Table 1.

Cross-sectional and longitudinal epidemiological studies of early life risk factors and Alzheimer’s Disease

Marker	Ref.	Sampling Method	Study Design	Strengths	Co-variates	Outcome Measure for AD Diagnosis	Results from fully adjusted model (95% CI)	Key Limitations
Genetics
Cross-Sectional
White matter myelin water fraction and gray matter volume	9	Community- based	Cross-sectional	APO-E4 carrier and non-carrier groups matched for age, gestational duration, birth weight, sex ratio, maternal age, education, and socioeconomic status.	Age, gestational duration, birth weight, maternal age and SES	Positive APOE4 genotype	Infants carrying APOE4 had lower white matter myelin water fraction and gray matter volume than noncarriers (p<0.05)	Unclear whether these metrics are also lower in APOE4 carriers after infancy (perhaps a temporary effect only)
Standardized achievement tests and R-O complex figure administered to children	16	Community- based	Retrospective cohort	Analyzed effects of APOE4 status on cognition in children	Age	Positive APOE4 genotype and positive family history of AD	Children with both an APOE4 allele and +FH scored significantly lower on reading (p=0.032), language (p=0.044), and the R-O complex figure test (p=0.015)	Small sample size (n=109)
The SORL1 gene and convergent neural risk for Alzheimer’s disease across the human lifespan.	10	Community- based	Cross-sectional	Analyzed white matter microstructure	Age, APOE4 status, sex	SORL1 SNP rs689021	Lower frontotemporal white matter fractional anisotropy in carriers of the SORL1 SNP (p=0.008)	Only one metric of white matter integrity used
Thickness of left entorhinal cortex in adolescence	13	Community- based	Cross-sectional	Use of specific cortical region for thickness measurements	Age, sex, race	Positive APOE4 genotype	APOE4 carriers had thicker left entorhinal cortex (3.79 mm) than non-carriers (3.94 mm) (p=0.03)	High SES of participants could bias results
Right hippocampal volume	14	Clinic-based	Cross-sectional	Use of extensive testing to rule out AD symptoms	Age, sex, education	Positive APOE4 genotype	APOE4 carriers had smaller right hippocampi (P=0.09)	Hippocampal volume measured in adulthood
Mitochondrial activity in posterior cingulate cortex	15	Population- based	Case-control	Use of pathological histology for mitochondrial activity	Age at death and postmortem interval	Positive APOE4 genotype	APOE4 carriers had reduced posterior cingulate mitochondria activity (p=0.009)	None noted.
Bilateral hippocampal volume in young adults	18	Population- based	Cross-sectional	Large, population- based sample	Age, sex, total brain volume	AD-associated SORL1 SNPs	Individuals with AD-associated SORL1 SNPs had smaller bilateral hippocampal volumes (p=0.01)	Relatively homogenous Netherlandish population.
Learning Disability
Cross-Sectional
Self-report: Family history of learning disability	27	Community- based	Case-control		None	McKhann, Neary and Mesulam criteria including consensus of neurologist and a neuropsychologist.	16% of PPA (of any type) vs. 6% of behavioral variant, 7% typical AD, 5% controls.
Self-reported personal history of delay in speaking or reading	11	Clinic-based	Consecutive clinicopathological series		Age, gender, handedness, scanner and total intracranial volume	Consensus diagnostic criteria for PPA supported by imaging.	25% of logopenic PPA patients vs. 3% of semantic and 3% of non-fluent PPA.
Educational and developmental history from neuropsychological evaluation	30	Clinic-based	Restrospective case-control	First to examine connection between LD and atypical dementia	Age, gender, handedness, education and symptom duration	Consensus diagnosis of PPA and AD	Patients with probable learning disability 13 times more likely to be diagnosed with dementia (OR 13.1 95% CI 1.3–128.4)	Uncertainty of learning disability presence due to self-report of symptoms
Longitudinal
Self-report: Family history of learning disability	29	Clinic-based	Case-control	Use of pathological criteria for AD and FTLD	Not discussed.	Autopsy-based diagnoses	LD prevalence in PPA was about 50% with no difference between PPA from AD and FTLD.	Small sample size and only estimated prevalence.
