Which Severe Mental Illnesses Most Increase the Risk of Developing Dementia? Comparing the Risk of Dementia in Patients with Schizophrenia, Major Depressive Disorder and Bipolar Disorder

Abstract

Objective

Previous studies have shown that certain severe mental illnesses (SMIs) increase the risk of dementia, but those that increase the risk to a greater degree in comparison with other SMIs are unknown. Furthermore, physical illnesses may alter the risk of developing dementia, but these cannot be well-controlled.

Methods

Using the Taiwan National Health Insurance Research Database, patients with schizophrenia, bipolar disorder and major depressive disorder (MDD) were recruited. We also recruited normal healthy subjects as the control group. All subjects were aged over 60 years, and the duration of follow-up was from 2008 to 2015. Multiple confounders were adjusted, including physical illnesses and other variables. Use of medications, especially benzodiazepines, was analyzed in a sensitivity analysis.

Results

36,029 subjects (MDD 23,371, bipolar disorder 4,883, schizophrenia 7,775) and 108,084 control subjects were recruited after matching according to age and sex. The results showed that bipolar disorder had the highest hazard ratio (HR) (HR 2.14, 95% confidence interval [CI] 1.99−2.30), followed by schizophrenia (HR 2.06, 95% CI 1.93−2.19) and MDD (HR 1.60, 95% CI 1.51−1.69). The results remained robust after adjusting for covariates, and sensitivity analysis showed similar results. Anxiolytics use did not increase the risk of dementia in any of the three groups of SMI patients.

Conclusion

SMIs increase the risk of dementia, and among them, bipolar disorder confers the greatest risk of developing dementia. Anxiolytics may not increase the risk of developing dementia in patients with an SMI, but still need to be used with caution in clinical practices.

INTRODUCTION

The pathogenesis of major mental illnesses mostly originates from certain neurological deficits, including monoamine dysregulation [1], neuroplasticity impairment [2], cell damage by cytokines or free radicals [3], and even protein misassembly and aggregation [4]. Cognitive impairment in psychiatric illnesses, such as memory and executive functions, has also been commonly reported in many studies [5-7]. In addition, several major mental disorders have been found to increase the risk of dementia, as reported in recent literature [8-13]. For example, patients with schizophrenia have altered brain volumes in specific regions, and exhibit cognitive dysfunction even in a drug-naïve state [14]; these patients also have an increased risk of developing dementia [13]. Furthermore, patients with bipolar disorder, whether under a stable condition or within an acute mood episode [15], have altered cognitive function, eventually leading to permanent neurocognitive disorder [9]. Long-term mood disorder disturbance also leads to a higher risk of dementia. Major depressive disorder (MDD) is a well-known risk factor for dementia [16], and studies have proved a bidirectional relationship between MDD and dementia [17].

Besides psychiatric illnesses, many physical illnesses alter the risk of dementia. For example, cardiovascular disease [18], metabolic disorders [19] and even medication use [20] have been reported to be correlated with the risk of dementia. Previous research has debated hypnotics use [20], but inconsistent findings regarding correlations between benzodiazepines (BZD; hypnotics/anxiolytics) and dementia have been reported [21]. A number of studies observed insignificant correlations between BZD use and dementia in specific psychiatric illnesses, but a comprehensive comparison of the correlations between the use of hypnotics and the risk of dementia in different kinds of major psychiatric illnesses is still lacking. As a large number of variables alter the risk of dementia, and may mask the relationships between psychiatric disorders and dementia, a well-designed and variable-controlled study is needed in order to identify the risk of dementia in major psychiatric illnesses. We hypothesized that only psychiatric illnesses themselves increase the risk of dementia. Therefore, our study focused on exploring the risk of dementia in three severe mental illnesses (SMIs), schizophrenia, MDD and bipolar disorder, and comparing the risk with that in normal elderly subjects. In addition, we examined whether the usage of BZD increases the risk of dementia in both the SMI groups and in elderly healthy subjects.

METHODS

Data Source

The research protocol was approved by the Institutional Review Board (IRB) of National Cheng Kung University Hospital (IRB no. B-ER-106-074). This study employed de- identified secondary data from National Health Insurance Database (NHID) and thus was exempted from informed consent.

