The association of depression with subsequent dementia diagnosis: A Swedish nationwide cohort study from 1964 to 2016.

BACKGROUND: Depression is associated with an increased risk of dementia. However, short follow-up times and lack of adjustment for familial factors in previous studies could influence this association. The purpose of the present study was to investigate the association between depression and subsequent dementia, while controlling for familial factors and with a follow-up of over 35 years. METHODS AND
FINDINGS: Two cohorts were formed from all individuals aged 50 years or older living in Sweden as of 31 December 2005 (n = 3,341,010). The Swedish National Patient Register was searched from 1964 through 2016 to identify diagnosis of depression and dementia. In the first cohort, individuals diagnosed with depression (n = 119,386) were matched 1:1 with controls without depression diagnosis. The second cohort was a sibling cohort (n = 50,644) consisting of same-sex full sibling pairs with discordant depression status. In the population matched cohort study, a total of 9,802 individuals were diagnosed with dementia during a mean follow-up time of 10.41 (range 0-35) years (5.5% of those diagnosed with depression and 2.6% of those without depression diagnosis (adjusted odds ratio [aOR] 2.47, 95% CI 2.35-2.58; p < 0.001), with a stronger association for vascular dementia (aOR 2.68, 95% CI 2.44-2.95; p < 0.001) than for Alzheimer disease (aOR 1.79, 95% CI 1.68-1.92; p < 0.001). The association with dementia diagnosis was strongest in the first 6 months after depression diagnosis (aOR 15.20, 95% CI 11.85-19.50; p < 0.001), then decreased rapidly but persisted over follow-up of more than 20 years (aOR 1.58, 95% CI 1.27-1.98; p < 0.001). Also in the sibling cohort, the association was strongest in the first 6 months (aOR 20.85, 95% CI 9.63-45.12; p < 0.001), then decreased rapidly but persisted over follow-up of more than 20 years (aOR 2.33, 95% CI 1.32-4.11; p < 0.001). The adjusted models included sex, age at baseline, citizenship, civil status, household income, and diagnoses at baseline. The main limitation of the study methodology is the use of observational data; hence, the associations found are not proof of causal effects.
CONCLUSIONS: Depression is associated with increased odds of dementia, even more than 20 years after diagnosis of depression, and the association remains after adjustment for familial factors. Further research is needed to investigate whether successful prevention and treatment of depression decrease the risk of dementia.

Introduction

Dementia is common among the elderly, causing severe individual suffering as well as societal strain [1]. As the proportion of people aged 65 years and above is rapidly increasing in the world population, the number of individuals with dementia is expected to double within 20 years, and this condition was estimated to have a worldwide cost of US$604 billion in 2010 [2]. Effective treatments for dementia remain scarce [3]; however, a preventive approach may be possible through the identification of high-risk individuals and potentially modifiable risk factors [4-6].

An association between depression and subsequent dementia diagnosis has been suggested in at least 4 independent meta-analyses—indicating close to a 2-fold higher risk of developing dementia following depression [7-10]—as well as several reviews [11-14]. However, there is a lack of studies with large samples and long follow-up, as well as adjustment for important covariates, particularly familial factors. Depression preceding dementia has been discussed both as a risk factor and possible prodrome [9,11], especially with regard to early versus late onset of depression. Thus, several reviews suggest that depression in early life may be a risk factor of subsequent dementia diagnosis [11,13,14] and that depression in late life is rather a prodrome of dementia [11,13], but the literature is not consistent [12]. Some studies have found an association with dementia more than a decade after onset of depression [15], whereas other studies have failed to demonstrate such an association [16,17]. Furthermore, there is evidence suggesting that the association between depression and dementia has a dose–response relationship [12,13,18-20], and the association may be dependent on the type of dementia [13,21,22]. Given the sex differences in the prevalence of depression [23-25] and dementia [26,27], the association may also differ between men and women. Finally, although it is well known that both depression and dementia are influenced by both genetic and familial factors [28-31], it has not been evaluated whether genetic and familial factors influence the association between depression and dementia.

The aim of the present study was to investigate the association between depression and dementia in a large nationwide cohort that included about 120,000 individuals diagnosed with depression and matched controls. Based on a previous study, where we investigated the association between traumatic brain injury and dementia [32], we hypothesized in particular that the association between depression and dementia would persist decades after the depression diagnosis, that the association would be independent of familial factors, and that the association would be stronger with more severe depression for the outcome of vascular dementia.

Methods

This study is reported as per the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) guideline (S1 STROBE Checklist).

Sample

The total cohort from which individuals were considered for inclusion consisted of all Swedish residents ≥50 years old who were alive on 31 December 2005 (n = 3,341,010). A total of 813 individuals were excluded because of missing data on sex, date of birth, or citizenship, or because the date of diagnosis for depression or dementia was postmortem in the national registers used in the present study. From the remaining total cohort (n = 3,340,197), 2 component cohorts were formed and included in the study.

The first cohort consisted of individuals diagnosed with major depression (n = 119,386), with no prior dementia diagnosis, each matched with 1 individual without a diagnosis of depression during follow-up, on the basis of birth year and month, sex, and citizenship (Swedish or non-Swedish). Baseline for both the individuals diagnosed with major depression and their corresponding controls was at the date of depression diagnosis of the individual diagnosed with depression. Controls who died before baseline or were diagnosed with dementia before baseline were excluded, and a new control was searched from the remaining cohort of potential controls. This procedure was repeated 3 times.

The second cohort was a sibling cohort (n = 50,644), consisting of same-sex full sibling pairs with discordant depression status during follow-up. Baseline for both the affected sibling and the unaffected sibling was set at the date of depression diagnosis of the affected sibling. Sibling pairs where the unaffected sibling died before baseline or at least 1 of the siblings was diagnosed with dementia before baseline were excluded. A few sibling pairs had discordant citizenship status. The purpose of the sibling cohort was to control for the effect of early environment and genetic factors (i.e., familial factors).

Procedure

The Swedish National Patient Register (SNPR), controlled by the National Board of Health and Welfare, was searched from 1 January 1964 through 31 December 2016 to identify diagnosis of major depression and dementia, classified according to the International Classification of Diseases (ICD, 8th, 9th, and 10th revisions). Diagnosis of major depression included the diagnosis codes F32 and F33 (ICD-10), 311 (ICD-9-SWE and ICD-8-SWE), and 296B (ICD-9-SWE). For the F32 and F33 diagnosis codes, depression was coded as mild (F320, F330), moderate (F321, F331), or severe (F322, F323, F332, F333). Depression was considered recurrent for diagnosis code F33 (recurrent depression; ICD-10) or at least 2 diagnoses of depression at least 6 months apart; otherwise, depression was considered as single episode. Diagnosis of dementia was coded as Alzheimer dementia (ICD-10: F00, G30) or vascular dementia (ICD-10: F01) and also included the diagnosis codes F039 (ICD-10) and 290 (ICD-8 and ICD-9). Based on their association with the main exposure in the present study (major depression), the main outcome (dementia), or death, the following covariates were selected and searched in the SNPR using appropriate ICD codes: diabetes, myocardial infarction, alcohol intoxication, drug intoxication, chronic pulmonary disease, renal failure, stroke, and hypothyroidism. The SNPR has national coverage of close to 90% from 1970 to 1986 for inpatient care, and complete coverage since 1987, and in a validation study showed predictive values of 85%–95% for most diagnoses [33]. All Swedish physicians are obliged to report data for all care (except for primary care visits) to the SNPR, including psychiatric care from 1973 and hospital-based outpatient care from 2001 [33]. Diagnosis of death was collected from the Swedish Cause of Death Register, complete since 1961, also controlled by the National Board of Health and Welfare and linked to each individual in the cohort. Data on citizenship (Swedish or non-Swedish), civil status, and household income in 2005 were derived from Statistics Sweden (https://www.scb.se), and linked to each individual in the cohort. There was no written prospective protocol for the analysis; however, all analyses were preplanned based on the results of a previous study [32].

