The study of disease progression in dementia syndromes is important to identify proper preventive strategies. Neuropsychiatric symptoms, such as hallucinations, agitation, and sleep disturbances, are factors associated with poor prognosis for patients with Alzheimer’s disease (AD) (
1). Other anthropometric, demographic, and clinical features have been associated with variable disease progression in Brazilian patients with AD, such as sex, education, age at dementia onset, sanitary conditions, history of traumatic brain injury with loss of consciousness, lifetime physical activity and occupational complexity (
2), variations in cerebrovascular risk factors (
3), body mass index, and coronary heart disease risk (
4).
Allelic forms of the APOE gene comprise ε4, ε3, ε2, and the rare ε3r (
5). APOE ε4 carriers have inefficient neural repair mechanisms particularly related to reduced clearance of amyloid-β, leading to synaptic dysfunction and emerging memory failure (
6); such mechanisms are potentially modifiable by epigenetic factors and compensated by healthier lifestyles. Nonetheless, APOE is a moderately penetrant gene that is neither a prerequisite nor a sufficient agent for development of AD (
7). APOE ε4 carrier status modulates cognitive decline in AD (
2), but its direct effects may be restricted to risk of subjective cognitive decline (
8) and incidence and age at dementia onset, as well as to development of behavioral features (
9). APOE ε4 carrier status also modulates the biochemical profile of the cerebrospinal fluid and its associations with behavioral symptoms (
10), but it does not seem to interfere with longevity in successful aging (
11).
Methods
Ethics Approval
The protocol was approved by the Ethics Committee of Hospital São Paulo, Federal University of São Paulo (UNIFESP), São Paulo, Brazil, according to the research project 1067/10 (CAAE 0540.0.174.000-10) and followed the code of ethics of the World Medical Association (Declaration of Helsinki) for experiments involving humans. All invited patients and their legal representatives agreed to participate in the research and signed an informed consent form before the evaluation.
Clinical Assessment
Consecutive outpatients with late-onset AD according to National Institute on Aging–Alzheimer’s Association criteria (
16) were recruited from November 2010 to May 2014 at the Behavioral Neurology Section of Hospital São Paulo, UNIFESP. Late-onset AD was considered when the dementia syndrome began after age 60 (
9). Participants could not have other structural brain diseases (such as brain tumors or stroke) and had to be willing to join the research along with their caregivers. All patients had magnetic resonance imaging of the brain to assess for medial parietal or medial, basal, or lateral temporal atrophy; in cases of claustrophobia or use of a pacemaker, a computed tomography scan excluded vascular lesions. Considering the high prevalence of vascular cognitive impairment in Brazilian university hospitals (
17) and the fact that small-vessel disease is very common in brains of older people, the Hachinski ischemic score, as modified by Loeb and Gandolfo (
18), was employed to exclude vascular dementia.
In this uncontrolled cohort, screening was conducted in the first consultation, when patient evaluations documented sex, education, current age, estimated age at dementia onset, cerebrovascular risk factors, use of a pacemaker, and number of different medications in use. Age at dementia onset was determined following a review of medical records for cognitive and functional decline and confirmed after an interview with the caregiver. Each patient was followed for 1 year and prospectively assessed for psychotropic therapy, body mass index, waist circumference, blood pressure, and scores on the Mini-Mental State Examination (MMSE) (
19) and the Severe Mini-Mental State Examination (SMMSE) (
20), and caregivers were queried for the Clinical Dementia Rating sum of boxes (CDR-SOB) (
21), the Brazilian version of the Zarit Caregiver Burden Interview (ZARIT) (
22), the Index of Independence in Activities of Daily Living (ADL) (
23), and Lawton’s Instrumental Activities of Daily Living (IADL) Scale (
24). The combination of cognitive and functional assessments was in accordance with traditional Brazilian guidelines (
25), and scoring guidelines for functional assessments have been previously described (
1). All evaluations were conducted on weekdays during morning hours by the same examiner (F.F.O.), who was blinded to genetic data. Although patients had at least three consultations during the year in which they were followed, only baseline and final scores were taken into account for the statistical analysis.
