Prevalence, incidence and mortality from cardiovascular disease in patients with pooled and specific severe mental illness: a large‐scale meta‐analysis of 3,211,768 patients and 113,383,368 controls

Abstract

People with severe mental illness (SMI) – schizophrenia, bipolar disorder and major depressive disorder – appear at risk for cardiovascular disease (CVD), but a comprehensive meta‐analysis is lacking. We conducted a large‐scale meta‐analysis assessing the prevalence and incidence of CVD; coronary heart disease; stroke, transient ischemic attack or cerebrovascular disease; congestive heart failure; peripheral vascular disease; and CVD‐related death in SMI patients (N=3,211,768) versus controls (N=113,383,368) (92 studies). The pooled CVD prevalence in SMI patients (mean age 50 years) was 9.9% (95% CI: 7.4‐13.3). Adjusting for a median of seven confounders, patients had significantly higher odds of CVD versus controls in cross‐sectional studies (odds ratio, OR=1.53, 95% CI: 1.27‐1.83; 11 studies), and higher odds of coronary heart disease (OR=1.51, 95% CI: 1.47‐1.55) and cerebrovascular disease (OR=1.42, 95% CI: 1.21‐1.66). People with major depressive disorder were at increased risk for coronary heart disease, while those with schizophrenia were at increased risk for coronary heart disease, cerebrovascular disease and congestive heart failure. Cumulative CVD incidence in SMI patients was 3.6% (95% CI: 2.7‐5.3) during a median follow‐up of 8.4 years (range 1.8‐30.0). Adjusting for a median of six confounders, SMI patients had significantly higher CVD incidence than controls in longitudinal studies (hazard ratio, HR=1.78, 95% CI: 1.60‐1.98; 31 studies). The incidence was also higher for coronary heart disease (HR=1.54, 95% CI: 1.30‐1.82), cerebrovascular disease (HR=1.64, 95% CI: 1.26‐2.14), congestive heart failure (HR=2.10, 95% CI: 1.64‐2.70), and CVD‐related death (HR=1.85, 95% CI: 1.53‐2.24). People with major depressive disorder, bipolar disorder and schizophrenia were all at increased risk of CVD‐related death versus controls. CVD incidence increased with antipsychotic use (p=0.008), higher body mass index (p=0.008) and higher baseline CVD prevalence (p=0.03) in patients vs. controls. Moreover, CVD prevalence (p=0.007), but not CVD incidence (p=0.21), increased in more recently conducted studies. This large‐scale meta‐analysis confirms that SMI patients have significantly increased risk of CVD and CVD‐related mortality, and that elevated body mass index, antipsychotic use, and CVD screening and management require urgent clinical attention.

People with severe mental illness (SMI) – including schizophrenia, bipolar disorder, major depressive disorder, and their related spectrum disorders – have a life expectancy shortened of 10‐17.5 years compared to the general population1, 2. While suicide explains some of this reduced life expectancy3, it is now established that physical diseases account for the overwhelming majority of premature mortality4, 5. Among physical conditions, cardiovascular disease (CVD) is the main potentially avoidable contributor to early deaths in patients with SMI⁴.

Given the importance of understanding the magnitude, contributors to and relative distribution of CVD risk in people with SMI, a number of disease‐specific meta‐analyses investigated if people with major depressive disorder, bipolar disorder or schizophrenia are at an increased risk of CVD compared to controls. These meta‐analyses reported that people with depression (defined by the presence of depressive symptoms or a diagnosis of major depressive disorder) are at increased CVD risk6, 7, including stroke (risk ratio, RR=1.34, 95% CI: 1.17‐1.54), myocardial infarction (hazard ratio, HR=1.31, 95% CI: 1.09‐1.57), coronary heart disease (RR=1.36, 95% CI: 1.24‐1.49) and coronary heart disease‐related death (HR=1.36, 95% CI: 1.14‐1.63)6, 7, 8. While clearly informative, results concerning CVD were not specific for major depressive disorder defined according to established diagnostic criteria, possibly biasing such observed association towards a lower risk9. Another meta‐analysis of longitudinal studies, which utilized standardized criteria to define bipolar disorder, reported mixed results, since people with that disorder were actually not at increased risk of myocardial infarction (RR=1.09, 95% CI: 0.96‐1.24), whereas the risk of stroke was higher compared to controls (RR=1.74, 95% CI: 1.29‐2.35)10. Among individuals with schizophrenia, previous meta‐analyses11, 12 reported an overall increased risk of CVD compared to controls (RR=1.53, 95% CI: 1.27‐1.86). This risk increase included stroke (up to RR=1.71, 95% CI: 1.19‐2.46) and heart failure (RR=1.81, 95% CI: 1.42‐2.29), but not coronary heart disease (RR=1.20, 95% CI: 0.93‐1.53).

While the existing literature has provided relevant insights, several limitations are to be highlighted and important questions remain unanswered. First, some of the previous meta‐analyses did not use standardized clinical assessments to identify and categorize SMI and/or cardiovascular events. Second, the exact prevalence and incidence of each type of CVD among people with SMI, both within and across major diagnostic SMI subgroups, remains unclear. Third, the magnitude of premature CVD‐related mortality risk in people with SMI versus controls is to be specified. Fourth, potential risk factors for increased CVD and related mortality risk across the SMI groups have not been elucidated via meta‐analytic techniques, which could help identify targets for treatment guidelines, clinical standards and development of preventive and therapeutic programs. In this regard, large‐scale pooled analyses in the SMI population can provide relevant information, allowing the investigation of potentially shared risk factors across many studies and participants, thus dissecting CVD risk factors associated with SMI and/or treatments for these disorders from factors which are non‐specific or shared with the general population13. Additionally, pooling of data allows for the investigation of demographic, regional and treatment variables, both within and across major diagnostic categories.

Given the caveats mentioned above, the current gaps within the literature and the need to better understand CVD risk among people with SMI, we conducted a large scale meta‐analysis investigating the prevalence, incidence and mortality attributed to CVD and their correlates among people with SMI, both within and across major diagnostic groups.

METHODS

This systematic review and meta‐analysis adhered to the PRISMA statement14, following a predetermined, but unpublished protocol.

Search strategy

An electronic literature search was conducted in PubMed, Embase and Scopus from database inception until August 2, 2016 by two independent reviewers, using the search terms (“bipolar disorder” OR mania OR schizophrenia OR schizoaffective OR psychosis OR “major depression” OR “serious mental illness”) AND (cardiovascular OR stroke OR cerebrovascular OR “transient ischemic attack” OR “transient ischaemic attack” OR “peripheral vascular” OR “myocardial infarction” OR “coronary heart disease” OR” coronary artery disease” OR “ischemic heart disease” OR “ischaemic heart disease” OR “hypertensive heart disease” OR angina OR “cardiac failure” OR “heart failure” OR “congestive heart failure” OR “atrial fibrillation” OR “pulmonary embolism” OR “cardiovascular mortality”). Furthermore, bibliographies of included papers were reviewed.

Inclusion and exclusion criteria

We included studies with the following characteristics: a) reporting on patients with schizophrenia, schizophrenia spectrum or schizoaffective disorder, bipolar disorder or bipolar spectrum disorders, major depressive disorder or depressive episodes, or SMI (defined as at least two among major depressive spectrum, bipolar spectrum and schizophrenia spectrum disorders) according to DSM‐III, DSM‐IV, DSM‐5, ICD‐8, ICD‐9 or ICD‐10, or a medical record diagnosis based on a clinical interview; b) having a cross‐sectional or a retrospective/prospective longitudinal design, either with or without a control group; c) using a standardized definition of CVD; d) reporting RR, HR or odds ratio (OR) comparing patients with region‐specific controls, percentage or number of events at baseline (data used for cross‐sectional analysis = prevalence) and/or follow‐up (data used for longitudinal analysis = cumulative incidence).

