Literature Review Resources

Contents lists available at ScienceDirect

Journal of Affective Disorders

journal homepage: www.elsevier.com/locate/jad

Review article

Risk of Dementia in persons who have previously experienced clinically- significant Depression, Anxiety, or PTSD: A Systematic Review and Meta- Analysis

J.K. Kuring, J.L. Mathias⁎, L. Ward School of Psychology, Faculty of Health & Medical Sciences, University of Adelaide, Adelaide, Australia

A R T I C L E I N F O

Keywords: Risk depression anxiety PTSD dementia Alzheimer's

A B S T R A C T

Background: Depression, anxiety and PTSD appear to be linked to dementia, but it is unclear whether they are risk factors (causal or prodromal) for, comorbid with, or sequelae to (secondary effect of) dementia. Existing meta-analyses have examined depression or anxiety in all-cause dementia, Alzheimer's disease (AD) and vascular dementia (VaD), but have not considered post-traumatic stress disorder (PTSD), dementia with Lewy bodies (DLB), or frontotemporal dementia (FTD). The current meta-analysis examined the risk of developing dementia (AD, VaD, DLB, FTD, all-cause) in people with and without a history of clinically-significant depression, anxiety or PTSD in order to better understand the link between mental illness and dementia (PROSPERO number: CRD42018099872). Methods: PubMed, EMBASE, PsycINFO and CINAHL searches identified 36 eligible studies. Results: There is a higher risk of developing all-cause dementia and AD in people with previous depression, and a higher risk of all-cause dementia in people with prior anxiety, than in persons without this history. Prior PTSD was not associated with a higher risk of later being diagnosed with dementia. Limitations: The data for anxiety, PTSD, DLB and FTD were limited. Conclusions: Depression and anxiety appear to be risk factors for dementia, but longitudinal studies across adulthood (young adult/mid-life/older adult) are needed to evaluate the likely causal or prodromal nature of this risk. The link between PTSD and dementia remains unclear. Regular screening for new onset mental illness and for cognitive changes in older adults with a history of mental illness may assist with earlier identification of dementia.

1. Introduction

As the population ages, the number of people with dementia and the associated personal and health care costs are predicted to rise, with over 130 million people worldwide estimated to be affected by 2050 (Alzheimer's Disease International, 2015). Although the search for a cure continues, research has increasingly attempted to identify poten- tially modifiable risk factors in order to prevent or slow the onset of dementia using early interventions (Deckers et al., 2015; Livingstone et al., 2017). The recent Lancet Commission into dementia prevention, intervention and care calculated that more than a third of dementia cases could be prevented by eliminating reversible risk factors (Livingstone et al., 2017). Even if not eliminated entirely, it is estimated that approximately three million cases of Alzheimer's disease (AD) could be prevented by 2050 if modifiable risk factors were reduced by

as little as 10% per decade (Barnes & Yaffe, 2011; Norton et al., 2014). There is now a growing body of research to suggest that these

modifiable risk factors may include depression, anxiety and post-trau- matic stress disorder (PTSD). This research has been examined in a number of recent meta-analyses, which have reported a higher risk of dementia in people with a history of depression (Cherbuin et al., 2015; Diniz et al., 2013; Ford et al., 2018; Xu et al., 2015) and anxiety (Becker et al., 2018; Ford et al., 2018; Gulpers et al., 2016; Santabarbara et al., 2019). A recent systematic review has additionally reported that there is a greater risk of dementia in military veterans who have a history of PTSD (Rafferty et al., 2017). Most recently, a meta-analysis also reported that there is an association between the four most common types of dementia (AD, vascular dementia [VaD], dementia with Lewy bodies [DLB], frontotemporal dementia [FTD]) and depression, anxiety and PTSD (Kuring et al., 2018). However, it

https://doi.org/10.1016/j.jad.2020.05.020 Received 27 August 2019; Received in revised form 16 April 2020; Accepted 10 May 2020

⁎ Corresponding author. E-mail address: [email protected] (J.L. Mathias).

Journal of Affective Disorders 274 (2020) 247–261

Available online 21 May 2020 0165-0327/ © 2020 Elsevier B.V. All rights reserved.

T

remains unclear whether depression, anxiety and PTSD are causal risk factors, prodromal symptoms of, or comorbid with dementia (Kuring et al., 2018).

There are multiple reasons why dementia may be linked to de- pression, anxiety, and PTSD. One possibility is that mental illness plays a causal role in the development of dementia, with the inflammatory processes that are thought to be associated with depression, anxiety, and PTSD hypothesized to underpin the subsequent neurodegeneration that leads to dementia (Leonard, 2017). Specifically, hypothalamic-pi- tuitary-adrenal (HPA) axis hyperactivity, dysfunction in the trypto- phan-kynurenine pathway, and dysfunctional brain glucose metabolism have all been suggested to lead to neurodegeneration as a consequence of the stress that is associated with mental illness (Rafferty et al., 2017; Leonard, 2017). Alternatively, depression and anxiety may be pro- dromal symptoms of dementia, such that the underlying disease pro- cesses that cause dementia also lead to depression and anxiety (en- dogenous cause), with these mental illnesses potentially manifesting earlier (or identified earlier) than the cognitive and functional symp- toms required for a diagnosis of dementia (Stella et al., 2014). Another possibility is that depression and anxiety are comorbid with, or the sequelae to (secondary effect of), dementia; with dementia indirectly resulting in the secondary symptoms of depression and anxiety (exo- genous/reactive cause) due to the cognitive and other forms of decline that are experienced (Kales et al., 2015; Lawlor, 1996). Although en- dogenous and exogenous causes do not readily account for the link between PTSD and dementia, it has been suggested that dementia may act as a trigger for a relapse of PTSD in people with a history of this mental illness (Hiskey et al., 2008).

If depression, anxiety and PTSD are risk factors (causal or pro- dromal) for dementia, it would be expected that prior mental illnesses would be associated with an increased likelihood of dementia at a later time, when compared to persons with no history of mental illness. Alternatively, if these mental illnesses are either comorbidities of, or sequelae to, dementia then persons with and without a prior history of depression, anxiety or PTSD should not differ in their risk of later de- veloping dementia.

Several meta-analyses have examined the temporal relationship between depression and/or anxiety that has been experienced prior to a diagnosis of dementia; all of which have reported a positive association between depression or anxiety and subsequent dementia (Becker et al., 2018; Cherbuin et al., 2015; Diniz et al., 2013; Ford et al., 2018; Gulpers et al., 2016; Jorm et al., 1991; Jorm, 2000; Jorm, 2001; Ownby et al., 2006; Santabarbara et al., 2019; Xu et al., 2015). How- ever, there are several important limitations to these meta-analyses. First, none of them confined their examination to clinically-significant symptoms of depression and anxiety, instead combining persons with clinically significant mental illness with others who were experiencing milder/subclinical symptoms. This prevents an evaluation of whether it is only clinically-significant mental illness that is a risk factor for de- mentia, as suggested by the inflammatory neurodegeneration hypoth- esis (Leonard, 2017), rather than any level of depression or anxiety symptoms (including subclinical levels). Second, some examined both state and trait anxiety (Becker et al., 2018; Cherbuin et al., 2015; Santabarbara et al., 2019), potentially confounding mental illness with more stable personality traits. Third, others failed to consistently in- clude healthy controls (Cherbuin et al., 2015; Jorm, 2001; Jorm, 2000; Jorm et al., 1991) or separately report the findings for those with de- mentia and those with milder cognitive decline (Diniz et al., 2013). Some are also now quite dated (Jorm, 2001; Jorm, 2000; Jorm et al., 1991) and consequently fail to include recent research. Finally, none of them have examined the temporal relationship between PTSD and de- mentia, or DLB and FTD when examining depression or anxiety as possible risk factors for dementia.

The current meta-analysis therefore sought to build on these ex- isting studies in order to further examine whether depression, anxiety and PTSD are likely to be risk factors for (causal or prodromal),

comorbidities of, or sequelae to, dementia. Unlike existing studies, the current meta-analysis: (1) only examined clinically-significant levels of depression/anxiety/PTSD; (2) excluded studies that assessed trait (ra- ther than state) anxiety; (3) ensured that the groups were cognitively ‘normal’ at baseline; (4) ensured that dementia was identified on the basis of published diagnostic/research criteria and that persons with milder cognitive decline were not included; and (5) investigated de- pression, anxiety and PTSD, along with the four most common types of dementia (AD, VaD, DLB, FTD) and all-cause dementia. The latter ca- tegory was included because many studies examine dementia, more generally, rather than specific subtypes; the assumption being that ‘all- cause dementia’ primarily comprises AD, VaD, DLB and FTD.

2. Method

This research adhered to two sets of guidelines: the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) (Moher et al., 2009) and the Meta-analysis Of Observational Studies in Epidemiology (MOOSE) (Stroup et al., 2000). The PROSPERO regis- tration number for this systematic review and meta-analysis is: CRD42018099872.

2.1. Search strategy

PubMed, EMBASE, PsycINFO, and CINAHL electronic databases were all searched, under the guidance of an expert research librarian, using multiple terms for dementia, mental illness (depression, anxiety, and PTSD), and risk (see Online Resource 1 for detailed search strate- gies). The search was confined to publications dated from 1980, which is when ‘major depressive disorder’ first appeared in the DSM-III (American Psychiatric Association, 1980). The final search was com- pleted on 03 January 2019 and all references were managed in EndNote X9.

2.2. Selection criteria

Studies were eligible for inclusion if they: (1) diagnosed dementia, or a subtype of dementia, on the basis of published diagnostic criteria (e.g., DSM-III) or a diagnostic tool (e.g., Cambridge Mental Disorders of the Elderly Examination); (2) evaluated clinically-significant levels of depression, anxiety and/or PTSD, which was defined on the basis of a clinical diagnosis made using published diagnostic criteria (e.g., DSM- IV) or a diagnostic tool (e.g., Geriatric Mental State Schedule - Automated Geriatric Examination for Computer Assisted Taxonomy), or using a cut-off score from a published screening tool (e.g. Geriatric Depression Scale) that has been demonstrated to identify clinically- significant levels of symptoms; (3) measured depression/anxiety/PTSD prior (Time 1) to the diagnosis of dementia (Time 2); (4) examined adults (≥ 18 years); (5) reported original, peer-reviewed research that was published in English; (6) used a longitudinal study design and in- cluded a healthy control group, the latter being defined as one that was cognitively normal and not experiencing clinically-significant levels of depression, anxiety or PTSD at Time 1; (7) screened for dementia at Time 1 (unless it was a young sample, where cognitive decline was less likely: mean age + 2 SD < 60 years); and (8) provided data that would enable the calculation of an odds ratio (OR) to estimate the risk of dementia (Time 2) in people with a prior diagnosis (or clinically-sig- nificant symptoms) of depression/anxiety/PTSD (Time 1). These data could be in the form of ORs or raw cross-tabulated data, from which an OR could be calculated (i.e. number of persons with and without de- pression/anxiety/PTSD at Time 1, and the number from each of these groups who had dementia at Time 2).

