|Home | About | Journals | Submit | Contact Us | Français|
To determine whether pre-diagnosis vascular risk factors are associated with Alzheimer’s disease progression.
Inception cohort followed longitudinally for a mean of 3.5 (up to 10.2) years.
Washington Heights Inwood Columbia Aging Project, New York City
156 incident AD patients (mean age 83 years at diagnosis)
Vascular factors including medical history (heart disease, stroke, diabetes, hypertension), smoking, and pre-diagnosis blood lipid measurements (total cholesterol, High density lipoproteins (HDL-C), Low density lipoproteins (LDL-C) and triglycerides).
Change in a composite score of cognitive ability from diagnosis on.
In Generalized Estimating Equation (GEE) models (adjusted for age, race/ethnicity and education), higher cholesterol (total and LDL-C), and diabetes history were associated with faster cognitive decline. Each 10-unit increase in cholesterol and LDL-C was associated with a 10% of a standard deviation decrease in cognitive score per year of follow-up (p<0.001 for total cholesterol, p=0.001 for LDL-C). HDL and triglycerides were not associated with rate of decline. Diabetes history was associated with an additional 50% of a standard deviation decrease in cognitive score per year (p=0.05). History of heart disease and stroke were associated with cognitive decline among APOE-ε4 carriers only. In a final GEE model that included HDL-C, LDL-C and diabetes, only higher LDL-C was independently associated with faster cognitive decline.
Higher pre-diagnosis total cholesterol, LDL-C, and diabetes were associated with faster cognitive decline among incident AD patients, providing further evidence for the role of vascular risk factors in Alzheimer’s disease course.
Few treatment options are available to improve AD prognosis. Controlling vascular conditions may be one way of delaying the disease course. Vascular risk factors and vascular disease are associated with higher risks of vascular dementia (VaD) 1, 2 and AD 3, 4. We previously reported associations between stroke 5, hyperinsulinemia 6, diabetes 7, current smoking 8 and hypertension 9 and higher AD risk. We found that high total cholesterol and low density lipoprotein (LDL) were related to increased VaD risk 10, 11, but not AD.
While vascular risk factors have been studied as predictors of AD 4, 12 few studies have assessed their influence on disease progression 13, 14. We examined the interplay between vascular factors and AD course among participants from the Washington Heights Inwood Columbia Aging Project (WHICAP), a multi-ethnic, community-based, prospective study of aging in Northern Manhattan.
Participants in WHICAP come from two population-based cohorts of Medicare enrollees. Recruitment for the first cohort began in 1992. The study area comprised fourteen census tracts in Manhattan between 155th and 181st streets. Lists of Medicare recipients in the study area were obtained from the Health Care Financing Administration. Potential participants were selected by systematic random sampling into one of six strata based on ethnicity (Hispanics, non-Hispanic Blacks, non-Hispanic whites) and age (65–74, 75+). 2125 subjects were interviewed at baseline. A cohort of 2183 additional participants was formed in 1999 using similar methods, with several exceptions: new lists of beneficiaries were obtained but those drawn into the 1992 cohort were excluded; subjects who reported a dementia diagnosis in the course of arranging for the initial evaluation were excluded; the study area was extended to encompass Manhattan north of 145th Street.
The sample for this analysis was restricted to individuals with lipid assessments prior to dementia diagnosis. During follow-up, 417 individuals developed AD. Of these, 319 had pre-diagnosis vascular risk factor data, 156 of whom also had post-diagnosis follow-up data (Figure 1). Of the incident AD cases with vascular risk factors, 44 (14%) died, 55 (17%) were diagnosed in the most recent interview wave and therefore had no post-diagnosis information available, and 64 (20%) were not followed due to refusal or study drop-out. Excluded AD patients were similar to the analysis sample in cognitive status at diagnosis, age, race/ethnicity, gender, APOE- ε4 status, total cholesterol, HDL, LDL, hypertension, stroke and heart disease. However, excluded participants had more education (8.1 years vs 6.6 years; p=0.002), a higher prevalence of diabetes (29% vs 18% of the analysis; p=0.02), and lower triglycerides (mean 154.9 mg/dl vs. 173.2 mg/dl; p=0.04).
The study was approved by the Columbia University institutional review board. Written informed consent was obtained from all subjects.
