Non-Genetic Risk Factors for Degenerative and Vascular Young Onset Dementia: Results from the INSPIRED and KGOW Studies

Abstract

Background:

Several brain reserve, vascular risk, and other modifiable factors have been associated with late-onset dementia, but their association with young onset dementia (YOD) has not been adequately explored.

Objective:

To examine the association of cognitive reserve enhancing factors, cardiovascular risk factors (including smoking), depression, alcohol use, and traumatic brain injury (TBI) with non-autosomal dominant degenerative and/or vascular YOD.

Methods:

Data for this matched case-control study were taken from two larger studies conducted in NSW, Australia. One comprised all people with YOD within a geographical region, while the other exclusively included Aboriginal and Torres Strait Islander participants. Dementia diagnosis was confirmed by clinical consensus, and risk exposure was retrospectively self- and/or informant-reported.

Results:

Participants were 96 people with YOD (58.4% with probable Alzheimer’s disease) and 175 age-group, sex, and sample matched control participants. Poor educational attainment, low participation in cognitive leisure activity, stroke, transient ischemic attack, and self-reported very heavy alcohol use were related to the risk of primary degenerative and/or vascular YOD. The effect of hypertension and depression varied depending on when they occurred relative to dementia onset. Current smoking was significantly associated with risk in univariate analyses but did not retain significance in multivariate modelling. There was no association with hypercholesterolemia, diabetes, or TBI of any kind. Some compensation for low educational attainment was possible via a complex occupation later in life.

Conclusion:

Non-genetic factors have a role in YOD, though the relative importance of each factor may be different to late onset dementia. The timing and severity of exposure, as well as the potential for compensation with later protective exposures, are important considerations for potential prevention strategies.

Keywords

Alzheimer’s disease early onset dementia epidemiology frontotemporal Lewy body lifestyle modifiable protective risk vascular young onset dementia

INTRODUCTION

Young onset dementia (YOD), with onset of symptoms before 65 years of age, accounts for up to 9% of all dementia cases [1, 2]. Despite being associated with greater personal and family burden and higher overall costs than late onset dementias (LOD) [3, 4], the causes of YOD are not well understood. Some misconception exists that all cases are directly inherited via autosomal-dominant genes or genetic polymorphism. Although most autosomal-dominant dementias do emerge prior to age 65, these account for only a minority of YOD cases. The majority are non-autosomal dominant [5].

It is well established that non-genetic factors have an important role in LOD. Poor educational attainment, cardiovascular risk factors, and other factors are noted to account for up to a third of cases of Alzheimer’s disease (AD) worldwide [6 –8]. Several studies highlight the importance of the timing, chronicity, and severity of risk exposure given the lengthy prodromal period of cumulating pathology that precedes symptom onset [9]. For example, hypertension only confers risk for LOD when it occurs many years before the onset of dementia. The lack of effect when occurring close in time to dementia onset appears to be influenced by cumulating neuropathology causing a decline in blood pressure [10]. Additionally, positive health behaviors in young adulthood and midlife may provide some compensation for early life disadvantage [11].

A systematic review identified 14 previous studies of non-genetic risk factors for YOD [12], providing some evidence of an association with cardiovascular risk factors, psychiatric illness, heavy alcohol use, and low estrogen in women. Evidence for other factors was inconsistent. However, none of the previous studies considered the details of exposure, nor did they actively exclude autosomal-dominant cases of YOD or those occurring secondarily to another illness. This obscures efforts to understand the influence of non-genetic factors on non-autosomal dominant primary degenerative (i.e., AD, frontotemporal dementia, Lewy body disease) and/or vascular dementias emerging in midlife. Understanding this influence may help to identify good candidates for dementia delay more generally.

The aim of this study was to assess the risk for non-autosomal dominant primary degenerative and/or vascular YOD associated with selected non-genetic risk and protective factors. Detailed exposure data across the life course were obtained and the impact of its timing and/or severity considered. It was hypothesized that increasing educational attainment, occupational complexity, participation in cognitive leisure activity, and mild to moderate alcohol consumption would significantly reduce the risk for YOD, and lifetime exposure to stroke, hypertension, hypercholesterolemia, diabetes mellitus (DM), smoking, depression, heavy alcohol consumption, and moderate to severe traumatic brain injury (TBI) would significantly increase the risk for YOD. The timing of exposure was expected to influence risk. It was also hypothesized that an interaction would exist wherein the risk for YOD associated with low educational attainment in early life would be attenuated by complex occupation and/or regular participation in cognitive leisure activity later in life.

MATERIALS AND METHODS

A case-control design was selected for this study given the relatively low prevalence of YOD. Data were derived from two cross-sectional, population-based studies conducted in New South Wales (NSW), Australia: The ‘Improving Service Provision in Younger Onset Dementia’ (INSPIRED) study of YOD epidemiology, and the ‘Koori Growing Old Well Study’ (KGOWS) of urban Aboriginal Australian aging. The samples were considered suitable for pooling because urban-dwelling Aboriginal people are demographically similar to non-Aboriginal people of the same region (i.e., English-speaking and literate, with access to similar education, health, and housing systems), albeit with lower socio-economic status [13]. Both studies are described in detail elsewhere [13, 14].

Briefly, people with a pre-existing diagnosis of YOD and/or their supporters were invited to participate in INSPIRED via health professionals (including general practitioners and specialists), service providers, and general advertising. Control participants were recruited from a selection of the same GPs or via general advertising. Participants and/or their primary support person (case participants only) completed a structured interview on a rolling basis from September 2011 to May 2016. For KGOWS, all Aboriginal and Torres Strait Islander people aged 60 and over living in five metropolitan and regional sites were identified via community census in collaboration with local Aboriginal community controlled organizations and research assistants. Consenting participants completed a structured interview that included cognitive testing. Participants who scored below designated cut-offs were reviewed for dementia and those scoring above the cut-offs were retained as potential control participants (see below).

Case participants from both cohorts were included in this study if they: 1) had a consensus diagnosis of degenerative or vascular YOD according to criteria described in Draper et al. [14] and Radford et al. [15] (including AD, vascular dementia (VaD), frontotemporal dementia (FTD), LBD (Lewy body disease), posterior cortical atrophy, progressive supranuclear palsy, a mix of these, or an uncertain differential diagnosis of two of these); 2) reported symptom onset prior to 65 years old; 3) could report adequate information about lifetime exposure to risk factors.

