A brain scan taken in midlife carried information about how blood markers changed over the following decade. In 224 adults who returned for a second blood draw 7 to 15 years later, those with higher scores on an MRI-based aging measure tended to show greater changes on a set of blood markers associated with aging. The link was modest and was observed across a group of participants.
Rebecca Reed and colleagues at the University of Pittsburgh and the University of Virginia used a measure called DunedinPACNI, short for Dunedin Pace of Aging Calculated from Neuroimaging. The peer-reviewed report in npj Aging tested whether the score relates to changes across several body systems beyond the brain. The researchers looked at blood samples taken at two points in time, giving them a way to compare the baseline MRI score with subsequent changes in physiology.
What the MRI score estimates
Chronological age is the time since birth. Biological aging describes gradual changes in how the body’s systems function, and those changes can vary among people born in the same year. Structural MRI shows features such as the thickness and shape of the brain’s outer layer. A conventional brain-age model uses such features to estimate a person’s age, then compares the estimate with their actual age. Earlier coverage in The Lifespan Brief examined how that age gap related to mortality in UK Biobank.
DunedinPACNI was trained for a different task. Its developers used brain scans collected at age 45 in the Dunedin Study in New Zealand. They trained the model to estimate a pace of change derived from 19 measures of physiology collected repeatedly from young adulthood to midlife. Those measures tracked functions including lung capacity, cardiovascular health and metabolism. Reed’s team applied the existing model to structural scans from two Pittsburgh-area cohorts.
What changed over the next decade
At baseline, the researchers examined 451 adults in the Adult Health and Behavior study and 295 in the Neurobiology of Adult Health study. The groups averaged about 43 and 42 years of age, respectively. Participants provided blood samples and had structural MRI scans. The researchers combined eight blood markers in the first cohort and six in the second. The panels included indicators of inflammation and metabolism, such as C-reactive protein and insulin. The first cohort also included cystatin C, a marker of kidney function. Most of the blood markers differed from the measures used to train DunedinPACNI.
Higher MRI scores were associated with higher blood-based aging scores at baseline in both cohorts after adjustment for age and sex. The standardized estimates were 0.13 in the first cohort and 0.44 in the second. The authors analysed the cohorts separately because they differed in recruitment, screening, scan equipment and the blood markers included. Those differences make it hard to know why the association was stronger in one group.
The longer-term result comes from the 224 participants in the first cohort who returned for a second blood draw, on average 10.5 years later. After accounting for age, sex and time between blood draws, a one-standard-deviation higher MRI score corresponded to a 0.22-standard-deviation greater change in the blood-based score. The 95% confidence interval was 0.09 to 0.36, with a p value of 0.001. The unit is a statistical measure of variation in the sample. It does not translate into a count of years gained or lost.
The association with follow-up blood markers persisted when the team accounted for each of three conventional brain-age measures. The conventional measures themselves had no statistically significant association with change in the blood score in that analysis. In one of the baseline comparisons, adjusting for a conventional brain-age measure weakened the link with DunedinPACNI enough that it was no longer statistically significant.
How far the finding travels
Both cohorts came from the Pittsburgh area, and only one supplied follow-up data for this analysis. The people who returned were older at baseline than those who did not, and the cohorts screened out people with some chronic conditions and medication use. Education was also associated with the MRI score in both cohorts. Differences in schooling and other social or health exposures could influence brain structure and blood measurements together.
The original DunedinPACNI study linked the score to later disease and mortality in other datasets. This follow-up study tracked a composite of blood markers. It did not test how accurately an MRI could forecast a particular person’s diagnosis or how clinicians should act on a score. The authors declared no competing interests and reported support from three US National Institutes of Health institutes.
An independent follow-up cohort could show whether the pattern recurs with different scanners, among people with a wider range of health conditions, and with other measures of biological aging such as DNA methylation. In this study, the evidence about change over time rests on one cohort with two blood draws.
Original research Reed and colleagues, npj Aging (2026). Open accepted manuscript (PDF).