Alzheimer’s disease (AD), the leading cause of dementia, is a progressive neurodegenerative disorder that typically first presents clinically as deficits in cognition. It is estimated that over five million people in the United States are currently living with AD, and by 2050, that number is expected to climb to 13.8 million (Hebert et al., 2013). The two major risk factors for the more common late-onset form of AD are increasing age and a first-degree family history of dementia (FH) (Braak et al., 2011; Green et al., 2002). FH is known to encompass both heritable and nonheritable risk factors for AD (Chang et al., 2012; Yi et al., 2018). FH has been associated with changes in multiple cognitive domains previously in children (13 years old) and young adults (35 years old); however, fewer cognitive domain changes were reported in middle-age (53 years old) and older (65–78 years old) FH adults (Aschenbrenner et al., 2016; Bloss et al., 2008; Honea et al., 2009; La Rue et al., 2008; Miller et al., 2005; Parra et al., 2015; Zeng et al., 2013). These effects were unrelated to the apolipoprotein (APOE) ε4 allele. Despite these FH findings, to the authors’ knowledge, there has not been a well-powered study of the effects of FH on memory across the lifespan. Furthermore, it is unknown if FH status interacts with demographics, common health and APOE genotype. Identification of factors potentiating or ameliorating an effect of FH is particularly crucial as one approach to identifying new avenues of risk reduction for AD. Indeed, risk reduction for AD is now more critical than ever due to the continued lack of a cure or effective disease-slowing treatments for AD (Barnes and Yaffe, 2011; Sperling et al., 2011; Takeda and Morishita, 2018). This fact underscores risk reduction as one of the most practical means to attenuate the enormous burden AD places on society (Baumgart et al., 2015).

To overcome challenges associated with underpowered studies, we created MindCrowd (www.mindcrowd.org) as an accessible and easy-to-use web-based cognitive and demographic assessment; specifically, to elucidate the effects of FH on cognition in healthy participants. Participants were recruited across the entire adult lifespan to understand better how FH may alter cognition at different ages. The cognitive assessment consisted of an online, verbal paired-associates learning (PAL) task. PAL was chosen because it is a medial temporal lobe (MTL)-dependent learning and memory task affected early in the onset of AD (Pike et al., 2008). In addition, PAL is sensitive to changes in performance associated with healthy aging (Pike et al., 2013). Data on multiple demographic, medical, health, and lifestyle factors were also obtained (see Supplementary file 2). FH status was determined by self-report of having (FH+), or not having (FH-) based on the question, ‘Have you, a sibling, or one of your parents been diagnosed with Alzheimer's disease?’ Further, a subset of FH participants was solicited to send us dried blood spots so that we could determine their apolipoprotein (APOE) genotype. APOE is a well-known genetic risk factor for AD. The risk of developing AD is higher by ~6% in ε4 allele carriers, while total genetic heritability risk for AD is 33% based on single nucleotide polymorphism data (Ridge et al., 2013), and ~60–75% in twin studies (Reynolds and Finkel, 2015). In addition, a recent study found a synergistic effect of FH and APOE ε4 on higher levels of amyloid-beta deposition measured via positron emission tomography (PET) imaging (Yi et al., 2018). As a result of these data and numerous other reports, we hypothesized that the APOE ε4 allele would relate to lower PAL scores in FH participants.

As of August 15, 2018, MindCrowd, has recruited 59,571 qualified participants (see Data Quality Control in Materials and methods) from around the world (Figure 1A). The sample was 62.46% female and 37.54% male (Figure 1B). An overrepresentation of women has been previously described in studies drawn from the general population (Krokstad et al., 2013) as well as for AD (Roberts et al., 2004). The breakdown of race was American Indian or Alaska Native = 0.62%, Asian = 5.13%, Black/African American = 1.75%, Native Hawaiian/Pacific Islander = 0.39%, Mixed = 0.0009%, and White = 92.03%. In terms of years of education, we collapsed across education milestones for visualization purposes. Here we found that 10.78% reported ≤12 y, 29.64% reported ≤14 y, 35.32% ≤ 16 y, and 24.25% reported ≥20 y (Figure 1C). Across the entire sample, a FH of AD is present in 22.76% with the overall percentage swelling with age (Figure 1D).

Figure 1 with 2 supplements see all

Download asset Open asset

In women and men 18–85 years old, our general linear model (GLM, see Figure 1—figure supplement 1 for regression diagnostic plots and Supplementary file 1 for a table all coefficient ns) revealed a significant Age coefficient (B_Age = −0.20 word pairs, p_Age <2e-16). Age was associated with a lowered PAL performance of two word-pairs per decade of life (Figure 2). Sex was also a significant predictor of PAL scores (B_Sex = −1.82, p_Sex <2e-16). Women were associated with nearly a two word-pair higher PAL score compared to men (Figure 2). Propensity score matching (PSM, see Statistical Methods in Materials and methods) revealed that the associated disparity between women and men’s PAL scores markedly grew around the 5^th decade of life. PSM revealed an estimated effect size (i.e., average treatment effect among treated, ATT) of being a woman grew from the 5^th to the 6^th decade of life (50 s: 2.68_ATT ±0.35 _SD-word pairs, 60 s: 3.72_ATT ±0.32 _SD-word pairs). Educational Attainment was another significant predictor of PAL scores (B_Education = 0.31, p_Education <2e-16, Figure 3). For both women and men, each milestone of Educational Attainment was associated with around a third of a word pair higher PAL score. In women and men, PSM revealed an approximate effect size of one-word pair higher performance across each level of Educational Attainment (e.g., Male 12y vs. 14y [_ATT0.72 ± _SD0.28], Female 12y vs. 14y=[_ATT1.06 ± _SD0.24]). However, the magnitude of this effect was not consistent for men and women across levels of Educational Attainment. Indeed, PSM estimated that women have a higher PAL score related to education than men, at all except the highest level of Educational Attainment (16y vs. 20y: Women=[_ATT0.97 ± _SD0.12], Men=[_ATT1.29 ± _SD0.16]).

Figure 2

Download asset Open asset

Figure 3

Download asset Open asset

Notably, there was a significant main effect of FH (B_FH = −2.39, p_FH = 3.47e-06, Figure 4A.). This coefficient indicated that for women and men, having FH was associated with nearly a two-and-a-half-word pair lower PAL scores when compared to FH- participants. Further our model revealed a significant Age x Diabetes (B_Age*Diabetes = 0.03, p_Age*Diabetes = 0.03) and Age x FH (B_Age*FH = 0.02, p_Age*FH <0.01, Figure 4A) interaction. The significant interactions between Age and Diabetes as well as Age and FH indicate a greater association between Diabetes and FH with lower PAL scores at younger ages as compared to older ages. Indeed, the linear trend lines for diabetes cross at 50 years of age (data not shown) and at age 65 for FH (Figure 4A). Due to the significant Age x FH interaction, our analyses evaluating interactions with FH included only participants ≤ 65 years old. A significant Sex x FH interaction (B_Sex*FH = −0.79, p_Sex*FH = 7.97e-06, Figure 4B) was found. Follow-up analyses of the estimated marginal mean (EMM, see Statistical Methods in Materials and methods) revealed that FH +women and men had lower PAL scores compared to FH- women and men (women: EMM = 1.01, t(53763) = 4.18, p<0.01; men: EMM = 1.58, t(53763) = 6.03, p<0.001); however, the magnitude of this effect was different between women and men. Indeed, PSM revealed that the estimated word pair reduction in PAL scores associated with FH status was larger in men (e.g., 20 s: _ATT1.93 ± _SD0.60) as compared to women (e.g., 20 s:_ATT0.60 ± _SD0.67), except for the 6^th decade of life (60 s: Women[_ATT-0.50 ± _SD0.18], Men[_ATT0.09 ± _SD−0.36], Figure 4 Inset).