Statistical analysis

In order to evaluate if the association between diagnosis of depression and subsequent diagnosis of dementia was time dependent, Schoenfeld’s residuals were calculated in Stata (version 12.1; StataCorp). Since the proportional hazards assumption was violated, the associations were analyzed in time intervals between baseline and the end of follow-up. For this purpose, multivariable conditional logistic regression analysis was performed in the 2 cohorts, with dementia as the dependent variable. SPSS (version 25.0; IBM) was used for the data analysis. The first model was unadjusted, but by design adjusted for sex, age at baseline, and citizenship in the population matched cohort, and for sex and familial factors in the sibling cohort. A second model was further adjusted for civil status, household income, and diagnoses at baseline in the population matched cohort, and for citizenship, age at baseline, civil status, household income, and diagnoses at baseline in the sibling cohort (see Table 1). In the population matched cohort study, separate analyses were performed by type of dementia (Alzheimer or vascular dementia) and for different severity (mild, moderate, or severe) and frequency (1 diagnosis of depression or recurrent depression) of depressive episodes. Separate analyses were also performed for men and women. The time dependency of the association between depression and dementia in the 2 different cohorts was graphically illustrated using flexible parametric models with 5 degrees of freedom (Fig 1) [34]. The knots were set in default positions. The differences between individuals based on depression diagnosis at baseline, in the population matched cohort and sibling cohort, were investigated using chi-squared tests for categorical data and independent samples t tests for continuous data. For all statistical analyses, significance level was set at p < 0.05.

Table 1

Cohort characteristics.

Characteristic	Matched cohort (n = 238,772)			Sibling cohort (n = 50,644)
	Depression (n = 119,386)	No depression (n = 119,386)	Group difference: depression versus no depression	Depression (n = 25,322)	No depression (n = 25,322)	Group difference: depression versus no depression
Age at baseline in years, mean (SD)	63.79 (11.89)			59.10 (8.85)	59.97 (8.91)	t (50,639.38) = 11.01, p < 0.001 (2-tailed); 95% CI of the difference = 0.71–1.02
Citizenship						χ2 (df = 1, n = 50,644) = 0.026, p = 0.87 (2-sided)
Non-Swedish	42,928 (18.0%)			1,698 (6.7%)	1,689 (6.7%)
Swedish	195,844 (82.0%)			23,624 (93.3%)	23,633 (93.3%)
Sex
Male	93,758 (39.3%)			21,416 (42.3%)
Female	145,014 (60.7%)			29,228 (57.7%)
Civil status			χ2 (df = 4, n = 238,040) = 4,120.52, p < 0.001 (2-sided)			χ2 (df = 4, n = 55,540) = 815.07, p < 0.001 (2-sided)
Married	53,478 (44.8%)	67,873 (56.9%)		12,108 (47.8%)	15,165 (59.9%)
Unmarried	18,009 (15.1%)	13,934 (11.7%)		4,614 (18.2%)	3,695 (14.6%)
Divorced	30,333 (25.4%)	20,600 (17.3%)		7,117 (28.1%)	5,006 (19.8%)
Widowed	17,297 (14.5%)	16,394 (13.7%)		1,413 (5.7%)	1,370 (5.4%)
Other	78 (0.1%)	44 (<0.1%)		21 (0.1%)	13 (0.1%)
Missing	191 (0.2%)	541 (0.5%)		31 (0.1%)	73 (0.3%)
Yearly household income in Swedish krona, mean (SD); approximate equivalent in US dollars*	255,808 (339,534); $26,368 (35,350)	311,620 (489,856); $32,429 (50,977)	t (211,706.04) = 32.31, p < 0.001 (2-tailed); 95% CI of the difference = 52,425.27–59,197.42	291,265 (438,927); $30,324 (45,705)	340,819 (346,022); $5,485 (36,027)	t (47,977.85) = 14.01, p < 0.001 (2-tailed); 95% CI of the difference = 42,664.91–56,443.49
Diagnoses at baseline
Diabetes	6,281 (5.3%)	4,368 (3.7%)	χ2 (df = 1, n = 238,772) = 359.70, p < 0.001 (2-sided)	1,172 (4.6%)	816 (3.2%)	χ2 (df = 1, n = 50,644) = 66.36, p < 0.001 (2-sided)
Myocardial infarction	5,634 (4.7%)	4,177 (3.5%)	χ2 (df = 1, n = 238,772) = 225.65, p < 0.001 (2-sided)	800 (3.2%)	644 (2.5%)	χ2 (df = 1, n = 50,644) = 17.35, p < 0.001 (2-sided)
Alcohol intoxication	7,803 (6.5%)	1,736 (1.5%)	χ2 (df = 1, n = 238,772) = 4,019.31, p < 0.001 (2-sided)	1,977 (7.8%)	605 (2.4%)	χ2 (df = 1, n = 50,644) = 768.21, p < 0.001 (2-sided)
Drug intoxication	1,763 (1.5%)	346 (0.3%)	χ2 (df = 1, n = 238,772) = 960.54, p < 0.001 (2-sided)	418 (1.7%)	115 (0.5%)	χ2 (df = 1, n = 50,644) = 174.08, p < 0.001 (2-sided)
Chronic pulmonary disease	3,028 (2.5%)	1,477 (1.2%)	χ2 (df = 1, n = 238,772) = 544.25, p < 0.001 (2-sided)	594 (2.3%)	315 (1.2%)	χ2 (df = 1, n = 50,644) = 87.20, p < 0.001 (2-sided)
Renal failure	1,325 (1.1%)	708 (0.6%)	χ2 (df = 1, n = 238,772) = 188.86, p < 0.001 (2-sided)	245 (1.0%)	122 (0.5%)	χ2 (df = 1, n = 50,644) = 41.52., p < 0.001 (2-sided)
Stroke	5,461 (4.6%)	2,967 (2.5%)	χ2 (df = 1, n = 238,772) = 765.02, p < 0.001 (2-sided)	857 (3.4%)	454 (1.8%)	χ2 (df = 1, n = 50,644) = 127.17, p < 0.001 (2-sided)
Hypothyroidism	2,622 (2.2%)	1,589 (1.3%)	χ2 (df = 1, n = 238,772) = 257.95, p < 0.001 (2-sided)	495 (2.0%)	299 (1.2%)	χ2 (df = 1, n = 50,644) = 49.15, p < 0.001 (2-sided)
Survival time in years, mean (SD)	10.04 (6.91)	10.79 (6.86)	t (238,754.64) = 26.80, p < 0.001 (2-tailed); 95% CI of the difference = 0.70–0.81	10.78 (7.00)	11.30 (6.90)	t (50,631.57) = 8.39, p < 0.001 (2-tailed); 95% CI of the difference = 0.40–0.64

Data are number (percent) unless otherwise indicated. Information on citizenship, civil status, and household income was derived from Statistics Sweden for the year of 2005.

*Using 2019 exchange rate: 1:0.104.

Fig 1

Association between depression and the risk of dementia during follow-up for 238,772 individuals.

Individuals with follow-up time of less than 1 month were excluded. The figure was constructed using a flexible parametric model with 5 degrees of freedom and knots in default positions. The black line represents the hazard ratio, and the grey areas represent 95% confidence intervals.

Ethical considerations

The Regional Ethical Review Board in Umeå (Nr 2013-86-31) and the National Board of Health and Welfare approved the present study. The study was funded by the Swedish Research Council (grant number 2016–02589). The funders had no role in the design, the interpretation of the results, the decision to publish, or the preparation of the manuscript. As data were analyzed anonymously, informed consent was not obtained from study participants.

Results

Characteristics of the sample

The characteristics of the population matched cohort and the sibling cohort are displayed in Table 1. Chi-squared test and t tests revealed statistically significant differences between cases and controls for civil status, household income, and diagnoses at baseline in the population matched cohort, and for age at baseline, civil status, household income, and diagnoses at baseline in the sibling cohort (all p < 0.001; see Table 1).

Time dependency of the association between depression and subsequent dementia

Schoenfeld’s residuals showed that the proportional hazards assumption was violated in the population matched cohort, indicating a time-dependent association between depression and subsequent dementia.

Incident cases of dementia in the population matched cohort

In the population matched cohort study, a total of 9,802 individuals were diagnosed with dementia during a mean follow-up time of 10.41 (SD 6.89; range 0–35) years (5.7% of those diagnosed with depression and 2.6% of those without depression diagnosis; adjusted odds ratio [aOR] 2.47, 95% CI 2.35–2.58; p < 0.001; Table 2). The cumulative incidence of dementia during follow-up is illustrated in Fig 2. The association with dementia diagnosis was strongest in the first 6 months after depression diagnosis (aOR 15.20, 95% CI 11.85–19.50; p < 0.001), then decreased rapidly but persisted for more than 20 years (aOR 1.58, 95% CI 1.27–1.98; p < 0.001; Table 2).

Table 2

Associations between depression and the risk of subsequent dementia diagnosis during follow-up in 119,386 individuals diagnosed with depression and 119,386 matched controls without depression diagnosis: Population matched cohort (n = 238,772).

