The impact of modifiable risk factor reduction on future dementia burden: a microsimulation modeling study

Age- and sex-specific dementia data were derived from participants of the Rotterdam Study, an ongoing population-based cohort study of 17,931 individuals aged > 40 years who reside in the Ommoord suburb of Rotterdam, the Netherlands. Details of the study design and of case ascertainment for dementia and mortality have been described previously [30, 31]. For the current study, we included all 10,209 participants who were alive and non-demented before the start of the fourth examination cycle on 1 January 2002. Information on vital status was obtained regularly from the municipal health authorities in Rotterdam. Participants were screened for dementia at each four-yearly center visit using the Mini-Mental State Examination and the Geriatric Mental Schedule organic level, and screen-positives underwent further investigation and informant interviews, including the Cambridge Examination for Mental Disorders of the Elderly. All participants also underwent routine cognitive assessment, including a verbal fluency test (animal categories), 15-word learning test, letter-digit substitution task, Stroop test, and Purdue pegboard task. In addition, they were under continuous surveillance for the detection of interval cases between center visits, through electronic linkage of the study base with medical records from general practitioners and the regional institute for outpatient mental healthcare. A consensus panel headed by a consultant neurologist established the final diagnosis in accordance with standard criteria for dementia (DSM-III-R). As such, 1,118 of 10,209 participants were diagnosed with dementia during follow-up until 1 January 2016.

In this study, we simulated the life histories of the 1980 Dutch birth cohort, using sex-specific life tables based on demographic data from Statistics Netherlands [24]. Furthermore, the life tables were adjusted for dementia-related mortality [32]. Dementia prevalence and overall life expectancy in the Netherlands lie in the middle of the European distribution [33, 34]. Compared to the United States, the Netherlands has slightly lower age-standardized dementia prevalence [35] and slightly higher life expectancy [34].

This analysis focuses on mid-life hypertension and late-life smoking as modifiable risk factors for dementia. Mid-life hypertension is defined as hypertension between the ages 40 and 59, and late-life smoking as being a current smoker after the age of 65. Prevalence data for hypertension and smoking was taken from Statistics Netherlands [26, 27]. The resulting prevalences are comparable to those of other Western European countries [36, 37]. Mid-life hypertension was adjusted for medication use [38], to best reflect unmedicated hypertension. Of note, fewer than one in seven people with hypertension in this age group use blood pressure lowering medication (Table 1). The co-occurrence of the risk factors was estimated based on data from the Rotterdam Study [31], where 24% of female and 31% of male late-life smokers also had mid-life hypertension. Risk-factor mortality and dementia risks were taken from systematic reviews for the respective age groups [4, 39, 40]. We assumed the effect of both risk factors to be additive. Table 1 summarizes the risk factor parameters.

For this analysis, MISCAN-Dementia was further developed to include mid-life hypertension and late-life smoking as risk factors for dementia. The simulated population was divided into four risk groups: none, smoking, hypertension, and both. To reflect the characteristics of these groups, we calibrated the prevalence, dementia and mortality risk of the risk factor groups as outlined in Table 1. Model calibration refers to the process of adjusting model parameters so that the simulation outputs closely match real-world, observed data (called calibration targets). The prevalence of the risk factors was calibrated for the respective age groups (hypertension 40–59 years, smoking 65–100 years). The prevalences in Table 1 were taken as calibration targets and the sizes of the risk groups were the calibrated parameters. For the dementia incidence, the baseline risk for the “none” risk group was the calibrated parameter, which was then multiplied by the relative risks of the risk groups. The calibration target was the overall dementia incidence equal to observed incidence rates from the Rotterdam Study [31], adjusted for a linearly declining trend as observed in a recent meta-analysis of North American and European cohort studies [28, 41]. For the mortality risk, the mortality rates were decreased for the “none” risk group and increased for the other risk groups. The calibration target was the overall mortality rate equal to the projected mortality rate of the 1980 birth cohort based on data of Statistics Netherlands [24]. To optimally capture sex differences in the risk factor characteristics and mortality, the calibration was performed separately for men and women. A genetic algorithm (DEAP [42]) was used for all calibrations. More details on the calibration can be found in Supplement A.