Figure 4

Download asset Open asset

There was a significant Educational Attainment x FH interaction (B_Education*FH = 0.08, p_Education*FH <0.01). Follow-up analyses of the EMMs showed that FH was associated with lower PAL score at each milestone of educational attainment (ts(53758) > 3.45, ps <0.01); the magnitude of the associated decline in PAL scores attributed to FH was greater at lower milestones of educational attainment as compared to higher milestones (e.g., 12y [High school diploma]: FH- EMM = 16.4, FH +EMM = 14.8, difference = 1.6; 20y [Post graduate degree]: FH- EMM = 20.8, FH +EMM = 19.6, difference = 1.2). In FH +or FH- women and men 18–65 years old, our model revealed a significant FH x Diabetes (B_FH*Diabetes = −0.71, p_FH*Diabetes = 0.04, Figure 5) interaction. Follow-up analyses of the EMM revealed that participants with FH and diabetes were associated with lower PAL scores compared to participants with FH but no diabetes (FH +Diabetes + EMM = 16.9, FH +Diabetes- EMM = 18.1, t(53763) = 2.73, p=0.03). Lastly, two models which included either: 1) the Number of APOE ε4 alleles (i.e., 0, 1, or two copies) or 2) APOE Genotype (e.g., ε2-ε2, ε2-ε3, ε4-ε4, etc.) were evaluated. There was a significant main effect for the Number of APOE ε4 Alleles variable (B_ε4Allele = −1.30, p_ε4Allele = 0.02, Figure 6). These data indicated that in FH +participants, there was an association of one and a third word pair lower PAL score per each ε4 allele. Moreover, there was a significant main effect in the model comparing the ε4-ε4 to the ε2-ε3 genotype (B_{ε4-ε4 vs.ε2-ε3}3−5.32, p_{ε4-ε4 vs.ε2-ε3}30.03). Similarly, FH +participants with the ε4-ε4 genotype were associated with a five-word pair lower PAL scores compared to FH +participants with the ε2-ε3 genotype.

Figure 5

Download asset Open asset

Figure 6

Download asset Open asset

Overall, this study finds that having a first-degree relative diagnosed with Alzheimer’s disease (FH) is associated with lower verbal learning and memory performance (i.e., paired-associates learning; PAL) below the age of 65. Notably, FH men showed a greater reduction in cognition compared to FH women. This effect is not surprising since women, regardless of FH, perform better on PAL compared to men (Kaushanskaya et al., 2011), an effect replicated in this study. To that end, we demonstrate the sex-effect for PAL extends across the entire adult lifespan; indeed, the first time to the authors’ knowledge that this has been demonstrated: 1) in a large cohort, 2) using the same test, and 3) in a single study. It is interesting that our study found the associated disparity between women’s and men’s PAL scores enlarged around the 5^th decade of life. The 5^th decade of life is the approximate age when women undergo menopause in developed countries. Menopause-related changes to women’s hormonal milieu, either endogenously or via hormone treatment or gynecological surgery have been found to alter cognition during this period (discussed in Koebele and Bimonte-Nelson, 2016). Future studies of this cohort will dissect which medical choices at menopause, and medical choices earlier in a woman’s lifespan, may underlie better preservation of verbal memory in middle aged women as compared to men. Lastly, educational attainment was associated with milestone-dependent higher learning and memory performance. Regardless of FH status, education in women was associated with better PAL scores compared to men, at all except the highest reported level of educational attainment (i.e., 16 vs. 20 years of education). The effect of education on PAL scores is in line with observations in a clinical study demonstrating that higher levels of education are associated with delayed AD onset (Vemuri et al., 2014; Wang et al., 2012).

In terms of health, lifestyle, and genetic factors, this study found that diabetes modified the effect of FH on PAL. Specifically, in both women and men, diabetes and FH were associated with reduced PAL performance. It is not surprising that diabetes exacerbates the effects of FH on cognition since diabetes has been linked to worse cognitive deficits in AD (Takeda and Morishita, 2018). Several factors may underlie this effect: 1) there are differences in the risk of dementia for type one diabetes and type two diabetes, 2) specific type two diabetes treatments may reduce age-related declines in neural metabolism (Hamed, 2017; Kuo et al., 2018), and 3) vascular damage may combine with other genetic and environmental factors. As for genetic factors that modify FH, we supported our hypothesis whereby the presence of an APOE ε4 allele was associated with lower PAL performance in a dose-dependent-like manner in FH individuals. These data suggest that APOE genotype is an important genetic factor that influences memory. Our findings are in line with results from a voxel-wise study in humans noting a synergistic effect of FH and the APOE ε4 allele to intensify amyloid-beta deposition and reduce glucose use in regions of the MTL and other AD-related brain regions (Yi et al., 2018).

At the systems level, these results suggest that the collection of heritable and non-heritable changes due to FH status alter the functioning of the MTL and associated structures (Yi et al., 2018). The fact that these effects were only observed in participants that were ≤65 years old could be due to age-related differences in the effect of FH or participation differences between younger and older FH test takers. The ≤65 years old effect is partially in line with earlier smaller and more age-specific studies of FH status and cognition demonstrating an effect of FH, but only in children and younger adults (Bloss et al., 2008; Parra et al., 2015; Zeng et al., 2013). Since the prevalence of AD rises after age 60, it is also possible that older FH participants consenting to our study are those that have remained cognitively intact. Thus, participants that were experiencing noticeable age- or disease-related cognitive impairment may choose not to participate.

It is important to acknowledge some limitations of our work. FH risk is known to vary depending on the relationship of the diagnosed relative, and previous reports have demonstrated that first-degree FH results in higher risk for dementia compared to second- and third-degree FH (Cannon-Albright et al., 2019). In our study, we asked about the first-degree FH only; therefore, it is possible that individuals who have other FH risk from extended family members were included in our non-FH group. Additionally, it is possible that the form of AD, late-onset versus the rarer early-onset form, may encode different levels of FH risk. Future work is planned to investigate the FH effect in the study cohort, including an improved ability to separate late and early onset FH for each participant as well as to inquire about additional extended family member FH status.

It is likely that we did not measure all demographic, lifestyle, and health factors that are associated with differential PAL performance. One such example is socioeconomic status (SES). SES has been shown to have an association with brain structure and cognitive measures during development (reviewed in Brito and Noble, 2014) and work also suggests SES could play a role in AD risk (Qian et al., 2014; Stępkowski et al., 2015; reviewed in Seifan et al., 2015). Importantly, while we did not measure SES directly, we did assess factors commonly used to construct the SES composite (e.g., Educational Attainment). In addition, due to the international recruitment of our study cohort, normalization of the SES construct is complicated due to differing definitions of the factors used to calculate SES across nations (Rose et al., 2005).