We excluded studies that investigated cardiovascular risk estimates and/or factors, subclinical CVD, or SMI rates in populations with CVD. In case of multiple publications from the same study, only the most recent paper or the article with the longest follow‐up was included. When required, we contacted the primary/corresponding authors of potential studies to confirm eligibility or acquire unpublished variables of interest.

Data extraction

Seven authors divided in four pairs independently extracted data in a standardized Microsoft Excel sheet, with reciprocal validation of data extraction results. The extracted data included: authors, year and country; geographic region; study design; data source; period of data collection; SMI diagnostic criteria; CVD diagnostic criteria; specific SMI and CVD diagnosis; case and control inclusion criteria; number of cases and controls; percentage or number with CVD, coronary heart disease, cerebrovascular disease and congestive heart failure at baseline; number of events at follow‐up; follow‐up duration; number and type of covariates considered in the analyses; OR, RR, rate ratio and HR with their respective 95% upper and lower CIs; mean age with standard deviation; mean body mass index with standard deviation; proportion of males; co‐occurring obesity, alcohol and substance related disorders, diabetes, hypertension, and hyperlipidemia; married status; employment status; percentage of patients with poorest income and least urbanized; and percentage of patients taking antipsychotics. Rate ratios calculated with Cox regression models were included in HR analyses. When authors did not specify whether or not a rate ratio had been calculated with Cox regression models, we contacted them seeking clarification.

Outcomes

Primary outcomes were CVD prevalence and cumulative incidence plus CVD‐related mortality in people with SMI, as well as adjusted OR for prevalence and HR for incidence rates in SMI versus controls. Secondary outcomes were the same measures for specific CVDs (i.e., coronary heart disease, cerebrovascular disease, congestive heart failure) in SMI patients, as well as adjusted OR and HR versus controls.

Prevalence and OR were calculated from cross‐sectional studies and from baseline results of longitudinal studies. Where available, incidence, RR and HR were calculated from longitudinal studies.

Quality assessment

For the purpose of this meta‐analysis, a checklist (yes versus no) was used to assess the methodological quality of included studies. The evaluation of methodological quality across studies was based on the following factors: clear diagnostic criteria, presence of a control group, matching of the control group, covariate‐adjusted outcomes, reported cardiovascular risk factors at baseline, and follow‐up ≥5 years.

Data analysis

This meta‐analysis was performed using Comprehensive Meta‐Analysis V315. All outcomes were meta‐analyzed when at least two studies provided data. A random effects model16, 17 was used to account for between‐study heterogeneity. We calculated pooled CVD prevalences and pooled CVD cumulative incidences, each with SMI subgrouping. For dichotomous primary and secondary outcomes comparing pooled SMI and SMI subgroups with controls, we calculated unadjusted as well as adjusted pooled OR for cross‐sectional data, and unadjusted pooled RR, as well as adjusted pooled HR, for longitudinal data. Funnel plots were visually inspected, and Egger's test18 and Begg‐Mazumdar Kendall's tau19 were used to determine if publication bias was likely. When publication bias was present, the trim and fill20 procedure was run to evaluate if the results changed after imputing potentially missing studies.

Between‐study heterogeneity was measured using the chi‐squared and I‐squared statistics, with chi‐squared p<0.05 and I‐squared ≥50% indicating significant heterogeneity21. To identify potential moderators, meta‐regression was run with Comprehensive Meta‐Analysis V3 for unadjusted outcomes where heterogeneity was significant.

Since CVD rates in the general population vary across the world, we also performed a stratified analysis across geographic regions (Asia, Europe, North America, Oceania) regarding raw CVD prevalence and incidence in SMI populations, and compared patients to their respective region‐specific general population controls (calculating RRs as well as adjusted ORs and HRs for the four regional strata and comparing them across the different regions whenever at least two studies provided data per each region).

The following study and patient characteristics were explored as potential moderators and mediators in addition to SMI status: geographical region of the sample; time of data collection; percentage of patients taking antipsychotics; and the difference between patient and control samples regarding age, body mass index, proportion of males and of those with married status, unemployed, with poorest income, least urbanized, and having co‐occurring obesity, alcohol and substance‐related disorders, diabetes, hypertension or hyperlipidemia.

RESULTS

Search results

Out of 18,064 initial hits across the searched electronic databases, 11,878 unduplicated hits were screened, and 11,576 were excluded through title/abstract reading. Altogether, 302 full texts were reviewed, and 210 were excluded with specific reasons. Among 92 studies meeting inclusion criteria, 27 had a cross‐sectional design22, 23, 24, 25, 26, 27, 28, 29, 30, 31, 32, 33, 34, 35, 36, 37, 38, 39, 40, 41, 42, 43, 44, 45, 46, 47, 48 and 65 studies had a retrospective or prospective longitudinal design49, 50, 51, 52, 53, 54, 55, 56, 57, 58, 59, 60, 61, 62, 63, 64, 65, 66, 67, 68, 69, 70, 71, 72, 73, 74, 75, 76, 77, 78, 79, 80, 81, 82, 83, 84, 85, 86, 87, 88, 89, 90, 91, 92, 93, 94, 95, 96, 97, 98, 99, 100, 101, 102, 103, 104, 105, 106, 107, 108, 109, 110, 111, 112, 113 (Figure 1).

Figure 1.

PRISMA flow chart

Characteristics of included studies

We included 92 studies, with a total population of 3,211,768 patients (mean age 50 years, 49% male) with SMI and 113,383,368 controls (mean age 51 years, 49% male), with a total of 116,595,136 subjects when summing those studies where patient and control sample sizes were not separately reported. Altogether, 27 studies (N=27,037,943) were cross‐sectional and 65 studies (N=89,557,193) were longitudinal. Overall, 38 studies included patients with schizophrenia (of which 29 were longitudinal), 30 with bipolar disorder (21 longitudinal), 30 with major depressive disorder (22 longitudinal), and 14 with SMI (8 longitudinal). Taken together, six studies included only patients with SMI (N=884,412), 16 studies included only patients with bipolar disorder (N=71,832), 20 studies included only patients with major depressive disorder (N=111,360), and 29 studies included only patients with schizophrenia (N=1,591,106), while 19 studies included different subgroups of SMI, providing data for each of them separately (some studies included more than one diagnostic group, see Tables 1 and 2 for details).

Table 1.

