The British Journal of Psychiatry (2022)

221, 410–416. doi: 10.1192/bjp.2021.217

Association between a history of clinical

depression and dementia, and the role of
sociodemographic factors: population-based
cohort study
Kaarina Korhonen, Lasse Tarkiainen, Taina Leinonen, Elina Einiö and Pekka Martikainen

Background Results
Depression is associated with an increased dementia risk, but A history of depression was related to an adjusted hazard ratio of
the nature of the association in the long-term remains unre- 1.27 (95% CI 1.23–1.31) for dementia in the conventional Cox
solved, and the role of sociodemographic factors mainly model and of 1.55 (95% CI 1.09–2.20) in the sibling fixed-effects
unexplored. model. Depression was related to an elevated dementia risk
similarly across all levels of education (test for interaction,
Aims P = 0.84), but the association was weaker for the widowed than
for the married (P = 0.003), and stronger for men than women
To assess whether a history of clinical depression is associated
(P = 0.006). The excess risk among men attenuated following
with dementia in later life, controlling for observed sociodemo-
covariate adjustment (P = 0.10).
graphic factors and unobserved factors shared by siblings, and
to test whether gender, educational level and marital status Discussion
modify the association.
This study shows that a history of depression is consistently
associated with later-life dementia risk. The results support the
Method hypothesis that depression is an aetiological risk factor for
We conducted a national cohort study of 1 616 321 individuals dementia.
aged 65 years or older between 2001 and 2018 using adminis-
trative healthcare data. A history of depression was ascertained Keywords
from the national hospital register in the period 15–30 years prior Dementia; depressive disorders; epidemiology; socioeconomic
to dementia follow-up. We used conventional and sibling fixed- status; life course.
effects Cox regression models to analyse the association
between a history of depression, sociodemographic factors and Copyright and usage
dementia. © The Author(s), 2022. Published by Cambridge University Press
on behalf of the Royal College of Psychiatrists.

Background Although many of the previous studies have controlled for

Depression and dementia are highly disabling disorders contribut- sociodemographic factors, the modifying effects remain mainly
ing to the global disease burden and challenging the sustainability unexplored. Prior studies have investigated gender and education
of healthcare systems.1,2 Depressive symptoms commonly coexist modification in prodromal depression8,15 but, to the best of our
with dementia,3 and recently, late-life depression was estimated to knowledge, no previous study has explicitly analysed the role of
contribute to 4% of all worldwide dementia cases.4 Despite system- sociodemographic factors in the association between depression
atic review evidence on depression doubling the risk of dementia and dementia in the long term. Identifying susceptible population
onset,5 the question about whether depression is an aetiological subgroups may elucidate potential mechanisms linking depression
risk factor for dementia remains unresolved. This is because and dementia and help target dementia interventions effectively. In
late-life depression or depressive symptoms in close proximity to Finland, depression is among the leading causes of disability in the
dementia onset may in fact be prodromal features of dementia working-age population and about 10% of adults have experienced
itself.6,7 An important limitation of previous studies is that they depression within the past 12 months.16 At the same time, the
have not considered the temporal dimension in the association number of people living with dementia is rapidly increasing, demen-
between depression and dementia onset. Only a few studies have tia already being the third leading cause of death in Finland. The high
analysed the association separately for earlier- and late-life depres- prevalence of depression and the growing elderly population suggest
sion.8 Although some such studies reported a consistent associ- an urgent need to enhance understanding and awareness of the
ation between depression and dementia as long as 20 years association between depression and dementia.
apart,9–12 other studies only observed an association when depres-
sion occurred less than about 10 years before dementia diagnosis6,7
or after the age of 45 or 50 years.13,14 The inconsistency in findings Aims of study
may relate to differences in measurement of depression, follow-up We conducted a register-based cohort study on Finnish older adults
times and characteristics of the study populations. Furthermore, to estimate the association between a history of clinical depression
only a few studies9,14 have been able to control for childhood and dementia, and to assess whether sociodemographic factors
family background to account for the strong familial aggregation including gender, educational level and marital status modify the
of both depression and dementia. For example, the low socio- association. We also estimated sibling fixed-effects models control-
economic position of the childhood family may increase the risk ling for unobserved early-life familial conditions and genetic factors
of both depression and dementia. shared by siblings that might confound the association between

410
https://doi.org/10.1192/bjp.2021.217 Published online by Cambridge University Press
 Association between a history of clinical depression and dementia

