(C) PLOS One [1]. This unaltered content originally appeared in journals.plosone.org.
Licensed under Creative Commons Attribution (CC BY) license.
url:https://journals.plos.org/plosone/s/licenses-and-copyright

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



Temporal trends in associations between severe mental illness and risk of cardiovascular disease: A systematic review and meta-analysis

['Amanda M Lambert', 'Institute Of Applied Health Research', 'University Of Birmingham', 'Birmingham', 'United Kingdom', 'Helen M Parretti', 'Norwich Medical School', 'University Of East Anglia', 'Norwich', 'Emma Pearce']

Date: 2022-04

Abstract Background Severe mental illness (SMI; schizophrenia, bipolar disorders (BDs), and other nonorganic psychoses) is associated with increased risk of cardiovascular disease (CVD) and CVD-related mortality. To date, no systematic review has investigated changes in population level CVD-related mortality over calendar time. It is unclear if this relationship has changed over time in higher-income countries with changing treatments. Methods and findings To address this gap, a systematic review was conducted, to assess the association between SMI and CVD including temporal change. Seven databases were searched (last: November 30, 2021) for cohort or case–control studies lasting ≥1 year, comparing frequency of CVD mortality or incidence in high-income countries between people with versus without SMI. No language restrictions were applied. Random effects meta-analyses were conducted to compute pooled hazard ratios (HRs) and rate ratios, pooled standardised mortality ratios (SMRs), pooled odds ratios (ORs), and pooled risk ratios (RRs) of CVD in those with versus without SMI. Temporal trends were explored by decade. Subgroup analyses by age, sex, setting, world region, and study quality (Newcastle–Ottawa scale (NOS) score) were conducted. The narrative synthesis included 108 studies, and the quantitative synthesis 59 mortality studies (with (≥1,841,356 cases and 29,321,409 controls) and 28 incidence studies (≥401,909 cases and 14,372,146 controls). The risk of CVD-related mortality for people with SMI was higher than controls across most comparisons, except for total CVD-related mortality for BD and cerebrovascular accident (CVA) for mixed SMI. Estimated risks were larger for schizophrenia than BD. Pooled results ranged from SMR = 1.55 (95% confidence interval (CI): 1.33 to 1.81, p < 0.001), for CVA in people with BD to HR/rate ratio = 2.40 (95% CI: 2.25 to 2.55, p < 0.001) for CVA in schizophrenia. For schizophrenia and BD, SMRs and pooled HRs/rate ratios for CHD and CVD mortality were larger in studies with outcomes occurring during the 1990s and 2000s than earlier decades (1980s: SMR = 1.14, 95% CI: 0.57 to 2.30, p = 0.71; 2000s: SMR = 2.59, 95% CI: 1.93 to 3.47, p < 0.001 for schizophrenia and CHD) and in studies including people with younger age. The incidence of CVA, CVD events, and heart failure in SMI was higher than controls. Estimated risks for schizophrenia ranged from HR/rate ratio 1.25 (95% CI: 1.04 to 1.51, p = 0.016) for total CVD events to rate ratio 3.82 (95% CI: 3.1 to 4.71, p < 0.001) for heart failure. Incidence of CHD was higher in BD versus controls. However, for schizophrenia, CHD was elevated in higher-quality studies only. The HR/rate ratios for CVA and CHD were larger in studies with outcomes occurring after the 1990s. Study limitations include the high risk of bias of some studies as they drew a comparison cohort from general population rates and the fact that it was difficult to exclude studies that had overlapping populations, although attempts were made to minimise this. Conclusions In this study, we found that SMI was associated with an approximate doubling in the rate ratio of CVD-related mortality, particularly since the 1990s, and in younger groups. SMI was also associated with increased incidence of CVA and CHD relative to control participants since the 1990s. More research is needed to clarify the association between SMI and CHD and ways to mitigate this risk.

Author summary Why was this study done? Previous research has identified higher incidence and mortality from cardiovascular disease (CVD) in people with severe mental illness (SMI) than those without SMI.

It is not known whether the gap has changed over time. What did the researchers do and find? We conducted a comprehensive systematic review and meta-analysis of 108 cohort and case–control studies, including over 30 million participants.

Compared to the general population, people with SMI, particularly schizophrenia, generally had higher levels of cardiovascular-related mortality and a higher incidence of CVD.

For schizophrenia and bipolar disorder (BD), the rate ratios for coronary heart disease (CHD) mortality and cardiovascular mortality were larger in the 1990s and 2000s than in earlier studies, and the rate ratios were larger in younger people. What do these findings mean? We confirmed a strong association between SMI and incidence of and mortality from CVD.

This association became stronger in the 1990s and 2000s.