Cross‐sectional studies: characteristics of included studies and samples

Study	Country	No. cases	No. controls	Period of data collection	SMI definition	Inclusion criteria for cases	No. covariates
Beyer et al22	USA	1,379	‐	2001‐2002	Medical records	Bipolar disorder	‐
Bresee et al23	Canada	28,775	2,281,636	1995‐2006	ICD‐9,10	Schizophrenia	4
Bresee et al24	Canada	399	120,044	2005	Medical records	Schizophrenia	11
Chen et al25	Taiwan	80	‐	2015	DSM‐IV	Bipolar disorder, >60 years	‐
Curkendall et al26	Canada	3,022	12,088	1994‐1999	ICD‐9	Schizophrenia	7
Devantier et al27	Denmark	28	27	2009‐2011	ICD‐10	Major depressive disorder, late onset	‐
Hagg et al28	Sweden	269	‐	2000‐2003	DSM‐IV	Schizophrenia, 20‐69 years	‐
Herbst et al29	USA	10,573 total population		2001‐2002	DSM‐IV	Major depressive disorder, >60 years	11
Huang et al30	Taiwan	117,987	21,356,304	2000‐2003	ICD‐9	Bipolar disorder or major depressive disorder	1
Hyde et al31	Australia	355	‐	2008‐2012	Medical records	Severe mental illness, prescribed clozapine	‐
Kilbourne et al32	USA	8,083	‐	2001	ICD‐9	Severe mental illness, >60 years	‐
Kilbourne et al33	USA	9,705	5,353	2000‐2001	ICD‐9	Bipolar disorder or severe mental illness, male	3
Lindegard34	Sweden	368	87,176	1966‐1979	ICD‐9, DSM‐III	Major depressive disorder or bipolar disorder	‐
Maina et al35	Italy	185	‐	2006‐2008	DSM‐IV	Severe mental illness	‐
Morden et al36	Canada	65,362	65,362	2000‐2007	ICD‐9	Schizophrenia	4
Munoli et al37	India	120	‐	2011	ICD‐10	Bipolar disorder	‐
Nielsen et al38	Denmark	937	‐	1969‐2014	ICD‐10	Schizophrenia	‐
Niranjan et al39	USA	5,695	34,979	2007	DSM‐IV	Major depressive disorder	6
Oreski et al40	Croatia	289	192	2011	ICD‐10	Bipolar disorder or schizophrenia	‐
Prieto et al41	USA	988	‐	2009‐2013	DSM‐IV	Severe mental illness	‐
Scherrer et al42	USA	628	6,903	1990‐1992	DSM‐III	Major depressive disorder, male twins	‐
Scott et al43	Multicenter	52,095 total population		2001‐2011	DSM‐IV	Bipolar disorder or major depressive disorder	6
Shen et al44	Taiwan	203	2,036	2005‐2007	ICD‐9	Schizophrenia, in intensive care unit	6
Smith et al45	UK	9,677	1,414,701	2007	Medical records	Schizophrenia	3
Smith et al46	UK	2,582	1,421,796	2007	Medical records	Bipolar disorder	2
Swain et al47	Multicenter	45,288 total population		2001‐2011	DSM‐IV	Bipolar disorder or major depressive disorder	7
Zilkens et al48	Australia	656	349	2000‐2009	ICD‐8,9,10	Major depressive disorder, 65‐84 years, developing dementia	‐

SMI – severe mental illness

Table 2.

Longitudinal studies: characteristics of included studies and samples

Study	Country	No. cases	No. controls	Period of data collection	SMI definition	Inclusion criteria for cases	No. covariates
Almeida et al49	Australia	1,503	35,691	1996‐2010	ICD‐9	Schizophrenia, bipolar disorder or major depressive disorder, 65‐85 years, male	8
Bremmer et al50	The Netherlands	41	2,080	1992‐2000	DSM‐III	Major depressive disorder, >55 years	13
Butnoriene et al51	Lithuania	184	369	2003‐2004	DSM‐IV	Major depressive disorder, >45 years	4
Callaghan et al52	Canada	5,999	5,999	2002‐2006	Medical records	Bipolar disorder	6
Callaghan et al53	Canada	9,815	9,815	2002‐2006	ICD‐10	Bipolar disorder	8
Carney et al54	USA	1,074	726,262	1996‐2001	ICD‐9	Schizophrenia or schizoaffective disorder	4
Chen et al55	Taiwan	63,913	63,913	2002‐2008	ICD‐9	Schizophrenia	8
Clouse et al56	USA	16	60	1982‐1992	DSM‐III	Major depressive disorder with diabetes	7
Coryell et al57	USA	903	‐	1998‐1999	RDC	Severe mental illness	‐
Crump et al58	Sweden	6,618	6,580,418	2003‐2009	ICD‐10	Bipolar disorder	6
Crump et al59	Sweden	8,277	6,097,834	2003‐2009	ICD‐10	Schizophrenia, >25 years	6
Davis et al60	Hawaii	280	39,000	1999‐2005	Medical records	Major depressive disorder	5
Davydow et al61	Denmark	68,137	5,912,158	1999‐2013	ICD‐9	Schizophrenia, schizoaffective disorder or bipolar disorder	5
Enger et al62	USA	1,920	9,600	1995‐1999	ICD‐9	Schizophrenia, on antipsychotic treatment, 15‐64 years	‐
Fiedorowicz et al63	USA	288	147	1978‐1981	RDC	Bipolar disorder	8
Filik et al64	UK	482	1,998	1999‐2002	DSM‐IV	Schizophrenia, schizophreniform or schizoaffective disorder	6
Fors et al65	Sweden	255	1,275	1981‐1991	DSM‐II	Schizophrenia	3
Gasse et al66	Denmark	873,898	52,693,301	1995‐2009	ICD‐8,10	Severe mental illness (affective psychoses)	22
Goldstein et al67	USA	5,835	26,266	2001‐2005	DSM‐IV	Bipolar disorder or major depressive disorder	8
Healy et al68	UK	1,429	‐	1875‐1924; 1994‐2010	Medical records	Schizophrenia	‐
Hendrie et al69	USA	757	30,831	1999‐2008	ICD‐9	Schizophrenia, >65 years	‐
Hou et al70	Taiwan	8,264	‐	1985‐2008	DSM‐III or IV, ICD‐9	Schizophrenia	‐
Hsieh et al71	Taiwan	9,715	‐	2001‐2009	ICD‐9	Schizophrenia	10
Huang et al72	Taiwan	7,937	31,748	1996‐2006	ICD‐9	Major depressive disorder	9
Ifteni et al73	Romania	7,189	‐	1989‐2011	DSM‐IV	Schizophrenia, inpatients	‐
Jakobsen et al74	Denmark	74,759	338,747	1977‐2000	ICD‐8,10	Schizophrenia or major depressive disorder	2
Janszky et al75	Sweden	646	48,675	1969‐2007	ICD‐8	Major depressive disorder, 18‐20 years	7
Jokinen & Nordstrom76	Sweden	346	‐	1980‐2005	DSM‐IV	Major depressive disorder or bipolar disorder	‐
Joukamaa et al77	Finland	606	8,000	1977‐1994	Medical records	Schizophrenia, mood disorder or severe mental illness	1
Kendler et al78	Sweden	5,647	24,727	1998‐2003	ICD‐10	Major depressive disorder, twins	‐
Kiviniemi et al79	Finland	6,987	‐	1998‐2003	ICD‐9	Schizophrenia, first onset	‐
Lahti et al80	Finland	204	11,880	1969‐2004	ICD‐8,9,10	Schizophrenia	5
Lan et al81	Taiwan	3,681	‐	2001‐2006	ICD‐9	Bipolar disorder	‐
Laursen et al82	Denmark	22,294	2,411,852	1995‐2007	ICD‐8,10	Schizophrenia or bipolar disorder, 15‐52 years	3
Laursen et al83	Denmark	1,454	59,256	1995‐2006	ICD‐8,10	Schizophrenia or bipolar disorder	4
Lemogne et al84	France	4,336	16,621	1990‐2010	ICD‐9,10	Depression or severe mental illness (bipolar disorder, psychosis)	6
Li et al85	Taiwan	1,003	4,012	1996‐2009	ICD‐9	Major depressive disorder	6
Lin et al86	Taiwan	7,353	22,059	2000‐2006	ICD‐9	Schizophrenia	8
Lin et al87	Taiwan	2,289	16,413	1998‐2003	ICD‐9	Bipolar disorder	10
Lin et al88	Taiwan	5,001	10,002	1998‐2003	ICD‐9	Schizophrenia, <45 years	9
Maina et al89	Italy	309	‐	2003‐2011	DSM‐IV	Bipolar disorder	‐
McDermott et al90	USA	503	2,083	1990‐2003	ICD‐9	Schizophrenia or severe mental illness	9
Murray‐Thomas et al91	UK	232,132	193,920	1997‐2001	ICD‐10	Schizophrenia, bipolar disorder or major depressive disorder	2
Olfson et al92	USA	1,138,853	‐	2001‐2007	ICD‐10	Schizophrenia, 20‐64 years	4
Osborn et al93	UK	38,824	‐	1995‐2010	Medical records	Bipolar disorder or severe mental illness, 30‐90 years	‐
Pratt et al94	USA	73	1,107	1981‐1994	DSM‐III	Major depressive disorder	11
Prieto et al95	USA	334	334	1966‐1996	DSM‐IV	Bipolar disorder	4
Rahman et al96	Sweden	6,822	29,832	1998‐2002	ICD‐7,8,9,10	Major depressive disorder, twin population study	7
Ramsey et al97	USA	129	1,339	1981‐1982	DSM‐III	Bipolar disorder or major depressive disorder	6
Saint Onge et al98	USA	548	10,821	1999‐2006	ICD	Major depressive disorder	11
Scherrer et al99	USA	77,568	214,749	1999‐2007	ICD‐9	Major depressive disorder, 25‐80 years	4
Schoepf & Heun100	UK	1,418	14,180	2000‐2012	ICD‐10	Schizophrenia, inpatients	‐
Schoepf et al101	UK	621	6,210	2000‐2012	ICD‐10	Bipolar disorder	‐
Shah et al102	USA	538	7,103	1988‐2006	DSM‐III	Major depressive disorder or bipolar disorder, 17‐39 years	14
Stewart et al103	USA	235	‐	NA	ICD‐9	Major depressive disorder	‐
Surtees et al104	UK	3,057	16,592	1996‐2008	DSM‐IV	Major depressive disorder, 45‐80 years	11
Ting et al105	China	153	7,682	1996‐2008	DSM‐IV	Major depressive disorder with diabetes	18
Torniainen et al106	Sweden	21,492	214,920	2006‐2015	ICD‐10	Schizophrenia, 17‐65 years	2
Tsai et al107	Taiwan	80,569	241,707	1999‐2003	ICD‐9	Schizophrenia	8
Tsan et al108	USA	49,173	‐	2002‐2009	ICD‐9	Schizophrenia	‐
van Marwijk et al109	The Netherlands	143	139	2002‐2003	DSM‐IV	Major depressive disorder, >55 years	‐
Weeke et al110	Denmark	3,795	‐	1950‐1957; 1969‐1977	ICD‐8	Bipolar disorder	‐
Westman et al111	Sweden	17,101	10,631,208	1987‐2006	ICD‐10	Bipolar disorder	3
Wu et al112	Taiwan	16,821	67,284	1999‐2010	ICD‐9	Bipolar disorder	9
Wu et al113	Taiwan	70,225	207,592	1996‐2007	ICD‐9	Schizophrenia or bipolar disorder	8