Due to the large, distributed, and electronic nature of our study cohort, we rely on self-report answers to demographic, lifestyle, and health questions. Current studies comparing self-report data given over the internet versus in-person collected data show anywhere from a 0.3–20% discrepancy for height and weight measurements (Maukonen et al., 2018; Nikolaou et al., 2017). To investigate the potential role that such error may play on our FH AD effect, we re-analyzed the FH effect after introducing additional error into the coding of the FH self-report response. Additional error was added by randomly re-assigning FH status to various percentages of the cohort (stepwise from 2–30% of individuals) and re-analyzing the effect of FH using our complete statistical model. This was performed a total of 10,000 times for each error percentage, and the resulting influences on the p-value were reported using boxplots. With 8% additional introduced error we are able to show statistically significant effect of FH on PAL in 100% of the 10,000 tested iterations while an additional 24% error in FH status would still result in a statistically significant effect of FH on PAL in over 50% of the iterations (see Figure 1—figure supplement 2). These results suggest that it is unlikely that FH self-report error is driving the significant effect of FH on PAL. Lastly, PAL was tested cross-sectionally in the cohort; therefore, determinations about the influence of collected factors on trajectories of change in performance across time within an individual subject are not possible. Additional longitudinal-based studies will be necessary to identify this class of variables.

Collectively, this study supports recommendations underscoring the importance of living a healthy lifestyle, properly treating disease states such as diabetes, and building cognitive reserve through education (i.e., risk reduction) to attenuate age- and AD-risk-related cognitive declines. Further, our findings specifically highlight the positive effects of such interventions on FH-associated risk, opening the door to the development of more targeted risk reduction approaches to combat AD. In addition, this work underscores the utility of web-based cohort recruitment and study; thus, facilitating large sample sizes in a cost- and time-effective fashion. Statistical power concerns are common in many scientific studies, and underpowered study designs can lead to several potential issues associated with false positive and negative rates. It should be acknowledged that web-based studies are not without concerns, however, we propose that the advantage of considerably larger sample sizes and enriched cohort diversity in online research mostly diminishes the potential disadvantages.

The data that support the findings of this study are freely available at Dryad (https://datadryad.org) doi:10.5061/dryad.2867k2m).

1. 1. Talboom J
   2. Håberg A
   3. De Both M
   4. Naymik M
   5. Schrauwen I
   6. Lewis C
   7. Bertinelli S
   8. Hammersland C
   9. Fritz M
   10. Myers A
   11. Hay M
   12. Barnes C
   13. Glisky E
   14. Ryan L
   15. Huentelman M

   (2019) Dryad Digital Repository

   Data from: Family history of Alzheimer's disease alters cognition and is modified by medical and genetic factors.

   https://doi.org/10.5061/dryad.j1fd7

1. 1. Aschenbrenner AJ
   2. Balota DA
   3. Gordon BA
   4. Ratcliff R
   5. Morris JC

   (2016) A diffusion model analysis of episodic recognition in preclinical individuals with a family history for Alzheimer's disease: The adult children study

   Neuropsychology 30:225–238.

   https://doi.org/10.1037/neu0000222

   • PubMed
   • Google Scholar
2. 1. Barnes DE
   2. Yaffe K

   (2011) The projected effect of risk factor reduction on Alzheimer's disease prevalence

   The Lancet Neurology 10:819–828.

   https://doi.org/10.1016/S1474-4422(11)70072-2

   • PubMed
   • Google Scholar
3. 1. Baumgart M
   2. Snyder HM
   3. Carrillo MC
   4. Fazio S
   5. Kim H
   6. Johns H

   (2015) Summary of the evidence on modifiable risk factors for cognitive decline and dementia: a population-based perspective

   Alzheimer's & Dementia 11:718–726.

   https://doi.org/10.1016/j.jalz.2015.05.016

   • PubMed
   • Google Scholar
4. 1. Bloss CS
   2. Delis DC
   3. Salmon DP
   4. Bondi MW

   (2008) Decreased cognition in children with risk factors for Alzheimer's disease

   Biological Psychiatry 64:904–906.

   https://doi.org/10.1016/j.biopsych.2008.07.004

   • PubMed
   • Google Scholar
5. 1. Braak H
   2. Thal DR
   3. Ghebremedhin E
   4. Del Tredici K

   (2011) Stages of the pathologic process in Alzheimer disease: age categories from 1 to 100 years

   Journal of Neuropathology & Experimental Neurology 70:960–969.

   https://doi.org/10.1097/NEN.0b013e318232a379

   • PubMed
   • Google Scholar
6. 1. Brito NH
   2. Noble KG

   (2014) Socioeconomic status and structural brain development

   Frontiers in Neuroscience 8:276.

   https://doi.org/10.3389/fnins.2014.00276

   • PubMed
   • Google Scholar
7. 1. Cannon-Albright LA
   2. Foster NL
   3. Schliep K
   4. Farnham JM
   5. Teerlink CC
   6. Kaddas H
   7. Tschanz J
   8. Corcoran C
   9. Kauwe JSK

   (2019) Relative risk for Alzheimer disease based on complete family history

   Neurology 92:e1745–e1753.

   https://doi.org/10.1212/WNL.0000000000007231

   • PubMed
   • Google Scholar
8. 1. Chang TS
   2. Coen MH
   3. La Rue A
   4. Jonaitis E
   5. Koscik RL
   6. Hermann B
   7. Sager MA

   (2012) Machine learning amplifies the effect of parental family history of Alzheimer's disease on list learning strategy

   Journal of the International Neuropsychological Society 18:428–439.

   https://doi.org/10.1017/S1355617711001834

   • PubMed
   • Google Scholar
9. 1. Green RC
   2. Cupples LA
   3. Go R
   4. Benke KS
   5. Edeki T
   6. Griffith PA
   7. Williams M
   8. Hipps Y
   9. Graff-Radford N
   10. Bachman D
   11. Farrer LA
   12. MIRAGE Study Group

   (2002) Risk of dementia among white and African American relatives of patients with Alzheimer disease

   JAMA 287:329–336.

   https://doi.org/10.1001/jama.287.3.329

   • PubMed
   • Google Scholar
10. 1. Hamed SA

    (2017) Brain injury with diabetes mellitus: evidence, mechanisms and treatment implications

    Expert Review of Clinical Pharmacology 10:409–428.

    https://doi.org/10.1080/17512433.2017.1293521

    • PubMed
    • Google Scholar
11. 1. Hebert LE
    2. Weuve J
    3. Scherr PA
    4. Evans DA

    (2013) Alzheimer disease in the united states (2010-2050) estimated using the 2010 census

    Neurology 80:1778–1783.

    https://doi.org/10.1212/WNL.0b013e31828726f5

    • PubMed
    • Google Scholar
12. 1. Ho DE
    2. Imai K
    3. King G
    4. Stuart EA

    (2007)

    Matching as nonparametric preprocessing for reducing model dependence in parametric causal inference

    Political Analysis 15:199–236.

    • Google Scholar
13. 1. Ho DE
    2. Imai K
    3. King G
    4. Stuart EA

    (2011)

    MatchIt : nonparametric preprocessing for parametric causal inference

    Journal of Statistical Software 42:1–28.

