Alteration of fractional anisotropy in preterm-born individuals: a systematic review and meta-analysis

Abstract

Backgroud

Neurological disorders are common in preterm (PT) born individuals. Diffusion tensor imaging (DTI) studies using tract-based spatial statistics (TBSS) effectively detect microstructural white matter (WM) abnormalities in the brain. We conducted this systematic review to integrate the findings of TBSS studies to determine the most consistent WM alterations in PT born individuals.

Methods

PubMed, Embase, Web of Science and Science Direct were searched. DTI studies using TBSS in PT born individuals were screened up to October 2022. The systematic review included studies reporting alterations in FA values for the entire brain in a stereotactic space, with three coordinates (x, y, z), according to the seed-based d mapping method.

Results

The search strategy identified seventeen studies that fulfilled our inclusion criteria, with a total of 911 PT-born individuals and 563 matched controls were analysed. Of the seventeen studies, eight were dedicated to 650 adults, five to 411 children and four to 413 infants. Ten studies recruited 812 individuals born very prematurely (GA <29 weeks), six studies recruited 386 moderately premature individuals (GA = 29–32 weeks) and one study recruited 276 individuals born late prematurely (GA >32 weeks). This meta-analysis of six studies including 388 individuals highlighted four brain regions in which fractional anisotropy (FA) was lower in PT group than in people born at term. The quantitative meta-analysis found that the most robust WM alterations were located in the corpus callosum (CC), the bilateral thalamus and the left superior longitudinal fasciculus (SLF) II. Significant changes in FA reflect WM abnormalities in PT born individuals from infant to young adulthood.

Conclusions

Significant changes in FA reflect WM abnormalities in individuals born PT from infancy to young adulthood. The abnormal development of the CC, bilateral thalamus and left SLF may play a vital role in the neurodevelopment of PT individuals.

PLAIN LANGUAGE SUMMARY

Neurological disorders are prevalent in preterm (PT) born individuals. The use of tract-based spatial statistics (TBSS) in diffusion tensor imaging (DTI) studies has proven effective in detecting microstructural abnormalities of the white matter (WM) of the brain. In order to determine the most consistent alterations in WM among those born prematurely, we have screened DTI studies using TBSS in this PT born population up until October 2022. The meta-analysis identified four brain regions where fractional anisotropy (FA) was lower in the PT group than in those born at term. The quantitative meta-analysis identified the corpus callosum, the bilateral thalamus and the left superior longitudinal fasciculus II. As the most robust WM alterations. Various studies have demonstrated the links between PT birth, intelligence quotient, gestational age and subject age.

Keywords:

• Diffusion tensor imaging (DTI)
• tract-based spatial statistics (TBSS)
• preterm born
• voxel-based morphometry (VBM)
• seed-based d mapping (SDM)

Introduction

About 15 million infants are born preterm (PT) (defined as < 37 weeks’ gestational age [GA]) each year, and this incidence is increasing). As survival rates increase, PT birth has become an important public health issue (Allotey et al. 2018). Most of previous studies on PT born individuals have highlighted the increased neurological risks associated with prematurity (Saigal and Doyle 2008). These individuals are at risk for adverse neurodevelopmental outcomes, including deficits in language (Aeby et al. 2013), attention (Bora et al. 2014), executive function, working memory (Menegaux et al. 2017) and poor educational and social outcomes (Anderson 2014). Neurological disorders such as attentional deficit/hyperactivity disorder (ADHD) (McNally et al. 2010) and autism spectrum disorders (ASD) (Solomon et al. 2011) are common in PT-born individuals.

