1. Introduction
Major depressive disorder (MDD) is a leading cause of disease burden worldwide [Reference Ferrari, Charlson, Norman, Patten, Freedman and Murray1]. Although MDD aetiology remains elusive, a large proportion of its genetic covariance is attributable to neuroticism [Reference Kendler, Neale, Kessler, Heath and Eaves2,Reference Jardine, Martin and Henderson3], suggesting a causal relationship. Neuroticism is a partially-heritable personality trait representing high emotionality and stress sensitivity [Reference Matthews, Deary and Whiteman4], which correlates highly with MDD [Reference Jylhä and Isometsä5]. Cross-sectional studies suggest a strong positive association between neuroticism and MDD [6–Reference Roelofs, Huibers, Peeters, Arntz and van Os8], whilst higher neuroticism prospectively associates with depression longitudinally [Reference Kendler, Neale, Kessler, Heath and Eaves2,9–Reference Fanous, Neale, Aggen and Kendler12], even when controlling for overlapping criteria [13–Reference Fergusson, Horwood and Lawton15] and demographics [Reference Kendler, Kuhn and Prescott16,Reference Neeleman, Ormel and Bijl17]. Whist the public health impacts of neuroticism are wide-ranging (for a comprehensive review see Lahey [Reference Lahey18]), neuroticism may be an indirect measure of later MDD risk, rather than the causative risk factor itself. Whereas MDD is often recurrent [Reference Hardeveld, Spijker, De Graaf, Nolen and Beekman19], neuroticism is a stable trait [Reference Conley20] suggesting that their correlation is unlikely to be substantially attributable to an effect of MDD on neuroticism.
General intelligence (g) is a latent construct theorized to explain the common observation that people who excel in one type of cognitive task tend to excel in others [Reference Humphreys21]. When reduced to a single factor (g) these correlations explain approximately 50% of the covariance between tests. Lower intelligence in early life has been found to be a risk factor for poor physical health [Reference Wraw, Deary, Gale and Der22] and early mortality in adulthood [Reference Deary, Weiss and Batty23,Reference Calvin, Deary, Fenton, Roberts, Der and Leckenby24]. Although research specifically regarding MDD is relatively sparse [Reference Gale, Deary, Boyle, Barefoot, Mortensen and Batty25], there is evidence to suggest that g is impaired in depression [Reference Marazziti, Consoli, Picchetti, Carlini and Faravelli26,Reference Sackeim, Steif, Georgotas and Cancro27] with longitudinal studies suggesting lower g in childhood or adolescence confers vulnerability to psychopathology in adulthood [28–Reference Zammit, Allebeck, David, Dalman, Hemmingsson and Lundberg31].
Psychological distress represents a cluster of emotional symptoms linked to depression [32–Reference Goldberg, Bridges, Duncan-Jones and Grayson34]. Although symptoms of distress are common in population samples [Reference Kessler and Wang35,Reference Singleton, Bumpstead, O’Brien, Lee and Meltzer36], they indicate only subthreshold mental health problems. With self-report measures of distress [Reference Goldberg and Hillier37,Reference Kroencke, Spitzer and Williams38] freely available in epidemiological research, their measurement provides greater detective power to make distinctions between syndrome and subthreshold symptoms. Longitudinal research suggests neuroticism has a strong, direct effect on psychological distress [Reference Ormel and Wohlfarth39]. Low childhood intelligence strongly associates with increased psychological distress in adulthood [Reference Gale, Hatch, Batty and Deary28,Reference Hatch, Jones, Kuh, Hardy, Wadsworth and Richards40], which may precede MDD onset [Reference Gulliver, Griffiths, Christensen and Brewer41]. However, this is not a universal observation, particularly in studies accounting for socioeconomic status (SES).
