Changing story of the dementia epidemic

The common perception that little progress has been made in stemming the expected tsunami of dementia in the U.S. may be incorrect. The tsunami is expected because the large baby boom generation—born 1946–1964—will soon reach the ages where the risks of dementia are highest. The population aged 80 years or above (80+) will double between 2025 and 2050, with the age 85+ population 2.5 times larger and the age 95+ population 3.0 times larger than today.¹ If age-specific dementia prevalence rates do not change over time, then the number of persons with dementia at these ages will be 2.0, 2.5, and 3.0 times larger, respectively, in just 25 years, justifying the tsunami metaphor.

Actually, the age-specific dementia prevalence rates declined by about two-thirds in the U.S. during 1984–2024.^2,3,4,5 The declines were unexpected and mostly ignored; they occurred at a relative rate of 2.5–3.0% per year over 40 years across three large nationally representative surveys designed to estimate such declines—with no slowdown at the end of the 40-year period. Dementia was defined as clinically significant cognitive impairment using algorithms developed specifically for each study; while the algorithms differed between studies they remained consistent within each study over time.^3,4,5

The declines in dementia prevalence rates in the studies cited above were accompanied by corresponding declines in dementia incidence rates not only in the U.S. but also in Sweden, Netherlands, United Kingdom, and France.^6,7,8 One U.S. study of 3,010 participants found that dementia incidence rates declined across successive birth cohorts, finishing 77% lower (95% confidence interval [CI]: 55–88%) for cohorts born 1932–1941 compared to those born 1902–1911.⁶ A meta-analysis of 49,202 individuals across seven population studies in the U.S. and Europe found that the incidence rates of dementia declined by 13% per calendar decade (95% CI: 7–19%) during 1988–2015.⁷ A systematic review of 14 studies (incl. six in Ref. 7) that examined changes in dementia incidence rates over time found that most, but not all, U.S. and European studies reported declines in incidence rates.⁸ In this systematic review, there were some populations that showed stable or even increasing prevalence rates such as in Japan or in African Americans in the U.S., suggesting the need to examine more racially diverse samples.

Age is the most important risk factor for late-life dementia and, barring effective prevention, the number of persons with dementia will rise as the population ages. Figure 1A displays the age-specific prevalence rates for dementia in the U.S. estimated for 1984 and 2004 using the National Long Term Care Survey (NLTCS),³ extrapolated to 2024 at an average relative rate of decline of 2.7% per year—midway between the 2.5% and 3.0% per year relative rates of decline estimated for 2000–2012 using the Health and Retirement Study (HRS)⁴ and 2011–2021 using the National Health and Aging Trends Study (NHATS)⁵—yielding a cumulative decline of 67% over 40 years (95% CI: 61–73%) in age-standardized prevalence rates at age 65+, assuming the effects of COVID-19 on dementia were mostly resolved by 2024 (though long-term effects may remain).

Figure 1. Declining U.S. Dementia Prevalence Rates, by Age, Year and Birth Cohort.

(A) Percent of the U.S. population with dementia by age in 1984 (blue line) versus 2004 (red line) based on the National Long Term Care Survey data. Data for 1984 and 2004 were obtained from Table 2.16 in Ref. 3. The figure extrapolates from 2004 to 2024 (purple dotted line) at the age-specific relative rates of decline in effect during 1984–2004. (B) Age-specific prevalence rates for 6 birth cohorts extracted from Panel (A) after interpolating 1984–2024 using age-specific relative rates of decline in effect during 1984–2004. This was done in two steps: (1) Generate prevalence data for 1989, 1994, and 1999 by interpolating Fig. 1A between 1984 and 2004; do the same for 2009, 2014, and 2019 by interpolating between 2004 and 2024. (2) The data for the 6 birth cohorts in Fig. 1B were then extracted from the diagonals of the age × calendar-year data table generated in step (1).

The age increases for 1984, 2004, and 2024 in Figure 1A reflect the joint effects of age and calendar year, not the increases actually experienced by any single individual or group of individuals born in a given calendar period (i.e., birth cohort). The monotonic progression in Figure 1B indicates that calendar-year declines will continue if younger birth cohorts continue to have lower age-specific prevalence rates than older birth cohorts, stopping only if younger birth cohorts stop having lower rates, at which point a gradual transition from declining to constant age-specific prevalence rates would be seen as the younger cohorts transited ages 65–69 to 95+.

We can only speculate as to why younger birth cohorts have lower age-specific prevalence rates than older birth cohorts. To our knowledge only one study has examined changes in attributable risk factors across birth cohorts and that study found declines in smoking and rising education levels to be possible contributors.⁸ The Lancet Commission identified 14 potential risk factors—11 at early/mid-life: education, hearing loss, LDL cholesterol, depression, head injury, physical inactivity, diabetes, smoking, hypertension, obesity, and alcohol consumption; and three at late life: social isolation, air pollution, and vision loss.⁹ Thus, given their timing relative to age 65, the early/mid-life risk factors (e.g., nutritional deficiencies and early life trauma during the Depression and World Wars, rising education levels, declining smoking, introduction of statins, wider use of antidepressants, and better control of hypertension) could account for most of the differences between birth cohorts. Additional factors may include increased use of adult vaccines, e.g., for pneumonia or Herpes Zoster. Other factors, however, increased dementia risk over time (e.g., obesity, diabetes, and oophorectomy) which may have dampened past declines and may even contribute to rising incidence rates in the future.

Limitations of these various studies include survey sampling biases, imprecise ascertainment of dementia and its subtypes, changes in diagnostic criteria over time, and inability to determine causality. Despite these limitations, the similarity of findings across U.S. and European studies suggests the 40-year decline in age-specific dementia prevalence rates in the U.S. was real and that it was mainly a birth-cohort phenomenon.^2–8

If the birth-cohort effect persists, then the future burden of dementia will be substantially lower than current projections. Given the long lead time between action of early/mid-life risk factors and onset of dementia, the answers to three questions become critical for developing policies to mitigate the future burden: What will be the true net increase in dementia in 2050 due to population aging after correcting for birth-cohort related changes in risk? What risk factor changes explain the differences between birth cohorts underlying the past 40-year decline in dementia prevalence? How do we create a global observatory to track future global trends and changes in attributable risks in diverse samples? In an era where cognitive skills are more crucial than ever, the time to act is now.