It is currently known that the circumstances that occur throughout adult life influence the risk of dementia during old age, including physical and subjective health and socioeconomic status (Sindi et al., Reference Sindi2020; Van Patten and Bellone, Reference Van Patten and Bellone2021). Changes in objective cognitive performance, in the context of adult development and normal cognitive aging, have also turned out to be relevant in understanding cognitive health and identifying risk profiles for cognitive decline. The literature shows through different longitudinal studies that memory decline is one of the greatest predictors of incident dementia. Using longitudinal data on episodic memory performance from five cohort studies with two to 15 years of follow-up, Lee et al. (Reference Lee2018) found two main memory trajectories, stables (constant or improved memory function) and decliners. The highest incident rates of dementia were observed in the oldest group, in those of Caribbean-Hispanics origins and among those Decliners with higher rates of memory declines. Using a composite cognitive test score based on performance in episodic memory, executive function, and global cognition over 18 years from the Chicago Health and Aging Project, Rajan et al. (Reference Rajan, Wilson, Weuve, Barnes and Evans2015) observed that average cognitive test scores were significantly lower among those participants who developed dementia. The tests of episodic memory included in the study (Immediate and Delayed Recall of the East Boston Story) significantly predicted incident dementia, although the decrease in OR estimates over time was greater than for global cognition and executive function. Using data from the Betula longitudinal cohort, Boraxbekk et al. (Reference Boraxbekk, Lundquist, Nordin, Nyberg, Nilsson and Adolfsson2015) found that age- and education-corrected performance on two free-recall episodic memory tests (recall of sentences with enactment and without enactment) significantly predicted dementia 10 years prior to clinical diagnosis. Free-recall performance also predicted dementia 11–22 years prior to diagnosis when controlling for education, but not when age was added to the model. In the study object of this commentary (Joseffson et al., Reference Josefsson, Sundström, Pudas, Nordin Adolfsson, Nyberg and Adolfsson2019), the authors study 1062 adults also from the Betula study, aged 45–80 years at baseline and followed over 23–28 years. In contrast with the cross-sectional classification of the memory performance in Boraxbekk et al. (Reference Boraxbekk, Lundquist, Nordin, Nyberg, Nilsson and Adolfsson2015), in this study three different trajectories of memory change (maintainers, averagers, and decliners) were established based on their longitudinal performance over 15–20 years on five episodic memory tests. Dementia and mortality status were compared in these three trajectories. Based on previous findings that free-recall performance is impaired decades before the onset of dementia, the authors hypothesize that episodic memory deficits are present, in form in different memory trajectories, years before the appearance of clinical dementia symptoms.
To study the risk of dementia and death, conceptualized as the two central vital events in the aging process, the authors developed multi-state models. In these models, the lifespan trajectories of the participants were modeled from a healthy state (“alive”) to an absorbing state in which individuals never exit once they enter (“dead”), but also considering transient states that alter their health status and their risk of dead, in this case “dementia.” Multi-state models are representations for health-related processes over time, applied in life sciences to represent the probability of an event through lifespan among a finite number of states, with the events or diagnostic transitions being the changes between states (Hougaard, Reference Hougaard1999; Meira-Machado et al., Reference Meira-Machado, de Uña-Alvarez, Cadarso-Suárez and Andersen2009). This class of models allows for a flexible approach that can model almost any kind of transition experienced during lifespan, characterizing the progression of an individual through a number of transient states until the absorbing state, also allowing to calculate transition rates and the impact in these rates of sociodemographic variables such as age, sex, and socioeconomic circumstances. The most frequently used multi-state models in health sciences are the Markov models, in which the transition rates are modeled to only depend on the previous state. The Markov models represent a simple approach from a probability point of view, allowing an easy understanding of transition probabilities, and simplifying the likelihood evaluation. Facal et al. (2015) used Markov models to model the