Magnetic resonance imaging (MRI) has been widely used to study brain development in PT-born individuals. Compared with MRI to detect morphometric changes, diffusion tensor imaging (DTI) enables researchers to examine WM properties at a microstructural level (Basser 1995). One important measure of DTI is fractional anisotropy (FA), which is sensitive to the alignment of white matter fibres and their structural integrity (Schmithorst et al. 2002). Several original TBSS studies have observed altered FA values in PT-born individuals. They had a cluster of lower FA located within several major WM fibre tracts compared with the controls, including in the corpus callosum (CC) (Vangberg et al. 2006, Skranes et al. 2007, Allin et al. 2011, Eikenes et al. 2011), cingulum (Eikenes et al. 2011), external capsule (EC) (Skranes et al. 2007, Eikenes et al. 2011), internal capsule (IC) (Vangberg et al. 2006, Skranes et al. 2007), inferior fronto-occipital/longitudinal fasciculus (IFOF/ILF) (Skranes et al. 2007, Eikenes et al. 2011), superior longitudinal fasciculus (SLF) (Skranes et al. 2007, Allin et al. 2011) and occipital WM/optic radiation (Vangberg et al. 2006). Increased FA has also been found in various WM tracts, including the anterior corona radiata (CR) and the inferior fronto-occipital fasciculus (IFOF) (Allin et al. 2011, Padilla et al. 2014). However, some studies brought about inconsistent conclusions, reporting no FA group differences in abovementioned structures (Feldman et al. 2012a, Loe et al. 2013).

To resolve the inconsistency across different studies and figure out the most consistent WM alterations in PT born individuals, we conducted a coordinate-based meta-analysis of the whole-brain studies using seed-based d mapping (SDM) software, which has a technical advantage over other meta-analytical tools in that the same map including positive and negative findings from contributing studies.

Methods

Data source

Comprehensive online searches for literatures were performed in PubMed, Embase, Web of Science and Science Direct databases published before 31th October 2022. The keywords searched were ‘premature labor’ ‘preterm infant’ ‘premature birth’ ‘premature delivery’ or ‘preterm born’; ‘tract-based spatial statistical’ or ‘TBSS’; ‘diffusion tensor’ or ‘DTI’; and ‘fractional anisotropy’ or ‘FA’. Moreover, all the review articles and their references were manually searched to obtain additional papers and raw data. Authors were contacted for any inaccessible raw data of concern. PROSPERO registration is not applicable to this systematic review and meta-analysis.

Study selection and data extraction

Two authors with experience in systematic review independently searched the literature, examined the retrieved articles, and extracted and crosschecked data. The authors then assessed the eligibility of these studies for systematic review. Any discrepancy was discussed and resolved by consensus or a third author.

The included studies should meet the following criteria: (1) analysis of DTI images using TBSS; (2) data were acquired from human infants, adolescents, and adults in the original studies; (3) participants born before 37 weeks’ gestation were included in the preterm born group; (4) healthy age-matched term-born participants were included as a comparison group; and (5) a whole-brain analysis were performed.

The exclusion criteria for this study were as follows: (1) reporting only region of interest (ROI) findings were reported instead of whole-brain results; and (2) reporting only the results of correlation analysis without reporting the results of whole-brain analysis.

To be included in the meta-analysis, studies drawn from the systematic review were required to meet the additional criterion of reporting whole-brain results of FA value alteration in a stereotactic space in three coordinates (x, y, z) according to the SDM method (Radua et al. 2014).

Studies that used multiple independent patient samples were separately compared with matching HC groups, and the appropriate coordinates were regarded as separate datasets. The correspondence authors of articles without whole-brain coordinates but were otherwise suitable for meta-analysis were contacted for additional information. There was no language restriction, although all included articles were written in English.

The following data were extracted from eligible studies for meta-analysis: name of the first author, year of publication, sample size, gender, mean age of subjects, GA, birth weight, the scores of Wechsler intelligence scales (either the Wechsler Intelligence Scale for Children-Revised [WISC-R] or the Wechsler Adult Intelligence Scale [WAIS]-III, depending on the subjects’ age), strength of the magnetic field, number of directions, method of voxel-based analysis, threshold for significance and the method used to correct whole-brain results for multiple comparisons.

Quality assessment checklist

The qualities of the studies were assessed using a set of nine weighted criteria specifically for the review and similar to the previously published quality assessment. The checklist was developed from previous meta-analyses focusing on MRI variables (Kmet et al. 2004, Welton et al. 2015). Although it was not designed as an assessment tool, the checklist provided an objective indication of the rigour in the individual studies. Articles that scored less than a predefined threshold of 50% were excluded from the meta-analysis.

