Hostname: page-component-7c8c6479df-p566r Total loading time: 0 Render date: 2024-03-18T01:01:18.329Z Has data issue: false hasContentIssue false

Longitudinal associations of affective symptoms with mid-life cognitive function: evidence from a British birth cohort

Published online by Cambridge University Press:  21 March 2019

Amber John
Affiliation:
PhD Student, EDGE Lab, School of Psychology, University of Sussex, UK
Sarah-Naomi James
Affiliation:
Postdoctoral researcher, MRC Unit for Lifelong Health and Ageing at UCL, UK
Urvisha Patel
Affiliation:
MSc Student, EDGE Lab, School of Psychology, University of Sussex, UK
Jennifer Rusted
Affiliation:
Professor of Experimental Psychology, School of Psychology, University of Sussex, UK
Marcus Richards
Affiliation:
Programme Leader, MRC Unit for Lifelong Health and Ageing at UCL, UK
Darya Gaysina*
Affiliation:
Senior Lecturer in Psychology, EDGE Lab, School of Psychology, University of Sussex, UK
*
Correspondence: Darya Gaysina, EDGE Lab, School of Psychology, University of Sussex, Pevensey 1 Building, Brighton BN1 9QH, UK. Email: d.gaysina@sussex.ac.uk
Rights & Permissions [Opens in a new window]

Abstract

Background

Affective disorders are associated with poorer cognition in older adults; however, whether this association can already be observed in mid-life remains unclear.

Aims

To investigate the effects of affective symptoms over a period of 30 years on mid-life cognitive function. First, we explored whether timing (sensitive period) or persistence (accumulation) of affective symptoms predicted cognitive function. Second, we tested how different longitudinal trajectories of affective symptoms were associated with cognitive function.

Method

The study used data from the National Child Development Study. Memory, verbal fluency, information processing speed and accuracy were measured at age 50. Affective symptoms were measured at ages 23, 33, 42 and 50 and used to derive longitudinal trajectories. A structured modelling approach compared a set of nested models in order to test accumulation versus sensitive period hypotheses. Linear regressions and structural equation modelling were used to test for longitudinal associations of affective symptoms with cognitive function.

Results

Accumulation of affective symptoms was found to be the best fit for the data, with persistent affective symptoms being associated with poorer immediate memory (b = −0.07, s.e. = 0.03, P = 0.01), delayed memory (b = −0.13, s.e. = 0.04, P < 0.001) and information processing accuracy (b = 0.18, s.e. = 0.08, P = 0.03), but not with information processing speed (b = 3.15, s.e. = 1.89, P = 0.10). Longitudinal trajectories of repeated affective symptoms were associated with poorer memory, verbal fluency and information processing accuracy.

Conclusions

Persistent affective symptoms can affect cognitive function in mid-life. Effective management of affective disorders to prevent recurrence may reduce risk of poor cognitive outcomes and promote healthy cognitive ageing.

Declaration of interest

None.

Type
Papers
Copyright
Copyright © The Royal College of Psychiatrists 2019 

Affective disorders, such as depression and anxiety, have been associated with impaired cognition, dementia and accelerated cognitive decline in late adulthood.Reference John, Patel, Rusted, Richards and Gaysina1Reference Gulpers, Ramakers, Hamel, Köhler, Oude Voshaar and Verhey3 However, there are significant gaps in our understanding of this association. The majority of studies have been conducted in older adults, usually over the age of 60 and have only considered this association in the context of short follow-up periods (<10 years).Reference Singh-Manoux, Akbaraly, Marmot, Melchior, Ankri and Sabia4 Therefore, it is unclear how affective symptoms are associated with cognitive function over time. There is some evidence that risk of dementia increases with the number of affective episodes experienced over the life course; whereby rate of dementia is estimated to increase by around 13% with each additional depressive episode.Reference Kessing and Andersen5 Additionally, evidence shows that individuals with persistent depressive symptoms may be at greater risk for subsequent cognitive deficits than individuals with fewer episodes.Reference Singh-Manoux, Akbaraly, Marmot, Melchior, Ankri and Sabia4 Another line of evidence suggests that timing, rather than frequency, of affective symptoms can be an important predictor of cognitive outcomes. For example, depressive symptoms in older age only were shown to be associated with increased risk of dementia.Reference Singh-Manoux, Dugravot, Fournier, Abell, Ebmeier and Kivimäki6 However, there is limited inconsistent evidence for associations between affective disorders and cognitive function before old age in two studies using data from the Medical Research Council (MRC) National Survey of Health and Development. The first studyReference Richards, Barnett, Xu, Croudace, Gaysina and Kuh7 found no associations of affective symptoms measured between ages 13 to 53 with objectively measured cognitive outcomes at age 60–64, whereas the most recent study reported that persistent case-level affective symptoms were associated with poorer cognitive state, verbal memory and letter search speed and accuracy at age 69.Reference James, Davis, O'Hare, Sharma, John and Gaysina8

In this current study we tested for associations of affective symptoms across adulthood (ages 23, 33, 42 and 50) with aspects of cognitive function at age 50. Specifically, two complementary advanced statistical approaches were employed. First, a structured modelling approach comparing a set of nested models was used to investigate whether accumulation or timing of case-level affective symptoms can predict mid-life cognition. Second, a latent growth mixture modelling approach was employed to derive longitudinal trajectories of affective symptoms and to examine how different longitudinal trajectories of affective symptoms were associated with mid-life cognition.