Intelligence and neuroticism may interact to influence indices of health. A longitudinal study of war veterans [Reference Weiss, Gale, Batty and Deary42] found high neuroticism and low cognitive ability were separate risk factors for mortality. Specifically, a 1-standard deviation increase in neuroticism resulted in a 33% increase in mortality; a 1-standard deviation decrease in intelligence associated with a 27% increase in mortality. An interaction (hazards ratio of 0.89) suggested that high neuroticism with low cognitive ability associates with high risk of poor health and reduced lifespan. Furthermore, high cognitive ability moderates the adverse effects of neuroticism on adjustment [Reference Leikas, Mäkinen, Lönnqvist and Verkasalo43]. Whether similar interactions exist with regard to their effects on depression remains unknown. No investigation has yet examined how intelligence and neuroticism influence risk for MDD and how they may moderate each other's associations in depression and psychological distress. Such an analysis may serve to clarify the mechanisms underlying MDD.
In this study, two large population-based cohorts were examined – Generation Scotland: Scottish Family Health Study (GS: SFHS) [Reference Smith, Campbell, Linksted, Fitzpatrick, Jackson and Kerr44,Reference Smith, Campbell, Blackwood, Connell, Connor and Deary45] and UK Biobank [Reference Allen, Sudlow, Downey, Peakman, Danesh and Elliot46,Reference Sudlow, Gallacher, Allen, Beral, Burton and Danesh47]. As previous studies suggest strong associations of neuroticism with risk of MDD [Reference Kendler, Neale, Kessler, Heath and Eaves2,Reference Jylhä and Isometsä5], the same effect was hypothesised here. We hypothesised that higher intelligence may reduce MDD risk by mitigating the adverse effects of neuroticism, similarly to the interaction identified for mortality [Reference Weiss, Gale, Batty and Deary42]. This reasoning transfers to psychological distress, hypothesising a positive association between neuroticism and psychological distress would be ameliorated by higher intelligence.
2. Method
2.1. GS:SFHS Overview
GS:SFHS is a family and population-based cohort recruited throughout Scotland between 2006 and 2011 [Reference Smith, Campbell, Linksted, Fitzpatrick, Jackson and Kerr44]. During clinic assessment, participants aged 18–98 (n = 24,084) provided clinical, cognitive and biological data. Full details are provided elsewhere [Reference Smith, Campbell, Linksted, Fitzpatrick, Jackson and Kerr44,Reference Smith, Campbell, Blackwood, Connell, Connor and Deary45]. The GS:SFHS sample is predominately female (59%), and generally healthier and wealthier than the Scottish population [Reference Smith, Campbell, Linksted, Fitzpatrick, Jackson and Kerr44]. This study includes 19,200 individuals with complete data of interest. Demographic information from this cohort is provided in Table 1 and within the Supplementary materials.
Table 1 Demographic, clinical, and cognitive characteristics of GS:SFHS and UK Biobank individuals in the current study.
* Significantly different from controls at P < 0.05.
Study assessments: during clinic assessment, participants were screened for lifetime history of MDD using a structured clinical interview [Reference First, Spitzer, Gibbon and Williams48]. Diagnosis of MDD follows DSM-IV criteria; if either symptoms of depressive mood or anhedonia are endorsed, a minimum of four further symptoms must also be endorsed. Clinical significance must be endorsed, too (ie., symptoms lasting nearly all day, every day for a minimum of two weeks). This study includes 2481 individuals meeting criteria for lifetime history of MDD (13%), and 16,719 non-MDD cases (87%).
Four cognitive tests measuring intelligence were administered during clinic assessment [Reference Smith, Campbell, Linksted, Fitzpatrick, Jackson and Kerr44,Reference Smith, Campbell, Blackwood, Connell, Connor and Deary45]. The Wechsler Digit Symbol Substitution Task [Reference Wechsler49] measured processing speed. One paragraph from The Weschler Logical Memory Test I & II [Reference Wechsler50] measured verbal declarative memory. The Verbal Fluency Test measured executive function [Reference Wechsler49] using phonemic lists of C, F and L. Vocabulary was measured with The Mill-Hill Vocabulary Test [Reference Raven51], using combined junior and senior synonyms. General intelligence (g) was extracted from these tests, as the first un-rotated principal component [Reference Marioni, Batty, Hayward, Kerr, Campbell and Hockling52], explaining 41% of the variance. Loadings for processing speed, vocabulary, verbal declarative memory and executive function were 0.57, 0.68, 0.63 and 0.69 respectively.
