Neurodevelopment and Growth 2 Years After Probiotic Supplementation in Extremely Preterm Infants: A Randomised Trial

ABSTRACT

Aim

To assess neurodevelopment and growth at 2 years corrected age in children born extremely preterm, supplemented with Limosilactobacillus reuteri or placebo.

Methods

Design: Secondary analysis of register data after randomised double‐blind placebo‐controlled trial.

Setting: Inclusion at two neonatal intensive care units in Sweden. Local follow‐up per national guidelines.

Subjects: 134 infants born at 23–27 weeks' gestation with a birth weight < 1000 g.

Intervention: Daily enteral supplementation with placebo or Limosilactobacillus reuteri DSM 17938 (2.5 × 10⁸ colony forming units), from within 72 h after birth to postmenstrual week 35–36.

Main outcome measures: Bayley‐III developmental assessment and auxology at 24 ± 3 months corrected age.

Results

Follow‐up data were available for 110 of 117 surviving participants, and 85 children were assessed with Bayley‐III. Mean language score after supplementation with L. reuteri was 90 (95% confidence interval: 85–95) compared to 83 in the placebo group (95% CI: 78–88), p = 0.04. No other significant differences in neurodevelopment or growth were observed.

Conclusion

Supplementation with L. reuteri in extremely preterm infants was not associated with altered growth nor general neurodevelopmental outcomes at 2 years while language index was improved. This is interesting given previously reported early improved growth of head circumference.

Summary.

• Probiotics have the potential to affect neonatal health and gut microbiome in extremely preterm infants, which might have positive effects on further growth and development.

• In a secondary analysis of a randomised placebo‐controlled trial, Limosilactobacillus reuteri supplementation was associated with improved language assessment results at 2 years of age in children born extremely preterm.

• Other neurological outcomes and growth were not affected.

Abbreviations

ATCC

american type culture collection

CI

confidence interval

DSM

deutsche sammlung von mikroorganismen und zellkulturen

EPT

extremely preterm, i.e., born before postmenstrual week 28

M‐CHAT

modified checklist for autism in toddlers

NDI

neurodevelopmental impairment

PMW

postmenstrual week

PROPEL

prophylactic probiotics to preterms of extremely low birth weight trial

RCT

randomised controlled trial

1. Introduction

While chances of surviving extremely preterm (EPT) birth before 28 full postmenstrual weeks (PMW) have improved [1], prevalence of long‐term disability appears more constant [2]. A population‐based study in Sweden has observed cognitive, language or motor difficulties in 44.5% among surviving infants at 2.5 years corrected age born EPT [3]. Long term disabilities after EPT birth have been described to affect quality of life in childhood and to persist into adulthood [4, 5]. Among EPT infants there is a well‐documented association between poor early growth and later neurodevelopmental impairment with cranial growth being particularly important [6].

Probiotics are live bacteria that are health‐promoting when administered to the host [7]. Prophylactic probiotics to preterm infants have been used with some success to prevent common complications in EPT infants associated with the bacterial colonisation of the gut, such as feeding intolerance, necrotising enterocolitis and late onset sepsis [8, 9, 10]. An increasing bulk of evidence supporting the microbiota gut‐brain axis hypothesis also suggests an effect of probiotics on brain development [11]. Inflammation due to colonising bacteria has been described as being capable of disrupting growth, synaptical development and maturation of the central nervous system in preterm infants [12]. There is sparse knowledge about actual long‐term effects and if there is a possibility to reduce neurodevelopmental impairment. Previous research has suggested that neonatal probiotic supplementation may lower the risk of developing autism spectrum disorders [13].