SDM meta-analysis

The meta-analysis was performed using the SDM software package (version 6.11, http://www.sdmproject.com), following a standardised process to compare the FA changes between PT and HC. SDM created effect size and variance maps based on the peak coordinates of each included study. Both negative and positive coordinates were reconstructed in the same map to prevent a particular voxel from appearing to be significant in opposite directions. The re-creation was based on converting the peak t value to Hedges’ effect size and then applying a non-normalised Gaussian kernel to the voxels close to the peak. We used an uncorrecte P < 0.005 as the main threshold (Radua et al. 2013). To reduce the possibility of false-positive results further, we used an additional peak-height threshold of Z > 1 and extent threshold of 10 voxels (Radua et al. 2013). In particular, a TBSS template based on the FMRIB58 FA skeleton (heep://www.fmrib.ox.ac.uk/fsl) was provided to restrict the analysis to the same WM tracts that were available in the TBSS studies.

Furthermore, to assess the robustness of findings, sensitivity and heterogeneity analyses were conducted. First, to test the replicability of the results, we used a systematic, whole-brain, voxel-based, jack-knife sensitivity analysis. We repeated the main analysis six times, discarding a different study each time. If a previously significant brain region remained significant in all or most of the combinations of the studies, it could be concluded that this finding was highly replicable (Radua et al. 2011). Second, between-study variance was analysed to assess the significant heterogeneity of brain abnormalities by using a random-effects model with Q statistics (χ2 distribution converted to z values) and tested with a permutation approach. The potential publication bias was formally assessed with Egger’s tests. We created funnel plots of the peaks of the main findings in order to discard gross abnormalities.

The meta-regressions were analysed with the mean age of subjects at testing, GA, birth weight, percentage of males and the scores of Wechsler intelligence scales (full IQ) in each study as independent variables. To minimise the detection of spurious relationships, we decreased the probability threshold to 0.0005 and required abnormalities to be detected both in the slope and in one of the extremes of the regression, and discarded findings in regions other than those detected in the main analyses. Furthermore, we inspected regression plots to discard fittings driven by too few studies (Radua et al. 2013).

Results

Characteristics of the studies included in the systematic review

The search strategy identified 1546 studies, 17 of which met our inclusion criteria. A total of 1065 PT-born individuals and 563 matched HCs in these studies were analysed. The clinical and demographic data was summarised in Table 1. The scanning methods and altered FA regions from all included studies were showed in Table 2. The flow chart of the search strategy was presented in Figure 1. The studies by Vangberg et al. (2006) and Skranes et al. (2007), as well as part of Feldman et al. (2012b) and Loe et al. (2013) used overlapping patient samples.

Figure 1. Search strategy used. DTI, diffusion tensor; FA, fractional anisotropy, ROI, region of interest; TBSS, tract-based spatial statistical.

Table 1. Demographic and clinical characteristics of subjects in the 17 PT-TBSS studies included in the systematic review.