The Taiwan NHID and Cause of Death database were used in this study. These databases are maintained by the Health and Welfare Data Science Center, Ministry of Health and Welfare, Taiwan, and are linked at individual level by the personal identification number (PIN) [22].

The NHID is derived from a universal health insurance program in Taiwan, which covers 99% of the entire population; it therefore exhibits representativeness and longitudinal follow-up properties for population-based study. The NHID includes a registry of beneficiaries, ambulatory care claims, inpatient claims, and prescription dispensing claims of pharmacies. Each claim contains diagnosis codes according to the International Classification of Dis-eases, Ninth Edition, Clinical Modification (ICD-9-CM, up to 2015) and Tenth Edition (ICD-10-CM, after 1/1/2016). The medication codes were according to National Health Insurance codes and the Anatomical Therapeutic Chemi-cal (ATC) Classification System for all reimbursed pharmaceutical products. Details of medication issue, such as date of prescription and days of supply, are also included in the NHID. Cause of Death data contain the PIN, date of death, and cause of death, which is recorded by applying ICD diagnosis codes (ICD-9-CM to 2008; ICD-10-CM after 2008). For details of the NHID and Cause of Death data, refer to a review paper published by Hsieh et al. [22].

Design, Setting and Cohort Identification

We conducted a retrospective cohort study to identify a cohort of patients diagnosed by psychiatrists with SMIs, including bipolar disorder (ICD-9-CM: 296.4-296.5, 296.7, 296.8), MDD (ICD-9-CM: 296.2, 296.3) and schizophrenia (ICD-9-CM: 295), from the NHID in 2009. The index date was defined as the first visiting day of patients who already had been diagnosed before 2009. Patients were excluded if they had no record of date of birth or sex, were aged below 60 years, had two or more diagnoses of bipolar disorder, MDD, and schizophrenia, or received a diagnosis of dementia during the one-year period prior to the index date. The control group was selected from patients who made outpatient visits who did not have an SMI diagnosis in 2009; they were excluded if they had no record of date of birth or sex, had a diagnosis of any mental disorder (ICD-9-CM: 290-319) (Supplementary Table 1), or had a diagnosis of dementia during the one-year period prior to the index date. The control group then was matched to the SMI group at a 3:1 ratio in terms of age and sex (Fig. 1).

Fig. 1.

Algorithm of the study.

Outcomes, Follow-up and Covariates

The major outcome of this study was the incidence of dementia (ICD-9-CM: 290, 294, 331.0-331.2, 331.7-331.9). Follow-up began from the index date and was censored at death after the index date or at the end of the study period (2015), whichever came first. The baseline covariates were obtained from the NHID during the 1-year pre-index date period, including demographics, comorbidities and concomitant medications. Details of the baseline covariates are presented in Table 1.

Table 1.

