Serial diffusion tensor imaging data were available for a total of 176 infants, 97 enrolled between April 2001 and March 2008 at UCSF and 79 enrolled from April 2006 to December 2008 at UBC. During the study period, at UCSF there were an additional 54 infants enrolled who either had only 1 scan (36 infants) or the diffusion tensor imaging data were motion degraded (18 infants) and they were not included in this analysis. At UBC there were an additional 27 infants enrolled that were not imaged serially and so they were not included in the analysis. At UCSF, 27% of the subjects received sedation. There were no differences in the clinical characteristics such as gestational age at birth, birth weight, and neonatal morbidities at either institution between those with serial diffusion tensor imaging data and those without.
There were no important differences in preterm newborns at each site in terms of gestational age at birth, number of infants born at less than 26 weeks, perinatal variables such as mode of delivery, treatment with antenatal steroids, or surfactant administration, or exposure to postnatal infection (). More infants in the UBC cohort had a diagnosis of chronic lung disease and patent ductus arteriosus. This may reflect differences in clinical practice, as at UCSF prophylactic indomethacin is administered to all newborns less than 28 weeks gestation. There was a difference between the sites in the timing of the second MRI, with those in the UBC cohort being imaged about 4 weeks later than those at UCSF. The presence of moderate to severe brain injury on the first scan was similar in neonates at both sites.
Clinical characteristics by site
When comparing premature newborns born at less than 26 weeks gestation with those born at or greater than 26 weeks, there was a significant increase in the number of days intubated (median of 38 days vs 1 day, p < 0.001), presence of a PDA (70% vs 28%, p < 0.001), NEC (28% vs 10%, p < 0.007), postnatal infection (74% vs 26%, p < 0.001) and chronic lung disease (80% vs 19%, p < 0.001). The timing of the scans and the proportion of those with moderate to severe injury (36% vs 31%, p = 0.6) on the first MRI were not different between these groups. More extremely premature infants had an abnormal neuromotor assessment at time of the second scan (70% vs 41%, p = 0.003).
Effect of Gestational Age at Birth on diffusion tensor imaging parameters
The results of our analyses are presented in and . To understand the effect of gestational age at birth and extreme premature birth on diffusion tensor imaging parameters we first demonstrated how the mean diffusivity and fractional anisotropy values change with each week increase in PMA. In white matter regions of interest, for each week increment in PMA at scan, mean diffusivity decreased by 0.021 mm2/sec per week, (95% CI -0.24 to -0.018, p=<0.0001) and fractional anisotropy increased by 0.008 per week (95%CI 0.007 to 0.009, p=<0.0001). Then we examined the effect of gestational age at the time of birth as a linear variable on the change in diffusion parameters in white and gray matter. In our primary model we accounted for the postmenstrual age at the time of scan, region of interest, an interaction term to account for scans at two sites, and for the presence of moderate to severe brain injury. There was no effect of gestational age at birth as a linear variable on the change in mean diffusivity in white matter or in gray matter mean diffusivity with increasing postmenstrual age (). In these multivariate models, the presence of moderate to severe brain injury was significantly associated with higher mean diffusivity (0.03 mm2/sec, 95% CI 0.003 to 0.058, p=0.03) and lower fractional anisotropy (-0.012, 95% CI -0.19 to -0.005, p=0.002). In gray matter regions, for each week increment in the PMA at scan mean diffusivity decreased by 0.018 mm2/sec (95%CI -0.02 to -0.016, p<0.001). Neither gestational age at birth nor the presence of moderate to severe MRI abnormalities had a significant effect on gray matter mean diffusivity.
Table 2 Association of gestational age (as a linear variable) at birth with the change in diffusion parameters in white and gray matter regions of interest in a multivariate regression model accounting for post-menstrual age (PMA) at scan, region of interest, (more ...)
Table 3 Association of birth <26 weeks with the change in diffusion parameters in white and gray matter regions of interest in a multivariate model including terms for post-menstrual age (PMA) at scan, region of interest, and an interaction term to account (more ...)
Effect of Extreme Premature Birth on diffusion tensor imaging Parameters
Next we examined the effect of extremely premature birth (birth at < 26 weeks gestation) on the diffusion parameters in the same primary model. In the model adjusting for postmenstrual age at time of scan, region of interest, an interaction term for site (UCSF or UBC) by region of interest, and the presence of moderate to severe brain injury on the first scan, birth at < 26 weeks gestation was significantly associated with lower white matter fractional anisotropy (-0.01, 95% CI -0.018 to -0.003, p=0.008) (). The effect of extremely premature birth on fractional anisotropy did not differ significantly by white matter regions of interest (p=0.2). In contrast, extremely premature birth had no effect on mean diffusivity values in the white or gray matter regions.
In order to determine the independent effect of extreme premature birth, we added the important markers of early neonatal illness most strongly associated with extreme premature birth to the model: days of mechanical ventilation, presence of a PDA and NEC (). Adding these variables to the model resulted in a loss of effect of extreme premature birth on the change in mean fractional anisotropy (-0.002, 95% CI - 0.02 to 0.004, p=0.53). Moderate to severe brain injury on the first scan continued to have a significant effect, resulting in a decrease in mean fractional anisotropy (-0.012, 95% CI -0.02 to -0.004, p=0.002). Adding post-natal infection and CLD to this model did not affect these associations of extremely premature birth or moderate to severe brain injury on fractional anisotropy values. In addition, there was no interaction between sex and extreme premature birth, p=0.6.
We then examined whether extremely premature birth was associated with different fractional anisotropy and mean diffusivity values at term-equivalent age. We repeated the multivariate regressions, limiting the analysis to scans occurring at >36 weeks postmenstrual age (N=120). In these analyses, extremely premature birth was not associated with significantly different white matter fractional anisotropy (-0.008, 95% CI -0.03 to 0.01, p=0.4) or mean diffusivity values (-0.002 mm2/sec, 95% CI -0.06 to 0.05 p=0.9).