Does brainstem size in extremely preterm infants predict later neurodevelopment?

A simple midsagittal measurement at term-equivalent age

The study asked whether the size of the brainstem, the lower part of the brain that regulates breathing, heart rate and motor coordination, carries information about later development in infants born extremely preterm, before 28 weeks of gestation. Researchers retrospectively analyzed infants born between 22+1 and 27+6 weeks' gestation and admitted to a level IV neonatal intensive care unit in Vienna between August 2017 and December 2022, keeping only those who had a routine cerebral MRI at term-equivalent age before discharge.

The measurement itself is deliberately plain. On a midsagittal T2-weighted MRI slice, the plane that cuts the brainstem lengthwise down the middle, the team outlined four compartments: the basis mesencephali and tectum mesencephali (midbrain regions in front of and behind the aqueduct of Sylvius), the pons, and the medulla oblongata. Each was measured as a two-dimensional area in square millimetres, and the four were summed into a total brainstem area. Scans were done without pharmacological sedation during natural sleep, and intra- and interrater reproducibility was classified as very good (at least 0.81).

This matters because the paper is not proposing a new scanning protocol or a complex segmentation pipeline. It is testing whether an area that a radiologist can trace on a standard clinical scan carries usable information, which is what would make it practical in a neonatal unit.

Reference values came from 242 infants without severe injury

Of 315 extremely preterm infants scanned during their hospital stay, 295 had a scan at term-equivalent age, defined here as 35+0/7 to 41+6/7 weeks' gestation. The reference values and z-scores were built only from the 242 infants without severe brain pathology, where severe means high-grade intraventricular hemorrhage, high-grade cerebellar hemorrhage, or cystic white matter injury.

Reference tables were split by gestational age at birth (22–23, 24–25 and 26–27 weeks) and by gestational age at the time of MRI (35+0/7–36+6/7, 37+0/7–38+6/7 and 39+0/7–41+6/7 weeks). For example, among infants born at 26–27 weeks and scanned at 37–38 weeks, total brainstem area had a mean of 372.3 mm² with a standard deviation of 31.5 mm², and the 10th to 90th percentile range ran from 336.5 to 412.1 mm². The paper also reports a cut-off for impaired brain growth set at −2 z-scores.

The z-score is the key unit to keep straight. It expresses how far one infant's brainstem area sits from the average of infants of the same gestational age at birth and same age at scanning, in units of standard deviation. A z-score of −0.5 is therefore not a fixed number of square millimetres; it means half a standard deviation below the matched reference group, and the same physical shortfall translates into different z-scores depending on which reference stratum applies.

Severe injury, especially high-grade IVH, tracks smaller brainstems

Brainstem z-scores differed significantly by grade of intraventricular hemorrhage, bleeding into the fluid-filled spaces of the brain (p < 0.001). Median z-scores were 0.13 in infants with no IVH (n = 201), 0.06 in those with low-grade IVH (n = 64), and −0.51 in those with high-grade IVH (n = 30). Infants with no IVH and low-grade IVH did not differ from each other (p = 0.083), but both groups measured significantly larger than the high-grade group (no IVH versus high-grade p < 0.001; low-grade versus high-grade p = 0.005).

The same pattern held when severe pathologies were pooled. Infants with high-grade IVH, high-grade cerebellar hemorrhage or cystic white matter injury had a median z-score of −0.38 versus 0.13 in the reference group (p < 0.001), a large effect size (Cohen's d = 0.96, 95% CI 0.53–1.15).

The stepwise pattern is the informative part: the gap appears at the severe end, not between no bleeding and mild bleeding. That is consistent with the authors' interpretation that brainstem size is sensitive to serious neonatal brain injury, but the design cannot show that the injury caused the smaller measurement.

Cognitive and motor scores rose with each standard deviation

Within the reference cohort of infants without severe injury, the team used multivariable regression to relate brainstem z-scores to Bayley-III scores at 2 years' corrected age, adjusting for gestational age at birth and at MRI, sex, birthweight z-score, sepsis, patent ductus arteriosus, necrotizing enterocolitis, high-grade retinopathy of prematurity, bronchopulmonary dysplasia, IVH grade, cerebellar hemorrhage and white matter injury.