Education & Intellect
Cross-Sectional
Self-assessed school performance (“below” or “above” average)	23	Population based	Case-control	Large sample size; adjusted for presence of APOE4	Age, gender, race, presence of APOE4	NINCDS / ADRDA criteria and DSM III and IV criteria	Participants with “below average” self-assessed school performance were more likely to have AD (OR 4.0; 1.2–14 95% CI)
Longitudinal
Young adult (∼22 years) linguistic ability (idea density and grammatical complexity)	25	Community based	Longitudinal cohort	Use of neuropathology to confirm AD pathology Dissociation of idea content and grammar	Age, education occupation	AD neuropathology at autopsy	Nuns with AD had decreased idea density (P<0.001) but not decreased grammatical complexity in early writings (P=0.61)	Catholic nuns are well-educated and thus may not represent the premorbid mental abilities of the general population
Idea density in handwritten autobiographies from 19 to 32 years old	24	Community- based	Longitudinal	Use of childhood written accounts (rather than self- report) and autopsy pathology	Age at death and location of convent (population of nuns)	Neurofibrillary tangle and senile plaque count in frontal, temporal, and parietal lobes	Greater idea density in childhood writing inversely correlated with neurofibrillary tangle and senile plaque count at autopsy (∼-0.5 for tangles and ∼-0.3 for plaques) (p<0.0001 for tangles and p<0.001 for plaques)	Only used text analysis for idea density and not for other linguistic or writing measures
SES
Cross-Sectional
Self-reported childhood rural residence	32	Community- based	Case-Control	Randomized sample	Age, gender, education	NINCDS-ADRDA criteria	OR 6.5 (2.6 to 16.7) for low education/rural residence vs. high education/urban residence	Not adjusted for major medical comorbidities (as would be expected in rural and urban populations). Cutoff for “low education” (grade 6 or lower) may be problematic as effect of education was only seen in rural residents.
Informant reported mother’s age at subject’s birth, birth order, sibship size, and area of residence before the age of 18 years	34	Community- based	case-control	Stratified by presence of APOE4 Significant linear trend for number of siblings	Age, gender, education, APOE	NINCDS/ADRDA or Definite AD by neuropathologic criteria	OR 1.4 (1.0 to 2.0) for sibship size of five or more; OR 0.5 (0.3 to 0.8) for suburban childhood residence	Proxies used for both case and control interviews, potentially producing misclassification of information. Greater response rate from cases.
Father’s occupation, parents’ age, household size, birth order, sibship size, and home ownership	33	Community- based	Case control	Stratified by presence of APOE4. Use of objective data (census and birth certificates) rather than interviews.	Age, gender, education, APOE	NINCDS / ADRDA criteria.	OR 1.8 (1.2 to 2.7). Strong interaction with APOE4.	Use of father’s occupation as a surrogate for quality of early home environment is limited (analysis did not include maternal occupation, area of residence, etc).
Longitudinal
Self-reported parental highest years of schooling, paternal occupational prestige, family financial status and cognitive milieu	35	Population- based	Longitudinal Cohort	Longitudinal 5 year follow-up time	Age, gender, race, education	Symbol Digit Modalities Test, East Boston Story, MMSE	No association with cognitive change over time (beta −0.005, p<0.10)	Self-reported recall of childhood cognitive milieu Even though the follow-up was five years, perhaps a longer period is necessary to show significant decline
Self-reported parental education, occupational prestige, sibship size	36	Community- based	Longitudinal clinic- pathologic		Age, gender, education, county-level SES	NINCDS / ADRDA criteria.	No association: RR 1.1 (0.9 to 1.4) for higher household SES composite score.
Paternal occupation, number of public rooms in childhood home, and number of people in home per sanitation facility	37	Community based	Cross-sectional and longitudinal	Analysis of HPC using volumetric MRI	Mental ability at 11 years old adult SES gender education	Hippocampal volume from MRI	Low childhood SES is associated with lower adult hippocampal volume (p=0.032)	Individuals with higher mental ability at 11 years selectively participated possibly leading to a systematic bias
Reading level and early SES	38	Community based	Prospective cohort	Adjusted for APOE4 status Use of resilience metric, rather than AD clinical or pathological criteria only	Age, gender education	AD pathology at autopsy Cognitive testing for memory Disparity between metrics = “resilience”	Adult reading level associated with greater resilience (p<0.0001) and accounts for effects of early life SES	Study population (Caucasian volunteers agreeing to post-mortem examination) may not represent entire aging population
Body Growth
Cross-sectional
Arm and leg length	41	Populatio n-based,	Case-control	Collected culturally relevant measures of early life environment; MMSE & KDRS scores from t-2 years for 64% of current sample (paired assessment). Observed sex differences in APO-E4 effect: Intrasex group risk difference (inc risk of dementia w e4) for men; intra-sex group risk difference was the opposite for women with e4 providing a protecting factor. Inclusive sampling source; cases and controls screened prior to inclusion.	Age, gender, education, menarche, menopause	NINCDS-ADRDA criteria, using two independent teams	In women only: OR 2.5 (1.6 to 4.0) for 5cm decrease.	Variable methods for measuring leg length (gold standard?). No assoc of sitting height with dementia. Anthropometric measures were not treated as continuous variables, why? In models adjusted for age, education, female gender, risk of Vas dementia 4x that of men, yet hypertension and diabetes were not associated w risk of dementia in this sample. Report of AD diagnosis: No formal investigation of APO-E4 status, limb length and AD. Reported higher rate of AD than population estimates for East Asia.