We have previously reported the results of the randomised double‐blind placebo controlled PROPEL trial in which EPT infants were supplemented with Limosilactobacillus reuteri (formerly Lactobacillus reuteri ) DSM 17938 or placebo, during the neonatal period. Although the primary outcome, time to full enteral feeding, was not shortened by L. reuteri , we observed an improved head growth during the first month of life [14]. The current study presents predefined secondary outcomes of the PROPEL trial on neurodevelopment (language, cognition and motor performance assessed by Bayley‐III) and growth until 2 years of age. We hypothesized that the neonatal supplementation of the probiotic L. reuteri would improve neurodevelopment and growth after EPT birth until 2 years corrected age.

2. Methods

2.1. Participants

Newborn EPT infants were recruited to the double‐blind randomised placebo controlled PROPEL‐trial between June 2012 and August 2015 at the tertiary level neonatal intensive care units in Linköping and Karolinska hospital Solna, Sweden. From 239 eligible infants, 134 were included in the trial (Figure 1). A full description of the trial and the neonatal outcome was previously published [14]. Inclusion criteria were birth in PMW 23 + 0–27 + 6, birth weight < 1000 g, written consent from both parents and introduction of both feeding and supplementation within 72 h after birth. Exclusion criteria were known gastrointestinal malformations, severe congenital disorders, or no realistic hope of survival. Infants were exclusively breast milk fed until reaching 2000 g.

FIGURE 1.

Flowchart for the PROPEL‐trial.

Linköping Academic Research Centre, Linköping, Sweden, which was independent of the research group, constructed the randomisation sequence using Stata v12 statistical software (StataCorp LLC, College Station, TX, USA). Bottles containing the study product Limosilactobacillus reuteri DSM 17938 or placebo were randomised to receive study ID numbers in a 1:1 allocation ratio using blocks of 2, 4, 6 or 8 and stratified for inclusion site.

The study plan was reviewed and approved by Ethics Committee for Human Research in Linköping (Dnr 2012/28–31, 2016/503–32, 2019–04975). It is registered at ClinicalTrials.gov (NCT01603368).

2.2. Intervention

Included infants were assigned to consecutive ID numbers to receive oral supplementation with five drops daily from the corresponding container (Limosilactobacillus reuteri DSM 17938 1.25 × 10⁸ colony forming units or placebo). The bottles and the content were identical regarding smell, taste and visual appearance. Supplementation started within 72 h after birth and ended at PMW 35 + 0–36 + 0.

2.3. Outcome

Routine follow‐up after EPT birth in Sweden is an evaluation at 24 ± 3 months corrected age [15]. This includes parents' report, physical examination, auxology, standardised neurologic examination, general assessment of gross‐ and fine motor function and whether the child is perceived to have normal general development, assessment with Bayley Scales of Infant and Toddler Development, 3rd edition (Bayley‐III) [16] and the autism‐screening tool M‐CHAT (The modified checklist for autism in toddlers) classified into a positive or negative outcome [17]. All is reported to the Swedish Neonatal Quality Register (SNQ), from where data was retrieved for this study. A planned addition of a questionnaire to study feeding problems was removed from the trial due to organisational problems in one study region.

The Bayley‐III test was the predefined neurodevelopmental assessment method at 2 years of age in the PROPEL trial. It produces three index scores for the cognitive, language and motor domains respectively. A composite of neurodevelopment impairment (NDI) was also calculated. If present, the Bayley‐III composite scores were used for this. Impairment was graded as none if all scores ≥ −1 standard deviation (SD), mild (lowest score ≥ −2SD to −1SD), moderate (lowest score ≥ −3 SD to −2 SD) or severe (any score < −3 SD). The results from term born children in a recent Swedish trial were used as reference mean score with SDs for the Bayley‐III assessment, where higher mean scores were found than in the original reference material [3, 16]. When Bayley‐III results were partially or totally missing, available data from the following variables were used: Hammersmith Infant Neurological Examination, diagnosed cerebral palsy, gross motor function classification score, physician's report on general development, head control, sitting, walking and fine motor function, parents' report of speech, walking, vision and hearing. Table S1 contains the template used for the grading.