Citation: Lambert AM, Parretti HM, Pearce E, Price MJ, Riley M, Ryan R, et al. (2022) Temporal trends in associations between severe mental illness and risk of cardiovascular disease: A systematic review and meta-analysis. PLoS Med 19(4): e1003960. https://doi.org/10.1371/journal.pmed.1003960 Academic Editor: Alex Dregan, King’s College London, UNITED KINGDOM Received: July 12, 2021; Accepted: March 8, 2022; Published: April 19, 2022 Copyright: © 2022 Lambert et al. This is an open access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited. Data Availability: All relevant data are within the manuscript and its Supporting Information files. Funding: This report presents independent research funded by the National Institute for Health and Care Research (NIHR). AML and TM are supported by the NIHR Applied Research Collaboration (ARC) West Midlands. HMP (NIHR Academic Clinical Lectureship) was funded by the NIHR during some of this research. MJP was supported by the NIHR Birmingham Biomedical Research Centre at the University Hospitals Birmingham NHS Foundation Trust and the University of Birmingham. The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript. Competing interests: I have read the journal’s policy and the authors of this manuscript have the following competing interests: CUC has been a consultant and/or advisor to or has received honoraria from: AbbVie, Acadia, Alkermes, Allergan, Angelini, Aristo, Axsome, Damitsa, Gedeon Richter, Hikma, IntraCellular Therapies, Janssen/J&J, Karuna, LB Pharma, Lundbeck, MedAvante-ProPhase, MedInCell, Medscape, Merck, Mitsubishi Tanabe Pharma, Mylan, Neurocrine, Noven, Otsuka, Pfizer, Recordati, Rovi, Servier, Sumitomo Dainippon, Sunovion, Supernus, Takeda, Teva, and Viatris. He provided expert testimony for Janssen and Otsuka. He served on a Data Safety Monitoring Board for Lundbeck, Rovi, Supernus, and Teva. He has received grant support from Janssen and Takeda. He is also a stock option holder of LB Pharma. MS received honoraria/has been a consultant for Angelini, Lundbeck, Otsuka. MLO receives funding from The Liaison Committee for education, research and innovation in Central Norway. Abbreviations: BD, bipolar disorder; CHD, coronary heart disease; CI, confidence interval; CVA, cerebrovascular accident; CVD, cardiovascular disease; DALY, disability-adjusted life year; DSM, Diagnostic and Statistical Manual of Mental Disorders; FGA, first-generation antipsychotic; HR, hazard ratio; ICD, International Classification of Diseases; MOOSE, Meta-analysis Of Observational Studies in Epidemiology; NOS, Newcastle–Ottawa scale; OR, odds ratio; PRISMA, Preferred Reporting Items for Systematic Review and Meta-Analyses; RR, risk ratio; SGA, second-generation antipsychotic; SIR, standardised incidence ratio; SMI, severe mental illness; SMR, standardised mortality ratio; WHO, World Health Organization

