Abstract
Objective: To investigate the association of early-life factors with AD.
Background: The early-life environment and its effect on growth and maturation of children and adolescents are linked to many adult chronic diseases (heart disease, stroke, hypertension, and diabetes mellitus), and these effects are also linked to maternal reproduction. AD may have an early-life link. The areas of the brain that show the earliest signs of AD are the same areas of the brain that take the longest to mature during childhood and adolescence. A poor-quality childhood or adolescent environment could prevent the brain from reaching complete levels of maturation. Lower levels of brain maturation may put people at higher risk for AD.
Methods: In a community-based case-control study (393 cases, 377 controls), we investigated the association of early-life factors and AD. Early-life variables include mother’s age at patient’s birth, birth order, number of siblings, and area of residence before age 18 years. Patient education level and apolipoprotein E (APOE) genotypes were also included in the analysis.
Results: Area of residence before age 18 years and number of siblings are associated with subsequent development of AD. For each additional child in the family the risk of AD increases by 8% (OR = 1.08, 95% CI = 1.01 to 1.15). More controls compared with cases grew up in the suburbs (OR = 0.45, 95% CI = 0.25 to 0.82). APOE ε4 and the patient’s education level did not confound or modify the associations.
Conclusions: The early-life childhood and adolescent environment is associated with the risk of AD.
The early-life environment and its effect on growth and maturation in children and adolescents are linked to many adult chronic diseases (heart disease, stroke, hypertension, diabetes mellitus, and chronic obstructive lung disease)1 and to female reproductive outcomes.2 AD may also have an early-life link.3-5 Understanding growth, maturation, and aging of the brain may be the key to this link. The brain grows most in size during the prenatal period and in childhood6-8 but continues to complete its maturation during adolescence.6-9 Brain maturation refers to the development of connectivity patterns, synapses, branching of dendrites, and myelination.7-9 The areas of the brain that take longest to mature during childhood and adolescence (e.g., hippocampal formation, intracortical association areas, reticular formation)7,9 are the same areas of the brain that show the earliest signs of AD.10-13 An association between early-life growth and development and later-life cognitive decline was first suggested by Conel14 in 1939.
Environmental factors can affect brain maturation. Studies on rats15-19 show that brain maturation can be retarded with only mild malnutrition and that catch-up growth is not always attainable. Mild malnutrition has a slowing effect on development and myelination patterns and interferes with normal development by decreasing dendritic growth. Increasing nutrition later in the rats improves brain maturation (as measured by amount of myelin) but not to the level of those who were never malnourished. This finding parallels studies on human children20,21 showing that children who were marginally malnourished are shorter, lighter, and score lower on cognitive ability than their larger, heavier, and better nourished peers. Improved nutrition and environment later in childhood modified, but did not eliminate, the difference between marginally undernourished and well-nourished children, although both groups of children scored within the normal intelligence range. Therefore, poor growth early in life could increase the risk of AD. The effects of impaired development could produce a brain that is normal but functions less efficiently because of less myelin, less branching of dendrites, and less developed connectivity patterns. This impaired development affects speed and specificity of nerve transmissions and requires increased energy to function properly.22,23 The negative effects of this less efficient brain would likely be marginal until aggravated by the aging process.
We investigated the association of AD and early-life factors: mother’s age at subject’s birth, birth order, sibship size, and area of residence before the age of 18 years. Babies born to mothers who are younger than 20 or older than 35 years of age tend to have lower birth weights.24 Although there is an increase in birth weight with each successive birth, in cross-sectional studies babies born to mothers over 35 years of age also tend to be smaller because mothers who continue having babies later in life usually are in lower socioeconomic levels.25 The number of children in a family is related to socioeconomic level.26-29 During the early 1900s when the subjects in this study were children, the optimal/preferred family size was three or four children.27,29 Families with five or more children were more likely to be from the lower socioeconomic levels27,28 and therefore were more likely to have poor growth rates.25,30-32 If deficient maturation is associated with a less developed brain, then these measures that influence early growth could be associated with AD. We also investigated whether the potential association of these early-life factors and AD changed after adjusting for education level and apolipoprotein E genotype (APOE), or whether the potential associations are modified by APOE (i.e., whether the strength of the association differs between those with and those without one or more APOE ε4 alleles).
Methods.
Study population and design.