    • Google Scholar
14. 1. Honea RA
    2. Vidoni E
    3. Harsha A
    4. Burns JM

    (2009) Impact of APOE on the healthy aging brain: a voxel-based MRI and DTI study

    Journal of Alzheimer's Disease 18:553–564.

    https://doi.org/10.3233/JAD-2009-1163

    • PubMed
    • Google Scholar
15. 1. Imai K
    2. King G
    3. Lau O

    (2008) Toward a common framework for statistical analysis and development

    Journal of Computational and Graphical Statistics 17:892–913.

    https://doi.org/10.1198/106186008X384898

    • Google Scholar
16. Software

    1. Imai K
    2. King G
    3. Lau O

    (2009)

    Zelig: Everyone’s statistical software

    R Package Version.
17. 1. Kaushanskaya M
    2. Marian V
    3. Yoo J

    (2011) Gender differences in adult word learning

    Acta Psychologica 137:24–35.

    https://doi.org/10.1016/j.actpsy.2011.02.002

    • PubMed
    • Google Scholar
18. 1. Koebele SV
    2. Bimonte-Nelson HA

    (2016) Modeling menopause: the utility of rodents in translational behavioral endocrinology research

    Maturitas 87:5–17.

    https://doi.org/10.1016/j.maturitas.2016.01.015

    • PubMed
    • Google Scholar
19. 1. Krokstad S
    2. Langhammer A
    3. Hveem K
    4. Holmen TL
    5. Midthjell K
    6. Stene TR
    7. Bratberg G
    8. Heggland J
    9. Holmen J

    (2013) Cohort profile: the HUNT study, Norway

    International Journal of Epidemiology 42:968–977.

    https://doi.org/10.1093/ije/dys095

    • PubMed
    • Google Scholar
20. 1. Kuo CL
    2. Lu CL
    3. Chang YH
    4. Li CY

    (2018) Population-Based cohort study on dementia risk in patients with type 1 diabetes mellitus

    Neuroepidemiology 50:57–62.

    https://doi.org/10.1159/000486719

    • PubMed
    • Google Scholar
21. 1. La Rue A
    2. Hermann B
    3. Jones JE
    4. Johnson S
    5. Asthana S
    6. Sager MA

    (2008) Effect of parental family history of Alzheimer's disease on serial position profiles

    Alzheimer's & Dementia 4:285–290.

    https://doi.org/10.1016/j.jalz.2008.03.009

    • PubMed
    • Google Scholar
22. 1. Maukonen M
    2. Männistö S
    3. Tolonen H

    (2018) A comparison of measured versus self-reported anthropometrics for assessing obesity in adults: a literature review

    Scandinavian Journal of Public Health 46:565–579.

    https://doi.org/10.1177/1403494818761971

    • PubMed
    • Google Scholar
23. 1. Miller KJ
    2. Rogers SA
    3. Siddarth P
    4. Small GW

    (2005) Object naming and semantic fluency among individuals with genetic risk for Alzheimer's disease

    International Journal of Geriatric Psychiatry 20:128–136.

    https://doi.org/10.1002/gps.1262

    • PubMed
    • Google Scholar
24. 1. Nikolaou CK
    2. Hankey CR
    3. Lean MEJ

    (2017) Accuracy of on-line self-reported weights and heights by young adults

    European Journal of Public Health 27:898–903.

    https://doi.org/10.1093/eurpub/ckx077

    • PubMed
    • Google Scholar
25. 1. Oh MJ
    2. Chung EK EK
    3. Shin YM
    4. Lee KO
    5. Park YS

    (1997)

    The apolipoprotein E genotyping using the PCR-RFLP was useful to linkage analysis of Alzheimer's disease families

    Experimental & Molecular Medicine 29:161–164.

    • Google Scholar
26. 1. Parra MA
    2. Saarimäki H
    3. Bastin ME
    4. Londoño AC
    5. Pettit L
    6. Lopera F
    7. Della Sala S
    8. Abrahams S

    (2015) Memory binding and white matter integrity in familial Alzheimer's disease

    Brain 138:1355–1369.

    https://doi.org/10.1093/brain/awv048

    • PubMed
    • Google Scholar
27. 1. Pike KE
    2. Rowe CC
    3. Moss SA
    4. Savage G

    (2008) Memory profiling with paired associate learning in Alzheimer's disease, mild cognitive impairment, and healthy aging

    Neuropsychology 22:718–728.

    https://doi.org/10.1037/a0013050

    • PubMed
    • Google Scholar
28. 1. Pike KE
    2. Kinsella GJ
    3. Ong B
    4. Mullaly E
    5. Rand E
    6. Storey E
    7. Ames D
    8. Saling M
    9. Clare L
    10. Parsons S

    (2013) Is the WMS-IV verbal paired associates as effective as other memory tasks in discriminating amnestic mild cognitive impairment from normal aging?

    The Clinical Neuropsychologist 27:908–923.

    https://doi.org/10.1080/13854046.2013.809149

    • PubMed
    • Google Scholar
29. 1. Qian W
    2. Schweizer TA
    3. Fischer CE

    (2014) Impact of socioeconomic status on initial clinical presentation to a memory disorders clinic

    International Psychogeriatrics 26:597–603.

    https://doi.org/10.1017/S1041610213002299

    • PubMed
    • Google Scholar
30. 1. Reynolds CA
    2. Finkel D

    (2015) A meta-analysis of heritability of cognitive aging: minding the "missing heritability" gap

    Neuropsychology Review 25:97–112.

    https://doi.org/10.1007/s11065-015-9280-2

    • PubMed
    • Google Scholar
31. 1. Ridge PG
    2. Mukherjee S
    3. Crane PK
    4. Kauwe JS
    5. Alzheimer’s Disease Genetics Consortium

    (2013) Alzheimer's disease: analyzing the missing heritability

    PLOS ONE 8:e79771.

    https://doi.org/10.1371/journal.pone.0079771

    • PubMed
    • Google Scholar
32. 1. Roberts JS
    2. Barber M
    3. Brown TM
    4. Cupples LA
    5. Farrer LA
    6. LaRusse SA
    7. Post SG
    8. Quaid KA
    9. Ravdin LD
    10. Relkin NR
    11. Sadovnick AD
    12. Whitehouse PJ
    13. Woodard JL
    14. Green RC

    (2004) Who seeks genetic susceptibility testing for Alzheimer's disease? Findings from a multisite, randomized clinical trial

    Genetics in Medicine 6:197–203.

    https://doi.org/10.1097/01.GIM.0000132688.55591.77

    • PubMed
    • Google Scholar
33. Book

    1. Rose D
    2. Pevalin DJ
    3. O'Reilly K

    (2005)

    The National Statistics Socio-Economic Classification: origins Development And Use

    Basingstoke: Palgrave MacMillan.

    • Google Scholar
34. 1. Seifan A
    2. Schelke M
    3. Obeng-Aduasare Y
    4. Isaacson R

    (2015) Early life epidemiology of Alzheimer's Disease--A Critical Review

    Neuroepidemiology 45:237–254.

    https://doi.org/10.1159/000439568

    • PubMed
    • Google Scholar
35. 1. Sperling RA
    2. Jack CR
    3. Aisen PS

    (2011) Testing the right target and right drug at the right stage

    Science Translational Medicine 3:111cm33.

    https://doi.org/10.1126/scitranslmed.3002609

    • PubMed
    • Google Scholar
36. 1. Stępkowski D
    2. Woźniak G
    3. Studnicki M

    (2015) Correlation of Alzheimer's disease death rates with historical per capita personal income in the USA

    PLOS ONE 10:e0126139.

    https://doi.org/10.1371/journal.pone.0126139

    • PubMed
    • Google Scholar
37. Book

    1. Takeda S
    2. Morishita R

    (2018)

    Diabetes and Alzheimer’s Disease

    In: Yamagishi S. I, editors. Diabetes and Aging-Related Complications. Singapore: Springer. pp. 101–111.