AD was diagnosed using physician-administered physical and neurological examinations, along with a standardized neuropsychological battery 15. All assessments were administered at baseline and at follow-up visits, which occurred at approximately 18-month intervals. Evaluations were conducted in English or Spanish, based on participant preference. All available information, including medical charts and imaging studies, was considered in the evaluations.
Consensus diagnosis of dementia was made at conferences attended by neurologists and neuropsychologists, using the neuropsychological battery and evidence of social or occupational function deficits, per Diagnostic and Statistical Manual of Mental disorders, Revised Third Edition (DSM-III-R) criteria. Diagnosis of probable or possible AD was based on the National Institute of Neurological and Communicative Disorders and Stroke – Alzheimer’s Disease and Related Disorders Association criteria16.
Five cognitive domains were assessed: 1) Memory: total and delayed recall of the Selective Reminding Test 17; the recognition component of the multiple choice Benton Visual Retention Test 18. 2) Abstract Reasoning: WAIS-R Similarities subtest 19; identities and oddities subtest of the Dementia Rating Scale 20. 3) Visual-Spatial: five items from the Rosen drawing test 21; the matching component of the multiple choice Benton Visual Retention Test 18. 4) Language: 15 item Boston Naming Test 22; the eight high probability items from the Repetition subtest of the Boston Diagnostic Aphasia Examination (BDAE) 22; the first six items of the BDAE Comprehension subtest. 5) Executive-Speed: average scores for phonemic fluency assessed by the Controlled Oral Word Association Test; category fluency (Animals, Food, Clothing) mean scores 23.
The 12 test scores were transformed into Z-scores. Means and standard deviations were calculated from baseline scores of age-, education- and ethnicity-matched non-demented subjects. Z-scores were averaged within cognitive domains, which were subsequently averaged to produce the composite cognitive score24. The outcome measure was the rate of change in the composite score from diagnosis onward.
Baseline age, sex, ethnicity, and education years were collected by interview. Stroke was defined by World Health Organization criteria 25. Diabetes and hypertension were defined based on self-report or documented treatment of either disorder at any time up to AD diagnosis. Heart disease was defined as a history of myocardial infarction, congestive heart failure, or angina pectoris at any time up to AD diagnosis.
Fasting plasma total cholesterol and triglyceride levels were measured an average of 3.3 (SD 2.2) years before diagnosis using standard enzymatic techniques. HDL cholesterol levels were determined after precipitation of apolipoprotein B–containing lipoproteins with phosphotungstic acid. LDL cholesterol levels were recalculated using the Friedewald formula 26. Because HDL and triglyceride values were not normally distributed, log-transformed versions of these variables were used.
Apolipoprotein E (APOE) genotype, determined using established methods 27, was available for 132 participants (85%), and was classified based on the presence of at least one ε4 allele.
Generalized estimating equations (GEE)28 were used to examine the relation of vascular factors to rates of cognitive change. By treating each subject’s repeated measures as a cluster, GEE accounts for the correlation of repeated measures in the same individual. Two models were developed:
Separate models were developed to assess the association of each vascular variable (heart disease, stroke, diabetes, hypertension, smoking, pre-diagnosis lipid measurements) with post-diagnosis cognitive decline. Taking diabetes as an example, the dependent variable was the composite cognitive score, and predictor variables were diabetes, time (in years), and a diabetes*time interaction. Age at incidence, sex, race/ethnicity, education (years) and study cohort were simultaneously introduced into models. A significant diabetes effect would suggest a diabetes-associated difference in cognition at diagnosis. A significant time effect would suggest change in cognitive scores over time (regardless of diabetes status). A significant interaction term would suggest differential rates of post-diagnosis cognitive change associated with diabetes.
The time interval between lipid measurement and AD incidence was included as a covariate in the lipid models.
To determine whether any vascular factor was independently associated with cognitive decline, we constructed a post-hoc model that simultaneously included all variables associated with cognitive decline after age- or multivariable-adjustment (cholesterol and diabetes)(Model 1), their interactions with time, and demographic factors. Diabetes, LDL-C and HDL-C, (but not total cholesterol), were included in this model.
We repeated our analyses within APOE-ε4 strata in the subsample with genotyping data (n=132). We also examined whether use of lipid lowering agents (LLAs) was associated with cognitive decline, and whether cholesterol predicted rate of decline similarly among LLA users vs. non-users.