Intracranial area by CT scan	43	Clinic- based	Retrospective case series	Gender-specific, brain- imaging (MRI/ CT scans) for confirmation of Probable AD. CT scan suitability determined by location of anatomical structures. Brain size correlated positively with age at first symptom.	Education, height ethnic group	Self-reported date of onset of symptoms of AD (diagnosed using NINCDS- ADRDA criteria using consensus panel)	Correlation 0.48, p 0.009 between head size and age of onset	Use of brain imaging at diagnosis for proxy of pre-morbid brain size; age at first symptom was imputed with no indication of a statistical measure). No accounting for height, weight, or overall volume with respect to brain size. Aside: Wouldn’t bigger brains just produce more plagues and neurofibrillary tangles??
Head circumference	52	Commun ity-based	Case-control	sampling from multiple cognitive categories for balancing	Age, gender, education	NINCDS-ADRDA criteria	OR 0.9 (0.3 to 1.9) for HC (treated linearly)	Statistical manipulation of head circumference and unbalanced weighting of categories, renders findings questionable. `Specific population; Japanese-American. Accuracy of head circumference measurements at birth. Head circumference was not a significant predictor of AD for prevalent AD. Incorrect assumption of head circumference as proxy for cognitive reserve. No adjustment for height or weight. Sub-sample analyses among patients diagnosed with probable AD didn’t attenuate the effect of “THC with CASI score” when adjusted for height.
Arm length	42	Populatio n-based	cross- sectional	Population based sample; multi-tier diagnosis of dementia based on cognitive tests and blinded neurological assessments.	Age, gender, education, smoking, alcohol consumption, pulse pressure, hypertension and diabetes	NINCDS-ADRDA criteria using consensus between a physician and neurologist; also change in Korean MMSE over three years	OR 1.2 (1.0 to 1.3) for 1cm decrease in arm length	None other than weakness of causal or correlative evidence
Head circumference	44	Populatio n-based	Case-control		Height, weight, education, APOE	NINCDS-ADRDA criteria	OR 2.9 (1.4 to 6.1) for women and 2.3 (0.6 to 9.8) for men for lowest quintile
Intracranial volume by CT scan	54	Clinic- based	Clinic-based, case-control	Use of gold-standard measure of premorbid brain size: total intracranial volume	Gender	NINCDS-ADRDA criteria with consensus	No significant differences in intracranial volume	Why is brain size a proxy for cognitive reserve? While women with AD had smaller head size on average in comparison to female controls, controlling for years of education, decreased that difference. The finding of no association between APO-E4, age, birth year and TIV. Even lowest tertiles of TIV not predictive of AD.
Intracranial volume by CT scan	53	Clinic based	case-control study	Matched controls, blinded MRI image analysis, results assessed for inter-tester variability	Age at scan, gender, familial vs. sporadic AD	NINCDS-ADRDA criteria	No significant differences in intracranial volume	None noted.
Prenatal sex hormone exposure (measured through 2D:4D length ratio proxy)	45	Clinic- based	Case-control	Gender-specific; determined that high estrogen:testosterone ratios are a risk factor in men but protective in women. Ability to estimate prenatal hormone exposure	Age and years of education	NINCDS-ADRDA criteria (consensus diagnosis)	AD males had higher 2D:4D ratio (high E:T) than controls (p<0.001) AD females had lower 2D:4D ratios (low E:T) than controls (p<0.001)	Accuracy of 2D:4D ratio as a proxy for prenatal hormone exposure Could not describe effects of estrogen and testosterone individually
Longitudinal
Height	47	Commun ity-based	Case-control	Longitudinal data on height.	Age, body mass index, years of childhood lived in Japan, level of education and father’s occupation	NINCDS-ADRDA criteria using consensus panel (study neurologist and two other dementia experts)	In men only: Prevalence of AD higher (4.7% vs. 2.9%, p =0.18) in men shorter than 154cm	Standing height at baseline used as anthropometric measure, with no adjustment for age related changes in height; Japanese heritage vs. multiracial status not addressed. No report on standardization of height for race.