Method

Participants

Participants were from the National Child Development Study (NCDS), a population representative sample of 18 558 people born during 1 week of 1958 in England, Scotland and Wales (i.e. British 1958 birth cohort). This sample comprised 98.2% of total births that week. Data were subsequently collected from participants at ages 7, 11, 16, 23, 33, 42, 46 and 50 years. Further information about the study sample, respondent profiles and data collection methods are described in depth by Power & Elliott.Reference Power and Elliott9 Data are available from the UK Data Service.1014 Cohort members gave written informed consent to participate. Ethical approval for the data collection in 2008 was provided by London Medical Research and Ethics Committee (REC reference: 08/H0718/29). Ethical approval for the present study was also provided by the University of Sussex (Reference number: ER/AJ316/1).

Detailed information about the sample analysed in the present study is presented in supplementary File 1, supplementary Fig. 1 and supplementary Table 1 available at https://doi.org/10.1192/bjp.2019.24. There were 9790 people who participated in the 2008 data collection (at age 50). A total of 9385 people in this sample completed the cognitive tests at this time point (95.9%). Of the 9385 participants who completed the cognitive tests at age 50, 4625 had complete information for all predictors and covariates (49.3% of those with cognitive measures).

Measures

Cognitive function

Cognitive measures collected at age 50 were verbal memory, verbal fluency, information processing speed and accuracy.Reference Brown and Dodgeon15 Memory was assessed using a word-recall test with immediate and delayed components. Verbal fluency was assessed using an animal naming task, in which participants named as many animals as possible in a 1 min period. Information processing was assessed using a letter cancellation task. In this task, cohort members crossed as many target letters as possible from a grid of letters in a 1 min period. The results of this task were split into two scores. The total number of letters scanned represented information processing speed, and the number of target letters missed up to the final letter searched represented information processing accuracy. Accuracy was negatively scored, so higher scores represent poorer performance.

Affective symptoms

Affective symptoms were assessed using the Malaise Inventory Scale, administered at ages 23, 33, 42, and 50.Reference Hope, Power and Rodgers16Reference Lacey, Bartley, Pikhart, Stafford, Cable and Coleman18 The items in the malaise inventory scale are available online at: https://www.ukdataservice.ac.uk/teaching-resources/malaise/questions.asp. This is a measure of psychological distress and comprises 24 self-completion items, which are combined to assess levels of emotional disturbance and associated somatic symptoms. The total number of questions answered ‘yes’ was summed, creating a sum score of affective symptoms out of 24. This score was dichotomised using a standard cut-off threshold score of 7 out of 24 to represent ‘case-level’ affective symptoms, indicating clinically relevant affective symptoms.Reference Hope, Power and Rodgers16Reference Lacey, Bartley, Pikhart, Stafford, Cable and Coleman18 At age 50, a short form of the Malaise Inventory Scale was administered, comprising nine items. For this time point, a recommended cut-off score of three was used to categorise case-level affective symptoms.Reference Bowling, Pikhartova and Dodgeon19

As the short form of the Malaise Inventory Scale was used at age 50, the nine items from this short form were extracted from the longer forms used at ages 23, 33, and 42, in order to make malaise scores over time more comparable to be used for modelling longitudinal trajectories.

Covariates

Based on previous research, the following factors were included as potential confounders: gender, childhood cognition,Reference McGurn, Deary and Starr20 childhood emotional adjustment,Reference Lupien, McEwen, Gunnar and Heim21 childhood socio-economic position,Reference Kaplan, Turrell, Lynch, Everson, Helkala and Salonen22 adult socio-economic positionReference Sattler, Toro, Schönknecht and Schröder23 and education.Reference Brayne, Ince, Keage, McKeith, Matthews and Polvikoski24 Childhood cognition was assessed at age 11, using a general ability test administered at the child's school. Childhood emotional adjustment was assessed at age 11 using the Bristol Social Adjustment Guides.Reference Mcdermott and Watkins25 This questionnaire was completed by teachers and is designed to assess behaviour that may be indicative of maladjustment and emotional disturbance.

A measure of household socio-economic position at age 11 was derived using guidelines from the Centre for Longitudinal Studies,Reference Elliott and Lawrence26 based on measures of father's occupation, mother's occupation and household tenure. Highest educational attainment was derived by combining education data from 1991, 2000, 2004 and 2008 to ascertain the highest academic qualification the cohort member had achieved by age 50. Adult socio-economic position was based on occupation, with three categories (working, intermediate and middle class). Additional information about these covariates are included in supplementary File 2.

Analytical procedure

First, a structured modelling approach was used to compare a set of nested models corresponding to accumulation and sensitive period hypotheses to a saturated model including all main effects and all possible interactions.Reference James, Davis, O'Hare, Sharma, John and Gaysina8, Reference Mishra, Nitsch, Black, De Stavola, Kuh and Hardy27, Reference Murray, Mishra, Kuh, Guralnik, Black and Hardy28 The sensitive period model included three measures indicating whether an individual experienced case-level affective symptoms at three time windows across the life course: early adulthood (age 23); middle adulthood (ages 33 and/or 42); and mid-life (age 50). The accumulation model included a measure of the number of time points (of the three windows of time described above) in which each individual experienced case-level affective symptoms. The saturated model was also compared with a ‘null’ model that assumed no effect of affective symptoms on mid-life cognition. Partial F-tests were used to compare each hypothesis to the fully saturated model. Where multiple P-values were >0.05, the model with the highest P-value and lowest F-statistic were selected as the best fit for the data. This analysis was conducted in Stata (V14.2).