The self-reported Eysenck Personality Questionnaire Short Form-Revised (EPQ-SF) [Reference Eysenck53] measured neuroticism. Twenty-four questions assessed neuroticism and extraversion, with total scores on each subscale ranging from 0–12. Higher scores indicate higher levels of each trait. This scale has been concurrently validated [Reference Gow, Whiteman, Pattie and Deary54] with high reliability [Reference Eysenck, Eysenck and Barrett55].
Psychological distress was self-reported using the General Health Questionnaire (GHQ-28) [Reference Goldberg and Hillier37]. Twenty-eight items were scored from 0 (“not at all”) to 3 (“much more than usual”) with a total score ranging from 0–84. Higher scores indicate increased psychological distress.
The Scottish Index of Multiple Deprivation (SIMD) [Reference Payne and Abel56] is an official tool which identifies deprivation by combining different indicators (eg., income, crime) into a single index. The SIMD divides Scotland into 6505 small areas based on participant postcode, and assigns them a relative ranking from 1 (most deprived) to 6505 (least deprived).
2.2. UK Biobank Overview
UK Biobank is a population cohort recruited across the UK from 2006–2010. During an extensive baseline assessments [Reference Smith, Nicholl, Cullen, Martin, Ul-Haq and Evans57] participants aged 40–69 (n = 502,682) provided biological, physical, and touch-screen questionnaire measures of socio-demographics (e.g., age, sex), psychosocial factors (e.g., mental health), and cognitive function. UK Biobank represents a wide range of exposures typical within the UK population [Reference Biobank58], and has been described in detail elsewhere [Reference Allen, Sudlow, Downey, Peakman, Danesh and Elliot46,Reference Sudlow, Gallacher, Allen, Beral, Burton and Danesh47]. In this study, 147 individuals were removed from analysis due to participation in GS:SFHS. In total, 90,529 individuals with complete data of interest were included. Demographic information is provided in Table 1 and in the Supplementary materials.
Study assessments: between 2008–2010, a touch-screen questionnaire was added to the protocol to assess probable depression (n = 172,751) [Reference Biobank59]. Although depression was not assessed using a precise diagnostic tool, the classification followed a self-report approach within the guidelines of the ICD- 10 [Reference World Health Organisation60] and the DSM-IV [Reference American Psychiatric Association61]. Lifetime history of depression was assessed using items relating to the lifetime experience of depressive symptoms and help-seeking for mental health. A detailed description of how this phenotype was derived is provided elsewhere [Reference Smith, Nicholl, Cullen, Martin, Ul-Haq and Evans57]. This study included 30,127 (33%) individuals self-reporting lifetime history of depression, and 60,402 (67%) non-depressed cases.
Three novel cognitive tests were administered via touch-screen questionnaire measuring reaction time, verbal-numerical reasoning, and visual memory [Reference Smith, Nicholl, Cullen, Martin, Ul-Haq and Evans57]. A timed symbol matching test measured reaction time as the mean response time in ms over 12 trials; higher reaction times equate to poorer performance. Thirteen logic/reasoning-type questions assessed verbal-numerical reasoning - the total number of correct answers given within two-minutes was analysed. A visuo-spatial memory task measured the number of errors made when matching card pairs, higher scores reflect poorer cognitive function. From these tests, g was extracted as the first un-rotated principal component [Reference Marioni, Batty, Hayward, Kerr, Campbell and Hockling52], explaining 44% of the variance in scores. Loadings onto g were: −0.61 (verbal-numeric reasoning), 0.57 (visual memory), and 0.55 (reaction time).
Neuroticism was assessed using 12 questions from the Eysenck Personality Questionnaire Short Form-Revised (EPQ-SF) [Reference Eysenck53], administered via a touch-screen questionnaire. A total score from 0–12 was produced, with higher scores reflecting increasing neuroticism.
The first four questions of the Patient Health Questionnaire–9 (PHQ9) [Reference Kroencke, Spitzer and Williams38] were administered by touch-screen questionnaire to measure psychological distress. Responses on a scale from 0 (“Not at all”) to 3 (“Nearly every day”) were aggregated and a higher total score denoted higher levels of psychological distress.