depression and dementia. As the preclinical and prodromal stages of Information on educational level and marital status was collected
dementia generally begin 10–15 years before the clinical stage,17 we from the period 15 to 30 years before baseline. Education was indi-
used information on depression diagnoses observed in hospital regis- cated as the highest achieved qualification, categorised as tertiary
ters in the period from 15 to 30 years before the follow-up for demen- (generally ≥13 years of education), secondary (10–12 years) and
tia in order to reduce bias arising from depressive symptoms basic education or less (up to 9 years). Marital status was classified
reflecting preclinical and prodromal stages of dementia. as married, divorced, widowed and never married. If education or
marital status changed during the observation period, we considered
the most recent status. We further included other medical conditions
Method as covariates to indicate vascular risk factors (for example smoking,
excessive alcohol use, diabetes) and cardiovascular diseases (CVDs)
Data and variables that are shown to be comorbid to depression23 and to influence
We used population register data on all Finns born between 1900 and dementia risk.24,25 These conditions were measured from 15 to 30
1950 obtained from Statistics Finland. The cohort was followed, years before baseline from the hospital care register, and included
through record linkage using personal identification codes assigned alcohol-related diseases and accidental poisoning by alcohol,
to all permanent residents, for incident dementia at the age of 65 chronic obstructive pulmonary disease (COPD) or asthma, diabetes,
years and above in administrative health registers from 2001 coronary heart disease (CHD), and other non-stroke CVDs. Stroke
through 2018. The baseline for follow-up thus varied from year was excluded because of the direct short-term effect on (post-
2000 to 2015 according to the year of birth. Information on sociode- stroke) dementia (for ICD-codes see Supplementary Table 2).
mographic characteristics was obtained from population censuses Because depression history and the covariates could only be
conducted in 1970, 1975, 1980 and 1985 and from the population defined for Finnish residents, we excluded those living abroad over
register from 1987 to 2018. The sibling analysis was based on a the period 15 to 30 years prior to baseline (n = 14 914).
10% household sample from the 1950 Finnish population census
that has been linked to subsequent census records and the population Statistical analyses
register. Children living in the same family at the age of 0 to 15 years
in 1950 were identified as siblings by means of unique family identi- We used Cox proportional hazards regression to estimate hazard
fiers. The family is defined based on the youngest generation in the ratios (HRs) and 95% CI for the association between a history of
household dwelling-unit, and thus siblings could only be identified clinical depression and dementia between 1 January 2001 and 31
if they did not have children of their own. We therefore limited the December 2018. Cohorts turning 65 years in year 2001 or later
age range to 15 years, when childbearing was still rare. The sibling entered the analysis on the 1 January following their 65th birthday.
subsample thus included cohorts born between 1935 and 1950. Attained age in years was used as the underlying timescale in the
We identified dementia using the medication reimbursement analyses. Individuals were censored on the date of death, at the
register of the Social Insurance Institution of Finland and the end of the year preceding emigration or at the end of 2018, which-
hospital care register of the Finnish Institute for Health and ever came first.
Welfare. Thus our definition of dementia only included people on We first used conventional Cox models to estimate the associ-
dementia medication or receiving hospital care. We collected the ation between a history of clinical depression and dementia, con-
beginning (month and year) of entitlement to special state reim- trolling for observed characteristics. In addition to underlying
bursement of antidementia medication, and the dates of state-reim- attained age, model 1 adjusted for gender and calendar year.
bursed purchases of antidementia medication using the World Model 2 accounted additionally for education and marital status,
Health Organization Anatomical Therapeutic Chemical (ATC) and model 3 further adjusted for comorbid medical conditions.
code N06D. Specialised out-patient care and in-patient hospital Second, we used Cox regression with sibling fixed-effects by assum-
admissions with a dementia diagnosis were collected with World ing a separate baseline hazard for each childhood family (n = 23
Health Organization ICD-1018 codes F00–03, F05.1 and G30. The 626) to control for all time-invariant characteristics shared by sib-
earliest entry in any of these registers was set as the date of dementia lings.26 We further controlled for the same observed characteristics
incidence. These data sources present good sensitivity and high pre- as in the conventional Cox models. Third, modification by gender,
cision for dementia diagnosis,19 and the age-specific incidence rates educational level and marital status was analysed in the full cohort,
observed in our data (Supplementary Table 1 available at https://doi. with basic (model 1) and full adjustments (model 3). The signifi-
org/10.1192/bjp.2021.217) are consistent with reports for screened cance of interaction was analysed using the likelihood ratio test.
community samples.20 All analyses were performed with Stata 16.0.
To restrict the study population to initially dementia-free
individuals, we excluded those with pre-baseline reimbursements Sensitivity analyses
of antidementia medication (reimbursement became available in As selection for health and survival may bias the estimates in studies
1999) or hospital admissions with a dementia diagnosis indicated of older people,24,27 we conducted age-stratified analyses to assess
by ICD-10 codes in 1996–2000 and ICD-921 codes 290, 2912A, whether the associations changed with increasing age at study inclu-
2928C, 2941A, 3310, 3311 and 4378A in 1987–1995 (n = 24 543). sion. We used three groups: cohorts born in 1935–1950 (aged 65
Because dementia is among the main indications of institutional years at baseline), cohorts born in 1920–1934 (66–80 years) and
residence, we further restricted the study population to those living cohorts born in 1900–1919 (81–100 years).
in private households at baseline (n = 49 384 excluded).
Data on clinical depression 15 to 30 years before baseline was
collected from the hospital care register. In-patient care episodes Ethics of research
with a depression diagnosis were identified using ICD-822 codes The authors assert that all procedures contributing to this work
2960, 2980, 3004 and 3011 in 1971–1986, ICD-9 codes 2961, comply with the ethical standards of the relevant national and insti-
2968, 3004, 3009 and 3090 in 1987–1995, and ICD-10 codes F32, tutional committees on human experimentation and with the
F33, F34.1 and F38.1 in 1996–2000. Individuals with at least one Helsinki Declaration of 1975, as revised in 2008. All procedures
care episode with a depression diagnosis were classified as having involving human participants were approved by Statistics Finland
a history of clinical depression. Board of Ethics (permit no. TK-53-1490-18). Participant consent

