Estimated lifetime risk (LTR) for Alzheimer disease among male and female first-degree relatives of probands (vertical bars show SE at each age value).
Estimated lifetime risk (LTR) for Alzheimer disease among first-degree relatives of probands with ϵ3/ϵ3, ϵ3/ϵ4, and ϵ4/ϵ4 genotypes (vertical bars show SE at each age value).
Estimated lifetime risk (LTR) for Alzheimer disease among first-degree relatives of probands with ϵ3/ϵ3, ϵ3/ϵ4, and ϵ4/ϵ4 genotypes (vertical bars show SE at each age value). Top, Female relatives. Bottom, Male relatives.
Martinez M, Campion D, Brice A, Hannequin D, Dubois B, Didierjean O, Michon A, Thomas-Anterion C, Puel M, Frebourg T, Agid Y, Clerget-Darpoux F. Apolipoprotein E ϵ4 Allele and Familial Aggregation of Alzheimer Disease. Arch Neurol. 1998;55(6):810-816. doi:10.1001/archneur.55.6.810
Copyright 1998 American Medical Association. All Rights Reserved. Applicable FARS/DFARS Restrictions Apply to Government Use.1998
To investigate the relationship among risk for Alzheimer disease (AD), familial aggregation of AD, and the apolipoprotein E (apoE) ϵ4 allele in first-degree relatives of probands with AD and known apoE genotype.
Two hundred ninety subjects fulfilling the criteria of the National Institute of Neurological Communicative Disease and Stroke–Alzheimer's Disease and Related Disorders Association for probable AD were ascertained from March 1, 1992, to December 31, 1996, through consecutive admissions in several university hospitals.
Design and Methods
Family data were collected on 1176 first-degree relatives (parents and siblings), aged 40 to 90 years. Most living relatives underwent a clinical examination, whereas we relied on family history for clinical data for deceased or unavailable relatives. First, we conducted standard survival analyses to estimate cumulative lifetime risk (LTR) for AD among relatives and to investigate for sex and apoE genotype effects on LTR. Then, we assessed to what extent clustering of secondary AD could be explained by the apoE ϵ4 allele by deriving the expected proportions of relatives with 0, 1, or 2 apoE ϵ4 alleles conditionally on the proband's genotype.
Cumulative LTR for AD among first-degree relatives increased significantly with the number of ϵ4 alleles present in the proband. By 90 years of age, LTRs in relatives of probands with ϵ3/ϵ3, ϵ3/ϵ4, and ϵ4/ϵ4 genotypes were 29.2%, 46.1%, and 61.4%, respectively. Significant sex-by-apoE genotype interaction effects on LTR were observed. Women had about a 2-fold higher risk for AD than men among relatives of ϵ4 carriers but not among relatives of non-ϵ4 carriers. The predicted proportion of ϵ4 carriers in relatives of probands with ϵ3/ϵ3 genotype remains about 50% lower than the corresponding LTR for AD, indicating that familial clustering of AD is largely due to other factors than the apoE ϵ4 allele. Although aggregation of AD in families of probands with the ϵ4 allele is more prominent, we estimated that AD would not develop in about 30% of female and up to 60% of male relatives carrying at least 1 ϵ4 allele, even by 90 years of age.
Our results support the hypothesis that the apoE ϵ4 allele enhances AD susceptibility, but putative factors enhancing risk for AD remain to be found.
APOLIPOPROTEIN E (apoE) genotype is a major risk factor for Alzheimer disease (AD). Numerous studies have reported an association between the apoE ϵ4 allele and late-1 and early-onset AD.2,3 A dose-related effect of the apoE ϵ4 allele has been suggested,4- 8 but not in a limited number of series of patients who had early-onset AD or who were nonwhite.2,9- 11 Results of Kaplan-Meier analyses suggested a decrease in age of onset of disease ranging from 8 to 16 years4,12,13 when the number of ϵ4 alleles increased from 0 to 2.
Other risk factors for AD such as sex, age, and family history also have been suggested. Several studies have assessed for sex differences in apoE-associated risk for AD, with conflicting results.8,14,15 However, the amount of AD risk due to the apoE ϵ4 allele is not precisely known. Indeed, due to ascertainment biases and severe truncation of data, current data derived from case-control designs might overestimate the magnitude of the effects of apoE ϵ4. Alternative approaches, such as linkage or segregation family studies, have not clarified the influence of apoE on AD risk.6,16,17 Indeed, these studies face difficult issues such as the presence of statistical interactions between risk factors and the absence of apoE genotypes among most relatives.