transition probabilities in the Compostela Aging Study (CompAS) between normal cognitive aging, multi-domain amnestic mild cognitive impairment (MCI), single-domain amnestic MCI, non-amnestic MCI and dementia (absorbing state) in different time intervals, constructing six different mid-point time intervals according to the time between the baseline and the follow-up assessment (9–15, 12–18, 15–21, 18–24, 21–27, 23–30 months). Multi-domain amnestic MCI presented the lowest rates of transition to normal cognitive aging and the higher rates of conversion to dementia. Sanz-Blasco et al. (Reference Sanz-Blasco, Ruiz-Sanchez de Leon, Avila-Villanueva, Valenti-Soler, Gomez-Ramirez and Fernandez-Blazquez2022) studied transitions from MCI to normal cognitive aging also using Markov models, showing similar probability of transitions from MCI to normal cognitive aging and dementia, and different factors related to transitions to normal cognitive aging, including age, socioeconomic status, presence of apolipoprotein E, and absence of affective symptoms. Marioni et al. (Reference Marioni, Valenzuela, van den Hout and Brayne2012) studied the link between lifestyles, cognitive transitions, and death using longitudinal data from the Medical Research Council Cognitive Function and Ageing Study. Using the MMSE scores, they classified the participants in no impairment, slight impairment, and moderate/severe impairment. Higher educational attainment and a more complex mid-life occupation predicted positive transitions, from slight to no impairment, but also increased mortality from moderate/severe states. In contrast, using data from six independent longitudinal studies, Robitaille et al. (Reference Robitaille2018) found higher levels of educative attainment only significantly related with early transitions, from no cognitive impairment to mild impairment.
In the study of Joseffson et al. (Reference Josefsson, Sundström, Pudas, Nordin Adolfsson, Nyberg and Adolfsson2019), the participants started in a healthy state, and transitions were recorded to dementia, considering Alzheimer’s disease (AD) and vascular dementia (VD), and/or an absorbing state of death. The individuals could stay as healthy aging, transition to AD and VD or, having remained as healthy aging or developing dementia, transition to the final state of death. Stratified analyses were conducted for old (71–86 years) and old-old (87–107 years) participants. In the old subgroup, 74% of the average participants were non-demented and alive, 81% of the Maintainers, and 51% of the Decliners, whereas in the old-old subgroup the percentages only reached the 15% in the Average participants, 23% in the Maintainers, and 3 % in the decliners. In the old cohort, Decliners had a 6.5 increased risk of developing dementia compared to Averages, with no maintainers developing dementia. In the old-old cohort, decliners had a 4.0 increased risk of developing dementia compared to averages, and averages 2.6 times increased risk compared to maintainers. The memory trajectories were not significantly related with mortality.
Two different cognitive scores were used, the episodic memory score (EMS) obtained from the sum of five scores in episodic memory tests, and the MMSE score as a general index of cognitive performance. Linear mixed models showed differences in episodic memory measured through the EMS, but not using MMSE score as dependent variable. These results point to the importance of neuropsychological assessments for the detection of clinically relevant profiles of cognitive decline, opposite to the lack of sensibility of short, general measures of the cognitive status such as the MMSE to detect early changes in cognitive performance. Boraxbekk et al. (Reference Boraxbekk, Lundquist, Nordin, Nyberg, Nilsson and Adolfsson2015) also highlight the potential relevance of easy-to-apply neuropsychological tests, such as the episodic verbal memory tests, in the promotion of reliable, and early procedures of dementia detection. The most demanding tasks from a recall and retrieval point of view, such as free-recall verbal tests, would have a higher predictive value.
In Joseffson et al. (Reference Josefsson, Sundström, Pudas, Nordin Adolfsson, Nyberg and Adolfsson2019), the authors state that memory declines prior to dementia provide a critical window for intervention, with decliners according to their EMS showing significantly higher risk of incidence of dementia than averages and maintainers. These observations are currently in line with what is indicated by scientific organizations and health authorities regarding dementia prevention (Anstey et al., Reference Anstey2017; Livingston et al., Reference Livingston2020). Nevertheless, little is known in the literature about which kind of preventive interventions can be prescribed for individuals with MCI.