Author	Year	Participants, n (males)		Age, year		Gestational age, week		Birth weight, g		Full IQ
		PT	HC	PT	HC	PT	HC	PT	HC	PT	HC
Vangberg et al.	2006	34 (16)	47 (18)	15.2	15.5	29.3	39.5	1218	3670	NA	NA
Skranes et al.	2007	34 (16)	47 (18)	15.2	15.5	29.3	39.5	NA	NA	77.9	94.7
Anjari et al.	2007	26 (15)	6 (4)	TEA	NB	28.9	39.7	1084	3300	NA	NA
Rose et al. (1)	2008	12 (5)	10 (3)	TEA	NB	26.9	39.5	1022	3353	NA	NA
Rose et al. (2)	2008	11 (6)	10 (3)	TEA	NB	30.9	39.5	1530	3353	NA	NA
Nagy et al.	2009	74 (23)	69 (16)	14.9	14.3	28.54	39.72	1069.54	3530	NA	NA
Allin et al.	2011	80 (41)	41 (27)	19.2	18.6	28.9	40.2	1263.6	3317.9	95.5	106.3
Eikenes et al.	2011	49 (20)	59 (22)	20.2	20.3	29.2	39.7	NA	NA	NA	NA
Feldman et al. (site1)	2012	25 (12)	40 (18)	12.6	13	28.6	NA	1209	NA	NA	NA
Feldman et al. (site2)	2012	33 (16)	23 (9)	12.6	12.8	29.6	NA	1349	NA	NA	NA
Loe et al.	2013	25 (12)	20 (10)	12.7	12.9	28.6	39.3	1191	3171	109	113
Lee et al.	2013	18 (9)	18 (9)	1.24	1.22	30.2	39.3	1520	3056	NA	NA
Salvan et al.	2014	21 (11)	10 (4)	20.85	20	28.24	39.33	1244.95	3145	99.86	108.9
Padilla et al.	2014	20 (11)	20 (10)	1	1	30.04	39.82	1531.15	3272	NA	NA
Alexandrou et al.	2014	58 (33)	14 (5)	TEA	NB	25.53	38.76	NA	3580	NA	NA
Kelly et al.	2016	193 (92)	83 (49)	TEA	NB	34.4	39.7	2171	3587	NA	NA
Kelly et al.	2016	145 (71)	33 (16)	7.5	7.6	27.5	38.9	974	3244	99.1	109.5
Saunavaara et al. (with FGR)	2017	43 (24)	21 (11)	9	9	28.8	39.4	1269	3356	NA	NA
Saunavaara et al. (without FGR)	2017	13 (6)	21 (11)	9	9	28.4	39.4	808	3356	NA	NA
Young et al.	2018	31 (18)	28 (13)	4.2	4.6	28.62	NA	NA	NA	NA	NA

Note: PT, preterm; HC, healthy control; TBSS, tract-based spatial statistical; IQ, intelligence quotient; HC, healthy control; NA, not available; TEA, term equivalent age; and NB, newborn.

Table 2. The main research results in the 17 PT-TBSS studies included in the systematic review.

Author	year	MRI Telsa	Statistical threshold	coordinate system	diffusion directions	Major findings (PT relative to HC)
Vangberg et al.	2006	1.5	FWE	MNI	6	reduced FA values in the CC, IC and superior fasciculus
Skranes et al.	2007	1.5	FWE	NA	6	reduced FA values in the IC, EC, CC and superior, middle superior and inferior fasciculus.
Anjari et al.	2007	3	Uncorrected	NA		dreduced FA values in genu of the CC, CS
Rose et al. (1)	2008	1.5	multiple comparisons	NA	NA	reduced FA in the genu, splenium of the CC, EC
Rose et al. (2)	2008	1.5	multiple comparisons	NA	NA	reduced FA in posterior commissural tracts of the CR
Nagy et al.	2009	1.5	multiple comparisons	NA	30	reduced FA in external capsule, posterior body, isthmus, CC
Allin et al.	2011	1.5	multiple comparison	MNI	uniform	reduced FA in CC, bilateral SLF and left superior CR, increased FA in bilateral IFOF and bilateral anterior CR
Eikenes et al.	2011	1.5	TFCE	MNI	12	reduced FA in the cerebellar peduncle, CSTt, uncinate fasciculus, SLF, cingulum
Feldman et al. (site1)	2012	3	multiple comparisons	MNI	30	no significant group difference in FA value at either site
Feldman et al. (site2)	2012	3	multiple comparisons.	MNI	30	no significant group difference in FA value at either site
Loe et al.	2013	3	TFCE	MNI	30	no significant group difference in FA value at either site
Lee et al.	2013	1.5	FWE/TFCE	NA	NA	In subgroup A (<1 year old), reduced FA in the CS, IC, EC, crus of the fornix, ILF, ILOF, and CC. In subgroup B (1 to 2 years old), reduced FA in the CS, ILOF, and CC In subgroup C (>2 years old), reduced FA in the CS and genu of CC
Salvan et al.	2014	1.5	FWE	MNI	uniform	reduced FA in left hippocampal, fornix, splenium of the CC, ILF and IFOF
Padilla et al.	2014	3	FWE	MNI	30	reduced FA in the CC increased FA in the anterior CR
Alexandrou et al.	2014	1.5	multiple comparisons	NA	15	reduced FA bilaterally in the frontal WM, CR, CS, genu and splenium of the CC, ILF and the left EC
Kelly et al. (3.9)	2016	3	FWE/TFCE	NA	45	reduced FA in the cerebral peduncle, IFOF, posterior thalamic, uncinate fasciculus, CC, cingulum, EC, IC, fornix, SLF and CR
Kelly et al. (4.5)	2016	3	FWE/TFCE	NA	45	reduced FA was in the right cingulum, right ILF, right IFOF, right uncinate fasciculus, right anterior thalamic radiation and EC
Saunavaara et al. (without FGR)	2017	1.5	FWE	NA	40	no significant differences between the two groups
Saunavaara et al. (with FGR)	2017	1.5	FWE	NA	40	reduced FA values in forceps major, forceps minor, bilaterally anterior thalamic radiation, CST, cingulum, IFOF, ILF, SLF and uncinate fasciculus
Young et al.	2018	3	multiple comparisons	NA	6	reduced FA values in the lateral frontal, middle and superior temporal, cingulate, precuneus and lateral occipital regions