Statistical analyses of growth used z‐score since body measurements were not measured at an exact age. Recent Swedish normative growth data for preterm children until 3 months corrected age [18] and children [19] were used.

This was a double‐blind trial. Allocation was revealed to investigators when analysing the neonatal clinical outcome. Staff and parents were not aware of the allocation at the follow‐up. The grading for neurodevelopmental impairment was performed without knowledge of allocation.

2.4. Statistical Analyses

Sample size was determined when designing the original PROPEL trial evaluating the time to full enteral feeding [14]. Outcomes with continuous variables were analysed with t‐test for independent samples. Data were analysed according to intention to treat. Multiple linear regression was used for correlating multiple independent factors of continuous data with outcome. Analysis of variance was used for multiple independent categorical data with continuous outcomes. For categorical values χ ²‐ or Fisher's exact test was used as appropriate for observed numbers. Ordinal data was analysed with ordinal regression analysis. Baseline characteristics were summarised by means and SD for continuous data and counts and percentages for categorical data. Means with 95% confidence interval (CI) were calculated for continuous data. Counts and percentages for categorical data were calculated along with 95% CI using Wilson confidence intervals for binomial outcomes. A probability level of < 0.05 was considered statistically significant. IBM SPSS Statistics software, version 28 (IBM Corp, Armonk, NY, USA) was used for statistical analyses.

3. Results

Survival until 2 years corrected age in the L. reuteri and placebo groups was 58/68 (85%; 95% CI 75–92) versus 59/66 (89%; 95% CI 80–95) respectively, p = 0.6. Data from 110 of those 117 surviving participants could be retrieved (flow‐chart Figure 1). Three children in each group were lost to follow up for unknown reasons and one in the L. reuteri group declined follow‐up.

Bayley‐III assessment was performed for 85/110. Background characteristics are shown in Table 1. Table S2 describes the subgroup of 85 infants assessed with Bayley‐III and Table S3 compares the background factors among the children in the follow‐up who were or were not assessed with Bayley‐III. Birth after chorioamnionitis and by caesarean section was more common among infants randomised to L. reuteri supplementation, but in the Bayley‐III tested group infants born by caesarean section were not significantly unequally distributed. There were more boys in the placebo group.

TABLE 1.

Background data, perinatal and family factors, in infants participating in the follow up at 2 years corrected age.

Perinatal data	L. reuteri (n = 54)	Placebo (n = 56)	p a
Gestational age, weeks, mean (SD)	25.4 (1.2)	25.5 (1.2)	0.84
Birth weight, g, mean (SD)	738 (124)	748 (143)	0.72
Birth weight z‐score, mean (SD)	−1.0 (1.1)	−1.0 (1.3)	0.95
Birth length z‐score, mean (SD)	−1.5 (1.7)	−1.3 (1.7)	0.58
Birth head circumference z‐score, mean (SD)	−0.8 (0.7)	−0.7 (0.7)	0.30
Apgar at 10 min, mean (SD)	7.7 (1.9)	7.9 (1.9)	0.47
Small for gestational age, n (%)	12 (22)	11 (20)	0.74
Female sex, n (%)	29 (55)	21 (38)	0.09
Infants from multiple pregnancy, n (%)	25 (46)	16 (29)	0.06
Chorioamnionitis, n (%)	18 (34)	8 (14)	0.02
Caesarean section, n (%)	40 (74)	30 (60)	0.02
Maternal smoking at inclusion, n (%)	3 (6)	5 (9)	0.72
Prenatal steroids administered, n (%)	53 (98)	55 (98)	1.0
Inclusion site
Linkoping, n (%)	16 (30)	20 (36)	0.50
Stockholm, n (%)	38 (70)	36 (64)
Neonatal complications
Bronchopumonary dysplasia, n (%)	32 (59)	34 (61)	0.88
Sepsis, culture positive, n (%)	18 (33)	17 (30)	0.74
Retinopathy of prematurity, grade 3–5, n (%)	10 (19)	6 (11)	0.25
Intracerebral haemorrhage, grade 3–4, n (%)	5 (9)	6 (11)	0.80
Periventricular leucomalacia, n (%)	0 (0)	4 (7)	0.12
Necrotising enterocolitis, grade II–III, n (%)	6 (11)	3 (5)	0.32
Family
Postgraduate parent, data found n (%)	29 (54)	34 (61)
Postgraduate parent, no/yes n (%)	11/18 (38/62)	17/17 (50/50)	0.45
Family constellation, data found, n (%)	46 (85)	46 (82)
Lives with both parents, n (%)	42 (91)	39 (85)	0.77
Moves between separated parents, n (%)	3 (7)	5 (11)
Lives with one parent, n (%)	1 (2)	2 (4)
Mother tongue, data found, n	44 (81)	43 (77)
Swedish or Scandinavian language, n (%)	33 (73)	32 (73)	0.75
Non‐Scandinavian language, n (%)	11 (24)	11 (26)
Bi‐ or multilingual family, data found, n	48 (89)	54 (96)
Bi‐ or multilingual family, n (%)	15 (31)	13 (24)	0.50