Higher brainstem z-scores were independently associated with better cognitive scores (adjusted β = 3.56, 95% CI 0.44–6.69, p = 0.026) and better motor scores (adjusted β = 2.80, 95% CI 0.39–5.22, p = 0.023). Language showed a borderline association that did not reach the conventional threshold (adjusted β = 3.58, 95% CI −0.014 to 7.171, p = 0.050). The β values are in Bayley score points per one-unit increase in brainstem z-score, so the cognitive estimate corresponds to roughly 3.6 points per standard deviation of brainstem size.

When the same analyses were repeated in the full cohort including infants with severe injuries, the associations were stronger rather than weaker: cognitive adjusted β = 5.09 (95% CI 2.60–7.57), language 4.29 (95% CI 1.42–7.16) and motor 5.73 (95% CI 3.63–7.83). The authors present this as a robustness check, and it is worth reading as one: the direction survives the inclusion of the sickest infants, though the larger coefficients also reflect the wider spread of brainstem sizes once injured infants are included.

Only the pons stayed independently associated in regional analyses

Because the total area combines four compartments, the team also tested each one separately within the reference cohort. After multivariable adjustment, only the pons remained independently associated with cognitive outcomes (adjusted β = 4.65, 95% CI 1.09–8.21, p = 0.011) and motor outcomes (adjusted β = 3.09, 95% CI 0.49–5.69, p = 0.020).

The authors offer a measurement-based explanation rather than a purely biological one: the pons is the largest of the compartments, which allows more robust and reproducible tracing than the smaller midbrain and medulla regions. That reading fits the reference table, where pons means are around 160–176 mm² across strata while the basis mesencephali sits near 30–33 mm². A small structure with a proportionally larger tracing error is a plausible reason for a weaker independent signal, and the paper does not separate that explanation from a genuine regional difference in developmental relevance.

For language, the authors suggest that the trend-level result may reflect different developmental trajectories of language networks, which is an interpretation rather than a tested mechanism.

Lower birthweight, higher lactate and lower Apgar marked smaller brainstems

Among infants in the reference collective, the team compared those whose brainstem size fell below the 10th percentile with the rest. Three factors differed: birthweight z-score was lower (median −0.52 versus 0.05, p = 0.004), first neonatal lactate was higher (median 4.6 versus 3.8 mmol/L, p = 0.008), and the first-minute Apgar score was lower (median 7 versus 8). No difference appeared for the other neonatal and demographic parameters examined.

The authors read these as signs of fetal growth restriction and early hemodynamic stress, and conclude that both antenatal and early postnatal influences contribute to brainstem development. Notably, factors that had been linked to brainstem development in earlier work, such as infection and bronchopulmonary dysplasia, were not statistically associated with brainstem size in this cohort.

These comparisons describe which infants tended to have smaller brainstems; they do not establish that any of the three factors produced the difference.

An association at age 2, not a prediction rule

Several boundaries sit close to these results. The study is retrospective and single-center, so unmeasured confounders could contribute to the adjusted associations, and the authors state that the observational design precludes causal inference. The reference values apply to the gestational age and imaging windows studied, and the authors note that applicability outside those ranges remains uncertain. External validation in independent cohorts or imaging protocols is required before clinical implementation.

The measurement method is also a boundary in itself. Two-dimensional midsagittal areas are practical landmarks, but the authors acknowledge they may not fully capture the three-dimensional complexity of brainstem anatomy. The outcome is the Bayley-III at a single time point, 2 years' corrected age, so nothing here speaks to school-age cognition or motor function.

Follow-up data were available for 261 of 315 infants (82.9%), and the reference analyses rest on the 242 without severe injury, so the adjusted coefficients describe that subgroup rather than the whole extremely preterm population.

The clinical signal concentrates at the severe end

Read together, the results point to a specific priority rather than a general screening claim. The clearest separation in brainstem size was between infants with high-grade IVH and everyone else, with no meaningful difference between no IVH and low-grade IVH. The outcome associations, meanwhile, were estimated in infants who had no severe injury at all, which is where the authors argue that even subtle variation may carry functional relevance.

That combination defines what the measurement is currently good for: flagging brains that are visibly undergrown alongside severe injury, and adding a quantitative layer to the MRI assessment of infants whose scans look otherwise unremarkable. Turning a z-score into an individual prognosis would require the external validation and three-dimensional or multimodal imaging the authors call for, and the language result, which did not reach significance in the reference cohort, is the weakest of the three domains.

The pons is where the regional evidence is strongest, and the severe-injury comparison is where the effect size is largest. Those two findings, rather than the headline cognitive and motor coefficients, are the most defensible basis for deciding whether brainstem measurement belongs in a term-equivalent MRI report.

Sources used in this answer