Four variables with case-level affective symptoms were summed up to create a variable for accumulation of adult affective symptoms (ranging from 0 to 4). Linear regression models were fitted to test for associations of accumulation of adult affective symptoms and cognitive function at age 50. For these analyses, four models were fitted: unadjusted (model 1); adjusted for gender only (model 2); additionally, adjusted for child cognition, emotional disturbance and socio-economic position (model 3); and additionally, adjusted for highest educational attainment, and adult socio-economic position (model 4). This analysis was conducted in R (V3.5.1).

To account for missing data, multiple imputation analysis was conducted on the sample with cognitive data. All analyses described above were re-run using imputed variables for the key predictors. Multiple imputation was conducted in R using the MICE package.Reference van Buuren and Groothuis-Oudshoorn29, Reference Azur, Stuart, Frangakis and Leaf30 Twenty imputations were conducted using data across seven sweeps over the life course. This multiple imputation approach includes a large number of covariates and auxiliary variables in the models, which maximises the plausibility of the missing at random assumption, and limits possibility of missing not at random data.Reference Coley, Gardette, Cantet, Gillette-Guyonnet, Nourhashemi and Vellas31 These multiple imputation techniques have been used extensively to address missing data in the NCDS.Reference Solís C, Kelly-Irving, Fantin, Darnaudéry, Torrisani and Lang32Reference Blane, Wahrendorf, Webb and Netuveli34 Technical details of the multiple imputation process are reported in supplementary File 3.

Second, a confirmatory factor analysis was conducted to generate latent factor scores of affective symptoms at each time point (see supplementary Table 2 for details about model fit). Linear mixed models were then used to examine affective symptom trajectories over time, using factor scores from each time point. Linear and quadratic models were fitted and compared and the model with the best fit according to the Akaike information criterion (AIC) and Bayesian information criterion (BIC) was selected for subsequent analyses. Next, growth mixture models were fit to the data to identify trajectory classes. Models with a two-class, three-class, four-class and five-class solution were fitted and compared using AIC, BIC, and the Lo-Mendel-Rubin adjusted likelihood ratio test. Once longitudinal trajectories were identified using growth mixture models, a structural equation model was used to investigate whether class membership predicted cognitive outcomes at age 50, after controlling for key covariates. One-step estimation approaches have been criticised in recent years on the basis that including distal outcomes into the measurement model in one step may lead to an unintended and problematic circular relationship in which the classes from the trajectory modelling are determined in part by the distal outcome that they are meant to be predicting.Reference Vermunt35Reference Zhu, Steele and Moustaki37 This analysis was therefore conducted in a step-wise fashion to avoid drawbacks associated with one-step estimation methods. Missing data were dealt with using full-information maximum-likelihood estimation (FIML). Analysis was conducted in Mplus V.8. According to Mplus defaults, variances across classes were held equal.

Results

Missing data and descriptive statistics

Of the participants who completed the cognitive tests at age 50 (n = 9385), those with complete information (n = 4625) were compared with those with missing data (n = 4760). Results revealed that participants with missing data did not differ by gender (P = 0.54). However, participants with missing data had significantly lower childhood cognitive scores (P < 0.001), higher level of childhood psychological maladjustment (P < 0.001), lower level of education (P < 0.001), lower socio-economic position in childhood (P < 0.001) and adulthood (P < 0.001), and more case-level affective symptoms at ages 23 (P < 0.001), 33 (P < 0.001), 42 (P < 0.001) and 50 (P < 0.001).

The imputed data-set was created for the sample of 9385 participants. Sociodemographic information for the participants with complete information for cognitive outcomes, main predictors and covariates (n = 4625; 50.5% women) and the imputed sample (n = 9385; 50.8% women) is presented in supplementary Table 3.

Accumulation of affective symptoms and mid-life cognitive function

Results from the analysis using a structured modelling approach revealed that the accumulation model was the best fit for the data. For this model, there were no significant differences observed from the saturated model for any of the cognitive outcomes (Table 1). This was particularly prominent for memory outcomes (immediate memory: F = 1.81; P = 0.09; delayed memory: F = 1.39, P = 0.22), for which none of the other models (i.e. ‘no effect’ model, or sensitive period models) fit the data well (P < 0.05 for all other models).

Table 1 A structured modelling approach comparing life course hypotheses of the association between lifetime affective symptoms and mid-life cognitive function (no effect hypothesis, accumulation hypothesis, sensitive period hypothesis)a

a. Values in italic represent estimates that are non-significant at the P < 0.05 level. Non-significant P-values represent a good fit for the data. The hypothesis with the smallest F statistic is taken as the hypothesis with the best fit for the data.

Fully adjusted regression models revealed significant linear associations, whereby as the number of case-level affective symptoms increased over the life course, mid-life immediate memory (b = −0.07, s.e. = 0.03, P = 0.01), delayed memory (b = −0.13, s.e. = 0.04, P < 0.001) and information processing accuracy (b = 0.18, s.e. = 0.08, P = 0.03) decreased (Table 2). No effects were apparent for verbal fluency or information processing speed in fully adjusted models. Results of the analyses using imputed data were similar (supplementary Tables 4 and 5).