411
https://doi.org/10.1192/bjp.2021.217 Published online by Cambridge University Press
 Korhonen et al

was not required as administrative register data can be used for with dementia. Table 3 presents the results from the sibling fixed-
scientific purposes under the Personal Data Act and the Statistics effects models. Adjusting for the unobserved factors shared by sib-
Act. Statistics Finland pseudonymised the data prior to providing lings, the association between a history of clinical depression and
it to researchers. dementia was stronger than in the conventional model but the con-
fidence interval also became wider (HR = 1.65, 95% CI 1.17–2.33;
model 1). In addition, having never married (HR = 1.21; 95% CI
Results 1.03–1.42), alcohol-related conditions (HR = 1.82; 95% CI 1.27–
2.61) and diabetes (HR = 3.40; 95% CI 1.97–5.88) were associated
In the full cohort of 1 616 321 individuals (55.9% female), we iden- with dementia. In the model further adjusting for observed educa-
tified 23 959 (1.5%) individuals with a history of clinical depression. tion and marital status (model 2), the excess hazard related to
In the subsample of 63 445 siblings (51.3% female), depression had depression slightly attenuated (HR = 1.60; 95% CI 1.13–2.27).
been diagnosed for 948 (1.5%) individuals. Mean (s.d.) baseline age Adjusting additionally for observed medical conditions (model 3),
of the full cohort was 68.9 (6.2) years, and the sibling subsample was depression was associated with a HR of 1.55 (95% CI 1.09–2.20)
by definition 65 years at baseline. Table 1 shows the frequencies and for dementia.
descriptive values of sociodemographic and health variables. Compared with individuals without a history of clinical depres-
During 14 725 473 person-years at risk (mean follow-up time sion, the excess hazard of dementia related to depression was
9.1 years), 274 817 individuals were identified with dementia in stronger among men (HR = 1.41, 95% CI 1.34–1.49) than women
the full cohort. (HR = 1.29, 95% CI 1.24–1.34; test for interaction x2(1) = 7.72,
Table 2 displays the results from the conventional Cox regres- P = 0.006) (Fig. 1(a)). Adjustment for other sociodemographic
sion for the full cohort. In the minimally adjusted model 1, depres- and medical conditions, however, attenuated the association to
sion was associated with a significantly increased hazard of the same level for both genders (x2(1) = 2.69, P = 0.10). Further
dementia (HR = 1.32, 95% CI 1.28–1.36). Also, secondary (HR = inspection revealed that alcohol-related conditions in particular
1.11, 95% CI 1.10–1.13) and basic education (HR = 1.18, 95% CI explained the stronger association among men (results not
1.16–1.19), and all unmarried statuses (HR = 1.12; 95% CI 1.10– shown). A history of clinical depression was associated with an ele-
1.13 for divorced; HR = 1.04, 95% CI 1.03–1.06 for widowed; vated hazard of dementia similarly across all levels of education
HR = 1.05, 95% CI 1.04–1.07 for never married) were associated (x2(2) = 0.34; P = 0.84) (Fig. 1(b)). By contrast, the association
with dementia at the 95% confidence level. between depression and dementia differed by marital status
The hazard of dementia was also elevated for those with alcohol- (x2(3) = 13.97; P = 0.003) (Fig. 1(c)). Specifically, the association
related conditions (HR = 1.33, 95% CI 1.28–1.37), COPD/asthma was weaker among the widowed (HR = 1.16, 95% CI 1.07–1.26)
(HR = 1.07, 95% CI 1.04–1.09), diabetes (HR = 1.52, 95% CI 1.45– than the married (HR = 1.36, 95% CI 1.31–1.41). The moderation
1.60), CHD (HR = 1.10, 95% CI 1.08–1.12) and other CVDs (HR was not explained by the covariates (x2(3) = 13.35, P = 0.004).
= 1.04, 95% CI 1.03–1.05). In model 2, adjusting for education
and marital status, the associations of depression, education and
marital status with dementia remained substantially unchanged. Sensitivity analyses
In the fully adjusted model further adjusting for medical conditions Age-stratified analyses indicated that our main results primarily
(model 3), depression was associated with a 27% (95% CI 1.23–1.31) reflect associations observed in cohorts born in 1920–1934, who
excess hazard of dementia. experienced most of the observed incident dementia cases
In the sibling subsample, 4508 individuals during 546 129 (Supplementary Table 3). Among cohorts born in 1935–1950, the
person-years at risk (mean follow-up time 8.6 years) were identified association between depression and dementia (HR = 1.53, 95% CI