Case participants were excluded if they: 1) were considered at clinical consensus to have a dementia secondarily to another illness (e.g., alcohol-related dementia, dementia in Parkinson’s disease). Cases were included if their diagnosis was foremost considered a primary degenerative and/or vascular dementia, even where secondary causes were also present (e.g., AD and TBI-related dementia); or 2) had previous evidence of a directly-inherited genetic mutation causing the dementia (e.g., genetic testing ordered by the treating clinician) or reported a family history of YOD greater than one generation. This threshold was identified by Jarmolowicz et al. [5] as able to reliably predict autosomal-dominant inheritance of YOD. Genetic testing was not conducted as part of either study protocol.

Control participants were included if they: 1) had not been diagnosed with dementia or any condition at high risk for dementia (e.g., Parkinson’s disease, motor-neurone disease); 2) could be matched by five-year age-group and sex to a case participant; and 3) passed all screening methods for cognitive impairment described below.

Dementia diagnosis

Case participants in both studies were subject to a clinical consensus review to confirm their diagnosis of YOD, dementia type, and age at symptom onset. For INSPIRED, up to three expert clinical academics who were not involved in the participant’s care reviewed medical records, imaging and biomarker assessments (where ordered by the treating clinician) and neuropsychological testing results (where available), and assigned a diagnosis. Control participant medical history and self-reported cognition were reviewed in detail to identify signs of impairment. In addition, all INSPIRED controls were administered the Addenbrooke’s Cognitive Examination – Revised (ACE-R) [16]. The ACE-R is a 28-item test of cognitive function with a maximum score of 100; a cut-off score of 88 has good sensitivity (0.94) and specificity (0.89). Any control participant who scored under this cut-off and/or was perceived by the study team as exhibiting signs of cognitive impairment was excluded from the current study and referred back to their GP for management.

All KGOWS participants completed a cognitive screen that included the Mini-Mental State Examination (MMSE), Rowland Universal Dementia Assessment Scale (RUDAS), and modified version of the Kimberly Indigenous Cognitive Assessment (mKICA). Details of how these scales were validated in KGOWS appear elsewhere [17]. Participants who scored RUDAS≤25, MMSE≤26, and/or mKICA≤35 additionally completed a comprehensive medical assessment with an aged care physician that included social and medical history, neurological exam, cognitive assessment, and assessment of functional status with an informant. Medical imaging and biomarker assessment was not conducted in KGOWS. All information was reviewed by an in-person panel of three or more clinical academics. Two control participants for each case, matched by age and sex, were chosen from among those who scored above the cognitive screening cut-offs and/or were assessed by the study team as not exhibiting signs of dementia.

Risk factor exposure

A similar detailed, retrospective demographic and medical history interview was designed for and implemented in both studies including: years of completed formal education; main lifetime occupation (coded according to the Australian and New Zealand Standard Classification of Occupations); lifetime diagnosis of stroke or transient ischemic attack (TIA), hypertension, hypercholesterolemia, and DM, including age at diagnosis; smoking history including pack-years; lifetime diagnosis of depression including age at onset; and history of TBI with loss of consciousness (LOC), including length of LOC. Lifetime participation in cognitive leisure activity was collected in INSPIRED (imputed for KGOWS) using an amended version of a scale developed by Wilson et al. [18] Participants were asked to report how often they visited a library, read a book or magazine, wrote a letter, or played ‘brain games’ in three time periods (adolescence, early adulthood, middle adulthood) on a five-point scale from 1 (less than once a year) to 5 (daily). Higher scores indicate more participation (maximum 75). Finally, estimated alcohol consumption over the lifetime (including changes over time) was reported using the World Health Organisation Alcohol Use Disorders Identification Test – shortened version (AUDIT-C). All information provided by case participants was verified by an informant (close family member or friend) and via medical record audit where available. The best available data was used where conflicting information was provided.

Statistical analysis

Data were analyzed using SPSS statistical software version 24 [19]. There was 2.1% missing data, which were not missing completely at random according to Little’s test (χ²(1682) = 2075.96; p < 0.001). Data were appropriate for substitution because they were related to the study sub-sample (INSPIRED versus KGOWS) and dementia status (case versus control; both p < 0.05). Accordingly, multiple imputation using a fully conditional model specification was used to replace missing values.

Lifetime diagnosis of depression was used for power analysis because its community prevalence (14%) [20] is roughly centred among all the risk factors. An initial target of 137 participants with YOD was set with α= 0.05 and 80% power. Significance was set at p < 0.05, but variables with p = 0.05–0.09 were retained and examined for themes given the practical barriers to achieving this sample target and the lower community prevalence of other risk factors.

Univariate conditional logistic regression modelling was used to assess the independent effect of each exposure on risk for YOD. Any variable with p < 0.25 was then selected for inclusion in multivariate modelling [21] with family history of dementia (parent or sibling) as a covariate.

Years of education and main lifetime occupation code were entered as continuous variables. Cognitive leisure activity score violated the linearity assumption and so was split into ‘high’ and ‘low’ groups at±52 based on a median split. Comparison groups for stroke or TIA, hypertension, hypercholesterolemia, DM, and depression were based on lifetime exposure (yes versus no). TBI was assessed according to any exposure and also per severity as defined in Corriger et al. [22]: mild (LOC <30 min), and moderate to severe (LOC≥30 min). Current, never, and former smokers were compared, as well as the impact of smoking pack-years. Mild to moderate alcohol use was defined as an AUDIT-C score below 5 (but more than lifetime abstinence) for all of the participant’s time drinking. Heavy alcohol use was considered any period of drinking above AUDIT-C = 8 over the lifespan, per recommendations by Bush [23]. To assess the impact of exposure timing, stroke or TIA, hypertension, hypercholesterolemia, DM, and depression exposure were separated into ‘distal’ (>10 years from dementia onset) and ‘proximal’ exposure (within 10 years of dementia onset). This cut-off was chosen based on conventions in LOD studies. Risk exposure that occurred after dementia symptom onset but before interview was not considered in the analyses. Given that matching occurred by age at interview, it was necessary to assign control participants an exposure cut-off equivalent to the age at onset of their matched case. Exposure that occurred after this cut-off but before interview was not considered in analyses.

