We have demonstrated spatially widespread alterations in white matter microstructure in VPT young adults, affecting areas likely to represent the corpus callosum, sensorimotor tracts and many long association tracts in both hemispheres. These microstructural alterations are associated with birth weight and gestational age, and with adult cognitive function.
Our results are consistent with other studies that have examined FA in VPT individuals at different ages – microstructural white matter abnormalities have been reported in preterm or low birth weight neonates 
, infants 
, children 
, and adolescents 
. For example, Constable et al. (2008) 
demonstrated FA reductions in inferior fronto-occipital fasciculus, anterior uncinate and splenium of the corpus callosum in preterm 12-year-olds. Similar patterns of FA reductions have been reported in VLBW adolescents, affecting the internal capsule, corpus callosum, and hemispheric association tracts 
. Eikenes et al. (2010) 
have studied FA and mean diffusivity (MD) in very-low-birth-weight (VLBW≤1500 g) adults at a comparable age to our sample (18–22 years). They also found decreased FA in many regions/tracts, including the corpus callosum. Similarly, Mullen et al. (2011) 
have demonstrated reduced FA in preterm 16-year-olds in several regions, including uncinate fasciculus, external capsule, corpus callosum (splenium), and frontal white matter.
Our findings differ from other studies in that we found clusters where FA was increased in the VPT group relative to the Term group. This was unexpected, and merits further discussion. Similar FA increases were reported by Vangberg et al. (2006) 
in very-low-birth-weight adolescents and by Eikenes et al. (2010) 
, although not by Skranes et al. (2007) 
in a similar sample. Methodological differences may underlie this discrepancy. First, Skranes et al. (2007) 
studied individuals selected by birth weight rather than gestational age – and although overlapping, these two populations are not identical. Second, their participants were scanned at a younger age than ours. There is active growth of white matter and increase in FA during adolescence 
. In a previous study, we have shown a striking pattern of increased growth of the corpus callosum between adolescence and adulthood in VPT individuals 
. Differential growth of white matter between VPT and term groups could alter the pattern of group differences that are observed at different age-ranges. In a comparably-aged group of very-low-birth-weight adults, Eikenes et al (2010) 
reported one area of increased FA in the very-low-birth-weight group relative to controls, which is consistent with our findings. In our study, the areas of increased FA in the VPT group were relatively small (796 voxels) compared to the areas of FA decrease (4688 voxels) – a similar pattern to that reported by Eikenes et al (2010) 
. Given the consistency of this finding across studies, and the relationships that we have demonstrated between these areas of FA increase and neonatal brain injury, we suggest that areas of increased FA in VPT and very-low-birth-weight adults may be biologically meaningful. There is some evidence that, in children and adolescents, there are negative correlations between corpus callosum thickness and cognitive function 
– so, for white matter at least, more does not always mean better.
Regions of increased FA in the VPT group could also represent an ‘unmasking’ effect. In regions of crossing fibres, an individual voxel is likely to contain fibres of more than one orientation. The mean FA of such a voxel could be relatively low (especially if the vectors of the crossing tracts diverge significantly). If this voxel were then to lose some fibres its apparent FA could increase, as the remaining fibres would have a more ‘coherent’ mean orientation. In this model, FA increases in the VPT group relative to the Term group could actually represent regions of white matter loss in the VPT group. The exact microstructural and anatomical correlates of DT-MRI, and how they change during development, are as yet not fully known 
Alternatively, clusters of increased FA in the VPT group might be indicative of compensatory changes – where plasticity of white matter has allowed function to be spared, although white matter integrity has been disturbed by early brain insults. Some of our findings are compatible with this explanation: VPT participants with more severe neonatal brain injury (by ultrasound) had higher FA in several clusters; one cluster of increased FA was associated with lower gestational age (right superior longitudinal fasciculus). In a structural MRI (not DT-MRI) study, Nosarti et al. (2008) 
demonstrated white matter increases in VPT adolescents who had experienced more severe grades of perinatal brain injury.