The Townsend Deprivation Index [Reference Townsend62] is a census-based measure of deprivation, incorporating unemployment, non-car ownership, non-home ownership and household overcrowding into a single index. Small geographical areas based on postcode information are allocated Townsend Scores. Higher scores represent greater deprivation.
2.3. Statistical analysis
In GS:SFHS, the MCMCglmm package was used. The Markov Chain Monte Carlo estimator produces generalised linear mixed models for binary outcomes (using the “threshold” family with a probit link function). The threshold link is unique to MCMCglmm, and although produces very similar results to a logit function, threshold links most closely match the underlying assumptions of latent normal errors in pedigree-based mixed effect models [Reference Lynch and Wasl63]. MCMCglmm was essential to control for genetic relatedness of the sample, which was fitted as a random effect using an inverse pedigree matrix. Due to limitations within MCMCglmm with missing predictor variables, only complete data can be used. An interaction was fitted to estimate the moderating effect of g on the contribution of neuroticism to MDD. Another model examined this interaction while conditioning on deprivation. Regression coefficients are reported as Odds-Ratios. In a second set of analyses, GHQ was modelled as a normally distributed outcome variable. Neuroticism and GHQ were standardised to have a mean of zero and a standard deviation of 1. Age (standardised) and sex were used as fixed effects throughout.
In UK Biobank, generalized linear regression analyses were conducted as kinship need not be accounted for. The main effects of neuroticism and g were examined as predictors for self-reported depression. The interaction between neuroticism and g on depression was modelled. Another model examined this interaction while adjusting for deprivation. Generalized linear regressions were fitted with a logit link function and Odds-Ratios reported. A second set of analyses examined psychological distress (PHQ) using linear regression models. Neuroticism and PHQ were standardised to have a mean of zero and a standard deviation of one. Reaction time was log transformed due to a significantly positive skew. Visual memory was transformed with a log + 1 transformation because it was significantly skewed and zero-inflated. All regression analyses co-varied for age, and sex.
3. Results
3.1. GS:SFHS
As seen in Table 1, MDD cases were younger, predominately female, and had higher GHQ and neuroticism scores. No group differences were found in general intelligence; (t(3243.38) = −1.39, P =0.17, Cohen's d = 0.03). Group differences were found in processing speed and executive function. MDD cases were from less deprived areas; (t(3171.20) = 9.93, P = 2.20 × 10−16, Cohen's d = 0.22). Full statistical output can be found in the Supplementary materials.
3.1.1. Associations of neuroticism and g with MDD status
Higher neuroticism was strongly associated with increased risk for MDD. A 1SD-increase in neuroticism increased MDD risk by an odds-ratio of 3.61 (95% CIs = [3.28, 4.01], P< 1.00 × 10−4). Although no age effects were found, being female increased risk for MDD by an Odds-Ratio of 1.76 (95% CIs = [1.52, 2.03], P< 1.00 × 10−4). g had no independent effect on risk for MDD (OR = 1.02, 95% CIs = [0.99, 1.07], P = 0.53).
3.1.2. Interaction between neuroticism and g on MDD
No interaction was found between neuroticism and g (OR = 1.03, 95% CI = [0.98, 1.08], P = 0.32), see Fig. 1 and Table 2, even after co-varying for SIMD. However, the main effect of neuroticism was strongly associated with MDD risk (OR = 3.71, 95% CI = [3.37, 4.12], P< 1.00 × 10−4) whilst g was associated with a small increase in MDD risk (OR = 1.14, 95% CIs = [1.07, 1.20], P< 1.00 × 10−4). A main effect was found whereby higher deprivation confers risk for MDD (OR = 0.80, 95% CIs = [0.75, 0.86], P< 1.00 × 10−4).
Fig. 1 Predicted risk for MDD and self-reported depression from the interaction of neuroticism and g in both GS:SFHS and UK Biobank. Regression lines reflect the interaction at mean g (black line) and 2SD above (blue line) and below mean g (pink line).