You have
Access
It is currently known that the circumstances that occur throughout adult life influence the risk of dementia during old age, including physical and subjective health and socioeconomic status (Sindi et al., Reference Sindi2020; Van Patten and Bellone, Reference Van Patten and Bellone2021). Changes in objective cognitive performance, in the context of adult development and normal cognitive aging, have also turned out to be relevant in understanding cognitive health and identifying risk profiles for cognitive decline. The literature shows through different longitudinal studies that memory decline is one of the greatest predictors of incident dementia. Using longitudinal data on episodic memory performance from five cohort studies with two to 15 years of follow-up, Lee et al. (Reference Lee2018) found two main memory trajectories, stables (constant or improved memory function) and decliners. The highest incident rates of dementia were observed in the oldest group, in those of Caribbean-Hispanics origins and among those Decliners with higher rates of memory declines. Using a composite cognitive test score based on performance in episodic memory, executive function, and global cognition over 18 years from the Chicago Health and Aging Project, Rajan et al. (Reference Rajan, Wilson, Weuve, Barnes and Evans2015) observed that average cognitive test scores were significantly lower among those participants who developed dementia. The tests of episodic memory included in the study (Immediate and Delayed Recall of the East Boston Story) significantly predicted incident dementia, although the decrease in OR estimates over time was greater than for global cognition and executive function. Using data from the Betula longitudinal cohort, Boraxbekk et al. (Reference Boraxbekk, Lundquist, Nordin, Nyberg, Nilsson and Adolfsson2015) found that age- and education-corrected performance on two free-recall episodic memory tests (recall of sentences with enactment and without enactment) significantly predicted dementia 10 years prior to clinical diagnosis. Free-recall performance also predicted dementia 11–22 years prior to diagnosis when controlling for education, but not when age was added to the model. In the study object of this commentary (Joseffson et al., Reference Josefsson, Sundström, Pudas, Nordin Adolfsson, Nyberg and Adolfsson2019), the authors study 1062 adults also from the Betula study, aged 45–80 years at baseline and followed over 23–28 years. In contrast with the cross-sectional classification of the memory performance in Boraxbekk et al. (Reference Boraxbekk, Lundquist, Nordin, Nyberg, Nilsson and Adolfsson2015), in this study three different trajectories of memory change (maintainers, averagers, and decliners) were established based on their longitudinal performance over 15–20 years on five episodic memory tests. Dementia and mortality status were compared in these three trajectories. Based on previous findings that free-recall performance is impaired decades before the onset of dementia, the authors hypothesize that episodic memory deficits are present, in form in different memory trajectories, years before the appearance of clinical dementia symptoms.
To study the risk of dementia and death, conceptualized as the two central vital events in the aging process, the authors developed multi-state models. In these models, the lifespan trajectories of the participants were modeled from a healthy state (“alive”) to an absorbing state in which individuals never exit once they enter (“dead”), but also considering transient states that alter their health status and their risk of dead, in this case “dementia.” Multi-state models are representations for health-related processes over time, applied in life sciences to represent the probability of an event through lifespan among a finite number of states, with the events or diagnostic transitions being the changes between states (Hougaard, Reference Hougaard1999; Meira-Machado et al., Reference Meira-Machado, de Uña-Alvarez, Cadarso-Suárez and Andersen2009). This class of models allows for a flexible approach that can model almost any kind of transition experienced during lifespan, characterizing the progression of an individual through a number of transient states until the absorbing state, also allowing to calculate transition rates and the impact in these rates of sociodemographic variables such as age, sex, and socioeconomic circumstances. The most frequently used multi-state models in health sciences are the Markov models, in which the transition rates are modeled to only depend on the previous state. The Markov models represent a simple approach from a probability point of view, allowing an easy understanding of transition probabilities, and simplifying the likelihood evaluation. Facal et al. (2015) used Markov models to model the transition probabilities in the