Risk Estimates of Dementia by Apolipoprotein E Genotypes From a Population-Based Incidence Study: The Rotterdam Study FREE

ALTHOUGH THE association between the ϵ4 allele of the apolipoprotein E (APOE) gene and Alzheimer disease (AD) is well established, reported risks vary tremendously among studies.^1- 7 Most of these investigations are of clinic-based patient series. In some studies, selection bias may have played a role because the ϵ4 frequency in memory clinic patients was found to be increased compared with patients in the general population.^8- 9 Furthermore, all but 2 studies^4- 5 are based on prevalent cases. Because the APOE genotype may be a determinant of survival in patients with dementia,^10- 11 cross-sectional studies are subject to bias. In addition, it is still controversial whether the ϵ2 allele exerts a protective effect,^11- 12 whether a family history of dementia further increases the ϵ4-related risk,^2- 3 and whether the APOE genotype is a determinant of the patient's age at the onset of dementia.^1,13 Thus, reliable estimates of the risk of dementia associated with the APOE genotypes are currently not available, and it is not yet clear which part of all patients with dementia is attributable to the APOE genotypes.

The aims of this study were to provide risk estimates of dementia and AD as a function of the APOE genotype and to estimate the proportion of dementia and AD that can be attributed to the APOE genotypes. To overcome selection bias, we investigated these issues in a prospective study of a general population initially free of demetia.

The Rotterdam Study is a population-based prospective cohort study among persons aged 55 years or older¹⁴ for which approval was given by the local medical ethics committee. Participants were recruited from 10275 eligible residents of a suburb of Rotterdam, the Netherlands, including institutionalized persons. A total of 7983 participants (response rate, 77.7%) were examined at baseline. Of these, 7528 persons (94.3%) were evaluated for dementia, of whom 474 received diagnoses and were excluded from the present study. At follow-up, after an average (SD) of 2.1 (0.9) years, 6315 persons (79.1%) were reexamined; 768 subjects (9.6%) had died, 106 (1.3%) were unavailable, and 794 (10%) refused to participate. Follow-up of all subjects who were not examined in person was completed by evaluating medical files. Of the 7054 participants who did not have dementia at baseline, 162 were found to have dementia at follow-up.

The analyses presented here are restricted to persons from whom blood specimens were available for APOE typing. This was successfully determined in 134 patients with incident dementia and in a random sample of 997 nondemented control subjects.

For the assessment of dementia, we used the same protocol at baseline and at follow-up.^15- 16 Briefly, all subjects were screened on cognitive functioning. Persons with abnormal results on screening underwent further neuropsychological testing, and an informant was interviewed on the daily functioning of the participant. Persons who were suspected of having dementia were examined by a neurologist, underwent neuropsychological testing, and if possible, had a magnetic resonance imaging scan of the brain.^15- 16 For subjects who could not be reexamined in person, information was obtained from general practitioners and the Regional Institute for Outpatient Mental Health Care, Rotterdam, which covers the entire study population. In the Netherlands, the Regional Institute for Outpatient Mental Health Care is consulted for social and psychiatric problems and is responsible for assessing indications for going to a dementia care facility. Examinations by the institute include (informant) interviews, neurologic and neuropsychological testing, blood biochemistry tests, and serologic tests for syphilis.

The diagnosis of dementia was made by a panel of study physicians, a neurologist, and a neuropsychologist, who reviewed all existing information and used criteria of the Diagnostic and Statistical Manual of Mental Disorders, Third Edition, Revised.¹⁷ A diagnosis of AD and vascular dementia was based on criteria of the National Institute of Neurological and Communicative Disorders and Stroke and the Alzheimer's Disease and Related Disorders Association,¹⁸ and of the National Institute of Neurological Disorders and Stroke and the Association Internationale pour la Recherche et l'Enseignement en Neurosciences.¹⁹ Cerebrovascular disease was recognized by a history of stroke, an evident transient ischemic attack, or evidence of vascular damage on brain scans.

