APOE genotype, plasma lipids, lipoproteins, and AD in community elderly

Methods.

The population for this study was derived from a random sample of 2,128 Medicare beneficiaries aged 65 years or older and residing in the community of Washington Heights in New York City. Of these, 273 (12.8%) refused to participate, 159 (7.5%) were lost to follow-up, 231 (10.8%) died before follow-up was completed, and 16 (0.75%) were seen but did not meet study criteria because they were too young or had moved. A total of 1,449 people (68%) were eligible. We additionally excluded 190 with stroke (8.9%), 15 with a diagnosis of PD (0.7%), and 6 with a diagnosis of both stroke and PD (0.3%). This left 1,238 individuals (58%) who were relatively healthy or had AD established at the initial visit or during the 2.5 years of study follow-up. The diagnosis of AD was based on National Institute of Neurological Disorders and Stroke–Alzheimer’s Disease and Related Disorders Association criteria.15

For this analysis we also excluded individuals with a history of diabetes mellitus (n = 131), using either insulin or oral hypoglycemic agents, antihyperlipidemics (n = 60), or oral estrogen replacement therapy (n = 14) because of their effect on cholesterol metabolism.16 Of these 190 individuals excluded, 13 were taking both antihyperlipidemics and insulin, and 2 were taking both insulin and oral estrogen. This left 1,048 individuals (49%) available for study. Because of refusal or inadequate fasting before blood draw, lipid levels were available on only 987 of the 1,048 people included in this study. This included roughly equal proportions from each diagnostic group. The final 987 accounted for 46% of the initial 2,128 individuals.

Data were also collected on age, gender, ethnicity, APOE polymorphism, body mass index (BMI), both systolic and diastolic blood pressure, smoking history, ischemic heart disease, aspirin use, previous stroke (CVA), or MI. Age was recorded at the beginning of the study for all subjects, and the interval at which the diagnosis of AD was established for incident cases. CVA was defined as history or clinical evidence of stroke (including imaging), and ischemic heart disease was defined as history of MI or angina. EKG abnormalities were not used in diagnosis. BMI was calculated (BMI = weight [kg]/height [m²]). Blood pressure measurements were obtained and hypertension defined as resting systolic >140 or diastolic >90 mm Hg. Blood pressure and BMI were measured at baseline and annually, but only baseline measurements were used in the analysis.

Participants were instructed to fast overnight before blood samples were drawn, usually in the morning. TC and triglyceride (TG) plasma levels were determined using standardized enzymatic procedures in a Hitachi 705 automated spectrophotometer (Boehringer Mannheim, Germany). High-density lipoprotein (HDL) cholesterol was analyzed after precipitation of apoB-containing lipoproteins and phosphotungstic acid.17 LDL cholesterol levels were calculated using a standard formula.18 The laboratory participated in the Centers for Disease Control Lipid Standardization Program, and interassay coefficients of variation were 2% for TC and TGs and 3% for HDL cholesterol.

APOE genotype was determined from genomic DNA essentially as described by Hixson and Vernier.19 APOE polymorphisms were initially analyzed separately, then grouped according to presence or absence of the ε4 allele.

We used the chi-square test to analyze categorical variables and analysis of variance for continuous variables. A Bonferroni correction was used where multiple comparisons were made in post hoc tests to determine specific differences. A p value of 0.05 was used in all analyses. When multiple comparisons were made, the p value, which varied depending on the number of comparisons made, was corrected to p = 0.05 in all final results.

Each lipid and lipoprotein fragment was first analyzed as a continuous variable. Subsequent analyses included each plasma level separated into quartiles to compare the lowest and highest groups. The latter analysis was employed to conserve statistical power. For prevalent cases, logistic regression was used to compute the odds ratio (OR) associated with AD, controlling for APOE genotype, age, sex, ethnicity, BMI, and other covariates. For incident cases, the Cox proportional hazards model was used with the same covariates to compute relative risk (RR) for incident AD. However, the time-to-event variable was age at onset of AD for the incident cases, eliminating the need for age adjustment. The end point for prospective analysis was determined as age at time of diagnosis or age at the end of the study period. Terms that did not significantly influence results were removed from the model in a stepwise fashion. Final models were based on lipid levels and variables that met p value <0.05 criterion for inclusion in stepwise analyses.

Results.

There were 309 men (31%) and 678 women (69%). The majority were of Caribbean Hispanic origin (60.3%), with the remaining defining themselves as white (22.1%), African American (17.2%), or other (0.4%) (table 1). Although the ethnicity defined as “other” was only 0.4% of the population (4 people), the group was kept separate to maintain the integrity of the sample. The group is excluded from the tables because small size precluded meaningful analysis. The total population of men was slightly but significantly younger than women (men 74.9 ± 6.0 years versus women 76.1 ± 6.6 years, p < 0.05, table 2). The total Hispanic population was significantly younger than the whites (see table 2).

There were 178 patients with prevalent AD and 129 patients with incident AD (see table 1). Twenty percent of women were diagnosed with AD at study onset, but their incident case percentage was similar to the overall population (see table 1). The percentage of AD in African Americans was significantly higher than that of whites (see table 1). A significant age difference existed between controls and prevalent and incident cases (controls 74.2 ± 5.7 years, prevalent AD 80.4 ± 7.1 years, and incident AD 77.5 ± 6.0 years; p < 0.05; see table 2). This difference in age in AD patients as compared with controls was significant across both gender and ethnicity (see table 2).

