary Scale of Intelligence-fourth edition (Wechsler, 2014); SD, Standard deviation; SDQ, Strengths and Difficulties Questionnaire (Goodman, 1997). ^aInverse Probability of Treatment Weighting, Generalised Estimation Equation regression analysis (GEE) to account for the non-independence of outcomes between neonatal units. ^bCut-off of the distribution related to a reference group born at term (Pierrat, 2021).

Discussion

In the French nationwide population-based cohort study EPIPAGE-2 of preterm children born at less than 32 weeks, exposure to SSC the first week of life varied widely across neonatal units and was associated with an increased likelihood of a higher FSIQ at 5^½ years among survivors. Behavioural difficulties, assessed with the SDQ scores, were not different between groups exposed and non-exposed to SSC.

Our results from a recent large cohort of extremely and very preterm children support cumulative evidence for the long-term effect of SSC intervention at birth^7,9,10 and the strategy of “zero separation” among these vulnerable children.⁸ These findings are consistent with the better long-term developmental outcome reported in preterm children, mostly very preterm.^7,9,10 They are also in accordance with a study of 97 extremely preterm children exposed to more frequent SSC with mothers and fathers, who were more likely to have higher scores on the cognitive and communication scales of the Bayley-III at 12-months.¹¹ Improvements in cognitive and behaviour outcome in infancy might persist into early adulthood with significant impact on daily life.²⁴ This potential effect might be strengthened in case of immediate SSC as improvement in the quality of mother children interaction has been suggested with this strategy in very preterm children.²⁵ Indeed, the twenty year follow-up of preterm children exposed to early SSC demonstrated that long-lasting social and behavioural protective effects increased considerably from infancy to adulthood.¹⁰

The effects observed in our study may be relevant at population level, particularly among the population of extremely and very preterm children who remain at high risk of severe/moderate neurodevelopmental disabilities. Given the wide variation in the implementation of evidence-based practices,²⁶ all strategies that can improve outcomes in this vulnerable population, particularly those included in infant and family centered developmental care programs, should be encouraged and promoted in neonatal units. However, evidence-based practices remain underutilized, and their widespread implementation is still far from being achieved. For example, only 58% of very preterm infants admitted for neonatal care in 19 European regions received all four key evidence-based practices for which they were eligible.²⁶ These infants had increased survival without severe morbidity, suggesting that more comprehensive provision of evidence-based practices could yield substantial gains. Variations in patterns of care across neonatal units are likely due to hospital organizations and clinical styles of practices.²⁷ In 2011, there were no French national clinical guidelines for extremely preterm children. Pierrat et al. found that the variation in SSC across French units was partly attributable to organizational factors, such as NIDCAP implementation, rather than to maternal or neonatal characteristics.²⁷ Nevertheless, multifaceted interventions are required to ensure the effective implementation of SSC, even in countries with numerous facilitating factors, such as Sweden.⁴ While parents request implementation of early and continuous SSC, hospital staff attitudes and environment were described by the parents as both supportive and barriers for their application of SCC.²⁸ Involving families in NICU quality improvement may optimize implementation of SSC at population level and long-term outcome of these vulnerable children.²⁹

This study has several strengths. These strengths include the population-based cohort design and the prospective enrolment of all children born preterm in France (2011) using a standardised questionnaire and developmental assessments referenced to contemporary term-born children, examined with the same protocol. Finally, this study produced unique large-scale national data for a topic where randomized trial would now be unethical in view of the many proven benefits of SSC and the desire of most parents.²⁸

This study also has limitations, primarily potential uncontrolled confounding. To minimize indication bias inherent in this type of study, we included multiple recognized individual (including clinical illness severity) and unit characteristics in the IPWT approach, ensuring a well-balanced distribution between groups. Consistent cognitive outcomes across sensitivity analyses further reduce the risk of misattributing the observed association to SSC. Statistical significance for sensitivity and subgroup analyses should be interpreted as exploratory, as no multiple testing adjustments were made. A limitation is the loss to follow-up. To address this attrition bias, we used multiple imputations, which produced results consistent with the complete cases analysis. Sensitivity analyses, particularly instrumental variable approaches, suggest some uncertainty regarding the precision and robustness of this association. A potential weakness of using NICU-level SSC preference as an instrument is that it may not be specific to SSC exposure and could reflect other differences in care practices and resources. Although the instrumental variable analysis yielded a point estimate similar to that of the primary analysis using IPTW propensity scores, the wide confidence intervals suggest a lack of performance and risk of residual confounding. Furthermore, SSC the first week of life was less practiced among the smaller subgroup of preterm children born at 24–27 weeks, which might limit statistical power and explain the absence of association between SSC and cognitive and behaviour outcome in this subgroup. We also have no information on the date of the first SSC among the group not exposed in the first week of life, nor on the duration of SSC after the first of week of life in both groups (exposed and non-exposed). This may have weakened the observed association and limits our ability to fully explore the overall role of SSC on neurodevelopmental outcomes. However, this limitation should be interpreted in light of the fact that this is the first study to investigate the association between SSC and neurodevelopmental outcomes in a recent cohort of extremely and very preterm children. A limitation of this study is that data on race and ethnicity were not collected, which may affect the generalizability of the findings. Finally, given the observational nature of the study, we cannot draw conclusions about causal relationship between SCC and cognitive development at 5^½ years.

Improving outcomes for children born extremely and very preterm remains a challenge, as no improvement in cognitive outcomes has been observed in recent years.³⁰ Early SSC, recognized and promoted as a safe intervention to improve short-term outcomes among this high-risk population,^1,2,4 offers an opportunity to reduce the FISQ gap compared to term peers. Although progress has been made since the 2011 data, SSC remains poorly translated into clinical practice for extremely and very preterm children,⁴ highlighting ongoing disparities and indicating that there is still room for improvement during this early, sensitive postnatal period.⁷ Moreover, as SSC represents a low-cost and easily accessible intervention, it is essential to continue efforts to standardize and promote its use across all care settings for such vulnerable populations. Future studies are needed to evaluate the dose–response relationships and timing effects of SSC, including among both mothers and fathers of extremely and very preterm-born children.

In this nationwide population-based cohort study of children born at less than 32 weeks, SSC during the first week of life was associated with an increased likelihood of higher FSIQ at 5^½ years. Variability of practices across units deserves attention. Further evaluation of the dose–response relationships and timing effects is needed.

Contributors

AM, LM and VP have access to and verify the underlying study data, and take full responsibility for the integrity of the data, the accuracy of the data analysis, and for the decision to submit for publication. They are the guarantors. AM, LM, and VP conceived and designed the experiments. LM analysed the data. AM wrote the first draft of the manuscript. All authors contributed to the writing of the manuscript, interpretation of data for the work, reviewing it critically for important intellectual content, approved the final version of the manuscript and agreed to be accountable for all aspects of the work. VB coordinated data collection and provided administrative, technical, and material support. The corresponding author attests that all listed authors meet authorship criteria and that no others meeting the criteria have been omitted.

Data sharing statement

The procedures carried out with the French data privacy authority (CNIL, Commission Nationale de l’Informatique et des Libertés) do not provide for the transmission of the database. Nevertheless, consultation by the editorial board or interested researchers may be considered, subject to prior determination of the terms and conditions of such consultation and in respect for compliance with the applicable regulations. Data access procedures can be found on the EPIPAGE-2website [https://epipage2.inserm.fr/index.php/en/related-research/access-to-epipage-2-data].

Declaration of interests

All authors declare no competing interests.