Interpretation of present findings
Given the insidious nature of dementia development [
11,
12], it cannot be ruled out that the associations identified in the present study are results of reverse causality (i.e. were early changes due to dementia impacts mental distress), even with a follow-up of between 11 and 27 years. There is evidence that pathology related to Alzheimer’s disease starts decades before clinical diagnosis [
53,
54]. In addition, although we were able to include several potentially important covariates in our analyses, residual confounding cannot be ruled out. Our findings are, however, in line with two previous studies, which both found a prospective association between midlife mental distress and dementia in women [
38,
39]. Also, another study found that proneness to mental distress was associated with increased risk for Alzheimer’s disease in old age [
55], and a fourth study found an association between mental distress and dementia deaths [
40]. However, the results from our study are not consistent with those from a retrospective cohort study where self-reported depressive symptoms in late-life were more strongly associated with increased risk of dementia than symptoms in mid-life [
31]. In contrast, our findings indicate that mental distress in early mid-life is more strongly associated with an increased risk for dementia, than mental distress measured in late mid-life. Participants aged >60 years were excluded from our study, as it was envisaged to be difficult to differentiate early symptoms of dementia from symptoms of mental distress in this group.
A number of neurological and cardiovascular mechanisms could explain the observed association between mental distress and risk of dementia, including structural and functional hippocampal damage [
55], and other negative effects on the hypothalamic-pituitary-adrenal axis [
38,
55], as well as hypertension as a potential mediator between mental distress and dementia [
38]. Studies have indicated that there may be a two-way/bidirectional relationship between cardiovascular disease and depression [
56,
57], and this may also be true for mental distress in general. Further, midlife mental distress may be a result of increased vulnerability to stressors and dysfunctional coping strategies related to previous negative experience or genetic influences, which also constitute risk factors for dementia [
39,
58-
60]. The finding from the sensitivity analysis that “Nervousness” was the only item independently associated with dementia after adjustment for age and gender, may suggest that this symptom of anxiety was the primary driver for dementia risk in our sample. Although this finding should be interpreted with caution, it does lend support to previous findings suggesting cortisol and exitotoxicity as a mechanism linking stress with dementia [
61,
62].
Associations between midlife mental distress and dementia may also be due to mental distress being a marker of rather than a risk factor for dementia [
38]. Even for longer follow-up periods, there is evidence suggesting that the prodromal phase of dementia in some cases may be present several decades before dementia is manifest [
54]. However, our age-stratified results, in which we found increased risk for dementia in the younger rather than the older group who reported mental distress, is not consistent with this. This may, however, be due to important differences in the characteristics of early (before 65 years of age) versus late-onset dementia (65 years or more), such as the disease progression, and symptom profile. During the past decades, it has become more common to differentiate between early- and late-onset dementia, each potentially associated with different epidemiology, genetics, clinical presentation and disease progression [
63,
64].
The differential association between mental distress and risk of dementia by age could potentially be related to differences in types and/or timing of onset of dementia. For example, midlife mental distress may be a more potent risk factor for earlier onset dementia. This notion may be supported by our finding that the younger group had a symptom onset on average 16 years earlier than the older group, while the time-to-dementia-onset was similar. The impact of mental distress on the development of dementia, if causal, might be affected by the duration of the distress, and as such, it may be that only prolonged mental distress has an effect on the risk of dementia [
38,
55]. In support of the greater impact of experiencing mental distress at younger age, early onset depressive disorder have been found to be a greater illness burden across a range of indicators compared to later onset depression [
65]. Therefore, it is possible that mental distress has to be both present in early adult life and sustained into late adulthood and early old age [
66]. Another explanation may be that the increased risk observed in the younger age group is a function of the interval between exposure and dementia, as the younger participants are less likely to have died during follow-up.