Studies were excluded if they: (1) assessed trait (i.e. a stable per- sonality characteristic) rather than state (i.e. transient mood symptoms) anxiety; (2) reported a case study or case series; or (3) recruited par- ticipants from highly selected groups with medical conditions known to

J.K. Kuring, et al. Journal of Affective Disorders 274 (2020) 247–261

248

be independent risk factors for dementia (e.g. hypertension, diabetes, high cholesterol, Mild Cognitive Impairment). Where there were mul- tiple publications that examined the same or overlapping samples (i.e. non-independent data), only those data from the paper that used the most stringent diagnostic criteria were included or, where equal, the study with the largest sample size or, where equal, the most recent study. In two cases, the data from two non-independent publications were included to maximize the sample size (Lin et al., 2017 + Wang et al., 2016; Byers et al., 2012 + Yaffe et al., 2010), with depression data taken from the former study of each pair (largest sample) and PTSD data from the latter (largest sample). Published re- views, systematic reviews/meta-analyses and conference abstracts were all excluded; however, their bibliographies, together with those of the included studies, were examined to identify any potentially relevant studies that were not captured by the above searches. None of the ad- ditional references that were identified via this method (N = 139)

proved to be eligible for inclusion, indicating that the original search criteria were effective. The first author (JKK) screened all studies, first by title and abstract, after which the full-texts of all potentially relevant studies underwent detailed review (see Figure 1 for details). In the event of any ambiguity, the second and third authors (JLM, LW) ad- ditionally reviewed the study and a consensus reached regarding elig- ibility. The authors of six papers were emailed to request additional data to enable the calculation of ORs (Becker et al., 2009; de Bruijn et al., 2014; Mawanda et al., 2017; Meziab et al., 2014; Saczynski et al., 2010; Singh-Manoux et al., 2017). Unfortunately, none of them re- sponded, necessitating their exclusion.

2.3. Data extraction and quality assessment

The following data was extracted (JKK) from all studies using a standardized template: (1) study characteristics (author, publication

Fig. 1. PRISMA flow chart documenting the literature search and screening process.

J.K. Kuring, et al. Journal of Affective Disorders 274 (2020) 247–261

249

year, sample size, length of follow-up, recruitment setting, geographical region), (2) sample demographics (sex, age, education, ethnicity, Mini Mental Status Examination [MMSE] score, percentages living in in- stitutions & living alone, socioeconomic status), (3) dementia details (type of dementia, criteria/tool used to diagnose dementia), (4) mental illness details (type of mental illness [depression, anxiety, PTSD], cri- teria/tool/clinical cut-off score used to diagnose mental illness), and (5) OR for the risk of dementia associated with prior depression/anxiety/ PTSD or raw data to calculate an OR. If studies reported multiple measures for the same mental illness (e.g., multiple measures of de- pression, such as DSM-III criteria and Neuropsychiatric Inventory [NPI]), the OR that was based on the most stringent measure/criteria or, where similar, the largest sample was used. If studies reported multiple subtypes of a mental illness, only the more severe subtype was included (e.g., depression, rather than dysthymia, included).

The reporting quality of the included studies was assessed using an adaptation of the National Institute of Health (NIH) Quality Assessment Tool for Observational Cohort and Cross-Sectional Studies (National Institute of Health, 2014). Each study was rated by the first author (JKK) in terms of whether they met (score = 2), partially met (score = 1), did not meet (score = 0), or did not report (score = 0) each of the 17 criteria (see Figure 2). The intra-rater reliability of these scores was assessed by having the same rater (JKK) re-rate a random sample of 15 studies after 6 to 8 weeks. An intra-class correlation (ICC) measured the consistency between the two total scores for each study (Rankin & Stokes, 1998) and a percentage score assessed the agreement between the scores assigned to the 17 individual criteria on the two occasions.

2.4. Effect size and statistical analysis

Summary descriptive statistics and the ICC were calculated using SPSS (Version 25). All other analyses were completed using Comprehensive Meta-Analysis (CMA; Version 3). Funnel plots were generated by CMA and forest plots were created using Microsoft Excel.

Odds ratios (ORs) (and 95% Confidence Intervals [CIs]) were used to calculate the risk of developing dementia (at Time 2), having pre- viously experienced clinically-significant depression, anxiety or PTSD (Time 1). ORs were calculated for all dementia types combined and, where possible, separately for AD, VaD, DLB and FTD (random-effects model). Although ORs are less intuitive to interpret than risk ratios (RRs), they are the preferred statistic for meta-analyses that are ana- lyzing cross-tabulated dichotomous data due to their superior statistical properties (Chen et al., 2010; Lipsey & Wilson, 2001). ORs can range in value from 0 to infinity. An OR = 1 indicates there is no relationship between depression/anxiety/PTSD at Time 1 and a subsequent diag- nosis of dementia (Time 2). ORs that fall between 0 and 1 indicate a negative association, such that depression/anxiety/PTSD at Time 1 is associated with a lower risk of subsequently being diagnosed with de- mentia (Time 2). ORs above 1 indicate a positive relationship, with depression/anxiety/PTSD at Time 1 being associated with a higher risk of later dementia (Lipsey & Wilson, 2001). As an example, an OR = 2 indicates the odds of developing dementia having previously experi- enced depression/anxiety/PTSD are 2 to 1. In other words, for every 2 people who have previously suffered from depression/anxiety/PTSD and gone on to develop dementia, there will be 1 person who did not have a history of depression/anxiety/PTSD but will develop dementia.

Between-study heterogeneity in the ORs was analyzed using both Q (to test for significant heterogeneity) and I2 (to measure the magnitude of heterogeneity: higher values indicate greater heterogeneity) statis- tics, and T2 was used to calculate the 95% prediction intervals (range within which we would expect to find the true effect/OR). Duval and Tweedies’ (1998) trim-and-fill procedure was used to assess publication bias, based on a random-effects model and looking for studies with lower ORs (left of the mean effect). This procedure adjusts for pub- lication bias by estimating the number of missing studies that may exist,

but were not published (or not captured by the searches), and calcu- lating the impact that these studies may have had on the current findings (Sutton et al., 2000).

Where possible, subgroup analyses were conducted to determine whether reporting quality or recruitment setting contributed to the observed heterogeneity in the ORs reported by different studies. Where there was missing data for one or more variables, these studies were excluded from the subgroup analyses. Studies were divided into higher and lower reporting quality (total score), based on a median split (be- cause this tool does not have a cut-off score to define high quality studies). Recruitment setting was additionally examined in the sub- group analyses because healthcare samples (e.g. health databases, in- patient/outpatient medical settings) are likely to have higher base rates of dementia than community samples, potentially impacting on ORs. Where possible, meta-regression (random-effects, method of moments, with z-distribution and log OR) was also used for moderator analyses to examine whether specific explanatory variables (mean age, percentage female) and covariates (mean length of follow-up) contributed to the observed heterogeneity. Age and sex were used as explanatory variables because the risk of dementia is known to increase with age and is higher in females (Podcasy & Epperson, 2016). Length of follow-up was ad- ditionally examined because studies that followed participants for re- latively short intervals (e.g. 1 to 5 years) would not capture later-de- veloping cases of dementia, particularly in younger cohorts (e.g. mean age: 60 to 70 vs 80 to 90s), potentially impacting on the findings.

3. Results

3.1. Study screening and summary details

As seen in Figure 1, a total of 29,963 references were identified through the literature searches, with an additional 139 references lo- cated from the bibliographies of review articles and included studies. Of these, 7,695 were duplicates which, when removed, reduced the number to 22,268 references. Preliminary screening of the titles and abstracts of these references against the eligibility criteria further re- duced the number to 249, all of which then underwent full-text review. A further 213 articles were excluded at this stage, leaving 36 studies that were eligible for inclusion in this meta-analysis.

Table 1 summarizes key demographic information for the 36 studies that were meta-analyzed (see Table 2 for specific details for each study). As expected, participants were elderly and, on average, had some high school education. Although depression and anxiety are usually more prevalent in females (Seedat et al., 2009), the inclusion of two large scale studies of military veterans (comprising ≥ 96% males; Byers et al., 2012, Nparticipants = 279,368; Yaffe et al., 2010, Nparticipants = 181,093), resulted in the overall sample comprising marginally more males than females. Although most studies recruited from community settings, the vast majority of participants were from healthcare settings. Less than half of the studies reported their mean length of follow-up (Nstudies = 16), with the mean interval between assessing depression/anxiety/PTSD (Time 1) and dementia (Time 2) ranging from 3.1 to 20.2 years (overall M = 6.3, SD = 4.0). A further 19 studies reported the minimum and/or maximum follow up interval, with the midrange interval varying between 1 to 13 years (M = 5.64, SD = 2.53), but did not provide a median from which to estimate the mean length of follow-up (Hozo et al., 2005). Very few studies provided information about socio-economic status (Nstudies = 4) and ethnicity (Nstudies = 6).

Most studies investigated all-cause dementia (Nstudies = 26), with only nine providing data for AD and two studies for VaD. Depression was the most commonly examined mental illness (Nstudies = 31), with four studies each evaluating the risk of dementia in relation to previous anxiety or PTSD. Thirty-three of the 36 studies reported controlling for confounding variables, the most common of which were age (Nstudies = 28), education (Nstudies = 24) sex (Nstudies = 23), and

J.K. Kuring, et al. Journal of Affective Disorders 274 (2020) 247–261

250

vascular risk factors (Nstudies = 21). Fewer adjusted for baseline cog- nition (Nstudies = 13) or reported controlling for comorbid mental ill- ness (Nstudies = 8), lifestyle factors (e.g. smoking; Nstudies = 6), or psychotropic medication (Nstudies = 6). Consideration was given to conducting a subgroup analysis to compare studies that controlled for specific confounding variables with those that did not. However, too few studies adjusted for the same combination of variables (Nstudies ≤ 4), which is much lower than the recommended minimum of 10 studies for a subgroup analysis (Deeks, Higgins, & Altman, 2019).

3.2. Study reporting quality

Figure 2 shows the percentage of studies that met each of the 17

criteria from the adapted NIH Quality Assessment Tool for Observa- tional Cohort and Cross-Sectional Studies (National Institute of Health, 2014). Most studies at least partially met more than half of the 17 criteria, with the average reporting quality score being 62% (SD = 7.9%). The criteria with the lowest ratings related to: study power, the severity and subtype of mental illness under investigation (e.g. major depressive disorder vs dysthymia), whether mental illness was assessed on multiple occasions prior to a dementia diagnosis or across age groups (e.g. young adults/mid-life/older adults), and whe- ther the outcome (dementia) assessors were blinded to baseline results. The first three items were likely to have a limited impact on the current meta-analysis because meta-analyses are designed to address the pro- blem of underpowered single studies. Second, the inclusion criteria

Fig. 2. Percentage of studies meeting each of the adapted NIH study quality criteria.

