In a recent article published in The Lancet Public Health, researchers analyzed the temporal trend in dementia incidence in England and Wales from 2002 to 2019.
Study: Dementia incidence trend in England and Wales, 2002–19, and projection for dementia burden to 2040: analysis of data from the English Longitudinal Study of Ageing. Image Credit: SewCreamStudio/Shutterstock.com
Background
The number of dementia cases is increasing worldwide, with implications for affected people, their families, social policy frameworks, and economies. Based on the predicted annual increase of 57%, the projected number of dementia cases in England and Wales will be as high as 1.2 million in 2040, incurring healthcare and social care costs worth billions of pounds (£).
Most previous studies completed in high-income countries, including the United States (US), United Kingdom (UK), the Netherlands, Sweden, and France, utilized data series ending in 2010 to estimate dementia incidence trends.
Studies have found an age-specific decline in dementia incidence in the 2000s, which could partly offset the effect of this upward trend. However, evidence on how the dementia incidence trend has evolved post-2010 is lacking, especially for the UK.
Evidence before the study
Researchers thoroughly searched the PubMed database for studies evaluating temporal trends in dementia incidence between January 1, 2010, and April 1, 2023. They used keywords such as Trend or Trends, Dementia or Cognitive Dysfunction, and Incidence.
They found that after 2010, studies estimating dementia incidence trends in the UK did not use population-based data and a consistent diagnostic definition.
Moreover, these studies had other methodological flaws, such as not accounting for potential bias introduced in dementia incident trends by under-ascertainment of dementia cases due to mortality risk between survey waves.
Study definitions and methodology
The present study used data from the English Longitudinal Study of Ageing (ELSA) to estimate dementia incidence trends in England and Wales from 2002-2019.
They also explored whether the temporal trend of dementia incidence across nine waves of ELSA spanning 17 years varied with sex, age, and educational status.
The panel data of ELSA represented people aged ≥50 who live in England. A consistent algorithmic case definition based on the Diagnostic and Statistical Manual of Mental Disorders, Fourth Edition (DSM-IV), was used to assess cognitive and functional impairment, which helped capture moderate to severe incident dementia.
For statistical analyses, the researchers created seven subcohorts with overlapping four-year-long follow-up periods between 2002 and 2019.
It comprised participants who did not have dementia at the start of the follow-up period. The follow-up culminated with a dementia diagnosis, the end of four years, or death, whichever came first.
The number of dementia cases divided by the number of person-years in each subcohort was the crude dementia incidence rates expressed per 1000 person-years, which the researchers further standardized by age, sex, and educational status.
Further, the team applied a Cox proportional hazards model aggregating data from all subcohorts to estimate the overall time trend. Additionally, they used a three-state Markov model to model the transitions between no-dementia, dementia, and mortality and estimate incident rates for sex, age, and calendar time separately.
In the Markov model, they incorporated an interaction term between the subgroup variables and calendar time and then conducted likelihood ratio tests to evaluate the statistical significance of these interactions.
In sensitivity analyses, the team used the UK's Office for National Statistics (ONS) mortality and population data from 2001 through 2018 to project mortality rates up to 2040 and their impact on the future dementia burden.
In these analyses, they made conservative and optimistic assumptions. In the first, they assumed mortality rates from 2018 would persist unchanged until 2040, and in the second, a log-linear association between mortality and calendar year.
Results
In total, 19,806 individuals were analyzed, with participants in later ELSA waves being older but exhibiting balanced sex distribution across waves. ELSA data linked to the mortality register across wave one to wave nine resulted in 90,073 person observations.
The results pointed towards a non-linear trend in dementia incidence from 2008 onwards.
The age- and sex-standardized dementia rates plummeted between 2002 and 2010, from 10·7 to 8·6 per 1,000 person-years, but rebounded between 2010 to 2019, from 8·6 to 11·3 per 1,000 person-years.
This non-linear trend was consistent in subcohorts stratified by age and sex but varied significantly by educational status (p=0·0086).
Both statistical models followed a similar non-linear trend in the dementia incidence rate. The age- and sex-adjusted dementia incidence rate decreased between 2002 and 2010 by 14·7% (hazard ratio [HR]=0·85) and surged again between 2010 and 2014 by 19·9% (HR=1·20) in the Cox model.
Likewise, in the Markov model, age- and sex-adjusted dementia incidence declined between 2002 and 2008 by 28·8% [incidence rate ratio (IRR)=0·71], followed by a 25·2% increase between 2008 and 2016 (IRR=1·25).
The Markov model estimated a 2·8% relative annual increase in dementia incidence after 2008. Assuming this upward trend continued after 2018, the projected number of dementia cases in 2040 was 1·7 million, i.e., 70% more cases than in 2018.
Even if dementia incidence rates became constant after 2018, the case numbers would have reached 1·3 million in 2040. Importantly, the dementia burden would have surged over time regardless of conservative or optimistic mortality trend scenario.
Conclusions
Overall, this longitudinal study found that the decline in dementia incidence followed a non-linear trend, with a sharper decline of 28.8% between 2002 and 2008, followed by a slower increase equivalent to 25·2% up to 2016 per a central estimate by a multistate Markow model.
If this upward trend continues in the future, the dementia burden will increase in England and Wales to 1.7 million by 2040, which will have societal implications, particularly for healthcare services. Thus, public health policies targeted to mitigate this trend are urgently needed.
Flaws in study design may limit detection of cognitive effects in diet trials