Time after depression diagnosis	Individuals at risk	Diagnosed with dementia					Unadjusted*			Adjusted**
		Total	Cases		Controls		OR	95% CI	p-Value	OR	95% CI	p-Value
			Count	Percent	Count	Percent
Overall	238,772	9,802	6,752	5.66%	3,050	2.55%	2.37	2.27–2.48	<0.001	2.47	2.35–2.58	<0.001
0–5.9 months	238,772	1,043	972	0.81%	71	0.06%	14.06	11.01–17.95	<0.001	15.20	11.85–19.50	<0.001
6–11.9 months	232,274	750	652	0.57%	98	0.08%	7.24	5.80–9.03	<0.001	7.83	6.24–9.83	<0.001
1–1.9 years	226,605	1,006	816	0.73%	190	0.17%	4.48	4.09–5.73	<0.001	5.15	4.33–6.12	<0.001
2–4.9 years	215,680	2,129	1,474	1.40%	655	0.59%	2.58	2.34–2.86	<0.001	2.62	2.36–2.91	<0.001
5–9.9 years	180,192	2,289	1,384	1.59%	914	0.98%	1.83	1.67–2.01	<0.001	1.85	1.68–2.05	<0.001
10–20 years	112,027	2,095	1,187	2.21%	908	1.55%	1.54	1.40–1.70	<0.001	1.58	1.43–1.75	<0.001
20 years or more	26,044	481	267	2.19%	214	1.55%	1.60	1.30–1.97	<0.001	1.58	1.27–1.98	<0.001

ORs and 95% confidence intervals were derived from conditional logistic regression analysis.

*By design adjusted for sex, age at baseline, and citizenship.

**Further adjusted for civil status, household income, and diagnoses at baseline: diabetes, myocardial infarction, alcohol intoxication, drug intoxication, chronic pulmonary disease, renal failure, stroke, and hypothyroidism.

OR, odds ratio.

Fig 2

Cumulative incidence of dementia after depression diagnosis in the matched cohort of 119,386 individuals with depression and 119,386 individuals with no depression during follow-up.

The number of individuals at risk at each time point is presented below the graph.

Incident cases of dementia in the sibling cohort

In the sibling cohort, a total of 1,161 individuals were diagnosed with dementia during a mean follow-up of 11.04 (SD 6.96; range 0–31) years, 3.4% of those diagnosed with depression and 1.2% of those without depression diagnosis (aOR 3.75, 95% CI 3.24–4.34; p < 0.001; Table 3). The cumulative incidence of dementia during follow-up in the sibling cohort is illustrated in Fig 3. As in the population matched cohort, the association with dementia diagnosis was strongest the first 6 months after depression diagnosis (aOR 20.85, 95% CI 9.63–45.12; p < 0.001), then decreased rapidly but persisted over follow-up of more than 20 years (aOR 2.33, 95% CI 1.32–4.11; p < 0.001; Table 3).

Table 3

Associations between depression and the risk of subsequent dementia diagnosis during follow-up for 25,322 full sibling pairs with discordant depression status at baseline: Sibling cohort (n = 50,644).

Time after depression diagnosis	Individuals at risk	Diagnosed with dementia					Unadjusted*			Adjusted**
		Total	Cases		Controls		OR	95% CI	p-Value	OR	95% CI	p-Value
			Count	Percent	Count	Percent
Overall	50,644	1,161	864	3.41%	297	1.17%	3.16	2.75–3.64	<0.001	3.75	3.24–4.34	<0.001
0–5.9 months	50,644	135	128	0.51%	7	0.03%	18.29	8.55–39.13	<0.001	20.85	9.63–45.12	<0.001
6–11.9 months	49,600	86	82	0.33%	4	0.02%	20.25	7.42–55.26	<0.001	24.39	8.82–67.45	<0.001
1–1.9 years	48,609	136	116	0.48%	20	0.08%	5.80	3.61–9.32	<0.001	7.07	4.33–11.54	<0.001
2–4.9 years	46,666	228	174	0.76%	54	0.23%	3.49	2.54–4.79	<0.001	4.25	3.04–5.94	<0.001
5–9.9 years	39,894	248	156	0.80%	92	0.45%	1.86	1.41–2.45	<0.001	2.03	1.51–2.72	<0.001
10–20 years	25,708	252	159	1.27%	93	0.70%	1.96	1.49–2.58	<0.001	2.50	1.86–3.35	<0.001
20 years or more	6,296	76	49	1.61%	27	0.83%	1.95	1.15–3.30	0.013	2.33	1.32–4.11	<0.001

ORs and confidence intervals were derived from conditional logistic regression analysis.

*By use of a sibling design, adjusted for sex and familial factors.

**Further adjusted for citizenship, age at baseline, civil status, household income, and diagnoses at baseline: diabetes, myocardial infarction, alcohol intoxication, drug intoxication, chronic pulmonary disease, renal failure, stroke, and hypothyroidism.

OR, odds ratio.

Fig 3

Cumulative incidence of dementia after depression diagnosis in 25,322 sibling pairs discordant for depression diagnosis.

The number of individuals at risk at each time point is presented below the graph.

Severity and frequency of depression, type of dementia, and patient sex in the population matched cohort

The severity of the depression influenced the association with dementia (Table 4). Thus, after a mild depression, the risk of subsequent dementia remained elevated 5–9.9 years after depression diagnosis (aOR 1.96, 95% CI 1.39–2.75; p < 0.001) but not at ≥10 years after depression diagnosis (aOR 1.39, 95% CI 0.83–2.36; p = 0.212). In contrast, the association remained significant at ≥10 years after depression diagnosis for both for moderate depression (aOR 1.41, 95% CI 1.02–1.95; p = 0.037) and for severe depression (aOR 1.75, 95% CI 1.29–2.36; p < 0.001). No clear differences were found between a single episode of depression and recurrent depression or between men and women (Table 5). The type of dementia also influenced the magnitude of the association. The association was stronger for vascular dementia than for Alzheimer disease, and remained at ≥20 years only for vascular dementia (aOR 1.64, 95% CI 1.02–2.62; p = 0.040, Table 5).

Table 4

The risk of dementia diagnosis during follow-up for individuals diagnosed with mild depression (n = 10,030), moderate depression (n = 24,746), and severe depression (n = 20,540), compared to matched controls.

Time after depression diagnosis	Individuals at risk	Diagnosed with dementia	Unadjusted*			Adjusted**
			OR	95% CI	p-Value	OR	95% CI	p-Value
Mild depression (n = 20,060)
Overall	20,060	728	2.75	2.32–3.26	<0.001	2.85	2.38–3.39	<0.001
0–5.9 months	20,060	75	17.75	6.48–48.60	<0.001	18.02	6.53–49.71	<0.001
6–11.9 months	19,495	69	5.70	2.91–11.16	<0.001	5.61	2.83–11.11	<0.001
1–1.9 years	18,921	82	4.57	2.56–8.15	<0.001	4.26	2.33–7.80	<0.001
2–4.9 years	17,840	222	3.74	2.66–5.25	<0.001	3.98	2.79–5.68	<0.001
5–9.9 years	14,269	191	1.91	1.38–2.65	<0.001	1.96	1.39–2.75	<0.001
10+ years	7,117	89	1.33	0.81–2.20	0.258	1.39	0.83–2.36	0.212
Moderate depression (n = 49,492)
Overall	49,492	1,426	2.07	1.84–2.32	<0.001	2.16	1.92–2.43	<0.001
0–5.9 months	49,492	124	11.40	5.97–21.76	<0.001	11.66	6.06–22.43	<0.001
6–11.9 months	48,309	108	5.63	3.31–9.57	<0.001	5.93	3.46–10.18	<0.001
1–1.9 years	47,092	178	3.50	2.44–5.02	<0.001	3.70	2.56–5.36	<0.001
2–4.9 years	44,802	367	2.14	1.70–2.69	<0.001	2.16	1.71–2.74	<0.001
5–9.9 years	36,401	417	1.58	1.27–1.96	<0.001	1.56	1.25–1.95	<0.001
10+ years	19,177	229	1.36	1.00–1.85	0.053	1.41	1.02–1.95	0.037
Severe depression (n = 41,080)
Overall	41,080	1,479	2.32	2.07–2.61	<0.001	2.36	2.10–2.66	<0.001
0–5.9 months	41,080	127	20.17	8.88–45.78	<0.001	20.99	9.21–47.89	<0.001
6–11.9 months	39,951	130	6.38	3.76–10.80	<0.001	6.77	3.96–11.58	<0.001
1–1.9 years	38,896	152	6.42	3.96–10.41	<0.001	6.08	3.71–9.97	<0.001
2–4.9 years	36,939	396	2.32	1.85–2.90	<0.001	2.43	1.92–3.06	<0.001
5–9.9 years	30,558	403	1.78	1.43–2.22	<0.001	1.77	1.41–2.22	<0.001
10+ years	17,095	281	1.80	1.35–2.39	<0.001	1.75	1.29–2.36	<0.001

Classification of the severity of the depression based of the International Classification of Diseases–10th revision. ORs and confidence intervals were derived from conditional logistic regression analysis.