Baseline characteristics of the study population

Variable	Controlgroup	Mental illness cohort
		Major depressive disorder	Bipolar disorder	Schizophrenia
Total (n)	108,084	23,371	4,883	7,775
Demographics
Age (yr)	68.7 ± 7.3	69.7 ± 7.4	68.6 ± 7.2	65.9 ± 6.1
60−64	39,440 (36.5)	7,075 (30.3)	1,775 (36.4)	4,297 (55.3)
65−69	25,440 (23.5)	5,576 (23.9)	1,223 (25.0)	1,681 (21.6)
70−74	19,482 (18.0)	4,668 (20.0)	855 (17.5)	971 (12.5)
75−79	12,875 (11.9)	3,227 (13.8)	549 (11.2)	516 (6.6)
≥ 80	10,847 (10.0)	2,825 (12.1)	481 (9.9)	310 (4.0)
Sex
Female	69,173 (64)	15,664 (67)	2,769 (56.7)	4,625 (59.5)
Male	38,911 (36)	7,707 (33)	2,114 (43.3)	3,150 (40.5)
Comorbidities
Substance use	0 (0.0)	355 (1.5)	123 (2.5)	72 (0.9)
Anxiety disorder (without OCD)	0 (0.0)	8,605 (36.8)	1,399 (28.7)	654 (8.4)
OCD	0 (0.0)	279 (1.2)	56 (1.1)	58 (0.7)
Hypertension	37,496 (34.7)	13,004 (55.6)	2,426 (49.7)	2,701 (34.7)
Hyperlipidemia	17,074 (15.8)	6,750 (28.9)	1,275 (26.1)	1,155 (14.9)
Stroke	6,021 (5.6)	3,321 (14.2)	612 (12.5)	593 (7.6)
Diabetes mellitus	15,347 (14.2)	6,010 (25.7)	1,323 (27.1)	1,639 (21.1)
Parkinson’s disease	598 (0.6)	1,139 (4.9)	291 (6)	546 (7.0)
Epilepsy	290 (0.3)	205 (0.9)	84 (1.7)	132 (1.7)
Chronic obstructive pulmonary disease	5,712 (5.3)	2,704 (11.6)	541 (11.1)	862 (11.1)
Traumatic brain injury	1,142 (1.1)	656 (2.8)	154 (3.2)	149 (1.9)
Co-medication
Anxiolytics	11,538 (10.7)	19,831 (84.9)	3,624 (74.2)	3,784 (48.7)
Hypnotics and sedatives	3,365 (3.1)	17,974 (76.9)	3,675 (75.3)	4,515 (58.1)
RAAS inhibitors	6,612 (6.1)	7,606 (32.5)	1,445 (29.6)	1,536 (19.8)
Beta-blocker agents	6,513 (6.0)	11,195 (47.9)	2,019 (41.3)	2,172 (27.9)
CCBs	9,727 (9.0)	10,560 (45.2)	1,921 (39.3)	2,068 (26.6)
Diuretics	4,702 (4.4)	5,543 (23.7)	1,119 (22.9)	1,117 (14.4)
Statins	3,723 (3.4)	4,946 (21.2)	866 (17.7)	711 (9.1)
Non-statin lipid lowering agents	1,140 (1.1)	1,460 (6.2)	284 (5.8)	313 (4.0)

Values are presented as mean ± standard deviation or number (%).

OCD, obsessive compulsive disorder; RAAS, renin-angiotensin-aldosterone system; CCBs, calcium channel blockers.

Sensitivity Analyses

In order to confirm that our conclusion was not biased by the study population selection, we restricted patients with bipolar disorder to those who were being treated with lithium (ATC code: N05AN01), and compared the risk of developing dementia with the risks of the other groups.

Statistical Analyses

Descriptive statistics were used to summarize the baseline characteristics of the study cohort. Continuous variables were described as the mean with standard deviation, and categorical variables were presented as the number and proportion. The distribution of duration to developing dementia from the index date was estimated using the Kaplan–Meier method. In order to consider that “death” could be a competing risk that would bias the estimation, hazard ratios (HRs) for developing dementia were estimated using Fine and Gray’s sub-distribution hazard model [23] in the SMI groups and control group. To reduce the effects of confounding factors on the associations between SMIs and the risk of developing dementia, multiple variables regression adjustment was employed; the variables included in the regression model are presented in Table 1. According to previous studies [24,25], BZD and other related drugs could increase the risk of dementia. Thus, we further examined the effects of the interactions between concomitant anxiolytic use and SMIs on the risk of developing dementia. All significance levels were two-sided; p < 0.05. Statistical analyses were performed using SAS software (version 9.4 for Windows; SAS Institute Inc.).

RESULTS

Demographic Details of Healthy Subjects and Patients with Severe Mental Illnesses

Table 1 presents the basic demographic data of the patients, including the sex and age distributions. The propor-tions of patients with psychiatric comorbidities, physical illnesses and medications use are also listed in Table 1. Higher proportions of the patients with MDD had anxiety disorder (36.8%), hypertension (55.6%), stroke (14.2%), use of anxiolytics (84.9%), renin-angiotensin-aldosterone system inhibitor use (32.5%), beta-blocker agent use (47.9%) and statin use (21.2%) as compared with the other groups. A higher proportion of the patients with bipolar disorder had diabetes mellitus (27.1%) as compared with the other groups.

Risk of Developing Dementia in Different Groups

Table 2 presents a pairwise comparison of HRs for developing dementia in the patients with MDD, bipolar disorder, and schizophrenia. The results showed that bipolar disorder had the highest unadjusted HR, whichever disorder was used as the reference group (for example, HR: 4.33, 95% confidence interval [CI]: 4.10−4.58 when the control group was used as the reference group) (Fig. 2). The findings remained consistent after adjusting for sex, age, comorbidities and co-medications (adjusted HR [aHR]: 2.13, 95% CI: 1.98−2.30 when the control group was used as the reference group) (Table 2).

