Abstract
Background:
Apathy is among the neuropsychiatric symptoms frequently observed in people with cognitive impairment. It has been postulated to be a potential predictor of conversion from mild cognitive impairment (MCI) to Alzheimer’s disease (AD).
Objective:
To detect conversion rates from MCI to AD, and to determine the effect of apathy on the progression to AD in patients with MCI enrolled in the Texas Alzheimer’s Research and Care Consortium (TARCC) cohort.
Methods:
Apathy was determined by a positive response to the respective item in the Neuropsychiatric Inventory –Questionnaire (NPI-Q) completed by family members or caregivers. The final dataset included 2,897 observations from 1,092 individuals with MCI at the baseline. Kaplan-Meier survival curves were estimated to provide indices of the probability of conversion to AD over time across all individuals as well as between those with and without apathy. Cox proportional hazards regression measured the hazard associated with apathy and several other predictors of interest.
Results:
Over a period of 8.21 years, 17.3% of individuals had conversion from MCI to AD (n = 190 of 1,092 total individuals) across observations. The median time-to-conversion across all participants was 6.41 years. Comparing individuals with apathy (n = 158) versus without apathy (n = 934), 36.1% and 14.2% had conversion to AD, respectively. The median time-to-conversion was 3.79 years for individuals with apathy and 6.83 years for individuals without apathy. Cox proportional hazards regression found significant effects of several predictors, including apathy, on time-to-conversion. Age and cognitive performance were found to moderate the relationship between apathy and time-to-conversion.
Conclusions:
Apathy is associated with progression from MCI to AD, suggesting that it might improve risk prediction and aid targeted intervention delivery.
INTRODUCTION
Mild cognitive impairment (MCI) is associated with a higher risk of progression to Alzheimer’s disease (AD) [1]. Despite heterogeneous findings, the reported annual conversion rates from MCI to all-cause dementia is generally between 11–33% within 2 years [2, 3], and somewhere between 10–15% from MCI to AD dementia [4, 5]. Although not all MCI patients will develop dementia [6, 7], it has been estimated that approximately 80% of MCI patients develop AD in a 6-year period [8–10]. Given the impact of AD, it is paramount to define parameters associated with distinct trajectories of MCI patients, especially identifying those who are more prone to develop AD, such as those with lower education level and amnestic profile carriers [11].
Neuropsychiatric symptoms (NPS) have been reported in 35% to 85% of adults with MCI [12–14]. NPS often present in the early clinical stages of AD and may precede the onset of measurable cognitive decline [15, 16]. Several reports have pointed out the possibility of using NPS as predictors of MCI conversion to dementia [17, 18]. Apathy is one of the most common NPS in patients with MCI, and often co-occur with other NPS, especially depression [19–21]. Apathy is defined as reduction of self-initiated or environment-stimulated goal-directed behaviors alongside emotional flattening [19]. The reported prevalence of apathy in MCI is between 10.7% and 44.8% [22–24]. Patients with apathy are more likely to have functional impairment in activities of daily living, independent of other factors such as age, cognitive function, and depression [25–27]. A number of studies have also supported apathy as risk factor for conversion from MCI to AD [28–32]. While many of these studies were community-based and sizeable, they did not specifically evaluate the role of moderating factors of the conversion from MCI to AD (e.g., age, MCI type). Moreover, these studies have been conducted in primarily non-Hispanic White populations, limiting the generalizability of the findings. As racial/ethnic differences have been reported in dementia presentation and progression, this is an important issue to be investigated [33, 34]. Interestingly, recent studies have shown that biomarkers reflecting molecular pathways implicated in AD are differently expressed in Hispanics and non-Hispanics with dementia, suggesting the influence of both biological (e.g., genetic, burden of comorbidities) and environmental (e.g., education, access to care) factors [35, 36].
Using subjects enrolled in the Texas Alzheimer’s Research and Care Consortium (TARCC) cohort with a significant percentage of Hispanics (∼50%), the objectives of this study were: 1) to explore the conversion rates of MCI to AD in this cohort; and 2) to determine the predictive role of apathy on the progression from MCI to AD by means of survival analysis exploring moderating factors. Our results demonstrated that apathy was strongly associated with the conversion from MCI to AD in this diverse population. While age and cognition moderated this association, non-Hispanics were more likely to progress to AD, highlighting a potential role for ethnicity.