*By design adjusted for patient sex, age at baseline, and citizenship.

**Further adjusted for civil status, household income, and diagnoses at baseline: diabetes, myocardial infarction, alcohol intoxication, drug intoxication, chronic pulmonary disease, renal failure, stroke, and hypothyroidism.

OR, odds ratio.

Table 5

Risk of dementia diagnosis during follow-up for individuals with diagnosed single depression (n = 56,522) and recurrent depression (n = 62,864), for men with depression (n = 46,879) and women with depression (n = 75,507), and for Alzheimer disease and vascular dementia separately, compared to matched controls.

Time after depression diagnosis	Individuals at risk	Diagnosed with dementia	Unadjusted			Adjusted
			OR	95% CI	p-Value	OR	95% CI	p-Value
Single depressive episode (n = 113,044)*
Overall	113,044	4,992	2.32	2.18–2.48	<0.001	2.48	2.32–2.65	<0.001
0–5.9 months	113,044	837	16.77	12.49–22.50	<0.001	18.82	13.92–25.45	<0.001
6–11.9 months	107,693	506	7.45	5.66–9.80	<0.001	8.17	6.17–10.82	<0.001
1–1.9 years	103,787	547	4.42	3.53–5.52	<0.001	5.06	4.01–6.39	<0.001
2–4.9 years	97,340	1,070	2.46	2.13–2.84	<0.001	2.63	2.26–3.06	<0.001
5–9.9 years	78,049	987	1.51	1.31–1.74	<0.001	1.51	1.30–1.75	<0.001
10–19.9 years	47,002	839	1.13	0.97–1.32	0.108	1.14	0.97–1.34	0.107
20 years or more	12,736	206	1.56	1.13–2.15	0.007	1.61	1.14–2.26	0.006
Recurrent depression (n = 125,728)*
Overall	125,728	4,810	2.43	2.27–2.59	<0.001	2.47	2.31–2.64	<0.001
0–5.9 months	125,728	206	8.27	5.32–12.88	<0.001	8.47	5.41–13.27	<0.001
6–11.9 months	124,581	244	6.84	4.69–9.97	<0.001	7.54	5.11–11.13	<0.001
1–1.9 years	122,818	459	5.41	4.20–6.99	<0.001	5.38	4.15–6.99	<0.001
2–4.9 years	118,340	1,059	2.70	2.35–3.11	<0.001	2.66	2.31–3.07	<0.001
5–9.9 years	102,143	1,311	2.10	1.86–2.38	<0.001	2.14	1.87–2.44	<0.001
10–19.9 years	65,025	1,256	1.90	1.67–2.16	<0.001	1.96	1.72–2.23	<0.001
20 years or more	13,308	275	1.63	1.24–2.15	<0.001	1.55	1.16–2.09	<0.001
Men (n = 93,758)**
Overall	90,375	3,383	2.52	2.33–2.72	<0.001	2.63	2.43–2.85	<0.001
0–5.9 months	90,375	380	17.90	11.41–28.08	<0.001	19.59	12.32–30.98	<0.001
6–11.9 months	90,855	284	7.18	5.01–10.27	<0.001	7.92	5.48–11.44	<0.001
1–1.9 years	89,239	378	4.68	3.57–6.15	<0.001	4.87	3.67–6.46	<0.001
2–4.9 years	83,664	725	2.83	2.37–3.37	<0.001	2.81	2.34–3.38	<0.001
5–9.9 years	68,972	777	2.07	1.75–2.44	<0.001	2.06	1.73–2.45	<0.001
10–19.9 years	42,485	693	1.68	1.42–1.99	<0.001	1.78	1.49–2.12	<0.001
20 years or more	9,752	146	1.46	1.01–2.12	0.050	1.41	0.95–2.11	0.091
Women (n = 145,014)**
Overall	145,014	6,419	2.30	2.18–2.43	<0.001	2.38	2.25–2.52	<0.001
0–5.9 months	145,014	663	12.49	9.34–16.71	<0.001	13.42	9.97–18.05	<0.001
6–11.9 months	144,419	446	7.27	5.49–9.64	<0.001	7.83	5.86–10.45	<0.001
1–1.9 years	138,312	628	4.94	3.99–6.12	<0.001	5.35	4.30–6.67	<0.001
2–4.9 years	132,016	1,404	2.47	2.19–2.79	<0.001	2.54	2.24–2.89	<0.001
5–9.9 years	111,220	1,521	1.73	1.54–1.93	<0.001	1.76	1.57–1.98	<0.001
10–19.9 years	69,542	1,402	1.48	1.31–1.66	<0.001	1.50	1.33–1.69	<0.001
20 years or more	16,292	335	1.67	1.30–2.15	<0.001	1.66	1.27–2.17	<0.001
Alzheimer disease (n = 238,772)*
Overall	238,772	4,201	1.65	1.54–1.75	<0.001	1.79	1.68–1.92	<0.001
0–5.9 months	238,772	470	10.19	7.42–13.99	<0.001	11.32	8.20–15.61	<0.001
6–11.9 months	232,274	343	5.66	4.19–7.65	<0.001	6.39	4.69–8.71	<0.001
1–1.9 years	226,605	462	3.96	3.13–4.99	<0.001	4.56	3.58–5.80	<0.001
2–4.9 years	215,680	918	1.94	1.68–2.24	<0.001	2.13	1.84–2.47	<0.001
5–9.9 years	180,192	978	1.21	1.05–1.38	0.007	1.28	1.11–1.47	0.001
10–19.9 years	112,027	825	0.89	0.77–1.04	0.13	0.97	0.83–1.23	0.657
20 years or more	26,044	205	0.99	0.73–1.34	0.938	0.97	0.70–1.34	0.854
Vascular dementia (n = 238,772)*
Overall	238,772	2,329	2.74	2.50–3.01	<0.001	2.68	2.44–2.95	<0.001
0–5.9 months	238,772	254	20.17	11.30–36.01	<0.001	21.19	11.51–39.00	<0.001
6–11.9 months	232,274	188	9.17	5.64–14.91	<0.001	9.54	5.81–15.66	<0.001
1–1.9 years	226,605	241	6.00	4.15–8.67	<0.001	5.61	3.84–8.17	<0.001
2–4.9 years	215,680	523	3.06	2.48–3.77	<0.001	2.72	2.19–3.37	<0.001
5–9.9 years	180,192	529	2.37	1.93–2.90	<0.001	2.36	1.91–2.91	<0.001
10–19.9 years	112,027	491	1.72	1.42–2.10	<0.001	1.69	1.37–2.07	<0.001
20 years or more	26,044	103	1.58	1.02–2.44	0.041	1.64	1.02–2.62	0.040

Depression was considered as recurrent for diagnosis code F33 (recurrent depression; ICD-10) or at least 2 diagnoses of depression at least 6 months apart; otherwise, depression was considered as a single episode. ORs and confidence intervals were derived from conditional logistic regression analysis.

*Unadjusted model is by design adjusted for patient sex, age at baseline, and citizenship. Adjusted model is further adjusted for civil status, household income, and diagnoses at baseline: diabetes, myocardial infarction, alcohol intoxication, drug intoxication, chronic pulmonary disease, renal failure, stroke, and hypothyroidism.

**Unadjusted model is by design adjusted for age at baseline and citizenship. Adjusted model is further adjusted for civil status, household income, and diagnoses at baseline: diabetes, myocardial infarction, alcohol intoxication, drug intoxication, chronic pulmonary disease, renal failure, stroke, and hypothyroidism.

OR, odds ratio.

Discussion

The results of the present nationwide study indicated increased odds for dementia diagnosis after depression, even when the depression occurred 20 years or more before the dementia diagnosis and while adjusting for a set of covariates including sex, age at baseline, citizenship, civil status, household income, and diagnoses at baseline. The association was strongest when dementia occurred within the first 6 months after depression diagnosis, and then decreased rapidly but remained statistically significant, with an OR of close to 1.6 at a follow-up time of 20 years or more. The results were consistent when controlling for familial factors in a nationwide sibling cohort. The results supported our hypothesis of an association between depression and dementia persisting decades after the depression diagnosis date, independent of familial factors.