Characteristics of the overall sample are presented in Table 1. At diagnosis, most participants (93%) had a Clinical Dementia Rating of 1.0 (mild) 29. The mean follow-up time between diagnosis and last follow-up was 3.5 years (SD 2.2; range 1.0 – 10.3), with an average of 1.6 post-diagnostic assessments (SD 0.9; range 2 – 5).
There were no sex-related differences in predictor variables, except for higher total cholesterol among women (201.16 vs. 184.53 mg/dl; p=0.01).
At diagnosis, non-Hispanic Whites were on average 2.6 years older than African-Americans (p= 0.26) and 3.4 years older than Hispanic participants (p = 0.03). Prevalent disease variables (diabetes, heart disease, stroke, and hypertension) did not vary by race/ethnicity. African-Americans had higher mean HDL-C (55.3 mg/dl) than Whites (43.8 mg/dl; p=0.004) and Hispanics (43.2 mg/dl; p<0.001), and lower triglycerides (mean 139.5 mg/dl) than Whites (207.1 mg/dl; p= 0.001) and Hispanics (186.1 mg/dl; p <0.001). There were no race/ethnic differences in sex, total cholesterol, LDL-C or APOE-ε4 status (data not shown).
A GEE model indicated an overall decline in composite cognitive score of 8% of a standard deviation (SD) per year (β=−0.08; p<0.001). A quadratic term for time, added to the model to test whether cognitive change was non-linear, was non-significant.
Results of Model I for each predictor are presented in Table 2. Higher total cholesterol and LDL were associated with faster cognitive decline. Each 10mg/dl unit increase in total cholesterol or LDL was associated with an additional 10% SD decline in cognitive score per year (p<0.001 for total cholesterol; and p=0.003 for LDL-C). Neither HDL nor triglycerides were associated with cognitive decline. Of the medical history variables, only diabetes was associated with faster decline (β=−0.050; p=0.045).
In a post-hoc model that simultaneously included HDL, LDL and diabetes and their interactions with time, only LDL was independently associated with faster cognitive decline (Table 3).
Among those with genotyping data, the presence of at least one APOE-ε4 allele was associated with faster cognitive decline (Model 1) (multivariable-adjusted beta for interaction with time: β= −0.080, p=0.05). APOE-ε4-stratified models revealed that, among ε4 non-carriers (n=89), higher pre-diagnosis total cholesterol, LDL, and HDL were associated with faster cognitive decline (multivariable-adjusted betas for interactions with time: total cholesterol β= −0.001, p<0.001; LDL β=−0.001, p=0.02; HDL β= −0.126, p=0.01). Triglycerides and medical history variables were not associated with cognitive decline (data not shown). Among ε4 carriers (n=41), higher total cholesterol (β=−0.002, p=0.03), higher LDL (β=−0.002, p=0.03), stroke (β=−0.178, p=0.02), and heart disease (β= −0.187, p=0.001) were associated with faster decline. No associations were seen with the remaining vascular variables (data not shown).
The use of LLAs was not associated with rate of cognitive decline. Models of the association of cholesterol with cognitive decline were repeated stratified by LLAs use and results were similar (data not shown).
There has been intense interest in identifying modifiable AD risk factors, such as cardiovascular risk factors, with an eye toward prevention or at least delaying disease onset 4. Yet little attention has been given to the influence of these factors on disease progression. Cerebrovascular lesions, common in AD patients, may accelerate the clinical manifestation of AD30. Vascular risk factors may increase oxidative stress or activate a neuroinflammatory response, triggering amyloid production. Thus it has been suggested that Alzheimer pathology and cerebrovascular disease may work synergistically to cause cognitive decline 31.
Consistent with previous research 32, we found that higher total cholesterol and LDL were associated with faster cognitive decline among AD patients. An earlier study of the WHICAP cohort found higher total cholesterol was associated with lower risk of incident AD 10. The prodromal stage of AD is associated with decreased plasma cholesterol, possibly due to dietary insufficiency and weight loss. By limiting the current study to incident cases, we may have eliminated bias and/or confounding in cholesterol levels associated with frailty and preclinical AD. This may explain the difference with our findings for incident AD.