Strengths and limitations
The present study has several strengths. First, it employed a linkage between a large population-based study with a high overall participation rate (92.3% for those aged 30–60 at participation) [
42] and a comprehensively ascertained dementia registry in the same catchment area [
47]. Also, the participation rate in HUNT1 for patients diagnosed with dementia in Nord-Trøndelag County was 86% [
47], indicating that the linkage between HUNT1 and HMS can be considered valid for research purposes. Second, the linkage enabled a follow-up ranging from 11 to 27 years, decreasing the likelihood that self-reported mental distress is a marker of early stages of dementia. Third, the established dementia registry used multiple sources of information to detect and validate the dementia diagnoses, securing a high degree of certainty/reliability related to dementia diagnosis. Finally, we were able to include a range of potentially relevant confounders for the main analysis, as well as investigating the potential two-way interaction between age and gender, and mental distress on later dementia.
Some limitations are pertinent for the interpretation of the results from the present study, and should be borne in mind. First, the measure of mental distress employed in this study is not a well-validated instrument, and the psychometric properties are somewhat unclear. The range of symptoms related to mental distress that was assessed was limited, and therefore the phenomenon of mental distress as measured may be somewhat faceted and have lower specificity than validated instruments. The ADI-4, has, however, been used in several other studies and is considered as an acceptable, but crude measure of symptoms of mental distress [
44,
45]. Aware of this issue, we also investigated the separate items, three of which were positively associated with dementia, as well as investigating the main association using SEM-analysis with “mental distress” as a continuous latent construct. The internal consistency of the instrument was good, and an exploratory factor analysis indicated that all four items measured the same latent construct. In addition, any deficiencies in the measurement’s accuracy will have obscured rather than exaggerated associations of interest.
Second, the limited number of dementia cases decreased the precision of our estimates, restricting the ability to adjust for potential confounders in the age-stratified analyses. There were also too few cases to investigate meaningfully the impact of mid-life mental distress on different sub-types of dementia. It is, however, difficult to discern the different types of dementia conditions, and in many cases the discrimination is poor, even between Alzheimer’s and vascular dementia [
38,
67,
68].
Third, the dementia cases which are included in the dementia registry cannot be assumed to be representative of all dementia cases residing in Nord-Trøndelag; for example, they may well be more severe [
47] than the community cases not identified [
69]. Also, individuals institutionalized (e.g. at mental health institutions) were not invited to participate. Although not all people suffering from dementia in Nord-Trøndelag County during the HMS study period 1995–2011 have been identified (for example those not yet diagnosed or diagnosed by other health care workers in primary care) [
47], the risk estimates should only be moderately deflated, since undetected dementia cases are likely to constitute only a small fraction of the subjects considered non-cases, with misclassification unlikely to have substantial effects on observed differences in risk [
47]. In addition, false positive dementia cases included in the HMS study are unlikely to have been common enough to exert an influence.
Fourth, onset of symptoms was only available for a sub-sample of the dementia cases, and was assessed in retrospect: At inclusion in the study, a next of kin was asked to report retrospectively when they noticed the first symptoms of dementia. Based on this information, the age of onset of dementia was estimated. Obviously, this information could be biased, and serves as only a crude indication for time of onset. Also, the date of diagnosis was not available, precluding any event-related analysis.
Fifth, there were few dementia cases (n = 48) among those aged under 45 years at baseline, and associations of interest in this group were thus imprecise, with wide confidence intervals. The small number of cases in the younger group also precluded any meaningful adjustment for potential confounders beyond age and gender.
Sixth, we are unable to assess and control for differential attrition, migration or mortality during follow-up, an issue which could lead to bias. There is, however, little reason to believe that these factors would lead to an inflation of the associations of interest.
Finally, there are a number of risk factors active between ascertainment of exposure and development of dementia – and these may potentially modify the association between mental distress and later dementia. The intervening risk factors could, however, reduce the likelihood of finding an association in a study with this long follow-up and controlling for subsequent factors might lead to an over-adjustment [
70].