Patients for this case-control study were drawn from the Group Health Cooperative (GHC), a large health maintenance organization in Seattle, WA. GHC was established in 1949; the Seattle area membership of people aged 60 years or over included about 23,000 people. The attrition rate, excluding deaths, is about 1% per year. Most of the GHC population are longtime members who originally enrolled through their employers and remain after retirement. GHC members are representative of the surrounding community with respect to age distribution, gender, and ethnicity, but have a slightly higher education level in this age group.
The AD cases for this study were obtained from patients enrolled in the University of Washington/GHC AD Patient Registry (ADPR) (U01 AG 06781) from 1987 to 1996. Specifically, they were probable AD cases also enrolled in the Genetic Differences Case-Control Study (R01 AG 07584). Cases were patients in whom dementia had been diagnosed according to the Diagnostic and statistical manual of mental disorders, 3rd ed., revised,33 and who had a diagnosis of probable AD as defined by the National Institute of Neurological and Communicative Disorders and Stroke–Alzheimer’s Disease and Related Disorders Association34 working group criteria, or definite AD if they died after ADPR enrollment and had a neuropathologic diagnosis of AD. Controls were patients selected at random from GHC enrollment lists during approximately the same period as the cases (case selection started 6 months earlier than control selection) and frequency matched on gender and age within 2 years. Potential controls were excluded if they had dementia or other neurologic disease causing dementia. To verify that the potential control subjects did not have dementia, they were required to achieve a score of 28 of 30 on the Mini-Mental State Examination35 (27 if over age 80 years) and to have no other indications of dementia based on other test results, medical record review, or observations of the research nurse-interviewer. For more details on the ADPR case surveillance, enrollment, and diagnostic protocol, see Larson et al.36; for description of the Genetic Differences Case-Control Study see Kukull et al.37
Collection of early-life variables.
Early-life information and other types of epidemiologic information were obtained by in-person interviews between research nurses and proxy informants for both case and control subjects. The variables used for this study include mother’s, patient’s, and siblings’ birth dates, patient’s education level, and patient’s area of residence before age 18 years. If a patient lived in more than one type of area prior to age 18, he or she was asked to choose the area in which he or she lived in the longest. The primary source of birth date information was a family history questionnaire. This questionnaire was completed by the proxy with consultation with other family members.
A greater proportion of case, compared with control, proxies answered fewer of the family history questions or did not complete the family history questionnaire. Missing information is unlikely to have biased reporting of mother’s age and other early-life variables because the reasons for not completing the questionnaire were not related to mother’s age or the early-life environment. If the exact birth date of the mother was unknown or missing from the family history questionnaire, approximate information from the in-person epidemiologic interview was used. At the epidemiologic data interview, the proxy was asked whether the mother’s age at patient’s birth was less than 20, 20 to 24, 25 to 29, 30 to 34, or greater than 35 years, rather than the date of birth as on the family history questionnaire.
APOE genotypes had been determined as part of the Genetic Differences Case-Control Study by the restriction enzyme digestion method of Hixson and Vernier,38 using DNA prepared from blood and brain tissue samples. Laboratory personnel were blinded to case/control status. APOE genotypes were unavailable for 74 of the case subjects and for 11 of the control subjects because either the family (n = 36) or the patient (n = 23) refused to give a blood sample, they discontinued participation (n = 11), the study ended before a blood sample was taken (n = 7), or samples were degraded or failed the PCR (n = 9).
Statistical analysis.
The strength of association between early-life factors and AD is described by ORs and 95% CIs.39,40 For categorical variables, ORs and 95% CIs were calculated in the conventional manner from frequency tables, and Mantel-Haenzsel adjusted ORs were calculated for stratified data.39,40 For continuous variables, unconditional logistic regression was used to estimate the crude and adjusted ORs and 95% CIs.39,40 Multiple logistic regression (unconditional) was also used to obtain adjusted effect estimates.39,40 For crude and adjusted ORs the variables are coded in the following ways: number of siblings was analyzed as a continuous variable, as a single dummy variable (<5 versus >5), and as three dummy variables (5 to 6 siblings, 7 to 9 siblings, and >10 siblings), using less than five siblings as the reference category. This value was chosen because families with five or more children were more likely to be from the lower socioeconomic levels26-29 and therefore were more likely to have poor growth rates.25,30-32 Education was coded as high school graduate or less versus more than high school. APOE genotype was coded as one or more ε4 alleles versus none.
Results.