    • Google Scholar
38. 1. Vemuri P
    2. Lesnick TG
    3. Przybelski SA
    4. Machulda M
    5. Knopman DS
    6. Mielke MM
    7. Roberts RO
    8. Geda YE
    9. Rocca WA
    10. Petersen RC
    11. Jack CR

    (2014) Association of lifetime intellectual enrichment with cognitive decline in the older population

    JAMA Neurology 71:1017–1024.

    https://doi.org/10.1001/jamaneurol.2014.963

    • PubMed
    • Google Scholar
39. 1. Wang H-X
    2. Gustafson DR
    3. Kivipelto M
    4. Pedersen NL
    5. Skoog I
    6. Windblad B
    7. Fratiglioni L

    (2012) Education halves the risk of dementia due to apolipoprotein ε4 allele: a collaborative study from the swedish brain power initiative

    Neurobiology of Aging 33:1007.e1–101007.

    https://doi.org/10.1016/j.neurobiolaging.2011.10.003

    • Google Scholar
40. 1. Yi D
    2. Lee Y
    3. Byun MS
    4. Lee JH
    5. Ko K
    6. Sohn BK
    7. Choe YM
    8. Choi HJ
    9. Baek H
    10. Sohn CH
    11. Kim YK
    12. Lee DY
    13. KBASE research group

    (2018) Synergistic interaction between APOE and family history of Alzheimer's disease on cerebral amyloid deposition and glucose metabolism

    Alzheimer's Research & Therapy 10:84.

    https://doi.org/10.1186/s13195-018-0411-x

    • PubMed
    • Google Scholar
41. 1. Zeng Y
    2. Chang W
    3. Shu C
    4. Ma L
    5. Huang Y
    6. Wang R
    7. Zhang J
    8. Zhu C
    9. McClintock SM

    (2013) Decreased cognitive function in extended family members from the single late-onset-Alzheimer's-disease pedigree

    Journal of the International Neuropsychological Society 19:809–819.

    https://doi.org/10.1017/S1355617713000581

    • PubMed
    • Google Scholar

Thank you for submitting your article "Family history of Alzheimer's disease alters cognition and is modified by medical and genetic factors" for consideration by eLife. Your article has been reviewed by two reviewers, one of whom is a member of our Board of Reviewing Editors, and the evaluation has been overseen by a Senior Editor. The reviewers have opted to remain anonymous.

As standard practice in eLife, the reviewers have discussed their critiques with one another in an online panel moderated by the Reviewing Editor. After consensus was reached, the Reviewing Editor has drafted this decision to help you prepare a revised submission.

Summary:

This is an interesting and potentially important study investigating the effects of family history (FH) of Alzheimer's disease on memory function across the lifespan. Using an online assessment platform (MindCrowd) the authors probed paired-associate verbal learning as an index of MTL-dependent learning from 18-85 years. The sample size of ~60,000 people is impressive, and permits a well-powered exploration of factors that potentially mediate the effects of FH on memory across the lifespan. The inclusion of APOE genotype data in a large subgroup is a further strength of the study. The most significant finding from this study is that FH of Alzheimer's disease was associated with poorer verbal paired-associate learning performance in individuals <65 years, and this effect appeared to be particularly pronounced in men. In general, the paper is well written and the findings are potentially of clinical importance, suggesting that distinct alterations on the PAL may be a useful marker of memory decline from middle-age.

Essential revisions:

1) A significant question concerns how the construct of FH was defined for participants. The exact wording for this measure must be provided as there are a number of possible options. Further, was FH related to all members in the family or not? Was FH defined in terms of both sporadic and familial AD? FH could be defined as a binary value (as in the study) or as numeric variable related to familial AD and these are important distinctions to state upfront. Please clarify throughout the manuscript.

2) The findings presented hinge upon the reliability of the self-reported data regarding FH and lifestyle conditions, however the validity and reliability of these data are not presented. Exactly what type of information was requested from participants to determine a FH of AD – important variables to consider include age of onset, specific symptom profiles, whether there was functional decline (i.e., differentiating between aMCI and AD).

Further, the reports on lifestyle characteristics are unclear. We would like to see further details regarding how the questions were phrased and how they were coded in the study. What is the reliability of these questions – is it possible that a specific answer might be interpreted and responded to differently across cultures/languages etc.?

3) While the demographic questionnaires cover a number of key factors, socioeconomic status (SES) was not explored. This is not a trivial omission as a number of the factors of interest could potentially be explained by lower SES. For example, educational attainment, dietary/medical risk factors, alcohol/drug consumption, diabetes, and even family history of AD may all relate to, or interact with, lower SES. How lower SES potentially contributes to, or interacts with, these risk factors is an important issue to address as it might be a common underlying variable that modulates the risk for cognitive decline. Further, SES may relate to selection bias of individuals who volunteer to take part in such studies. Please clarify and include these data, if possible, as a covariate of interest.

4) The potential role of other risk factors/disease does not appear to have been taken into account. Please clarify or state as a limitation.

5) The result that PAL performance was linearly associated with participant's age (Figure 2) seems not to be in agreement with findings of memory decline across age which is not described by linear functions but a sinuous curve (cf. Jack et al. on dynamics in AD development). Can the authors please comment on this discrepancy?

6) The discussion of sex-differences around the 5th decade in relation to menopause and modulating effects of estrogen is particularly interesting. Estrogen has been suggested to offer a neuroprotective effect and it would be interesting to know whether sharp declines in PAL performance are evident in women from the 6th decade onwards (when presumably hormonal levels are much lower, and hysterectomy may have occurred). This relates to the somewhat arbitrary age cut-off point of 65 years – in truncating the age-range of participants, the authors may conceal important hormonally-driven transitions in cognitive function in women. Is it possible to look at these factors earlier in the lifespan using these data?

Essential revisions:

1) A significant question concerns how the construct of FH was defined for participants. The exact wording for this measure must be provided, as there are a number of possible options. Further, was FH related to all members in the family or not? Was FH defined in terms of both sporadic and familial AD? FH could be defined as a binary value (as in the study) or as numeric variable related to familial AD and these are important distinctions to state upfront. Please clarify throughout the manuscript.

We thank the review team for their thoughtful comments regarding FH. We recognize the importance of properly defining the FH construct for the study participant and for this reason we directly indicate what is meant by “first-degree relative” in the wording of the question on the MindCrowd site. A screenshot is included in Author response image 1 showing exactly what each test taker sees when answering this question.