MATERIALS AND METHODS
Study population, definitions, and clinical tools
This study analysis is based on the TARCC database. TARCC is a state funded longitudinal multi-site study involving 9 academic institutions in Texas with a cohort of older adults (≥50), comprised of healthy (non-cognitively impaired) subjects, MCI patients, and AD patients [37, 38]. Annually, TARCC participants underwent a standard examination that included medical assessment, neuropsychological testing, and a laboratory workup. As part of the clinical assessment, the Neuropsychiatry Inventory Questionnaire (NPI-Q) was completed by family members or caregivers. NPI-Q is a brief informant-based assessment of 12 NPS, including delusions, hallucinations, agitation/aggression, dysphoria/depression, anxiety, euphoria/elation, apathy/indifference, disinhibition, irritability/lability, aberrant motor behaviors, night-time behavioral disturbances, and appetite/eating disturbances [39, 40]. Participants also underwent assessments of cognition (Clinical Dementia Rating, CDR; and Mini-Mental State Examination, MMSE) [41, 42]. A complete description of the study is reported elsewhere [37, 43]. The TARCC study excluded any individual with a history of major psychiatric disorders. Institutional Review Board approval was obtained at each TARCC site and written informed consent was obtained from all participants and/or legally authorized representatives.
To investigate the role of apathy on MCI progression to AD, our analyses were restricted to TARCC participants diagnosed with MCI at any time point during the study. The diagnosis of MCI in TARCC study was established using Petersen’s criteria [44]. The diagnosis of possible or probable AD was established through a consensus panel using NINCDS-ADRDA criteria [45]. Presence of NPS was assessed using the NPI-Q. A positive response to the item ‘Apathy’ on NPI-Q, regardless its severity, was regarded as presence of apathy. Presence of depression among the TARCC population was assessed using the Geriatric Depression Scale (GDS). GDS scores range from zero-30 with higher scores indicating greater depressive symptomatology. Scores equal or above 10 were considered to indicate clinically significant depression [46].
Data analytic strategy
From an original set of 14,655 observations across N = 3,670 unique individuals, data cleaning first involved removing all observations where the individual was not diagnosed as MCI or AD. Data were further reduced to include only those individuals who had been diagnosed with MCI for at least one observation. Finally, observations following an individual’s first observed diagnosis of AD were removed. Thus, for each individual, data consisted of all MCI diagnosis observations over time up to their first AD observation (if any). This final data set included 2,897 observations from N = 1,092 individuals. All data cleaning and analysis was performed in the R statistical computing environment [47].
The time-to-conversion outcome was created by coercing data into person-period format, requiring three unique columns to be generated: time1, time2, and event. Each individual’s first observation was set to time1 = 0 and time2 = 1. The second observation per individual was then set to time1 = 1 and time2 to the number of days since the first observation was recorded. Subsequent observations carried this pattern forward until each individual’s final observation. The event column was then coded as “0” for all observations in the data except for those individuals that had a final observation indicating a diagnosis of AD, wherein that final observation was coded as “1.”
Kaplan-Meier survival curves were estimated to provide indices of the probability of conversion to AD over time across all individuals as well as between those with and without apathy. First, a univariate Cox proportional hazards model was used to fit time-to-conversion as a function of apathy (absent versus present). Potential confounders of the apathy-conversion relationship were tested via guidelines from the literature [48, 49], including handedness, race, level of independence, maritalstatus, and a paternal or maternal history of dementia. Any baseline demographic variable that separately demonstrated a relationship with both apathy and time-to-conversion met criteria as a potential confounder and included as covariates if they resulted in different inferences with respect to the effect of apathy on time-to-conversion. Cox regression coefficients were exponentiated to provide hazard ratios (HR). All Cox regression models included a term for cluster ID to provide robust estimates of standard errors. The proportional hazards assumption was evaluated via graphical analysis of Schoenfeld residual plots. The R package survival [50] was used to estimate Kaplan-Meier survival curves and perform Cox proportional hazards modeling.
Univariate Cox proportional hazards models also evaluated time-to-conversion as functions of key predictors of interest considered a priori to be relevant to the apathy-conversion relationship: age (years), MMSE score (0 to 30), sex (male versus female), ethnicity (Hispanic versus non-Hispanic), and MCI type (non-amnestic versus amnestic) and depression (absent versus present). Multiple predictor models were then fit to test the relative contribution of apathy (absent versus present) while adjusting for the others. A final model excluding depression was then evaluated, and follow-up models investigated potential moderating effects of age, MMSE score, sex, ethnicity, and MCI type on the relationship between apathy and time-to-conversion by testing interaction terms.
