The Gale readResearch

New dementia worldwide: 9.7 cases per 1,000 person-years, 22 cohorts

A pooled analysis of 38,418 older adults in 22 cohorts across 18 countries puts standardized dementia incidence near 1% a year, higher in women and in lower-income countries, with intervals wide enough that the gaps are less firm than they look.

By Gale Staff · October 16, 2026 · JAMA Neurology

Not reviewed by a clinician.

Illustration: Brain health & aging

The short answer

Pooling individual-level data from 22 community-based cohorts, researchers estimated that about 9.7 people per 1,000 person-years develop dementia worldwide once rates are standardized to the 2023 global age and sex mix (95% confidence interval 7.89 to 11.97). Across 287,632 person-years of follow-up, 3,001 of 38,418 participants developed dementia; the point estimates were higher in women (10.6) than in men (8.6) and in low- and middle-income countries (13.6) than in high-income countries (8.3), but each pair of confidence intervals overlaps. The 24.7 million new cases projected for 2050 assumes incidence stays constant, which is a scenario rather than a forecast of what will happen.

Who to see

Neurologist

New dementia rates were a little higher in women and in low- and middle-income countries, though the ranges overlap; a neurologist can check memory or thinking changes you notice.

Find care →

The question behind a single global number

A family in São Paulo, another in Seoul and a third in Lagos each hear the same word from a memory clinic, and each lives in a country with a different history of schooling, blood-pressure control, infectious disease and clinic access. A reasonable question follows the headline figure: how many people in the world start to develop dementia each year, and is the answer the same everywhere?

A paper in JAMA Neurology, published online on 12 October 2026 by Ashleigh Vella, Rachel Visontay, Darren Lipnicki and colleagues, tries to answer both halves. The authors say in their abstract why it has been hard: earlier global estimates underrepresent low- and middle-income countries and use methods that differ from one source to the next, which leaves the picture incomplete.

What the researchers did

The data come from the Cohort Studies of Memory in an International Consortium, known as COSMIC. The team included longitudinal, population-based cohorts in which dementia had been diagnosed clinically. Recruitment at baseline began between January 1990 and December 2011, follow-up ran through the end of 2021, and the analysis was carried out from June 2024 to June 2025. In all, 22 cohorts from 18 countries or territories on 6 continents contributed individual participant-level data for 38,418 people.

Using participant-level records instead of each study's published summary lets the analysts apply one method to every cohort. They counted person-years at risk, meaning the time each person was followed while still free of dementia, and used inverse-probability censoring weights, a technique that adjusts for people who left a study early in ways that could skew the count. Poisson models then produced incidence rates by age and sex. A two-step meta-analysis with random intercepts for country and for study combined the cohorts at the global level, by region and by country income level, and the final rates were standardized to the 2023 global age and sex population. Standardizing answers a specific question: what would the rate be if every place had the same age and sex makeup, so that a country with an older population does not look worse for that reason alone. The authors report following MOOSE guidelines for observational meta-analyses.

What they found

Over 287,632 person-years at risk, 3,001 participants developed dementia. The standardized global incidence was 9.7 per 1,000 person-years, with a 95% confidence interval of 7.89 to 11.97. Put in everyday terms, that is a little under 1 new case per 100 people per year in this standardized population. As a rough cross-check using only the two counts, 3,001 cases over 287,632 person-years is about 10.4 per 1,000 before any standardizing, which sits close to the adjusted figure.

Women had a higher age-standardized rate than men: 10.6 per 1,000 person-years (95% CI 8.61 to 13.11) against 8.6 (95% CI 6.94 to 10.72). Standardized incidence was also higher in low- and middle-income countries, at 13.6 (95% CI 9.67 to 19.24), than in high-income countries, at 8.3 (95% CI 6.57 to 10.54).

The abstract describes both comparisons as higher, but in each case the two confidence intervals overlap. For women and men the ranges share a large part of their span, and the low- and middle-income interval stretches from 9.67 up to 19.24 while the high-income one runs from 6.57 to 10.54. Overlapping intervals do not by themselves show that there is no difference, and the abstract gives no formal test of the gaps. They do mean that the paired point estimates are better read as a direction than as a measured size, and the abstract does not say how many cohorts sit in each income group.

The caveat that governs the projection

The headline projection is that applying the age- and sex-specific rates to the projected 2050 population yields 24.7 million new cases that year. The abstract attaches a condition to it: incidence is assumed to stay constant over time. That is a calculation of what today's measured rates would imply for the projected 2050 population, not a prediction. The paper's own cohorts were recruited between 1990 and 2011 and followed to 2021, so the rates describe that period, not 2050. The abstract also does not give a present-day count of new cases to set beside the 24.7 million, so the size of the projected rise cannot be read from it.