Regarding cerebrovascular risk factors, smoking and alcohol use were averaged for the year, whereas blood pressure, body mass index, and waist circumference were prospectively followed. Diagnosis of arterial hypertension followed traditional guidelines (
26), and variations in pulse pressure (the difference between systolic blood pressure and diastolic blood pressure) were also measured (
3). Diagnoses of diabetes mellitus (
27) and hypercholesterolemia (
28) were based on the results of blood tests, with specific guidelines for management. All efforts were directed to reduce levels of glucose or cholesterol to normal as much as possible.
Regarding psychotropic therapy, particular attention was given to categorize medications as follows: cholinesterase inhibitors, memantine, antidepressants, second-generation antipsychotics, and antiepileptic drugs.
Genotyping
Blood was collected from all patients in tubes with ethylenediaminetetraacetic acid 0.1%. After genomic DNA extraction by a standard salting-out procedure, all samples were genotyped for rs7412 and rs429358 (APOE) by way of real-time polymerase chain reactions by using TaqMan SNP Genotyping Assays on the Applied Biosystems 7500 Fast Real-Time PCR System (Applied Biosystems, Foster City, Calif.), following the standard protocols of the manufacturer. Genotyping procedures were carried out only after clinical data were collected from all patients.
Outcome Measures
The main outcome measure was the score variation in 1 year regarding cognition, functionality, dementia staging, or caregiver burden, taking into account the use of psychotropic drugs or the number of different medications in use and APOE ε4 carrier status. Secondarily, we assessed variations in the number of different medications in use, anthropometric measures, and test scores according to APOE ε4 carrier status.
Statistical Analyses
Student’s t test was used for variations in 1 year of anthropometric variables and test scores (taking baseline and final scores after 1 year into account), as well as to compare the number of different medications in use and variations in 1 year of anthropometric variables and test scores according to APOE ε4 carrier status. Fisher’s exact test was used to compare rates of psychotropic use according to APOE ε4 carrier status. A univariate general linear model, separately for APOE ε4 carriers and APOE ε4 noncarriers, was used for test score variations in 1 year, according to the number of different medications in use, and use or not of a cholinesterase inhibitor, memantine, an antidepressant, a second-generation antipsychotic, or an antiepileptic drug. The general linear model was adjusted for sex, years of education, baseline age, estimated baseline disease duration, number of cerebrovascular risk factors (up to six: need of a pacemaker, alcohol use, smoking, diagnoses of arterial hypertension, diabetes mellitus, and hypercholesterolemia), and variations in 1 year of pulse pressure, body mass index, and waist circumference. Secondarily, the same general linear model was used to assess whether the combination of a cholinesterase inhibitor with memantine, as well as combined antidepressants, second-generation antipsychotics, and antiepileptic drugs affected test score variations in 1 year for all patients by including copies of APOE ε4 in the adjustment. Univariate analyses disclosed how test scores varied according to the adjusted effects of psychotropic drugs or their combinations. Significance was set at p<0.05.
Results
Overall, 217 patients were included. During follow-up, 14 died (6.5%), and 10 abandoned the study (4.6%), resulting in a final sample of 193 patients. No significant difference was noted in number of different medications in use between APOE ε4 carriers and APOE ε4 noncarriers (mean=6.06, SD=2.2; and mean=6.32, SD=2.2, respectively; p=0.406). There was also no significant difference in use of any psychotropic class according to APOE ε4 carrier status, except for more frequent second-generation antipsychotic drug therapy among APOE ε4 carriers (p=0.027).
Table 1 shows clinical and demographic data for all patients. Of 193 patients, two still used alcohol (one used around 180 L per year, and the other used around 25 L per year), and seven patients smoked (mean=101.43, SD=62.0 packs per year). Regarding behavioral pharmacological therapy, 181 patients used a cholinesterase inhibitor (93.8%), 143 used memantine (74.1%), 93 used an antidepressant (48.2%), 56 used a second-generation antipsychotic (29.0%), and 25 used an antiepileptic drug (13.0%). None of the patients had epilepsy, and thus use of antiepileptic drugs was for behavioral improvement only.