^a t‐test for independent samples for continuous data. χ²‐test for categorical data (Fisher's exact test when appropriate).

Infants supplemented with L. reuteri reached higher Bayley‐III language scores (mean (95% CI); 90 (85—95), n = 39; vs. placebo group 83 (78—88), n = 41; p = 0.04). No significant differences were observed in cognition, or motor composite scores (Table 2). The Bayley‐III assessment was incomplete for 15/85 children. Five subjects (two in the L. reuteri and three in the placebo group, respectively) lack language results. Three were not possible to test following extensive developmental problems (two in the placebo group and one in the L. reuteri group). One child was exhausted, and one was not tested for unknown reason (both from the placebo group). There were 12 children (eight in the L. reuteri and four in the placebo group, respectively) that were not or only partially assessed for motor function for unknown reasons.

TABLE 2.

Bayley scale for infant development, 3rd edition (Bayley‐III), the Modified checklist for autism in toddlers (M‐CHAT) and auxology expressed as z‐score at 2 years corrected age in extremely preterm infants supplemented with L. reuteri or placebo.

Bayley‐III index score	L. reuteri			Placebo
	Mean	95% CI	n	Mean	95% CI	n	p
Language	90.0	85.0–95.1	39	83.0	78.3–87.7	41	0.04 a
Cognition	90.9	86.5–95.3	40	89.2	84.5–93.8	45	0.60 a
Motor	92.1	85.3–98.9	33	87.3	82.2–92.4	41	0.24 a

M‐CHAT	n/N	% (95% CI)	n/N	% (95% CI)	p
Positive c	7/25	28 (14–48)	8/25	32 (17–52)	1.0 b

Auxology, z‐score	Mean	95% CI	n	Mean	95% CI	n	p
Weight	−1.2	−1.6 to −0.9	46	−1.6	−1.9 to −1.2	39	0.22 a
Length	−0.6	−1.0 to −0.3	46	−0.8	−1.2 to −0.5	39	0.40 a
Head circumference	−1.9	−2.1 to −1.8	36	−1.9	−2.0 to −1.8	29	0.81 a

^a t‐test for independent samples.

^b Fisher's exact test.

^c Fails ≥ 3, or ≥ 2 central, questions.

The effects of the unevenly distributed background factors sex, multiple pregnancy, caesarean section and chorioamnionitis were further explored. None was significantly associated with the language score.

In the L. reuteri group 24% were perceived as without NDI, and 17% as having a severe NDI, as compared to 13% and 30% respectively in the placebo group (p = 0.16) (Figure 2). M‐CHAT screening results were similar for both study groups (Table 2).

FIGURE 2.