Compostela Aging Study (CompAS) between normal cognitive aging, multi-domain amnestic mild cognitive impairment (MCI), single-domain amnestic MCI, non-amnestic MCI and dementia (absorbing state) in different time intervals, constructing six different mid-point time intervals according to the time between the baseline and the follow-up assessment (9–15, 12–18, 15–21, 18–24, 21–27, 23–30 months). Multi-domain amnestic MCI presented the lowest rates of transition to normal cognitive aging and the higher rates of conversion to dementia. Sanz-Blasco et al. (Reference Sanz-Blasco, Ruiz-Sanchez de Leon, Avila-Villanueva, Valenti-Soler, Gomez-Ramirez and Fernandez-Blazquez2022) studied transitions from MCI to normal cognitive aging also using Markov models, showing similar probability of transitions from MCI to normal cognitive aging and dementia, and different factors related to transitions to normal cognitive aging, including age, socioeconomic status, presence of apolipoprotein E, and absence of affective symptoms. Marioni et al. (Reference Marioni, Valenzuela, van den Hout and Brayne2012) studied the link between lifestyles, cognitive transitions, and death using longitudinal data from the Medical Research Council Cognitive Function and Ageing Study. Using the MMSE scores, they classified the participants in no impairment, slight impairment, and moderate/severe impairment. Higher educational attainment and a more complex mid-life occupation predicted positive transitions, from slight to no impairment, but also increased mortality from moderate/severe states. In contrast, using data from six independent longitudinal studies, Robitaille et al. (Reference Robitaille2018) found higher levels of educative attainment only significantly related with early transitions, from no cognitive impairment to mild impairment.
In the study of Joseffson et al. (Reference Josefsson, Sundström, Pudas, Nordin Adolfsson, Nyberg and Adolfsson2019), the participants started in a healthy state, and transitions were recorded to dementia, considering Alzheimer’s disease (AD) and vascular dementia (VD), and/or an absorbing state of death. The individuals could stay as healthy aging, transition to AD and VD or, having remained as healthy aging or developing dementia, transition to the final state of death. Stratified analyses were conducted for old (71–86 years) and old-old (87–107 years) participants. In the old subgroup, 74% of the average participants were non-demented and alive, 81% of the Maintainers, and 51% of the Decliners, whereas in the old-old subgroup the percentages only reached the 15% in the Average participants, 23% in the Maintainers, and 3 % in the decliners. In the old cohort, Decliners had a 6.5 increased risk of developing dementia compared to Averages, with no maintainers developing dementia. In the old-old cohort, decliners had a 4.0 increased risk of developing dementia compared to averages, and averages 2.6 times increased risk compared to maintainers. The memory trajectories were not significantly related with mortality.
Two different cognitive scores were used, the episodic memory score (EMS) obtained from the sum of five scores in episodic memory tests, and the MMSE score as a general index of cognitive performance. Linear mixed models showed differences in episodic memory measured through the EMS, but not using MMSE score as dependent variable. These results point to the importance of neuropsychological assessments for the detection of clinically relevant profiles of cognitive decline, opposite to the lack of sensibility of short, general measures of the cognitive status such as the MMSE to detect early changes in cognitive performance. Boraxbekk et al. (Reference Boraxbekk, Lundquist, Nordin, Nyberg, Nilsson and Adolfsson2015) also highlight the potential relevance of easy-to-apply neuropsychological tests, such as the episodic verbal memory tests, in the promotion of reliable, and early procedures of dementia detection. The most demanding tasks from a recall and retrieval point of view, such as free-recall verbal tests, would have a higher predictive value.
In Joseffson et al. (Reference Josefsson, Sundström, Pudas, Nordin Adolfsson, Nyberg and Adolfsson2019), the authors state that memory declines prior to dementia provide a critical window for intervention, with decliners according to their EMS showing significantly higher risk of incidence of dementia than averages and maintainers. These observations are currently in line with what is indicated by scientific organizations and health authorities regarding dementia prevention (Anstey et al., Reference Anstey2017; Livingston et al., Reference Livingston2020). Nevertheless, little is known in the literature about which kind of preventive interventions can be prescribed for individuals with MCI.
Conflict of interest
Authors declare that they have no conflicts of interest.