Although an association between depression and risk of subsequent dementia has been established previously [7,8,11,12,35], the nature of the association has not been explored in great detail. In studies with short follow-up times, the association may reflect several different potential underlying relationships: depressive symptoms might be a characteristic of the prodromal phase of dementia, the patient might in fact already have dementia although yet undiagnosed, or the depression might be discovered and diagnosed when the patient is under evaluation for dementia. Additionally, lack of evaluation of important covariates increases the risk of confounding and bias. In the present study, the maximum follow-up time was 35 years, and the association between depression and subsequent risk of dementia was consistent for more than 20 years of follow-up. In addition, the association remained after adjusting for an array of important covariates including medical conditions and socioeconomic factors. The results were similar for men and women, suggesting an association independent of patient sex.

To the best of our knowledge, it has not been previously established whether the association between dementia and depression remains after controlling for familial factors. This is of importance as early environmental or genetic factors may mediate the association between depression and subsequent dementia. Interestingly, in a previous study using a twin design, there was an association between depression and dementia [16]. However, 40% of the dementia diagnoses occurred within 4 years of the depression, and the study sample was small, which limited the possibility of evaluating the association during longer follow-up times; the study also lacked covariates. Thus, the association found may well represent depression in individuals with undiagnosed dementia, or dementia in an early phase. In the present study we controlled for familial factors using a sibling design including more than 25,000 sibling pairs. The results from the sibling cohort showed an association between depression and subsequent risk of dementia that lasted for more than 20 years. Thus, the association was similar to that found in the population matched cohort study, suggesting an association between the 2 diagnoses independent of genetic and early environmental factors.

It is also not clear from previous studies whether the association between depression and dementia is dependent on dementia type [13,21,22]. Our results indicate that the association is stronger and covers a wider timespan for vascular dementia than for Alzheimer disease: For vascular dementia, the risk remained statistically significant over a follow-up of more than 20 years, whereas for Alzheimer disease the association remained statistically significant for less than 10 years. The association of depression with Alzheimer disease may best be described as reflecting prodromal depressive symptoms [36]. With respect to the stronger association with vascular dementia, it could be that depression induces the inflammatory process of atherosclerosis that is known to begin already early in life [37]. This hypothesis has support from a recent study in which Setiawan and others [38] found signs of increased activation of microglia in certain areas of the brain in individuals after a major depression, as well as from a study noting cerebrovascular disease in depressed elderly individuals to be associated with cognitive decline [39]. Yet, given the observational design of these studies and our study, no causal inferences can be made. Previous studies [18-20] and reviews [12,13] have suggested a dose–response relationship between depression and the risk of subsequent dementia. In the present study the association between depression and risk of subsequent dementia remained statistically significant for moderate and severe depression for a follow-up of more than 10 years, in contrast to mild depression. However, the association was similar for a single episode of depression and recurrent depression. Thus, our hypothesis of a dose–response relationship between depression and subsequent dementia was only partly supported. Again, whether this dose–response relationship indicates a causal relationship cannot be determined in the present study.

Thus, a limitation of the present study is the use of observational data. Possible explanations for the found association between depression and dementia include depression as a risk factor for dementia, and depression and dementia being associated based on meditating pathogenesis. In particular, there is increased risk of reverse causality for those diagnosed with dementia shortly after a depression diagnosis, where the association may reflect depressive symptoms as a prodrome of yet-undiagnosed dementia, or individuals who are evaluated for dementia also being diagnosed with depression due to more intense healthcare. Moreover, even though we controlled for civil status and household income, we were not able to account for additional socioeconomic factors, e.g., education level. Additionally, since the SNPR doesn’t include data from general practice, some of the individuals included in the study may have been diagnosed with either dementia or depression in a primary care setting. Additionally, a limitation of our study is that the results here may not necessarily generalize to individuals who were diagnosed in a primary care setting because the individuals included in the study were diagnosed in specialty care, and as such may represent more severe cases of depression/dementia. This may also explain the lower than expected rates of depression and dementia in our study, since many patients are diagnosed exclusively in primary care: Previous studies have indicated that the prevalence of both depression and dementia is around 10% in older community-dwelling individuals [40,41]. Nevertheless, our results distinctly reveal a long-lasting association between depression and subsequent risk of developing dementia, information that may be useful in the identification of individuals potentially at high risk for development of dementia in late life. A potential limitation in the present study is the use of ORs instead of risk ratios, since the use of ORs can result in non-collapsibility, i.e., that estimated ORs are not similar to the relative risks. However, this potential problem seems to be of minor importance as the estimated relative risks and ORs for the different time intervals were similar in the present study. Strengths of the present study include the use of clinically confirmed diagnoses of depression and dementia made in specialist care. Other strengths include the nationwide cohort studied, the long follow-up of up to 35 years, and the use of a sibling design to control for familial factors, all together increasing the external validity.

In conclusion, depression diagnosis is associated with increased odds of subsequent dementia diagnosis, even when the depression diagnosis occurs more than 20 years before the dementia diagnosis. The association was stronger for vascular dementia than for Alzheimer disease, which might suggest that depression contributes to the inflammatory process characteristic of atherosclerosis specifically in the brain. It would be of interest if future longitudinal studies could evaluate whether any inflammatory process associated with depression is reduced locally in the brain if depression is successfully treated, and whether the successful treatment of a depressive episode reduces the later risk of dementia.

(DOCX)

Click here for additional data file.

16 Oct 2019

Dear Dr. Nordström,

Thank you very much for submitting your manuscript "Depression and the risk of subsequent dementia diagnosis: A Nationwide Cohort Study" (PMEDICINE-D-19-03130) for consideration at PLOS Medicine.

Your paper was evaluated by a senior editor and discussed among all the editors here. It was also discussed with an academic editor with relevant expertise, and sent to three independent reviewers, including a statistical reviewer. The reviews are appended at the bottom of this email and any accompanying reviewer attachments can be seen via the link below:

[LINK]

In light of these reviews, I am afraid that we will not be able to accept the manuscript for publication in the journal in its current form, but we would like to consider a revised version that addresses the reviewers' and editors' comments. Obviously we cannot make any decision about publication until we have seen the revised manuscript and your response, and we plan to seek re-review by one or more of the reviewers.

In revising the manuscript for further consideration, your revisions should address the specific points made by each reviewer and the editors. Please also check the guidelines for revised papers at http://journals.plos.org/plosmedicine/s/revising-your-manuscript for any that apply to your paper. In your rebuttal letter you should indicate your response to the reviewers' and editors' comments, the changes you have made in the manuscript, and include either an excerpt of the revised text or the location (eg: page and line number) where each change can be found. Please submit a clean version of the paper as the main article file; a version with changes marked should be uploaded as a marked up manuscript.

In addition, we request that you upload any figures associated with your paper as individual TIF or EPS files with 300dpi resolution at resubmission; please read our figure guidelines for more information on our requirements: http://journals.plos.org/plosmedicine/s/figures. While revising your submission, please upload your figure files to the PACE digital diagnostic tool, https://pacev2.apexcovantage.com/. PACE helps ensure that figures meet PLOS requirements. To use PACE, you must first register as a user. Then, login and navigate to the UPLOAD tab, where you will find detailed instructions on how to use the tool. If you encounter any issues or have any questions when using PACE, please email us at PLOSMedicine@plos.org.

We expect to receive your revised manuscript by Nov 06 2019 11:59PM. Please email us (plosmedicine@plos.org) if you have any questions or concerns.

***Please note while forming your response, if your article is accepted, you may have the opportunity to make the peer review history publicly available. The record will include editor decision letters (with reviews) and your responses to reviewer comments. If eligible, we will contact you to opt in or out.***

We ask every co-author listed on the manuscript to fill in a contributing author statement, making sure to declare all competing interests. If any of the co-authors have not filled in the statement, we will remind them to do so when the paper is revised. If all statements are not completed in a timely fashion this could hold up the re-review process. If new competing interests are declared later in the revision process, this may also hold up the submission. Should there be a problem getting one of your co-authors to fill in a statement we will be in contact. YOU MUST NOT ADD OR REMOVE AUTHORS UNLESS YOU HAVE ALERTED THE EDITOR HANDLING THE MANUSCRIPT TO THE CHANGE AND THEY SPECIFICALLY HAVE AGREED TO IT. You can see our competing interests policy here: http://journals.plos.org/plosmedicine/s/competing-interests.