Table 1 Descriptive characteristics of the full cohort and the sibling subsample

Study cohort
Full cohort Sibling subsample
Variable History of clinical depression History of clinical depression
Yes (n = 23 959) No (n = 1 592 362) Yes (n = 948) No (n = 62 497)
Age, years: mean (s.d.) 68.5 (5.6) 68.9 (6.2) 65.0 (−) 65.0 (−)
Gender, n (%)
Men 8865 (37.0) 704 429 (44.2) 413 (43.6) 30 454 (48.7)
Women 15 094 (63.0) 887 933 (55.8) 535 (56.4) 32 043 (51.3)
Education, n (%)
Tertiary 3415 (14.3) 302 677 (19.0) 175 (18.5) 14 872 (23.8)
Secondary 5855 (24.4) 374 279 (23.5) 302 (31.9) 19 107 (30.6)
Basic or less 14 689 (61.3) 915 406 (57.5) 471 (49.7) 28 518 (45.6)
Marital status, n (%)
Married 13 128 (54.8) 1 134 560 (71.3) 528 (55.7) 45 513 (72.8)
Divorced 5579 (23.3) 174 547 (11.0) 234 (24.7) 8344 (13.4)
Widowed 2229 (9.3) 120 902 (7.6) 45 (4.7) 1759 (2.8)
Never married 3023 (12.6) 162 353 (10.2) 141 (14.9) 6881 (11.0)
Medical conditions, n (%)
Alcohol relateda 2994 (12.5) 20 785 (1.3) 135 (14.2) 916 (1.5)
COPD/asthma 1004 (4.2) 32 244 (2.0) 30 (3.2) 1133 (1.8)
Diabetes 361 (1.5) 10 972 (0.7) 14 (1.5) 462 (0.7)
CHD 1420 (5.9) 45 525 (2.9) 34 (3.6) 993 (1.6)
Other CVDs 4420 (18.5) 213 597 (13.4) 147 (15.5) 7586 (12.1)
COPD, chronic obstructive pulmonary disease; CHD, coronary heart disease; CVD, cardiovascular disease.
a. Alcohol-related diseases and accidental poisoning by alcohol.

412
https://doi.org/10.1192/bjp.2021.217 Published online by Cambridge University Press
 Association between a history of clinical depression and dementia

Table 2 Hazard ratios for dementia by a history of clinical depression Discussion

and covariates in the conventional Cox regression on the full cohort

Exposure Hazard ratio (95% CI) In a large register-based cohort study of 1 616 321 individuals, we
variable Model 1a Model 2b Model 3b found an elevated risk of dementia related to having been diagnosed
History of clinical with depression 15 to 30 years before baseline. Cohort members
depression with a history of clinical depression had around a 30% higher
No 1 (Reference) 1 (Reference) 1 (Reference) hazard of developing dementia compared with those with no
Yes 1.32 (1.28–1.36) 1.30 (1.27–1.34) 1.27 (1.23–1.31)
history of depression, even after accounting for differences in edu-
Education
Tertiary 1 (Reference) 1 (Reference) 1 (Reference)
cational level, marital status and several comorbid medical condi-
Secondary 1.11 (1.10–1.13) 1.11 (1.09–1.13) 1.11 (1.09–1.12) tions. A similar association was found when we compared siblings
Basic 1.18 (1.16–1.19) 1.17 (1.16–1.19) 1.17 (1.16–1.18) to each other, adjusting for unobserved characteristics shared by
Marital status siblings in addition to the observed characteristics. Our results
Married 1 (Reference) 1 (Reference) 1 (Reference) thus align with the hypothesis that depression is an aetiological
Divorced 1.12 (1.10–1.13) 1.11 (1.09–1.12) 1.10 (1.09–1.12)
risk factor for dementia. Furthermore, our results show that depres-
Widowed 1.04 (1.03–1.06) 1.04 (1.02–1.05) 1.03 (1.02–1.05)
Never married 1.05 (1.04–1.07) 1.06 (1.04–1.07) 1.06 (1.04–1.07)
sion is related to an elevated dementia risk similarly across all levels
Medical of education, but the association is generally stronger for men than
conditionsc women, and weaker for the widowed than for the married.
Alcohol 1.33 (1.28–1.37) – 1.25 (1.21–1.30) To our knowledge, this is the first study to explicitly assess
relatedd whether sociodemographic factors modify the long-term associ-
COPD/asthma 1.07 (1.04–1.09) – 1.05 (1.02–1.08) ation between depression and dementia. Our results show that the
Diabetes 1.52 (1.45–1.60) – 1.50 (1.43–1.58)
association was not specific to any of the assessed sociodemographic
CHD 1.10 (1.08–1.12) – 1.08 (1.06–1.10)
Other CVDs 1.04 (1.03–1.05) – 1.04 (1.02–1.05) subpopulations, although the strength of association varied in terms
of gender and marital status. The stronger association among men
COPD, chronic obstructive pulmonary disease; CHD, coronary heart disease; CVD, car-
diovascular disease. attenuated, however, to a statistically non-significant level following
a. Each exposure variable modelled separately; adjusted for gender and calendar year. covariate adjustment. Our supplementary analyses indicated that
b. Exposure variables mutually adjusted; adjusted for gender and calendar year.
c. Reference: not having this particular condition. the attenuation was mainly the result of adjustment for alcohol-
d. Alcohol-related diseases and accidental poisoning by alcohol.
related conditions. Alcohol use disorder is frequently comorbid
with major depression, especially among men,28 and heavy
alcohol use is associated with an increased risk of dementia.25
1.44–1.61; Supplementary Table 4) reflected estimates obtained in Nevertheless, our sensitivity analyses suggested that in younger
the sibling analysis (Table 3). In these cohorts, depression was ten- cohorts born in 1935–1950, the association between a history of
tatively more strongly associated with dementia among women than depression and dementia may in fact be stronger among women
men (Supplementary Figure 1). Furthermore, the association was than men, despite the higher prevalence of alcohol-related condi-
stronger for those with a basic education compared with those tions among men. Because the different cohorts were followed for
who were more highly educated (Supplementary Figure 2) but no incident dementia partly at different ages, these analyses do not dis-
differences emerged by marital status (Supplementary Figure 3). close whether the stronger association among women in these
Among the oldest cohorts born in 1900–1919, a history of clinical cohorts reflects a cohort effect (the stronger association among
depression was not associated with dementia (Supplementary women will persist with age) or an age effect (the stronger
Table 5).