Author response image 1

Download asset Open asset

We spent time reviewing prior studies on FH to come to what we consider the most accurate question to probe the potential influence of FH on cognitive performance in our cohort. Prior studies have demonstrated a stronger influence on dementia risk if the FH of Alzheimer’s disease (AD) was within a first-degree relative versus other relatives (Cannon-Albright et al., 2019). Due to this knowledge and our interest in balancing our ability to recruit a large-sized cohort with the parallel goal of collecting as much demographic, medical, and lifestyle data as would be palatable to a participant, we restricted the question regarding FH on the main MindCrowd site to that of first-degree relatives only. We added in exactly how the FH question was asked:

“FH status was determined by self-report of having (FH⁺), or not having (FH-) based on the question, “Have you, a sibling, or one of your parents been diagnosed with Alzheimer's disease?”

Like the reviewers, we also have an interest in further understanding FH risk in early vs. late AD diagnosed relatives as well as second-degree and third-degree relatives. We chose not to ask questions related to these on the main MindCrowd site to facilitate overall cohort recruitment.

However, in January of 2018, we initiated a follow-up survey questionnaire for all study participants who voluntarily provided us with contact information. In this survey, we ask additional questions related to FH – including extended family member AD diagnoses and the age of diagnosis for their first-degree relative. To date, over 6,000 individuals have answered these survey questions and we have identified 1.3% (n=84) in this survey cohort with a self-reported first-degree relative diagnosed with early-onset AD (defined for the survey participant as a diagnosis under the age of 55). We explored these data and found that early onset FH resulted in a lowered PAL performance as well – mirroring what we report in the manuscript for the first-degree FH regardless of the age of diagnosis. While the proportion of early-onset AD approximates the expected population frequency for this rare form of AD, we believe this cohort is still undersized for appropriately powered statistical analyses and therefore is beyond the scope of the current manuscript.

We have included a Supplementary file in the manuscript that lists all the demographic, medical, and lifestyle questions as they were worded for the participant as well as how the answers were coded for our statistical analyses (Supplementary file 2). In the manuscript, we include an additional Discussion of FH AD risk:

“It is important to acknowledge some limitations of our work. FH risk is known to vary depending on the relationship of the diagnosed relative, and previous reports have demonstrated that first-degree FH results in higher risk for dementia compared to second- and third-degree FH (Cannon-Albright et al., 2019). In our study, we asked about the first-degree FH only; therefore, it is possible that individuals who have other FH risk from extended family members were included in our non-FH group. Additionally, it is possible that the form of AD, late-onset versus the rarer early-onset form, may encode different levels of FH risk. Future work is planned to investigate the FH effect in the study cohort, including an improved ability to separate late and early onset FH for each participant as well as to inquire about additional extended family member FH status.”

2) The findings presented hinge upon the reliability of the self-reported data regarding FH and lifestyle conditions, however the validity and reliability of these data are not presented. Exactly what type of information was requested from participants to determine a FH of AD – important variables to consider include age of onset, specific symptom profiles, whether there was functional decline (i.e., differentiating between aMCI and AD).

Further, the reports on lifestyle characteristics are unclear. We would like to see further details regarding how the questions were phrased and how they were coded in the study. What is the reliability of these questions – is it possible that a specific answer might be interpreted and responded to differently across cultures/languages etc?

We agree that self-report data are reliant on the accuracy of each of those self-reported variables. We did not specifically study the reliability of our particular self-report data as there have been many studies to examine this question previously (Maukonen, Mannisto, and Tolonen, 2018; Nikolaou, Hankey, and Lean, 2017). The conclusion from these previously published examinations of self-report reliability is that the error of the self-reported answer is unique for each variable and can be related to the complexity of the question and answer options (e.g. dichotomous vs. quantitative) as well as to social influences related to the answer, however, the range of error reported in these studies was between 0.3-20% for quantitative traits like height and weight (Elgar and Stewart, 2008; Ikeda, 2016; Shields et al., 2008; Yoon et al., 2014). We do believe it is possible that a self-report answer could also vary across cultures, as the review team pointed out; however, we would like to emphasize that only individuals who self-declared English as their primary language were included in the analyses reported in the manuscript. Therefore, we believe that this potential cultural effect is likely minimized in the cohort used for analysis in our manuscript.

We addressed the details of the wording and rationale for our FH question above in the answer to Comment #1 and we agree with the reviewers that publishing exactly how each demographic, medical, and lifestyle question was asked and how the data were coded during statistical analysis is important and we have now included all of these details in Supplementary file 2.

We performed two additional analyses to test the impact of the self-report error on the FH finding. First, we re-analyzed the FH effect after introducing additional error into the coding of the FH self-report response. This was done by randomly re-assigning FH status to various percentages of the MindCrowd cohort (stepwise from 2-30% of individuals) and re-analyzing the effect of FH using our complete statistical model. This was performed a total of 10,000 times for each error percentage and the results on the p-value were reported using boxplots. It is important to note that this error is an additional amount introduced on top of whatever may already be present in the MindCrowd data due to self-report of FH. With 8% additional introduced error we are able to show that we would still report a statistically significant effect of FH on PAL in 100% of the 10,000 tested iterations. An additional 24% error in FH status would still result in a statistically significant effect of FH on PAL in over 50% of the 10,000 iterations. These data suggest that it is highly unlikely that FH self-report error is driving the significant effect of FH on PAL we report in the manuscript.

Secondly, we examined the FH effect on PAL through the use of permutation testing. This is an approach utilized to determine the probability of a false positive finding under the null hypothesis. The FH data label for every participant was randomly assigned and the t-statistic for the FH effect on PAL was re-calculated. This was performed one million times. In every case, not a single t-statistic arising from a permutation of the FH data label was observed to be more extreme than our reported t-statistic (permuted p-value = <1e-6). This suggests that the odds of our FH finding being observed due to chance alone is less than one in one million.

These findings are illustrated in Author response image 2. Figure A from Author response image 2 was added to the manuscript as “Figure 1—figure supplement 2”.

Author response image 2

Download asset Open asset

Further, while acknowledging the potential drawbacks with web-based self-report approaches, we agree with the authors of a recent book chapter who suggest some advantages to web-based self-report data collection compared to traditional in-person methods, "The presence of the researcher at the completion of the questionnaire may affect answers as well. For example, subjects may change their behavior or demonstrate an improvement in their outcome because they know they are being observed. More specifically, in the presence of the examiner, a responder may not feel comfortable selecting the extreme choices (Demetriou, Ozer, and Essau; 2015).” This is a known effect of self-report data collected in the presence of a medical or research professional. In short, societal pressures can lead to over- and under-reporting of certain variables (e.g. reporting less body weight) and there is suggestive evidence that the anonymity afforded by an internet self-report collection approach may actually lead to lower rates of self-report error due to this fact.

The addition of self-report error is added as a potential limitation to our study in the Discussion section of the manuscript:

“Due to the large, distributed, and electronic nature of our study cohort, we rely on self-report answers to demographic, lifestyle, and health questions. Current studies comparing self-report data given over the internet versus in-person collected data show anywhere from a 0.3-20% discrepancy for height and weight measurements (Maukonen, Mannisto, and Tolonen, 2018; Nikolaou, Hankey, and Lean, 2017). To investigate the potential role that such error may play on our FH AD effect, we re-analyzed the FH effect after introducing additional error into the coding of the FH self-report response. Additional error was added by randomly re-assigning FH status to various percentages of the cohort (stepwise from 2-30% of individuals) and re-analyzing the effect of FH using our complete statistical model. This was performed a total of 10,000 times for each error percentage, and the resulting influences on the p-value were reported using boxplots. With 8% additional introduced error we are able to show statistically significant effects of FH on PAL in 100% of the 10,000 tested iterations while an additional 24% error in FH status would still result in a statistically significant effect of FH on PAL in over 50% of the iterations (see Figure 1—figure supplement 2). These results suggest that it is unlikely that FH self-report error is driving the significant effect of FH on PAL.”