Table 1

Participant demographics

	$\dot{x}$ (SD) or n (%)
All cases	All controls	INSPIRED cases	INSPIRED controls	KGOWS cases	KGOWS controls
Age at interview	63.6 (5.5)	63.8 (5.0)	63.5 (5.8)	63.6 (5.2)	64.5 (3.5)	64.9 (4.0)
Age at dementia onset	56.9 (5.0)	N/A	56.5 (5.2)	N/A	59.0 (3.5)	N/A
Male	62 (64.6)	103 (57.5)	56 (69.1)	91 (61.1)	6 (40.0)	12 (40.0)
Urban-dwelling*	61 (63.5)	156 (87.2)	56 (69.1)	146 (98.0)	5 (33.3)	10 (33.3)
Living at interview*
Community	77 (80.2)	179 (100)	62 (76.6)	149 (100)	15 (100)	30 (100)
Residential aged care	17 (17.7)	0 (0)	17 (21.0)	0 (0)	0 (0)	0 (0)
Deceased	2 (2.1)	0 (0)	2 (2.5)	0 (0)	0 (0)	0 (0)
Born in Australia*	63 (65.6)	140 (78.2)	48 (59.3)	110 (73.8)	15 (100)	30 (100)
English as first language*	83 (86.5)	172 (96.1)	68 (84.0)	143 (96.0)	15 (100)	29 (96.7)
Family history of dementia^a	48 (50.0)	78 (43.6)	43 (53.1)	66 (44.3)	5 (33.3)	12 (40.0)
Family history of YOD^a	10 (10.5)	10 (5.6)	8 (9.9)	9 (6.1)	2 (13.4)	1 (3.4)
MMSE Score*	11.8 (10.9)	29.3 (1.1)	9.7 (10.4)	29.4 (0.9)	20.5 (8.7)	28.3 (1.7)

*Significant difference between all cases and all controls to p < 0.05. ^aParent or sibling.

Sensitivity analyses were conducted after the removal of all cases of VaD or mixed dementias with a VaD component to compare effects for primary degenerative dementias (i.e., AD, FTD, and LBD) and VaD. Participants with history of stroke or TIA but without a consensus diagnosis of VaD were retained to examine the effect of these factors on non-vascular degenerative YOD.

Ethics

The INSPIRED study was approved by the Human Research Ethics Committees (Health and Medical) of the University of Wollongong and the former South Eastern Sydney and Illawarra Area Health Service (HE11/007). Ethics approval for KGOWS was granted by the Aboriginal Health and Medical Research Council (615/07), the University of New South Wales Human Research Ethics Committee (08003), and the NSW Population & Health Services Research Ethics Committee (AUREDRef: HREC/09/CIPHS/65; Cancer Institute NSW Ref:2009/10/187).

RESULTS

The final sample included 96 people with non-autosomal dominant, primary degenerative YOD (INSPIRED n = 81, KGOWS n = 15) and 179 sex-, study-, and age-group matched control participants (INSPIRED n = 149, KGOWS n = 30; Supplementary Figure 1). Of 95 participants with consensus-confirmed YOD in INSPIRED, four had dementia secondarily to another condition (Huntington’s disease n = 3, alcohol abuse n = 1), four had prior genetic testing that confirmed an autosomal-dominant inheritance (FTD n = 3, AD n = 1), and four were excluded after no agreement could be reached about the cause of their dementia due to multiple contributing factors. Two additional cases were excluded because they were too impaired to be interviewed and their informants could provide too little information about risk exposure history. Fifty-three participants were interviewed with their informant, and one was interviewed alone. The remaining 27 were entirely informant reported because they were too impaired to provide any information (n = 25) or had died (n = 2) after recruitment but prior to interview. Only one of 150 control participants who completed data collection scored below 88 on the ACE-R and was excluded.

As described in Radford et al. [15], 153 KGOWS cases were reviewed by a consensus panel and 41 cases of dementia were diagnosed. Twenty of these were aged 60–64 at the time of consensus or reported an onset of symptoms prior to 65 years. Fifteen were eligible for inclusion in this study as five were diagnosed with dementia secondarily to another illness (TBI n = 3, alcohol abuse n = 1, and a combination of the two n = 1). Thirty control participants were randomly matched and included.

Demographic details appear in Table 1. Case participants were more likely to reside in regional areas than controls (χ²(1) = 21.0, p < 0.001), to be born outside Australia (χ²(1) = 5.13, p = 0.03), and to have a first language other than English (χ²(1) = 8.60, p = 0.003). There were no differences between case and control groups in parent or sibling history of dementia or YOD (both p > 0.05). KGOWS participants were more likely than INSPIRED participants to be female (χ²(1) = 8.97, p = 0.003) and regional-dwelling (χ²(1) = 67.16, p < 0.001). There were no differences in age at interview (t = –1.74, df = 82.64, p = 0.09) or age at dementia symptom onset (t = –1.88, df = 94, p = 0.07) between studies. One KGOWS participant identified as both Aboriginal and Torres Strait Islander; all others identified as Aboriginal.

Dementia etiology was diverse (Table 2). KGOWS cases were more likely to have a non-Alzheimer dementia than INSPIRED cases (χ²(1) = 7.34, p = 0.007). No agreement could be reached about differential diagnosis for three INSPIRED case participants. These were classified as unspecified degenerative dementias and included two cases with a differential diagnosis of FTD or LBD, and one of FTD or VaD.

Table 2

Case etiology by study

Dementia type	INSPIRED (n = 81)	KGOWS (n = 15)
Alzheimer’s disease^a	52 (64.2)	4 (26.7)
Frontotemporal dementia	10 (12.3)	1 (6.7)
Behavioral variant	7 (8.7)	1 (6.7)
Language variant	3 (3.8)	0
Mixed dementias	6 (7.4)	6 (40.0)
AD/VaD	4 (4.9)	3 (20.0)
AD/LBD	1 (1.2)	0 (0)
AD/ARBD	1 (1.2)	0 (0)
AD/TBI	0 (0)	3 (20.0)
Vascular dementia	5 (6.2)	4 (26.7)
Progressive supranuclear palsy	3 (3.7)	0 (0)
Unspecified degenerative dementia	3 (3.7)	0 (0)
Lewy Body Disease	2 (2.5)	0 (0)

AD, Alzheimer’s disease; ARDB, Alcohol-related brain damage; FTD, frontotemporal dementia; LBD, Lewy body disease; TBI, traumatic brain injury; VaD, vascular dementia. ^aIncludes one case of logopenic progressive aphasia.