Table 3 Hazard ratios for dementia by a history of clinical depression and covariates in the sibling fixed-effects Cox regression on the sibling subsample

Hazard ratio (95% CI)

Exposure variable Model 1a Model 2b Model 3b
History of clinical depression
No 1 (Reference) 1 (Reference) 1 (Reference)
Yes 1.65 (1.17−2.33) 1.60 (1.13−2.27) 1.55 (1.09−2.20)
Education
Tertiary 1 (Reference) 1 (Reference) 1 (Reference)
Secondary 1.05 (0.89−1.24) 1.03 (0.87−1.22) 1.02 (0.86−1.20)
Basic 1.11 (0.94−1.31) 1.09 (0.92−1.29) 1.08 (0.92−1.28)
Marital status
Married 1 (Reference) 1 (Reference) 1 (Reference)
Divorced 1.12 (0.97−1.31) 1.11 (0.96−1.29) 1.09 (0.94−1.27)
Widowed 1.17 (0.90−1.51) 1.16 (0.90−1.50) 1.14 (0.88−1.47)
Never married 1.21 (1.03−1.42) 1.20 (1.02−1.41) 1.19 (1.01−1.40)
Medical conditionsc
Alcohol-relatedd 1.82 (1.27−2.61) - 1.63 (1.13−2.35)
COPD/asthma 0.92 (0.65−1.30) - 0.88 (0.62−1.26)
Diabetes 3.40 (1.97−5.88) - 3.28 (1.89−5.71)
CHD 1.31 (0.90−1.90) - 1.23 (0.85−1.80)
Other CVDs 1.05 (0.91−1.21) - 1.04 (0.90−1.20)
COPD, chronic obstructive pulmonary disease; CHD, coronary heart disease; CVD, cardiovascular disease.
a. Each exposure variable modelled separately; adjusted for gender and calendar year.
b. Exposure variables mutually adjusted; adjusted for gender and calendar year.
c. Reference: not having this particular condition.
d. Alcohol-related diseases and accidental poisoning by alcohol.

413
https://doi.org/10.1192/bjp.2021.217 Published online by Cambridge University Press
 Korhonen et al

(a) and susceptible to developing dementia do so before reaching the

1.60 2(1) = 7.72; P = 0.006 2(1) = 2.69; P = 0.10 oldest old age.
These findings add to the existing evidence base showing a long-
1.50 term association between depression and dementia.9–12 Our ana-
1.40 1.41 lysis on siblings suggest that the association between depression
Hazard ratio

and dementia is not likely to arise from early-life familial back-

1.30 1.32
1.29 ground such as childhood adversity or genetic factors shared by sib-
1.25 lings. Similar findings have previously been reported by a Swedish
1.20
register study9 and a US twin study, although the latter did not
1.10 observe statistically significant results for early-onset depression
in either conventional (risk ratio 1.4, 95% CI 0.7–2.9) or sibling
1.00
stratified (risk ratio 1.6, 95% CI 0.5–5.3) analysis.14 In contrast to
0.90 the previous studies,9,14 we also adjusted for differences in educa-
Basic adjustment Full adjustment tional level between those with and without a history of depression.
Men Women Education is a strong determinant of both depression and dementia,
(b) and thus it was important to explicitly assess the extent to which
1.60 2(2) = 0.37; P = 0.84 2(2) = 0.23; P = 0.89 education confounded the association between the two conditions.
Our analysis shows that the association was not explained by shared
1.50 family background or educational level, thus providing more
1.40 evidence for a non-spurious relationship between depression and
Hazard ratio

1.34 dementia.
1.30 1.30 1.31 1.29 In light of these and previous results, it seems likely that both
1.27
1.25 explanations for an association between depression and dementia
1.20
are valid: depressive symptoms reflect prodromal stages when
1.10 they appear in close proximity to dementia onset, and depression
earlier in life may increase the risk of neurodegeneration itself.
1.00 This is also reflected in the study by Singh-Manoux et al,6 which
0.90 found that individuals with dementia had more depressive symp-
Basic adjustment Full adjustment toms not only less than 10 years before dementia diagnosis but
Tertiary Secondary Basic
also more than 20 years before diagnosis compared with individuals
(c) without dementia. The results of their prospective analysis did not,
1.60 2(3) = 13.97; P = 0.003 2(3) = 13.35; P = 0.004 however, reach statistical significance.6 This highlights the consider-
able demands on data, especially in cohort studies of older people.
1.50 In such studies, the most vulnerable and frail individuals tend to
drop-out from follow-up; consequently, the numbers of cases of
1.40
Hazard ratio