SMI – severe mental illness, RDC – Research Diagnostic Criteria

Meta‐analysis: cross‐sectional results

The pooled CVD prevalence in SMI was 9.9% (95% CI: 7.4‐13.3; 38 studies). Individual rates were 8.4% for people with bipolar disorder (95% CI: 5.4‐12.6, 12 studies, N=66,911); 11.7% for those with major depressive disorder (95% CI: 3.6‐32.2, 7 studies, N=83,965); 11.8% for those with schizophrenia (95% CI: 7.1‐19.0, 13 studies, N=191,982), and 11.8% for those with SMI (95% CI: 4.1‐29.4, 6 studies, N=17,286) (p<0.001 for SMI diagnostic subgroup comparisons).

Adjusting for a median of seven potential confounders, the adjusted pooled OR for CVD in SMI compared to controls was 1.53 (95% CI: 1.27‐1.83, p<0.001, 11 studies). For specific CVDs, pooled together, people with SMI had an increased risk of coronary heart disease (OR=1.51, 95% CI: 1.47‐1.55, p<0.001, 5 studies) and cerebrovascular disease (OR=1.42, 95% CI: 1.21‐1.66, p<0.001, 6 studies), with a strong statistical trend for congestive heart failure (OR=1.28, 95% CI: 0.99‐1.65, p=0.06, 4 studies). Considering separately single types of SMI and CVD, in adjusted OR analyses, bipolar disorder was not significantly associated with CVD or its subtypes; major depressive disorder was significantly associated with CVD and coronary heart disease; and schizophrenia was significantly associated with coronary heart disease, cerebrovascular disease and congestive heart failure (Table 3). No adjusted ORs were available for mixed SMI groups.

Table 3.

Meta‐analysis of cross‐sectional studies: unadjusted and adjusted odds ratios

	Meta‐analysis of unadjusted odds ratios								Meta‐analysis of covariate adjusted odds ratios
	No. studies	No. participants		Unadjusted odds ratios				Hetero‐geneity	No. studies	No. participants		Adjusted odds ratios				Hetero‐geneity
Disorder		Patients	Controls	OR	95% CI		p	I2		Patients	Controls	OR	95% CI		p	I2
Cardiovascular disease
Bipolar disorder	4	19,562	1,526,110	1.73	1.11	2.71	0.02	91	4	2,640	1,423,135	1.28	0.90	1.80	0.17	52
Major depressive disorder	3	1,577	47,851	2.08	1.51	2.88	<0.001	58	7	7,050	43,570	1.75	1.36	2.26	<0.001	69
Schizophrenia	10	190,584	4,100,315	1.23	0.92	1.65	0.16	99	5	42,076	3,860,505	1.38	0.93	2.05	0.11	96
Severe mental illnesses	1	146	2,083	1.59	0.87	2.88	0.13	‐	‐	‐	‐	‐	‐	‐	‐	‐
Pooled	14	211,869	7,808,603	1.59a	1.32	1.91	<0.001	99	11	51,766	5,325,871	1.53e	1.27	1.83	<0.001	94
Coronary heart disease
Bipolar disorder	3	19,504	1,524,771	1.75	1.11	2.77	0.02	94	1	2,582	1,421,796	0.94	0.79	1.11	0.49	‐
Major depressive disorder	1	958	35,691	2.44	2.13	2.79	<0.0001	‐	3	6,323	41,882	2.52	1.81	3.52	<0.001	93
Schizophrenia	8	187,359	4,086,191	1.03	0.85	1.25	0.76	98	1	399	120,044	1.52	1.48	1.56	<0.001	‐
Severe mental illnesses	1	146	2,083	1.02	0.56	1.83	0.96	‐	‐		‐	‐	‐	‐	‐	‐
Pooled	8	207,967	4,160,030	1.80b	1.62	2.00	<0.001	98	5	9,304	1,583,722	1.51f	1.47	1.55	<0.001	90
Cerebrovascular disease
Bipolar disorder	3	2,741	1,458,826	1.68	1.07	2.63	0.03	47	2	2,582	1,421,796	1.06	0.85	1.31	0.62	0
Major depressive disorder	3	1,577	47,851	2.24	1.33	3.79	0.003	81	2	656	349	1.64	0.96	2.78	0.07	72
Schizophrenia	5	41,071	37,77,039	1.63	1.19	2.24	0.003	96	3	32,196	2,413,768	2.05	1.59	2.64	<0.001	61
Severe mental illnesses	1	146	2,083	1.02	0.56	1.83	0.96	‐	‐	‐	‐	‐	–	‐	‐	‐
Pooled	10	45,535	5,454,785	1.63c	1.31	2.02	<0.0001	93	6	35,434	3,835,913	1.42g	1.21	1.66	<0.001	90
Congestive heart failure
Bipolar disorder	1	2,582	1,421,796	1.38	1.03	1.84	0.03	‐	1	2,582	1,421,796	1.11	0.80	1.54	0.53	0
Major depressive disorder	‐	‐	‐	‐	‐	‐	‐	‐	‐	‐	‐	‐	‐	‐	‐	‐
Schizophrenia	5	40,984	3,743,431	1.71	1.36	2.15	<0.001	92	3	41,474	5,708,425	1.60	1.06	2.40	0.02	97
Severe mental illnesses	1	146	2,083	1.59	0.87	2.88	0.13	‐	‐	‐	‐	‐	‐	‐	‐	‐
Pooled	6	43,712	5,167,189	1.57d	1.32	1.87	<0.001	88	4	44,056	7,130,221	1.28h	0.99	1.65	0.06	96