So far, 2 studies have assessed influence of apoE ϵ4 status on lifetime risk (LTR) for AD.8,18 Both analyses used estimates of LTRs associated with apoE genotypes; these were based on bayesian derivations using population-based apoE distributions and a priori AD risks derived from age-specific annual incidence rates reported by large epidemiological studies. Seshadri et al18 estimated that presence and absence of the apoE ϵ4 allele were associated with a 1.9 times greater risk and a 60% lower risk, respectively, than baseline. The LTR remained below 30%, however, even when an apoE ϵ4 allele was present. Similar variations in AD risks were observed by Bickeböller et al8; sex-specific risk estimates were higher for individuals with ϵ3/ϵ4 and ϵ4/ϵ4 genotypes than for individuals with ϵ3/ϵ3 genotypes, but lifetime risk estimates for ϵ3/ϵ4 genotypes reached only 30% in men and 28% in women by 85 years of age. However, incidence rates for genotype ϵ4/ϵ4 were not uniformly larger than those for genotype ϵ3/ϵ4 across all age groups, and apoE ϵ4 was shown to exert its maximal effects in the group aged 60 to 79 years. Similar age variations for effects of apoE ϵ4 have been reported.17,19 As indicated,18,20 the role of the apoE ϵ4 allele in AD across the age span needs to be clarified if apoE ϵ4 is to be used to predict disease in relatives of patients with AD. To further explore this issue, we investigated whether familial distribution of the apoE ϵ4 allele could account for the observed aggregation of AD in relatives of 275 probands with AD (aged 41-89 years) with known apoE genotypes. The LTRs for disease were evaluated first in 1176 first-degree relatives (parents and siblings, aged 40-90 years), stratified by the proband's apoE genotype to evaluate effects of sex and sex-by-apoE interaction on AD risks. Then, given the proband's apoE genotype and the observed age and sex distributions of relatives, we inferred expected proportions of apoE genotypes in relatives to evaluate the amount of secondary AD explained by apoE ϵ4.
Our study is based on the French Alzheimer Collaborative Group data. Unrelated patients (n=308) were ascertained, during 3 years, through consecutive admissions in several university hospitals.8 They all fulfilled the criteria of the National Institute of Neurological Communicative Disease and Stroke–Alzheimer's Disease and Related Disorders Association21 for probable AD. Age of onset for all patients with AD was assessed by interviewing next of kin. Age of onset was defined as the age at which the patient or the patient's family first noted the symptoms required for diagnosis. Family data were collected on first-degree relatives of 290 probands. Most living relatives (64.8%) were examined directly, whereas, for deceased or unavailable relatives, we relied on family history. Data were collected from a primary informant using a standardized questionnaire and were confirmed by interview of other informants. Medical records, when available, were systematically consulted. In all our analyses, relatives were considered to be unavailable for follow-up after 90 years of age.
In probands with early-onset AD (before 60 years of age), we delineated, using segregation analysis,22 a group of probands (n=18) explained by autosomal inheritance of a dominant AD gene with complete penetrance by 60 years of age (type 1 probands). These results were confirmed in subsequent molecular analyses; most patients with type 1 early-onset AD had positive results of testing for mutations in the presenilin 1 gene23; a few, for the mutation 717 val→ile in the amyloid precursor protein gene.24 These type 1 probands were excluded from our analysis.
The characteristics of the total cohort of non–type 1 probands with available data in first-degree relatives (n = 290) are given in Table 1. Our cohort of French patients with AD has a female-male ratio value of 1.7 and a mean ± SD age of onset of 62.9 ± 9.2 years. The characteristics of our French cohort are quite similar to those obtained in other cohorts. For instance, in a sample of 550 subjects with AD, including a population-based cohort of 171 subjects, the reported sex ratio and mean ± SD age of onset values were 1.6 and 65.3 ± 9.2, respectively.25
Genomic DNA was amplified using polymerase chain reaction (PCR) with amplification conditions and primers described by Hixson and Vernier.26 Polymerase chain reaction products were digested with restriction enzyme Hha I and subjected to electrophoresis on polyacrylamide gels.
