Cholesterol Level, Statin Use and Alzheimer's Disease in Adults with Down Syndrome

Abstract

Adults with Down syndrome (DS) are at significantly higher risk of Alzheimer's disease (AD) than the general population, but there is considerable variability in age at onset. This study tested the hypothesis that total cholesterol (TC) levels are related to vulnerability, and that the use of statins may decrease risk. The relation of TC level and statin use to risk of AD was investigated in 123 Caucasian adults with DS. Evaluations included serial assessments of cognitive, adaptive and maladaptive behavior, medical records, and neurological examinations. Mean length of follow-up was 5.5 years [1.2-7.1] for the entire sample, 5.1 years [1.2-7.1] for subjects who developed dementia, and 5.6 years [1.5-7.1] for those who did not develop dementia. Controlling for covariates, participants with TC ≥ 200mg/dL were more than two times as likely to develop AD than subjects with lower TC [Hazard Rate (HR) = 2.59, p = .029, 95% CI: 1.1,6.1]. In contrast, participants with higher TC levels who used statins during the study, had less than half the risk of developing AD than participants with higher TC levels who did not use statins (HR = .402, p = .095, 95% CI: .138, 1.173). If the protective effects of statins can be further validated, these findings suggest that their use may delay or prevent AD onset in vulnerable populations.

A number of reports have suggested that there is a link between serum cholesterol level and risk of AD. Findings have been inconsistent, however, with various measures of cholesterol not always significantly related to increased risk of AD [3, 4, 6, 10, 11, 14, 15, 17, 24, 35]. The relationship between cholesterol and AD-type neuropathology has led to a working hypothesis, based upon human, cellular and animal studies, that cholesterol metabolism plays a major role in the pathophysiological development, deposition, as well as clearance of beta-amyloid [22, 33]. Launer and colleagues [13] found an association between increased late-life levels of high-density lipoprotein cholesterol and increased numbers of neuritic plaques and tangles, while low mid-life TC was associated with fewer numbers of neuritic plaques and tangles. Kuo et al. [12] found a significant relationship between increased TC and increased Aβ1-42 in brain in sera obtained immediately postmortem from 64 neuropathologically diagnosed AD cases and 36 controls.

HMG-CoA (3 hydroxy-3methylglutaryl-coenzyme A) reductase inhibitors, commonly known as statins, are the most widely prescribed class of cholesterol-lowering drug. Statins have been found to be related to a lower risk of dementia and/or AD in case control studies, while prospective studies have found inconsistent results [5, 27-29, 36]. Non-statin cholesterol lowering medications are not associated with a decreased risk of AD, which suggests that it may not be the cholesterol lowering effect of statin use, per se, that reduces the risk of AD [5, 20]. Pedrini has posited, for example, that statins may disrupt certain enzyme reactions that lead to amyloid deposition and plaque formation [20].

Virtually all adults with DS over 35 years of age have neuropathology consistent with a diagnosis of AD, and most will develop dementia [32], starting around 50 years of age. Adults with DS have both an increased neural beta-amyloid load and onset of dementia that is 10-20 years earlier than peers in the general population [9]. Age at onset varies widely, and few factors, with the exception of age, APOE genotype, and age at menopause (for women), are known to influence the development of AD in this population [25, 26]. Because of the high incidence and earlier onset of AD in this group, study of adults with DS provides a unique opportunity to examine the hypothesis that increased TC levels are related to an increased risk of AD, and that the use of statins may decrease risk of AD.

The relation of TC level and statin use to risk of AD was compared in 123 Caucasian adults with DS. The study hypotheses were that high levels of TC would be related to an increased risk of AD, and that the use of statins would be related to a reduced risk of AD. An initial pool of 281 individuals with DS was ascertained from a large developmental disabilities service system. Adults 40 years of age and older were selected using age- and sex-stratified random sampling as well as samples of convenience. [A related study is focused on women's heath issues, therefore, women participants were oversampled.]

The Institutional Review Boards of participating institutions approved this study, and family members/correspondents of participants were contacted by staff of social service agencies to initiate recruitment. A family member/correspondent then provided written informed consent, with participants providing consent or signing a form acknowledging their assent and willingness to participate (in cases where capacity to provide valid informed consent was at issue).