Risk factor exposure

In univariate conditional logistic regression analyses (Table 3), increasing years of education and regular participation in cognitive leisure activity were significantly protective against YOD. Both distal and proximal stroke/TIA were significantly associated with risk for YOD. Hypertension conferred risk only when first diagnosed more than 10 years before dementia symptom onset. Conversely, depression was significantly associated with YOD only with onset within 10 years of dementia onset. With each year increase between the onset of depression and dementia, the risk for YOD decreased by 1% (COR = 0.99, 95% CI:0.99–1.00, p = 0.04). Although ‘current’ smokers (at the time of symptom onset) were at higher risk for YOD than never smokers, the potential for recovery was evident in that former smokers were at no higher risk for YOD than never smokers. Smoking pack-years was also positively associated with YOD risk (COR = 1.02, 95% CI:1.01–1.03, p = 0.03). There was no evidence of a protective effect of mild to moderate alcohol consumption. Rates of ‘risky’ drinking were very high in both case and control groups, and no differences between groups were detected at AUDIT-C≥8. Given these high rates, post-hoc analysis was conducted raising the comparison to AUDIT = 12 (at least 10 standard drinks at least 4 times a week). A strong risk for YOD was noted at this level. Main lifetime occupation code and distal hypercholesterolemia were retained for multivariate modelling (both p < 0.25). There was no effect of DM or TBI of any kind.

Table 3

Results of univariate conditional logistic regression modelling

	Cases $\dot{x}$ (SD) or n (%)	Controls $\dot{x}$ (SD) or n (%)	Unadjusted COR (95% CI)	χ ²	p
Years of education	12.7 (4.3)	14.5 (4.3)	0.86 (0.80–0.93)	17.0	<0.001
Main lifetime occupation code	7 (0–8)^a	7 (0–8)^a	0.93 (0.83–1.04)	1.94	0.17
Low lifetime cognitive leisure activity	59 (61.5)	74 (41.4)	3.12 (1.47–6.61)	14.9	<0.001
Stroke or TIA	21 (21.9)	12 (6.8)	7.99 (2.42–26.4)	18.6	<0.001
Distal	11 (11.5)	5 (2.8)	8.78 (1.81–42.7)	9.9	0.008
Proximal	10 (10.4)	7 (3.9)	7.55 (2.06–27.8)	10.0	0.002
Lifetime hypertension	45 (46.9)	71 (39.7)	1.40 (0.84–2.33)	1.90	0.17
Distal	23 (24.0)	23 (12.9)	1.94 (1.02–3.80)	4.62	0.04
Proximal	22 (23.0)	48 (26.9)	0.89 (0.43–1.86)	0.16	0.75
Lifetime hypercholesterolemia	28 (29.2)	45 (25.2)	1.19 (0.69–2.08)	0.41	0.53
Distal	11 (11.5)	8 (4.5)	2.05 (0.80–5.26)	2.13	0.14
Proximal	17 (17.7)	37 (20.7)	0.97 (0.51–1.85)	0.02	0.93
Lifetime diabetes mellitus	8 (8.4)	17 (9.5)	0.94 (0.35–2.56)	0.03	0.87
Distal	3 (3.1)	4 (2.2)	2.10 (0.31–14.6)	0.66	0.46
Proximal	5 (5.2)	13 (7.3)	0.67 (0.19–2.43)	0.26	0.54
Smoking status
Never	33 (34.4)	83 (46.4)	–	–	–
Ever	63 (65.7)	96 (53.7)	1.61 (0.95–2.74)	3.10	0.08
Former	42 (43.8)	75 (41.9)	1.38 (0.79–2.41)	1.23	0.27
Current	21 (21.9)	21 (11.8)	2.69 (1.24–5.86)	6.21	0.02
Lifetime diagnosis of depression	41 (53.2)	54 (30.2)	1.74 (1.03–2.95)	4.24	0.04
Distal	11 (11.5)	31 (17.3)	0.79 (0.35–1.78)	0.36	0.57
Proximal	30 (31.3)	23 (12.8)	2.90 (1.48–5.67)	11.0	0.002
Heavy drinking
Ever AUDIT≥8	31 (32.3)	76 (42.5)	0.68 (0.39–1.19)	2.06	0.16
Ever AUDIT≥10	19 (19.8)	42 (23.6)	0.80 (0.41–1.57)	0.43	0.52
Ever AUDIT = 12	11 (11.5)	8 (4.5)	3.46 (1.18–10.2)	5.51	0.02
TBI with LOC
Never	71 (74.0)	124 (69.3)	–	–	–
Mild	16 (16.7)	43 (24.1)	0.65 (0.31–1.38)	1.75	0.26
Moderate or severe	9 (9.4)	12 (6.8)	0.92 (0.35–2.44)	0.06	0.86

AUDIT-C, Alcohol Use Disorders Identification Test; CI, confidence interval; COR, conditional odds ratio; LOC, loss of consciousness; LR, likelihood ratio; SD, standard deviation; TBI, traumatic brain injury; TIA, transient ischemic attack. ^aMedian (Range).

Years of education, main occupation code, cognitive leisure activity, lifetime stroke or TIA, distal hypertension and hypercholesterolemia, current smoking, proximal depression, and very heavy alcohol use (ever AUDIT-C = 12) were entered into a multivariate conditional logistic regression model with family history of dementia as a covariate. Occupation code, distal hypercholesterolemia, and current smoking were not significantly related to YOD in the base model (Supplementary Table 1) and were not confounds for other variables when removed. Results of the final model are displayed in Table 4. Years of education, lifetime participation in cognitive leisure activity, lifetime stroke or TIA, distal hypertension, and proximal depression were related to risk for YOD. The effect of very heavy alcohol use approached significance (p = 0.05).