Neurodevelopmental impairment at 2 years corrected age for extremely preterm born children who received Limosilactobacillus reuteri DSM 17938 or placebo in the neonatal period. Ordinal regression analysis for testing of statistically significant difference.

Auxological data from the cohort is presented both as change in z‐score at 2 years of age (Table 2) and the change in z‐score from birth (Figure 3). Catch up growth for weight and head circumference were observed, while z‐score for height did not reach the z‐score at birth.

FIGURE 3.

Change in weight (A), length (B) and head circumference (C) in z‐score from birth to 28 days, to postmenstrual week (PMW) 36 + 0 and to two years corrected age respectively, in infants receiving Limosilactobacillus reuteri DSM 17938 (solid) and placebo (broken lines). z‐score calculated from standardised growth charts (18, 19). Error bars show 95% confidence intervals. **p < 0.01; t‐test for independent samples.

The relationship between growth and the Bayley‐III results was explored. Among all growth parameters recorded (weight, height and head circumference at birth, 28 days, PMW 36 and 2 years; z‐scores from the same occasions; and growth expressed as change in z‐scores from birth to the different time points), growth of the head from birth to PMW 36 expressed as change in z‐score had the strongest correlation to the Bayley‐III result. After adjusting for this factor, no other growth parameters contributed significantly to the Bayley‐III results. Head growth to PMW 36 correlated positively with language index score (B‐coefficient = 6.5; R² = 0.11; p = 0.003), motor score (B‐coefficient = 7.6; R² = 0.24; p = 0.001) and cognition (B‐coefficient 3.9; R² = 0.12; p = 0.001). Head growth to PMW 36 was combined with L. reuteri supplementation in a linear regression model. Supplementation of L. reuteri had an additional positive effect on language score (B‐coefficient for L. reuteri 6.8; p = 0.04; B‐coefficient for head growth to PMW 36 3.8; p = 0.003; total model fit R² = 0.15; p = 0.002) but neither for cognition nor motor scores.

4. Discussion

Extremely preterm infants who received L. reuteri as newborns had an improved mean score in the Bayley‐III language assessment at 24 months corrected age compared with placebo treated controls. No statistically significant differences were found for cognition, motor development, general assessment of NDI or growth parameters.

The mean language score of 83 in the placebo group is in line with previous studies [3, 20]. L. reuteri supplementation was associated with a 7‐point higher language score. This could indicate a significant improvement of the child's communication skills. EPT infants in previous trials have achieved language composite scores 11 to 14 points lower than term infants [3, 20]. Any observed increase is important considering that speech impairment after EPT birth contributes to a lower quality of life [5].

This is to our knowledge the first trial reporting an effect of probiotics on language development. Jacobs et al. presented a follow up after the ProPrems‐trial, where neonatal supplementation of a compound probiotic was found to reduce the incidence of necrotising enterocolitis in infants born ≤ 32 PMW. There were no effects on language development or any other Bayley‐III outcomes [21]. Akar et al. evaluated the effects of the same probiotic as in our study— L. reuteri DSM 17938—in preterms born at ≤ 32 PMW using the previous Bayley‐II assessment tool at 18–24 months corrected age without observing any difference [22]. Neither did Agrawal et al. observe effects on development among infants born at ≤ 32 PMW after supplementation of Bifidobacterium breve on neurodevelopment at 3–5 years of age using Mullen's Scale of Early Learning scores including language testing [23].

Our study is unique by exclusively including children born extremely preterm. They are at a greater risk for poor neurodevelopment than other preterm born infants and this is reflected by the generally lower Bayley‐III scores in our trial compared to others [24]. This might explain why there was an effect in our trial as compared to previous studies that included more mature infants.