3) While the demographic questionnaires cover a number of key factors, socioeconomic status (SES) was not explored. This is not a trivial omission as a number of the factors of interest could potentially be explained by lower SES. For example, educational attainment, dietary/medical risk factors, alcohol/drug consumption, diabetes, and even family history of AD may all relate to, or interact with, lower SES. How lower SES potentially contributes to, or interacts with, these risk factors is an important issue to address as it might be a common underlying variable that modulates the risk for cognitive decline. Further, SES may relate to selection bias of individuals who volunteer to take part in such studies. Please clarify and include these data, if possible, as a covariate of interest.

We agree with the review team that SES is an essential factor to consider. Our initial 26 demography and health questions included educational attainment, but not income. Hence, it is not possible to evaluate SES across all of our participants. However, all analyses included education as well as all other covariates (see Supplementary file 2) in the statistical model.

Since the submission of this manuscript, we have been collecting responses from a follow-up survey allowing us to calculate both current (your current income and educational attainment) and childhood SES (your parent’s income and educational attainment). Initial evaluations of current and adult SES within this cohort of 6,187 individuals did not reveal a statistically significant association between the SES composite and PAL. The individual factors that contribute to the SES composite score – e.g. the income and educational attainment values – were also tested individually and only educational attainment was associated with PAL performance.

For reference, included as Author response image 3 is a figure illustrating the current income distributions within the MindCrowd survey cohort demonstrating that ~12% of the cohort has a current income equal to or less than $35,000. This amount approximates the U.S. national estimate for poverty rate (reported as 13.5% in 2015 by U.S. Census Bureau; 2017).

Author response image 3

Download asset Open asset

Further, we acknowledge that SES is a complex measure which is not defined the same across all disciplines (e.g., includes or excludes measures of education quality, access to health care, employment, health variables, etc.), and there appears to be marked cultural differences as to which factors best describe the SES construct as well (e.g., employment in the UK, see Rose et al., 2005). Because our cohort includes participants from all over the world, and the second largest portion of our data (n = 2060), next to the US, is based on participants in the UK, we think that traditional SES measurements may not adequately capture the SES construct in our data.

We now address the potential limitation of not measuring SES in our cohort in the Discussion section. Please see the response to review team critique #4 below for the exact added wording we utilize.

4) The potential role of other risk factors/disease does not appear to have been taken into account. Please clarify or state as a limitation.

We want to note for the review team that when a variable was tested for association with PAL, we included other collected demographic, lifestyle, and medical factors as covariates in order to control for their influence. In re-reading the description of the statistical methods it is evident that this fact wasn’t fully explained. To address this, we have updated the Materials and methods section:

“It is important to note that each of the above demographic variables and indicated interaction terms were included in every analysis unless stated otherwise due to model limitations.”

However, we do agree that it is possible that factors that may influence PAL were not collected via our list of demographic, lifestyle, and medical factors. We have added this limitation in the Discussion section:

“It is likely that we did not measure all demographic, lifestyle, and health factors that are associated with differential PAL performance. One such example is socioeconomic status (SES). SES has been shown to have an association with brain structure and cognitive measures during development (reviewed in Brito and Noble, 2014) and work also suggests SES could play a role in AD risk (Qian, Schweizer, and Fischer, 2014; Stepkowski, Wozniak, and Studnicki, 2015) reviewed in Seifan and colleagues (2015). Importantly, while we did not measure SES directly, we did assess factors commonly used to construct the SES composite (e.g., Educational Attainment). In addition, due to the international recruitment of our study cohort, normalization of the SES construct is complicated due to differing definitions of the factors used to calculate SES across nations (Rose, Pevalin, and O'Reilly, 2005).”

5) The result that PAL performance was linearly associated with participant's age (Figure 2) seems not to be in agreement with findings of memory decline across age which is not described by linear functions but a sinuous curve (cf. Jack et al. on dynamics in AD development). Can the authors please comment on this discrepancy?

The direct comparison with Jack et al. is problematic because those data (and the resulting model) are based on disease progression. Their model indicates how AD-relevant biomarkers can increase to non-normative levels decades before the cognitive decline is clinically registered. Therefore, the alteration of these disease-associated biomarkers could alter the cognitive decline curve for a disease cohort but may not be present – or if they are present, may exert a different magnitude of effect in a normative aging cohort like MindCrowd. We agree with the review panel that the linear decline with age in MindCrowd was surprising; however, it is clearly the best fit for what we observe in the study. In fact, another cross-sectional study of 1250 participants noted a similar linear decline in PAL performance (i.e., from the Wechsler Memory Scale III) from 18 to 80 years of age (Salthouse, 2003).

Future work leveraging longitudinal assessments of large normative aging cohorts like MindCrowd will help to further elaborate on the aging trajectory and will also help us better appreciate individual deviations from this trajectory. We view the non-longitudinal nature of MindCrowd as a current limitation. Indeed, a longitudinal study of our cohort may reveal a non-linear age-related decline in PAL performance. We have discussed this limitation further in the manuscript:

“Lastly, PAL was tested cross-sectionally in the cohort; therefore, determinations about the influence of collected factors on trajectories of change in performance across time within an individual subject are not possible. Additional longitudinal-based studies will be necessary to identify this class of variables.”

6) The discussion of sex-differences around the 5th decade in relation to menopause and modulating effects of estrogen is particularly interesting. Estrogen has been suggested to offer a neuroprotective effect and it would be interesting to know whether sharp declines in PAL performance are evident in women from the 6th decade onwards (when presumably hormonal levels are much lower, and hysterectomy may have occurred). This relates to the somewhat arbitrary age cut-off point of 65 years – in truncating the age-range of participants, the authors may conceal important hormonally driven transitions in cognitive function in women. Is it possible to look at these factors earlier in the lifespan using this data?

We are in agreement with the review team that the sex-difference around the 5th decade of life is exciting; however, we are confused by the review team comment regarding the cut-off point of 65 years for this sex effect. In Figure 2 in the manuscript, we demonstrate the sex effect across all ages 18 and above in the study.

The review team is likely referring to the age cut-off we utilized for the FH analysis. This was rationalized in the manuscript (see quote below) due to the convergence of the trend lines for the FH effect. We did not employ an age cut-off for any of the other analyzed variables, including sex.

“Indeed, the linear trend lines for diabetes cross at 50 years of age (data not shown) and at age 65 for FH (Figure 4A.). Due to the significant Age x FH interaction, our analyses evaluating interactions with FH included only participants ≤65 years old.”

In our follow-up survey mentioned earlier, we have started to look at menopause (hormone therapy and surgery options) and hormonal contraception choices. Due to the lower size of the female survey cohort (~4,000 female respondents to date), we think that the results of the survey thus far are outside of the realm of this particular publication. However, a further discussion of this effect and how it will be addressed in future studies was expanded in the Discussion section:

“It is interesting that our study found the associated disparity between women’s and men’s PAL scores enlarged around the 5th decade of life. The 5th decade of life is the approximate age when women undergo menopause in developed countries. Menopause-related changes to women’s hormonal milieu, either endogenously or via hormone treatment or gynecological surgery have been found to alter cognition during this period (discussed in Koebele and Bimonte-Nelson, 2016). Future studies of this cohort will dissect which medical choices at menopause, and medical choices earlier in a woman’s lifespan, may underlie better preservation of verbal memory in middle aged women as compared to men.”

