Highlight
– Lifetime and 10‑year absolute risk of incident mild cognitive impairment (MCI) and dementia increase monotonically with continuous amyloid PET centiloid values among individuals who are cognitively unimpaired at baseline.
– Centiloid value was the strongest single biological predictor of lifetime risk; however, starting age and APOE ε4 carriership materially influence absolute risk, and sex modifies cumulative risk estimates.
– Accounting for outcomes that occur out of study (through multistate hidden Markov modelling and external ascertainment) materially changes lifetime risk estimates; participants who left the study had higher observed dementia rates.
Background
Biomarker-based definitions of Alzheimer’s disease (AD) have shifted the field toward biological staging, with in vivo amyloid PET widely used to detect cerebral β‑amyloid deposition years before cognitive decline. Clinically relevant questions remain about how an abnormal amyloid PET signal translates into absolute risk of future cognitive impairment for an individual who is still cognitively unimpaired. Absolute risk estimates (both over a 10‑year horizon and across remaining lifetime) are important for counseling, planning, and weighing potential benefits and harms of emerging disease‑modifying therapies that target amyloid biology.
Study design
Jack et al. report a retrospective, longitudinal cohort analysis using data from the Mayo Clinic Study of Aging (Olmsted County, MN). The analysis included participants aged ≥50 years who were cognitively unimpaired at enrolment and who underwent amyloid PET quantified on the centiloid scale. The primary predictor was continuous amyloid PET centiloid value (biological AD severity). Additional predictors included starting age, sex, and APOE ε4 genotype. Outcomes were incident MCI, dementia, and death.
To estimate lifetime and 10‑year absolute risks, the authors applied multistate hidden Markov models that allowed transition among cognitive states (cognitively unimpaired → MCI → dementia) and accounted for competing risk of death. Crucially, the modelling also estimated outcomes that occurred out of study (for participants lost to follow‑up) to reduce bias from differential attrition.
Key findings
Population and follow‑up: Between Nov 29, 2004, and Dec 2, 2024, a total of 5158 participants who were cognitively unimpaired and 700 participants with MCI were included. The cognitively unimpaired sample was balanced by sex (51% women). The authors evaluated lifetime and 10‑year absolute risks across a range of centiloid values (example centiloids reported: 5, 25, 50, 75, 100) and stratified by sex and APOE ε4 carriership.
Lifetime risk of MCI and dementia
Lifetime risk estimates for MCI increased monotonically with higher centiloid values (p<0.0001). Centiloid was the predictor with the largest effect on lifetime risk among the variables assessed. Representative estimates for participants who were cognitively unimpaired at a starting age of 75 years are shown below (95% confidence intervals reported by the authors):
Male APOE ε4 carriers (starting age 75): lifetime risk of MCI
– Centiloid 5: 56.2% (95% CI 50.5–61.9)
– Centiloid 25: 60.2% (54.9–65.6)
– Centiloid 50: 71.0% (65.2–76.7)
– Centiloid 75: 75.2% (69.1–81.2)
– Centiloid 100: 76.5% (70.5–82.4)
Female APOE ε4 carriers (starting age 75): lifetime risk of MCI
– Centiloid 5: 68.9% (63.7–74.1)
– Centiloid 25: 71.3% (66.6–76.0)
– Centiloid 50: 77.6% (72.5–82.7)
– Centiloid 75: 81.2% (76.7–85.7)
– Centiloid 100: 83.8% (78.5–89.1)
Within each centiloid stratum, APOE ε4 carriers had consistently higher lifetime and 10‑year risks than non‑carriers (p<0.0001). Females showed higher lifetime risk than males for comparable centiloid and APOE status, particularly at older starting ages in these estimates.
10‑year absolute risk and effect of starting age
Biological AD severity (centiloid) predicted 10‑year absolute risk of MCI and dementia (p<0.0001), but starting age had an even more prominent effect on 10‑year risk estimates (p<0.0001). In other words, while high amyloid burden increases near‑term risk, advancing chronological age remains a dominant determinant of short‑term risk of clinical decline.
Out‑of‑study outcomes and attrition
Notably, the rate of incident dementia was approximately two times greater among participants who had left the study compared with those who remained actively followed. The modelling approach estimated out‑of‑study transitions to reduce downward bias in lifetime risk estimates that would result from selective attrition of higher‑risk individuals.