Table 2.

Pairwise comparison of hazard ratios for the development of dementia in patients with major depressive disorder, bipolar disorder, and schizophrenia

Comparison	Main analysis						Sensitivity analysis
	Number	Person- years	Incidence (per 1,000 person-years)	Crude HR (95% CI)	Adjusted HR (95% CI)		Number	Person- years	Incidence (per 1,000 person-years)	Crude HR (95% CI)	Adjusted HR (95% CI)
Ref. = control group	9,284	663,672	14	1.00	1.00		9,284	663,672	14	1.00	1.00
Major depressive disorder	6,776	120,167	56	3.92 (3.80−4.04)	1.59 (1.5−1.68)		6,776	120,167	56	3.92 (3.80−4.05)	1.52 (1.43−1.61)
Bipolar disorder	1,531	24,403	63	4.33 (4.10−4.58)	2.13 (1.98−2.30)		275	4,792	57	3.99 (3.53−4.50)	2.52 (2.20−2.89)
Schizophrenia	1,657	41,700	40	2.68 (2.54−2.82)	2.05 (1.92−2.19)		1,657	41,700	40	2.68 (2.54−2.82)	2.01 (1.89−2.15)
Ref. = major depressive disorder	6,776	120,167	56	1.00	1.00		6,776	120,167	56	1.00	1.00
Bipolar disorder	1,531	24,403	63	1.11 (1.05−1.17)	1.34 (1.26−1.43)		275	4,792	57	1.02 (0.90−1.15)	1.66 (1.46−1.89)
Schizophrenia	1,657	41,700	40	0.68 (0.65−0.72)	1.29 (1.21−1.37)		1,657	41,700	40	0.68 (0.65−0.72)	1.33 (1.25−1.41)
Control group	9,284	663,672	14	0.26 (0.25−0.26)	0.63 (0.59−0.67)		9,284	663,672	14	0.26 (0.25−0.26)	0.66 (0.62−0.70)
Ref. = bipolar disorder	1,531	24,403	63	1.00	1.00		275	4,792	57	1.00	1.00
Major depressive disorder	6,776	120,167	56	0.90 (0.85−0.96)	0.75 (0.70−0.79)		6,776	120,167	56	0.25 (0.22−0.28)	0.6 (0.53−0.69)
Schizophrenia	1,657	41,700	40	0.62 (0.58−0.66)	0.96 (0.89−1.04)		1,657	41,700	40	0.98 (0.87−1.11)	0.8 (0.70−0.92)
Control group	9,284	663,672	14	0.23 (0.22−0.24)	0.47 (0.44−0.51)		9,284	663,672	14	0.67 (0.59−0.76)	0.4 (0.35−0.45)
Ref. = schizophrenia	1,657	41,700	40	1.00	1.00		1,657	41,700	40	1.00	1.00
Major depressive disorder	6,776	120,167	56	1.46 (1.39−1.55)	0.78 (0.73−0.83)		6,776	120,167	56	1.47 (1.39−1.55)	0.75 (0.71−0.80)
Bipolar disorder	1,531	24,403	63	1.62 (1.51−1.74)	1.04 (0.96−1.12)		275	4,792	57	1.49 (1.31−1.69)	1.25 (1.09−1.44)
Control group	9,284	663,672	14	0.37 (0.36−0.39)	0.49 (0.46−0.52)		9,284	663,672	14	0.37 (0.35−0.39)	0.50 (0.47−0.53)

Crude HR: unadjusted hazard ratio (HR). Adjusted HR: hazard ratio adjusted for age, sex, substance use, anxiety disorder, obsessive compulsive disorder, hypertension, hyperlipidemia, stroke, diabetes mellitus, Parkinson’s disease, epilepsy, chronic obstructive pulmonary disease, traumatic brain injury, anxiolytics, hypnotics and sedatives, renin-angiotensin-aldosterone system inhibitors, beta-blocker agents, calcium channel blockers, diuretics, statins and non-statin lipid lowering agents.

Ref., reference; CI, confidence interval.

Fig. 2.