Introduction Mortality in people with mental disorders is more than 2.2 times that of the general population and more than 2.5 times in those with psychoses, according to a systematic review and meta-analysis of 148 studies [1]. A recent estimate suggested that mental health disorders could be associated with 14.3% of all deaths across the globe, corresponding to around 8 million deaths each year [1]. The increased premature mortality seems to be the highest in schizophrenia spectrum disorders (schizophrenia), who die around 14.5 years earlier than people without schizophrenia, according to data from 11 studies and 247,603 patients [2]. In a large national cohort study, the mortality rate ratio was 3.2 in 104,000 people with schizophrenia [3]. In bipolar disorder (BD), figures are also concerning, with life expectancy decreased by 9 to 12 years in males and 8 to 10.5 years in females [4]. In a large cohort study, the mortality rate ratio was around 2.4 in 58,000 participants with BD compared with the general population [3]. More than two-thirds of the increased mortality in people with mental health disorders has been estimated to be due to natural causes [1]. When looking at cause-specific mortality in those with mental health disorders, cardiovascular disease (CVD) plays a major role [5]. According to a recent large national cohort study in Taiwan, the standardised mortality ratio (SMR) for CVD-related mortality was 3.02 (95% confidence interval (CI): 2.80 to 3.24) in those with schizophrenia and 1.87 (95% CI: 1.70 to 2.05) for BD [3]. While excess mortality from CVD for people with severe mental illness (SMI) is established, the relationship with CVD incidence is less clear. CVD is associated with modifiable risk factors such as smoking, hypertension, obesity, dyslipidemia, diabetes, alcohol consumption, and physical inactivity. People with SMI are known to have higher prevalence of risk factors that are associated with CVD [6] and face barriers to healthcare that reduce their ability to benefit from effective treatment [7–11]. Despite high CVD-related mortality rates, CVD may remain undiagnosed in people with schizophrenia and BD. A recent study found that people with schizophrenia were 66% less likely to be diagnosed with CVD prior to CVD-related death than individuals without SMI [12]. Women with BD were 38% less likely than control participants to be diagnosed. The disease burden of both SMI and CVD is greater among higher-income countries compared with middle- or lower-income countries. Mortality from CVD is particularly high among high-income countries (294 deaths per 100,000 compared with 239.9 globally) [13]. There is also variation in the prevalence of schizophrenia, with a higher prevalence in developed countries compared to less developed countries (median 3.3 versus 2.6 per 1,000) [14]. The course, severity, and outcome of schizophrenia are generally worse in high-income compared with low-income countries. For example, in a World Health Organization (WHO) 10-country study, remission of their condition was experienced by 62.7% of patients with schizophrenia in developing countries, compared with only 36.8% in developed countries [15]. The rates of disability-adjusted life years (DALYs) from schizophrenia and BD are higher among World Bank high-income countries than other countries. In 2019, the rate of DALYs in high-income countries was 244.18 per 100,000 for schizophrenia and 174.69 for BD. This compares with 195.25 and 100.84 globally [16]. There have been modest, but steady reductions in CVD incidence in the general populations of high-income countries in the last 3 decades [17]. According to the Global Burden of Disease study, incidence of coronary heart disease (CHD) and stroke decreased by 13% and 7%, respectively, in World Bank high-income countries [18] between 1990 and 2019. This may be due to overall improvements in levels of some risk factors, in particular smoking and blood pressure levels [19]. Improved treatments for CVD have also reduced mortality [20]. However, it remains unclear whether the risk of CVD diagnosis and mortality for SMI patients has changed over time. According to one meta-analysis, the gap in premature mortality between people with SMI and the general population does not seem to have changed over time [2]. It remains unknown whether the gap has remained unchanged also for CVD-related mortality. A recent large cohort study did not describe substantial changes from 2005 to 2010 regarding CVD-related increased mortality in schizophrenia compared with the general population (SMR 3.16 to 3.02), although there was a 10% reduction in SMR for BD (2.08 to 1.87) [3]. However, another cohort study found a widening gap in CVD-related mortality between SMI and the general population [21]. Yet, to the best of our knowledge, a comprehensive systematic review and meta-analysis on the trend over time of increased CVD-related mortality and incidence in schizophrenia and BD is currently missing. Indeed, previous reviews of the association between SMI and CVD have mainly focused on either a single SMI diagnosis (schizophrenia and BD) or a single cardiovascular outcome (myocardial infarction and stroke) or considered CVD mortality separately from incidence, and some excluded unpublished or non-English language studies [5,22–29]. These previous reviews have shown inconsistencies in included studies and differences in the scale and direction of associations. There are also recent large studies not included in previous reviews [30]. Most importantly, none of the previous reviews specifically investigated time trend of CVD-related mortality risk in those with schizophrenia and BD. An up-to-date, comprehensive, systematic review of the association between SMI and CVD incidence and mortality is therefore indicated. The aim of this systematic review and meta-analysis was to specifically investigate changes over time of CVD-related mortality and incidence, in addition to providing updated pooled estimates of increased CVD mortality and incidence in those with schizophrenia and BD compared with the general population in high-income countries.