The reference scenario in the base case analysis simulates the risk factors using the point estimates outlined in Table 1. The risk factor reduction scenarios decrease the prevalence of the risk factors compared to the reference scenario, while maintaining the dementia and mortality risks of the risk factor groups. To study the maximum effect of reductions in smoking and hypertension on the future dementia burden, three extreme scenarios were modeled: complete elimination of (1) both risk factors, (2) hypertension only and (3) smoking only. As complete elimination of a risk factor is often not feasible in the population, we also modeled reductions by 50%, 25% and 10% for both risk factors as well as for each risk factor separately.

Each risk factor reduction scenario was run with 100 million simulated individuals born in 1980, to be able to model the effect of present-day interventions in mid-life on future disease burden. The analysis was performed for men and women separately due to their considerable differences in mortality and dementia incidence in general as well as risk factor characteristics. Every risk factor reduction scenario was evaluated in terms of incidence and prevalence, total dementia cases, and total life years with and without dementia. All outcomes were compared to the reference scenario of no reduction. Furthermore, we calculated the difference in dementia cases and life years with and without dementia by age. Since the reduction scenarios influence the size of the population in each age group (i.e. individuals live longer in the reduction scenarios and therefore more often are part of the older age groups), we also calculated the difference in dementia cases, life years with and without dementia relative to 100,000 life years lived by age.

In univariate sensitivity analyses, we varied several uncertain parameters to investigate their influence on the future dementia burden. We re-calibrated the model using the bounds of the 95% confidence intervals of the risk factor prevalence, excess dementia risk, and excess mortality risk of the risk factors (Table 1), resulting in twelve additional models for each sex. The same outcomes for every risk factor reduction scenario were calculated and compared to the main analysis.

In view of average life-expectancies and the larger impact of dementia on loss of quality-adjusted life years before age 85 [43], we performed a sub-group analysis of outcomes until age 85.

Complete elimination of smoking or hypertension led to fewer dementia cases. For women, the reductions were 1.4% for smoking, 1.1% for hypertension, and 2.5% for both. For men, the reductions were 2.5% for smoking, 3.3% for hypertension, and 6.2% for both (Table 2). The effect increases almost linearly with the magnitude of the reduction (Fig. 1). The complete elimination of both risk factors reduced the lifelong probability of developing dementia for an individual free of dementia at age 65 from 54.5 to 53.1% for women and from 35.5 to 33.3% for men (Supplemental Table C3). All reduction scenarios reduced the number of dementia cases in younger ages and then increased for the oldest old (Fig. 2, top panels).The shift from fewer to more dementia cases takes place at age 88 to 90 for women and at age 91 to 96 for men depending on the reduction scenario (Fig. 2, top panels). Whereas we found that absolute cases increased at older ages compared to the reference scenario, the relative rate adjusting for life years lived was reduced for each age and sex group (Fig. 2, bottom panels), except for women over 95 years (Fig. 2, bottom left panel). The complete elimination of both risk factors simultaneously has the largest impact (up to 188 fewer cases for women and 267 cases for men per 100,000 life years; Fig. 2, bottom panels). With a considerable gap, complete elimination of smoking, 50% reduction of both and complete elimination of hypertension have the second to fourth largest effect. To illustrate the size of the population to which these findings apply at various ages, Supplemental tables C1 and C2 provide the corresponding population sizes of all reduction scenarios.

All reduction scenarios increase the number of life years without dementia in both absolute terms (blue lines in Fig. 3, top panels) and relative to 100,000 life years lived in each age group (blue lines in Fig. 3, bottom panels). The peak in additional life years without dementia is reached around age 90 for both sexes. Similar to reductions in dementia cases, the shift from fewer to more life years with dementia takes place at ages 92 to 94 for women and at ages 92 to 98 for men depending on the reduction scenarios (red lines in Fig. 3, top panels). When considering the life years lived in each age group, there are fewer life years with dementia for both sexes across all ages (red lines in Fig. 3, bottom panels).