Please use the following link to submit the revised manuscript:

https://www.editorialmanager.com/pmedicine/

Your article can be found in the "Submissions Needing Revision" folder.

To enhance the reproducibility of your results, we recommend that you deposit your laboratory protocols in protocols.io, where a protocol can be assigned its own identifier (DOI) such that it can be cited independently in the future. For instructions see http://journals.plos.org/plosmedicine/s/submission-guidelines#loc-methods.

Please ensure that the paper adheres to the PLOS Data Availability Policy (see http://journals.plos.org/plosmedicine/s/data-availability), which requires that all data underlying the study's findings be provided in a repository or as Supporting Information. For data residing with a third party, authors are required to provide instructions with contact information for obtaining the data. PLOS journals do not allow statements supported by "data not shown" or "unpublished results." For such statements, authors must provide supporting data or cite public sources that include it.

We look forward to receiving your revised manuscript.

Sincerely,

Caitlin Moyer, Ph.D.

Associate Editor

PLOS Medicine

plosmedicine.org

-----------------------------------------------------------

Requests from the editors:

1. Did your study have a prospective protocol or analysis plan? Please state this (either way) early in the Methods section.

a) If a prospective analysis plan (from your funding proposal, IRB or other ethics committee submission, study protocol, or other planning document written before analyzing the data) was used in designing the study, please include the relevant prospectively written document with your revised manuscript as a Supporting Information file to be published alongside your study, and cite it in the Methods section. A legend for this file should be included at the end of your manuscript.

b) If no such document exists, please make sure that the Methods section transparently describes when analyses were planned, and when/why any data-driven changes to analyses took place.

c) In either case, changes in the analysis—including those made in response to peer review comments—should be identified as such in the Methods section of the paper, with rationale.

2. You have noted restrictions apply regarding the availability of your data. PLOS Medicine requires that the de-identified data underlying the specific results in a published article be made available, without restrictions on access, in a public repository or as Supporting Information at the time of article publication, provided it is legal and ethical to do so. Please see the policy at

http://journals.plos.org/plosmedicine/s/data-availability

and FAQs at

http://journals.plos.org/plosmedicine/s/data-availability#loc-faqs-for-data-policy

3. Please clearly state somewhere in the main text the primary outcome of the study, and secondary outcomes if applicable (e.g. diagnosis of dementia or death).

4. Title: Please change “risk of” to “association with” in the title, as this is an observational cohort study and we want to avoid any implication of causality.

5. Abstract Background: The final sentence should clearly state the study question.

6. Abstract Methods and Findings: Please define the years during which the study took place.

7. Abstract Methods and Findings: Please quantify the main results comparing dementia diagnoses in those with and without depression, and the sub-analyses, with p values in addition to the 95% CIs.

8. Abstract Methods and Findings: Please include the important dependent variables that are adjusted for in the analyses.

9. Abstract Methods and Findings: In the last sentence, please describe the main limitation(s) of the study’s methodology.

10. Methods: Please indicate the form of consent obtained, or the reason that consent was not obtained (e.g. the data were analyzed anonymously).

11. Methods: For all observational studies, in the manuscript text, please indicate: the specific hypotheses you intended to test, and when reported analyses differ from those that were planned, transparent explanations for differences that affect the reliability of the study's results. If a reported analysis was performed based on an interesting but unanticipated pattern in the data, please be clear that the analysis was data-driven.

12. Results: Please provide 95% CIs and p values associated with the analysis of differences between cohort demographic factors (data presented in Table 1).

13. Results: Please provide the p values in addition to the 95% CIs for the odds ratios of depression and risk of subsequent dementia diagnosis in the non-sibling cohort (Table 2 data).

14. Results: Please provide the p values in addition to the 95% CIs for the odds ratios of depression and risk of subsequent dementia diagnosis in the sibling cohort (Table 3 data).

15. Results: Please provide the p values in addition to the 95% CIs for the odd ratios of the associations between depression severity/type and dementia diagnosis (Table 4 and 5 data).

16. Results: For all the adjusted analyses described, please also provide the unadjusted analyses.

17. Figure 1: Please indicate in the figure caption the meaning of the black and gray lines. Please display the 95% CIs

18. Table 1: Please provide statistics to illustrate the factors for which the individuals with and without depression were significantly different, as stated in the Results.

19. Table 2 and Table 3: “Confident intervals” should be “confidence intervals” in the Table legend. Please provide p values along with the 95% CIs for all comparisons. Please define the abbreviations “OR” and “CI” used in the tables.

20. Table 4 and Table 5: For the adjusted analyses described, please also provide the unadjusted analyses. “Confident intervals” should be “confidence intervals” in the Table legend. Please define the abbreviations “OR” and “CI” used in the tables. Please provide p values along with the 95% CIs for all comparisons.

21. Table 5: The footnotes denoted by * and ** are repeated.

22. Please use the "Vancouver" style for reference formatting, and see our website for other reference guidelines:

https://journals.plos.org/plosmedicine/s/submission-guidelines#loc-references

23. Please ensure that the study is reported according to the STROBE guideline (or the most relevant guideline: http://www.equator-network.org/), and include the completed STROBE checklist as Supporting Information. When completing the checklist, please use section and paragraph numbers, rather than page numbers. Please add the following statement, or similar, to the Methods: "This study is reported as per the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) guideline (S1 Checklist)."

Comments from the reviewers:

Reviewer #1: The work uses routinely-collected data sources (EHR records from a country with UHI) to ascertain timing and incidence of dementia and depression, and then estimates the differences in risk of a record of dementia according to a previous diagnosis of depression. My role in the review process is as a statistical reviewer, and I have made some comments on the interpretation of the results with regard to the data sources as well. I can appreciate the work that has gone into this study, creating matched cohorts of this size from EHR data poses many challenges.

Statistical Issues:

The first statistical issue comes from the comparison of odds ratios between the general population cohort and the sibling-matched cohort. The incidence of the key outcome is different in the two populations (9,802/238,772 in the pop cohort, vs 1,161/50,644 in the sibling cohort). The odds ratio is not collapsible, so that a similar odds ratio between the two cohorts at the same time after diagnosis may actually reflect dissimilar relative risk (or a dissimilar odds ratio, and a similar relative risk). This issue affects comparison of ORs over time, and between single vs. multiple episodes, men vs. women etc. Estimating a rate ratio (estimated by log-binomial or log-Poisson with robust SE) is a way to avoid this issue, as is estimating relative risk from marginal estimates of risk (e.g. predictive margins in Stata).

The second statistical issue relates to the selection of covariates. What criteria was used for the selection of comorbidities subsequently used as adjustments for confounding? P-values for univariable association or effect size? Univariable pre-filtering can have unintended consequences - there is a good review by Heinz and Dunkler (Transplant International 2017; 30: 6-10) on this subject.

Data sources:

The foundation of this work is the quality of the data collection used to identify the outcome, exposure, and the potential confounders. Is there a key reference that about the timing and composition of data collection in the SNPR? It looks like it is composed of hospital and specialist care - is there general practice (family practice) collection included? Many of the PLoS Medicine readers won't be familiar with the health system behind the data collection and may need some context to understand what is accomplished in this work.

Was there missing data in the original data collection that precluded matching? How many records couldn't be used if so?

There is good evidence overall the SIR data accurately captures many conditions (from Ludvigsson et al 2011) and that previous validation studies have shown that a general dementia diagnosis is captured accurately. How well is depression exposure captured if the records are from acute (hospitalised) and specialist care episodes? In my country the majority of depression is treated in general practice (family practitioner) or by clinical psychologists and only severe cases of depression would be treated as an in-patient or by a psychiatrist. Is this a similar situation to Sweden? Are there any validation studies for PPV and sensitivity of timing of depression from the SNPR for depression of differing severity?

There is clearly a time-dependence, and the odds of dementia diagnosis are very high immediately following a diagnosis of depression. This is hinted at in the discussion but I think needs to be discussed clearly - could the depression actually an unrecognised sign or symptom of early dementia (potentially reversing the causal mechanism) or is it due to confounding according to more intense health service use increasing the diagnosis of both diseases?

Are there any known confounders of the depression-dementia relationship unavailable in the SNPR data? What level of confounding would be needed to account for the observed relationship? I have found the E-value is a useful tool (Ann Intern Med. 2017;167(4):268-274.) to consider how sensitive an observed outcome-exposure is to an unmeasured confounder.