Head circumference	55	Community- based	Retrospective cohort	Apoe4	Age, education, gender	NINCDS-ADRDA criteria	Combination of small head circumference and APOE4 positivity predicted earlier onset of AD (p=0.0007)
Head circumference	56	Community based	Prospective Cohort study	Apoe4; enhanced follow-up for subjects with marked cognitive changes (CASI ≤87)	Head circumference, height, verbal IQ, income, education, age at growth cessation, household characteristics seated BP, anthropometrics		APOE 4 Hetero- and homozygosity had a differential effect on AD for men vs women: HRheterozygosity men= 1.9 (95% CI 0.7–5.4) vs HRheterozygositywomen= 4.2 (95% CI 2.1–8.6); HRhomozygositymen = 5.3 (95% CI 0.7–41) vs HRhomozygositywomen = 18.3 (95% CI 2.3–144)
Height	48	Community- based,	Nested case- control	Blind confirmatory diagnoses of dementia by neurologists + Consensus diagnoses (by a blinded 2nd neurologist)	Age, gender, education, occupation, and area of birth	NINCDS-ADRDA criteria	OR 0.6 (0.4–0.9) for highest quartile vs. lowest quartile	TICS-m used for initial dementia diagnosis; Healthy volunteer effect (1999 sample overall had lower risks factors)
BMI and HOMA- IR	98	Clinic-based	Prospective cohort	Use of serum-based biomarkers of AD:	Age, gender, fasting lipid panel, glucose, and WBC	Aβ-42 and PSEN1	RR: 7.1, p-value= 0.002 for Aβ- 42
Arm length	49	Population- based	Longitudinal (prospective cohort)	Gold standard assessment tools: MRI, Genetic testing MMSE and 3MSE; Representative cohort (across race and age)	Age, gender, ethnicity, education, income, self-reported health, APOE4 status	NINCDS-ADRDA criteria with consensus (one neurologist and one psychiatrist)	HR 1.7 (1.1–2.6) in women vs. 0.9 (0.8–1.0) in men	Potential misclassification (non- differential/differential?), Conclusion of lower knee height and arm spans associated with increased risk of dementia troublesome bc: phenomenon seen only among women for knee height and when assessed for lowest quartile of knee height, found not significant. However, arm span was significantly associated with dementia (men and women) and AD (women only) No record of assessment of childhood nutritional deficiencies though!
Fetal head circumference and adult head circumference	50	Community- based	Prospective case control study	Comparison of cognitive function at study enrollment and at 3.5 year follow-up	Age, sex, education, social class at birth, history of cerebrovascular disease, Nottingham Health Profile emotion subscale score, gestational age	AH4 Intelligence Test, Logical Memory subtest of the Wechsler Memory Scale	OR for delayed recall 0.3 (0.1 to 0.9) for highest quartile	Accuracy of childhood head circumference measurements. Categorized adult head circumference and used the lowest quartile as the reference group for effects of the measure on the Logical memory test; once again due to nonstandardization of anthropometric measurements. When comparing furthermore sample size for observed decline in WLM were far to small and unmatched for comparisons.
Intracranial volume (ICV) and total brain volume (TBV) from MRI	51	Clinic-based	Longitudinal cohort	ICV is better predictor of premorbid brain size than head circumference Use of APOE status	Age, gender, education APOE genotype, CV disease presence	NINCDS-ADRDA criteria for AD MMSE, ADAS- cog, and CDR scores for longitudinal follow-up	Atrophy and APOE4 allele had reduced impact on cognitive and clinical decline in MCI with larger ICV (p<0.05)	ICV measurements taken after diagnosis (not in childhood or mid- adulthood) ICV is an imperfect approximation of premorbid brain size
Childhood Adversity
Longitudinal
Early parental death and remarriage of widowed parents	58	Population based	Prospective cohort	Use of consensus AD diagnosis Large, population based sample	Age, gender and education	NINCDS-ADRDA criteria; consensus diagnosis	Maternal death from age 11 17 associated with 2x risk of AD	Adjustment for later SES from parental death was difficult due to missing data