References

Demetriou, C., Ozer, B. U., and Essau, C. A. (2015). Self-Report Questionnaires. In The Encyclopedia of Clinical Psychology.

Elgar, F. J., and Stewart, J. M. (2008). Validity of self-report screening for overweight and obesity. Evidence from the Canadian Community Health Survey. Can J Public Health, 99(5), 423-427.

Ikeda, N. (2016). Validity of Self-Reports of Height and Weight among the General Adult Population in Japan: Findings from National Household Surveys, 1986. PloS one, 11(2), e0148297. doi:10.1371/journal.pone.0148297

Jack, C. R., Jr., Knopman, D. S., Jagust, W. J., Shaw, L. M., Aisen, P. S., Weiner, M. W.,. . . Trojanowski, J. Q. (2010). Hypothetical model of dynamic biomarkers of the Alzheimer's pathological cascade. Lancet Neurol, 9(1), 119-128. doi:10.1016/S1474-4422(09)70299-6

Jack, C. R., Jr., Knopman, D. S., Jagust, W. J., Petersen, R. C., Weiner, M. W., Aisen, P. S.,. . . Trojanowski, J. Q. (2013). Tracking pathophysiological processes in Alzheimer's disease: an updated hypothetical model of dynamic biomarkers. Lancet Neurol, 12(2), 207-216. doi:10.1016/S1474-4422(12)70291-0

Salthouse, T. A. (2003). Memory aging from 18 to 80. Alzheimer Dis Assoc Disord, 17(3), 162-167.

Shields, M., Connor Gorber, S., and Tremblay, M. S. (2008). Estimates of obesity based on self-report versus direct measures. Health Rep, 19(2), 61-76.

U.S. Census Bureau data (2017); U.S. Department of Commerce, Bureau of Economic Analysis, Survey of Current Business; and DataQuick Information Systems, a public records database company located in La Jolla, San Diego, CA.

Yoon, K., Jang, S. N., Chun, H., and Cho, S. I. (2014). Self-reported anthropometric information cannot vouch for the accurate assessment of obesity prevalence in populations of middle-aged and older Korean individuals. Arch Gerontol Geriatr, 59(3), 584-592. doi:10.1016/j.archger.2014.08.008

Author details

1. Joshua S Talboom

   1. The Translational Genomics Research Institute, Phoenix, United States
   2. Arizona Alzheimer’s Consortium, Phoenix, United States

   Contribution

   Conceptualization, Data curation, Formal analysis, Supervision, Investigation, Visualization, Methodology, Writing—original draft, Project administration, Writing—review and editing

   Competing interests

   No competing interests declared

   "This ORCID iD identifies the author of this article:" 0000-0002-4327-4103

2. Asta Håberg

   Norwegian University of Science and Technology, Trondheim, Norway

   Contribution

   Conceptualization, Data curation, Formal analysis, Supervision, Visualization, Methodology, Writing—original draft, Project administration, Writing—review and editing

   Competing interests

   No competing interests declared

3. Matthew D De Both

   1. The Translational Genomics Research Institute, Phoenix, United States
   2. Arizona Alzheimer’s Consortium, Phoenix, United States

   Contribution

   Conceptualization, Resources, Data curation, Software, Formal analysis, Supervision, Validation, Investigation, Visualization, Methodology, Writing—original draft, Project administration, Writing—review and editing

   Competing interests

   No competing interests declared

4. Marcus A Naymik

   1. The Translational Genomics Research Institute, Phoenix, United States
   2. Arizona Alzheimer’s Consortium, Phoenix, United States

   Contribution

   Data curation, Software, Formal analysis, Validation, Investigation, Visualization, Methodology, Writing—original draft, Writing—review and editing

   Competing interests

   No competing interests declared

5. Isabelle Schrauwen

   1. The Translational Genomics Research Institute, Phoenix, United States
   2. Arizona Alzheimer’s Consortium, Phoenix, United States

   Contribution

   Conceptualization, Data curation, Formal analysis, Supervision, Investigation, Visualization, Methodology, Writing—original draft, Project administration, Writing—review and editing

   Competing interests

   No competing interests declared

6. Candace R Lewis

   1. The Translational Genomics Research Institute, Phoenix, United States
   2. Arizona Alzheimer’s Consortium, Phoenix, United States

   Contribution

   Conceptualization, Formal analysis, Investigation, Visualization, Methodology, Writing—original draft, Writing—review and editing

   Competing interests

   No competing interests declared

7. Stacy F Bertinelli

   The Translational Genomics Research Institute, Phoenix, United States

   Contribution

   Conceptualization, Data curation, Project administration, Writing—review and editing

   Competing interests

   No competing interests declared

8. Callie Hammersland

   The Translational Genomics Research Institute, Phoenix, United States

   Contribution

   Supervision, Project administration

   Competing interests

   No competing interests declared

9. Mason A Fritz

   The Translational Genomics Research Institute, Phoenix, United States

   Contribution

   Data curation, Formal analysis, Investigation

   Competing interests

   No competing interests declared

10. Amanda J Myers

    University of Miami, Miami, United States

    Contribution

    Conceptualization, Methodology, Writing—review and editing

    Competing interests

    No competing interests declared

11. Meredith Hay

    1. Arizona Alzheimer’s Consortium, Phoenix, United States
    2. University of Arizona, Tucson, United States

    Contribution

    Conceptualization, Methodology, Writing—original draft, Writing—review and editing

    Competing interests

    No competing interests declared

12. Carol A Barnes

    1. Arizona Alzheimer’s Consortium, Phoenix, United States
    2. University of Arizona, Tucson, United States

    Contribution

    Conceptualization, Visualization, Methodology, Writing—review and editing

    Competing interests

    No competing interests declared

13. Elizabeth Glisky

    1. Arizona Alzheimer’s Consortium, Phoenix, United States
    2. University of Arizona, Tucson, United States

    Contribution

    Conceptualization, Methodology, Writing—original draft, Writing—review and editing

    Competing interests

    No competing interests declared

14. Lee Ryan

    1. Arizona Alzheimer’s Consortium, Phoenix, United States
    2. University of Arizona, Tucson, United States

    Contribution

    Conceptualization, Methodology, Writing—original draft, Writing—review and editing

    Competing interests

    No competing interests declared

15. Matthew J Huentelman

    1. The Translational Genomics Research Institute, Phoenix, United States
    2. Arizona Alzheimer’s Consortium, Phoenix, United States

    Contribution

    Conceptualization, Resources, Data curation, Software, Formal analysis, Supervision, Funding acquisition, Validation, Investigation, Visualization, Methodology, Writing—original draft, Project administration, Writing—review and editing

    For correspondence

    mhuentelman@tgen.org

    Competing interests

    No competing interests declared

    "This ORCID iD identifies the author of this article:" 0000-0001-7390-9918

• 7,544

  Page views

• 585

  Downloads

• 20

  Citations

Article citation count generated by polling the highest count across the following sources: Crossref, PubMed Central, Scopus.