Of all parameters investigated in the sensitivity analysis, the magnitude of the risk factor dementia risk ratios had the largest effect on the results (Fig. 4). The percentage of potentially avoidable dementia cases decreased with a lower dementia risk associated with the risk factors and increased with a higher dementia risk. For hypertension, the percentage remained negative for the entire range of investigated excess dementia risk (1.6 RR, 95% CI 1.2, 2.2, Table 1), meaning also a lower dementia risk associated with hypertension resulted in fewer dementia cases (range − 0.56 to − 2.63% for women and − 1.35 to − 4.59% for men). However, with the lower and upper bounds of the dementia risk associated with smoking (95% CI 1.2–2.2, Table 1), there were more (+) dementia cases compared to the point estimate RR of 1.6 (range + 0.39 to − 3.74% for women and + 1.40 to − 4.77% for men). Smoking related mortality had the opposite effect; assuming a higher smoking related mortality, reducing smoking prevalence resulted in a lower percentage of potentially avoidable dementia cases. For men, reducing smoking prevalence under the assumption of higher smoking related mortality, even caused an increase (+) in dementia cases compared to the reference (range − 2.82% to + 0.19% for men and − 2.07% to − 0.84% for women). All other parameters had a smaller effect on the outcomes.

When restricting the analysis until age 85, the effect of the risk factor reductions becomes more favorable for all outcome measures, except for dementia-free life years (Supplemental Table B1). For example, the beneficial effect of complete risk factor elimination on the number of dementia cases amounts to 7.5% before age 85 for women, as compared to 2.5% across the entire age span (10.6% vs. 6.2% for men). Dementia-free life years are influenced to a lesser extent when excluding the oldest-old (0.58% all ages vs. 0.16% <85 for complete elimination of both risk factors for women).

Legend: The main analysis is presented as the middle bar (dashed). The analysis using the lower bound of the parameter is presented as the left bar and the analysis using the upper bound is presented as the right bar. The upper and lower bounds are given in Table 1.

The reduction of late-life smoking and mid-life hypertension results on a population level in a modestly lower incidence and prevalence of dementia, fewer dementia cases, fewer life years with dementia and more dementia-free life years. These interventions reduce the number of dementia cases and life years with dementia particularly before the age of 90, after which there is a small increase due to increasing life expectancy outweighing dementia incidence reductions. On balance, when considering the life years lived across age groups, reduction scenarios result in fewer dementia cases and life years with dementia.

Several studies estimated that reducing modifiable risk factors might prevent up to 30–50% of dementia cases [3, 4, 6‐10]. These estimates account for risk factor prevalence and dementia risk, but not for risk factor related mortality or the co-occurrence of multiple risk factors. The 2020 Lancet Commission report [4] estimated that the elimination of smoking could potentially prevent 5% of all dementia cases and the elimination of hypertension could potentially prevent 2%. However, the 2024 update [3] revised these estimates to 2% for both smoking and hypertension. This change may reflect the classification of smoking as a mid-life rather than a late-life risk factor in the 2024 report, as well as reduced relative risk estimates for both factors. In the current study, we found similar percentages as the 2024 Lancet Commission report. One prior study from France [44], which also accounted for risk factor mortality when simulating effects of mid-life hypertension elimination, found that the prevalence of dementia reduced by 2%, compared to around 4% in the current study, and reductions in lifetime risk of around 4% compared to around 2% in the current study. Some of the remaining variation between studies may be explained by differences in the modeling strategy and baseline dementia incidence.