The Kaplan-Meier method27 was used to estimate cumulative risk for AD among first-degree relatives, aged 40 years or older, of probands with AD. Standard errors for the cumulative escape were calculated using the Greenwood formula.28 Differences in cumulative risk curves were assessed for significance using a log-rank test. Effects of sex were evaluated using stratification of the relatives by sex. Probands were stratified by apoE genotype, sex, or age of onset.
Estimation of the familial aggregation of AD accounted for by the apoE ϵ4 allele depended, ideally, on the direct examination of apoE genotype for each first-degree relative. For a common late-onset disease such as AD, this investigation is more likely to be approached indirectly, since most parents and siblings are deceased or otherwise unavailable for study. We evaluated the amount of the observed familial AD risk that could be accounted for by the apoE ϵ4 allele by deriving likelihoods of apoE status in relatives. Indeed, under given assumptions (relatives are likely to share apoE alleles), missing apoE status in parents and siblings can be inferred using the proband's apoE genotype.
Proportions of the expected apoE genotype in first-degree relatives of patients with AD can be derived from the conditional probabilities of the possible mating types among parents of the probands with a specified apoE genotype. For our analyses, we assumed that the frequency of the ϵ4 allele in the parental chromosomes not transmitted to the proband was the same as in the general population. We used the apoE allele frequencies estimated in a large sample of French control subjects stratified by sex and age.8 Frequency of the ϵ4 allele in men vs women within each age group were as follows: 0.15 vs 0.19, respectively, for younger than 60 years; 0.09 vs 0.19, respectively, for ages 60 to 69 years; 0.09 vs 0.17, respectively, for ages 70 to 79 years; and 0.13 vs 0.16, respectively, for older than 79 years. Sex- and age-specific apoE ϵ4 allele frequencies were considered because the ϵ4 allele was shown to occur significantly less frequently in men than women,8 and because its frequency decreases with increasing age.29,30 Because the ϵ2 allele accounts for less than 6% of the polymorphism in the population, its frequency was combined with that for the ϵ3 allele. Thus, we modeled the apoE genotype as a 2-allele system with frequencies p and 1−p for the ϵ4 and non-ϵ4 (ϵx = ϵ2 or ϵ3) alleles, respectively. Hardy-Weinberg equilibrium among allele frequencies was also assumed.
For our analysis, Gp indicates the observed proband's apoE genotype (Gp = A1A2, with A1,A2 = ϵ3 or ϵ4); A1 and A2 correspond to the transmitted alleles. When the proband has the apoE genotype ϵ3/ϵ4 (ie, A1‘A2), the paternal (tF) and maternal (tM) transmitted alleles can be ϵ3 and ϵ4, respectively, or ϵ4 and ϵ3, respectively. When the proband is homozygous (Gp = ϵ3/ϵ3 or ϵ4/ϵ4), tF and tM are known (ie, tF = tM = ϵ3 or ϵ4). The nontransmitted paternal and maternal alleles (ϵ4 or ϵx) are denoted by nF and nM, respectively. Age F and age M are the ages (at examination or at death) of the father and mother, respectively. When both parents have unknown apoE genotypes, the conditional probability of each possible mating type was derived as follows:
for nF=ϵx or ϵ4 and nM=ϵx or ϵ4 and P(nF|),P(nM|) corresponded to the apoE allele (ϵ4 or ϵx) frequencies within each age-by-sex stratum estimated by Bickeböller et al.8
When parents had known apoE genotypes (ie, nF or nM known), the probability of each possible mating type, given the Gp, was computed by setting, in equation 1, the corresponding allele frequencies P(nF|) and/or P(nM|) to equal 1. When additional apoE data were available in siblings (Gs), derivations of conditional probability of each possible mating type were obtained by extending the probability of Gp (equation 2) to the probability of Gp and Gs. The proportions of apoE genotypes in parents were obtained from the conditional probabilities of possible mating type. For each possible mating type, the proportions of siblings with apoE genotype ϵx/ϵx, ϵx/ϵ4, or ϵ4/ϵ4 were obtained using mendelian expectations. These proportions were then multiplied by the conditional probability of the mating type and summed over all possible mating types. Finally, within each proband genotype group, the expected apoE genotype frequencies were summed for the observed distributions of parents and siblings within each age-by-sex stratum.