Table 4

Results of final multivariate conditional logistic regression model with family history of dementia as a covariate

	Cases $\dot{x}$ (SD) or n (%)	Controls $\dot{x}$ (SD) or n (%)	Adjusted COR (95% CI)	Difference in LnL after exclusion (LRχ²)	p-value of LR Test
Years of education	12.7 (4.3)	14.5 (4.3)	0.87 (0.80–0.95)	11.2	<0.001
Low lifetime cognitive leisure activity	59 (61.5)	74 (41.4)	3.48 (1.46–8.33)	11.8	<0.001
Ever stroke or TIA	21 (21.9)	12 (6.8)	9.92 (2.24–44.0)	13.5	<0.001
Distal hypertension	23 (24.0)	23 (12.9)	2.55 (1.33–4.73)	4.6	0.04
Proximal depression	30 (31.3)	23 (12.8)	3.71 (1.63–8.46)	11.1	<0.001
Ever AUDIT-C = 12	11 (11.5)	8 (4.5)	3.53 (0.94–13.3)	4.0	0.05

Base -2LnL = 132.4, all DF = 1. AUDIT-C, Alcohol Use Disorders Identification Test; CI, confidence interval; COR, conditional odds ratio; LnL, log of the likelihood; LR, likelihood ratio; SD, standard deviation; TIA, transient ischemic attack.

To assess the possibility of compensation for low educational attainment with later life occupational complexity or participation in cognitive leisure activity, interaction terms were added to the final model. There was a significant interaction between education and occupation (χ² = 7.06, p = 0.008), but not between education and cognitive leisure activity (χ² = 0.61, p = 0.44). To explore the interaction further, participants were allocated to one of four groups according to their combination of educational attainment (less than 10 years versus greater than 10 years) and occupational complexity (high versus low). Post-hoc analyses (Table 5) revealed that those who completed less than 10 years of formal education and also reported a low complexity main lifetime occupation were at a more than 5-fold increased risk for YOD when compared with those with both high education and occupational complexity. Some compensation for low education was evident in that the risk for YOD was slightly attenuated in those who progressed from low educational attainment to high occupational complexity.

Table 5

Risk for YOD associated with combination of educational attainment and occupational complexity, with family history of dementia as a covariate

	Cases n (%)	Controls n (%)	Adjusted COR (95% CI)	χ ²	p
High education, high complexity occupation*	53 (55.2)	128 (71.5)	–	–	–
High education, low complexity occupation	11 (11.5)	16 (8.9)	1.70 (0.75–3.88)	1.58	0.21
Low education, high complexity occupation	18 (18.8)	23 (12.8)	3.24 (1.38–7.63)	7.57	0.005
Low education, low complexity occupation	14 (14.6)	12 (6.7)	5.85 (1.93–17.8)	9.73	0.002

CI, confidence interval; COR, conditional odds ratio; *Reference category.

Sensitivity analysis

The final model was re-analyzed after removing all cases of VaD or mixed dementia with a VaD component (n = 17) and their matched controls (n = 32; Table 6). The effect of midlife hypertension became non-significant. The difference between case and control groups regarding very heavy alcohol use did not notably change, but the drop in statistical power also caused this effect to become non-significant. The final model was also re-analyzed with INSPIRED (i.e., non-Aboriginal) participants only (cases n = 81, controls n = 149). The descriptive differences between case and controls groups did not noticeably change for any factor, but the comparisons were underpowered to detect significant effects for distal hypertension, heavy alcohol use, cognitive leisure activity, and education (data not shown).

Table 6

Results of sensitivity analysis of base model, with cases of VaD or mixed dementia with a VaD component removed

	Cases $\dot{x}$ (SD) or n (%)	Controls $\dot{x}$ (SD) or n (%)	Adjusted COR (95% CI)	Difference in LnL after exclusion (LRχ²)	p-value of LR Test
Years of education	12.9 (4.3)	15.0 (4.2)	0.87 (0.80–0.96)	10.1	0.001
Low lifetime cognitive leisure activity	62 (78.5)	82 (55.8)	3.05 (1.34–7.00)	8.1	0.005
Lifetime stroke or TIA	9 (11.4)	5 (3.4)	7.13 (1.22–41.7)	5.4	0.03
Distal hypertension	10 (12.7)	15 (10.2)	1.92 (0.62–5.92)	1.3	0.26
Proximal depression	23 (29.2)	18 (11.2)	3.97 (1.69–9.34)	11.1	<0.001
Ever AUDIT-C = 12	8 (10.1)	5 (3.4)	3.21 (0.85–3.44)	2.2	0.14

AUDIT-C, Alcohol Use Disorders Identification Test; CI = Confidence interval; COR, conditional odds ratio; LnL, log of the likelihood; LR, likelihood ratio; SD, standard deviation; TIA, transient ischemic attack. Base -2LnL = 116.1, all DF = 1.

DISCUSSION

The findings of this study suggest that non-genetic factors have a role in YOD. This is the first study to consider the impact of non-genetic factors in exclusively non-autosomal dominant primary degenerative and vascular YODs, while eliminating the confounding effect of directly inherited dementias and those occurring secondarily to another disease. It emphasizes the importance of the timing and severity of exposure in influencing risk. It is also the first study of YOD risk profiles to represent urban Aboriginal Australians, who report YOD at much higher rates than the non-Aboriginal population [15]. The main findings are summarized below.

Main findings

There was a strong effect of formal education in this study, and each additional year of education reduced the risk for YOD by 13% even after controlling for all other factors. Poor educational attainment was particularly detrimental when in combination with a low complexity main lifetime occupation, but some compensation was evident with a high complexity occupation. This is a novel finding that has not been previously explored in YOD groups. It is consistent with patterns demonstrated in studies of older people, in which participants who maintain a ‘low’ socio-economic position over the lifespan were at a higher risk for LOD than those who moved from low education to higher SES by midlife [24, 25]. The results emphasize the power of early life intellectual enrichment, but also support the possibility of recovery from disadvantage and the benefit of tapping into previously underutilized intellectual potential. Existing public health programs that aim to intervene and alleviate early life social disadvantage may have the added benefit of reducing the risk of midlife neurodegenerative disease.

Participation in a range of cognitive leisure activities at least once a month from adolescence to middle age was associated with a three-fold reduction in YOD risk. This study is consistent with results from LOD studies [9, 26]. Cognitive leisure activities like these can be accessed readily, cheaply, and can be delivered in a variety of forms to suit preferences and abilities. They are noted to provide benefits even when first implemented late in life or closer in time to the onset of dementia [27]. However, leisure activity was not powerful enough to compensate for low formal educational attainment.