The Kaplan–Meier curves for comparing the risk of dementia among patients with severe mental illnesses.

Sensitivity analysis produced similar results. Regarding MDD, the crude HR was higher than that for schizophrenia (MDD: 3.92, 95% CI: 3.80−4.04; schizophrenia: 2.68, 95% CI: 2.54−2.82) but the adjusted HR for schizophrenia leading to dementia was significantly higher when adjusting for multiple variables (MDD: 1.59, 95% CI: 1.5−1.68; schizophrenia: 2.05, 95% CI: 1.92−2.19). Furthermore, the patients with bipolar disorder were at significantly greater risk of developing dementia (aHR: 1.25, 95% CI: 1.09−1.44) as compared with the patients with schizophrenia (Table 2).

Risk of Developing Dementia When Considering Anxiolytic Use

Considering the effects of anxiolytics/hypnotics use on the risk of dementia, the results presented in Table 3 showed that all three SMIs resulted in a non-significant higher risk when comparing the non-BZD group and the BZD use group (MDD group - aHR: 0.99, 95% CI: 0.86−1.14; bipolar group - aHR: 1.01, 95% CI: 0.84−1.22; schizophrenia group - aHR: 1.17, 95% CI: 0.99−1.37). The results remained consistent when comparing the SMI groups with the control group, indicating that BZD use has a weak effect on the risk of dementia in patients with SMI, and BZD use in the MDD group did not have a significant effect on the risk of dementia.

Table 3.

Effects of concomitant anxiolytic use with major depressive disorder, bipolar disorder and schizophrenia on the development of dementia

Comparisons		Non-anxiolytic use			Anxiolytic use		Adjusted HR (95% CI) for anxiolytic vs. non- anxiolytic use within strata of mental disorders
		Number with outcome	Adjusted HR (95% CI)		Number with outcome	Adjusted HR (95% CI)
Control group		6,601	1.00 (reference)		2,683	2.66 (2.52−2.80)	2.66 (2.52−2.80)
Major depressive disorder		1,009	2.96 (2.73−3.22)		5,767	2.93 (2.34−3.66)	0.99 (0.86−1.14)
Bipolar disorder		372	3.72 (3.29−4.19)		1,159	3.76 (2.76−5.12)	1.01 (0.84−1.22)
Schizophrenia		766	2.94 (2.71−3.19)		891	3.43 (2.68−4.39)	1.17 (0.99−1.37)
Adjusted HR (95% CI) for mental disorder vs. control group within strata of anxiolytic use	Major depressive disorder vs. control group	2.96 (2.73−3.22)			1.10 (0.93−1.31)
	Bipolar disorder vs. control group	3.72 (3.29−4.19)			1.41 (1.09−1.83)
	Schizophrenia vs. control group	2.94 (2.71−3.19)			1.29 (1.06−1.57)

Adjusted HR: hazard ratio (HR) adjusted for age, sex, substance use, anxiety disorder, obsessive compulsive disorder, hypertension, hyperlipidemia, stroke, diabetes mellitus, Parkinson’s disease, epilepsy, chronic obstructive pulmonary disease, traumatic brain injury, anxiolytics, hypnotics and sedatives, renin-angiotensin-aldosterone system inhibitors, beta-blocker agents, calcium channel blockers, diuretics, statins and non-statin lipid lowering agents.

CI, confidence interval.

DISCUSSION

In this study, the risk of developing dementia in patients with one of three SMIs was examined. The results showed that after multiple corrections, bipolar disorder conferred the greatest risk of development of dementia as compared with MDD and schizophrenia. Further analysis revealed that the use of BZD may not alter the risk of dementia in patients with the three SMIs, meaning that the diseases themselves may have a greater impact on the risk of development of dementia than the influence of medications.

Previous studies have shown that in patients with SMIs, whether under medication control or not, cognitive function is more rapidly impaired than in healthy subjects [26,27]. Multiple possible etiologies have been reviewed in previous studies. Anatomically, the brain volume reduction rates in specific regions in patients with schizophrenia, bipolar disorder and MDD were observed to be faster than those in healthy elderly subjects [28,29]. Cognitively, functional decline increases with time, and a lack of protective factors (for example, educational level may be hindered by early disease onset) may increase the risk of neurodegenerative disorder development [30,31]. Genetically, although the influence of genes on the associations of psychiatric illnesses with dementia is limited, protein assembly or even epigenetic variations may increase the risk of dementia in psychiatric patients [32]. The etiologies listed above may not be reversed, but our study demonstrated again a higher risk of dementia development in patients with SMIs as compared with healthy elderly subjects, and the risk in each SMI was compared. Further disease-specific rehabilitation programs for elderly psychiatric patients should focus on functional maintenance and regular cognitive function evaluation.