At baseline, subjects were interviewed about the attained level of education and their family history of dementia. All first-degree relatives of the participants were listed, and for each relative, the presence of dementia was evaluated with the study physician. The family history was affirmed if 1 or more first-degree relatives had dementia.

Genotyping for APOE was performed on coded DNA specimens without knowledge of the diagnosis. A polymerase chain reaction was performed using previously published primers.²⁰ The reaction mixture contained 200-ng genomic DNA as the template, 25 pmol of both primers, 0.1 U of DNA polymerase, fluorescent sequencing (AmpliTaq, Applied Biosystems, Foster City, Calif), magnesium chloride (1.5 mmol/L), Tris(hydroxymethyl)-aminomethane hydrochloride (pH 9.0) (75 mmol/L), ammonium sulfate (20 mmol/L), and 10% dimethyl sulfoxide in a total volume of 25 µL. Thirty-five cycles of 30 seconds each at 94°C, 65°C, and 70°C were performed, and the polymerase chain reaction amplification products were digested with 5 U of HhaI for 3 hours at 37°C. The restriction fragments were separated on precast gels (ExcelGel, Pharmacia Biotech, Uppsala, Sweden) by electrophoresis (MultiPhorII, Pharmacia Biotech) for 1½ hours at 600 V. The restriction fragments were visualized by silver staining. The results were read by 3 persons independently. In case of discrepancies, APOE genotyping was repeated.

The relative risk of dementia or AD was estimated as an odds ratio (OR) in a multiple logistic regression model and presented with a 95% confidence interval (95% CI) using the ϵ3/3 genotype as the reference. To compute the OR associated with carriership of the ϵ2 or ϵ4 allele, persons with ϵ2/2 or ϵ2/3 were grouped, as well as subjects with ϵ3/4 or ϵ4/4; persons carrying both the ϵ2 and the ϵ4 allele were excluded from these models. To adjust for confounding, age, sex, and education level were added to the above models. Education was dichotomized as primary school or less, and more than primary school. The proportion of patients with dementia that is attributable to 1 of the APOE genotypes was assessed with the population-attributable risk.²¹

The age at the onset of dementia was taken to be the midpoint between the baseline age and the age at follow-up. Differences in age at the onset of dementia were examined in a multiple linear regression model, and the percentage of explained variance was estimated by the squared adjusted multiple correlation coefficient.²²

The lifetime risk of dementia was calculated for 55-year-old subjects with or without the ϵ4 allele using a matched Bayesian analysis, as described earlier.²³ As patients were on average older than the controls, and the frequency of the APOE genotypes differed across age strata, we first matched every patient with a control subject on age (5-year interval) and sex. The a priori lifetime risk of dementia for 55-year-old persons was derived from The Rotterdam Study. Overall, this risk was estimated to be 33% for women and 16% for men, taking into account the competing risk of death.¹⁶

Characteristics of the study population are shown in Table 1. Patients with dementia were on average older, more often female, and more often had primary education only compared with the unaffected controls. Overall, a family history of dementia was not more common in patients than in controls. Patients with dementia, however, more often had 2 or more first-degree relatives with dementia than the unaffected controls (5% vs 2%; χ²=4.1, P=.04). The distribution of the APOE genotypes in patients and controls was in Hardy-Weinberg equilibrium.

As shown in Table 2, the OR for all dementia associated with ϵ4 homozygosity was 11.2 (95% CI, 3.6-35.2) and for AD, 6.2 (95% CI, 1.4-28.2). Persons with the ϵ3/4 genotype had almost a 2-fold increased risk of all dementia and AD compared with those with the ϵ3/3 genotype. Carriers of the ϵ2/3 genotype had a decreased risk of these disorders, although this did not reach statistical significance.