Bold values represent significant results

Egger test for bias: ^a2.24, p=0.53; ^b6.41, p=0.03 (Duval & Tweedie trim‐and‐fill procedure adjusted OR: 1.35, 95% CI: 0.98‐1.83); ^c−1.69, p=0.24; ^d−2.39, p=0.15; ^e−1.73, p=0.18; ^f0.17, p=0.93; ^g−2.59, p=0.07; ^h−5.26, p=0.20

All significant results were significantly heterogeneous. After adjusting for publication bias with the trim‐and‐fill method, all pooled previously significant ORs remained statistically significant, confirming the association of CVD, coronary heart disease and cerebrovascular disease with SMI, while the OR for congestive heart failure became marginally significant (p=0.05).

Meta‐analysis: longitudinal adjusted results

Among patients with SMI, 3.6% (95% CI: 2.7‐5.3%) experienced a CVD event during a median follow‐up period of 8.4 years (range 1.8‐30.0) (65 studies). After adjusting for a median of six potential confounders, people with SMI were at significantly increased risk across longitudinal studies for CVD (HR=1.78, 95% CI: 1.60‐1.98) (31 studies, N=671,384 cases vs. N=14,335,203 controls) as well as for specific CVDs, including coronary heart disease (HR=1.54, 95% CI: 1.30‐1.82, 18 studies, N=194,017 cases vs. N=13,530,858 controls), cerebrovascular disease (HR=1.64, 95% CI: 1.26‐2.14, 11 studies, N=188,841 cases vs. N=13,113,564 controls), congestive heart failure (HR=2.10, 95% CI: 1.64‐2.70, 2 studies, N=409 cases vs. N=41,678 controls), peripheral vascular disease (only unadjusted RR=3.11, 95% CI: 2.46‐3.91, three studies), and CVD‐related death (HR=1.85, 95% CI: 1.53‐2.24, 16 studies, N=353,407 cases vs. N=7,317,053 controls).

According to adjusted HRs, schizophrenia was significantly associated with CVD in longitudinal studies (HR=1.95, 95% CI: 1.41‐2.70, 14 studies), as well as with coronary heart disease (HR=1.59, 95% CI: 1.08‐2.35, 5 studies), cerebrovascular disease (HR=1.57, 95% CI: 1.09‐2.25, 5 studies), and CVD‐related death (HR=2.45, 95% CI: 1.64‐3.65, 9 studies).

According to adjusted HRs, bipolar disorder was significantly associated with CVD in longitudinal studies (HR=1.57, 95% CI: 1.28‐1.93, 10 studies) as well as with CVD‐related death (HR=1.65, 95% CI: 1.10‐2.47, 3 studies), with a trend toward a significant association with cerebrovascular disease (HR=1.60, 95% CI: 0.99‐2.57, 4 studies), but no significant association with coronary heart disease (HR=1.16, 95% CI: 0.76‐1.78, 4 studies). One study reported a significant association with congestive heart failure (HR = 2.27, 95% CI: 1.49‐3.45).

According to adjusted HRs, major depressive disorder was significantly associated with CVD in longitudinal studies (HR=1.72, 95% CI: 1.48‐2.00, 18 studies) as well as with coronary heart disease (HR=1.63, 95% CI: 1.33‐2.00, 9 studies), cerebrovascular disease (HR=2.04, 95% CI: 1.05‐3.96, 3 studies), congestive heart failure (HR=2.02, 95% CI: 1.48‐2.75, 2 studies), and CVD‐related death (HR=1.63, 95% CI: 1.25‐2.13, 7 studies).

According to adjusted HRs, mixed SMIs were significantly associated with CVD in longitudinal studies (HR=3.24, 95% CI: 2.15‐4.88, 3 studies) as well as with CVD‐related death (HR=2.75, 95% CI: 1.32‐5.73, 3 studies).

All significant results were significantly heterogeneous, except for mixed SMI and CVD risk, as well as all the congestive heart failure results. After trim and fill procedure, all results remained unchanged, and Egger test did not show any evidence of publication bias influencing the results (see Table 4 for details).

Table 4.