Expert commentary and interpretation
This analysis addresses a critical and practical question for clinicians and patients: what is the absolute likelihood that a cognitively unimpaired person with a given amyloid PET burden will develop MCI or dementia? By using continuous centiloid values rather than a binary positive/negative threshold, the authors provide a more granular view of risk across the biological continuum.
Clinical implications
– Risk communication: These absolute risk estimates provide tangible probabilities that can inform shared decision‑making around monitoring strategies, lifestyle interventions, and potential consideration of disease‑modifying therapies when clinically appropriate.
– Trial design and preventive strategies: Quantifying how risk rises across centiloid levels can help identify individuals at sufficiently high short‑term risk for enrollment in prevention trials, where event rates matter for feasibility and statistical power.
– Personalized counseling: The results demonstrate that amyloid burden must be interpreted alongside age, APOE genotype, and sex. For example, a 75‑year‑old female APOE ε4 carrier with a centiloid of 75 has a markedly higher lifetime risk than a younger non‑carrier with a low centiloid, underscoring the importance of individualized risk estimation.
Strengths
– Large, population‑based cohort with decades of follow‑up and standardized cognitive diagnoses from the Mayo Clinic Study of Aging.
– Use of continuous centiloid values aligns with efforts to harmonize amyloid PET quantification across radiotracers and centers.
– Sophisticated multistate hidden Markov modelling to account for transitions, competing risks, and out‑of‑study outcomes reduces bias from selective attrition.
Limitations and caveats
– Generalizability: The cohort is from Olmsted County, Minnesota; demographics and risk factor distributions may differ from other regions and populations, including under‑represented minorities.
– Unmeasured modifiers: Other biomarkers (tau PET, neurodegeneration measures), vascular risk, education, and lifestyle factors also affect progression risk but were not the primary predictors here.
– Observational nature: Absolute risks are estimated from natural history data; the impact of future therapeutic interventions that alter amyloid burden on these absolute risks remains an open question.
– PET centiloid measurement variability: While centiloid harmonization reduces inter‑tracer differences, scanner and processing pipeline variability can still affect individual centiloid values; clinicians should interpret single measurements cautiously and in context.
Conclusion
Jack et al. provide robust, clinically actionable estimates showing that increasing amyloid PET centiloid values are associated with monotonically higher lifetime and 10‑year absolute risks of MCI and dementia among people who are currently cognitively unimpaired. APOE ε4 carriership, sex, and starting age modify these risks, with starting age exerting a particularly large influence on short‑term (10‑year) probabilities. Importantly, incorporating out‑of‑study outcomes into modelling is essential to avoid underestimating lifetime risk.
These findings inform clinical conversations about prognosis and risk stratification, and they offer quantitative inputs for the design and interpretation of prevention trials. Future work that integrates tau and neurodegeneration biomarkers, diverse populations, and the effects of interventions will be needed to refine individualized risk prediction further.
Funding and registration
This work was supported by the US National Institutes of Health, the GHR Foundation, and the Alexander Family. The primary report: Jack CR Jr et al. Lancet Neurol. 2025;24(12):1016–1025. doi:10.1016/S1474-4422(25)00350-3. PMID: 41240917.
References
1. Jack CR Jr, Hu M, Wiste HJ, Knopman DS, Vemuri P, Graff‑Radford J, et al. Lifetime and 10‑year absolute risk of cognitive impairment in relation to amyloid PET severity: a retrospective, longitudinal cohort study. Lancet Neurol. 2025 Dec;24(12):1016‑1025. doi:10.1016/S1474-4422(25)00350-3.
2. Jack CR Jr, Bennett DA, Blennow K, Carrillo MC, Dunn B, Haeberlein SB, et al. NIA‑AA Research Framework: Toward a biological definition of Alzheimer’s disease. Alzheimers Dement. 2018 Apr;14(4):535‑562. doi:10.1016/j.jalz.2018.02.018.
AI thumbnail prompt
Realistic photo of an older adult (age 60–80) seated at a hospital computer workstation viewing a brain PET image with a colored heatmap showing progressive amyloid uptake (blue to red). Soft clinical lighting, neutral clinic background, physician or technician silhouette blurred in the background, composition evokes informed concern and scientific clarity.