Participants were excluded if they had possible or definite dementia at baseline (n= 25), or if data were unavailable regarding TC level (n=72), APOE genotype (n=42), statin use (n=0), or body-mass index [BMI] (n=4). Ethnicity has been found to influence AD incidence, age at onset, and rate of decline [1, 2, 7], and there were too few non-Caucasian subjects in the sample to stratify the analyses (n=23). Therefore, only data for Caucasians were analyzed. (A re-analysis of data including all participants found essentially the same results; not shown.) The final sample included 123 nondemented adults with DS, 41.4 –78.1 years old at baseline. Univariate analyses of variance and chi-square analyses were used to determine whether missing data were associated with age, sex, level of mental retardation (LMR), BMI, or APOE and no significant effects were found. Cytogenetic confirmation of DS status was available for 87% of the participants (n = 107), with clinical and phenotypic confirmation used for the remaining 16 (13%).

A cross-sequential research design was employed, with longitudinal testing of subjects up to five times, between May 1998 and April, 2006. Mean follow-up was 5.5 years (S.D. = 1.4) [median = 5.9, range = 1.2-7.1] for the entire sample, 5.1 years (S.D. = 1.6) [median = 5.5, range = 1.2-7.1] for subjects who developed dementia, and 5.6 years (S.D. = 1.3) [median = 6.0, range = 1.5-7.1] for those who did not develop dementia. Evaluations included assessments of adaptive and maladaptive behavior, together with a review of all medications and medical records at intervals ranging from 14-21 months. Maladaptive behavior was measured to provide information on depression, psychosis and other changes that might mimic or be associated with dementia. Cognitive abilities were assessed using a battery that covered a wide range of intellectual disability. A full description of the instrument battery and study procedures has been previously published [37].

Classification of dementia status was determined at consensus conferences using all available data (but blind to APOE genotype). Subjects were considered demented if there was a history of progressive memory loss, disorientation, and functional decline over a period of at least one year, and if no medical or psychiatric conditions that might result in or mimic dementia were present. They were classified as nondemented if they were without cognitive or functional decline or if they showed only mild declines of insufficient severity to meet criteria for dementia. Neurological examinations determined differential diagnoses for subjects classified as having dementia. Each case was classified as: (a) non-demented, indicating that dementia was definitely not present or some uncertainty regarding status was indicated, or (b) demented, indicating that symptoms of dementia were present. Only participants classified as having AD were included in the analyses; participants with mixed AD diagnoses were excluded (n=4).

Blood samples for determination of DS karyotype and APOE genotype were collected from willing subjects via routine venipuncture by qualified program staff. APOE genotypes were determined using standard PCR-RFLP methods [8], and were coded dichotomously for analysis. Either the participant had one or more APOE ∈4 alleles or the participant did not have an APOE ∈4 allele.

TC levels were obtained from lab reports available in medical records, and therefore procedures and timing of blood draws (i.e., fasting versus non-fasting) varied from case to case. Recent findings, however, indicate that random TC levels are not significantly different from fasting levels [21] and can be measured accurately on non-fasting samples [31]. All TC tests were completed before baseline data collection. In addition, TC levels were determined independent of dementia assessments, and any variation should be random. In accordance with National guidelines [16], TC levels were initially coded as <200 mg/dL (desirable), 200–239 mg/dL (borderline high) or ≥240 mg/dL (high). Few subjects had levels ≥240 (n=6, who were not using statins), so final coding of TC for analysis was <200 mg/dL (desirable) and ≥200mg/dL (high).

BMI was calculated as weight in kilograms divided by height (in meters) squared. The distributions for age at baseline, sex, premorbid level of MR, BMI, TC level, and APOE genotype as a function of AD status and TC category are provided in Table 1.

Table 1.