No significant variation in APOE allele frequency existed with age,7 the APOE-ε4 allele frequency was similar in the total population of men and women7 but significantly higher in African Americans when compared with Hispanics and whites (table 3). As in our previous study, the APOE-ε4/ε2 genotype was associated with significantly higher HDL levels, and the APOE-ε3/ε2 genotype had significantly lower LDL levels compared with the ε3/ε3 genotype (data not shown, reference 7). The absence of the APOE-ε4 allele was predictive of lower TC and LDL levels in prevalent cases compared with controls.

We have previously described the distribution of lipids and lipoproteins in the various demographic subgroups in this population.7 Men had significantly lower TC, LDL, and HDL levels as compared with women (TC: men 188.5 ± 36.8 mg/dL, women 207.7 ± 39.4, p < 0.05; LDL: men 110.2 ± 33.4 mg/dL, women 122.1 ± 35.3, p < 0.05; HDL: men 43.3 ± 12.8 mg/dL, women 50.4 ± 15.4, p < 0.05). As compared with whites, Hispanics had significantly lower TC and LDL levels (table 4). As compared with whites and Hispanics, African Americans had significantly higher HDL and significantly lower TG levels (see table 4).

Compared with elderly controls, both incident and prevalent AD patients had significantly lower TC and TG levels, but no difference in HDL and LDL levels (TC: controls 204.1 ± 38.2 mg/dL, prevalent AD 196.3 ± 45.0, incident AD 196.0 ± 38.0, p < 0.05; TG: controls 182.0 ± 90.6 mg/dL, prevalent AD 156.5 ± 80.8, incident AD 166.4 ± 70.7, p < 0.05; HDL: controls 47.9 ± 14.8 mg/dL, prevalent AD 48.7 ± 16.8, incident AD 48.7 ± 13.2, p = 0.7; LDL: controls 119.8 ± 34.3 mg/dL, prevalent AD 115.9 ± 38.5, incident AD 113.9 ± 33.9, p = 0.13). When adjusting for age and education, only TG levels remained significantly different between controls and prevalent cases. When further adjusting for genotype and ethnicity, there was no difference between groups for any of the lipid profile.

Further multivariate analyses were done separately for prevalent and incident cases because the analysis of prevalent cases may reflect survival with AD rather than risk for the disease. Initial analysis by logistic regression of the prevalent AD group revealed no association with blood pressure, aspirin use, CVA or MI, BMI, or gender. Age, ethnicity, genotype, and education were significantly and independently associated with AD and were used as covariates in subsequent analyses. Separated into quartiles, only TC was significantly associated with AD (table 5). The lowest quartile of TC showed an increased OR for AD compared with the highest (OR = 1.6, 95% confidence interval [CI] = 1.0 to 2.4, p = 0.04; see table 5). When adjusted for age, the association diminished (OR = 1.3, 95% CI = 0.8 to 2.1, p = 0.3; see table 5). Ethnicity, genotype, and education had no effect on the OR for cholesterol quartiles.

For incident cases, we found no significant influence of blood pressure, aspirin use, CVA or MI, BMI, or gender on the risk for developing AD. The risk of incident AD was associated with the lowest TC quartile compared with the highest (RR = 1.8, 95% CI = 1.1 to 2.9, p = 0.02; table 6). The inclusion of education slightly decreased the association (see table 6). Ethnicity and APOE genotype also had significant effects on AD risk (genotype comparing ε4/ε4 with all other genotypes RR = 4.35, 95% CI = 1.7 to 11.7, p = 0.003; ethnicity comparing African Americans with all other ethnicities RR = 2.9, 95% CI = 1.6 to 5.2; p = 0.0005), but neither affected the risk of AD associated with low TC levels.

Discussion.

Initially, significantly lower TC and TG levels were seen in individuals with prevalent and incident AD compared with controls. Adjusted for age, education, APOE genotype, and ethnicity, patients with AD did not have plasma lipid levels that were significantly different from disease-free individuals. In our analysis only a small decrease in TC levels was observed in individuals diagnosed who developed AD during the follow-up interval. Genotype and ethnicity did not change the risk ratio for this association, suggesting that any changes in lipid levels associated with disease are independent.

The reasons for these findings are not clear. The short time course of our study may have precluded assessing changes in lipid levels as disease becomes more severe, which would reflect survival with AD. Given the time course of the study, and our knowledge of changes in relationship between APOE genotype and lipid levels with aging, it is unlikely that we have missed an earlier important relationship between lipid levels and subsequent development of AD.

ApoE is synthesized in the brain. Its physiologic and possible pathophysiologic role has not been clarified, although a number of possible functions, such as role in damage repair in the central and peripheral nervous system,20 induction of alterations in cholineacetyltransferase activity,21 processing of the amyloid precursor protein,22 and accumulation of the amyloid β protein,23 have been suggested. It should also be emphasized that a role of apoE in the nervous system may differ from its role in lipid and lipoprotein metabolism. Although at a population level, variation at the APOE locus affects plasma lipid levels to a considerable extent, more than 85% of the variation in cholesterol levels in the general population is not APOE genotype-dependent.4 Thus, potentially important genotype-specific effects of apoE in the nervous system may not necessarily be reflected in measurement of blood lipid levels. Further studies both at the population as well as the molecular level are needed to help clarify the role of apoE.