J.K. Kuring, et al. Journal of Affective Disorders 274 (2020) 247–261

251

ensured that the mental illness met published diagnostic criteria or exceeded clinical cut-offs. Third, depression and anxiety subtypes were not being examined. Finally, older adults were the only group relevant to this study. If outcome (dementia) assessors were not blinded to baseline results (presence or absence of mental illness), there is a po- tential for selection bias because some symptoms of mental illness mimic those of dementia (and vice versa), consequently participants with mental illness at Time 1 may have been more likely (than those without mental illness) to be screened for dementia. The majority of included studies did not report whether the outcome (dementia) as- sessors were blinded to a person's baseline results, making it difficult to know whether this had an impact on the current findings.

An intra-rater reliability analysis yielded an ICC of 0.81 (95% CIs 0.54 - 0.93), with 96% agreement between the reporting quality scores assigned to the 17 individual criteria on the two separate occasions (6-8 weeks apart). Thus, these ratings showed very good intra-rater relia- bility.

3.3. Depression

Thirty-one studies examined the risk of dementia in those with a history of depression. As seen in Figure 3, the majority of these studies had statistically significant ORs (Nstudies = 22) that ranged from 1.21 (Luppa et al., 2013) to 6.37 (Bae et al., 2015), with two additional outliers of 16.00 (Bartolini et al., 2004) and 130.73 (Lauriola et al., 2018). Most of the studies were prospective (Nstudies = 26), but there

was no obvious difference between the prospective and retrospective studies in terms of range of ORs or whether they were statistically significant. Notably the four studies with exceptionally large 95% confidence intervals (see Figure 3) were all prospective; however, this within study heterogeneity is unlikely to be explained by the pro- spective design because the other 22 prospective studies all had much narrower 95% confidence intervals. The average length of follow-up was only reported by 14 of these 31 studies and ranged from 2.4 (Kim et al., 2011) to 20.2 years (Zilkens et al., 2014).

The unadjusted pooled OR for a diagnosis of dementia (all types) in people who previously had depression was 1.91 (Nstudies = 31) (z=12.05, p<0.001) (see Figure 3). Thus, persons with a history of depression had almost twice the odds of developing dementia than those who did not have this history. Moreover, this OR did not change markedly (OR = 1.87) when the analyses were re-run after excluding the two outliers (ORs = 16.00 & 130.73). When AD was examined (Nstudies = 8), alone, the unadjusted pooled OR was 2.23 (95% CIs = 1.46 to 3.41, z=3.71, p<0.001), indicating that people with depression at Time 1 had more than twice the odds of later developing AD than those who had not suffered from depression (see Figure 4). There was one notable outlier in the ORs for AD (OR = 130.73; Lauriola et al., 2018), which was so large that it could not be shown graphically. However, when this study was excluded from the analysis, the OR proved to be only slightly lower (OR = 1.91). Unfortunately, the data for the other types of dementia were extremely limited, with a single study of VaD reporting an OR of 4.23 (95% CI 2.94 - 6.08;

Table 1 Summary demographic and study information for the mental illness and control groups.

All Participants Mental Illnessa Controlsa

Nstudies Nparticipants M SD Nstudies Nparticipants M SD Nstudies Nparticipants M SD

Sample size 36 828,767 23,021 39,715 35c 102,885 2,940 9,849 35c 725,116 20,718 54,233 Age (at baseline) 31 721,689 74.1 5.8 Education (years) 11 20,434 9.0 4.1 MMSE score 10 21,356 25.4 3.1 Length of follow-up (years) 16 247,190 6.3 4.0

% Sex: Female 34 809,169 47.9%b

Ethnicity 6 202,718 Caucasian 3 196,768 69.0%b

Black 5 200,786 6.8%b

Hispanic 5 191,234 7.8%b

Asian 2 190,573 10.8%b

Socioeconomic Status 4 516,792 Low 33.3%b

Medium 33.2%b

High 30.0%b

Nstudies Nparticipants %studiesd %particip.e Nstudies Nparticipants %studiesd %particip.e Nstudies Nparticipants %studiesd %particip.e

Dementia 36 828,767 100.0 100.0 36 102,885f 100.0 100.0 36 725,116f 100.0 100.0 All-cause 26 761,689 72.2 91.9 26 90,762f 72.2 88.2 26 670,161f 72.2 92.4 Alzheimer's disease 9 17,827 25.0 2.2 9 2,836 25.0 2.8 9 14,991 25.0 2.1 Vascular dementia 2 51,091 5.6 6.2 2 10,035 5.6 9.8 2 41,056 5.6 5.7 Dementia with Lewy bodies 1 1,136 2.8 0.1 1 44 2.8 <0.1 1 1,092 2.8 0.2 Frontotemporal dementia 1 1,136 2.8 0.1 1 44 2.8 <0.1 1 1,092 2.8 0.2

Mental Illness 36 828,767 100.0 100.0 36 102,885f 100.0 100.0 36 725,116f 100.0 100.0 Depression 31 446,669 86.1 53.9 31 45,423f 86.1 44.1 31 400,480f 86.1 55.2 Anxiety 4 17,226 11.1 2.1 4 2,294 11.1 2.2 4 14,932 11.1 2.1 PTSD 4 379,620 11.1 45.8 4 56,096 11.1 54.5 4 323,524 11.1 44.6

Recruitment Setting 36 828,767 100.0 100.0 36 102,885f 100.0 100.0 36 725,116f 100.0 100.0 Community 19 58,173 52.8 7.0 19 4,502f 52.8 4.4 19 53,671f 52.8 7.4 Health 17 774,086 47.2 93.4 17 98,383 47.2 95.6 17 675,703 47.2 93.2

MMSE = Mini Mental Status Examination. M = mean. SD = standard deviation. a Summary demographic and background data (age, education, MMSE, follow-up interval, sex, ethnicity, socioeconomic status) were not available for the mental

illness and control groups because studies did not routinely provide it. Only data for the combined samples was available. b Percentage based on total participant data available for that variable (not entire sample). c One study only provided the total sample size (n=766) and did not report the sample size for the mental illness and control groups separately. d Percentage based on total number of studies (n=36). e Percentage based on total available participant numbers (102,885 for mental illness and 725,116 for control). f Figure based on Nstudies – 1 because one study did not report separate sample sizes for those with mental illness and controls.

J.K. Kuring, et al. Journal of Affective Disorders 274 (2020) 247–261

252

Table 2 Summary details for individual studies that examined the risk of dementia (OR) associated with depression, anxiety and/or PTSD.

Study Dementia group

Dementia measure

Mental illness Mental illness measure

n Odds Ratio Age (years) M(SD) (all groups)

Female % (total)

Follow-up length (years) M (SD)

Setting Region

Almeida et al 2017 All-cause ICD-10 Depression GDS-15 294 1.70 77.2 (3.7) 0.0 8.9 (-) community Australia/NZ Controls 4628

Bae et al 2015 AD N-ADRDA Depression GDS short form

29 6.37 71.7 (5.1) 55.2 3.5 (0.3) community Asia

Controls 87 Baldwin et al 2006 All-cause ICD-10 Depression DSM-IV 32 5.37 73.3 (6.6) 63.5 3.1 (-) health Western

EuropeControls 30 Bartolini et al 2004 AD N-ADRDA Depression DSM-III-R 124 16.00 69.2 (4.8) 63.5 - health Western

EuropeControls 98 Billioti de Gage et al

2014 AD ICD-9 Depression ICD-9 224 1.22 - 67.0 - healthbc North

America/ Canada

Controls 8756 Anxiety ICD-9 1467 1.53 Controls 7513

Bonanni et al 2018 All-cause (AD/VaD/ DLB/FTD combined)

PTSD DSM-IV-TR 44 4.09 72.2 (4.1) 53.4 - healthc Western Europe

AD N-ADRDA Controls 1092 VaD N-AIREN DLB McKeith et al

1992 FTD McKhann et al

2001 Brommelhoff et al

2009 All-cause DSM-IV Depression ICD-7/8/9/

10 317 1.60 73.0 (6.4) 56.2 - communitybc Western

Europe Controls 12363

Buntinx et al 1996 All-cause ICHPPC-2 Depression ICHPPC-2 489 2.38 - - - healthbc Western EuropeControls 18614

Burke et al 2017 AD N-ADRDA Depression DSM-V 415 1.60 83.0 (9.2) 63.9 4.2 (-) healthb North America/ Canada

Controls 1336

Byers et al 2012 All-cause ICD-9-CM Depression ICD-9-CM 27594 1.85 69.7 (8.2) 4.0 - healthb North America/ Canada

Controls 251774

Dal Forno et al 2005 AD N-ADRDA Depression CES-D 149 2.03 65.5 (12.0) 45.5 6.1 (-) community North America/ Canada

Controls 1083

Flatt et al 2018 All-cause ICD-9 PTSD ICD-9 1147 0.33 71.1 (7.9) 54.7 8.0 (4.6) healthb North America/ Canada

Controls 187494

Fuhrer et al 2003 All-cause DSM-III-R Depression CES-D 527 2.00 75.2 (6.9) 58.3 - community Western EuropeControls 3250

Ganguli et al 2006 All-cause DSM-III-R Depression modified CES-D

128 1.38 74.6 (5.3) 60.8 7.4 (-) community North America/ CanadaControls 1137

Gatz et al 2005 All-cause DSM-III-R Depression CES-D 766a 2.37 74.5 (6.0) 61.7 - community North America/ Canada

Controls

Geerlings et al 2008 All-cause N-ADRDA Depression CES-D 35 3.42 73.5 (7.9) 49.0 5.9 (1.6) community Western EuropeControls 451

Goveas et al 2011 All-cause DSM-IV Depression CES-D short form

508 2.19 70.1 (3.8) 100.0 5.4 (1.6) community North America/ CanadaControls 5868

Gracia-Garcia et al 2015

AD DSM-IV Depression GMS- AGECAT

452 1.74 71.9 (9.0) 54.4 - community Western Europe

Controls 3412 Heser et al 2016 All-cause SIDAM Depression GDS-15 311 1.76 81.1 (3.5) 65.2 - health Western

EuropeControls 2391 Irie et al 2008 All-cause DSM-III-R Depression CES-D 11

items 169 2.35 76.3 (3.6) 0.0 - community North

America/ CanadaControls 1763

Kassem et al 2018 All-cause DSM-IV Depression GDS 121 2.07 82.8 (3.1) 100.0 4.9 (0.6) community North America/ Canada

Controls 1304 Anxiety Goldberg

Anxiety Scale

190 1.09

Controls 1235 Kim et al 2011 All-cause DSM-IV Depression GMS-

AGECAT 57 2.12 71.8 (5.0) 54.4 2.4 (0.3) community Asia

Controls 461 Lauriola et al 2018 AD DSM-V Depression DSM-V 115 130.73 74.5 (7.5) 59.7 - health Western

EuropeControls 66

(continued on next page)

J.K. Kuring, et al. Journal of Affective Disorders 274 (2020) 247–261

253

Lin et al., 2017) and no studies examining DLB or FTD. The Q statistics provided in Table 3 indicate that there was sig-

nificant between-study heterogeneity in the ORs when all dementia types were combined and for AD alone. The I2 statistics also show a moderate degree of between-study variability for all dementia types combined and slightly more variability for AD, as evidenced by the 95% prediction intervals (see Table 3). Thus, dementia type does not appear to be the main source of variability in the ORs reported by different studies because significant variability remained when AD was ex- amined alone.