Table 2 Life course accumulation of affective symptoms and cognitive function at age 50a

a. Results are presented for model 1 (unadjusted estimates), model 2 (estimates adjusted for gender), model 3 (estimates adjusted for gender, childhood socio-economic status, childhood emotional adjustment, and childhood cognition) and model 4 (estimates adjusted for gender, childhood socio-economic status, childhood emotional adjustment, childhood cognition, adult socio-economic status and highest educational attainment at age 50). Model 5 is a sensitivity analysis, using all the covariates in model 4, but excluding childhood emotional adjustment. Values in bold represent estimates significant at the P<0.05 level. Estimates are based on sample with complete data for all key factors and covariates (n = 4625).

Trajectories of affective symptoms and mid-life cognitive function

Linear mixed models revealed that a linear trajectory (AIC = 9020.291, BIC = 9065.883), rather than a quadratic trajectory (AIC = 9021.989, BIC = 9075.18), fitted adult affective symptom latent variables scores better and were therefore used to identify trajectory classes. When comparing models with a two-class, three-class, four-class and five-class solution, the five-class solution was found to be the best fit for the data (supplementary Table 6). Therefore, five different trajectories of adult affective symptoms were identified (Fig. 1): Class 1 – no affective symptoms (51.4%); Class 2 – consistent mild/moderate affective symptoms (28.3%); Class 3 – initially low and increasing to high affective symptoms (5.4%); Class 4 – initially high and persistently increasing affective symptoms (7.3%); Class 5 – initially high and decreasing to low affective symptoms (7.5%).

Class 1: no affective symptoms (51.4%); class 2: mild/moderate affective symptoms (28.3%); class 3: initially low and rapidly increasing affective symptoms (5.4%); class 4: initially high and persistently increasing affective symptoms (7.3%); class 5: initially high and steadily decreasing affective symptoms (7.5%).

Fig. 1 Longitudinal trajectories of affective symptoms (estimated from five-class growth mixture model).

A fully adjusted structural equation model revealed that class membership predicted cognitive function at age 50 (Table 3). Belonging to a trajectory with initially high and increasing affective symptoms was associated with lower mid-life immediate memory (b = −0.25, s.e. = 0.07, P < 0.001), delayed memory (b = −0.23, s.e. = 0.09, P = 0.006), and verbal fluency scores (b = −0.79, s.e. = 0.30, P = 0.01), as compared with the trajectory with no affective symptoms. Belonging to a trajectory with initially low and increasing affective symptoms was associated with significantly lower mid-life immediate (b = −0.16, s.e. = 0.07, P = 0.03) and delayed memory (b = −0.23, s.e. = 0.09, P = 0.007) and information processing accuracy (b = 0.50, s.e. = 0.21, P = 0.02). Belonging to a trajectory with initially high and decreasing affective symptoms was associated with poorer immediate memory (b = −0.15, s.e. = 0.07, P = 0.04). Finally, belonging to a trajectory with consistently mild/moderate affective symptoms was associated with poorer immediate memory (b = −0.09, s.e. = 0.04, P = 0.03). No associations were found for information processing speed and accuracy. Unadjusted and partially adjusted models are reported in supplementary Table 7.

Table 3 Fully adjusted structural equation model output for class membership predicting mid-life cognitive function

a. Values in bold represent estimates significant at the P < 0.05 level.

Discussion

The present study found that accumulation of affective symptoms across three decades of adulthood (from age 23 to age 50) was associated with poorer cognitive function in mid-life: a greater number of case-level affective symptoms was linearly associated with poorer memory and information processing accuracy in mid-life. Analysis of longitudinal trajectories of affective symptoms showed that belonging to a trajectory with a high and increasing level of affective symptoms across adulthood was significantly associated with poorer verbal memory and fluency in mid-life. Belonging to a trajectory with low and increasing affective symptoms was associated with lower verbal memory and information processing accuracy scores at age 50. Belonging to a trajectory with initially high and decreasing affective symptoms or a trajectory with consistently mild/moderate affective symptoms was associated with lower immediate memory scores in mid-life.

These findings suggest that associations between affective symptoms and cognitive function may be evident even by mid-life, an earlier stage in the life course than considered by many previous studies.Reference Ganguli, Du, Dodge, Ratcliff and Chang38 It is now believed that for those experiencing dementia in later life, there is a long preclinical period before cognitive impairment becomes evident.Reference Morris39 It is possible that older participants may have already developed cerebral pathology by the time of baseline assessment, even if they are not yet displaying symptoms of cognitive impairment. It is plausible that associations between affective disorders and cognition in older adults may be the result of reverse causality from subtle cognitive changes short of dementia. For this reason, it is important to explore this association earlier in the life course and in a population who have not developed dementia pathology. Mid-life may prove a better age to guarantee forward temporal associations between risk factors, such as affective disorders and subsequent cognitive impairments.Reference Lancaster, Tabet and Rusted40, Reference Exalto, Quesenberry, Barnes, Kivipelto, Biessels and Whitmer41 Additionally, if associations between affective disorders and cognitive function are already apparent by mid-life, this may be an important window for early intervention.Reference Sperling, Jack and Aisen42 These results advance previous findings, demonstrating for the first time that effects may be apparent even many years prior to development of any substantial cognitive deficits and also may be apparent even if no dementia develops. This observation has value in clarifying the temporal order of this association and minimises the issue of reverse causality inherent in studies focusing exclusively on older adults.