1.36 people with depression and dementia becomes small. We were in

1.33 1.32 a fortunate position to overcome these potential problems as we
1.30 1.30
1.26
1.22 used large register-based data without bias because of self-selection
1.20
or non-random attrition.
1.16 1.13
1.10 The causal mechanisms through which depression may contrib-
ute to dementia risk are still not fully known. The suggested path-
1.00 ways relate to hippocampal atrophy because of the activation of
0.90 the hypothalamic–pituitary–adrenal axis and increased glucocortic-
Basic adjustment Full adjustment
oid production, and raised levels of proinflammatory cytokines.30
Furthermore, depression is associated with several vascular risk
Married Divorced Widowed Never married factors and CVD,23 establishing a potential mediating pathway to
increased dementia risk.4,31 Our adjustments for comorbid medical
Fig. 1 Association between a history of clinical depression and
conditions cover part of these mediating factors but accounting for
dementia by (a) gender, (b) educational level, and (c) marital status. these attenuated the association only modestly. Although we cannot
exclude the possibility of residual confounding, our analysis pro-
Hazard ratio = 1.00 for no history of clinical depression. Error bars indicate 95% CIs.
vides strong evidence for a long-term association between the two
conditions.

association will attenuate with age). The nature of the gender differ-
ence will be an important point for future enquiry.
The association between a history of clinical depression and Impact and relevance
dementia was weaker among the widowed compared with the The consistent long-term association between a history of clinical
married. This finding may arise from selective survival: widowhood depression and later-life dementia risk highlight that patients pre-
is related to higher mortality,29 and thus the surviving widow(er)s senting with depression at working age, especially if severe,
may be more selected on health characteristics. Furthermore, our should be monitored for cognitive function in the long term.
sensitivity analyses suggest that selective survival may also attenuate Furthermore, information about a patient’s depression history
the overall association between depression and dementia. Similar should be utilised in the assessment of cognitive impairment and
attenuation and even reversal of association with increasing age timely detection of dementia. The results also suggest that any
has also been reported for other dementia risk factors including success in fighting depression in those of working age will have
smoking,24 suggesting that individuals with certain risk factors an outsized societal impact throughout adult life in terms of, for

414
https://doi.org/10.1192/bjp.2021.217 Published online by Cambridge University Press
 Association between a history of clinical depression and dementia

example, improved workability and delayed or avoided early exit

Author contributions
from the labour force and the need for long-term institutional care.
K.K. conceived the study. P.M. acquired the data. K.K. designed the analysis, with input from
L.T., T.L., E.E. and P.M.. The analysis and interpretation of the data was carried out by K.K.,
with contributions from L.T., T.L., E.E. and P.M.. The initial version of the manuscript was written
Strengths and limitations by K.K. All authors contributed to the final version, revising it critically for intellectual content
The long follow-up in population and health registers provided an and gave final approval of the version to be published.