A subgroup analysis that examined the impact of reporting quality (median split: lower vs higher ratings) on the ORs for depression and all dementia types combined revealed that there was no significant dif- ference (Q=0.008, df=1, p=0.930) between the findings from studies with lower (OR = 1.93, 95% CI 1.56 – 2.39, z=6.08, p<0.001) or higher quality ratings (OR=1.91, 95% CI 1.77 – 2.07, z=16.59, p<0.001). Thus, reporting quality did not explain the variability in ORs. Similarly, the pooled OR for studies that examined community samples (OR=1.88, 95%CI 1.70 – 2.09, z=11.79, p<0.001) did not differ significantly (Q=0.121, df=1, p=0.727) from those that re- cruited in healthcare settings (OR=1.96, 95% CI 1.63 – 2.34, z=7.33, p<0.001), indicating that recruitment setting did not account for the

between-study variability in the estimates of the risk of dementia as- sociated with previous depression.

A meta-regression (see Table 4), which examined whether the age and sex of the sample contributed to the risk of developing dementia, having previously suffered from depression, was not significant (Nstudies = 25, Qmodel=0.56, df=2, p=0.7555). Although we intended to examine a third covariate – the mean length of time between the depression assessment and diagnosis of dementia – the available data did not permit this (i.e., the subject/study [Nstudies = 14] to variable [3: age, sex, time] ratio was too low). Similarly, subgroup and meta-re- gression analyses could not be conducted for AD, due to insufficient data.

The potential impact of publication bias was assessed using Duval and Tweedies’ (1998) trim-and-fill procedure. This analysis re- vealed that the current estimate of the risk of developing dementia (all types combined) associated with prior depression was likely to be missing 10 studies (shown as solid dots in Online Resource 2) and that the imputed OR after adjusting for these missing studies was 1.73 (95% CI 1.54 – 1.95), which was only marginally lower than the aforemen- tioned OR of 1.91 (95% CI 1.72 – 2.12).

Table 2 (continued)

Study Dementia group

Dementia measure

Mental illness Mental illness measure

n Odds Ratio Age (years) M(SD) (all groups)

Female % (total)

Follow-up length (years) M (SD)

Setting Region

Lenoir et al 2011 All-cause DSM-IV Depression CES-D 1053 2.10 74.0 (5.4) 61.3 - community Western EuropeControls 6936

Li et al 2011 All-cause DSM-IV Depression CES-D 11 items

321 1.79 74.9 (6.2) 59.9 7.1 (-) community North America/ CanadaControls 3089

Lin et al 2017 VaD ICD-9-CM Depression ICD-9-CM 9991 4.23 - 61.2 - healthbc Asia Controls 39964

Lugtenburg et al 2016 All-cause GMS-AGECAT Depression GMS- AGECAT

186 2.73 - 62.3 - health Western Europe

Controls 1725 Luppa et al 2013 All-cause DSM-IV Depression DSM-III-R 128 1.21 81.3 (4.7) 73.4 4.3 (2.4) community Western

EuropeControls 760 Santabarbara et al

2018 All-cause DSM-IV Anxiety GMS-

AGECAT 54 2.61 72.1 (9.1) 54.9 - community Western

Europe Controls 2424

Vilalta-Franch et al 2013

AD DSM-IV Depression DSM-IV 41 2.89 76.7 (5.4) 63.7 - community Western EuropeControls 304

Wallin et al 2013 All-cause DSM-IV Depression DSM-IV 50 2.82 88.8 (4.1) 67.5 3.4 (1.6) community Western EuropeControls 162

Wang et al 2016 All-cause ICD-9-CM PTSD ICD-9-CM 1750 6.24 55.4 (9.2) 76.6 6.5 (2.7) healthbc Asia Controls 7000

Yaffe et al 2010 All-cause ICD-9-CM PTSD ICD-9-CM 53155 5.49 68.8 (8.6) 3.5 - healthbc North America/ Canada

Controls 127938

Zahodne et al 2016 All-cause DSM-III-R Depression CES-D 10 items

394 1.24 76.0 (6.1) 68.7 - healthb North America/ CanadaControls 2199

Zalsman et al 2000 All-cause DSM-IV Depression DSM-IV 164 1.94 77.3 (-) - - healthbc Asia Controls 338

Zilkens et al 2014 All-cause ICD-10-AM Depression ICD-10-AM 1005 1.92 - 56.6 20.2 (10.4) healthbc Australia/NZ Controls 26131 Anxiety ICD-10-AM 583 1.88 Controls 26553

a Total sample size indicated; separate sample sizes for those with mental illness and healthy controls were not reported. b Study based on registry data. c Retrospective study design. CES-D = Center for Epidemiologic Studies – Depression Scale. DSM III-R/IV/IV-TR/V = Diagnostic and Statistical Manual of Mental

Disorders. GDS= Geriatric Depression Scale. GMS-AGECAT= Geriatric Mental State – Automated Geriatric Examination for Computer Assisted Taxonomy. ICD 7/8/ 9/9AM/9CM/10/10AM = International Classification of Diseases ICHPPC-2 = International Classification of Health Problems in Primary Care. PTSD = post traumatic stress disorder. AD = Alzheimer's disease, VaD = vascular dementia. DLB = dementia with Lewy bodies. FTD = fronto-temporal dementia. N- ADRDA= National Institute of Neurological and Communicative Disorders and Stroke – Alzheimer's Disease and Related Disorders Association. N-AIREN= National Institute of Neurological and Communicative Disorders and Stroke – Association Internationale pur la Recherche et l'Enseignement en Neurosciences. SIDAM = Structured Interview for Diagnosis of Dementia of Alzheimer type, Multi-infarct Dementia and Dementia of other Aetiology according to DSM-III-R, DSM- IV, and ICD-

J.K. Kuring, et al. Journal of Affective Disorders 274 (2020) 247–261

254

3.4. Anxiety

Four studies examined the risk of dementia in elderly persons who had previously suffered from an anxiety disorder or clinically-sig- nificant levels of symptoms (see Table 1 for participant numbers). Three of them had significant ORs, ranging from 1.09 (Kassem et al., 2018) to 2.61 (Santabarbara et al., 2018) (see Figure 5). The average length of follow-up for the two studies that provided this information varied considerably, ranging from 4.9 (Kassem et al., 2018) to 20.2 (Zilkens et al., 2014) years.

As seen in Figure 5, the pooled OR for a dementia diagnosis (Time 2) in those with previous anxiety (Time 1) was 1.60 (95% CI = 1.29 to 2.00, z=4.211, p<0.001) for the combined dementia types (Nstudies = 4). Therefore, persons with clinically-significant anxiety had a 1.6 greater odds of being diagnosed with dementia at a later time, compared to those with no history of anxiety. However, it should be noted that, although there were a large number of participants, the

number of studies was small. Only one study (Billioti de Gage et al., 2014) reported an OR for AD (1.53; 95% CI 1.35 – 1.75), and no studies reported on VaD, DLB or FTD. No subgroup or meta-regression analyses could be conducted due to the small number of studies, preventing the planned exploration of some of the variables that may have impacted on these results (e.g. reporting quality, recruitment setting, age, sex, length of follow-up).

As with depression, the Q statistics indicated that there was sig- nificant between-study heterogeneity in the ORs for the combined de- mentia types (see Table 3) and the I2 statistics indicated moderate heterogeneity. However, it is important to note that these analyses are likely to be underpowered due to the small number of studies (Nstudies = 4). Although not feasible to test formally, it is possible that the sex of the participants may have contributed to the observed het- erogeneity. Whereas the one study with a non-significant OR included only females, the three studies that had significant ORs (ranging from 1.53 to 2.61) included both males and females (55% to 67% female).

Fig. 3. Forest plot for the risk (OR) of subsequent dementia diagnosis (all types) associated with previous clinically-significant symptoms of depression. *p<0.05 aLauriola et al 2018 OR could not be plotted because it exceeded the scale. Note: OR=1 indicates no association between dementia and previous depression. OR>1 indicates greater odds of a subsequent dementia diagnosis of dementia in people with previously diagnosed depression compared to people with no history of depression. OR<1 indicates lower odds of a subsequent diagnosis of dementia in people with previously diagnosed depression compared to people with no history of depression.

J.K. Kuring, et al. Journal of Affective Disorders 274 (2020) 247–261

255

There were no other obvious differences between the four studies – in terms of age, length of follow-up, recruitment setting, reporting quality, or whether prospective or retrospective data was used – that might assist in explaining the observed heterogeneity.

An examination of the potential impact of publication bias, using Duval and Tweedies’ (1998) trim-and-fill procedure, indicated that data from one study was likely to be missing (shown as a solid dot in Online Resource 3). The imputed OR after adjusting for this missing study was 1.55 (95% CI 1.24 – 1.93), which was only marginally lower than the aforementioned 1.60 (95% CI 1.29 – 2.00). However, with fewer than 10 studies this analysis was likely underpowered (Sterne et al., 2011).

3.5. PTSD

Just four studies examined PTSD and the risk of subsequently de- veloping dementia (see Table 1 for participant numbers), three of which were found to have significant ORs (see Figure 6). The ORs for these studies were very divergent, ranging from 0.33 (Flatt et al., 2018) – which indicates that persons who had previously suffered from PTSD were less likely to later develop dementia than persons who had no history of PTSD – to 6.24 (Wang et al., 2016), indicating that persons with PTSD were much more likely to subsequently develop dementia than those without such a history. The average length of follow-up for the two studies that provided this information ranged from 6.5 (Wang et al., 2016) to 8.0 years (Flatt et al., 2018) years.

As shown in Figure 6, the unadjusted pooled OR for a subsequent dementia diagnosis (all types combined) in people with PTSD was non-

significant (OR = 2.55, 95% CI 0.43 – 15.12, z = 1.302, p = 0.302), reflecting the small number of studies and divergent individual ORs. There were no obvious differences between the studies that might ex- plain the disparate findings (e.g. sample size, age, sex, length of follow- up, recruitment setting), except for the use of prospective or retro- spective data. More specifically, the single study that reported an OR indicating a lower odds of developing dementia in people with a history of PTSD was conducted prospectively. In contrast, the three remaining studies were conducted retrospectively (reviewed prior medical data to identify persons with PTSD) and had ORs (one non-significant) sug- gesting a higher odds of later developing dementia in people with a history of PTSD.