Demographic characteristics by APOE genotype and TC level

Characteristic	Total	Not Demented	Demented	Desirable TC	High TC
Sample size	123	85	38	53	70
Age at baseline  Mean (S.D.)	51.2 (5.4)	50.1** (5.0)	53.6** (5.5)	50.2* (4.9)	51.9* (5.7)
Sex n, (%)
Male	28 (22.8)	17 (20.0)	11 (28.9)	13 (24.5)	15 (21.4)
Female	95 (77.2)	68 (80.0)	27 (71.1)	40 (75.5)	55 (78.6)
Level of MR n (%)†
Mild/moderate	56 (48.3)	43 (53.8)	13 (36.1)	22 (44.9)	34 (50.7)
Severe/profound	60 (51.7)	37 (46.3)	23 (63.9)	27 (55.1)	33 (49.3)
Body Mass Index  Mean (S.D.)	28.6 (6.0)	29.4 (6.2)*	26.9 (5.2)*	28.5 (6.8)	28.7 (5.4)
TC level mean (S.D.)	205.7 (33.1)	204.8 (35.0)	207.7 (28.8)	177.0** (17.6)	227.4** (24.4)
APOE Genotype
∈4 Allele present	102 (82.9)	72 (84.7)	30 (78.9)	41 (77.4)	61 (87.1)
∈4 Allele not present	21 (17.1)	13 (15.3)	8 (21.1)	12(22.6)	9 (12.9)

^†N < 123 because of missing data

^*p < .10

^**p < .05

Univariate analyses of variance and chi-square analyses were used to determine whether there were differences in age at baseline, sex, LMR, TC level, BMI, or APOE genotype as a function of AD status and TC category. Cumulative incidence was computed using Kaplan-Meier survival analysis with the Mantel-Cox log-rank score for determining statistical significance, and Cox proportional hazards modeling (CPHM) to evaluate the association between: (a) TC level and risk of AD, and (b) statin use and risk of AD. The time-to-event variable was defined as time since baseline data collection in years. For demented participants, time of onset of AD was determined to be the halfway point between the time of data collection before the onset of dementia and the time of data collection when dementia was recognized. For nondemented participants, time-to-event was the time from baseline to the time at last data collection (in years).

To control for the effects of cholesterol lowering medication on risk of AD, analyses relating TC level to AD risk only included subjects who were not taking cholesterol-lowering medication (n=89; 30 demented, 59 nondemented). The analysis relating statin use to risk of AD included all subjects (n=123; 38 demented, 85 nondemented). All analyses were conducted using SPSS-13 [30].

There were no significant differences in sex, LMR, or APOE genotype, as a function of AD status. There was a significant difference in age at baseline as a function of AD classification (F (1,121) = 11.8, p = .01), with participants who developed AD being 3.5 years older than their nondemented peers. Mean age of the 38 participants who developed dementia was 56.1 (S.D.= 5.3, Range= 47.6-80.4). There was also a marginally significant effect of BMI, indicating that participants who developed AD had lower initial BMI's (26.9 versus 29.4).

There were no significant differences in sex, LMR, or APOE genotype related to TC. Participants with high TC levels were approximately 1.7 years older at the beginning of the study than participants with desirable TC levels (F (1,121) = 3.0, p < .09), and consequently all initial CPHM included age at baseline as a covariate to ensure that differences in age did not confound the findings. Final CPHM also included sex, LMR, BMI, and APOE genotype as covariates.

Participants with high TC levels were at a significantly increased age-specific risk of developing AD than participants with desirable TC levels (Mantel-Cox Log Rank (1) = 5.5, p < .019)(Fig. 1, Kaplan-Meier survival analysis cumulative hazard function). Controlling for age at baseline in a CPHM, participants with high TC levels were over two times more likely to develop AD over the course of the study than participants with desirable TC levels (Hazard Rate [HR] = 2.14, p = .052, 95% CI: .994,4.622). Age at baseline was related to risk of AD (HR = 1.1, p < .001, 95% CI: 1.04,1.15), with the expected finding of a significantly increased age-specific risk of developing AD. In a CPHM controlling for sex, LMR, BMI, and APOE genotype as covariates, participants with high TC levels were approximately 2.5 times more likely to develop AD over the course of the study than participants with desirable TC levels (HR = 2.59, p = .029, 95% CI: 1.1,6.1). In addition to finding a significantly increased age-specific risk of developing AD (HR = 1.1, p < .001, 95% CI: 1.04,1.2), a significant effect of BMI (HR = .932, p = .05, 95% CI: .87, .99) showed that participants with higher BMI's had significantly reduced risk of developing AD than participants with lower BMI's. (A final CPHM added the interaction effect of APOE genotype X TC category, but no significant interaction effect was found.)

