The main findings of this study are as follows: (1) an increased mortality rate was observed in elderly patients with AD, with HR of 2.19 after controlling for covariates, including cognitive function and disability; (2) the incidence rates of death per 1000 person-years were 50.5 and 12.9 for the AD and MCI groups, respectively, and the mean survival durations were 5.5 and 6.4 years for the AD and MCI groups, respectively; (3) NPS increased death risk in both the MCI and AD groups by 2.32- and 2.60-fold, respectively; and (4) among all the domains of NPS, mood domain, but not psychosis or frontal domain, increased mortality risk.
The mean estimate of survival in AD was 5.5 years in this study conducted in Asia. Overall, shortened life expectancy has been noted after first diagnosis at different ages and ranged from approximately 10 years at onset of 60 years to approximately 3 years at onset of 95 years [2]. Our finding of 5.5 years for AD in this population with a mean age of 78 years was compatible with the aforementioned observation. The literature has demonstrated several risk factors for mortality in AD, including age, sex, baseline cognitive function, and functional disabilities [30, 31], which were also noted in our study. Contrary to the general population, with cancer being the most common cause of death, our findings suggested that patients with dementia commonly die of cardiovascular or respiratory diseases, which is in line with a previous finding [32].
Mortality risk in MCI did not reach statistical difference in this study. Various definitions of MCI have been proposed. For instance, Peterson MCI Criteria emphasize the presence of subjective memory complaints and subclinical objective memory performance in the context of relatively intact everyday functioning [33]. Recent consensus criteria for mild neurocognitive disorder have been included in Diagnostic and Statistical Manual of Mental Disorders, Fifth Edition (DSM-5) [34], which specifies a modest decline in one or more cognitive domains but cognitive deficits that do not affect independence in everyday activities. Increased effort, compensatory strategies, or accommodation may be required to stay independent. In comparison, DSM-5 criteria emphasize objective cognitive impairment and subclinical functional impairment, whereas the Peterson criteria do not. DSM-5 MCI allows for greater compromise in functional independence than Peterson criteria do, indicating a severe form of MCI [35]. One study comparing the mortality risk of MCI using DSM-5 and Peterson criteria found that mortality risk was higher in DSM-5–diagnosed MCI than in Peterson criteria–diagnosed MCI [7]. As Peterson criteria are considered heterogeneous, a considerable proportion of patients do not develop dementia or may even return to normal cognition [36]. In this study, the case definition of MCI was according to Peterson MCI criteria, which is possibly one of the reasons for the low mortality rate in MCI.
However, we found shortened survival in the MCI group if NPS were considered. Mean survival period of 5.9 years in patients with MCI with significant NPS was similar to that of patients with AD without significant NPS. The relationship between NPS and mortality is complex. Our analyses were adjusted for age, sex, education, APOE, other medical conditions, baseline cognition, and function status, and even then, NPS affected mortality in both the AD and MCI groups. The findings indicate that factors beyond covariates mediate the relationships. A new phenotype, namely mild behavioral impairment (MBI), was proposed and defined as late-onset NPS in the context of predementia [37]. Elderly people with coexisting MBI and MCI have a higher risk of developing dementia than those with MCI alone [38] [39]. Furthermore, biomarker studies have suggested that patients with MBI share similar genetic profiles, brain beta-amyloid imaging, and plasma neurofilament light with those with AD [40-42]. In addition to conversion to AD, our findings highlight the clinical significance of NPS in MCI in terms of mortality and support the MBI concept, which allows for early identification and facilitates new possibilities for therapeutic intervention.
In addition to factors related to cognitive disorders per se, studies have suggested that antipsychotic treatment is one of the possible risk factors [43]. In our study, we added the precise variable of DDD of antipsychotics [29] to the survival analysis. DDD is used for calculating the assumed average maintenance dose per day for a drug used for its main indication. Using DDD allowed us to combine different antipsychotics to compare their mortality risk. However, the analysis did not reveal the mortality risk of DDD of antipsychotics (HR = 2.78, 95% CI = 0.66–11.69, p = 0.16). However, the risk of antipsychotic use in treating NPS for either AD or MCI could not be excluded as the DDD of antipsychotics was only at baseline and not over this study period.
An intriguing result was that the mood domain, but not psychosis or frontal domain, was associated with mortality. Mood domain of NPI-Q comprise anxiety, apathy, and dysphoria. Late-life depression could increase dementia risk [44, 45], and mood dysregulation is often the indicator of neurodegenerative diseases and progressive cognitive change [46]. Studies have suggested that affective symptoms and apathy are associated with early death in AD [47, 48]. Depression itself increases death risk [49]. Depression or apathy in dementia could aggravate failure to thrive and difficulty to manage chronic disease and may lead to social isolation [50, 51]. Therefore, these mood symptoms could accelerate cognitive and functional progression and lead to increased mortality risk. No association was noted between psychosis and survival time in this study, which is contrary to the finding of a study in community-dwelling elderly people [52]. A possible explanation for the disparity is the study population. This study was hospital-based; thus, treatment for psychosis was provided more actively than in the community. However, further studies are required to explain this discrepancy.
Limitations
Some limitations merit discussion. First, in this study, only baseline NPS were identified, but NPS may change over time. A further study using the NPS change trend will be beneficial. Second, as study participants were enrolled from teaching hospitals, the findings need to be extrapolated cautiously to other populations. Third, although DDD is a delicate measure of antipsychotics dose, we could not exclude the nonlinear association between drug dose and mortality risk. Finally, studies with large study samples are necessary in future to confirm the findings.