4. Discussion

This study built on existing meta-analyses by examining whether a history of clinically-significant depression, anxiety or PTSD is a risk factor (causal or prodromal) for a later diagnosis of one of the four most common types of dementia, namely AD, VaD, DLB, and FTD. The focus on clinically-significant symptoms is important because previous meta- analyses have examined a much broader range of symptoms, including subclinical and very mild levels that are commonly experienced in the general community and unlikely to pose a serious risk to health. All- cause dementia was also investigated because many studies do not identify the specific type, consequently it was assumed that this broader group primarily comprised the four most common subtypes. A total of 36 studies were eligible for inclusion; 31 examined depression, with an

Fig. 4. Forest plot for the risk (OR) of subsequent Alzheimer's disease (AD) diagnosis associated with previous clinically-significant symptoms of depression. *p<0.05 aLauriola et al 2018 OR could not be plotted because it exceeded the scale. Note: OR=1 indicates no association between dementia and previous depression. OR>1 indicates greater odds of a subsequent diagnosis of dementia in people with previously diagnosed depression compared to people with no history of depression. OR<1 indicates lower odds of a subsequent diagnosis of dementia in people with previously diagnosed depression compared to people with no history of depression.

Table 3 Heterogeneity results for depression and anxiety as risk factors (OR) for dementia.

Depression Q df(Q) p I2 T2 T 95% Prediction Interval

All dementia 84.951 30 <0.001 64.686 0.035 0.188 1.28 – 2.85 AD 25.448 7 0.001 72.493 0.193 0.439 0.67 – 7.40

Anxiety Q df(Q) p I2 T2 T 95% Prediction Interval

All dementia 8.676 3 0.034 65.421 0.027 0.165 0.68 – 3.78

J.K. Kuring, et al. Journal of Affective Disorders 274 (2020) 247–261

256

additional four studies each for anxiety and PTSD. The majority in- vestigated all-cause dementia (Nstudies = 26), with eight for AD, 1 for VaD, and none for DLB and FTD. One study provided data for all de- mentia types (combined) as well as separate results for AD, VaD, DLB and FTD; however, only the former data were analyzed to avoid using a single control group in multiple comparisons.

Older persons who had previously experienced clinically-significant depression were at a higher risk of subsequently being diagnosed with all-cause dementia (OR = 1.91, p < 0.001) and AD (OR = 2.23, p < 0.001), with approximately twice the odds of developing dementia compared to those with no history of depression. Although there was significant heterogeneity in the ORs from individual studies, subgroup analyses indicated that neither reporting quality (higher vs lower rat- ings) nor recruitment setting (healthcare vs community) adequately accounted for this variability. Similarly, a meta-regression found that neither age nor sex (percentage female) contributed to the findings. Unfortunately, it was not possible to determine whether the amount of time that had elapsed between the depression (earlier) and dementia (later) diagnoses affected the findings because the subject/study to variable ratio was too low. An examination of potential publication bias revealed a marginally lower result (OR = 1.73, 95% CI 1.54 – 1.95) after adjusting for imputed studies, while still indicating that there is a greater risk for developing dementia in persons who had previously suffered from depression.

A prior history of clinically-significant anxiety was additionally as- sociated with a higher risk of developing all-cause dementia at a later time (OR = 1.60, p < 0.001), compared with those with no such history. However, caution should be exercised when interpreting these results because of the small number of studies (Nstudies = 4) and sig- nificant heterogeneity in the ORs. It is possible that sampling differ- ences may have contributed to the observed heterogeneity: unlike the other studies, the one study with a non-significant OR examined an all- female sample. Publication bias is unlikely to have played a major role, with the OR only marginally lower after adjusting for imputed studies (OR = 1.55, 95% CI 1.24 – 1.93). Thus, both the unadjusted and ad- justed ORs indicate that there is a greater risk of being diagnosed with dementia in persons who had previously suffered from anxiety, than persons without such a history.

Unlike depression and anxiety, a prior diagnosis of PTSD was not associated with a subsequent dementia diagnosis (OR = 2.55, p > .05).

As with anxiety, this result should be interpreted cautiously because it is likely to have been affected by the small number of studies (Nstudies = 4) and the opposing findings. The one notable difference between the single study that had an OR indicating a lower risk of de- mentia in persons who had previously suffered from PTSD, and the three that indicated a higher risk, was that it used a prospective design; which is methodologically stronger.

These findings expand on those of previous meta-analyses by de- monstrating that clinically-significant levels of depression or anxiety in cognitively normal persons are associated with a greater risk of devel- oping dementia at a later time. Multiple explanations have been pro- posed to explain this link, including the causal inflammatory neuro- degeneration hypothesis, which asserts that chronic major depression (rather than mild/subclinical levels) leads to several pro-inflammatory reactions within the brain and, ultimately, the degenerative changes that characterize dementia (Leonard, 2017). Although a plausible ex- planation for the findings, the fact that the data are observational mean that it is not possible to rule out a competing explanation: namely, that depression and anxiety are prodromal symptoms of dementia, which share the same underlying disease processes, but manifest (or are identified) earlier than dementia. Importantly, these results indicate that depression and anxiety are not merely comorbidities or sequelae to dementia, because cognitive decline or dementia were not evident when these mental illnesses were assessed. Overall, the results support the hypothesis that previous depression and anxiety are risk factors for later development of dementia; however, the causal or prodromal nature of this risk remains to be determined. The relationship between PTSD and dementia remains unclear with existing studies reporting very divergent results.

Unlike other meta-analyses that have examined this topic (Becker et al., 2018; Cherbuin et al., 2015; Diniz et al., 2013; Ford et al., 2018; Gulpers et al., 2016; Jorm et al., 1991; Jorm, 2000; Jorm, 2001; Ownby et al., 2006; Santabarbara et al., 2019; Xu et al., 2015), the present study focused solely on clinically-significant mental illness (de- pression and anxiety), thus excluding mild/subclinical symptoms. This resulted in 14% to 45% overlap in the studies that were meta-analyzed here and elsewhere. Despite this, there were notable similarities in the findings, with the current findings for depression (OR = 1.91 for all- cause dementia, OR = 2.23 for AD) being comparable to those of earlier meta-analyses (OR = 1.96 all-cause dementia, OR = 1.85 AD,

Table 4 Random effects model meta-regression results for the impact of covariates age and sex on the risk of dementia associated with previous depression.

Random effects (method of moments), z-distribution, log odds ratio (OR) Covariate Coefficient Standard Error 95% lower 95% upper Z-value 2 sided p-value

Depression All dementia groups Intercept 1.4111 1.0496 -0.6461 3.4684 1.34 0.1788 Mean age -0.0106 0.0143 -0.0387 0.0175 -0.74 0.4593 Percentage female 0.0004 0.0026 -0.0046 0.0054 0.15 0.8785

Fig. 5. Forest plot for the risk (OR) of sub- sequent dementia diagnosis (all types) asso- ciated with previous clinically-significant symptoms of anxiety. *p<0.05 Note: OR=1 indicates no association between dementia and previous depression. OR>1 in- dicates greater odds of a subsequent diagnosis of dementia in people with previously diag- nosed depression compared to people with no history of depression. OR<1 indicates lower odds of a subsequent diagnosis of dementia in people with previously diagnosed depression compared to people with no history of de- pression.

J.K. Kuring, et al. Journal of Affective Disorders 274 (2020) 247–261

257

Diniz et al., 2013; OR = 1.64 all-cause dementia, OR = 1.40 AD, Ford et al., 2018). This was also true for anxiety (OR = 1.60, for the current study; OR = 2.05, Ford et al., 2018). Thus, when milder sub- clinical cases of depression, anxiety and cognitive decline are excluded, there remains a significant relationship between prior clinically-sig- nificant depression and anxiety and a subsequent diagnosis of de- mentia. However, the similarity between the ORs reported by meta- analyses that did and did not include mild/subclinical symptoms of depression or anxiety also raises the possibility that the risk of later dementia is not restricted to those who suffer from clinically-significant mental illnesses. If the risk of dementia (be it causal or prodromal) is associated with all symptoms of depression and anxiety (including mild/subclinical), regardless of severity, this may argue against the inflammatory neurodegeneration hypothesis (Leonard, 2017). It may also suggest that mental illness is only a partially modifiable risk factor because, although clinically-significant depression and anxiety can be reduced through targeted screening, treatment and preventative stra- tegies, milder symptoms usually represent normal responses to major life events and are, therefore, more difficult to prevent (Anusic et al., 2014).

Modifiability aside, the knowledge that depression and anxiety (and potentially PTSD) are risk factors – be they causal or prodromal in nature – for dementia highlights the importance of healthcare profes- sionals regularly screening for cognitive changes in older people who have a history of these mental illnesses. Regular screening for new- onset mental illness in older adults is also important, not only from a mental health perspective, but also because they may be prodromal symptoms of dementia. These screens may improve the early detection of cognitive decline and possible dementia, with the potential for ear- lier treatment and intervention. Earlier diagnosis and symptom treat- ment can help maintain functional independence for longer and reduce carer burden, thereby improving the quality of life for persons living with dementia and their carers, and reducing healthcare costs (de Vugt & Verhey, 2013).

There were several limitations to the current meta-analysis. First, it was necessary to exclude six studies after unsuccessful attempts to contact the corresponding authors in order to obtain data that would enable the calculation of ORs. Additionally, there was a further risk of omitting relevant studies because only one reviewer conducted the screening process. Reassuringly, the current publication bias analyses confirmed that missing studies are likely to have had a minimal impact on the ORs.

Second, less than half of the studies reported the mean length of follow-up, with most of the remainder only reporting ranges, which meant that its impact could not be examined statistically. For those that did report this information, the majority had an average of 6 years or

less. A longer interval between the onset of mental illness (e.g. mid-life) and later dementia diagnosis would provide more credibility for the hypothesis that depression/anxiety/PTSD are potentially causal risk factors, rather than merely prodromal symptoms; but would not rule out the possibility that mental illness and dementia share the same underlying risk factors.

Third, there was considerable missing data for several other de- mographic variables (i.e. education, ethnicity, socioeconomic status), again preventing a statistical examination of their impact on the ORs and making it difficult to comment on the generalizability of these re- sults.

Fourth, very few studies examined anxiety and PTSD. Although the findings for anxiety are based on a total of 2,294 participants, they came from only four studies and the majority (Nparticipants = 2,050) were recruited from healthcare settings, limiting the generalizability of the findings. In the case of PTSD, the majority of them used retro- spective data. Retrospective studies are typically thought to be more biased than prospective studies because the data collection is un- planned and, therefore, reliant on what is available (Norvell, 2010). Notably, the one PTSD study that reported an OR that was contrary to expectations (i.e. a lower risk of later dementia) used a more rigorous prospective design. It is therefore possible that study design sig- nificantly impacted on the ORs and the validity of the findings.