The distribution of 290 probands by apoE genotype, sex (female-male ratio), and AD family history are presented in Table 1. Of 290 probands, 162 (55.9%) were homozygous or heterozygous for the apoE ϵ4 allele; 11 (3.8%) were homozygous or heterozygous for the apoE ϵ2 allele; and 117 (40.3%) were homozygous for the apoE ϵ3 allele. The female-male ratio of probands was 1.7, with similar sex distribution by apoE genotype. The overall proportion of probands with AD and a positive family history was 38.3%; it was lower among probands without the apoE ϵ4 allele (0%, 10.0%, and 24.8% for ϵ2/ϵ2, ϵ2/ϵ3, and ϵ3/ϵ3 genotypes, respectively) than among those bearing the apoE ϵ4 allele (75.0%, 45.5%, and 58.7% for ϵ2/ϵ4, ϵ3/ϵ4, and ϵ4/ϵ4 genotypes, respectively). Because of the small number15 of probands with ϵ2/ϵ2, ϵ2/ϵ3, and ϵ2/ϵ4 apoE genotypes, all familial analyses were restricted to the family sample of the remaining 275 probands.
Within the 275 families of these probands, there were 1176 first-degree relatives (parents and siblings) aged 40 years or older (Table 2). The mean number of first-degree relatives by proband was 4.0, 4.5, and 4.4 for ϵ3/ϵ3, ϵ3/ϵ4, and ϵ4/ϵ4 genotypes, respectively, and the female-male ratio of relatives was close to 1. A total of 138 patients with secondary AD (38 men and 100 women) were identified in 107 (38.9%) of the 275 families (29, 51, and 27 families of probands with ϵ3/ϵ3, ϵ3/ϵ4, and ϵ4/ϵ4 genotypes, respectively). Thus, the crude rate of affected first-degree relatives in our cohort is 138 (11.7%) of 1176. Within each apoE genotype group, there were more affected women than men; this ratio varied from 2.4 (ϵ3/ϵ4 genotype) to 2.8 (ϵ3/ϵ3 genotype) to 2.9 (ϵ4/ϵ4 genotype). Similar trends were observed among parents and siblings, although, in families of probands with ϵ4/ϵ4 genotype, the ratio of affected mothers to affected fathers was the highest (5.2),5 whereas the ratio of affected sisters to affected brothers was the lowest (1.3).
Table 3 shows overall estimated cumulative risks for AD in the first-degree relatives of all probands and specific risks when families were stratified by the sex, age of onset, and apoE genotype of the proband. Table 3 also shows, for each family group, sex-specific cumulative risk estimates. By 90 years of age, the risk for AD in first-degree relatives of all 275 probands with AD was 42.1 ± 0.04%. Estimated cumulative risks in female and male relatives were equal to 46.7 ± 0.04% and 37.3 ± 0.10%, respectively; women had a significantly (log rank χ2= 7.9; P= .007) higher cumulative risk for disease than men (Figure 1). No significant differences in LTRs for AD among relatives were found when families were stratified by the proband's sex or age of onset.
Estimated cumulative risks for AD among relatives increased with the number of ϵ4 alleles present in the proband's apoE genotype (Figure 2); by 90 years of age, these estimates were equal to 29.2 ± 0.06%, 46.1 ± 0.07% and 61.4 ± 0.08% for relatives of probands with the apoE genotypes ϵ3/ϵ3, ϵ3/ϵ4, and ϵ4/ϵ4, respectively. For each proband's apoE genotype, cumulative risks were higher for female than for male relatives, but these sex differences were significant for the ϵ4/ϵ4 group of probands only (the relative risk for AD by 90 years of age for female relatives compared with that of male relatives was 2.1; log rank χ2=4.02; P=.04). Relatives of probands with ϵ3/ϵ3 genotype had a significantly lower cumulative risk for AD by 90 years of age than relatives of probands having at least 1 ϵ4 allele (log rank χ2=15.9; P<.001). Similar trends were obtained for female (log rank χ2=11.1; P<.001) and male (log rank χ2=4.6; P=.03) relatives. Pairwise comparisons showed significantly lower risk for AD among relatives of probands with ϵ3/ϵ3 genotype than among relatives of probands with ϵ3/ϵ4 (log rank χ2=8.5; P=.004) and ϵ4/ϵ4 (log rank χ2=24.5; P<.001) genotypes. Similar trends (Figure 3) were observed for female (ϵ3/ϵ3 vs ϵ3/ϵ4 genotype, log rank χ2=5.4 [P=.02]; ϵ3/ϵ3 vs ϵ4/ϵ4 genotype, log rank χ2=18.6 [P<.001]) and for male (ϵ3/ϵ3 vs ϵ3/ϵ4 genotype, log rank χ2=3.0 [P=.08]; ϵ3/ϵ3 vs ϵ4/ϵ4 genotype, log rank χ2=5.8 [P=.02]) relatives. Furthermore, cumulative risk curves among first-degree relatives of probands with ϵ4/ϵ4 and ϵ3/ϵ4 genotypes did not appear homogeneous (log rank χ2=5.7; P= .02); relatives of probands with ϵ3/ϵ4 genotype had a lower risk than relatives of probands with ϵ4/ϵ4 genotype. Sex-specific comparisons revealed that risks for AD in relatives of probands with ϵ4/ϵ4 vs ϵ3/ϵ4 genotype were significantly higher for female (log rank χ2=6.1; P=.01) but not male (log rank χ2=0.70; P=.40) relatives.