Specific comments:

p2, Conclusions. Logistic regression is used throughout this manuscript, so 'increased odds' of dementia is a more accurate description of the results.

p2, Conclusions. The phrasing part of this sentence 'of especially vascular dementia' is unclear. I would either just refer to dementia generally or reword this sentence to include the additional information about vascular dementia.

P15, Household income. Is there a way to indicate what the relative difference in household income is for international readers (i.e. converted to USD/Euros or an indication of household income distribution in Sweden at the time of the data collection)? The difference in household income between those diagnosed with depression and not diagnosed with depression in the sibling cohort looks extreme - but this could just be my unfamiliarity with the currency.

P16, Title of table. Although it is possible to identify the source of the data from the reported N, it would be easier if the source of the data (e.g. Population matched cohort) was also included as in Table 3.

P19, Table 5. There is a 'ns' indicator for single depressive episode and 10-20 years after diagnosis, but this doesn't appear elsewhere in the table where the 95% CI overlaps with 1. I would suggest to just remove the 'ns' as the CI already gives this information.

Reviewer #2: PMEDICINE-D-19-03130 "Depression and the risk of subsequent dementia diagnosis: A Nationwide Cohort Study"

In this study the authors examined the association between depression and dementia and reported depression to be associated with "an increased risk of especially vascular dementia, even

more than 20 years after the depression, and the association remained after

adjustment for familiar factors."

The data in the study come from two cohorts formed of all individuals above 50 years, living in Sweden. The first cohort was composed of individuals diagnosed with depression matched to controls without depression (n=238,772). The second was a co-sibling cohort (n=50,644) consisting of same-sex full sibling pairs with discordant depression status. This is an impressive study design, and the size of the study sample would lead us to believe that the results would be robust. Nevertheless, the following points need further reflection.

1.Novelty

The authors say

"The association between depression and subsequent dementia diagnosis has been established

in at least four independent meta-analysis indicating close to a twofold risk of developing

dementia followed by depression [7-10], as well as several reviews [11-14]."

Given the meta-analyses and reviews on this topic, the contribution of this study is unclear.

2.Depression diagnosis

Diagnosis of depression came from the Swedish National Patient Register. The manuscript provides the ICD codes for diagnosis but no detail on the register itself. Further information is required on the sources of data used in this register, in terms of consultations - family doctor, psychiatrists, hospitals? Is the diagnosis based on prescription of antidepressants or hospitalization for depression? This information is needed to understand what is being measured by the ICD codes. The target population was 3.3M and only 119,386 cases of depression were identified - this is a surprisingly low rate.

3.Matched controls

In the first cohort, controls were matched for age, sex, and citizenship but Table 1 shows cases and controls to differ systematically on sociodemographic and health status. It is possible that the results can be explained by systematic differences in these two groups that have not much to do with depression.

4.Dementia diagnosis

Further details are required on how dementia was diagnosed in both cohorts. Do the data on dementia come primarily from mortality registers? Both cohorts in the manuscript have a follow-up of over 20 years but the rate of dementia is 4.1% and 2.3%. This is again surprising low. Is it possible that most cases of dementia were missed in the register?

5.Conclusion

I do not think that the conclusion "The results of the present nationwide study indicated an increased risk for dementia diagnosis after depression, even when the depression occurred 20 years or more before the dementia diagnosis" is supported by the data presented in the manuscript. Figure 1 clearly shows reverse causation (or depression being a prodromal feature of dementia) as the excess risk drops at Year 5 of follow-up and remains unchanged thereafter. There are problems with measurement of exposure and the outcome, and the selection of the control group. Furthermore, the analysis of the 20-year follow up is based on a small number of cases (481 cases in 26,044 persons in cohort 1, and 76 cases in 6,296 persons in cohort 2), making it difficult to draw conclusions.

Reviewer #3: This study on risk of dementia among older depressed adults should be of substantial interest to readers of the journal.

In the Introduction, the authors note the high prevalence of dementia and the need to identify modifiable dementia risk factors. They briefly cite literature noting the increased risk of dementia among those with depression, noting that depression may be seen as a discrete risk factor or as a dementia prodrome.

In the final introductory paragraph, the authors note that the purpose of the study was to investigate if the association between depression and later dementia risk is time dependent, identify factors that influence the association, and whether the association is influenced by severity and frequency of depression. Given their prior literature review in the preceding paragraphs, one wonders whether the authors have any a priori hypotheses in these three areas, or if this is truly a completely exploratory study, which would require appropriate statistical correction. If the authors did have a priori hypotheses, they should state them.

In the Materials and Methods section, the authors describe reasonably well how the cohorts for the present study were derived. They refer to “depression” as opposed to “major depression.” If indeed they are focusing on major depression, they should be more precise. This is also the case for the next section, “Procedure,” in which they specify ICD-10 codes of F32 and F33 but only mention “diagnosis of depression.”

They note that “all diagnoses were set in specialist care.” It is not clear what that means.

In the statistical methods section, the authors quickly jump to time dependency analyses without first indicating what initial analysis, if any, that they undertook comparing depression and dementia risk. They should specify how the dependent and independent variables were constructed. They should justify why separate analyses were performed for men and women, as opposed to, for example, doing separate analyses for different age cohorts, given differences in typical ages of onset for Alzheimer’s disease and vascular dementia.

It is not clear what the following sentence means: The knots were set in default positions.

If this is an exploratory study without hypotheses, then it is not clear that appropriate statistical corrections were employed.

The results, in the text, tables and figures, are clearly presented.

In the Discussion, the authors summarize some of the main findings, but should mention their findings related to severity as well.

They then place their study in the context of prior literature. There are a few key studies that they should consider including:

Barnes et al. Arch Gen Psychiatry. 2006;63(3):273-279. Study focusing on depression as risk for MCI

Steffens et al. Biol Psychiatry. 1997;41(8):851-856. Large epidemiological study noting depression as risk for AD only when there is a few years between depression onset and dementia onset

Kokmen et al. Neurology. 1991;41(9):1393-1397. Large population study linking episodic depression and AD.

Becker et al. Am J Geriatr Psychiatry. 2009;17(8):653-663. A community based study that found that depressed mood is not a risk factor for incident dementia.

Steffens et al. Am J Geriatr Psychiatry. 2007 Oct;15(10):839-49. Clinical study noting depression as risk for non-AD dementias.

Inclusion of some of these and other studies might allow for a fuller discussion of their findings.

The authors identify appropriate limitations in their study. If this truly was an exploratory study (without a priori hypotheses), that might also be considered a limitation.

Any attachments provided with reviews can be seen via the following link:

[LINK]

31 Oct 2019

Submitted filename: Authors response_final.docx

Click here for additional data file.

25 Nov 2019

Dear Dr. Nordström,

Thank you very much for re-submitting your manuscript "Depression and the association with subsequent dementia diagnosis: A Nationwide Cohort Study" (PMEDICINE-D-19-03130R1) for review by PLOS Medicine.

I have discussed the paper with my colleagues and the academic editor and it was also seen again by three reviewers. I am pleased to say that provided the remaining editorial and production issues are dealt with we are planning to accept the paper for publication in the journal.

The remaining issues that need to be addressed are listed at the end of this email. Any accompanying reviewer attachments can be seen via the link below. Please take these into account before resubmitting your manuscript:

[LINK]

Our publications team (plosmedicine@plos.org) will be in touch shortly about the production requirements for your paper, and the link and deadline for resubmission. DO NOT RESUBMIT BEFORE YOU'VE RECEIVED THE PRODUCTION REQUIREMENTS.

***Please note while forming your response, if your article is accepted, you may have the opportunity to make the peer review history publicly available. The record will include editor decision letters (with reviews) and your responses to reviewer comments. If eligible, we will contact you to opt in or out.***

In revising the manuscript for further consideration here, please ensure you address the specific points made by each reviewer and the editors. In your rebuttal letter you should indicate your response to the reviewers' and editors' comments and the changes you have made in the manuscript. Please submit a clean version of the paper as the main article file. A version with changes marked must also be uploaded as a marked up manuscript file.

Please also check the guidelines for revised papers at http://journals.plos.org/plosmedicine/s/revising-your-manuscript for any that apply to your paper. If you haven't already, we ask that you provide a short, non-technical Author Summary of your research to make findings accessible to a wide audience that includes both scientists and non-scientists. The Author Summary should immediately follow the Abstract in your revised manuscript. This text is subject to editorial change and should be distinct from the scientific abstract.

We expect to receive your revised manuscript within 1 week. Please email us (plosmedicine@plos.org) if you have any questions or concerns.