Fifth, it is possible that people with more (i.e. frequent relapses) or longer episodes of depression, anxiety or PTSD have a greater risk of subsequent dementia diagnosis due to their greater exposure to neu- roinflammation, as suggested by the inflammatory neurodegeneration hypothesis (Leonard 2017). This hypothesis could not be examined due to insufficient data.

Finally, having only one reviewer assess the quality of the included studies increased the potential for information bias. An intra-rater re- liability analysis was conducted to partially address this risk, which revealed 96% agreement across the two separate occasions indicating that the reviewer was highly consistent in their decision-making when assessing study quality. Further bias may also have resulted from the inclusion of 12 studies that used registry data, 9 of which were retro- spective, with one sourcing their data via a review of medical records (see Table 2). Registry and retrospective data are likely to have come from multiple clinicians, with varying expertise, who may have used slightly different criteria to diagnose mental illness and dementia, po- tentially reducing the reliability of the data obtained from these sources. Moreover, retrospective studies are less likely than prospective studies to have systematically followed-up all participants in order to determine whether they developed dementia, potentially leading to an underestimate of the number of dementia cases. Unfortunately, it was not possible to determine whether the number of dementia cases

Fig. 6. Forest plot for the risk (OR) of subsequent dementia diagnosis (all types) associated with previous diagnosis of PTSD. *p<0.05 Note: OR=1 indicates no association between dementia and previous depression. OR>1 indicates greater odds of a subsequent diagnosis of dementia in people with previously diagnosed depression compared to people with no history of depression. OR<1 indicates lower odds of a subsequent diagnosis of dementia in people with previously diagnosed depression compared to people with no history of depression.

J.K. Kuring, et al. Journal of Affective Disorders 274 (2020) 247–261

258

differed between retrospective/registry and prospective studies because there were fewer than 10 retrospective/registry studies for each type of mental illness (depression/anxiety/PTSD); 10 being the minimum re- commended number for subgroup analyses (Deeks et al., 2019).

Future research should endeavor to better understand the likely causal or prodromal nature of the increased risk of dementia associated with prior depression and anxiety, and seek to clarify whether prior PTSD is also a risk factor. The necessary observational nature of this research prevents causal inferences and is further complicated by the knowledge that the disease processes underlying dementia (e.g. amy- loid β deposition and tau protein accumulation) may begin decades prior to clinical diagnosis (Bateman et al., 2012). However, there are several hypotheses that could be explored to add weight to the possi- bility of a causal link between mental illness and dementia. First, pro- spective, longitudinal studies that repeatedly assess for the presence of clinically-significant depression, anxiety or PTSD across the adult age range (young adult/ mid-life/ older adult) would help to determine whether the risk of dementia differs depending on the age at which these disorders occur. A prodromal link would be more likely if only late-life mental illness is associated with an increased risk of dementia; whereas a causal link may be more likely if this risk was present, re- gardless of the age at which the mental illness was experienced. Second, an evaluation of whether the severity and number of episodes of de- pression/anxiety/PTSD influence the risk of later dementia would assist in clarifying whether greater exposure to the potential neuro-in- flammatory sequelae of these mental illnesses increases the risk of de- mentia. Although there is evidence to suggest that greater duration and frequency of depression is associated with an increased risk of later developing dementia (Dal Forno et al., 2005; Dotson, Beydoun & Zonderman, 2010), further research is needed to examine whether this dose-dependent relationship also applies to anxiety and PTSD. Ad- ditionally, it is important to examine whether the risk of dementia only increases in cases of clinically-significant depression, anxiety or PTSD, or whether milder/subclinical symptoms also increase a person's risk of dementia. Even if these milder symptoms are risk factors, it is unlikely that they could be avoided entirely, although some modification to the frequency and duration of these milder symptoms may be possible. Similarly, the impact of different diagnostic approaches to mental ill- ness (e.g. DSM-IV vs ICD-10) and dementia (e.g. DSM-IV vs N-ADRDA) across studies should be explored as a potential confounding variable to ensure the findings reflect the disorder, rather than the measurement tool. Future research should also investigate the risk of developing the four most common types of dementia (AD, VaD, DLB and FTD), having previously experienced depression, anxiety or PTSD in order to de- termine whether the risk differs between dementia-types. Finally, traumatic brain injury (TBI) is a known risk factor for depression, an- xiety, and PTSD (Rogers and Read, 2007) as well as dementia (Fann et al., 2018). As a result, TBIs may artificially inflate estimates of the extent to which mental illness increases the risk of dementia. Re- searchers should therefore provide separate risk estimates for in- dividuals with and without a TBI in order to evaluate its impact.

In conclusion, depression and anxiety appear to be risk factors (causal or prodromal) for a later diagnosis of dementia, and are not just comorbidities or sequelae to dementia. This highlights the importance of healthcare professionals screening for cognitive impairment in older persons with a history of these mental illnesses in order to assist in the early identification of cognitive decline and dementia. The causal or prodromal nature of the risk between depression and anxiety with de- mentia remains to be clarified. Additionally, the relationship between prior PTSD and subsequent dementia requires further examination. A prospective, longitudinal study, ideally examining clinically-significant depression, anxiety and PTSD at multiple time points (young adult, mid-life and older adults) and their association with later dementia diagnosis is needed to further explore the nature of the relationship.

Author Declaration

We wish to confirm that there are no known conflicts of interest associated with this publication.

Contributors

We confirm that the manuscript has been read and approved by all named authors and that there are no other persons who satisfied the criteria for authorship but are not listed. All authors were involved in the design, interpretation, and report preparation. The first and second authors were additionally involved in the data collection and analysis. We further confirm that the order of authors listed in the manuscript has been approved by all of us.

We confirm that we have given due consideration to the protection of intellectual property associated with this work and that there are no impediments to publication, including the timing of publication, with respect to intellectual property. In so doing we confirm that we have followed the regulations of our institutions concerning intellectual property.

We understand that the Corresponding Author is the sole contact for the Editorial process (including Editorial Manager and direct commu- nications with the office). He/she is responsible for communicating with the other authors and progress, submissions of revisions and final approval of proofs. We confirm that we have provided a current, correct email address which is accessible by the Corresponding Author.

Role of Funding Source

We confirm that there has been no significant financial support for this work that could have influenced its outcome. This research is supported by an Australian Government Research Training Program (RTP) Scholarship. The funding source had no involvement in the de- sign, data collection, data analysis and interpretation, report writing, or the decision to submit the article for publication.

Declaration of Competing Interest

None

Acknowledgements

The authors would like to thank M. Bell (Research Librarian, Research and Reference Services, Barr Smith Library, University of Adelaide) for her expert assistance with developing the search grids for the literature searches.

Supplementary materials

Supplementary material associated with this article can be found, in the online version, at doi:10.1016/j.jad.2020.05.020.

References

Almeida, O.P., Hankey, G.J., Yeap, B.B., Golledge, J., Flicker, L., 2017. Depression as a modifiable factor to decrease the risk of dementia. Transl Psychiatry 7 (5), e1117. https://doi.org/10.1038/tp.2017.90.

Alzheimer's Disease International. (2015). World Alzheimer Report 2015 The Global Impact of Dementia: an analysis of prevalence, incidence, cost and trends. Retrieved from London, UK: https://www.alz.co.uk/research/WorldAlzheimerReport2015.pdf.

American Psychiatric Association, 1980. Diagnostic and statistical manual of mental disorders (3rd Ed.). Author, Washington, DC.

Anusic, I., Yap, S.C.Y., Lucas, R.E., 2014. Does personality moderate reaction and adap- tation to major life events? Analysis of life satisfaction and affect in an Australian national sample. Journal of Research in Personality 51, 69–77. http://dx.doi.org/10. 1016/j.jrp.2014.04.009.

. . .Bae, J.B., Kim, Y.J., Han, J.W., Kim, T.H., Park, J.H., Lee, S.B., Kim, K.W., 2015. Incidence of and risk factors for Alzheimer's disease and mild cognitive impairment in Korean elderly. Dementia and Geriatric Cognitive Disorders 39 (1-2), 105–115.

J.K. Kuring, et al. Journal of Affective Disorders 274 (2020) 247–261

259

Baldwin, R.C., Gallagley, A., Gourlay, M., Jackson, A., Burns, A., 2006. Prognosis of late life depression: A three-year cohort study of outcome and potential predictors. International Journal of Geriatric Psychiatry 21 (1), 57–63. http://dx.doi.org/10. 1002/gps.1424.

Barnes, D.E., Yaffe, K., 2011. The projected impact of risk factor reduction on Alzheimer's disease prevalence. Lancet Neurology 10, 819–828. https://doi.org/10.1016/S1474- 4422(11)70072-2.

Bartolini, M., Coccia, M., Luzzi, S., Provinciali, L., Ceravolo, M., 2004. Motivational Symptoms of Depression Mask Preclinical Alzheimer's Disease in Elderly Subjects. Dementia and Geriatric Cognitive Disorders 19 (1), 31–36. http://dx.doi.org/10. 1159/000080968.

Bateman, R.J., Xion, C., Benzinger, T.L.S., Fagan, A.M., Goate, A., Fox, N.C., ... Morris, J.C., 2012. Clinical and biomarker changes in dominantly inherited Alzheimer's disease. The New England Journal of Medicine 367, 795–804.

. . .Becker, J.T., Chang, Y.-F., Lopez, O.L., Dew, M.A., Sweet, R.A., Barnes, D., Reynolds III, C.F., 2009. Depressed mood is not a risk factor for incident dementia in a com- munity-based cohort. The American Journal of Geriatric Psychiatry 17 (8), 653–663. http://dx.doi.org/10.1097/JGP.0b013e3181aad1fe.

Becker, E., Orellana Rios, C.L., Lahmann, C., Rucker, G., Bauer, J., Boeker, M., 2018. Anxiety as a risk factor of Alzheimer's disease and vascular dementia. Br J Psychiatry 213 (5), 654–660. https://doi.org/10.1192/bjp.2018.173.

. . .Billioti de Gage, S., Moride, Y., Ducruet, T., Kurth, T., Verdoux, H., Tournier, M., Begaud, B., 2014. Benzodiazepine use and risk of Alzheimer's disease: case-control study. BMJ 349, g5205. https://doi.org/10.1136/bmj.g5205.

. . .Bonanni, L., Franciotti, R., Martinotti, G., Vellante, F., Flacco, M.E., Di Giannantonio, M., Onofrj, M., 2018. Post Traumatic Stress Disorder Heralding the Onset of Semantic Frontotemporal Dementia. Journal of Alzheimers Disease 63, 203–215. https://doi. org/10.3233/jad-171134.

Brommelhoff, J.A., Gatz, M., Johansson, B., McArdle, J.J., Fratiglioni, L., Pedersen, N.L., 2009. Depression as a risk factor or prodromal feature for dementia? Findings in a population-based sample of Swedish twins. Psychology and Aging 24 (2), 373–384. http://dx.doi.org/10.1037/a0015713.