Meta‐analysis of longitudinal studies with publication bias assessment

	Meta‐analysis of unadjusted relative risk								Meta‐analysis of covariate adjusted hazard ratio
	No. studies	No. participants		Unadjusted relative risk				Hetero‐geneity	No. studies	No. participants		Adjusted hazard ratio				Hetero‐geneity
Disorder		Patients	Controls	RR	95% CI		p	I2		Patients	Controls	HR	95% CI		p	I2
Cardiovascular disease
Bipolar disorder	12	66,549	9,606,575	1.50	1.28	1.75	<0.0001	76	10	91,187	6,967,728	1.57	1.28	1.93	<0.0001	91
Major depressive disorder	13	328,431	800,718	1.29	0.92	1.81	0.14	99	18	282,621	682,045	1.72	1.48	2.00	<0.0001	67
Schizophrenia	16	361294	16,096,125	1.21	1.006	1.45	0.04	98	14	296,778	7,176,374	1.95	1.41	2.70	<0.0001	99
Severe mental illnesses	2	874022	52,709,922	2.44	1.13	5.25	0.02	74	3	798	31,724	3.24	2.15	4.88	<0.0001	0
Pooled	33	1,630,296	76,031,192	1.38a	1.23	1.54	<0.0001	98	31	671,384	14,335,203	1.78g	1.60	1.98	<0.0001	95
Coronary heart disease
Bipolar disorder	4	25,286	9,200,196	1.95	1.20	3.17	0.007	96	4	19,129	6,789,683	1.16	0.76	1.78	0.49	87
Major depressive disorder	6	14,3671	515,187	1.15	0.71	1.85	0.57	98	9	99,028	392,210	1.63	1.33	2.00	<0.0001	80
Schizophrenia	8	169,507	15,446,625	0.93	0.81	1.08	0.33	87	5	75,860	6,348,965	1.59	1.08	2.35	0.02	95
Severe mental illnesses	1	873,898	52,693,301	1.80	1.74	1.86	<0.0001	‐	‐	‐	‐	‐	‐	‐	‐	‐
Pooled	17	1,212,362	75,235,865	1.75b	1.69	1.80	<0.0001	99	18	194,017	13,530,858	1.54h	1.30	1.82	<0.0001	92
Cerebrovascular disease
Bipolar disorder	6	32,898	9,082,511	1.92	1.13	3.26	0.02	97	4	23,831	6,649,375	1.60	0.99	2.57	0.05	85
Major depressive disorder	4	8,121	41,665	1.55	1.02	2.35	0.04	77	3	7,046	38,853	2.04	1.05	3.96	0.04	74
Schizophrenia	8	243,254	15,475,608	1.48	1.21	1.81	<0.0001	96	5	157,964	6,425,336	1.57	1.09	2.25	0.02	95
Severe mental illnesses	‐	‐	‐	‐	‐	‐	‐	‐	‐	‐	‐	‐	‐	‐	‐	‐
Pooled	17	284,273	22,187,932	1.53c	1.29	1.82	<0.0001	96	11	188,841	13,113,564	1.64i	1.26	2.14	<0.0001	90
Congestive heart failure
Bipolar disorder	1	6,215	2,411,852	11.52	9.37	23.14	<0.0001	‐	1	58	1,339	2.27	1.49	3.45	<0.0001	0
Major depressive disorder	‐	‐	‐	‐	‐	‐	‐	‐	2	351	40,339	2.02	1.48	2.75	<0.0001	0
Schizophrenia	3	85,290	9,050,272	1.80	1.15	2.79	0.009	84	‐	‐	‐	‐	‐	‐	‐	‐
Severe mental illnesses	‐	‐	‐	‐	‐	‐	‐	‐	‐	‐	‐	‐	‐	‐	‐	‐
Pooled	4	91,505	11,459,059	8.24d	6.84	9.94	<0.0001	99	2	409	41,678	2.10	1.64	2.70	<0.0001	0
Peripheral vascular disease
Bipolar disorder	1	6,215	2,411,852	3.44	2.70	4.38	<0.0001	‐	‐	‐	‐	‐	‐	‐	‐	‐
Major depressive disorder	‐	‐	‐	‐	‐	‐	‐	‐	‐	‐	‐	‐	‐	‐	‐	‐
Schizophrenia	3	85,290	9,050,272	0.96	0.43	2.17	0.92	93	‐	‐	‐	‐	‐	‐	‐	‐
Severe mental illnesses	‐	‐	‐	‐	‐	‐	‐	‐	‐	‐	‐	‐	‐	‐	‐	‐
Pooled	3	91,505	11,402,868	3.11e	2.46	3.91	<0.0001	98	‐	‐	‐	‐	‐	‐	‐	‐
Death due to cardiovascular disease
Bipolar disorder	5	37,144	356,298	1.31	0.94	1.83	0.11	75	3	17,420	162,231	1.65	1.10	2.47	0.02	88
Major depressive disorder	5	18,112	283,746	1.30	0.59	2.86	0.51	99	7	183,297	282,014	1.63	1.25	2.13	<0.0001	81
Schizophrenia	9	53,779	7,179,454	1.26	0.84	1.90	0.27	96	9	152,690	6,872,808	2.45	1.64	3.65	<0.0001	96
Severe mental illnesses	3	874,146	52,714,134	2.99	2.84	3.13	<0.0001	0	3	798	31,724	2.75	1.32	5.73	0.007	75
Pooled	18	1,151,181	60,287,400	2.89f	2.75	3.03	<0.0001	99	16	353,407	7,317,053	1.85j	1.53	2.24	<0.0001	95

Egger test for bias: ^a−0.44, p=0.80; ^b−1.24, p=0.71; ^c0.03, p=0.96; ^d8.07, p=0.37; ^e3.08, p=0.60; ^f−3.66, p=0.20; ^g1.16, p=0.31; ^h−0.13, p=0.92; ⁱ2.57, p=0.07; ^j−1.19, p=0.43

Quality assessment of included studies

Quality ratings of single studies are presented in Table 5. All studies used clear diagnostic criteria, by design. Among the 27 cross‐sectional studies, all except 9 studies had a control group, 5 studies used a matched control sample, 13 studies adjusted analyses for relevant covariates, and all except 6 studies reported cardiovascular risk factors. Among the 65 longitudinal studies, all had a control group, which was matched in all but 12 studies, only 19 studies adjusted for covariates, 38 studies reported on cardiovascular risk factors, and all except 12 studies had a follow‐up of at least 5 years.

Table 5.