Figure 1.

Legend. Kaplan–Meier estimates of cumulative hazard of AD by baseline level of TC and time (in years) since baseline assessment. Tic marks denote participants without an event (censored cases).

While statin use was unrelated to dementia risk for participants with desirable TC levels (Mantel-Cox Log Rank (2) < 1, p > .05), statin use was associated with reduced risk for participants with high TC levels (Mantel-Cox Log Rank (1) = 4.5, p < .033) (Fig. 2, Kaplan-Meier survival analysis cumulative hazard function). Subjects with high TC levels who used statins were approximately 40% as likely to develop AD than subjects with high TC levels who did not use statins (HR = .402, p = .095, 95% CI: .138, 1.173). Again, age at baseline was also related to risk of AD (HR = 1.1, p < .001, 95% CI: 1.05, 1.17), with older subjects more likely to develop AD. Sub-samples treated with any specific molecule were of insufficient size to permit statistical analysis, but were: 14 (41%) Atorvastatin, 11 (32%) Pravastatin, 6 (18%) Simvastatin, 2 (6%) Lovastatin and 1 (3%) Fluvastatin.

Figure 2.

Legend. Kaplan–Meier estimates of cumulative hazard of AD by statin use and time (in years) since baseline assessment in participants with TC ≥ 200 mg/dL. Tic marks denote participants without an event (censored cases).

To our knowledge, this is the first study to find an association between TC levels, statin use and risk of AD in adults with DS. The fact that a large proportion of adults with DS experience AD 10-20 years earlier than adults in the general population reduces the difficulties inherent in surveiling a cohort of subjects to determine the effects of putative risk factors of AD. Among subjects who were not demented at baseline, those with high levels of TC (i.e., ≥ 200 mg/dL) were more than two times as likely to develop AD as those with desirable (i.e., < 200 mg/dL) TC levels. Within the high TC group, individuals who used statins during the course of the study were approximately 40% as likely to develop AD than those who did not use statins. The adjusted risk of developing AD did not differ as a function of statin use for participants who were non-demented at baseline and who had TC levels < 200. These conclusions are in general agreement with many of the positive findings reported in previous studies regarding the relationship between increased TC level and increased risk of AD [17, 18, 23]. However, the expected interaction between APOE genotype and TC category (i.e., higher TC levels related to increased risk of AD only in subjects without an APOE ∈4 allele) found in some previous studies [3, 4, 6] was not evident, most likely because of small sample size and insufficient power to detect an interaction. The significant relationship between TC and AD risk, even within a group with a known high neural beta-amyloid load, provides strong evidence for the importance of cholesterol metabolism as a mediator of AD risk.

The protective effect of statin use on risk of AD among participants with high cholesterol levels parallels some findings from observational studies in the general population. However, the results of clinical trials examining the beneficial effect of statins on AD risk have been less encouraging (Panza et al., 2006). Recent data suggest a cholesterol-independent action of statins and inhibition of beta-amyloid deposition in the brain may be one possible pathway mediating the relationship between statin use and AD risk [20]. Statins also have anti-inflammatory and antioxidant properties that may afford some protection [5], but these results indicate that the cholesterol-lowering effect of statin use could be mediating risk of AD in and of itself.

The observed relationship between increased BMI and reduced risk of AD may be due, in part, to increased levels of estrogen. Patel et al. [19] reported that higher BMI was related to increased levels of bio-available estradiol, and to better neuropsychological function in postmenopausal women with DS.

This study has several limitations. Participant TC levels were obtained from medical records, and we had no control of procedural variation. Recent findings, however, indicate that random total cholesterol levels are not significantly different from fasting levels, probably the major factor of concern [21, 31], and whatever variation was introduced could not have had a systematic effect on our dementia classifications. Similarly, the specifics of statin use were not controlled and various dosages and treatment durations and compounds were combined in analyses. There may be important differences in the effectiveness of particular formulations in reducing risk of AD. While there are a number of statin trials underway in the general population [5, 34], results thus far have been inconclusive. The use of statins for reduction of AD risk in DS will need to be evaluated in carefully controlled clinical trials, and if the neuroprotective effects of statins on cognitive impairment and risk of AD can be further validated, these findings suggest one modifiable factor that may delay or prevent onset of AD.