Note: PT, preterm; TBSS, tract-based spatial statistics; MRI, magnetic resonance imaging; HC, healthy control; FEW, familywise error correction; MNI, Montreal Neurological Institute;; FA, fractional anisotropy; CC, corpus callosum; IC, internal capsule; NA, not available; EC, external capsule; CS, centrum semiovale; WM, white matter; CR, corona radiata; IFOF, inferior fronto-occipital fasciculus; SLF, superior longitudinal fasciculus; CST, corticospinal tract; TFCE, threshold-free cluster enhancement; and ILF, inferior longitudinal fasciculus.

A total of eight studies focused on adults, five on children and three on infants. Ten studies recruited the very PT–born (GA <29 weeks) individuals, five studies the moderate PT (GA = 29–32 weeks) and one study late PT (GA >32 weeks).

Only six studies reported the three-dimensional coordinates in stereotactic space (Vangberg et al. 2006, Allin et al. 2011, Feldman et al. 2012a, Loe et al. 2013, Padilla et al. 2014, Salvan et al. 2014). A meta-analysis was performed using SDM software based on them.

Quality assessment

All the articles surpassed the minimum quality threshold (Supplementary Table 1). The item that deducted the most points was the acquisition parameters, such as magnet strength and number of diffusion gradient directions.

Narrative review

TBSS study of preterm-born adults

A total of eight studies recruited 650 adults. Vangberg et al. (2006) and Skranes et al. (2007) used overlapping samples. The main finding was that the PT group had significantly reduced FA values in several WM regions, including the CC, IC and superior fasciculus, compared to the control group. The authors speculated that this may be caused by reduced myelination. Skranes reported in detail that low FA values in these areas were associated with perceptual, cognitive, motor and mental health impairments.

Eikenes (2011), Nagy (2009) and Salvan (2014) found that several major WM tracts had significantly reduced FA values in the PT group, for example in the hippocampal fornix and splenium of the CC, ILF and IFOF. Meanwhile, in the PT group, FA was found to positively correlate with birth weight and total IQ.

However, Allin’s study (2011) differed from others, suggesting FA values were increased in clusters included the bilateral IFOF and bilateral anterior CR. Meanwhile, two studies failed to find FA differences between PT-born adults and controls. The study by Feldman et al. (2012a) consisted of convenience samples of children born at two sites: Stanford University (site 1) and University of Pittsburgh (site 2). The participants in the study by Loe et al. study (2013) overlapped with those of site 1 in the Feldman’s study. They found no significant group difference in FA value at either site. The authors believed this may due to many PT-born participants did not have significant WM lesions on conventional MRI.

TBSS study of preterm-born children

Three studies including 411 individuals were focused on school-age and pre-school-age children.