We ask every co-author listed on the manuscript to fill in a contributing author statement. If any of the co-authors have not filled in the statement, we will remind them to do so when the paper is revised. If all statements are not completed in a timely fashion this could hold up the re-review process. Should there be a problem getting one of your co-authors to fill in a statement we will be in contact. YOU MUST NOT ADD OR REMOVE AUTHORS UNLESS YOU HAVE ALERTED THE EDITOR HANDLING THE MANUSCRIPT TO THE CHANGE AND THEY SPECIFICALLY HAVE AGREED TO IT.

Please ensure that the paper adheres to the PLOS Data Availability Policy (see http://journals.plos.org/plosmedicine/s/data-availability), which requires that all data underlying the study's findings be provided in a repository or as Supporting Information. For data residing with a third party, authors are required to provide instructions with contact information for obtaining the data. PLOS journals do not allow statements supported by "data not shown" or "unpublished results." For such statements, authors must provide supporting data or cite public sources that include it.

If you have any questions in the meantime, please contact me or the journal staff on plosmedicine@plos.org.

We look forward to receiving the revised manuscript by Dec 02 2019 11:59PM.

Sincerely,

Caitlin Moyer, Ph.D.

Associate Editor

PLOS Medicine

plosmedicine.org

------------------------------------------------------------

Requests from Editors:

1.Response to reviewers: Thank you for your response to Reviewer 1, point 3. However, please include in the Discussion section (lines 329-344) as a limitation the fact that because general/family practice data are not included in the SNPR collection, there is a possibility that some of the individuals included in the study may have been diagnosed with dementia/depression in a primary care setting. Please also include in your discussion that the fact that the diagnoses were made in specialty care may speak to lower than expected rates of dementia/depression in the cohorts (in reference to the points brought up by Reviewer 2).

2.Data Availability Statement: Thank you for your response regarding your limitations on making the study data available. If the data are owned by a third party but freely available upon request, please note this and state the owner of the data set and contact information for data requests (web or email address). Note that a study author cannot be the contact person for the data. If the data are not freely available, please describe briefly the ethical, legal, or contractual restriction that prevents you from sharing it. Please also include an appropriate contact (web or email address) for inquiries (again, this cannot be a study author).

3.Title: Please include: “A Swedish nationwide cohort study” to reflect the setting where the study took place, and also please include the dates of the study in the title.

4.Author summary: Please use bullets for each point in the Author Summary.

5.Author summary: “Why was this study done?”: In the first point, please clarify what is meant by “...leading cause of dependency…”

6.Author summary: “What did the researchers do and find?”: Lines 87-89, please clarify to “...after the diagnosis of depression” in both sentences.

7.Author summary: “What did the researchers do and find?”: Line 91: Please delete the word “one”

8.Author summary: “What do these findings mean?”: Please revise “after a depression” to “after the diagnosis of depression”

9. Introduction: Line 110: Please revise “...has been established…” to “...has been suggested…” or similar, because it is difficult to say something was established in a meta-analysis or review.

10.Methods, and Table 1: Please clarify in the methods for the sibling cohort whether some sibling pairs were discordant for citizenship (as indicated by the mismatched counts for Swedish and non-Swedish citizenship in Table 1 and the chi-squared test.

11.Methods: Lines 219-220: Please add the ethical approval number (e.g. Nr 2013-86-31) to the sentence describing the ethical approval.

12.Results: Line 238: Please also provide the standard deviation for the mean follow-up time.

13.Discussion: Paragraph 1: Line 270: Please revise to “...while adjusting for a set of covariates, including…” and deleting the phrase “a rich set” as this does not convey any specific information regarding the covariates.

14.Discussion: Line 272 (and throughout discussion where the term “significant” is used): Please clarify what you mean by "significant". If statistical significance is intended, please indicate that.

15.Table 1: Please define the abbreviation “SD” in the table legend.

16.References: Please use the "Vancouver" style for reference formatting, and see our website for other reference guidelines: https://journals.plos.org/plosmedicine/s/submission-guidelines#loc-references

17.Thank you for including the STROBE checklist. Please revise the checklist, using section and paragraph numbers, rather than page numbers, to refer to locations of checklist items.

Comments from Reviewers:

Reviewer #1: Thank you for the revised manuscript and responses you have given to mine and the other reviewers comments.

The main areas of concern from the original manuscript were the non-collapsibility of odds ratios, selection of covariates, and the suitability of the data for the research questions.

The selection of covariates has been clarified and this no longer a concern. The limitations to the data has been more clearly explained, the one area I feel that needs to be articulated is that the measurement of depression is most likely to be limited to acute cases, and so that only severe depression may be a risk factor for subsequent development of dementia. There is a note in the methods about the recording of major depression may exclude primary care visits - it should be made clear that the exposure measured here is the record of a diagnosis of major depression by a specialist and so is likely to be a record of severe cases. The amount of data excluded for missingness is small and not of concern given the final sizes of the cohorts.

You are correct that the issue with non-collapsibility is more severe as prevalence increases. Adding the incidence to T2 and T3 is reassuring and it does appear the non-collapsibility will have a limited impact here, and the appropriate terminology (risk -> odds) is used. I personally prefer that a risk ratio be estimated as it makes for a clearer evidence, but I think the changes are sufficient that if this is added as a minor limitation to the manuscript I think it is acceptable to present ORs over RRs.

Reviewer #2: "Depression and the association with subsequent dementia diagnosis: A Nationwide Cohort Study" # PMEDICINE-D-19-03130R1

I have the same concerns with the revised manuscript as that with the previous version.

1.Novelty

There are several meta-analyses on the association between depression and dementia, the authors do not make a clear case for what is new in yet another study.

Re reverse causation, the authors ought to look at this paper, https://www.ncbi.nlm.nih.gov/pubmed/28514478

2.Depression diagnosis

The authors say that depression rates are low because diagnosis of depression is drawn from "specialist care". What are the implications for generalizability? I would also recommend that this detail be indicated in the title, abstract, and conclusions of the paper.

3.Dementia diagnosis

The low rate of dementia is explained by the authors as being due to diagnosis in "specialist care". The implications for generalizability ought to be discussed in the limitations section.

4.Conclusion

I do not think that the conclusion "…depression diagnosis is significantly associated with increased risk of dementia diagnosis, even when depression occurs more than 20 years before the dementia diagnosis" is supported by the data presented in the manuscript. Figure 1 clearly shows reverse causation (or depression being a prodromal feature of dementia) as the excess risk drops at Year 5 of follow-up and remains unchanged thereafter.

Reviewer #3: The authors have addressed my concerns.

Any attachments provided with reviews can be seen via the following link:

[LINK]

11 Dec 2019

Submitted filename: Authors_response_191210.docx

Click here for additional data file.

13 Dec 2019

Dear Dr. Nordström,

On behalf of my colleagues and the academic editor, Dr. Carol Brayne, I am delighted to inform you that your manuscript entitled "Depression and the association with subsequent dementia diagnosis: A Swedish Nationwide Cohort Study from 1964 to 2016" (PMEDICINE-D-19-03130R2) has been accepted for publication in PLOS Medicine.

PRODUCTION PROCESS

Before publication you will see the copyedited word document (in around 1-2 weeks from now) and a PDF galley proof shortly after that. The copyeditor will be in touch shortly before sending you the copyedited Word document. We will make some revisions at the copyediting stage to conform to our general style, and for clarification. When you receive this version you should check and revise it very carefully, including figures, tables, references, and supporting information, because corrections at the next stage (proofs) will be strictly limited to (1) errors in author names or affiliations, (2) errors of scientific fact that would cause misunderstandings to readers, and (3) printer's (introduced) errors.

If you are likely to be away when either this document or the proof is sent, please ensure we have contact information of a second person, as we will need you to respond quickly at each point.

PRESS

A selection of our articles each week are press released by the journal. You will be contacted nearer the time if we are press releasing your article in order to approve the content and check the contact information for journalists is correct. If your institution or institutions have a press office, please notify them about your upcoming paper at this point, to enable them to help maximize its impact.

PROFILE INFORMATION

Now that your manuscript has been accepted, please log into EM and update your profile. Go to https://www.editorialmanager.com/pmedicine, log in, and click on the "Update My Information" link at the top of the page. Please update your user information to ensure an efficient production and billing process.

Thank you again for submitting the manuscript to PLOS Medicine. We look forward to publishing it.

Best wishes,

Caitlin Moyer, Ph.D.

Associate Editor

PLOS Medicine

plosmedicine.org