Table 2 Results of a MCMC generalized linear mixed model from GS:SFHS predicting Odds-Ratios of MDD status, beta-coefficients for psychological distress (GHQ), P-value, upper and lower 95% confidence intervals and the Deviance Information Criterion AND results of a logistic regression from UK Biobank predicting Odds-Ratios for MDD status, beta-coefficients for psychological distress (PHQ), P-value, upper and lower 95% confidence intervals, the Akaike Information Criterion and adjusted R 2 value for the model.
3.1.3. Associations of neuroticism and g with psychological distress
Neuroticism was associated with increased psychological distress; a 1SD increase in neuroticism was associated with an increase in GHQ of β 0.52 (95% CIs = [0.50, 0.53], P< 1.00 × 10−4). A small inverse relationship was found whereby higher g was associated with decreased levels of psychological distress (β = −0.08, 95% CIs = [−0.09, −0.07], P< 1.00 × 10−4).
3.1.4. Interaction between neuroticism and g on psychological distress
A small interaction suggested higher g interacts with neuroticism to mitigate neuroticism's detrimental association on GHQ (β = −0.05, 95% CIs = [−0.06, −0.04], P< 1.00 × 10−4), see Fig. 2 and Table 2. This interaction remained after co-varying for deprivation.
Fig. 2 Psychological distress scores from the interaction of neuroticism and g in both GS:SFHS (GHQ) and UK Biobank. Regression lines reflect the interaction at mean g (black line) and 2SD above (blue line) and below mean g (pink line).
3.2. UK Biobank
As reported in Table 1, MDD cases were younger, predominately female, and had higher psychological distress (PHQ) and neuroticism scores than non-depressed cases. Significant differences were found in verbal-numerical reasoning (in which non-depressed cases performed better) and reaction time (in which depressed cases performed better). g was higher in depressed cases (t(61357) = −2.65, P = 8.12 × 10−3, Cohen's d = 0.02). Non-depressed cases had lower deprivation scores than depressed cases; (t(57110) = −20.08, P = 2.2 × 10−16, Cohen's d = 0.14), although this difference was small. See the Supplementary materials for full statistical output.
3.2.1. Associations of neuroticism and g with MDD status
Higher neuroticism was associated with increased likelihood of self-reported depression. For every 1SD increase in neuroticism, the odds for depression increased by 2.39 (95% CIs = [2.35, 2.43], P< 2.00 × 10−16). No main effects of g were found (OR = 1.00, 95% CIs = [0.99, 1.01], P = 0.86). Small effects of age and sex were found.
3.2.2. Interaction between neuroticism and g on MDD
A small interaction was found in which high levels of intelligence and neuroticism associate with reduced self-reported depression (OR = 0.96, 95% CIs = [0.95, 0.98], P =1.09 × 10−7), see Table 2 and Fig. 1. This interaction remained after co-varying for deprivation.
3.2.3. Associations of neuroticism and g with psychological distress
Neuroticism was moderately associated with increased levels of psychological distress. For every 1SD increase in neuroticism, PHQ increased by β 0.52 (95% confidence intervals = [0.51, 0.52], P < 2.00 × 10−16). g was associated with a small reduction in PHQ (β = −0.08, 95% CIs = [−0.08, −0.07], P < 2.00 × 10−16).
3.2.4. Interaction between neuroticism and g on psychological distress
A small interaction was found in which g moderates the detrimental effects of neuroticism on psychological distress (β = −0.02, 95% CIs = [−0.03, −0.02], P < 2.00 × 10−16), see Table 2 and Fig. 2. This interaction remained after co-varying for deprivation.
4. Discussion
The cross-sectional associations between neuroticism, general intelligence (g), MDD, self-reported depression, and psychological distress were examined in two large population based cohorts; GS:SFHS and UK Biobank. Neuroticism was strongly associated with increased risk for both MDD diagnosis and self-reported depression, replicating previous findings [Reference Chan, Goodwin and Harmer6,Reference Muris, Roelofs, Rassin, Franken and Mayer7]. Intelligence conferred no consistent independent effects but associated with an increased risk for depression once neuroticism was adjusted for. UK Biobank data suggest an interaction whereby higher g has a small effect in reducing the impact of neuroticism on self-reported depression. This interaction was small, both absolutely, and in comparison to the main effects of neuroticism. No such interaction was found in GS:SFHS using a clinical measure of MDD. However, across samples, the risk conferred by neuroticism after co-varying for g appears to be increased in terms of the absolute OR value when compared to basic models. Overall, results demonstrate an association whereby intelligence provides modest protection against the risk-conferring effects of neuroticism on self-reported depression, but not clinical MDD.