These results also suggest that accumulation of multiple, repeated affective episodes can predict poorer cognitive function in mid-life. Although the late-life trajectory of cognitive decline in this current cohort cannot be modelled, the findings add to previous research that reports a monotonic increase in dementia risk with each additional affective episode, by demonstrating that the pattern of accumulating affective episodes is important in predicting mid-life cognitive function in a general population.Reference Kessing and Andersen5 Singh-Manoux et al,Reference Singh-Manoux, Dugravot, Fournier, Abell, Ebmeier and Kivimäki6 investigated depression trajectories over a period of 28 years for a population of individuals who developed dementia up to 2015 (participants aged 65–85 in 2015), and those who remained cognitively healthy, and found an accelerated growth in depressive symptoms during the decade before dementia diagnosis. Our results complement findings from recent studies using data from the MRC National Survey of Health and Development cohort born in 1946, which reported that recurrent case-level affective symptoms were associated with poorer cognitive state (as measured with Addenbrooke's Cognitive Examination third edition), verbal memory, and search speed and accuracy in early old age,Reference James, Davis, O'Hare, Sharma, John and Gaysina8 but not with objectively measured cognitive outcomes in late mid-adulthood.Reference Richards, Barnett, Xu, Croudace, Gaysina and Kuh7 Because prevalence of depression and psychological distress has been increasing in the UK over the past few decadesReference Fink, Patalay, Sharpe, Holley, Deighton and Wolpert43 it is possible that the effect of mental health problems on cognitive ageing can manifest earlier (by mid-life) in younger cohorts, such as individuals in the NCDS born in 1958.

Taken together, these results suggest that verbal memory is affected by the number of affective episodes experienced and by longitudinal trajectories of affective symptoms. Information processing accuracy is affected by the number of affective episodes experienced and by a trajectory of initially low and increasing affective episodes. Verbal fluency scores are affected by a pattern of initially high and increasing accumulation of affective episodes. Strongest effects of affective symptoms were therefore observed on mid-life verbal memory, verbal fluency and information processing accuracy, whereas information processing speed was consistently unaffected by affective symptoms. Several studies have found untreated depressive symptoms persistent over the entire life course may lead to cumulative hippocampal volume loss;Reference Sheline, Gado and Kraemer44 a key structure associated with verbal memory.Reference Bird and Burgess45 This could potentially explain effects of affective problems on memory. It is possible that mid-life may be too early for effects to be observed in information processing speed, and that effects of affective disorders may become apparent on this cognitive domain as individuals transition from mid-life into older adulthood.

Strengths and limitations

The key strength of the study is a large nationally representative sample with a long follow-up period (three decades). Moreover, multiple cognitive domains were assessed in mid-life, and prospective assessments of affective symptoms using the same instrument from early adulthood to mid-life were available for the analyses. However, results should be interpreted with consideration of a number of limitations. Cognitive function was only assessed at one time point in adulthood (age 50) in the NCDS, and therefore it was not possible to investigate the effects of affective problems on cognitive change over time. Cognitive assessments at this time point were also limited in breadth; single, rather than multiple, cognitive tests were used for each domain (immediate and delayed memory, verbal fluency, information processing speed and accuracy); and functions in other cognitive domains (such as inhibitory processes, attention) known to be affected in people with affective disorders, were not assessed in NCDS. Additionally, although cognitive ability in childhood was controlled for, this does not completely eliminate the possibility of reverse causality, whereby the association may operate in the opposite direction with lower cognitive function leading to higher affective symptoms across the adulthood.Reference Hatch, Jones, Kuh, Hardy, Wadsworth and Richards46 We were also unable to take the effect of medications into account, which may play an important role in this association.

Missing data are inevitable in the long-running cohorts such as NCDS, and indeed there was a lot of missing data that could potentially lead to biased estimates. We have dealt with this by imputing missing data using a multiple imputation approach. The benefits of the multiple imputation approach are that missing data can be dealt with prior to analysisReference Schafer and Graham47 using a large amount of additional information from other variables available in the data-set. Specifically, this allows extra information that is not included in our main models (such as birth weight and parental education) to be used as auxiliary variables to aid with imputing the missing data. This maximises plausibility of the missing at random assumption.Reference Coley, Gardette, Cantet, Gillette-Guyonnet, Nourhashemi and Vellas31 Notably, the results from imputed data were substantially identical to the ones obtained using a complete data-set and also were consistent and complementary with the analysis using FIML to account for missing data.

Plausible mechanisms

Both biological and sociobehavioural pathways have been proposed to explain the link between affective disorders and cognitive function. Specifically, hypothalamic–pituitary–adrenal (HPA) axis function has been proposed as one potential mechanism linking affective problems and subsequent cognitive dysfunction.Reference Byers and Yaffe48 Affective symptoms and chronic stress may give rise to HPA axis activity and lead to increased glucocorticoid production, which in turn may lead to hippocampal atrophy and cognitive dysfunction. Animal studies exploring response to stress have proposed that conditions of high stress and exogenous glucocorticoids can lead to cognitive impairment.Reference Park, Zoladz, Conrad, Fleshner and Diamond49