The potential moderating influence of each primary covariate (age, MMSE score, sex, ethnicity, and MCI type) on the apathy-conversion relationship was tested in separate models by including an interaction term between apathy and a given covariate while controlling for the constituent main effects, both with and without adjustment for the other predictors. These variables were chosen based on their reported role in the conversion from MCI to AD. Proportional hazards were supported in all models. Sex, ethnicity, and MCI type did not moderate the apathy-conversion relationship.
RESULTS
Sample characteristics
A total of 2,897 observations from 1,092 individuals were included in the present analysis. Across participants, mean average 2.65 (SD = 1.73) observations were included, with a range from 1 to 9 observations. The count of observations per participant was skewed, as lower counts were much more common than higher counts. The median observation count per participant was 2 (IQR = [1, 4]). Individuals were more often white (68.3%) and female (58.8%), with nearly equal representation across non-Hispanic and Hispanic ethnicities (50.1% and 49.9%, respectively). The mean average age and years of education were 71.8 (SD = 8.61) and 12.58 (SD = 4.38), respectively. Table 1 provides descriptive statistics for the sample and primary variables of interest.
Sociodemographic and clinical characteristics of TARCC sample with mild cognitive impairment (MCI, overall), and categorized by the presence of apathy and conversion from MCI to Alzheimer’s disease
*p-values for group differences in apathy were derived via multilevel logistic regression. **p-values for group differences in MCI to AD conversion were derived via Cox proportional hazards regression. ***Values for characteristics that may change (i.e., time-varying) are summarized by the final observation in the data (per person).
Kaplan-Meier survival curve
A Kaplan-Meier survival curve for time-to-conversion across all participants and between apathy groups is provided in Fig. 1. The x-axis describes the length of time (in years) to AD diagnosis, starting from a first MCI diagnosis in the data. The y-axis describes the probability of having not yet converted from MCI to AD. Over a period of 8.21 years, 17.3% of individuals had conversion from MCI to AD (n = 190 of 1,092 total individuals; 160 with probable AD, 30 with possible AD). The median time-to-conversion across all participants was 6.41 years. Between apathy present and absent groups, 36.1% and 14.2% of individuals had conversion, respectively. The median time-to-conversion was 3.79 years for apathy-present individuals and 6.83 years for apathy-absent individuals.
Kaplan-Meier survival curve. This plot displays the probability of converting from mild cognitive impairment (MCI) to Alzheimer’s disease (AD) across (a) the overall sample (solid line), (b) individuals without apathy at baseline (dashed line), and (c) individuals with apathy at baseline (dotted line). 95% confidence bands are provided in shaded regions around each line. Across the sample, the probability of not converting to AD from MCI fell to 50% around Year 6. However, the plot details the differences between apathy groups: those without took longer to reach the 50% probability mark, whereas those with apathy reached the 50% probability mark around 3.5 to 4 years.
Cox proportional hazards regression
Table 2 summarizes the results across all Cox proportional hazards regression models of time-to-conversion for each predictor of interest: age (years), MMSE score (0 to 30), sex (male versus female), ethnicity (non-Hispanic versus Hispanic), MCI type (non-amnestic versus amnestic), depression (absent versus present), and apathy (absent versus present). This table is organized into three sections: 1) the set of univariate models for each predictor of interest, 2) a model including all the predictors, and 3) a model with all the predictors except depression.
Cox proportional hazards regression model summary
The univariate models (Table 2) demonstrated that, without adjustment for covariates, each predictor of interest was significantly related to time-to-conversion (p < 0.05). Individuals who had apathy demonstrated a 2.40-fold greater hazard of conversion relative to those who did not have apathy. Higher hazards were also found for male sex (HR = 1.50), age (HR = 1.04; a 4% increase for each additional year of age), amnestic type MCI (HR = 1.63; 63% higher hazard than non-amnestic), and depression (HR = 1.63). As expected, higher MMSE scores were related to lower hazard of conversion (HR = 0.88; 12% lower for each additional point of MMSE). Hispanic ethnicity was also related to lower hazard of conversion (HR = 0.35; 65% lower hazard than non-Hispanic). The “full” multiple predictor model found that depression was not significant when adjusting for age, MMSE score, sex, ethnicity, MCI type, and apathy. The final multiple predictor model found that estimates were largely stable after removing depression. For each model, Schoenfeld residual plots supported proportional hazards, and the estimates reported in Table 2 reflect the average HR over time.