Table 2 shows variations in test scores and anthropometric measures in 1 year. Blood pressure and body mass index were significantly lower at the end of the follow-up for both APOE ε4 carriers and APOE ε4 noncarriers. Waist circumference was not significantly different after 1 year, regardless of APOE ε4 carrier status. All test scores were significantly worse after 1 year, both for APOE ε4 carriers and APOE ε4 noncarriers, with the exception of the ZARIT, which did not show a significant change for either group.
In separate analyses, we assessed potential differences in variations of anthropometric measures and test scores in 1 year between APOE ε4 carriers and APOE ε4 noncarriers. No differences were found regarding variations in pulse pressure, systolic blood pressure, diastolic blood pressure, body mass index, or waist circumference or in scores on the CDR-SOB, SMMSE, ZARIT, ADL, or IADL, although we found a faster decline of MMSE scores in APOE ε4 carriers (p=0.046).
Tables 3 and 4 show effects of psychotropic drugs or the number of different medications in use on test score variations in 1 year for APOE ε4 carriers and APOE ε4 noncarriers, respectively. For APOE ε4 carriers, no significant effects of any psychotropic class or number of different medications in use were found, although there was a nonsignificant association between antidepressant therapy and improved CDR-SOB scores. For APOE ε4 noncarriers, the following correlations were found: use of memantine was associated with faster worsening of SMMSE scores, and use of a second-generation antipsychotic had a nonsignificant effect in the same direction; antiepileptic therapy and the number of different medications in use were associated with faster worsening of ADL scores; use of a cholinesterase inhibitor was associated with improvement in ADL scores and slower worsening of ZARIT scores.
Effects of combined psychotropic drugs on test score changes were also assessed. Patients who used memantine combined with a cholinesterase inhibitor (N=134) were compared with patients who used neither (N=3), and a slower worsening of SMMSE scores was seen for those taking the combination (
=4.826, 95% CI=0.799, 8.854, p=0.019); effects on other test scores were nonsignificant. Patients who used an antidepressant with a second-generation antipsychotic (N=24) were compared with patients who used neither (N=68), and no significant effects of the combination on any score changes were noted. Patients who used an antidepressant with an antiepileptic drug (N=14) were compared with those who used neither (N=89), and the drug combination was associated with faster worsening of scores on the ADL (
=–0.919, 95% CI=–1.818, –0.020, p=0.045); effects on other test scores were nonsignificant. Patients who used a second-generation antipsychotic with an antiepileptic drug (N=8) were compared with patients who used neither (N=120), and the combination was associated with faster worsening of SMMSE scores (
=–3.594, 95% CI=–5.749, –1.439, p=0.001); in addition, a nonsignificant association was seen with faster worsening of scores on the ADL (
=–0.755, 95% CI=–1.618, 0.107, p=0.086). Effects on other test scores were also nonsignificant. Patients using a combination of three drugs (N=4)—an antidepressant, a second-generation antipsychotic, and an antiepileptic drug—were compared with those who used none of these drugs (N=61); a nonsignificant association was noted between the combination and faster worsening of SMMSE scores (
=–3.465, 95% CI=–7.426, 0.497, p=0.085); effects on other test scores were also nonsignificant.