Buntinx, F., Kester, A., Bergers, J., Knottnerus, J.A., 1996. Is depression in elderly people followed by dementia? A retrospective cohort study based in general practice. Age Ageing 25 (3), 231–233.

Burke, S.L., Cadet, T., Alcide, A., O'Driscoll, J., Maramaldi, P., 2017. Psychosocial risk factors and Alzheimer's disease: the associative effect of depression, sleep dis- turbance, and anxiety. Aging & Mental Health 1–8. https://doi.org/10.1080/ 13607863.2017.1387760.

Byers, A.L., Covinsky, K.E., Barnes, D.E., Yaffe, K., 2012. Dysthymia and depression in- crease risk of dementia and mortality among older veterans. The American Journal of Geriatric Psychiatry 20, 664–672. http://dx.doi.org/10.1097/JGP. 0b013e31822001c1.

Chen, H., Cohen, P., Chen, S., 2010. How big is a big odds ratio? Interpreting the mag- nitudes of odds ratios in epidemiological studies. Communications in Statistics – Simulation and Computation 39, 860–864.

Cherbuin, N., Kim, S., Anstey, K.J., 2015. Dementia risk estimates associated with mea- sures of depression: A systematic review and meta-analysis. BMJ Open 5 (12).

Dal Forno, G., Palermo, M.T., Donohue, J.E., Karagiozis, H., Zonderman, A.B., Kawas, C.H., 2005. Depressive Symptoms, Sex, and Risk for Alzheimer's Disease. Annals of Neurology 57 (3), 381–387. http://dx.doi.org/10.1002/ana.20405.

de Bruijn, R.F., Direk, N., Mirza, S.S., Hofman, A., Koudstaal, P.J., Tiemeier, H., Ikram, M., 2014. Anxiety is not associated with the risk of dementia or cognitive decline: The rotterdam study. The American Journal of Geriatric Psychiatry 22 (12), 1382–1390. http://dx.doi.org/10.1016/j.jagp.2014.03.001.

Deckers, K., van Boxtel, M.P., Schiepers, O.J., de Vugt, M., Munoz Sanchez, J.L., Anstey, K.J., ... Kohler, S., 2015. Target risk factors for dementia prevention: a systematic review and Delphi consensus study on the evidence from observational studies. International Journal of Geriatric Psychiatry 30, 234–246. https://doi.org/10.1002/ spgs.4245.

Deeks, J.J., Higgins, J.P.T., Altman, D.G., 2019. Chapter 10: Analysing data and under- taking meta-analyses. In: Higgins, J.P.T., Thomas, J., Chandler, J., Cumpston, M., Li, T., Page, M.J., Welch, V.A. (Eds.), Cochrane Handbook for Systematic Reviews of Interventions, version 6.0 (updated July 2019). Cochrane, 2019. Available from. www.training.cochrane.org/handbook.

De Vugt, M.E., Verhey, F.R.J., 2013. The impact of early dementia diagnosis and inter- vention on informal caregivers. Progress in Neurobiology 110, 54–62.

Diniz, B.S., Butters, M.A., Albert, S.M., Dew, M.A., Reynolds III, C.F., 2013. Late-life depression and risk of vascular dementia and Alzheimer's disease: Systematic review and meta-analysis of community-based cohort studies. The British Journal of Psychiatry 202 (5), 329–335. http://dx.doi.org/10.1192/bjp.bp.112.118307.

Dotson, V.M., Beydoun, M.A., Zonderman, A.B., 2010. Recurrent depressive symptoms and the incidence of dementia and mild cognitive impairment. Neurology 75, 27–34.

Duval, S., Tweedie, R., 1998. Practical estimates of the effect of publication bias in meta- analysis. Australasian Epidemiologist 5, 14–17.

Fann, J.R., Riisgaard Ribe, A., Schou Pedersen, H., Fenger-Gron, M., Christensen, J., Eriksen Benros, M., Vestergaard, M., 2018. Long-term risk of dementia among people with traumatic brain injury in Denmark: a population-based observational cohort study. Lancet Psychiatry 5, 424–431. http://dx.doi.org/10.1016/S2215-0366(18) 30065-8.

Flatt, J.D., Gilsanz, P., Quesenberry, C.P., Albers, K.B., Whitmer, R.A., 2018. Post-trau- matic stress disorder and risk of dementia among members of a health care delivery system. Alzheimer's and Dementia 14 (1), 28–34. https://doi.org/10.1016/j.jalz. 2017.04.014.

. . .Ford, E., Greenslade, N., Paudyal, P., Bremner, S., Smith, H.E., Banerjee, S., Cassell, J., 2018. Predicting dementia from primary care records: A systematic review and meta-

analysis. PLoS ONE 13, e0194735. https://doi.org/10.1371/journal.pone.0194735. Fuhrer, R., Dufouil, C., Dartigues, J.F., 2003. Exploring Sex Differences in the

Relationship Between Depressive Symptoms and Dementia Incidence: Prospective Results From the PAQUID Study. Journal of the American Geriatrics Society 51 (8), 1055–1063. http://dx.doi.org/10.1046/j.1532-5415.2003.51352.x.

Ganguli, M., Du, Y., Dodge, H.H., Ratcliff, G.G., Chang, C.-C.H., 2006. Depressive symptoms and cognitive decline in late life. A prospective epidemiological study. Archives of General Psychiatry 63 (2), 153–160. http://dx.doi.org/10.1001/ archpsyc.63.2.153.

Gatz, J.L., Tyas, S.L., St. John, P., Montgomery, P., 2005. Do Depressive Symptoms Predict Alzheimer's Disease and Dementia. The Journals of Gerontology: Series A: Biological Sciences and Medical Sciences 60 (6), 744–747. http://dx.doi.org/10. 1093/gerona/60.6.744.

Geerlings, M., den Heijer, T., Koudstaal, P., Hofman, A., Breteler, M., 2008. History of depression, depressive symptoms, and medial temporal lobe atrophy and the risk of Alzheimer disease. Neurology 70 (15), 1258–1264. http://dx.doi.org/10.1212/01. wnl.0000308937.30473.d1.

Goveas, J.S., Espeland, M.A., Woods, N.F., Wassertheil-Smoller, S., Kotchen, J.M., 2011. Depressive symptoms and incidence of mild cognitive impairment and probable de- mentia in elderly women: The Women's Health Initiative Memory Study. Journal of the American Geriatrics Society 59 (1), 57–66. http://dx.doi.org/10.1111/j.1532- 5415.2010.03233.x.

. . .Gracia-Garcia, P., de-la-Camara, C., Santabarbara, J., Lopez-Anton, R., Quintanilla, M.A., Ventura, T., Lobo, A., 2015. Depression and incident alzheimer disease: The impact of disease severity. The American Journal of Geriatric Psychiatry 23 (2), 119–129. http://dx.doi.org/10.1016/j.jagp.2013.02.011.

Gulpers, B., Ramakers, I., Hamel, R., Kohler, S., Oude Voshaar, R., Verhey, F., 2016. Anxiety as a predictor for cognitive decline and dementia: A systematic review and meta-analysis. The American Journal of Geriatric Psychiatry 24 (10), 823–842. http://dx.doi.org/10.1016/j.jagp.2016.05.015.

. . .Heser, K., Bleckwenn, M., Wiese, B., Mamone, S., Riedel-Heller, S.G., Stein, J., Wagner, M., 2016. Late-life depressive symptoms and lifetime history of major depression: Cognitive deficits are largely due to incipient dementia rather than depression. Journal of Alzheimer's Disease 54 (1), 185–199.

Hiskey, S., Luckie, M., Davies, S., Brewin, C.R., 2008. The emergence of posttraumatic distress in later life: A review. Journal of Geriatric Psychiatry and Neurology 21, 232–241.

Hozo, S.P., Djulbegovic, B., Hozo, I., 2005. Estimating the mean and variance from the median, range, and the size of a sample. BMC Medical Research Methodology 5, 13. https://doi.org/10.1186/1471-2288-5-13.

. . .Irie, F., Masaki, K.H., Petrovitch, H., Abbott, R.D., Ross, G., Taaffe, D.R., White, L.R., 2008. Apolipoprotein E epsilon4 allele genotype and the effect of depressive symp- toms on the risk of dementia in men: The Honolulu-Asia Aging Study. Archives of General Psychiatry 65 (8), 906–912. http://dx.doi.org/10.1001/archpsyc.65.8.906.

. . .Jorm, A.F., van Duijn, C.M., Chandra, V., Fratiglioni, L., Graves, A.B., Heyman, A., Hoffman, A., for the EURODEM Risk Factors Research Group, 1991. Psychiatric history and related exposures as risk factors for Alzheimer's disease: a collaborative re-analysis of case-control studies. EURODEM Risk Factors Research Group. International Journal of Epidemiology 20, S43–S47.

Jorm, A., 2000. Is depression a risk factor for dementia or cognitive decline? A review. Gerontology 46 (4), 219–227. http://dx.doi.org/10.1159/000022163.

Jorm, A.F., 2001. History of depression as a risk factor for dementia: An updated review. Australian and New Zealand Journal of Psychiatry 35 (6), 776–781. http://dx.doi. org/10.1046/j.1440-1614.2001.00967.x.

Kales, H.C., Gitlin, L.N., Lyketsos, C.G., 2015. Assessment and management of behavioral and psychological symptoms of dementia. BMJ 350, h369. https://doi.org/10.1136/ bmj.h369.

. . .Kassem, A.M., Ganguli, M., Yaffe, K., Hanlon, J.T., Lopez, O.L., Wilson, J.W., Cauley, J.A., 2018. Anxiety symptoms and risk of dementia and mild cognitive impairment in the oldest old women. Aging & Mental Health 22 (4), 474–482. https://doi.org/10. 1080/13607863.2016.1274370.

. . .Kim, J.M., Stewart, R., Kim, S.Y., Kim, S.W., Bae, K.Y., Yang, S.J., Yoon, J.S., 2011. Synergistic associations of depression and apolipoprotein E genotype with incidence of dementia. Int J Geriatr Psychiatry 26 (9), 893–898. https://doi.org/10.1002/gps. 2621.

Kuring, J.K., Mathias, J.L., Ward, L., 2018. Prevalence of Depression, Anxiety and PTSD in People with Dementia: a Systematic Review and Meta-Analysis. Neuropsychology Review 28, 393–416. https://doi.org/10.1007/s11065-018-9396-2.

. . .Lauriola, M., Mangiacotti, A., D'Onofrio, G., Cascavilla, L., Paris, F., Ciccone, F., Greco, A., 2018. Late-Life Depression versus Amnestic Mild Cognitive Impairment: Alzheimer's Disease Incidence in 4 Years of Follow-Up. Dementia and Geriatric Cognitive Disorders 46 (3-4), 140–153. https://doi.org/10.1159/000492489.

Lawlor, B.A., 1996. Environmental and social aspects of behavioural disturbances in dementia. International Psychogeriatrics 8, 259–261.