Saunavaara et al. (2017) divided the PT individuals into two subgroups according to the history of foetal growth restriction (FGR). Compared to term-born controls, PT-born children with FGR were found to have lower FA values in the forceps major, forceps minor, bilaterally anterior thalamic radiation, CST, cingulum, IFOF, ILF, SLF and uncinate fasciculus. However, there were no significant differences in FA between non-FGR and HC groups.

Padilla et al. (2014) conducted similar research. Compared to IUGR, PT-born children had reduced FA in the corpus callosum, and increased FA in the anterior corona radiata. No significant differences were identified in any other diffusivity parameter.

The study by Young et al. (2018) of four-year-old children found that lower FA was mainly in the left hemisphere. Within the preterm group, no significant associations were found between FA and any cognitive ability. In contrast, Kelly’s study found that the reduced FA was concentrated in the right hemisphere (Kelly et al. 2016b). FA was significantly associated to neurodevelopmental outcomes in PT-born children.

Lee et al. (2013) focused on neurodevelopmental trends in premature infants of different ages. A total of 18 PT-born children and 18 term-born children were divided into three subgroups according to the corrected age at the time of DTI scanning. In the preterm group, FA values of overall WM showed an increase with age. In the children older than two years old, no significant difference between groups was found, except in the CS and genu of CC regions, where the FA values in PT-born children were lower. In the full-term group, most WM structures had reached maturation at around one year of age. This trend indicated that WM maturation was a gradual occurrence during a child’s first two years.

TBSS study of preterm-born infants

Four studies included 413 infants born very preterm (born ≤28 weeks GA) (Anjari et al. 2007, Rose et al. 2008, Alexandrou et al. 2014). They found that the most immature infants displayed the most extensive reductions in FA, suggesting more severe WM microstructural abnormalities with increasing exposure to the ex-utero environment. The largest TBSS study in late-PT infants also found that they exhibited widespread brain WM microstructural alterations (Kelly et al. 2016a).

SDM meta-analysis results

Six studies with peak coordinates were included in the meta-analysis (Vangberg et al. 2006, Allin et al. 2011, Feldman et al. 2012a, Loe et al. 2013, Padilla et al. 2014, Salvan et al. 2014). They included a total of 213 PT-born individuals (107 men and 106 women with a mean age of 13.6 [range 1–20.85] years) and 169 matched HC participants (78 men and 91 women with a mean age of 13.5 [range 1–20] years). The mean GA of PT-born individuals was 29.1 weeks [range 28.2–30.0 weeks].

The results of the pooled meta-analysis were provided in Table 3 and Figure 2. Compared with controls, reduced FA was found in the CC, bilateral thalamus and left SLF II. None of the clusters showed significant publication bias based on Egger’s test (p > 0.05) (Egger et al. 1997).

Figure 2. Four clusters with decreased FA were identified in PT compared with HC: the corpus callosm, bilateral thalamus and the left superior longitudinal fasciculus II. FA, fractional anisotropy; PT, preterm; HC, healthy control. CC, corpus callosm; L, left; R, right; SLF, superior longitudinal fasciculus.

Table 3. Regional differences in FA values between PT and HC subjects identified by the present meta-analyses (voxelwise p < 0.005 and FWHM 20 mm).

WM Tract	Voxels, n	MNI Coordinates (peak voxel)			SDM z value	P, uncorrected	Cluster breakdown (voxels, n)
		x	y	z
PT < HC
Corpus callosum	336	18	−44	10	−2.271	∼0	Corpus callosum (336)
Left thalamus	157	−10	−30	6	−1.674	0.000441551	Left thalamus (120)
							Left anterior thalamic projections (37)
Right thalamus	146	12	−30	8	−2.458	∼0	Right thalamus (110)
							Right hippocampus, BA 27 (22)
							Right anterior thalamic projections (14)
Left superior longitudinal fasciculus II	107	−34	054	24	−1.412	0.001545429	Left superior longitudinal fasciculus II (75)
							Left inferior network, inferior fronto-occipital fasciculus (32)

Note: FA, fractional anisotropy; PT, preterm; HC, healthy control; FWHM, full width at half maximum; WM, white matter; MNI, Montreal Neurological Institute; SDM, seed-based d mapping; and BA, Brodmann area.

