Abstract
Early neuroimaging may provide a surrogate marker for brain development and outcome after preterm birth. Tract-Based Spatial Statistics (TBSS) is an advanced Diffusion Tensor Image (DTI) analysis technique that is sensitive to the effects of prematurity and may provide a quantitative marker for neuroprotection following perinatal brain injury or preterm birth. Here, we test the sensitivity of TBSS to detect diffuse microstructural differences in the developing white matter of preterm infants at term-equivalent age by modelling a 'treatment' effect as a global increase in fractional anisotropy (FA). As proof of concept we compare these simulations to a real effect of increasing age at scan. 3-Tesla, 15-direction diffusion tensor imaging (DTI) was acquired from 90 preterm infants at term-equivalent age. Datasets were randomly assigned to 'treated' or 'untreated' groups of increasing size and voxel-wise increases in FA were used to simulate global treatment effects of increasing magnitude in all 'treated' maps. 'Treated' and 'untreated' FA maps were compared using TBSS. Predictions from simulated data were then compared to exemplar TBSS group comparisons based on increasing postmenstrual age at scan. TBSS proved sensitive to global differences in FA within a clinically relevant range, even in relatively small group sizes, and simulated data were shown to predict well a true biological effect of increasing age on white matter development. These data confirm that TBSS is a sensitive tool for detecting global group-wise differences in FA in this population.
Introduction
Preterm infants are at high risk for lifelong cognitive and motor disabilities due to disruptions in brain development. While researchers are developing new neuroprotective treatments, evaluating their success is difficult. Waiting years to see if a child meets developmental milestones is slow and inefficient for clinical trials. Scientists need "surrogate markers"—biological signals visible on brain scans—that can tell them early on if a treatment is working. Diffusion Tensor Imaging (DTI) and specifically Tract-Based Spatial Statistics (TBSS) are often used to measure white matter integrity via Fractional Anisotropy (FA). However, it wasn't previously clear exactly how sensitive TBSS is to the subtle, diffuse changes a drug or intervention might produce in the neonatal brain, especially when sample sizes are small.
Methods
The study utilized 3-Tesla DTI scans from 90 preterm infants (born <36 weeks gestation) recruited from Queen Charlotte’s and Chelsea Hospital, London. All infants were scanned at term-equivalent age. To evaluate TBSS sensitivity, the researchers employed a simulation-first approach.
Datasets were randomly assigned to 'treated' or 'untreated' groups. Treatment effects were simulated by artificially increasing FA values in the 'treated' group's maps by 1% to 20%. The team tested different group sizes (n = 5 to 45) and modeled biological variation using Gaussian distributions to mimic real-world treatment responses. Finally, they validated these simulations by comparing them to the natural increase in FA that occurs as an infant ages (comparing infants scanned at 38 weeks vs. 44 weeks).
Key Results
The analysis demonstrated that TBSS is a highly sensitive tool for detecting widespread changes in white matter.
- The 5% Threshold: For group sizes ≥ 20, a global increase in FA of just 5% resulted in significant differences in over 50% of the white matter skeleton. This is a critical finding as 5% is within the range of changes expected in clinical interventions.
- Sample Size and Sensitivity: Even with only 10 infants per group, TBSS could detect a 10% change in FA. However, the coefficient of variation was significantly higher in groups smaller than 20, suggesting that larger cohorts are necessary for stable, reproducible results.
- Simulation vs. Reality: The simulated data accurately predicted real biological effects. The observed 4.14% FA increase between younger and older infants matched the curve generated by the 4% simulation model.
- False Positives: When comparing groups with no simulated changes, a 10% false-positive rate for small regional differences was observed, cautioning against over-interpreting tiny clusters in very small studies.
Discussion and Limitations
This study provides a "green light" for using TBSS in early-phase neonatal clinical trials, offering a roadmap for researchers to estimate necessary sample sizes. However, some limitations remain. The study primarily modeled global effects, whereas real-world treatments might affect specific brain regions. Additionally, the simulation used linear voxel-wise increases and did not account for potential cross-scanner noise variations. Future work should explore non-linear treatment effects and more complex regional variations to further refine the sensitivity profile of TBSS in neonatal research.
Summary updated