Quality assessment of included studies

Study	Clear diagnostic criteria	Control group	Matched controls	Coavariate adjusted analyses	Reported cardiovascular risk factors at baseline	Follow‐up at least 5 years
Cross‐sectional studies
Beyer et al22	Y	N	N	N	Y	N
Bresee et al23	Y	Y	N	Y	Y	N
Bresee et al24	Y	Y	N	Y	Y	N
Chen et al25	Y	N	N	N	Y	N
Curkendall et al26	Y	Y	Y	Y	Y	N
Devantier et al27	Y	Y	Y	N	Y	N
Hagg et al28	Y	N	N	N	Y	N
Herbst et al29	Y	Y	N	Y	N	N
Huang et al30	Y	Y	N	Y	Y	N
Hyde et al31	Y	N	N	N	Y	N
Kilbourne et al32	Y	N	N	N	Y	N
Kilbourne et al33	Y	Y	N	Y	Y	N
Lindegard34	Y	Y	N	N	N	N
Maina et al35	Y	N	N	N	Y	N
Morden et al36	Y	Y	Y	Y	Y	N
Munoli et al37	Y	N	N	N	Y	N
Nielsen et al38	Y	N	N	N	Y	N
Niranjan et al39	Y	Y	N	Y	Y	N
Oreski et al40	Y	Y	N	N	Y	N
Prieto et al41	Y	N	N	N	Y	N
Scherrer et al42	Y	Y	N	N	N	N
Scott et al43	Y	Y	N	Y	N	N
Shen et al44	Y	Y	Y	Y	Y	N
Smith et al45	Y	Y	N	Y	Y	N
Smith et al46	Y	Y	N	Y	Y	N
Swain et al47	Y	Y	N	Y	N	N
Zilkens et al48	Y	Y	Y	N	N	N
Longitudinal studies
Almeida et al49	Y	Y	Y	N	Y	Y
Bremmer et al50	Y	Y	Y	N	Y	Y
Butnoriene et al51	Y	Y	Y	N	N	Y
Callaghan et al52	Y	Y	Y	Y	Y	N
Callaghan et al53	Y	Y	Y	Y	Y	N
Carney et al54	Y	Y	Y	N	Y	N
Chen et al55	Y	Y	Y	Y	Y	Y
Clouse et al56	Y	Y	Y	N	Y	Y
Coryell et al57	Y	Y	N	N	N	Y
Crump et al58	Y	Y	Y	N	N	Y
Crump et al59	Y	Y	Y	N	Y	Y
Davis et al60	Y	Y	Y	N	Y	N
Davydow et al61	Y	Y	Y	N	N	Y
Enger et al62	Y	Y	Y	Y	Y	N
Fiedorowicz et al63	Y	Y	Y	N	Y	Y
Filik et al64	Y	Y	Y	N	Y	N
Fors et al65	Y	Y	Y	Y	N	Y
Gasse et al66	Y	Y	Y	N	N	Y
Goldstein et al67	Y	Y	Y	N	Y	N
Healy et al68	Y	Y	N	N	N	Y
Hendrie et al69	Y	Y	Y	N	Y	Y
Hou et al70	Y	Y	N	N	N	Y
Hsieh et al71	Y	Y	N	N	N	N
Huang et al72	Y	Y	Y	Y	Y	Y
Ifteni et al73	Y	Y	N	N	N	Y
Jakobsen et al74	Y	Y	Y	Y	N	Y
Janszky et al75	Y	Y	Y	N	Y	Y
Jokinen & Nordstrom et al76	Y	Y	N	N	N	Y
Joukamaa et al77	Y	Y	Y	N	N	Y
Kendler et al78	Y	Y	Y	N	N	Y
Kiviniemi et al79	Y	Y	N	N	N	Y
Lahti et al80	Y	Y	Y	N	Y	Y
Lan et al81	Y	Y	N	N	Y	Y
Laursen et al82	Y	Y	Y	N	N	Y
Laursen et al83	Y	Y	Y	N	N	Y
Lemogne et al84	Y	Y	Y	N	Y	Y
Li et al85	Y	Y	Y	Y	Y	Y
Lin et al86	Y	Y	Y	Y	Y	Y
Lin et al87	Y	Y	Y	Y	Y	Y
Lin et al88	Y	Y	Y	Y	Y	Y
Maina et al89	Y	Y	N	N	N	Y
McDermott et al90	Y	Y	Y	N	N	Y
Murray‐Thomas et al91	Y	Y	Y	N	N	N
Olfson et al92	Y	Y	N	N	N	Y
Osborn et al93	Y	Y	Y	N	Y	Y
Pratt et al94	Y	Y	Y	N	Y	Y
Prieto et al95	Y	Y	Y	Y	Y	Y
Rahman et al96	Y	Y	Y	Y	Y	N
Ramsey et al97	Y	Y	Y	N	Y	Y
Saint Onge et al98	Y	Y	Y	N	Y	Y
Scherrer et al99	Y	Y	Y	N	Y	Y
Schoepf & Heun100	Y	Y	Y	Y	N	Y
Schoepf et al101	Y	Y	Y	Y	Y	Y
Shah et al102	Y	Y	Y	N	Y	Y
Stewart et al103	Y	Y	N	N	N	Y
Surtees et al104	Y	Y	Y	N	N	Y
Ting et al105	Y	Y	Y	N	Y	Y
Torniainen et al106	Y	Y	Y	Y	N	Y
Tsai et al107	Y	Y	Y	Y	Y	Y
Tsan et al108	Y	Y	N	N	Y	Y
van Marwijk et al109	Y	Y	Y	Y	Y	N
Weeke et al110	Y	Y	N	N	N	N
Westman et al111	Y	Y	Y	N	N	Y
Wu et al112	Y	Y	Y	Y	Y	Y
Wu et al113	Y	Y	Y	N	Y	Y

N – no, Y – yes

Regional CVD prevalence, incidence and longitudinal risk

Raw CVD prevalence and incidence rates consistently increased from Asia, through Europe and North America, to Oceania (Asia: 5.4% and 2.6%; Europe: 9.7% and 3.4%; North America: 14.6% and 4.6%; Oceania: 20.6% and 26.3%; p<0.0001 for both prevalence and incidence). However, when comparing CVD risk in SMI patients in each region with their respective control groups, there was no statistically significant difference anymore across regions, with both RRs and adjusted HRs showing comparably increased CVD incidence risk in the SMI population (RRs ranging from 1.17 in Europe to 1.63 in Asia, p=0.08; and HRs ranging from 1.58 in Oceania to 1.88 in both Europe and North America, p=0.29) (Table 6). There were insufficient numbers of studies to perform this analysis for adjusted ORs regarding prevalence rates across regions, or for adjusted ORs, RRs or HRs pertaining to specific CVD subgroups.

Table 6.

Prevalence and incidence of cardiovascular disease (CVD) in severe mental illness stratified by region

Regional strata	Analysis details	Prevalence of CVD	Incidence of CVD	Risk ratios for incident CVD	Adjusted hazard ratios for incident CVD
Asia	Pooled estimate, % (95% CI)	5.4 (4.3‐6.7)	2.6 (1.9‐3.6)	1.63 (1.31‐2.04)	1.75 (1.38‐2.22)
	p value	<0.0001	<0.0001	<0.0001	<0.0001
	Heterogeneity, I2 (p value)	98 (<0.0001)	100 (<0.0001)	99 (<0.0001)	96 (<0.0001)
	No. comparisons	8	12	9	10
Europe	Pooled estimate, % (95% CI)	9.7 (6.5‐14.2)	3.4 (2.2‐5.3)	1.17 (0.96‐1.42)	1.88 (1.44‐2.46)
	p value	<0.0001	<0.0001	0.11	<0.0001
	Heterogeneity, I2 (p value)	97 (<0.0001)	100 (<0.0001)	97 (<0.0001)	96 (<0.0001)
	No. comparisons	9	35	20	22
North America	Pooled estimate, % (95% CI)	14.6 (12.0‐17.7)	4.6 (3.4‐6.2)	1.39 (0.91‐2.12)	1.88 (1.62‐2.19)
	p value	<0.0001	<0.0001	0.13	<0.0001
	Heterogeneity, I2 (p value)	97 (<0.0001)	100 (<0.0001)	97 (<0.0001)	62 (0.003)
	No. comparisons	17	15	11	11
Oceania	Pooled estimate, % (95% CI)	20.6 (10.9‐35.4)	26.3 (24.1‐28.6)	1.52 (1.40‐1.66)	1.58 (1.41‐1.78)
	p value	<0.0001	<0.0001	<0.0001	<0.0001
	Heterogeneity, I2 (p value)	97 (<0.0001)	100 (<0.0001)	0 (0.72)	0 (0.84)
	No. comparisons	4	3	3	3
	p (difference between regions)	<0.0001	<0.0001	0.08	0.29

Meta‐regression

Due to heterogeneous or partial reporting of possible moderator variables in the included studies, all meta‐regression analyses were based on a much reduced number of studies. Hence, all analyses were less powered in comparison with the large sets of data used for cross‐sectional prevalence and longitudinal incidence analyses. Nonetheless, CVD incidence increased significantly with a higher percentage of patients using antipsychotics (12 studies; β=0.04, 95% CI: 0.01‐0.08, p=0.008), higher baseline body mass index in patients vs. controls (6 studies; β=0.24, 95% CI: 0.06‐0.42, p=0.008), and higher CVD prevalence at baseline in patients vs. controls (7 studies; β=0.07, 95% CI: 0.01‐0.14, p=0.03). CVD prevalence increased in more recent studies (38 studies; β=0.07, 95% CI: 0.02‐0.12, p=0.007), whereas the same was not true for CVD incidence (65 studies; β=–0.02, 95% CI=–0.07 to 0.01, p=0.21).

DISCUSSION

To our knowledge, this is the first large scale meta‐analysis providing comprehensive quantitative data on the prevalence and incidence of CVD in people with SMI, including both pooled data and comparisons across CVD and SMI diagnostic subgroups. Our results establish that approximately 10% of people with SMI with a mean age of 50 years have at least one comorbid CVD. Moreover, our longitudinal analysis documents a 3.6% incidence rate of CVD during a median of 8.4 years of follow‐up. Patients with SMI show a 53% higher risk for having CVD, a 78% higher risk for developing CVD, and an 85% higher risk of death from CVD compared to the regionally matched general population.