Lenoir, H., Dufouil, C., Auriacombe, S., Lacombe, J.M., Dartigues, J.F., Ritchie, K., Tzourio, C., 2011. Depression history, depressive symptoms and incident dementia: The 3c study. Alzheimer's and Dementia 7 (4), S355–S356.

Leonard, B.E., 2017. Major Depression as a Neuroprogressive Prelude to Dementia: What Is the Evidence. Mod Trends Pharmacopsychiatry 31, 56–66. https://doi.org/10. 1159/000470807.

. . .Li, G., Wang, L.Y., Shofer, J.B., Thompson, M.L., Peskind, E.R., McCormick, W., Larson, E.B., 2011. Temporal relationship between depression and dementia: Findings from a large community-based 15-year follow-up study. Archives of General Psychiatry 68 (9), 970–977. http://dx.doi.org/10.1001/archgenpsychiatry.2011.86.

Lin, W.-C., Hu, L.-Y., Tsai, S.-J., Yang, A.C., Shen, C.-C., 2017. Depression and the risk of vascular dementia: a population-based retrospective cohort study. International

J.K. Kuring, et al. Journal of Affective Disorders 274 (2020) 247–261

260

Journal of Geriatric Psychiatry 32 (5), 556–563. https://doi.org/10.1002/gps.4493. Lipsey, M.W., Wilson, D.B., 2001. Applied social research methods series: Vol. 49.

Practical Meta-Analysis. Sage Publications Inc, CA, US. Livingstone, G., Sommerlad, A., Orgeta, V., Costafreda, S.G., Huntley, J., Ames, D., ...

Mukadam, N., 2017. Dementia prevention, intervention, and care. Lancet 390, 2673–2734. http://dx.doi.org/10.1016/S0140-6736(17)31363-6.

Lugtenburg, A., Zuidersma, M., Voshaar, R.C., Schoevers, R.A., 2016. Symptom dimen- sions of depression and 3-year incidence of dementia: Results from the Amsterdam study of the elderly. Journal of Geriatric Psychiatry and Neurology 29 (2), 99–107. http://dx.doi.org/10.1177/0891988715606235.

Luppa, M., Luck, T., Ritschel, F., Angermeyer, M.C., Villringer, A., Riedel-Heller, S.G., 2013. Depression and incident dementia. An 8-year population-based prospective study. PLoS ONE 8 (3) 2013, ArtID e59246, 8(3). http://dx.doi.org/10.1371/journal. pone.0059246.

Mawanda, F., Wallace, R.B., McCoy, K., Abrams, T.E., 2017. PTSD, psychotropic medi- cation use, and the risk of dementia among US veterans: a retrospective cohort study. Journal of the American Geriatrics Society 65, 1043–1050. https://doi.org/10.1111/ jgs.14756.

Meziab, O., Kirby, K.A., Williams, B., Yaffe, K., Byers, A.L., Barnes, D.E., 2014. Prisoner of war status, posttraumatic stress disorder, and dementia in older veterans. Alzheimer's & Dementia: The Journal of the Alzheimer's Association 10, S236–S241. http://dx. doi.org/10.1016/j.jalz.2014.04.004.

Moher, D., Liberati, A., Tetzlaff, J., Altman, D.G., The PRISMA Group, 2009. Preferred reporting items for systematic reviews and meta-analyses: the PRISMA statement. PLoS Medicine 6, e1000097. https://doi.org/10.1371/journal.pmed1000097.

National Institute of Health. (2014). Quality Assessment Tool for Observational Cohort and Cross-Sectional Studies. Retrieved from https://www.nhlbi.nih.gov/health- topics/study-quality-assessment-tools.

Norton, S., Matthews, F.E., Barnes, D.E., Yaffe, K., Brayne, C., 2014. Potential for primary prevention of Alzheimer's disease: an analysis of population-based data. The Lancet, Neurology 13, 788–794. https://doi.org/10.1016/s1474-4422(14)70136-x.

Norvell, D.C., 2010. Study types and bias – don't judge a study by the abstract's conclusion alone. Evidence-Based Spine-Care Journal 1, 7–10. https://doi.org/10.1055/s-0028- 1100908.

Ownby, R.L., Crocco, E., Acevedo, A., John, V., Loewenstein, D., 2006. Depression and Risk for Alzheimer Disease: Systematic Review, Meta-analysis, and Metaregression Analysis. Archives of General Psychiatry 63 (5), 530–538. http://dx.doi.org/10. 1001/archpsyc.63.5.530.

Podcasy, J.L., Epperson, C.N., 2016. Considering sex and gender in Alzheimer disease and other dementias. Dialogues in Clinical Neuroscience 18, 437–446.

Rafferty, L.A., Cawkill, P.E., Stevelink, S.A.M., Greenberg, K., Greenberg, N., 2017. Dementia, post-traumatic stress disorder and major depressive disorder: a review of the mental health risk factors for dementia in the military veteran population. Psychological Medicine 1–10. https://doi.org/10.1017/s0033291717001386.

Rankin, G., Stokes, M., 1998. Reliability of assessment tools in rehabilitation: an illus- tration of appropriate statistical analyses. Clinical Rehabilitation 12, 187–199.

Rogers, J.M., Read, C.A., 2007. Psychiatric comorbidity following traumatic brain injury. Brain Injury 21, 1321–1333. https://doi.org/10.1080/02699050701765700.

Saczynski, J.S., Beiser, A., Seshadri, S., Auerbach, S., Wolf, P.A., Au, R., 2010. Depressive symptoms and risk of dementia: the Framingham Heart Study. Neurology 75 (1), 35–41. https://doi.org/10.1212/WNL.0b013e3181e62138.

. . .Santabarbara, J., Lopez-Anton, R., de la Camara, C., Lobo, E., Gracia-Garcia, P., Villagrasa, B., Lobo, A., 2018. Clinically significant anxiety as a risk factor for de- mentia in the elderly community. Acta Psychiatrica Scandinavica 1–9. https://doi. org/10.1111/acps.12966.

Santabárbara, J., Lipnicki, D.M., Villagrasa, B., Lobo, E., Lopez-Anton, R., 2019. Anxiety and risk of dementia: Systematic review and meta-analysis of prospective cohort studies. Maturitas 119, 14–20. https://doi.org/10.1016/j.maturitas.2018.10.014.

. . .Seedat, S., Scott, K.M., Angermeyer, M.C., Berglund, P., Bromet, E.J., Brugha, T.S., Kessler, R.C., 2009. Cross-national associations between gender and mental disorders in the WHO World Mental Health Surveys. Archives of General Psychiatry 66, 785–795.

Singh-Manoux, A., Dugravot, A., Fournier, A., Abell, J., Ebmeier, K., Kivimaki, M., Sabia, S., 2017. Trajectories of Depressive Symptoms Before Diagnosis of Dementia: A 28- Year Follow-up Study. JAMA Psychiatry 74 (7), 712–718. https://doi.org/10.1001/ jamapsychiatry.2017.0660.

Stella, F., Radanovic, M., Balthazar, M.L.F., Canineu, P.R., de Souza, L.C., Forlenza, O.V., 2014. Neuropsychiatric symptoms in the prodromal stages of dementia. Current Opinions in Psychiatry 27, 230–235. https://doi.org/10.1097/YCO. 0000000000000050.

Sterne, J.A.C., Sutton, A.J., Ioannidis, J.P.A., Terrin, N., Jones, D.R., Lau, J., ... Higgins, J.P.T., 2011. Recommendations for examining and interpreting funnel plot asym- metry in meta-analyses of randomized controlled trials. BMJ 343, d4002.

Stroup, D.F., Berlin, J.A., Morton, S.C., Olkin, I., Williamson, G.D., Rennie, D., ... the Meta-analysis Of Observational Studies in Epidemiology (MOOSE) Group, 2000. Meta-analysis of observational studies in epidemiology; a proposal for reporting. Journal of the American Medical Association 283, 2008–2012.

Sutton, A.J., Duval, S.J., Tweedie, R.L., Abrams, K.R., Jones, D.R., 2000. Empirical as- sessment of effect of publication bias on meta-analyses. British Medical Journal 10, 1574–1577.

Vilalta-Franch, J., Lopez-Pousa, S., Llinas-Regla, J., Calvo-Perxas, L., Merino-Aguado, J., Garre-Olmo, J., 2013. Depression subtypes and 5-year risk of dementia and Alzheimer's disease in patients aged 70 years. International Journal of Geriatric Psychiatry 28 (4), 341–350. http://dx.doi.org/10.1002/gps.3826.

Wallin, K., Bostrom, G., Kivipelto, M., Gustafson, Y., 2013. Risk factors for incident de- mentia in the very old. International Psychogeriatrics 25 (7), 1135–1143. http://dx. doi.org/10.1017/S1041610213000409.

. . .Wang, T.-Y., Wei, H.-T., Liou, Y.-J., Su, T.-P., Bai, Y.-M., Tsai, S.-J., Chen, M.-H., 2016. Risk for developing dementia among patients with posttraumatic stress disorder: A nationwide longitudinal study. Journal of Affective Disorders 205, 306–310. http:// dx.doi.org/10.1016/j.jad.2016.08.013.

. . .Xu, W., Tan, L., Wang, H.-F., Jiang, T., Tan, M.-S., Tan, L., Yu, J.-T., 2015. Meta- analysis of modifiable risk factors for Alzheimer's disease. Journal of Neurology, Neurosurgery & Psychiatry 86 (12), 1299–1306.

. . .Yaffe, K., Vittinghoff, E., Lindquist, K., Barnes, D., Covinsky, K.E., Neylan, T., Marmar, C., 2010. Posttraumatic stress disorder and risk of dementia among US veterans. Archives of General Psychiatry 67 (6), 608–613. http://dx.doi.org/10.1001/ archgenpsychiatry.2010.61.

Zahodne, L.B., Schupf, N., Brickman, A.M., Mayeux, R., Wall, M.M., Stern, Y., Manly, J.J., 2016. Dementia Risk and Protective Factors Differ in the Context of Memory Trajectory Groups. Journal of Alzheimer's Disease 52 (3), 1013–1020.

Zalsman, G., Aizenberg, D., Sigler, M., Nahshony, E., Karp, L., Weizman, A., 2000. Increased risk for dementia in elderly psychiatric inpatients with late-onset major depression. Journal of Nervous and Mental Disease 188 (4), 242–243. http://dx.doi. org/10.1097/00005053-200004000-00010.

Zilkens, R.R., Bruce, D.G., Duke, J., Spilsbury, K., Semmens, J.B., 2014. Severe psychiatric disorders in mid-life and risk of dementia in late-life (age 65-84 years): A population based case-control study. Current Alzheimer Research 11 (7), 681–693. http://dx.doi. org/10.2174/1567205011666140812115004.

J.K. Kuring, et al. Journal of Affective Disorders 274 (2020) 247–261

261