Characteristics of the study population are shown in table 1. The mean birth order is similar in case and control subjects. The mean number of siblings is higher in case than in control subjects (3.8 versus 3.3). The area of residence the patient lived in prior to age 18 years is similar for farm, rural, and urban residence; however, more control compared with case subjects reported growing up in the suburbs. Cases had a higher frequency of APOE ε4–containing genotypes and an overall lower education level. The difference in level of education between case and control subjects is primarily due to response bias among controls. A substudy conducted as part of the Genetic Differences Case-Control Study showed that persons selected as potential controls who refused to participate in the study were more similar to the enrolled AD cases in their level of education, whereas control subjects who agreed to enter the study were more highly educated. Therefore, the appearance of an association between lower education and AD in this data is thought to be spurious.
Characteristics of case and control subjects in a study of early-life factors and the development of AD
Increased number of siblings is associated with an increased risk of AD (table 2). The risk of AD increases by 8% for each additional sibling in the family (OR = 1.08, 95% CI = 1.02 to 1.15). Growing up in a family with five or more siblings increases the risk of developing AD by 39% (OR = 1.39, 95% CI = 0.99 to 1.95). There is a linear trend of increasing risk with increasing sibship size. Compared with families with less than five siblings, having seven to nine siblings is associated with an almost twofold risk (OR = 1.72, 95% CI = 1.01 to 2.43), and in extremely large families with 10 or more siblings the risk is greater than twofold (OR = 2.66, 95% CI = 0.92 to 7.99). The area of residence prior to age 18 years is associated with AD. Specifically, more control compared with case subjects grew up in the suburbs (OR = 0.46, 95% CI = 0.25 to 0.83). The presence of at least one APOE ε4 allele is associated with a 3.6-fold increased risk of AD. We found no association between mother’s age at patient’s birth and subsequent onset of AD.
Crude associations of early-life factors and AD
Having more than a high school education is inversely correlated with number of siblings (Spearman’s r = −0.17, p = 0.0014). The presence/absence of APOE ε4 allele shows no correlation to growing up in the suburbs, level of education, or number of siblings. Table 3 shows the results of six different multiple logistic regression models used to obtain adjusted ORs and 95% CIs for the risk of AD. The ORs of the variables of interest are quite similar for each model and to the crude ORs shown in table 2. Model 1 shows the association of growing up in the suburbs while adjusting for patient’s education. Only a few patients are missing area of residence information and are not in this model. Model 2 shows the association of more than five siblings while controlling for presence of the APOE ε4 allele. The large amount of missing information is from sibling size and APOE; even so, the adjusted associations of models 1 and 2 shown in table 3 are very similar to the crude associations shown in table 2. Model 3 shows the associations of both area of residence and number of siblings adjusting for education, and model 4 shows the associations of area of residence and more than five siblings adjusting for both potential confounders (APOE and education). These four models individually and together show that the associations are stable for area of residence and sibship size when investigating individual associations or combined associations while adjusting for one or both of the potential confounders and across variations in the sample size. As can be seen in the variation in number of case and control subjects across the six models, there is notable but not complete overlap in the amount of missing information on each patient. Models 5 and 6 show the linear trend of sibship size in the presence of one or both confounders. Again the amount of variation between the associations of sibship size is minimal and the increasing risk with increasing sibship size is stable. Thus, the large amount of missing data on level of education or APOE ε4 allele does not appear to confound the association between AD, sibship size, or growing up in the suburbs.
ORs (95% CIs) relating various combinations of risk factors to AD
Stratifying by the presence/absence of APOE ε4 allele (table 4) revealed no significant variation in the strength of association between the variables of interest and AD. Increasing sibship size was not associated with AD in the presence of APOE ε4; however, the large percent of missing information for the number of siblings in the case subjects combined with the lower frequency of APOE ε4 in the control subjects may have compromised the ability to evaluate this association. The Breslow-Day test for homogeneity of ORs across strata was not significant (p > 0.1), indicating no statistical evidence for effect modification by APOE genotype.
Results of stratified analysis of early-life factors and AD by presence of any APOE ε4 allele
Discussion.
The relationship of the process of growth and development of the brain and the pathology of AD describes a biologic connection. We as well as others3-5 have concluded that the early-life environment may be associated with the development of AD. Each of these studies used different measures of early life (head circumference,3 adult height,5 and early-adult linguistic ability4) to investigate an early-life association with AD. We used information collected by interview to retrospectively collect factors in the early-life environment that influence growth to further explore the association between early life and AD.