Increasing research of the gut‐brain‐axis emphasises the impact of gut colonisation and probiotics for brain development of the preterm infant [11]. There are indications that the intestinal microbiome can initiate inflammatory reactions affecting the developing nervous system [12]. Germ free mice exhibit different behavioural patterns with reduced anxiety that can be reversed with establishing a normal microbiota [25]. Interestingly, microbiome analysis in the present cohort revealed an increased microbiome diversity and lower abundance of Enterobacteriaceae and Staphylococcaceae in stool in the L. reuteri group as compared to the placebo group during the neonatal period [26]. Human studies have also suggested a connection between early microbiome and neuropsychiatric problems [12, 27]. In a follow up of 13‐year‐old children who had participated in a randomised controlled trial with Lactobacillus rhamnosus GG (ATCC 53103) during their first 6 months in an allergy prevention trial, the diagnoses ADHD and Asperger's syndrome appeared in 17% of control subjects compared to none in the probiotic group (p < 0.01) [13]. The results in the present trial, however, did not indicate any effects on autism screening with M‐CHAT.

The previously reported improved head growth during L. reuteri supplementation in the first 4 weeks is interesting in conjunction with the finding of better language results [14], and our analysis showed a correlation between language development and head growth from birth until PMW 36, confirming previous research [6]. Head growth, along with weight and length gain from birth to follow‐up, was similar between the two groups in this study. Neither was there a correlation between z‐score of head circumference at birth and the Bayley‐III outcomes. It is well known that language development has a sensitive period [28]. When Hickey et al. performed a follow up of EPT infants at 8 years, language development correlated to cranial growth during the first year but not to growth during later periods of childhood [2].

This study has several strengths being a double‐blind randomised trial with continuous inclusion at public hospitals without socioeconomic selection. Only 7 of 117 living subjects were lost to follow up. We used Bayley‐III assessment, which is a validated instrument, developed for children born preterm and found to predict to later impairment in EPT infants [29].

There are also limitations of this study. Firstly, it represents a secondary outcome analysis of a trial where the primary outcome was not affected by the intervention and the trial is underpowered for the long‐term outcomes. The multi‐centre design using staff not specially hired for the study may have contributed a larger variation in Bayley III scores and more infants with incomplete assessments. Treatment effects might therefore have been underestimated. Furthermore, using Bayley‐III at the age of 2 years might underestimate impairment that becomes evident later and not adequately predict school age outcomes [20, 30]. The Bayley III's three scales also create a multiple comparisons issue, as the predefined outcome in our trial was not specified for any domain. In the absence of Bayley‐III testing, we relied on structured clinical observations to evaluate the secondary outcome NDI (Table S1), enabling the inclusion of all children in the developmental outcomes' comparison. The algorithm used for this purpose has been applied in a similar trial [3], though it remains unvalidated and did not include parameters for cognitive evaluation. Another limitation is the lack of objective evaluation of hearing and vision in the follow‐up programme.

Additional well‐powered trials are needed to further investigate the potential impact of probiotics on language and neurodevelopment in extremely preterm infants.

5. Conclusion

Neonatal probiotic supplementation with L. reuteri was associated with improved Bayley‐III language score evaluated at 24 months corrected age in children born extremely preterm. This may facilitate their communication skills. This is interesting given previously reported early improved growth of head circumference. No effects on other neurodevelopmental variables, growth or autism screening results were observed. The result should be interpreted with caution being a secondary outcome but should encourage further studying.

Author Contributions

Erik Wejryd: conceptualization, investigation, writing – original draft, methodology, formal analysis. Giovanna Marchini: investigation, conceptualization, writing – review and editing. Peter Bang: writing – review and editing, supervision, methodology. Baldvin Jonsson: conceptualization, writing – review and editing, supervision. Ulrika Ådén: supervision, methodology, writing – review and editing. Thomas Abrahamsson: conceptualization, investigation, writing – review and editing, methodology, supervision, project administration.

Conflicts of Interest

Thomas Abrahamsson has received honoraria for lectures and a study grant from BioGaia AB. The other authors declare no conflicts of interest.

Supporting information

Data S1.