Stroke was a strong risk factor for YOD, though the effect was unsurprisingly attenuated after removal of cases of VaD. The effect of hypertension was time dependent. While associated with significant risk when occurring more than 10 years prior to dementia onset, it had no effect within 10 years of dementia onset. The findings are consistent with the generally recognized time-dependent effect of hypertension in LOD [9 , 29] owing to a drop in blood pressure caused by prodromal degenerative disease [10]. However, that the effect of midlife hypertension may be limited to young onset VaD contrasts with evidence of its relationship to late onset AD [9]. There was also no effect of hypercholesterolemia or DM in this study. That these factors do not confer risk for YOD is plausible given that direct comparisons of YOD and LOD groups suggest that cardiovascular factors are less common in younger groups [30, 31] and are uncommon in those with early onset AD overall [32]. Evidence also suggests that, among those with neurodegenerative disease, a concurrent cerebrovascular disease is more common with increasing age [33]. It is possible that cardiovascular risk factors require a longer period of exposure to confer risk. However, it is also possible that the current study was underpowered to detect subtle effects. Two previous large retrospective cohort studies have demonstrated an increased risk for YOD with lifetime use of antidiabetic medication [34] and a concurrent diagnosis of DM in medical records [35].

Although current smokers were at higher risk for YOD than never smokers in univariate analyses, the effect became non-significant after controlling for other factors. It is possible that smoking confers risk for YOD via its relationship with cardiovascular risk factors like hypertension and stroke.

The risk for YOD associated with depression became stronger the closer in time the two conditions occurred to each other. This is congruent with several studies that have demonstrated this effect in LOD [36 –38], providing support for the suggestion that ‘proximal’ depression is a prodromal consequence of, rather than etiological risk factor for, YOD. To our knowledge, this is the first study to consider the timing of depression onset in moderating the risk for YOD.

Rates of alcohol consumption were very high in this sample. More than 30% of cases and 40% of controls reported consumption at the ‘extreme’ risk level designated by the World Health Organization as a score of 8 or higher on the AUDIT-C at some time during their lives. This is not surprising given the noted cohort effect that sees the ‘baby boomer’ generation (currently aged 53 to 71 years) drink more than both their parents and their children [39]. Significant risk for YOD appeared only when groups were categorized at the maximum possible score: regular consumption of at least ten standard drinks at least four times a week at some time over the lifespan. One previous retrospective cohort study reported a nearly four-fold increase in all-cause YOD risk associated with hospitalization for alcohol intoxication in the >40 years prior to dementia onset, but did not exclude cases of alcohol-related brain damage from their analyses [34]. The current results suggest that the risk of very heavy consumption extends to primary degenerative and vascular dementias. These analyses should be repeated in larger samples with the power to detect differences at lower consumption thresholds. There was no protective effect of mild to moderate alcohol consumption identified in this study.

Finally, there was no effect of TBI of any kind. This contrasts with one large retrospective cohort study that reported a 2-fold increase in risk for all-cause YOD associated with severe TBI noted in medical records [40]. The different results are likely related to the very low absolute risk for YOD associated with TBI overall. Authors of the prior study note that only 25 participants with severe TBI (of 5982, 0.5%) went on to develop YOD during the >30 years follow up. The large sample size of that study afforded the statistical power to identify such a subtle effect, but this may overstate the absolute importance of TBI for YOD risk. Such a low absolute risk was also evident here, with only 21 moderate to severe TBIs recorded overall.

Limitations

The results of this study should be interpreted in the context of important limitations. The sample size was not large enough to reliably determine whether there were significant differences between Aboriginal and non-Aboriginal participants which might occur given the possible social and economic differences between them. The KGOWS sample only included those aged 60 to 64 years, and this is problematic because Aboriginal Australians are known to experience an approximate 10-year health differential compared to non-Aboriginal Australians [41]. Although this differential is more pronounced in rural and remote-dwelling Aboriginal people, KGOWS cases included here may still resemble LOD more than YOD. Future studies focused on risk profiles for YOD in urban Aboriginal people (and including people under age 60) will be valuable to address this problem.

While the case-control design allowed for targeted recruitment of a difficult-to-access population, the reliance on self-reported retrospective data is vulnerable to recall bias. There were also demographic differences between case and control groups, especially in INSPIRED, that may have acted as confounds and were not controlled for in modelling. Control participants were more often born in Australia, spoke English as their first language, and lived in a metropolitan area than the general population. The use of health services to recruit most participants in both studies may have introduced a sampling bias in favor of those in worse health or with better access to services. This may also have influenced the types of dementia represented and rates of risk exposure. In addition, self-selection of control participants for INSPIRED via GP invitation or general advertising was vulnerable to a sampling bias where volunteers are highly conscious of (or worried for) their health but, because of these concerns, practice healthy behaviors and are actually in very good health [42]. This may have contributed to the differences in risk exposure between case and control groups.

Diagnosis by consensus experts was limited by the testing and imaging ordered by the treating clinician(s) and where it was available for review. Additionally, neither study implemented genetic testing as part of the study protocol. Although familial history of dementia was studied for signs of inheritance, the exclusion rate of 6% is lower than would be expected in a population of degenerative YOD (approximately 15%) [5]. This also meant that no information was available regarding APOE status, which is demonstrated to reduce the age of dementia onset [43] and has a mediating role in the relationship between non-genetic risk factors and LOD [44]. Some factors known to confer strong risk for LOD (e.g., lifetime obesity, physical inactivity) [7] were omitted from this study primarily due to the low quality of data when reported retrospectively. These should be assessed in future studies. Finally, recruiting people with YOD for research is difficult [45] and the sample of this study was relatively small. Some comparisons may have been underpowered to detect small differences. Statistical corrections were not possible and some Type I error may be present in the results. This also did not allow for stratified analyses by dementia type (i.e., risk for AD alone), and the risk associated with certain factors is likely to vary for each type.

Future directions

Non-genetic factors appear to operate in YOD, challenging the misconception that genetic factors are wholly accountable for YOD. However, the relative importance of certain factors and mechanisms by which they confer risk (especially cardiovascular risk factors) may differ from LOD. What causes the very early emergence of symptoms remains unknown and should be explored in future research. A larger sample size will facilitate sub-group analyses by dementia type and exploration of causal pathways for distinct etiologies. Results of this study can be used to inform power calculations. Important factors that were not included here should be prioritized in future work, and the role of socio-economic status across the lifespan should be explored and/or controlled for appropriately. Genetic screening must be considered as this will help to identify and exclude directly-inherited cases, and consideration of APOE status as a key confounding variable is crucial. Finally, an exploration of the interrelationships between risk factors is warranted given that individual factors do not occur in a vacuum. Detecting clusters of exposures may facilitate identification of those most at risk.