The design carries further limits that follow from how it was built. The abstract describes the cohorts as retrospective and community-based, and each comes from the COSMIC consortium, so the pooled rate describes the cohorts that consortium holds. Eighteen countries or territories cover six continents, but they do not cover every population in each income group, and a pooled rate for low- and middle-income countries is a pooled rate for the places in this consortium.

Why it matters

Dementia services are planned from incidence, the flow of new cases, as much as from prevalence, the number of people living with the condition. A harmonized benchmark computed one way across many cohorts gives planners and researchers a common yardstick, and that is the use the authors point to: planning for prevention, diagnosis and memory-care services, particularly in underrepresented low- and middle-income settings.

The result is also a reminder of how thinly low- and middle-income countries have been represented in earlier global estimates. This analysis brings more of the rest of the world into one set of estimates, with the honest consequence that its intervals are wide. A narrower interval would need more cohorts, more follow-up or both.

What this study can't tell you

  • Whether the female excess is real in size: the abstract gives 10.6 per 1,000 person-years for women and 8.6 for men, but the 95% intervals (8.61 to 13.11 and 6.94 to 10.72) overlap and no formal test of the difference is reported there.
  • Whether low- and middle-income countries truly have higher incidence than high-income countries: 13.6 against 8.3 per 1,000 person-years is the point estimate, and the intervals (9.67 to 19.24 and 6.57 to 10.54) overlap between about 9.7 and 10.5.
  • What causes dementia or why rates differ by sex or income level: this is a descriptive incidence estimate, and the abstract does not examine education, blood pressure, infection, diagnosis practices or any other explanation.
  • What will happen by 2050: the 24.7 million new cases come from multiplying today's age- and sex-specific rates by the projected population and assume no change in incidence, so they say nothing about whether rates will rise or fall.
  • How many of the 38,418 participants or 3,001 cases came from low- and middle-income countries, and which 18 countries or territories contributed: the abstract does not break either down.
  • Incidence by dementia subtype or by individual age band: the abstract reports all-cause dementia and the age- and sex-standardized totals, not Alzheimer disease, vascular dementia or the rate at a given age.
  • Who funded the work and what the authors' competing interests are: neither appears in the abstract that was read for this story.

The Gale read

This is a careful attempt to put one method under a body of cohorts that had been counted in different ways, and its strongest claim is the modest one: a standardized global incidence of 9.7 per 1,000 person-years, with an interval the authors do not hide. The sex and income-group comparisons are reported as higher, which is accurate for the point estimates, but the confidence intervals overlap in both pairs, so they are suggestive rather than settled gaps. The projection to 24.7 million new cases in 2050 comes with its assumption stated in the abstract, constant incidence, and it is best read as arithmetic about a projected population rather than a forecast about disease. The consortium covers 18 countries or territories, which is a wide sample but still not a map of the world. What the paper offers is a comparable benchmark and a clearer view of where the evidence is thin. It does not explain why any group has higher or lower incidence, and it does not estimate what any individual's risk is.

Common questions

how many people get dementia each year worldwide

In this pooled analysis of 22 cohorts, the standardized incidence was 9.7 new cases per 1,000 person-years (95% CI 7.89 to 11.97), which is a little under 1 per 100 people per year in the standardized population. It rests on 3,001 cases among 38,418 participants followed for 287,632 person-years, and it comes from the countries that had suitable cohorts. The abstract does not give a worldwide count of cases for today.

is dementia more common in women than men

The estimated rate was higher in women, at 10.6 per 1,000 person-years, than in men, at 8.6. The 95% confidence intervals, 8.61 to 13.11 and 6.94 to 10.72, overlap, and the abstract does not report a formal test of the difference. It also does not say why a difference would exist.

is dementia more common in low income countries

The pooled estimate was 13.6 per 1,000 person-years in low- and middle-income countries against 8.3 in high-income countries. The intervals, 9.67 to 19.24 and 6.57 to 10.54, overlap, so the data point toward a higher rate without fixing its size. The abstract does not say how many of the cohorts came from each group.

how many people will get dementia by 2050

The paper projects 24.7 million new cases in 2050 by applying age- and sex-specific incidence to the projected 2050 population. It assumes incidence stays constant over time, so it reflects the projected population and not any change in rates. The abstract does not give a present-day figure to compare it with.

Sources

  1. 1.Vella AS, Visontay R, Lipnicki DM, Nichols E, Steinmetz JD, Vos T, et al. Global Incidence of All-Cause Dementia. JAMA Neurology. 2026 (published online 12 October). doi:10.1001/jamaneurol.2026.3540 link

Read the study →

1 sources, numbered by first appearance. General health information, not medical advice. AI-assisted editorial content — every citation independently verified. Editorial policy

More from Gale Staff

The Latest Health Research →

Published October 16, 2026 · Gale Staff

All Gale Staff analyses →