opportunity to study the long-term association between depression

and dementia. All our measures came from administrative registers,
and were thus collected prospectively and were not subject to self- or Funding
surrogate reporting biases. Both depression and dementia were This study was supported by grants from the Eino Jutikkala Fund and the Finnish Cultural
diagnosed by a physician, entailing high specificity. Foundation, and travel grants from the Finnish Concordia Fund and the Doctoral School in
Humanities and Social Sciences of the University of Helsinki (K.K.). P.M., L.T. and K.K. were sup-
We also acknowledge the limitations of the study. First, the ported by the European Union Horizon 2020 Programme under grant agreement n° 667661
study is observational in design and thus cannot demonstrate a (Promoting mental wellbeing in the ageing population – MINDMAP). The study does not neces-
sarily reflect the Commission’s views and in no way anticipates the Commission’s future policy
causal relationship between depression and dementia. The long in this area. P.M., L.T. and K.K. were also supported by NordForsk under grant agreement n°
gap between depression diagnosis and dementia follow-up, 83540 (WELLIFE). P.M. has received funding from the Academy of Finland under grant agree-
however, reduces the likelihood that our results are biased by pre- ments n° 294861 and n° 1308247. L.T. is employed in URBARIA in a postdoctoral research pos-
ition funded by the municipalities of the Helsinki capital region. The funders had no role in the
clinical symptoms of dementia. Second, the use of hospital diagno- design and conduct of the study; collection, management, analysis, and interpretation of the
ses entails that less severe depressive episodes not requiring hospital data; and preparation, review, or approval of the manuscript; and decision to submit the manu-
script for publication.
care remained undetected, and thus the results reflect first and fore-
most an association for hospital-treated depression. Third, despite
the reasonably good sensitivity of our register-based measure of
dementia,19 not all people with dementia could be identified. Declaration of interest
However, since the age-specific incidence rates were consistent None.
with those obtained from screened community samples,20 we
believe that our results are not biased because of people with undiag-
nosed cases. Finally, although the sibling fixed-effects model con- References
trols for all unmeasured time-invariant characteristics shared by
siblings, it is likely that many other characteristics are not shared
1 Nichols E, Szoeke CE, Vollset SE, Abbasi N, Abd-Allah F, Abdela J, et al. Global,
by siblings. For example, only half of the genes are shared and regional, and national burden of Alzheimer’s disease and other dementias,
thus the model does not fully exclude the possibility of common 1990–2016: a systematic analysis for the global burden of disease study 2016.
genetic antecedents. Also, other non-shared characteristics might Lancet Neurol 2019; 18: 88–106.
confound the association, but we believe that these differences 2 Liu Q, He H, Yang J, Feng X, Zhao F, Lyu J. Changes in the global burden of
depression from 1990 to 2017: findings from the global burden of disease
were at least partly captured by controlling for observed individual
study. J Psych Res 2020; 126: 134–40.
characteristics including educational level and health
3 Zhao Q-F, Tan L, Wang H-F, Jiang T, Tan M-S, Tan L, et al. The prevalence of
characteristics. neuropsychiatric symptoms in Alzheimer’s disease: systematic review and
In conclusion, our results showed a consistent long-term asso- meta-analysis. J Affect Disord 2016; 190: 264–71.
ciation between a history of clinical depression and dementia, in 4 Livingston G, Huntley J, Sommerlad A, Ames D, Ballard C, Banerjee S, et al.
line with the aetiological risk-factor hypothesis. Although the Dementia prevention, intervention, and care: 2020 report of the lancet com-
mission. Lancet 2020; 396: 413–46.
causal mechanisms are not yet clear, better management and treat-
5 Cherbuin N, Kim S, Anstey KJ. Dementia risk estimates associated with mea-
ment of mental health problems when people are working age
sures of depression: a systematic review and meta-analysis. BMJ Open 2015;
should play a role in dementia prevention. 5: e008853.
6 Singh-Manoux A, Dugravot A, Fournier A, Abell J, Ebmeier K, Kivimäki M, et al.
Kaarina Korhonen , Population Research Unit, Faculty of Social Sciences, University Trajectories of depressive symptoms before diagnosis of dementia: a 28-year
of Helsinki, Finland; Lasse Tarkiainen, Population Research Unit, Faculty of Social follow-up study. JAMA Psychiatry 2017; 74: 712–8.
Sciences, University of Helsinki, Finland and Helsinki Institute of Urban and Regional 7 Brommelhoff JA, Gatz M, Johansson B, McArdle JJ, Fratiglioni L, Pedersen NL.
Studies (URBARIA), University of Helsinki, Finland; Taina Leinonen, Finnish Institute of
Depression as a risk factor or prodromal feature for dementia? Findings in a
Occupational Health, Finland; Elina Einiö, Population Research Unit, Faculty of Social
population-based sample of Swedish twins. Psychol Aging 2009; 24: 373–84.
Sciences, University of Helsinki, Finland; Pekka Martikainen, Population Research Unit,
Faculty of Social Sciences, University of Helsinki, Finland; Department of Public Health 8 da Silva J, Goncalves-Pereira M, Xavier M, Mukaetova-Ladinska EB. Affective
Sciences, Stockholm University, Sweden; and Laboratory of Population Health, Max disorders and risk of developing dementia: systematic review. Br J Psychiatry
Planck Institute for Demographic Research, Germany 2013; 202: 177–86.
Correspondence: Kaarina Korhonen. Email: kaarina.korhonen@helsinki.fi 9 Holmquist S, Nordström A, Nordström P. The association of depression with
subsequent dementia diagnosis: a Swedish nationwide cohort study from
First received 3 Jul 2021, final revision 7 Dec 2021, accepted 11 Dec 2021
1964 to 2016. PLoS Med 2020; 17: e1003016.
10 Barnes DE, Yaffe K, Byers AL, McCormick M, Schaefer C, Whitmer RA. Midlife vs
late-life depressive symptoms and risk of dementia: differential effects for
Alzheimer disease and vascular dementia. Arch Gen Psychiatry 2012; 69:
493–8.
11 Green RC, Cupples LA, Kurz A, Auerbach S, Go R, Sadovnick D, et al. Depression
Supplementary material
as a risk factor for Alzheimer disease: the MIRAGE study. Arch Neurol 2003; 60:
To view supplementary material for this article, please visit https://doi.org/10.1192/bjp.2021. 753–9.
217. 12 Pálsson S, Aevarsson Ó, Skoog I. Depression, cerebral atrophy, cognitive per-
formance and incidence of dementia. Br J Psychiatry 1999; 174: 249–53.
13 Li G, Wang LY, Shofer JB, Thompson ML, Peskind ER, McCormick W, et al.
Data availability Temporal relationship between depression and dementia: findings from a
large community-based 15-year follow-up study. Arch Gen Psychiatry 2011;
The data that support the findings of this study are available from Statistics Finland, the 68: 970–7.
Finnish Institute for Health and Welfare and the Social Insurance Institution of Finland.
Restrictions apply to the availability of these data, which were used under licence for this 14 Steffens DC, Plassman BL, Helms MJ, Welsh-Bohmer KA, Saunders AM,
study. Data are available from the authors with the permission of Statistics Finland and the Breitner JC. A twin study of late-onset depression and apolipoprotein E ε4 as
Social and Health Data Permit Authority Findata. risk factors for Alzheimer’s disease. Biol Psychiatry 1997; 41: 851–6.

