New research highlights the ages at which Alzheimer’s-related brain changes accelerate, offering critical clues about when screening may be most effective.
Study: Breakpoints in Alzheimer’s Disease Biomarkers and Cognition Across the Aging Spectrum: The Mayo Clinic Aging Study. Image credit: Orawan Pattarawimonchai/Shutterstock.com
A recent study published in Alzheimer’s and dementia investigated the specific ages at which Alzheimer’s disease biomarkers and cognitive measures show significant slope changes, providing insight into the timing of early pathological processes across the aging spectrum.
Molecular pathology and biomarker development in Alzheimer’s disease
Alzheimer’s disease (AD) is a progressive neurodegenerative disorder characterized by gradual cognitive decline, beginning with subtle memory loss and progressing to impairments in orientation, reasoning, language, and daily functioning. As the disease progresses, neuropsychiatric symptoms and loss of independence become increasingly common.
At the molecular level, AD is characterized by the accumulation of amyloid-beta plaques and neurofibrillary tangles composed of hyperphosphorylated tau protein, leading to extensive synaptic dysfunction, neuronal loss, and brain atrophy. These pathological features have catalyzed the development of biomarkers that directly quantify and stage AD pathology in vivo, thereby reshaping both clinical diagnostics and research protocols.
Blood-based biomarker (BBM) tests have become reliable, minimally invasive, and cost-effective tools for detecting molecular changes associated with amyloid, tau, and neurodegeneration, as well as for predicting cognitive decline. When combined with genetic, clinical, and demographic information, BBMs improve the accuracy of Alzheimer’s disease (AD) screening, guide advanced diagnostic procedures, and support personalized treatment strategies. BBM assays are now a standard component of preclinical AD trials, aiding both participant selection and ongoing disease monitoring.
However, most BBM research has used convenience samples or cohorts with above-average health, limiting generalizability and making it difficult to identify optimal screening windows for the broader population. Population-representative studies are needed to clarify how biomarker trajectories change with age and in different clinical backgrounds. Such data are essential to improve the timeliness, effectiveness, and equity of AD screening and intervention.
Identifying critical ages for AD-related screening and follow-up
Age-specific breakpoints identify periods of rapid biomarker change that may signal clinical relevance, helping to optimize screening and follow-up strategies. Biomarkers evaluated in this study include plasma Aβ42/40, p-tau181, GFAP (glial fibrillary acidic protein), NfL (neurofibular light chain), amyloid positron emission tomography (PET), tau PET, hippocampal tumor (adjusted for intracranial tumor). In a subset, additional plasma p-tau181, p-tau217 and their ratios to unphosphorylated tau proteins were analyzed using mass spectrometry.
Participants were drawn from the Mayo Clinic Study of Aging (MCSA), a population-based cohort designed to investigate cognitive decline and dementia risk among Minnesotans. Recruitment was random, using the Rochester Epidemiology Project to ensure a representative sample.
Each participant attended comprehensive clinical visits that included neuropsychological testing, physician evaluations, and age-appropriate blood draws. Neuroimaging procedures were performed on a subset of the cohort. The present analysis focuses on 2,082 subjects for whom blood-based biomarkers of AD in plasma (BBM) were available, which included subjects with cognitive impairment, subjects with mild cognitive impairment (MCI), and subjects with late-onset dementia. Demographic data, including age and gender, were self-reported.
Age-related patterns in biomarkers and cognition were analyzed using generalized additive models (GAMs) for smooth trends and breakpoint regression to identify key turning points. the cycle number was adjusted where necessary. Analyzes were focused on ages 45 to 90 years to avoid sparse data. As a sensitivity check, models were rerun on non-cognitively impaired subgroups using samples from the Quanterix and C2N biomarker platforms.
Cognitive decline and biomarker changes smallhow the age-related turning point occurs at the population level
The Quanterix sample included 2,082 participants (mean age: 71 years, 54% male). The C2N subsample included 462 participants (median age: 73 years, 54 % male), with 93 % cognitively unimpaired and 7.4 % with mild cognitive impairment (MCI).
Median global cognitive ability in the C2N subsample was 0.16, slightly lower than in the full cohort, although still within a generally unimpaired range. Hippocampal volume, amyloid PET SUVR, tau PET SUVR, and other plasma biomarkers were similar to those found in the full Quanterix cohort.
In the full Quanterix sample, plasma Aβ42/40, hippocampal volume, and global cognition decreased with age, while p-tau181, NfL, and GFAP increased, especially after age 70 years. Tau PET increased with age but did not show a clear break point.
In the C2N subsample, hippocampal volume and overall cognitive function decreased with age, with accelerated cognitive decline in older adults. p-tau181, NfL and GFAP increased more steeply after age 70 years, while amyloid and tau PET increased steadily. Plasma Aβ42/40 remained constant until about 75, increasing thereafter. For tau markers in the C2N subsample, p-tau217 and p-tau181 increased non-linearly with age, especially after age 72 years, while their ratio measures increased more gradually.
Breakpoint analysis in the full sample showed significant breakpoints for plasma Aβ42/40, GFAP, NfL, p-tau181, PET amyloid, hippocampal volume, and global cognition, with sharpest changes typically between ages 62–71. Aβ42/40 had an earlier turning point before age 50. Breakpoint models were stronger for NfL, GFAP and global cognition.
In the C2N subsample, breakpoints were found for plasma Aβ42/40, GFAP, NfL, and p-tau181, generally at older ages than in the full sample. No breakpoints were observed for hippocampal volume, global cognition, or PET amyloid. The NfL again showed the best model fit.
Among the plasma biomarkers unique to the C2N subsample, both p-tau217 and p-tau181 showed breakpoints at age 72.6 years, indicating steeper increases in late life. Aβ42/40 ratios did not show clear inflection points, and C2N-derived Aβ42/40 measurements did not show consistent breakpoint behavior across analyses.
It should be noted that the breakpoints identified in both the Quanterix and C2N groups were partly consistent across platforms, particularly for GFAP and NfL. Other markers, such as Aβ42/40, showed assay variability and cohort composition, and some breakpoints were not replicated across samples. Sensitivity analyzes of cognitively non-impaired participants showed that most biomarker breakpoints were similar to those in the full cohort, except that the NfL breakpoint occurred earlier. In the C2N subsample, most breakpoints remained stable, except for p-tau181 and p-tau217, which lost statistical support.
conclusions
This study shows that breakpoint modeling can identify age thresholds in AD biomarker trajectories, revealing key turning points, particularly for plasma GFAP, NfL and p tau markers, at approximately 68–72 years of age. These observed inflection points indicate a late middle age to early older age acceleration in population-level biomarker changes associated with neurodegeneration. The findings improve our understanding of the optimal timing for screening and follow-up strategies in Alzheimer’s disease.
Importantly, these breakpoint estimates do not imply a precise temporal sequence of disease progression or that biomarker changes occur in a fixed order within individuals. Age explains only a moderate proportion of the variability in biomarker levels, indicating that other factors, such as underlying pathology and comorbidities, also play a substantial role. These results are based on cross-sectional data and reflect population-level age associations rather than precise biological transition points within individuals or direct predictors of future cognitive decline.
However, interpretation of these results is limited by the cognitive and demographic composition of the cohort, underrepresentation of advanced dementia, and some missing data, which may limit generalizability and obscure later-stage associations.
Future research should validate the current findings in more diverse and advanced populations, incorporate newer biomarkers, and apply advanced statistical methods to optimize screening and staging.