Of the 275 probands, 23 had 1 or more first-degree relatives with known apoE genotype. This led to a total of 30 relatives with known apoE status (26 siblings and 4 parents). Thus, for the majority of probands, proportion of relatives with apoE genotypes ϵx/ϵx, ϵx/ϵ4, or ϵ4/ϵ4 had to be inferred. Table 4 shows for each apoE genotype group among probands, the expected proportion of siblings and parents with apoE ϵ4/ϵ4 and ϵx/ϵ4 genotypes, given the proband's apoE genotype and the observed distribution of relatives (parents and siblings) within each age-by-sex stratum. Among relatives aged 40 to 90 years who had the apoE ϵ3/ϵ3, ϵ3/ϵ4, or ϵ4/ϵ4 genotype, the total predicted proportion of members carrying at least 1 apoE ϵ4 allele was equal to .14, .58, and .90, respectively. These estimates were 0.5 lower and 1.3 and 1.5 times greater than the cumulative incidence of AD by 90 years of age in relatives of probands with ϵ3/ϵ3, ϵ3/ϵ4, and ϵ4/ϵ4 genotypes, respectively. Thus, apoE genotypes cannot account for whole familial aggregation of AD observed in families of probands with ϵ3/ϵ3 genotype. On the other hand, in families of probands with ϵ3/ϵ4 or ϵ4/ϵ4 genotype, AD did not develop in about half of relatives predicted to have at least 1 ϵ4 allele, even by 90 years of age. The same trends were observed among siblings and parents. For example, in families of probands with ϵ4/ϵ4 genotype, proportions of relatives carrying at least 1 ϵ4 allele were 1.7 (0.83/48%) for siblings and 1.6 (1.00/64%) for parents, greater than the corresponding cumulative LTRs. Table 4 also shows that whatever the proband's apoE genotype, more mothers than fathers of probands were expected to have 1 or 2 ϵ4 alleles, particularly within groups aged 60 to 69 and 70 to 79 years. In these age stratums, the estimated proportion of relatives with the apoE ϵ4/ϵ4 genotype was about 2 and 3 times higher in mothers than in fathers of probands with ϵ4/ϵ4 or ϵ3/ϵ4 genotype, respectively. Among parents of probands with ϵ3/ϵ3 genotype, the proportion of mothers heterozygous for ϵx/ϵ4 was more than 2 times higher than fathers.
The first goal of our study was to examine the relationship between familial aggregation of AD and apoE genotype in probands. Conflicting results have been published concerning this topic. Farrer et al25 reported a higher risk in relatives of probands with the ϵ4 allele, whereas Li et al31 found that the risk was lower than that found in relatives of probands without the allele. They suggest that the increased mortality due to heart disease in relatives associated with presence of the ϵ4 allele in probands with AD might have confounded an otherwise strong association between familial aggregation and the allele in AD. Our survival analyses results support the hypothesis that the apoE ϵ4 allele enhances AD susceptibility. In our sample of 275 families, the risk for AD among first-degree relatives increased significantly with the number of ϵ4 alleles present in the proband; by 90 years of age, the LTR for AD in relatives of probands with ϵ3/ϵ3, ϵ3/ϵ4, and ϵ4/ϵ4 genotypes was 29.2%, 46.1%, and 61.4%, respectively. Thus, our results agree with those of Farrer et al.25 Also, our LTR estimations are close to those estimated by Farrer et al,25 although our cohort has a relatively younger mean age of onset (62.9 ± 9.2 vs 65.3 ± 9.2 years). Our survival analyses provide further evidence of nonhomogeneous LTRs among relatives of probands with ϵ3/ϵ4 and ϵ4/ϵ4 genotypes; in our data, relatives of probands with ϵ3/ϵ4 genotype have a significantly lower risk than relatives of probands with ϵ4/ϵ4 genotype.