Discussion
In this study, we showed that psychotropic therapy differentially affected cognitive and functional changes, as well as caregiver burden, according to APOE ε4 carrier status, and that the number of different medications in use was associated only with worsening basic functionality for APOE ε4 noncarriers. One naturalistic study had already shown an association between the number of different medications in use and increased functional impairment in AD, but that study did not stratify participants according to APOE ε4 carrier status (
29). In our cohort, APOE ε4 carriers did not experience any effects of isolated psychotropic drugs on clinical changes, which could be due to disruptive mechanisms already inherent to this genetic heritability, such as disorders of amyloid-β and tau proteostasis, mitochondrial dysfunction, and neuroinflammation (
30). Other mechanisms that could potentially explain the lower susceptibility of APOE ε4 carriers to benefits of cholinesterase inhibitors or detrimental effects of second-generation antipsychotics and antiepileptic drugs include the impairment in blood-brain barrier integrity causing diminished clearance of amyloid-β, reduced suppression of inflammatory signaling, and reduced glutamate receptor function and synaptic plasticity (
31).
Only for APOE ε4 noncarriers did use of cholinesterase inhibitors benefit basic functionality and caregiver burden, possibly due to improvements in behavior or executive functioning (
32), whereas memantine was harmful in regard to SMMSE score changes. Nevertheless, memantine combined with a cholinesterase inhibitor benefited SMMSE score changes regardless of APOE ε4 carrier status, but it should be noted that few patients (N=3) used neither of these drugs. The heterogeneity of our sample regarding dementia stages could explain why therapeutic effects on other score changes remained nonsignificant. Efficacy of memantine has been shown primarily in patients with moderate or severe dementia (
33), although one small retrospective study showed that its cognitive effects depended on the presence of vitamin D deficiency (
34). The combination of memantine with a cholinesterase inhibitor has been shown to be particularly stronger for patients with behavioral symptoms (
35) and synergistic for slowing cognitive and functional decline in AD even in the long term (
36), but not for prolonging survival (
33). One meta-analysis confirmed the modest benefits of maximal doses of cholinesterase inhibitors and memantine for cognition, functionality, and behavior in patients with AD (
37). Another study observed better functional but worse cognitive response to cholinesterase inhibitors in younger patients (
29), and it has been suggested that these drugs do not benefit patients in the mild dementia stage (
38). Furthermore, cholinergic drugs may increase cerebral blood flow and modulate the sleep-wake cycle, correlating with cognitive improvement by such treatment, and it has been shown that APOE ε4 carriers have deficient hippocampal cholinergic compensatory sprouting and remodeling in response to cholinergic deafferentation, as well as reduced neuronal metabolic activity in the nuclei basalis of Meynert (
30). Harmful therapeutic effects in some clinical trials may reflect the existence of diverse caveats: selection bias by prescription of cognitive enhancement therapy to patients who are declining faster (
35); lacking information on APOEε4 carrier status; and insufficient duration of follow-up to allow proper assessments.
Only for APOE ε4 noncarriers was use of second-generation antipsychotics harmful regarding SMMSE score changes, although this association was nonsignificant, whereas for noncarriers antiepileptic therapy was harmful regarding basic functionality. Antipsychotic therapy has been associated with more behavioral symptoms in AD, according to increasing dementia severity, possibly due to confounding by indication (
33); despite the potential prospective behavioral benefits of antipsychotics, they have been consistently associated with faster cognitive decline in several studies of patients with AD (
39). One 6-month study did not detect cognitive variations according to whether patients used or did not use second-generation antipsychotics and attributed the faster deterioration seen in other studies to sedation, intrinsic antimuscarinic properties, and lower baseline cognition (
40). Considering that our patients were followed for 1 year, we believe it is more likely that our results reflect adverse cognitive effects of second-generation antipsychotics in the long term. Second-generation antipsychotics may have antioxidant effects and increase brain synthesis of dopamine, improving sleep-wake patterns and psychotic symptoms (
1), but they have also been associated with increased risk of thromboembolic events, metabolic syndrome, and death (
39).
Overall, combinations of a second-generation antipsychotic with an antiepileptic drug were harmful regarding basic functionality and SMMSE score changes, regardless of APOE ε4 carrier status. Antiepileptic drugs are usually prescribed off label for behavioral symptoms in AD, and one clinical trial showed that use of divalproex sodium did not delay emergence of behavioral symptoms or slow cognitive or functional decline in patients with moderate AD but was associated with more brain atrophy after 1 year and higher rates of somnolence (
41). In our study, antidepressants were harmful regarding basic functionality only when combined with an antiepileptic drug, possibly due to a major effect of the antiepileptic drug. Stimulation of 5-HT receptors helps to protect neurons of the hippocampi and neocortex and might have antiamyloidogenic effects (
42).