This study provides a worldwide epidemiologic representation of CVD prevalence and incidence rates in SMI, reporting the lowest absolute prevalence and incidence in Asia, increasing through Europe and North America, and reaching the highest levels in Oceania. However, in analyses with sufficient numbers of available studies, neither RRs nor adjusted HRs indicated significantly different CVD incidence risk across regions, meaning that SMI patients are at an increased risk across the world and that CVD risk‐reducing interventions in SMI are needed with the same urgency across all regions of the world. Moreover, while the prevalence and incidence of each CVD in people with SMI show some minor variations, people with major depressive disorder, bipolar disorder and schizophrenia are clearly all at an increased risk of CVD‐related deaths compared to population‐stratified controls, calling for urgent action.

We were able to identify some important and actionable moderators of increased CVD risk, including antipsychotic use, elevated body mass index and elevated baseline CVD. Based on these results, it is imperative that clinicians: a) only utilize antipsychotics, particularly for non‐psychotic conditions, when alternative treatment options with lower CVD risk potential have been tried sufficiently; and b) screen for and manage emerging and existing CVDs as well as their risk factors, including weight gain and elevated body mass index. Our data, adding to research demonstrating a significantly higher prevalence of metabolic syndrome in people with SMI compared to controls114, clearly suggest there is an urgent need to prevent and manage CVD risk in this population.

Our results demonstrating a higher CVD prevalence in SMI populations versus controls in more recent studies are also concerning, as they support accumulating data indicating that secondary prevention has been much less successful in the SMI population that in the general population, leading to a widening of the mortality gap in recent years49, 115, 116. Our findings confirm prior reports that antipsychotic medication use is associated with higher CVD risk13, 117, 118. However, due to limitations in the published data, we were unable to explore variations in CVD risk profiles between different antipsychotic medications13, 117, 118, 119, 120. Previous research has suggested that the highest cardio‐metabolic risks are associated with clozapine and olanzapine, whilst the lowest risk is with aripiprazole, ziprasidone, lurasidone, amisulpride and high potency typical antipsychotics13, 117, 118, 119, 120, 121, 122. However, in this context it is also important to note that antipsychotic medications can decrease CVD‐related mortality, as reported for example in Finnish79 and Swedish123 national database studies, that are highly generalizable. These data underscore that symptom control and functional improvement benefit both psychiatric and overall health, as severe psychiatric illness negatively affects lifestyle behaviors, medical care seeking and adherence to medical treatments. Thus, benefits of improved psychiatric status with antipsychotics and other psychotropic agents need to be carefully weighed against their potential for elevated cardiometabolic risk, which differs across available agents13, 117.

Since antipsychotic medication use moderates CVD risk and since antipsychotics are increasingly used as first line treatments for much more prevalent non‐psychotic conditions, including bipolar disorder124 and major depressive disorder with suboptimal response to antidepressant treatment125, the pool of people at an increased CVD risk is greatly enlarged. Therefore, research on the underlying mechanisms for the increased CVD risk after pharmacotherapy initiation is even more urgently needed to develop more effective and targeted preventive and interventional treatments. Studies should also examine whether different clinical subtypes of depression (i.e., melancholic, psychotic, atypical or undifferentiated) and bipolar disorder (e.g., type 1 or 2, cyclothymic disorder), certain mood states (manic, depressive, mixed or euthymic), or different antipsychotics, antidepressants or mood stabilizers13 significantly moderate CVD risk.

Furthermore, the pathophysiology underlying the association between SMI and CVD risk is complex and not well understood, clearly requiring further investigation. Emerging evidence suggests that SMI and CVD share pathophysiological features, including hypothalamic‐pituitary‐adrenal and mitochondrial dysfunction, peripheral immune activation, neuro‐inflammation, oxidative and nitrosative stress, as well as common genetic links and epigenetic interactions126. However, since these different mechanisms probably interact, research that integrates these pathways is urgently needed. Beyond mechanistic evaluations, such studies also need to investigate the general and specific effects of physical health improvements on SMI outcomes.

Future research should also investigate optimal monitoring regimens across stratified patient subgroups as well as the most effective timing and efficacy of primary, secondary and tertiary preventive interventions120, 127. In this regard, studies should comprehensively assess relevant moderator and mediator variables of CVD risk, including type and duration of specific psychotropic medications use, physical activity (including using passive monitoring via actimetry or mobile phone technology), diet, smoking, body mass index, personal and family history of CVD, in order to identify subgroups of patients who may require different monitoring and or interventions schemes. Long‐term follow‐up studies are also required to accurately document the emergence of more distal physical and mental health as well as health economic outcomes in relationship to the early identification and management of CVD risk factors and manifest CVD conditions in people with SMI.

Finally, since people with SMI engage in unhealthy lifestyle and often take psychotropic medication for extensive periods, long‐term follow‐up studies are needed that assess whether current predictor models based on the magnitude of traditional CVD risk factor effects observed in the general population apply or need to be adjusted for the SMI population93, in whom CVD risk factors also emerge at a far earlier age117, 128.

While this is the most comprehensive meta‐analysis of CVD risk in people with SMI conducted to date, we acknowledge several limitations that are largely related to factors in the primary data. First, lifestyle behavior information (e.g., physical activity) was inadequately reported, precluding meta‐analytic assessment of these important factors as moderating or mediating variables. People with SMI are less likely than the general population to engage in physical activity and have higher levels of sedentary behaviour129, smoke more130, consume diets that are high in saturated fats and refined sugars, while being low in fruit and vegetables131, all factors relevant for CVD risk. Second, variables such as clinical subtypes of major depressive disorder and bipolar disorder, negative symptom severity in people with schizophrenia, and concomitant or previous use of specific antipsychotics, antidepressants and mood stabilizers were not reported or were insufficiently reported or controlled for in most available studies. Third, as expected when combining observational data132, many of the results were moderately to highly heterogeneous. However, in accordance with the MOOSE guidelines133, we conducted meta‐regression analyses and were able to explain some of the observed heterogeneity. In addition, all of our results remained robust after adjustment for potential publication bias with the trim and fill analysis.

In conclusion, SMIs pooled together were significantly associated in cross‐sectional studies with CVD, coronary heart disease, cerebrovascular disease and CVD‐related death. Additionally, in longitudinal studies, each specific diagnostic SMI group was significantly associated with CVD and CVD‐related death. Furthermore, schizophrenia was associated with coronary heart disease and cerebrovascular disease, while bipolar disorder was associated with congestive heart failure, and major depressive disorder was associated with coronary heart disease, cerebrovascular disease, and congestive heart failure.

Importantly, our data confirm that CVDs are associated with an increased risk of mortality in people with SMI, which to a large part explains the shortened life expectancy of people with SMI compared to the general population2, 4, 5. Furthermore, we showed geographical variations in raw CVD prevalence and incidence risk in SMI populations, but no significant regional variance in the difference in CVD risk compared to the region‐specific general population. Finally, the fact that antipsychotic use, higher body mass index and baseline CVD significantly increased the risk for CVD morbidity and mortality underscores the urgent need to limit antipsychotic use to those populations truly requiring them, choosing the lowest risk antipsychotic agents first in the treatment algorithm, screening all SMI patients regularly for CVD risk factors and conditions, and addressing any identified abnormalities aggressively.