It was estimated that 5% (95% CI, 1%-8%) of cases of dementia could be attributed to the ϵ4/4 genotype, 13% (95% CI, 7%-19%) to the ϵ3/4 genotype, and 2% (95% CI, 0%-5%) to the ϵ2/4 genotype. Thus, if the ϵ4 allele did not have an effect on the dementia risk, the incidence would be reduced by 20%. Likewise, it was assessed that 3% (95% CI, 0%-6%) of AD cases were attributable to ϵ4 homozygosity, 14% (95% CI, 7%-21%) to the ϵ3/4 genotype, and 0% (95% CI, 0%-2%) to the ϵ2/4 genotype.

Among patients with dementia, ϵ2 carriers were on average older and ϵ4 carriers younger at the onset of dementia than persons homozygous for ϵ3 (Table 2). The APOE genotype explained 10% of the variance in age at the onset of dementia and 4% of the variance in age at the onset of AD.

The ϵ4-associated OR for dementia and AD was highest in the youngest age category and decreased with aging (Figure 1). The effects of the ϵ2 allele were similar across age strata (data not shown). The ϵ4-related OR for dementia in women (2.3; 95% CI, 1.2-4.5) was comparable with the corresponding OR in men (1.9; 95% CI, 0.8-4.2). Analyses of the sex-specific OR for AD yielded similar results as well (data not shown).

A family history of dementia further increased the ϵ4-associated OR for dementia and AD (Table 3). The risk of these disorders for ϵ4 carriers with no family history of dementia was nonsignificantly increased (for dementia: OR, 1.7 [95% CI, 1.0-3.1]; for AD: OR, 1.3 [95% CI, 0.7-2.7]). Among persons with a family history of dementia, however, ϵ4 carriers had a statistically significant 4.8-fold increased risk of dementia (95% CI, 1.5-15.6) and a 4.6-fold increased risk of AD (95% CI, 1.3-16.1).

The lifetime risk of dementia for a 55-year-old carrier of at least 1 ϵ4 allele was estimated to be 26% for men and 46% for women. For non-ϵ4 carriers at a similar age, the lifetime risk was estimated to be 11% for men and 28% for women. The numbers were too small to estimate the lifetime risk of dementia associated with each APOE genotype separately.

With this prospective population-based study, we provide risk estimates for incident dementia and AD for carriers of the various APOE genotypes. Our study shows that the proportion of patients with dementia that is attributable to the APOE genotype was approximately 20% and that the APOE genotypes explained not more than 10% of the variance in age at the onset of dementia.

The present investigation has several advantages over most previous studies on the APOE genotype and dementia. Selection bias was avoided because the study was population based and had a high response rate. The diagnostic workup was comprehensive and was similar at baseline and at follow-up. For participants who could not be reexamined in person, information was obtained from general practitioners and ambulatory psychiatry services to have a complete follow-up of the cohort. We minimized survival bias by studying the relation longitudinally and by including subjects who died during follow-up. This was possible because APOE genotyping was performed on blood specimens collected before the onset of dementia.

A limitation of this study is the lack of autopsy confirmation. Therefore, we cannot rule out that we misclassified the dementia subtype in some instances. With clinical examination alone, however, approximately 90% of cases of AD are diagnosed correctly.²⁴ Furthermore, any misclassification of dementia subdiagnoses will not affect analyses on dementia in general.

We found a higher ϵ4/4-associated OR for all dementia (11.2) than for AD (6.2). In addition, the APOE genotypes explained a higher proportion of the variance in age at the onset of dementia (10%) than of AD (4%). Although these observations are tentative given the lack of neuropathological confirmation of the dementia subtype and the overlapping confidence intervals, these findings suggest that the ϵ4 allele is strongly associated with the non-AD dementias.²⁵

We further found that carriers of ϵ2/3 had a nonsignificantly decreased risk of dementia and AD and were older at the onset of dementia. This is compatible with a protective effect, as has been reported previously.¹² Among patients with early-onset AD, we found an increased risk for ϵ2 carriers.¹¹ In the present investigation, the effect of the ϵ2 allele on early-onset AD could not be addressed because all but 3 of the patients were older than 65 years at the onset of dementia.