As Supplementary Table 2 showed, a jack-knife sensitivity analysis revealed that these clusters were highly reliable, especially the clusters in the CC and the right thalamus, which were significant for all iterations.

To characterise the potential effects of key demographic and clinical variables on WM, we explored the following variables by means of simple linear regression: percentages of male individuals, mean age at testing, GA, birth weight, and the scores of full IQ (available in three studies (Allin et al. 2011, Loe et al. 2013, Salvan et al. 2014)). However, these variables were not significantly linearly associated with PT-related WM FA changes.

Discussion

From 32 weeks of gestation, WM experienced an abrupt increase about five-fold in volume, with the majority of this augmentation happening after week 36 (Hüppi et al. 1998). Prior to full term, the foetus resides in the uterus, and is comparatively well shielded from external stimuli. On the contrary, premature infants are subjected to distinct auditory and visual stimuli, and their extra-uterine environment may accelerate WM maturation. This rapid development of the WM particularly in premature infants put it at increased risk of damage (Volpe 1981). Additionally, these infants are vulnerable to brain insults, such as maternal infections, high vulnerability to inflammatory mechanisms (Volpe 2009), oxidative processes (Haynes et al. 2005), and other perinatal complications (Volpe 2001). As a result, WM injury is the most common form of brain injury in very premature infants (Volpe 1981).

The current study pooled TBSS studies on PT to conduct a quantitative meta-analysis and provided a narrative review of current findings. This is the first SDM meta-analysis of PT-TBSS studies. To resolve the inconsistency across different studies, we conducted a coordinate-based meta-analysis of six whole-brain studies. The DSM analysis combines FA values in children, adolescents, and young adults born PT.

Most robust WM alterations: the CC, the bilateral thalamus and the left SLF II

The meta-analysis highlighted four brain regions in which FA was lower in the PT than in the HC. The quantitative meta-analysis found that the most robust WM alterations be located in the CC, the bilateral thalamus and the left SLF II.

The CC is the largest fibre bundle of the human brain connecting the left and right cerebral hemispheres (McNally et al. 2010). The genu of the CC connects the prefrontal and orbitofrontal regions, and the body connects the precentral frontal regions and parietal lobes (Liao et al. 2013). CC injury may be due to the inherent vulnerability of immature oligodendrocytes, which makes the developing CC vulnerable to hypoxic–ischemic injury and haemorrhage (Glickstein and Berlucchi 2008). Reductions in CC volumes have been described in PT-born children and adolescents (Nosarti et al. 2008, Chao et al. 2009). Deficits in FA have also been detected in the CC in PT-born infants (Abe et al. 2004, Vangberg et al. 2006, Skranes et al. 2007, Rose et al. 2008). The study by Dibble et al. (2021) presented analogous findings. The research included eleven studies that utilised DTI data. The corpus callosum was identified as a region of interest in both the early and moderate-late preterm groups, with statistically significant diffusion measures reported in all eleven studies. FA reductions in the CC may indicate the functional impairment of frontal lobe information transfer between hemispheres and may be associated with impairment of several motor, perceptual, and cognitive functions (Glickstein and Berlucchi 2008, Yamada et al. 2015). Low FA values of CC are also found in attention-deficit hyperactivity disorder (ADHD) (Skranes et al. 2007), major depressive disorder (MDD) (Yamada et al. 2015) and bipolar disorder (Yamada et al. 2015).

Our study found that the thalamus showed significantly lower FA in the preterm group. Brain volumes in the thalamus are smaller in PT infants than in full-term infants (Counsell et al. 2007). The thalamus is known to myelinate earlier, at around 24–25 weeks, and is noted to be more vulnerable to injury in the preterm population (Huang and Castillo 2008). The thalamus is involved in language processing and its lesions produce verbal fluency impairment (Crosson 1999).