While narratives around risk factor reduction often claim the prevention of dementia altogether, this framing can be misleading. For a multifactorial disease like dementia, it is important to acknowledge that risk factor interventions likely result in postponement rather than prevention of dementia [45, 46]. Consequently, such interventions may lead to a redistribution of the burden from younger to older age groups, yet still yielding substantial clinical and socioeconomic benefits [46]. Delaying the onset of dementia to older ages alleviates some strain on healthcare systems and caregiving networks, in particular if it leads to so-called “compression of morbidity”. Compression of morbidity refers to a reduction over time in the total lifetime spent with chronic disability [47], occurring when an intervention postpones the onset of disease by more than it increases lifespan. Consequently, in an aging population, life years with and without disease become more meaningful measures than absolute disease risk (i.e. disease incidence). If populations live long enough, eventually most will experience illness. The great success of public health and medicine has been in delaying the onset of illness further and further. It is therefore sensible to examine life years with and without dementia as well as present the analysis for individuals up to age 85. Examining life years with and without dementia shows that interventions shift the burden of dementia to older ages, as illustrated in Figs. 2 and 3 (top panels). Risk factor reductions result in better outcomes for individuals up to age 85 (Supplemental Table B1), since most life years with dementia and mortality gains occur in older age.

For all outcome measures, the effect of risk factor reductions on a population level was more pronounced in men than in women. This disparity is likely due to the higher prevalence of smoking among men, which provides more opportunities for risk factor reduction, and the longer life expectancy in women, which puts them at higher dementia risk in very old age [25]. On an individual and short-term basis, lifestyle interventions showed no significant differences between men and women [48, 49], and some studies reported larger effects for women [50]. These studies, however, featured balanced risk factor exposure between sexes and were generally conducted in age groups with low mortality (between 60 and 80 years). In addition, poor adherence to lifestyle interventions is widespread among men and women, particularly over the long-term [51]. Women tend to have higher adherence to antihypertensive treatments [52]. When translating these effects to the population level and considering risk factor prevalence and background mortality, men benefit more from risk factor reductions than women.

Primary prevention interventions targeting risk factors, particularly mid-life risk factors, require a long time horizon since dementia is particularly prevalent in older ages. For the 1980 birth cohort, as simulated in this study, hypertension interventions would take place in the 2020s–2030s and smoking interventions in 2040s, but the effect on dementia would not be fully visible until 2080. Given the long time horizon, the set up and evaluation of primary prevention interventions on a population scale is rarely done [54]. Modeling studies are useful to estimate the long-term effect of primary prevention interventions, and guide transgenerational, long-term policymaking.

MISCAN-Dementia synthesizes information on the natural history of dementia, birth cohort demographics as well as risk factor characteristics by using input from observational, population-based cohort data and peer-reviewed literature. Although we believe the model provides valid estimates of the effect of risk factor reduction on the future dementia burden, some limitations should be considered. First, the analysis relies on the assumption of causality between modifiable risk factors and the onset of dementia. Whilst such evidence is seldomly available from randomized trials, we believe our analysis represents the current best knowledge on the relationship between the risk factors and dementia and provides valid estimates for guiding preventive strategies. Second, our results are subject to uncertainties in the risk estimation and modeling approach, which we attempted to address by performing sensitivity analyses of the most uncertain model parameters. Risk factor related dementia risk had the biggest impact, and more precise evidence from observational studies may strengthen future modeling approaches. Third, we assumed that reducing a risk factor lowers the dementia risk to the level of non-exposed individuals. Additionally, we did not account for adherence to lifestyle interventions, which is generally low for individual-level interventions - particularly over the long term [51]. While this is a simplification of reality, it illustrates the maximum potential effect of risk factor reductions in a population. Fourth, we could not directly model quality-of-life effects as these depend on many factors beyond dementia outcomes. As a proxy for quality-of-life impact, in the analysis by age and the sub-group analysis focusing on individuals aged 85 or younger, we found that the effect of risk factor reductions was more favorable for younger age groups. Fifth, the model was developed based on Dutch data. Although risk factor and dementia occurrence are fairly comparable to other high-income countries in Europe and North America [36, 37], this limits generalizability of the results to other populations worldwide.