Psychotic symptoms and anxiety are associated with poorer prognoses in patients with depression (
43). Symptom chronicity, late improvement, and executive dysfunction are additional predictors of poor response to therapy with antidepressants (
44). Selective serotonin reuptake inhibitors, such as citalopram, have shown similar efficacy in the treatment of dementia-associated agitation, compared with second-generation antipsychotics, but with an improved safety profile (
45). Citalopram and fluoxetine have been associated with prevention of depression in addition to enhancement of motor recovery when combined with physiotherapy after an acute ischemic stroke, probably due to modulation of neuroplasticity (
46), a mechanism that may also benefit patients with dementia. Nevertheless, these findings have been challenged by recent large randomized studies that confirmed antidepressant effects but showed increased risk of bone fractures and unimproved functional outcomes with fluoxetine (
47). Use of different scales might explain these differences, but future analyses of more focused cohorts concerning motor impairment or functional status will elucidate the reasons for these differences.
Although some studies have shown adverse effects of all psychotropic drugs on cognitive and functional decline in AD (
48), with variable efficacy for behavioral symptoms, one study showed that only antipsychotics and benzodiazepines affect functional (but not cognitive) decline and mortality (
49). Intrinsic anticholinergic properties are common to several psychotropic drugs of different classes that were used in our study (such as paroxetine, trazodone, and risperidone but not venlafaxine) and have been associated with cognitive and functional impairment, as well as worse physical performance, in older people (
50).
APOE ε4 carriers used second-generation antipsychotics more often than did noncarriers, probably reflecting higher behavioral burden. All cognitive and functional scores (but not caregiver burden) were significantly changed after 1 year, but only MMSE score changes were different according to APOE ε4 carrier status. Although behavioral burden may be even higher in other dementia syndromes (
51), behavioral symptoms, rather than cognition, are the primary determinants of caregiver burden (
14) and correlate better with functionality than with cognitive decline (
33). Yet cognition has been known to have an impact on functionality in all stages of AD (
52). Education affects cognitive test scores but not functionality tests (
2), whereas instrumental activities of daily living are more susceptible than is basic functionality to aging and mild cognitive changes (
4). Whereas increasing body mass index may benefit functionality and cognition (
53), excess abdominal fat results in insulin resistance and higher susceptibility to the neurological effects of obesity, which explains why waist circumference should be added to body mass index as a surrogate marker for neurological disease modulation (
2).
One important limitation of the study is that we did not have prospective behavioral data for our patients. Even though they used several drugs for behavioral enhancement, behavioral symptoms are subject to change, and cross-sectional assessments are prone to error. Another limitation is related to generalizability, considering that all participants were recruited from a single medical setting and are thus unlikely to represent the full spectrum of the disease in the general population. It is controversial whether rates of clinical changes differ according to dementia stages; however, small subgroup sizes prevented us from examining our results in each dementia stage. This would have been particularly important for memantine, indicated only for patients with moderate and severe dementia. Furthermore, our modest sample size did not allow us to discriminate psychotropic classes into each specific medication or according to dosage or duration of specific therapy. Reproduction of our analyses in other centers with larger samples is recommended to ascertain our findings. Nevertheless, our cohort was carefully characterized across the whole spectrum of dementia severity with use of standardized assessments during a relatively long follow-up. The fact that most anthropometric variables and all cognitive and functional test scores were significantly different at the end of 1 year validates our findings regarding the suitability of the duration of follow-up. Nevertheless, the inclusion criterion of having a reliable caregiver maximized the likelihood that medication compliance was optimal, and detailed availability of clinical information also allowed proper statistical adjustments.