Footnotes

ACKNOWLEDGMENTS

The Australian National Health and Medical Research Council provided funding for both INSPIRED (1008267) and KGOWS (510347). Funding was provided for this combined sub-study by the Dementia Centre for Research Collaboration, Alzheimer’s Australia Dementia Research Foundation, and Brain Sciences UNSW. The studies were conducted at the Dementia Centre for Research Collaboration and at Neuroscience Research Australia, both at UNSW Sydney, Australia. A/Prof Low is supported by an NHMRC Career Development Fellowship and Dr Radford is supported by an NHMRC Dementia Research Development Fellowship.

The authors would like to thank all the participants and their families who participated in both INSPIRED and KGOWS. We acknowledge Apo Demirkol, Clement Loy, and Susan Quine for their contribution to the planning and clinical consensus for the INSPIRED study. We thank Nicole Denham, Fiona White, Christine Metusela, and Linda Nattrass for their assistance with study coordination and data collection. Similarly, we thank the KGOWS Aboriginal Reference Group; the Aboriginal Health and Medical Research Council; and the KGOWS Aboriginal community partners across NSW who continue to contribute to the research. We recognize all of the KGOWS Aboriginal research assistants, investigators, project officers and interviewers, medical doctors, knowledge translation and support team. In particular, we acknowledge Gail Daylight for her role in community engagement and Holly Mack for study management, as well as Simon Chalkley, Hayley Bennett, and Cecilia Minogue for their contribution to clinical consensus for KGOWS.

Authors’ disclosures available online ().

The supplementary material is available in the electronic version of this article: .

References

Draper

, Withall

(2016) Young onset dementia. Intern Med J 46, 779–786.

Withall

, Draper

, Seeher

, Brodaty

(2014) The prevalence and causes of younger onset dementia in Eastern Sydney, Australia. Int Psychogeriatr 26, 1955–1965.

Freyne

, Kidd

, Coen

, Lawlor

(1999) Burden in carers of dementia patients: Higher levels in carers of younger sufferers. Int J Geriatr Psychiatry 14, 784–788.

Brown

, Hansnata

, La

(2017) Economic cost of dementia in Australia 2016-2056, Alzheimer’s Australia, Canberra, Australia.

Jarmolowicz

, Chen

H-Y

, Panegyres

(2015) The patterns of inheritance in early-onset dementia, Alzheimer’s disease and frontotemporal dementia. Am J Alzheimers Dis Other Demen 30, 299–306.

Norton

, Matthews

, Barnes

, Yaffe

, Brayne

(2014) Potential for primary prevention of Alzheimer’s disease: An analysis of population-based data. Lancet Neurol 13, 788–794.

Livingston

, Sommerlad

, Orgeta

, Costafreda

, Huntley

, Ames

, Ballard

, Banerjee

, Burns

, Cohen-Mansfield

, Cooper

, Fox

, Gitlin

, Howard

, Kales

, Larson

, Ritchie

, Rockwood

, Sampson

, Samus

, Schneider

, Selbæk

, Teri

, Mukadam

(2017) Dementia prevention, intervention, and care. Lancet 390, 2673–2734.

Ashby-Mitchell

, Burns

, Shaw

, Anstey

(2017) Proportion of dementia in Australia explained by common modifiable risk factors. Alzheimers Res Ther 9, 1–8.

Prince

, Albanese

, Guerchet

, Prina

(2014) World Alzheimer Report 2014 - Dementia and risk reduction: An analysis of protective and modifiable factors, Alzheimer’s Disease International, London, UK.

10.

Power

, Weuve

, Gagne

, McQueen

, Viswanathan

, Blacker

(2011) The association between blood pressure and incident Alzheimer disease: A systematic review and meta-analysis. Epidemiology 22, 646–659.

11.

Wang

H-X

, MacDonald

SWS

, Dekhtyar

, Fratiglioni

(2017) Association of lifelong exposure to cognitive reserve-enhancing factors with dementia risk: A community-based cohort study. PLoS Med 14, e1002251.

12.

Cations

, Withall

, Low

L-F

, Draper

(2015) What is the role of modifiable environmental and lifestyle risk factors in young onset dementia? Eur J Epidemiol 31, 107–124.

13.

Radford

, Mack

, Robertson

, Draper

, Chalkley

, Daylight

, Cumming

, Bennett

, Jackson Pulver

, Broe

(2014) The Koori Growing Old Well Study: Investigating aging and dementia in urban Aboriginal Australians. Int Psychogeriatr 26, 1–11.

14.

Draper

, Cations

, White

, Trollor

, Loy

, Brodaty

, Sachdev

, Gonski

, Demirkol

, Cumming

, Withall

(2016) Time to diagnosis in young-onset dementia and its determinants: The INSPIRED study. Int J Geriatr Psychiatry 31, 1217–1224.

15.

Radford

, Mack

, Draper

, Chalkley

, Daylight

, Cumming

, Bennett

, Delbaere

, Broe

(2015) Prevalence of dementia in urban and regional Aboriginal Australians. Alzheimers Dement 11, 271–279.

16.

Mioshi

, Dawson

, Mitchell

, Arnold

, Hodges

(2006) The Addenbrooke’s Cognitive Examination Revised (ACE-R): A brief cognitive test battery for dementia screening. Int J Geriatr Psychiatry 21, 1078–1085.

17.

Radford

, Mack

, Draper

, Chalkley

, Delbaere

, Daylight

, Cumming

, Bennett

, Broe

(2015) Comparison of three cognitive screening tools in older urban and regional Aboriginal Australians. Dement Geriatr Cogn Disord 40, 22–32.

18.

Wilson

, Barnes

, Bennett

(2003) Assessment of lifetime participation in cognitively stimulating activities. J Clin Exp Neuropsychol 25, 634–642.

19.

IBM Corp (2015) IBM SPSS Statistics, IBM Corp, Armonk, NY.

20.

Slade

, Teesson

, Burgess

, others (2009) The mental health of Australians 2: Report on the 2007 National Survey of Mental Health and Wellbeing, Department of Health and Ageing, Canberra, Australia.