Another potential pathway involves cardio–metabolic risk factors. Affective problems and psychological distress across the life course have been associated with higher cardio–metabolic risk.Reference Winning, Glymour, McCormick, Gilsanz and Kubzansky50 Additionally, cardio–metabolic disorders have also been linked with Alzheimer's disease and cognitive decline.Reference Bhat51 Related to this, physical health and health behaviours,Reference Ströhle52 including physical activity and exercise may also play an important role in the association between affective problems and cognitive ageing. Beyond this, chronic inflammation plays a role in both depression and dementia, and as such may act as an important pathway between the two. Moreover, β-amyloid deposition is known to play an important role in the pathogenesis of dementiaReference Hardy and Selkoe53 and has additionally been associated with major depression.Reference Wu, Hsiao, Chen, Chen, Hsieh and Wai54 Educational attainmentReference Geerlings, Schmand, Braam and Jonker55 and socio-economic positionReference Sattler, Toro, Schönknecht and Schröder23 may also play a role in the association between affective problems and cognitive ageing. It is plausible that the association between affective problems and cognitive ageing is underpinned by a complex interaction of biological and sociobehavioural mechanisms, rather than by one single aetiological determinant.Reference Da Silva, Gonçalves-Pereira, Xavier and Mukaetova-Ladinska2

Implications

In conclusion, the present study suggests that individuals with affective symptoms in adulthood are at increased risk of poorer cognitive outcomes by mid-life. This finding has implications for prevention efforts, as the asymptomatic phase before development of dementia pathology may be a critical window to target for early intervention.Reference Sperling, Jack and Aisen42 Additionally, in the absence of pathological change, one further important avenue of research is to investigate whether effective treatment and management of affective problems early in life can reduce risk of poor cognitive outcomes and promote healthy cognitive ageing. Future research should also focus on determining the biological and sociobehavioural mechanisms that underpin the association between affective and cognitive factors. All interventions to promote and sustain healthy ageing are important to health policy development in an ageing population.

Funding

We would like to acknowledge the Economic and Social Research Council (ESRC) for supporting this project (Grant number: ES/J500173/1).

Acknowledgements

We would like to acknowledge the Economic and Social Research Council (ESRC) for supporting this project. We would also like to thank the Centre for Longitudinal Studies (CLS), UCL Institute of Education for use of the data. We also acknowledge the UK Data Service for making data freely available. Finally, we would like to thank the cohort members who have dedicated their time to the National Child Development Study.

Supplementary material

Supplementary material is available online at https://doi.org/10.1192/bjp.2019.24.