Consistent and replicable findings were found suggesting higher neuroticism associates with increased psychological distress, whereas higher intelligence associates with reduced psychological distress. A small interaction was found across samples such that lower distress associates with higher intelligence and lower neuroticism. Although these results are of small magnitude, they suggest an important interaction whereby higher g lessens the strength of the neuroticism-distress association.
This is the first study of intelligence's potential protective influence on MDD [Reference Robinson and Oishi64], self-reported depression, and psychological distress in high neuroticism individuals. Consistent with previous research the strong link between neuroticism with increased risk for depression and psychological distress was replicated with moderate effect sizes. Although longitudinal work suggests intelligence provides protection to mental health [Reference Gale, Deary, Boyle, Barefoot, Mortensen and Batty25,Reference Gale, Batty, Tynelius, Deary and Rasmussen29,Reference Maccabe30], we found g increased the risk for depression when adjusted for neuroticism. The magnitude of this risk was very small, however. Across cohorts, intelligence associated with decreased levels of psychological distress. A modest association of intelligence as a mitigating factor in reducing psychological distress in individuals with high neuroticism was found in both cohorts. Although this study suggests intelligence provides a protective function in self-reported depression and psychological distress (which mirrors previous research [Reference Calvin, Deary, Fenton, Roberts, Der and Leckenby24,Reference Weiss, Gale, Batty and Deary42,Reference Leikas, Mäkinen, Lönnqvist and Verkasalo43]), intelligence was not found to be protective against diagnosis of depression in those high in neuroticism.
It is unclear why intelligence associates with protection to risk for psychological distress, but not MDD. One supposition is that individuals with higher intelligence may be more likely to seek help, and therefore are more likely to receive a clinical diagnosis of depression. Another postulation could be that intelligence has an effect only during times of depressive episode. A state-dependent association of cognitive ability has been suggested in which variability in intelligence co-varies with depressive episode and remission (for a comprehensive review, see Sackeim and Steif [Reference Sackeim, Steif, Georgotas and Cancro27]). As such, subsequent investigations may benefit from addressing the same hypotheses examining individuals with current MDD in comparison to individuals in remission, and controls. Increased psychological distress is an established symptom of depression and often used in clinical diagnosis [Reference Beck, Rush and Shaw32,Reference Snaith33]. Goldberg [Reference Goldberg, Bridges, Duncan-Jones and Grayson34] described distress as representing the overall severity of depression and so it is likely that individuals scoring highly on measures of psychological distress may be more likely to self-report the disorder, irrespective of its clinical significance. However, we must be mindful of the complexities of causality; whilst it is likely that the neuroticism trait prospectively predicts later distress and self-reported depression, we cannot be certain that these factors are not manifestations of the same underlying risk.
Intelligence could be a marker of system integrity [Reference Deary65] in which increased intelligence circumvents negative mood biasing in individuals high in neuroticism that may lead to distress and disorder [Reference Hasler, Drevets, Manji and Charney66]. Alternatively, more intelligent individuals may be better able to employ successful coping mechanisms during times of distress: higher intelligence associates with increased resilience to adversity in children [Reference Fergusson, Horwood and Ridder67]. Research suggests that psychosocial factors are associated with resilience to mood disorders [Reference Garmezy, Masten and Tellegen68]. Proactive and psychosocial coping mechanisms may enable individuals decrease transient feelings of distress and to implement established, effective strategies learned from previous exposure to distress or depression [Reference LeDoux and Gorman69]. This possibility is consistent with the finding that whereas g and neuroticism interacted to associate with reduced psychological distress, the same interaction was not found in clinical MDD. It would be interesting to explore intelligence's influences on coping style [Reference Higgins and Endler70] and subsequent psychological distress and MDD diagnosis in future investigations. Intelligence may influence the adoption of specific coping strategies, and this could be a mediating factor in the ‘depressogenic’ process.