21.

Hosmer

Jr , Lemeshow

(2004) Applied logistic regression, John Wiley & Sons, New York, NY.

22.

Corrigan

, Bogner

(2007) Initial reliability and validity of the Ohio State University TBI identification method. J Head Trauma Rehabil 22, 318–329.

23.

Bush

, Kivlahan

, McDonell

, Fihn

, Bradley

(1998) The AUDIT Alcohol Consumption Questionnaire (AUDIT-C): An effective brief screening test for problem drinking. Arch Intern Med 158, 1789–1795.

24.

Scazufca

, Menezes

, Araya

, Di Rienzo

, Almeida

, Gunnell

, Lawlor

(2008) Risk factors across the life course and dementia in a Brazilian population: Results from the Sao Paulo Ageing & Health Study (SPAH). Int J Epidemiol 37, 879–890.

25.

Al Hazzouri

, Haan

, Kalbfleisch

, Galea

, Lisabeth

, Aiello

(2011) Life-course socioeconomic position and incidence of dementia and cognitive impairment without dementia in older Mexican Americans: Results from the Sacramento area Latino study on aging. Am J Epidemiol 173, 1148–1158.

26.

Stern

, Munn

(2010) Cognitive leisure activities and their role in preventing dementia: A systematic review. Int J Evid Based Healthc 8, 2–17.

27.

Valenzuela

, Sachdev

(2007) Assessment of complex mental activity across the lifespan: Development of the Lifetime of Experiences Questionnaire (LEQ). Psychol Med 37, 1015–1026.

28.

Sharp

, Aarsland

, Day

, Sønnesyn

, Ballard

(2011) Hypertension is a potential risk factor for vascular dementia: Systematic review. Int J Geriatr Psychiatry 26, 661–669.

29.

Barnes

, Yaffe

(2011) The projected effect of risk factor reduction on Alzheimer’s disease prevalence. Lancet Neurol 10, 819–828.

30.

Garre-Olmo

, Batlle

, del Mar Fernandez

, Daniel

, de Eugenio Huélamo

, Casadevall

, Recio

, Estrada

, López-Pousa

(2010) Incidence and subtypes of early-onset dementia in a geographically defined general population. Neurology 75, 1249–1255.

31.

Panegyres

, Chen

(2014) Early-onset Alzheimer’s disease: A global cross-sectional analysis. Eur J Neurol 21, 1149–1154.

32.

Chen

, Sillaire

, Dallongeville

, Skrobala

, Wallon

, Dubois

, Hannequin

, Pasquierthe

, Bombois

, Boutantin

, Cassagnaud

, Chen

, Delbeuck

, Delmaire

, Deramecourt

, Gele

, Houssein-Foucher

, Jacquemont

, Lebert

, Lebouvier

, Lopez

, Mackowiak

M-A

, Maureille

, Pasquier

, Petyt

, Pollet

, Rollin-Sillaire

, Schraen

, Semah

, Vanhoutte

(2017) Low prevalence and clinical effect of vascular risk factors in early-onset Alzheimer’s disease. J Alzheimers Dis 60, 1045–1054.

33.

Toledo

, Arnold

, Raible

, Brettschneider

, Xie

, Grossman

, Monsell

, Kukull

, Trojanowski

(2013) Contribution of cerebrovascular disease in autopsy confirmed neurodegenerative disease cases in the National Alzheimer’s Coordinating Centre. Brain 136, 2697–2706.

34.

Nordström

, Nordström

, Eriksson

, Wahlund

L-O

, Gustafson

(2013) Risk factors in late adolescence for young-onset dementia in men: A nationwide cohort study. JAMA Intern Med 173, 1612–1618.

35.

Kadohara

, Sato

, Kawakami

(2017) Diabetes mellitus and risk of early-onset Alzheimer’s disease: A population-based case–control study. Eur J Neurol 24, 944–949.

36.

Geerlings

, Den Heijer

, Koudstaal

, Hofman

, Breteler

MMB

(2008) History of depression, depressive symptoms, and medial temporal lobe atrophy and the risk of Alzheimer disease. Neurology 70, 1258–1264.

37.

, Wang

, Shofer

, Thompson

, Peskind

, McCormick

, Bowen

, Crane

, Larson

(2011) Temporal relationship between depression and dementia: Findings from a large community-based 15-year follow-up study. Arch Gen Psychiatry 68, 970–977.

38.

Heser

, Tebarth

, Wiese

, Eisele

, Bickel

, Köhler

, Mösch

, Weyerer

, Werle

, König

H-H

, Leicht

, Pentzek

, Fuchs

, Riedel-Heller

, Luppa

, Prokein

, Scherer

, Maier

, Wagner

(2013) Age of major depression onset, depressive symptoms, and risk for subsequent dementia: Results of the German Study on Ageing, Cognition, and Dementia in Primary Care Patients (AgeCoDe). Psychol Med 43, 1597–1610.

39.

Australian Institute of Health and Welfare (2014) 2013 National Drug Strategy Household Survey: Detailed findings, AIHW cat. no. PHE 183, Canberra, Australia.

40.

Nordström

, Michaëlsson

, Gustafson

, Nordström

(2014) Traumatic brain injury and young onset dementia: A nationwide cohort study. Ann Neurol 75, 374–381.

41.

Vos

, Barker

, Begg

, Stanley

, Lopez

(2009) Burden of disease and injury in Aboriginal and Torres Strait Islander Peoples: The Indigenous health gap. Int J Epidemiol 38, 470–477.

42.

Mann

(2003) Observational research methods - Research design II: Cohort, cross sectional, and case-control studies. Emerg Med J 20, 54–60.

43.

Blacker

, Haines

, Rodes

, Terwedow

, Go

RCP

, Harrell

, Perry

, Bassett

, Chase

, Meyers

, Albert

, Tanzi

(1997) APOE-ɛ4 and age at onset of Alzheimer’s disease: The NIMH genetics initiative. Neurology 48, 139–147.

44.

Low

L-F

, Yap

, Brodaty

(2010) Will testing for apolipoprotein E assist in tailoring dementia risk reduction? A review. Neurosci Biobehav Rev 34, 408–437.

45.

Withall

(2013) The highs and lows of determining the prevalence of younger onset dementia. Australas Epidemiol 20, 9.