References

1John, A, Patel, U, Rusted, J, Richards, M, Gaysina, D. Affective problems and decline in cognitive state in older adults: a systematic review and meta-analysis. Psychol Med 2018; 49(3): 353–65.Google Scholar
2Da Silva, J, Gonçalves-Pereira, M, Xavier, M, Mukaetova-Ladinska, EB. Affective disorders and risk of developing dementia: systematic review. Br J Psychiatry 2013; 202: 177–86.Google Scholar
3Gulpers, B, Ramakers, I, Hamel, R, Köhler, S, Oude Voshaar, R, Verhey, F. Anxiety as a predictor for cognitive decline and dementia: a systematic review and meta-analysis. Am J Geriatr Psychiatry 2016; 24: 823–42.Google Scholar
4Singh-Manoux, A, Akbaraly, TN, Marmot, M, Melchior, M, Ankri, J, Sabia, S, et al. Persistent depressive symptoms and cognitive function in late midlife: the Whitehall II study. J Clin Psychiatry 2010; 71: 1379.Google Scholar
5Kessing, LV, Andersen, PK. Does the risk of developing dementia increase with the number of episodes in patients with depressive disorder and in patients with bipolar disorder? J Neurol Neurosurg Psychiatry 2004; 75: 1662–6.Google Scholar
6Singh-Manoux, A, Dugravot, A, Fournier, A, Abell, J, Ebmeier, K, Kivimäki, M, et al. Trajectories of depressive symptoms before diagnosis of dementia. JAMA Psychiatry 2017; 74: 712.Google Scholar
7Richards, M, Barnett, JH, Xu, MK, Croudace, TJ, Gaysina, D, Kuh, D, et al. Lifetime affect and midlife cognitive function: prospective birth cohort study. Br J Psychiatry 2014; 204: 194–9.Google Scholar
8James, S-N, Davis, D, O'Hare, C, Sharma, N, John, A, Gaysina, D, et al. Lifetime affective problems and later-life cognitive state: over 50 years of follow-up in a British Birth Cohort Study. J Affect Disord 2018; 241: 348355.Google Scholar
9Power, C, Elliott, J. Cohort profile: 1958 British birth cohort (National Child Development Study). Int J Epidemiol 2006; 35: 3441.Google Scholar
10University of London, Institute of Education, Centre for Longitudinal Studies. National Child Development Study: Childhood Data, Sweeps 0-3, 1958–1974. [data collection] 3rd Edition: SN 5565. National Birthday Trust Fund, National Children's Bureau [original data producer(s)]. National Birthday Trust Fund, 2014. http://doi.org/10.5255/UKDA-SN-5565-2Google Scholar
11University of London, Institute of Education, Centre for Longitudinal Studies. National Child Development Study: Sweep 4, 1981, and Public Examination Results, 1978. [data collection] 2nd Edition: SN 5566. National Children's Bureau [original data producer(s)]. National Children's Bureau, 2008. http://doi.org/10.5255/UKDA-SN-5566-1Google Scholar
12University of London, Institute of Education, Centre for Longitudinal Studies. National Child Development Study: Sweep 5, 1991. [data collection] 2nd Edition: SN 5567. City University, Social Statistics Research Unit [original data producer(s)]. City University, Social Statistics Research Unit, 2008. http://doi.org/10.5255/UKDA-SN-5567-1Google Scholar
13University of London, Institute of Education, Centre for Longitudinal Studies. National Child Development Study: Sweep 6, 1999–2000. [data collection] 2nd Edition: SN 5578. Joint Centre for Longitudinal Research [original data producer(s)]. Joint Centre for Longitudinal Research, 2008. http://doi.org/10.5255/UKDA-SN-5578-12008Google Scholar
14University of London, Institute of Education, Centre for Longitudinal Studies. National Child Development Study: Sweep 8, 2008–2009. [data collection] 3rd Edition: SN 6137. UK Data Service, 2012. http://doi.org/10.5255/UKDA-SN-6137-2Google Scholar
15Brown, M, Dodgeon, B. NCDS Cognitive Assessments at age 50: Initial Results. CLS Cohort Studies, Working Paper. Centre for Longitudinal Studies, Institute of Education, University of London, 2010.Google Scholar
16Hope, S, Power, C, Rodgers, B. Does financial hardship account for elevated psychological distress in lone mothers? Soc Sci Med 1999; 49: 1637–49.Google Scholar
17Sacker, A, Cable, N. Do adolescent leisure-time physical activities foster health and well-being in adulthood? Evidence from two British birth cohorts. Eur J Public Health 2006; 16: 331–5.Google Scholar
18Lacey, R, Bartley, M, Pikhart, H, Stafford, M, Cable, N, Coleman, L. Parental separation and adult psychological distress: evidence for the “reduced effect” hypothesis? Longit Life Course Stud 2012; 3: 359–68.Google Scholar
19Bowling, A, Pikhartova, J, Dodgeon, B. Is mid-life social participation associated with cognitive function at age 50? Results from the British National Child Development Study (NCDS). BMC Psychol 2016; 4: 58.Google Scholar
20McGurn, B, Deary, IJ, Starr, JM. Childhood cognitive ability and risk of late-onset Alzheimer and vascular dementia. Neurology 2008; 71: 1051–6.Google Scholar
21Lupien, SJ, McEwen, BS, Gunnar, MR, Heim, C. Effects of stress throughout the lifespan on the brain, behaviour and cognition. Nat Rev Neurosci 2009; 10: 434–45.Google Scholar
22Kaplan, GA, Turrell, G, Lynch, JW, Everson, SA, Helkala, E-L, Salonen, JT. Childhood socioeconomic position and cognitive function in adulthood. Int J Epidemiol 2001; 30: 256–63.Google Scholar
23Sattler, C, Toro, P, Schönknecht, P, Schröder, J. Cognitive activity, education and socioeconomic status as preventive factors for mild cognitive impairment and Alzheimer's disease. Psychiatry Res 2012; 196: 90–5.Google Scholar
24Brayne, C, Ince, PG, Keage, HAD, McKeith, IG, Matthews, FE, Polvikoski, T, et al. Education, the brain and dementia: neuroprotection or compensation? Brain 2010; 133: 2210–6.Google Scholar
25Mcdermott, PA, Watkins, MW. Dimensions of maladaptive behavior among kindergarten level children. Counc Except Child Dimens 2017; 7: 11–7.Google Scholar
26Elliott, J, Lawrence, J. Refining Childhood Social Class Measures in the 1958 British Cohort Study in the 1958 British Birth Cohort Study. University of London: Centre for Longitudinal Studies, Institute of Education, 2014.Google Scholar
27Mishra, G, Nitsch, D, Black, S, De Stavola, B, Kuh, D, Hardy, R. A structured approach to modelling the effects of binary exposure variables over the life course. Int J Epidemiol 2009; 38: 528–37.Google Scholar