The present meta-analysis identified significantly decreased FA in the left SLF II. The SLF II occupies the central core of the white matter above the insula. It extends from the angular gyrus to the caudal-lateral prefrontal regions (Makris et al. 2005) contains long association fibres connecting all the lobes. To date, many DTI studies of PT-born patients have reported decreased FA in the SLF. For example, Hüppi et al. (2001) reported fewer fibre bundles in the SLF in PT infants with perinatal brain damage. Saunavaara et al. (2017) also reported that PT-born children had lower FA values in the SLF. The SLF tract is involved in inter-hemispheric communication (Fjell et al. 2010). Low FA values in these tracts may indicate fewer fibre bundles and disturbed fibre organisation that cause cognitive impairments and reduced functioning, for instance, of motor, language articulation and working memory (Makris et al. 2005, Skranes et al. 2007, Hatton et al. 2014).

Meta-regression analyses: no significant associations with age at testing, GA or IQ and PT-related WM FA changes

In our meta-analysis, there were no significant associations with age at testing, GA or IQ and PT-related WM FA changes, at least linearly, but many studies had different results, which may be related to the few studies included in the meta-analysis.

We did not find a linear relationship between FA values and age at testing in meta-analysis, but some recent studies have reported the difference in FA values between PTs and controls seems to increase between birth and young adulthood. At term-equivalent age, the differences in FA are mainly seen in the frontal and central part of the brain (Anjari et al. 2007, Rose et al. 2008). In childhood, association fibres also become affected (Anjari et al. 2007). In the studies of young adults, central WM were still affected, but other affected fibres were the IFOF and ILF (Vangberg et al. 2006).

The meta-analysis showed no correlation between different brain regions and GA, which may be because participants in the studies were concentrated between 28 and 30 weeks of gestation. Of all the studies we included in the systematic review, only one study of Kelly et al. (2016a) was of late-PT infants (born at 32–36 GA).

We did not find a linear relationship between FA values and cognitive function in meta-analysis, but some recent studies have reported an association between WM injury and disabilities in children born prematurely. Impairment of WM maturation in PT infants could result in inadequate neurological maturation and clinical dysfunction related to the specific region. FA was found to correlate positively with total IQ in the young PT-born adults (Eikenes et al. 2011).

Limitations and conclusions

This meta-analysis has some limitations that must be considered. First, the SDM analyses were conducted using a limited number of studies, because the remaining studies were excluded due to the exclusion criterion, such as control groups born not full-term, studies not targeted at whole-brain research, and individuals with missing coordinate data. Second, a subgroup analysis was not presented and discussed in this study because some subgroups contained only one study, which led to a lack of statistical value. Third, a lot of studies provided limited information about clinical characteristics and social background. Fourth, we utilised the FMRIB58 adult template, which is not designed for neonatal MRI analyses, potentially resulting in analysis deviations. Additionally, we solely assessed alterations in white matter FA values in the brain, with varying results for other indicators which will be explored in upcoming studies.

Conclusion

To our knowledge, this is the first SDM meta-analysis on PT-TBSS studies. Our results indicate that the significant changes in FA in PT-born individuals reflect WM abnormalities from childhood to young adulthood. Abnormal maturation of the CC, bilateral thalamus and left SLF II may be a key target in the neurodevelopment of PT-born individuals. In the future, we will need more data to study the dynamic brain microstructure changes of individuals born with PT, as well as the effects of external and internal factors on brain development.

Contribution to authorship

LZ, XL, and CS conceived and designed the study. LZ and GH searched in databases and conducted the study selection. LZ and MC independently screened studies following the inclusion and exclusion criteria and extracted the data from eligible studies. LZ and SZ assessed the risks of bias of all included studies. CS resolved discrepancies in the aforementioned processes. LZ and XL analysed the data. LZ wrote the manuscript, and then XL and CS critically revised the manuscript. All authors reviewed and approved the final manuscript.

Acknowledgements

We thank all the authors of the studies included and the subjects.

Disclosure statement

No potential conflict of interest was reported by the author(s).

Data availability statement

All data relevant to the study are included in the article or uploaded as supplementary information.

Additional information

Funding

This study was supported by the National Key Research and Development Program under ‘Reproductive Health and Prevention and Control of Major Birth Defects’ (No. 2016YFC1000400).