We have also run a model replacing the dichotomous apathy predictor with a four-level index of apathy severity (0 = absent or no apathy, 1 = mild, 2 = moderate, 3 = severe). This model evaluating the severity predictor as a set of dummy-coded comparators to non-apathy found that each level was significantly related to the outcome, with effects observed for mild apathy (HR = 2.10, p < 0.001), moderate apathy (HR = 2.31, p < 0.001), and severe apathy (HR = 2.51, p = 0.048).
Moderation models
Age demonstrated a significant moderating influence (p = 0.049 unadjusted; p = 0.002 adjusted). Figure 2 provides a plot of the adjusted interaction between age and apathy. Evaluating simple slopes demonstrated that for individuals who had apathy, the relationship between age and time-to-conversion was essentially null (HR = 0.99, p = 0.679), whereas for individuals who did not have apathy, there was a 5% increase in the hazard per additional year of age (HR = 1.05, p < 0.001).
Apathy by age interaction. Plot of group differences in the risk of converting from cognitive impairment (MCI) to Alzheimer’s disease (AD) as a function of age at baseline. For individuals who had apathy, there was no evidence for a relationship between age and time-to-conversion. For individuals who did not have apathy, the hazard was higher by 5% per year of age. The y-axis (risk score) describes the risk of an event (conversion) per unit of time at different levels of the predictor.
MMSE score also demonstrated a moderating influence (p < 0.001 for both adjusted and unadjusted models). Figure 3 provides a plot of the adjusted interaction between MMSE score and apathy. Simple slopes demonstrated that for each group, higher MMSE scores were related to lower hazards; however, this effect was stronger for individuals who had apathy (HR = 0.75, p < 0.001) compared to those who did not have apathy (HR = 0.87, p < 0.001). Evaluating the interaction within strata of MMSE scores provided the most insight; as shown in Fig. 3, the interaction was driven primarily by the disproportionately high risk of conversion for those individuals who had apathy and the lowest strata of MMSE scores. Finally, an exploratory follow-up model did not support the higher-order apathy by age by MMSE score 3-way interaction, and reducing this model to the constituent two-way interactions did not change inferences from the moderation models described above.
Apathy by Mini Mental Status Examination (MMSE) score interaction. Plot of group differences in the risk of converting from cognitive impairment (MCI) to Alzheimer’s disease (AD) as a function of MMSE score at baseline. Higher MMSE scores were related to lower hazard for each group, with a stronger effect observed for individuals who had apathy relative to those who did not. The y-axis (risk score) describes the risk of an event (conversion) per unit of time at different levels of the predictor.
DISCUSSION
Our results showed that MCI patients with apathy have significantly higher conversion rates to AD compared with MCI patients without apathy, even after adjustment for confounding factors. The presence of apathy in MCI patients was associated with 2.40-fold greater hazard of conversion relative to MCI patients without apathy, with a median time-to-conversion of 3.79 years for MCI with apathy as compared to 6.83 for non-apathy MCI group. Previous reports linked apathy in presence of depression as one of the risk factors for cognitive decline [28–32]. In our cohort, independent from depressive symptoms, apathy by itself was strongly associated with cognitive decline and increased conversion rates from MCI to AD.
Our study corroborates previous reports on the role played by apathy in the conversion from MCI to AD [13, 52]. Given the unique composition of TARCC cohort, our study confirms the generalizability of this finding to a diverse population including a significant percentage of Hispanics. Van Dalen and colleagues, for instance, found that apathy was associated with a 26% increased risk of dementia in over 3,000 community-dwelling older European adults aged 70–78 years old without dementia at baseline and after adjustment for age, sex, MMSE, disability and comorbidities [53]. More recently, in a 13-month prospective study involving 542 community dwelling adults, apathy was also found to be predictive factor for motoric cognitive risk syndrome (MCR) (a predementia syndrome characterized by the presence of cognitive complaints and slow gait in older individuals without dementia or mobility disability) [32, 54]. It is worth mentioning that apathy seems to predict worse cognitive and functional outcomes in other conditions as well, such as Parkinson’s disease and stroke [55]. Zhao and colleagues reported a prevalence of apathy in post-stroke MCI of 12.9% with a significant conversion rate to dementia of 17.2% for the apathy MCI group as compared to 3.4% in non-apathy MCI group [56].