28Murray, ET, Mishra, GD, Kuh, D, Guralnik, J, Black, S, Hardy, R. Life course models of socioeconomic position and cardiovascular risk factors: 1946 birth cohort. Ann Epidemiol 2011; 21: 589–97.Google Scholar
29van Buuren, S, Groothuis-Oudshoorn, K. mice: multivariate imputation by chained equations in R. J Stat Softw 2011; 45: 168.Google Scholar
30Azur, MJ, Stuart, EA, Frangakis, C, Leaf, PJ. Multiple imputation by chained equations: what is it and how does it work? Int J Methods Psychiatr Res 2011; 20: 40–9.Google Scholar
31Coley, N, Gardette, V, Cantet, C, Gillette-Guyonnet, S, Nourhashemi, F, Vellas, B, et al. How should we deal with missing data in clinical trials involving Alzheimer's disease patients? Curr Alzheimer Res 2011; 8: 421–33.Google Scholar
32Solís C, B, Kelly-Irving, M, Fantin, R, Darnaudéry, M, Torrisani, J, Lang, T, et al. Adverse childhood experiences and physiological wear-and-tear in midlife: findings from the 1958 British birth cohort. Proc Natl Acad Sci 2015; 112: E738–46.Google Scholar
33Kelly-irving, M, Lepage, B, Dedieu, D, Lacey, R, Cable, N, Bartley, M, et al. Childhood adversity as a risk for cancer: findings from the 1958 British birth cohort study. BMC Public Health 2013; 13: 767.Google Scholar
34Blane, D, Wahrendorf, M, Webb, E, Netuveli, G. Life course influences on quality of life at age 50 years: evidence from the National Child Development Study (1958 British birth cohort study). Longit Life Course Stud 2012; 3: 346–58.Google Scholar
35Vermunt, JK. Latent class modeling with covariates: two improved three-step approaches. Polit Anal 2010; 18: 450–69.Google Scholar
36Bakk, Z, Vermunt, JK. Robustness of stepwise latent class modeling with continuous distal outcomes. Struct Equ Model A Multidiscip J 2015; 23: 2031.Google Scholar
37Zhu, Y, Steele, F, Moustaki, I. A general 3-step maximum likelihood approach to estimate the effects of multiple latent categorical variables on a distal outcome. Struct Equ Model A Multidiscip J 2017; 24: 643–56.Google Scholar
38Ganguli, M, Du, Y, Dodge, HH, Ratcliff, GG, Chang, C-CH. Depressive symptoms and cognitive decline in late life. Arch Gen Psychiatry 2006; 63: 153.Google Scholar
39Morris, JC. Early-stage and preclinical Alzheimer disease. Alzheimer Dis Assoc Disord 2005; 19: 163.Google Scholar
40Lancaster, C, Tabet, N, Rusted, J. The elusive nature of APOE ε4 in mid-adulthood: understanding the cognitive profile. J Int Neuropsychol Soc 2017; 23: 239–53.Google Scholar
41Exalto, LG, Quesenberry, CP, Barnes, D, Kivipelto, M, Biessels, GJ, Whitmer, RA. Midlife risk score for the prediction of dementia four decades later. Alzheimer's Dement 2014; 10: 562–70.Google Scholar
42Sperling, RA, Jack, CR, Aisen, PS. Testing the right target and right drug at the right stage. Sci Transl Med 2011; 3: 111cm33.Google Scholar
43Fink, E, Patalay, P, Sharpe, H, Holley, S, Deighton, J, Wolpert, M. Mental health difficulties in early adolescence: a comparison of two cross-sectional studies in England from 2009 to 2014. J Adolesc Heal 2015; 56: 502–7.Google Scholar
44Sheline, YI, Gado, MH, Kraemer, HC. Untreated depression and hippocampal volume loss. Am J Psychiatry 2003; 160: 1516–8.Google Scholar
45Bird, CM, Burgess, N. The hippocampus and memory: insights from spatial processing. Nat Rev Neurosci 2008; 9: 182–94.Google Scholar
46Hatch, SL, Jones, PB, Kuh, D, Hardy, R, Wadsworth, MEJ, Richards, M. Childhood cognitive ability and adult mental health in the British 1946 birth cohort. Soc Sci Med 2007; 64: 2285–96.Google Scholar
47Schafer, JL, Graham, JW. Missing data: our view of the state of the art. Psychol Methods 2002; 7(2): 147–77.Google Scholar
48Byers, A, Yaffe, K. Depression and risk of developing dementia. Nat Rev Neurol 2011; 7: 323–31.Google Scholar
49Park, CR, Zoladz, PR, Conrad, CD, Fleshner, M, Diamond, DM. Acute predator stress impairs the consolidation and retrieval of hippocampus-dependent memory in male and female rats. Learn Mem 2008; 15: 271–80.Google Scholar
50Winning, A, Glymour, MM, McCormick, MC, Gilsanz, P, Kubzansky, LD. Psychological distress across the life course and cardiometabolic risk findings from the 1958 British birth cohort study. J Am Coll Cardiol 2015; 66: 1577–86.Google Scholar
51Bhat, NR. Linking cardiometabolic disorders to sporadic Alzheimer's disease: a perspective on potential mechanisms and mediators. J Neurochem 2010; 115: 551–62.Google Scholar
52Ströhle, A. Physical activity, exercise, depression and anxiety disorders. J Neural Transm 2009; 116: 777–84.Google Scholar
53Hardy, J, Selkoe, DJ. The amyloid hypothesis of Alzheimer's disease: progress and problems on the road to therapeutics. Science 2002; 297: 353–6.Google Scholar
54Wu, KY, Hsiao, IT, Chen, CS, Chen, CH, Hsieh, CJ, Wai, YY, et al. Increased brain amyloid deposition in patients with a lifetime history of major depression: evidenced on 18F-florbetapir (AV-45/Amyvid) positron emission tomography. Eur J Nucl Med Mol Imaging 2014; 41: 714–22.Google Scholar
55Geerlings, MI, Schmand, B, Braam, AW, Jonker, IC. Depressive symptoms and risk of Alzheimer's disease in more highly educated older people. J Am Geriatr Soc 2000; 48: 1092–7.Google Scholar
Figure 0

Table 1 A structured modelling approach comparing life course hypotheses of the association between lifetime affective symptoms and mid-life cognitive function (no effect hypothesis, accumulation hypothesis, sensitive period hypothesis)a

Figure 1

Table 2 Life course accumulation of affective symptoms and cognitive function at age 50a

Figure 2

Fig. 1 Longitudinal trajectories of affective symptoms (estimated from five-class growth mixture model).

Class 1: no affective symptoms (51.4%); class 2: mild/moderate affective symptoms (28.3%); class 3: initially low and rapidly increasing affective symptoms (5.4%); class 4: initially high and persistently increasing affective symptoms (7.3%); class 5: initially high and steadily decreasing affective symptoms (7.5%).
Figure 3

Table 3 Fully adjusted structural equation model output for class membership predicting mid-life cognitive function

Supplementary material: File

John et al. supplementary material

John et al. supplementary material
Download John et al. supplementary material(File)
File 88.8 KB
Submit a response

eLetters

No eLetters have been published for this article.