Lipid-lowering agents (LLAs) have previously been associated with slower cognitive decline among AD patients 33. Nonetheless, consistent with others 13, we found no association between LLAs and cognitive decline.
Among the medical history variables, only diabetes was associated with faster cognitive decline. Diabetes may influence AD progression via an inflammatory mechanism, or by contributing to amyloid plaque and neurofibrillary tangle formation 34. Stroke, heart disease, hypertension, and smoking history were not associated with disease progression in the overall sample.
The ApoE-ε4 genotype may contribute both to vascular disease and AD neuropathology, with effects of vascular risk factors more pronounced among ε4 carriers35. A previous study of this cohort found faster cognitive decline among participants with mild AD who were ε4 carriers36. In our ε4-stratified models, higher total cholesterol and higher LDL-C predicted faster decline in both groups, while history of heart disease and/or stroke predicted faster decline among ε4 carriers only.
Few studies have examined the simultaneous effects of multiple vascular risk factors on AD progression. A study of prevalent AD found stroke, but not other vascular factors, was associated with faster decline on the MMSE 13. A study of incident cases found faster decline on the Clinical Dementia Rating and the MMSE among those with a history of atrial fibrillation, systolic hypertension and angina at baseline, while diabetes was associated with slower decline 14. Our findings may differ from these studies for several reasons: we measured cognitive change using a comprehensive battery of cognitive tests, which is potentially more sensitive than the MMSE; one of the previous studies13 included some participants with relatively advanced AD; and both of the previous studies were limited to Caucasian participants who, on average, were more highly educated than our multiethnic sample. In our post-hoc model, only higher LDL emerged as an independent predictor. The negative finding for diabetes as an independent predictor may be due to reduced power in the post-hoc analysis, since the beta value associated with diabetes-related cognitive change was similar to the diabetes-specific model.
This study has limitations. Since disease history (stroke, heart disease, diabetes) was self-reported, the prevalence of these conditions in our sample was likely underestimated. Further, 44 of the participants with incident AD were not included in this analysis since they died before the next follow-up assessment. Cause of death was unavailable, however it is likely that many of these deaths were due to vascular disease. If so, the effect size we found may be an underestimate. We had only one lipid assessment, potentially resulting in measurement error. Although lipids were measured prior to diagnosis, some participants may have been in the prodromal phase of AD. We attempted to account for this by adjusting for the interval between lipid measurement and diagnosis; results were unchanged. Similarly, AD-associated reductions in blood pressure may have masked an association between hypertension and disease progression that might have been seen had blood pressure been measured at mid-life. In future longitudinal studies, it would be better to measure these factors earlier in life. Our AD diagnoses were not neuropathologically confirmed, and imaging was not consistently used as part of diagnosis. Previous studies using MRI and neuropathogy data suggest that, especially among the oldest old, mixed dementia (AD + Vascular dementia) is the most common cause of dementia 37 Therefore, it is likely that some of our AD patients actually had mixed dementia. Finally, there is always the possibility that our findings could be due to chance.
Despite these limitations, clinical diagnosis was based on uniform application of widely accepted criteria, and unlike cognitive screening instruments used in some studies, our cognitive assessment comprised a comprehensive battery of tests evaluating a range of cognitive domains. Our use of a population-based sample limited to incident cases reduced biases associated with convenience samples (disease registries, hospital- or clinic-based samples) that may not accurately reflect the course of the disease in the general population.
In conclusion, we found that higher pre-diagnosis cholesterol levels (total cholesterol and LDL-C) and prevalent diabetes were associated with accelerated post-diagnostic cognitive decline. Heart disease history and stroke history predicted faster decline among ε4 carriers only. Prevention or treatment of these conditions can potentially slow the course of AD.
Dr. Helzner prepared this manuscript. Drs. Stern and Luchsinger were instrumental in the study design. Drs. Helzner, Scarmeas, Cosentino, Luchsinger, Glymour, Brickman and Stern were involved in data analysis and interpretation, and critical manuscript review. Dr. Helzner had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis. This research was supported by federal grants P01-AG07232, AG00261, RR00645, 5T32NS007153-22, and the Taub Institute for Research in Alzheimer’s Disease and the Aging Brain. We thank Nicole Schupf, PhD for reviewing this manuscript.