**Background:** Modest structural brain changes occur in children and young adults with type 1 diabetes (T1D), and by age 60 some individuals show cognitive decline and reduced gray matter volumes. It is unclear whether these changes represent accelerated brain aging or early Alzheimer disease (AD)-like neurodegeneration. This study used machine learning-derived MRI indices—SPARE-BA (brain age) and SPARE-AD (AD-like atrophy)—to differentiate these processes in middle- and older-aged adults with long-standing T1D.
**Methods:** This cohort study included 416 participants from the Diabetes Control and Complications Trial (DCCT)/Epidemiology of Diabetes Interventions and Complications (EDIC) study (median age 60 years, range 44-74; median diabetes duration 37 years, range 30-51) and 99 demographically similar controls without diabetes. MRI scans were performed on 3 Tesla scanners across 24 centers, with scanner effects harmonized using Combat-GAM. SPARE-BA was derived using support vector regression trained on 2,764 cognitively healthy individuals from the iSTAGING consortium. SPARE-AD was derived using a support vector machine trained on 256 amyloid-negative controls and 221 amyloid-positive AD participants from ADNI. Cognitive assessments included psychomotor/mental efficiency tests (verbal fluency, digit symbol substitution, trail making part B, grooved pegboard) and memory tests (logical memory immediate and delayed recall). Linear mixed models and linear regression models were used, with false discovery rate (FDR) < 0.05 considered significant.
**Key Results:** EDIC participants had significantly higher SPARE-BA values compared to controls, indicating approximately 6 additional years of brain aging (EDIC: β=6.16, SE=0.71; controls: β=1.04, SE=0.04; P<.001). SPARE-AD values were comparable between groups, with no significant difference. Regional atrophy in T1D was most pronounced in the bilateral planum temporale, bilateral superior occipital gyrus, right transverse temporal gyrus, bilateral thalamus, putamen, and pallidum. Most temporal lobe regions did not show significant between-group differences. Among EDIC participants, SPARE-BA and SPARE-AD were not associated with glycemic measures (HbA1c: SPARE-AD β=−0.01, SE=0.06, P=.86; SPARE-BA β=0.31, SE=0.38, P=.41) or diabetes-related complications. Higher BMI was associated with less brain age-related atrophy (SPARE-BA: β=−0.23, SE=0.09, P=.007) and less AD-like atrophy (SPARE-AD: β=−0.04, SE=0.01, P=.01). In multivariable models, BMI remained significant for both SPARE-BA (β=−0.32, SE=0.09, P=.001) and SPARE-AD (β=−0.04, SE=0.01, P=.01), and higher diastolic blood pressure was associated with greater SPARE-BA (β=0.18, SE=0.07, P=.01). Greater brain aging (SPARE-BA) was associated with lower psychomotor and mental efficiency among EDIC participants (β=−0.04, SE=0.01, P<.001). Greater SPARE-AD was associated with decreased psychomotor and mental efficiency (β=−0.17, SE=0.04, P<.001), immediate memory (β=−0.13, SE=0.04, P=.001), and delayed recall (β=−0.11, SE=0.05, P=.02). Among controls, only SPARE-BA was associated with delayed recall (β=−0.04, SE=0.02, P=.03).
**Clinical Implications:** This study provides evidence that T1D is associated with accelerated brain aging—equivalent to approximately 6 years—without early signs of AD-related neurodegeneration in middle- and older-aged adults. The pattern of regional atrophy (thalamus, putamen, frontal and temporal regions) differs from the typical AD signature. These findings suggest that cognitive decline in T1D may be driven by accelerated aging rather than AD pathology, though the overall differences were modest even after a mean of 38 years of diabetes. The lack of association with glycemic control or diabetes complications leaves the mechanism unclear. Limitations include a predominantly non-Hispanic White cohort and the relatively young age of the cohort for AD pathology assessment. Future studies should investigate whether these brain aging patterns progress or interact with AD pathology at older ages.