Two clusters of reduced FA were associated with birth weight and gestational age, although the correlations between FA and birth weight were stronger than those between FA and gestational age. The large cluster involving corpus callosum and corticospinal tract was associated with birth weight only, and the cluster involving the right superior longitudinal fasciculus was associated with both birth weight and gestational age. Although there is considerable overlap, the categories of low birth weight and preterm birth may be indicative of different pathological processes. Our results may indicate that gestational age and birth weight have differential relationships to adult white matter, but we were not able to assess patterns of intrauterine growth which would be necessary properly to address this question. Andrews et al. (2010) 
demonstrated a relationship between birth weight and corpus callosum FA in preterm children (mean age at assessment 11 years). Eikenes et al. (2010) 
in very-low-birth-weight young adults also showed relationships between FA reduction and perinatal adversity, including gestational age, birth weight, days in Neonatal Intensive Care and length of mechanical ventilation. White matter injury is common in VPT and very-low-birth-weight neonates, and its prevalence and severity is related to perinatal adversity 
We also have demonstrated associations between structural neuroimaging measures and FA in several clusters. These clusters were all ones in which FA was reduced in the VPT group. Notably, corpus callosum size was associated with FA in clusters 1, 2 and 4. Two of these clusters anatomically involved the corpus callosum. Thus the structural MRI findings are consistent with the DT-MRI findings. Bassi et al (2011) 
report relationships between white matter lesions and reduced FA in corticospinal tracts or preterm infants, which is consistent with our findings. Lateral ventricular volume was also associated with FA in 3 of the VPT<Term clusters. Enlargement of the lateral ventricles is a known sequel of perinatal brain injury in VPT infants 
and is associated with white matter damage, and subsequently reduced white matter volumes 
Alterations of FA in the VPT group were also associated with cognitive outcome. IQ was associated with FA in three of the VPT<Term clusters. This relationship was specific for performance IQ rather than verbal IQ. We also found significant associations between higher FA and better global memory function, in all 4 clusters of reduced FA. Associations between FA and cognition have been observed in other VPT and very-low-birth-weight groups, and our results are consistent with this 
. The lack of an association between verbal IQ and FA is striking, particularly given that several of the tracts involved (notably the superior longitudinal fasciculus) are part of the anatomy underlying language function. Successful performance on the performance IQ subtests of the WASI requires an element of timed performance, with bimanual coordination and spatial reasoning (the block design subtest). This requires the coordinated action of several different brain areas, and is likely to be dependent on intact and functioning white matter. In a parallel study using DT-MRI tractography in this same group of VPT adults Kontis et al. (2009) 
showed that altered microstructure in the genu of the corpus callosum is associated with lower performance IQ. Andrews et al. (2010) 
found that reading skill in 11 year old preterm children was associated with FA of the genu and body of the corpus callosum. The pattern of association of IQ with multiple white matter regions or tracts has face-validity, since IQ is a composite of multiple cognitive processes 
associated with the structure and function of several connected brain regions 
. Our results are consistent with the concept that the distributed neural networks underlying cognition are altered in VPT adults. Consistent results are reported by Mullen at al. (2011) 
, who demonstrated correlations between uncinate fasciculus FA a semantic language task, and between arcuate fasciculus FA and a phonological task. Like us, Mullen et al. (2011) 
found these relationships only in their preterm participants, and not in their term-born control group. They suggest that their findings indicate that neural networks are altered in preterm individuals, or that this represents delayed maturation in the preterm group relative to the control group. Our findings are also consistent with these explanations. It is of interest is that correlations with IQ and memory in our study were only found in regions of FA reduction in the VPT group, and not in regions where FA was increased. A similar pattern of association between white matter volume (not DT-MRI) and cognition was reported by Nosarti et al (2008) 
. This would be consistent with the suggestion that areas of increased FA in the VPT group represent compensatory changes. Lubsen et al. (2011) 
have suggested that neurodevelopmental sequelae of preterm birth are due to altered patterns of neural connectivity. Evidence from functional MRI studies 
indicates that VPT adolescents and adults have altered neural networks underlying a variety of cognitive domains. The functional connectivity study of Myers et al. (2010) 
is also consistent with this concept.
We acknowledge a number of possible limitations in the interpretation of these results. First, co-registration of low-resolution, high-contrast FA maps may give rise to mis-registration and partial volume artefacts in regions of high and low anisotropy, for example, around the ventricles. In order to minimise such artifacts we used a two-step registration process and a masking procedure which restricted analyses to core white matter regions; we also excluded VPT participants with ventriculomegaly, in whom any such issues are likely to be exacerbated. Second, resolving a cluster into component tracts by reference to anatomical atlases is compromised by the limited resolution of the parametric maps and by the limited white matter detail such atlases contain. Third, the correlation analyses that we report could be vulnerable to type I errors by virtue of the number of comparisons made. Fourth, there are systematic differences between our participant groups, including an age difference between VPT and term comparison groups. Since FA is known to change with age, this could have introduced bias into our results. We have attempted to adjust statistically for this possibility in all the analyses.
There is much still to discover about the lifespan development of white matter after premature birth. Longitudinal studies remain the best method of addressing this kind of question, although they are not easy to carry out, and require commitment from funding bodies and research institutions if they are to be sustained for the length of time required. A recently-published MRI atlas of neonatal brain development was able to image maturational changes in the neonatal period in normally developing babies 
. Such methodology could provide information about brain development after preterm birth, and may be able to identify plausible sensitive developmental periods during which to target therapeutic interventions. The interaction between brain lesions, development and social and economic factors are not yet well studied. Nagy et al. (2009) 
found milder-than-expected brain abnormalities in very preterm adolescents born in the late 1980s and early 1990s, and speculate that such factors may be important in changing brain structural outcomes. New imaging techniques may also prove informative. For example, Driven Equilibrium Single Pulse Observation of T1 and T2 (quaintly known as “DESPOT”) can now be used to estimate characteristics such as myelination 
. Such emerging techniques have the potential to tell us more about the pathological and developmental processes affecting white matter after VPT birth.
VPT young adults have widespread alterations of fractional anisotropy, which are related to gestational age and birth weight. Some microstructural alterations may represent plastic reorganisation of white matter as well as the effects of early brain lesions. White matter microstructure is associated with cognitive outcome.