Our results showed that age, amnestic type of MCI, non-Hispanic ethnicity, and lower cognitive performance, as assessed by MMSE, were significantly associated with MCI-AD conversion. Higher hazards were also found for male sex in the univariate analysis, but not in the multiple predictor model. In line with our findings, age, MCI type and cognition have been consistently reported as risk factors for MCI-AD conversion [57]. In contrast, women have greater risk for progression than men [58]. This contradiction might reflect sociodemographic specificities of the current cohort.
It is worth noticing that the role of ethnicity has been overlooked in the context of MCI-AD conversion. While the overall rate of neuropsychiatric burden seems to be higher in Hispanics compared to non-Hispanics, a recent study showed that a cluster of behaviors marked by apathy and agitation/irritability was a better predictor for the conversion from normal cognition to MCI in non-Hispanics than Hispanics, corroborating our findings [59]. Furthermore, this result reinforces the emerging recognition that racial/ethnic differences can influence AD presentation and progression [33, 59]. It remains to be determined the multiple factors and their relative weight on this complex interplay among sociocultural/environmental and biological factors.
Our results also indicated that age and cognitive performance moderated the association between apathy and MCI-AD conversion. This is not surprising given that apathy has been closely related to aging and cognitive functioning— major risk factors for AD [21]. In this context, apathy might be understood as a behavioral expression of the underlying pathophysiological process and not necessarily playing a mechanistic role in the conversion from MCI to AD.
Studies have also investigated the role of other NPS in the conversion from MCI to dementia, with mixed findings. In a community-based study, Liew and colleagues showed that people with MCI and anxiety and depression had a high risk of conversion to dementia [60]. In contrast, Roberto et al. showed that apathy was a better predictor for MCI conversion as compared to depression and anxiety [61]. In our study, after adjusting for confounding variables, depression alone was not significantly associated with MCI-AD conversion. The mixed results regarding the role of depression on MCI-AD conversion deserves further investigation taking into account the diversity of sample profiles, including ethnic, social, medical (e.g., cardiovascular), among other factors.
Our study has limitations that must be acknowledged. Despite the fairly large sample, most patients were followed up for only 3 years, which may not be sufficient time to determine full conversion rates. The diagnosis of AD was based solely on clinical basis and lacked neuroimaging and cerebrospinal fluid biomarkers of the disease. Therefore, it is not possible to completely rule out other causes of dementia, including vascular dementia and other neurodegenerative diseases. The diagnosis of apathy was based on a single item of the NPI, not on a clinical diagnosis or an apathy-specific scale. We did not take into account severity either. The related scores include both severity and level of change from previous behavior pattern, therefore, are a more subjective rating than simple occurrence [35, 38]. Despite these shortcomings, the unidimensionality of the criterion did not seem to inflate prevalence of the condition. Our approach also prevented us to investigate subdomains of apathy (e.g., behavior, cognition, and emotional domains) what could have been possible with specific tools. We only investigated the potential role of apathy and depression on MCI-AD conversion. As future directions, we plan to perform latent class analyses (LCA) with possible clusters of NPS (e.g., irritability/agitation, delusions/hallucinations) [62] to better elucidate the relationship between patterns of NPS and progression to dementia. Furthermore, the integration of neurobiological measures (e.g., biomarkers, genetics) into the clinical framework is also planned.
In conclusion, our results have potential implications for early prediction of dementia in those with NPS, particularly apathy. Since patients with apathy are more likely to progress to AD diagnosis, they might be considered for early assessment/intervention. Unlike established clinical assessments such as amyloid and tau PET imaging, the evaluation of apathy is cost-effective, non-invasive, and does not rely on availability of expensive equipment. Therefore, assessments of apathy may provide an efficient and highly scalable method for improving risk stratification across clinical and research settings.
Footnotes
ACKNOWLEDGMENTS
The authors have no acknowledgment to report.
FUNDING
This study was supported by the Texas Alzheimer’s Research and Care Consortium (TARCC 2020-23-93-II) and the NIH (P30AG066546).
CONFLICT OF INTEREST
The authors have no conflict of interest to report.