415
https://doi.org/10.1192/bjp.2021.217 Published online by Cambridge University Press
 Korhonen et al

15 Geerlings MI, Bouter LM, Schoevers RA, Beekman ATF, Jonker C, Deeg DJH, 23 Penninx BW. Depression and cardiovascular disease: epidemiological evi-
et al. Depression and risk of cognitive decline and Alzheimer’s disease: results dence on their linking mechanisms. Neurosci Biobehav Rev 2017; 74: 277–86.
of two prospective community-based studies in The Netherlands. Br J 24 Hernán MA, Alonso A, Logroscino G. Commentary: cigarette smoking and
Psychiatry 2000; 176: 568–75. dementia: potential selection bias in the elderly. Epidemiology 2008; 19:
16 Markkula N, Suvisaari J, Saarni SI, Pirkola S, Peña S, Saarni S, et al. Prevalence 448–50.
and correlates of major depressive disorder and dysthymia in an eleven-year 25 Rehm J, Hasan OS, Black SE, Shield KD, Schwarzinger M. Alcohol use and
follow-up–results from the finnish health 2011 survey. J Affect Disord 2015; dementia: a systematic scoping review. Alzheimers Res Ther 2019; 11: 1–11.
173: 73–80.
26 Allison PD. Fixed Effects Regression Models (Vol. 160). SAGE Publications, 2009.
17 Vermunt L, Sikkes SA, Van Den Hout A, Handels R, Bos I, Van Der Flier WM, et al.
Duration of preclinical, prodromal, and dementia stages of Alzheimer’s dis- 27 Euser SM, Schram MT, Hofman A, Westendorp RG, Breteler MM. Measuring
ease in relation to age, sex, and APOE genotype. Alzheimers Dement 2019; 15: cognitive function with age: the influence of selection by health and survival.
888–98. Epidemiology 2008; 19: 440–7.

18 World Health Organization. International classification of diseases and related 28 Hunt GE, Malhi GS, Lai HMX, Cleary M. Prevalence of comorbid substance use
health problems, tenth revision 2nd ed. World Health Organization, 2004. in major depressive disorder in community and clinical settings, 1990–2019:
systematic review and meta-analysis. J Affect Disord 2020; 266: 288–304.
19 Solomon A, Ngandu T, Soininen H, Hallikainen MM, Kivipelto M, Laatikainen T.
Validity of dementia and Alzheimer’s disease diagnoses in finnish national 29 Shor E, Roelfs DJ, Curreli M, Clemow L, Burg MM, Schwartz JE. Widowhood
registers. Alzheimers Dement 2014; 10: 303–9. and mortality: a meta-analysis and meta-regression. Demography 2012; 49:
575–606.
20 Wolters FJ, Chibnik LB, Waziry R, Anderson R, Berr C, Beiser A, et al. Twenty-
seven-year time trends in dementia incidence in Europe and the United States: 30 Herbert J, Lucassen PJ. Depression as a risk factor for Alzheimer’s disease:
the Alzheimer cohorts consortium. Neurology 2020; 95: e519–31. genes, steroids, cytokines and neurogenesis–what do we need to know? Front
Neuroendocrinol 2016; 41: 153–71.
21 World Health Organization. International classification of diseases, 9th revision
(1975 revision). World Health Organization, 1977. 31 Kuźma E, Lourida I, Moore SF, Levine DA, Ukoumunne OC, Llewellyn DJ. Stroke
and dementia risk: a systematic review and meta-analysis. Alzheimers Dement
22 World Health Organization. Manual of the international statistical classification 2018; 14: 1416–26.
of disease, injuries, and causes of death. Based on the recommendations of
the eighth revision conference, 1965, and adopted by the Nineteenth World
Health Assembly. World Health Organization, 1967.

Charles Arthur Mercier

Psychiatry
C. V. Haldipur
in History
Charles Mercier (1852–1919), a contemporary of Dr Henry Maudsley, lectured on insanity at Westminster Hospital and London
School of Medicine for Women and later held a post at Charing Cross Hospital in London. He authored several books, whose
titles (Astrology in Medicine, Crime and Insanity, A New Logic) attest to the breadth of his interests and knowledge. He wrote
for the Journal of Mental Science and was President of the Medico-Psychological Association of Great Britain and Ireland in
1908–1909. Mercier addressed some thorny issues in psychiatry that continue to be controversial to this day: the nature of
insanity and the definition of disease.

Insanity, he argued, is a disorder of conduct. The statement does not seem that outlandish when one considers that we base our
diagnosis on the patient’s behaviour and what he or she says.

His two-part essay ‘What is a Disease?’ was published on both sides of the Atlantic, in England as well as in the USA, in 1917.
There is the commonly held notion that diseases exist in nature and that clinicians ‘discover’ them in much the same way as
Columbus discovered America; indeed, we sometimes honour the discoverer by naming the disease after the clinician.
Disease, Mercier averred, is a ‘mental construct or concept, consisting of a symptom or a group of symptoms, correlated
with or by a single intra-corporeal cause’.

As a forensic psychiatrist he was aware of the problem that our inability to define insanity presents in courts of law. With his
characteristic sense of humour he suggested that, if counsel were to ask a psychiatrist to define mental disease in the
courtroom, the psychiatrist should confound the lawyer by responding with a counterquestion to define law.

Mercier’s views of insanity as a conduct disorder and of diseases as mental constructs were controversial then as they are likely
to be today. Nevertheless, as the debate about how diseases are defined continues a century after these views were published,
his proposed definition may be worth re-examining.

Sir William Osler, in his obituary of Mercier, described him as having a ‘rich vocabulary and a keen wit, he had no equal among us
as a controversialist’.

© The Author(s), 2022. Published by Cambridge University Press on behalf of the Royal College of Psychiatrists

The British Journal of Psychiatry (2022)

221, 416. doi: 10.1192/bjp.2021.203

416
https://doi.org/10.1192/bjp.2021.217 Published online by Cambridge University Press