Area of residence and number of siblings are related to socioeconomic level and therefore to quality of the living environment. The association between living in the suburbs and AD could reflect the benefits of higher socioeconomic status and less exposure to infectious disease. During the early 1900s infectious diseases were known to be less frequent in less densely populated rural areas compared with urban areas.26 The suburbs were less densely populated and, especially during this era, were an area of at least middle to upper socioeconomic levels. Therefore, children growing up in the suburbs may have been more likely to have better nutrition and less exposure to infectious disease, leaving more energy for normal growth and development. Following this logic, we expected to see a protective association with living on a farm. We did not. However, many farming families during this era who experienced economic hardship migrated into other urban-based occupations. Without being able to differentiate between patients who grew up on economically successful farms versus those whose families lost their farms, we are unable to fully test this association.
The number of siblings in the family also reflects the economic level.26-29 Simply, more people living in one family stretches the resources more thinly. Also, it was more common for persons of lower socioeconomic status to marry younger and begin childbearing at an earlier age, continuing reproduction into their middle and late 30s.26,27 Persons of higher socioeconomic status tended to marry later in life after furthering their education and as a result tended to have fewer children.27
The association of education level needs to be interpreted with caution. In this study sample 39% of the case subjects and 56% of the control subjects have more than a high school education. Prior to 1940 when people of this age group were on average 28 years old, less than 25% of the population over 24 years of age had more than a high school education.41 GHC, the Seattle area health maintenance organization from which this study population is selected, has approximately 20% more members with more then a high school education42 compared with the surrounding population.43 Another factor complicating the analysis of education level is the effect of selection bias. There is likely selection bias for education in this study because potential control subjects who declined to participate were similar in education level to the enrolled case subjects. More highly educated control subjects participated; therefore, the larger proportion of control subjects with more than high school education is likely to be at least partially the result of selection factors, and thus the association with AD cannot be correctly estimated. Case subjects display less selection effect because they were motivated to participate to obtain expert diagnosis by ADPR clinicians of a very serious problem.
We see in our data that number of siblings, a socioeconomic indicator, and greater than high school education are correlated, albeit modestly (Spearman’s r = −0.17). The complexities of understanding the meaning of the association of education level with disease are not unique to this study. Level of education can be a risk factor, a confounder, and a reason for selection bias. Therefore, we suggest the association be viewed with caution. Adjusting for educational level in this study did not significantly change the association between AD and the variables of interest to the study’s hypothesis.
This study has other limitations. The early-life variables are obtained from proxy interviews; we interviewed proxies for both case and control subjects to avoid obvious asymmetry in data collection that would occur if we interviewed proxies of case subjects and then interviewed control subjects directly.44 We expect that proxy interviews, generally speaking, may lead to nondifferential misclassification of information obtained and potentially bias effect measures toward the null.44 Data on mother’s age at patient’s birth, birth order, and number of siblings are unavailable for many patients. Although missing data were more prevalent in case than in control subjects, we have no evidence that these missing data would bias our finding away from the null. However, better, more objective means of collecting these data are necessary to understand the association of these variables. We are currently developing another project to follow up on these preliminary findings and to further explore the association of the early-life environment and the development of AD.
The strengths of this study are that the results indicate associations consistent with a biologically plausible connection seen between growth and development and the pathology of AD. Although the associations of sibling size and area of residence before age 18 years are not particularly large, they are stable when controlling for potentially confounding effects, and in sibship size there is a consistent and significant linear trend.
Acknowledgments
Supported in part by grants R03 AG 15179, R01 AG 07584, and U01 AG 06781-06 from the National Institute on Aging, US Public Health Service.
- Received February 22, 1999.
- Accepted August 27, 1999.
References
Disputes & Debates: Rapid online correspondence
REQUIREMENTS
If you are uploading a letter concerning an article:
You must have updated your disclosures within six months: http://submit.neurology.org
Your co-authors must send a completed Publishing Agreement Form to Neurology Staff (not necessary for the lead/corresponding author as the form below will suffice) before you upload your comment.
If you are responding to a comment that was written about an article you originally authored:
You (and co-authors) do not need to fill out forms or check disclosures as author forms are still valid
and apply to letter.
Submission specifications:
- Submissions must be < 200 words with < 5 references. Reference 1 must be the article on which you are commenting.
- Submissions should not have more than 5 authors. (Exception: original author replies can include all original authors of the article)
- Submit only on articles published within 6 months of issue date.
- Do not be redundant. Read any comments already posted on the article prior to submission.
- Submitted comments are subject to editing and editor review prior to posting.