We confirmed that the ϵ4-associated risk may decrease with advancing age.¹² Although this is in contrast to findings from another population-based incidence study,⁵ the latter investigation is smaller than our study. We found, further, that the ϵ4-related risk of dementia was similar in men and women. This is in contrast with a study on familial AD.²⁶ The interpretation of that study is hampered by the fact that it is based on a cross-sectional design,²⁶ and sex-specific differences in cardiovascular mortality related to the ϵ4 allele may have played a role.²⁷ A family history of dementia may further increase the ϵ4-associated risk of dementia and AD. The difference in the ϵ4-related risk of dementia between persons with and without a family history of dementia did not result from an asymmetric distribution of ϵ4 homozygotes across these strata. Also, analyses on the 6 APOE genotypes yielded a higher OR for dementia in those with a family history of dementia than in those without, although numbers were small (data not shown). This suggests the presence of other genetic or shared environmental factors in the cause of dementia that may interact with the effect of the ϵ4 allele.^2- 3

Only 2 previous studies reported lifetime risks of dementia for carriers and noncarriers of the ϵ4 allele specifically.^7,23 These risk estimates for ϵ4 carriers were lower (0.01-0.29) than ours.^7,23 The 2 studies, however, based their estimates on a meta-analysis of different case series, but the a priori risk of dementia was drawn from yet another study.^7,23 We found a lifetime risk of dementia for 55-year-old ϵ4 carriers of 46% in women and 26% in men. In the absence of the ϵ4 allele, the lifetime risk was estimated to be 28% for women and 11% for men. In these calculations, the competing risk of death and the higher a priori risk in women were taken into account. In asymptomatic persons, the presence of the ϵ4 allele seems to have limited predictive value. Although there is consensus in the literature that APOE genotyping should not be used for presymptomatic testing,^28- 29 with the expanding use of APOE genotyping for diagnostic purposes, a growing number of unaffected offspring will be aware of their APOE genotype. Therefore, reliable estimates of the lifetime risk of dementia will become important.

Although the association between the ϵ4 allele and dementia is robust, the incidence would be reduced by only 20% as the effects of ϵ4 could be blocked. In addition, the variance in age at the onset of dementia was only partly (10%) determined by the APOE genotypes. These findings suggest that other factors play an important role in the cause of dementia. The higher ϵ4-related risk in persons with than in subjects without a family history of dementia emphasizes the need to identify other genes or shared environmental factors.

Accepted for publication November 18, 1997.

This study was supported by the NESTOR Stimulation Program for Geriatric Research in the Netherlands (Ministry of Health and Ministry of Education), the Netherlands Organization for Scientific Research, and the Netherlands Prevention Fund, The Hague, the Netherlands; the municipality of Rotterdam, the Netherlands; the Flemish Biotechnology Program, Brussels, Belgium; and by a postdoctoral fellowship with the Fund for Scientific Research–Flanders, Brussels, Belgium (Dr Cruts).

We thank the general practitioners of Ommoord, Rotterdam, the Regional Institute for Outpatient Mental Health Care, and Rijnmond Noord-Oost, Rotterdam, for providing medical information of participants. We also thank Hubert Backhovens, lng, Marleen Van den Broeck, Anita Wehnert, and Sally Serneels for APOE genotyping. Inge de Koning, MSc, Maarten de Rijk, MD, PhD, Alewijn Ott, MD, PhD, and Frans van Harskamp, MD, are acknowledged for their role in diagnosing dementia in patients.

Reprints: Arjen J. C. Slooter, MD, Department of Epidemiology and Biostatistics, Erasmus University Medical School, PO Box 1738, 3000 DR Rotterdam, the Netherlands (e-mail: slooter@epib.fgg.eur.nl).