The results of our study showed that bipolar disorder conferred the greatest risk of dementia development as compared with schizophrenia and MDD. Sensitivity analysis showed that the use of lithium may not alter this result. The exact mechanism that led to the results was unclear, but medical adherence to bipolar disorder treatment is inadequate as compared with treatment for schizophrenia [33,34]. Long-term alteration with a hyperdopaminergic state [35] and dysregulation of other neurotransmitters (for example, glutamate) [35,36] without medication may be correlated with cognition and other functional impairment, and may therefore hasten the development of dementia. Furthermore, patients with bipolar disorder may suffer from two influencing states, a manic (or hypomanic) state or a depressive episode, both of which may increase the risk of dementia [37]. One study even indicated a specific genetic influence on the relationship between bipolar disorder and dementia [38]. Further studies need to be conducted to analyze these associations.

In the current study, whether or not hypnotics use may increase the risk of dementia in all three groups of patients was assessed, and the results were consistent with those of other recent studies indicating that BZD may not increase the risk of dementia [21]. The pathogeneses of the SMIs themselves may have a stronger influence on the development of dementia than the influence of medications. However, our result could not be taken as encouragement for elderly psychiatric patients to use hypnotics. Even though the risk of developing dementia is weakened in patients using anxiolytics, greater risks of medication-related (especially hypnotics) consciousness disturbance, fall-related events, and even a hepatorenal burden are still evident [39]. Hypnotics use in patients with SMIs should be cautious, and a number of varying risks and benefits need to be evaluated, only one of which is the risk of development of dementia.

Another important finding should be noted. The results of this study showed that patients with MDD were at greater risk of cardiovascular illnesses, which was consistent with a previous study [40]; these factors are also risk factors for dementia [41]. However, the significance of the risk of developing dementia in the MDD group as compared with healthy subjects vanished under BZD use. The HR also decreased for the patients with bipolar disorder and schizophrenia in comparison with healthy subjects under BZD use. In healthy subjects without SMI, anxiolytics may still play an important role in cognitive impairment. Careful prescription of anxiolytics for non- SMI patients needs to be reiterated in clinical practice.

To conclude, the results of this study showed that bipolar disorder confers a greater risk of developing dementia than MDD and schizophrenia. Clinical evaluation of cognitive function decline in geriatric psychiatric patients should be performed with care, and appropriate medical management should be initiated as soon as possible once a dementia-related change has been observed. In addition, appropriate hypnotics use may be indicated in the clinical treatment of SMI patients. Proper psychoedu-cation with regards to hypnotics use and frequent monitoring of side effects of hypnotics may be a better way forward than complete avoidance of their use.

Limitations

Several limitations of this study should be mentioned. First, some variables were not obtainable in this study, for example, educational level, lifestyle modification and exercise habits, which may also be regarded as factors related to the development of dementia. Second, diagnosis of bipolar disorder varies according to different medical systems. We employed lithium use in the sensitivity analysis, and the results remained consistent. Third, although hypnotics may not alter the risk of dementia according to the results of this study, possible side effects of hypnotics that may also be related to dementia (for example, medication-related delirium) could not be measured. In addition, the use of other psychotropic medications, for example, antipsychotics, could not be controlled in the bipolar disorder and schizophrenia groups. Fourth, owing to NHID has been launched since 1995, and it is not long enough to evaluate effect of lifetime duration of illness from SMI onset to the date of cohort inclusion (index date) for risk of dementia. Also, we cannot extract the proportions of the anti-dementia medications use in this study because the limitations of the projects. Similar studies, however, revealed that about 16% patients with antidementia treatment in Taiwan [42].

Funding Statement

Funding This research was funded by the Taiwanese Society of Psychiatry (TSOP-2012-R1). The funders had no role in the study design, data collection and analysis, decision to publish, or preparation of the manuscript.