We also found sex-by-apoE genotype interaction effects on AD LTR; among relatives of probands with ϵ4/ϵ4 genotype, women had a 2-fold higher cumulative AD risk than men. The same trend was also observed among relatives of probands with ϵ3/ϵ4 but not ϵ3/ϵ3 genotypes. Furthermore, female relatives of probands with ϵ4/ϵ4 genotype had a significantly higher risk than those of probands with ϵ3/ϵ4 genotype, whereas risk for AD in male relatives of probands with ϵ4/ϵ4 and ϵ3/ϵ4 genotypes did not appear to be significantly different. Significant sex differences in AD risks among relatives of probands heterozygous for the ϵ4 allele have also been reported.25 Several studies, correcting for the greater life expectancy of women32 or not,33 have indicated that the risk for dementia is more important in mothers than in fathers of patients with AD. It has been hypothesized that mitochondrial inheritance or genomic imprinting could account for this finding. Our results suggest the following explanation: whatever the apoE genotype of the proband, the proportion of subjects carrying at least 1 apoE ϵ4 allele is greater in mothers than in fathers (Table 4).
Our second purpose was to assess the extent to which clustering of AD in first-degree relatives of probands with AD and known apoE genotype could be explained by the apoE ϵ4 allele. For each proband, we used the observed age- and sex-stratified distributions of siblings and parents to derive expected proportions of relatives with 0, 1, or 2 apoE ϵ4 alleles, conditionally on the proband's genotype. The predicted proportion of ϵ4 carriers in relatives of probands with ϵ3/ϵ3 genotype (0.15) remains about 50% lower than the corresponding LTR for AD (29.2%), indicating that other (genetic and/or nongenetic) factors explain AD risks in these families. On the other hand, in relatives of probands with ϵ4/ϵ4 or ϵ3/ϵ4 genotypes, proportions of ϵ4 carriers are much higher than corresponding LTRs, indicating that AD will not develop in about 30% of female and up to 60% of male relatives with 1 or 2 ϵ4 alleles, even by 90 years of age. Our results agree with those reported by Farrer et al25; although relatives were not stratified by sex, they estimated that about 50% of relatives having at least 1 apoE ϵ4 allele remain cognitively normal by 92 years of age.
Such a picture could result from underrepresentation of oldest relatives having 1 or 2 ϵ4 alleles, due to differential survival for ϵ4 and non-ϵ4 carriers, particularly in men. Indeed, a survival analysis has reported a greater likelihood of death due to heart disease in relatives of probands with ϵ3/ϵ4 or ϵ4/ϵ4 genotype than in relatives of non-ϵ4 carriers, especially before 70 years of age.31 These results are of major interest, since it has also been shown that the apoE ϵ4 allele exerts its maximal effects on AD within the sixth to the seventh decades of life, when the disease is relatively rare, and not in the eighth and ninth decades, when disease prevalence is the most common.8,17,34 Another finding favoring this explanation is the report of a decreased frequency of the ϵ4 allele among male controls after 60 years of age.8
Finally, in probands with ϵ3/ϵ3 genotype, the modest familial clustering of AD is in a large part due to other factors than the apoE ϵ4 allele. In families of probands carrying at least 1 ϵ4 allele, aggregation of AD is more prominent. However, in this case, AD does not develop in numerous relatives of these probands, particularly men, possibly as a consequence of censorship by heart disease. In addition, it has been suggested30 that sex-specific factors such as hormones may interact with the apoE genotype in women to modify the disease susceptibility. So far, however, this hypothesis has gained little support from epidemiological data. Contrary to the expected trend, a recent report35 indicates that, in women, estrogen use during menopause actually reduces the risk for AD. Thus, putative factors enhancing risk for AD in women remain to be found.
Accepted for publication October 2, 1997.
Supported by grants from the Institut National de la Santé et de la Recherche Médicale (Network 492002), the Caisse Nationale d'Assurance Maladie des Travailleurs Sociaux, and the Mutuelle Générale de l'Education Nationale, Paris, France.
Reprints: Maria Martinez, PhD, Institut National de la Santé et de la Recherche Médicale, U358, Hôpital St Louis, 1 Avenue Claude Vellefaux, 75010 Paris, France.