Some caveats merit comment. Different cognitive tasks were used to generate g across our samples. In GF:SFHS, pre-existing, standardized measures were used, whereas UK Biobank used bespoke cognitive tasks. Further replication utilising standardised measures would be beneficial. A second limitation is the differing MDD phenotypes used in each sample. In GS:SFHS, MDD was determined using a semi-structured interview [Reference First, Spitzer, Gibbon and Williams48], obtaining a robust MDD phenotype based on a standardised diagnostic tool. In UK Biobank, self-reported questionnaires were aggregated to form a depression phenotype; this data is not as comprehensive. Although it is of benefit to have conducted an independent replication within this study, the disparity in depression phenotypes may explain not only the difference in prevalence rates across samples, but also why an interaction was found in UK Biobank and not GS:SFHS. Thirdly, this investigation only examined neuroticism. Personality represents stable individual dispositions in emotional reactivity, behavioural tendencies, and cognitive styles [Reference Deary, Weiss and Batty23,Reference Roberts, Kuncel, Shiner, Caspi and Goldberg71], which may be moderated by intelligence in predicting mental health outcomes. Examining such associations between all major dimensions of personality in subsequent research is advised. As neuroticism and MDD share genetic aetiology [Reference Kendler, Neale, Kessler, Heath and Eaves2,Reference Jardine, Martin and Henderson3], causality cannot be inferred here, although the associations reported do make a significant contribution to the literature. Because neuroticism is a stable trait and MDD is a disease with a given age of onset, we can use neuroticism to predict an individual's risk for depression, without needing to infer causality.
In conclusion, this study fails to demonstrate that intelligence confers protection to clinical MDD in those with high neuroticism. However, in both samples, a modest interaction was found in which higher intelligence appears to ameliorate the detrimental association between neuroticism and psychological distress. It would be useful to determine this relationship prospectively in a sample where incident cases of MDD can be identified. An important corollary of this work may inform risk and resilience mechanisms in MDD. Future studies to disentangle the mechanisms driving depression are an important next step in further elucidating the aetiology of the disorder.
Author contributions
L.N. wrote the manuscript text and prepared all tables and figures. A.M. was the main supervisor for the project, with co-supervision provided by S.R. and S.C. M.J. aided in the statistical analysis. L.N. and E.H. contributed to the data entry for the project. D.K., D.P., I.D., C.G. and D.B. reviewed the manuscript.
Funding
The authors declare that they have no competing interest.
Disclosure of interest
The authors have not supplied their declaration of competing interest.
Acknowledgements
GS:SFHS received ethical approval from the NHS Tayside Committee on Medical Research Ethics (REC Reference Number: 05/S1401/89). Methods cited are in accordance with the relevant guidelines and regulations. Written consent for the use of data was obtained from all participants. Genotyping of the GS:SFHS samples was carried out by the Genetics Core Laboratory at the Wellcome Trust Clinical Research Facility, Edinburgh, Scotland and was funded by the Medical Research Council UK and the Wellcome Trust (Wellcome Trust Strategic Award “STratifying Resilience and Depression Longitudinally” (STRADL) Reference 104036/Z/14/Z). This investigation was supported by Dr Mortimer and Theresa Sackler Foundation in addition to core support being provided by the Chief Scientist Office of the Scottish Government Health Directorates [CZD/16/6] and the Scottish Funding Council [HR03006].
UK Biobank received ethical approval from the North West Multicentre Research Ethics Committee (REC Reference Number: 11/NW/0382), and all methods were conducted in accordance with the relevant guidelines. Written consent for the use of data was obtained from all participants. This study was conducted under UK Biobank application 4844 “Stratifying Resilience and Depression Longitudinally” (PI Andrew McIntosh).
Appendix A. Supplementary data
Supplementary data associated with this article can be found, in the online version, at http://dx.doi.org/10.1016/j.eurpsy.2016.12.012.
Open access
Comments
No Comments have been published for this article.