Methods The present systematic review follows a prepublished publicly available protocol registered with PROSPERO (CRD42017058068) and adheres to Preferred Reporting Items for Systematic Review and Meta-Analyses (PRISMA) guidelines [31] and to Meta-analysis Of Observational Studies in Epidemiology (MOOSE) [32]. PRISMA 2020 checklist and abstract checklist are available in S1 and S2 Files, respectively. Minor deviations from the protocol are listed in S3 File. Inclusion criteria Population. Eligible studies included participants aged 16 to 65 years at onset of psychotic disorder. The upper age limit was chosen because the onset of psychotic disorders is uncommon at older ages, but transient psychotic conditions are associated with dementia, making it difficult to distinguish between this and SMI [33,34]. Studies were included if at least 90% of participants were within the age range. Only studies in high-income countries (defined as International Monetary Fund “Advanced Economies” [35]: Western Europe, Canada, United States of America, Australia, New Zealand, Taiwan, South Korea, and Japan) were included due to differences between high- and low- or middle-income countries in the diagnosis, prevalence, treatment, and prognosis of SMI [7,8]. Exposure. The exposure of interest was a diagnosis of SMI (schizophrenia, other psychoses, and BD). SMI was defined by the Diagnostic and Statistical Manual of Mental Disorders (DSM), International Classification of Diseases (ICD), and other clinical coding systems (e.g., OPCRIT, OXMIS, and Read codes), operational systems, or clinical diagnosis recorded in patient notes [36–39]. Dissociative, multiple personality disorders, depression (in the absence of psychosis or BD), dementia, learning disabilities, and obsessive-compulsive disorder were excluded. Comparator. Eligible studies compared people with SMI to control participants without SMI or to the general population. Participants could be in an inpatient, outpatient, or community setting. Outcome. The primary outcomes were mortality from, or incidence of, CVD (i.e., CHD, cerebrovascular accident (CVA), or heart failure) subsequent to SMI diagnosis and its time trend. Studies that reported composite outcomes including both fatal and nonfatal CVD events were analysed together with incidence studies as the majority of CVD events in these studies were not fatal. Studies were eligible if they described methods to exclude CVD at baseline. However, some studies reporting CVD-related mortality did not exclude individuals with prevalent CVD. These studies were included only if a prevalence of CVD below 10% at baseline was reported. Studies reporting all circulatory disease mortality were included [40] because two-thirds of deaths from all circulatory disease (ICD-10 Chapter I00-I99) are from CHD or stroke. However, those reporting all incident circulatory disease were excluded as incident circulatory disease diagnoses are mainly CVD risk factors [41]. Studies reporting only CVD risk factors were excluded. Studies were eligible if the mean follow-up period was at least 1 year. Study design. Designs could include cohort studies or case–control studies but not cross-sectional, case series, case reports, or uncontrolled studies. Case–control studies compared the odds of SMI compared with no SMI in people who died from, or were newly diagnosed with, CVD compared with matched control participants. See S4 File for full inclusion criteria. Search strategy The following bibliographic databases were searched up to November 30, 2021: MEDLINE, EMBASE, and PsycINFO via Ovid and CINAHL via EBSCO. The British Library’s ZETOC and Web of Science Conference Proceedings Citation Index were searched to the same date for conference abstracts and the Cochrane Library CENTRAL database searched up to December 1, 2021 for relevant trials. Initial searches were performed in 2017 and updated in 2019 and again in 2021 using the same search methods but narrowing the searches to 2017 and 2019 onwards, respectively. No language restrictions were applied. Citations of relevant reviews and included studies were searched as was grey literature, including OpenGrey. Detailed search strategies are included in S5 File. Screening and data extraction Pairs of reviewers, working independently, firstly screened titles and abstracts and secondly screened full texts. Disagreements were resolved by discussion and if necessary, by reference to a third independent reviewer. Data extraction was conducted by a first reviewer (AML) and checked by a second (EP) using a form adapted from a previous review [5] (S6 File). If necessary, study authors were contacted for more information. Hazard ratios (HRs), relative risks, rate ratios, risk ratios (RR), odds ratios (ORs), SMRs, or standardised incidence ratios (SIRs) measuring the association between SMI and CVD were extracted or sufficient information to calculate them. Risk measures adjusted for factors related to cardiovascular risk (e.g., socioeconomic factors, comorbidities, smoking, and alcohol use) were extracted where available; otherwise, measures adjusted for age and sex only were extracted. Risk measures were extracted from graphs if not reported in the text. CIs for effect estimates were extracted or calculated from available information. For studies reporting results for the same cohort that were stratified over calendar time periods, estimates for the whole period were selected if available; otherwise, the earliest time period was selected to avoid a “healthy survivor” effect [30,42–44]. However, for studies that reported SMRs with multiple periods of follow-up [45], actual and expected deaths were summed and the ratio calculated to provide an estimated overall SMR. SMRs for males and females were combined to give a figure for persons if that was unreported. CIs were calculated using the method described by Kirkwood and Sterne [46]. Risk of bias assessment Risk of bias was assessed using a version of the Newcastle–Ottawa scale (NOS) [47] adapted to include extra criteria on accuracy of outcome measurement and identification of confounders from another checklist for cohort studies [48] (S7 File). Risk of bias was assessed with the NOS independently by 2 researchers. Any discrepancies were discussed, and consensus reached with the aid of a third reviewer when required. Data synthesis Narrative and quantitative synthesis of evidence were conducted, and study characteristics recorded in tables for all included studies. Where appropriate, meta-analysis was conducted using DerSimonian and Laird random effects models due to expected population, clinical, and methodological heterogeneity between studies [49]. Heterogeneity was measured using I2 and chi-squared statistics and 95% prediction intervals [50]. Associations between different SMIs, CVD outcomes, and type of effect estimate were presented in separate subgroups in forest plots and meta-analyses. Rate ratios were combined with HRs since they have the same interpretation. Studies that may have included the same patients in overlapping periods of recruitment were not included in the same meta-analyses to avoid double counting outcomes. Rules were devised to determine which studies should be included if the period of overlap was greater than 10% of the total follow-up time (S8 File). Those with fewer exposed cases, or shorter follow-up durations, were excluded to minimise the risk of double counting events. Where calculable, results for both sexes were included in the meta-analysis; otherwise, separate results for males and females were used. For each meta-analysis with more than 10 studies, potential publication bias or small study effects were assessed visually using funnel plots and formally tested using Egger test [51]. If evidence of significant bias was found, a random effects trim-and-fill procedure [52] was applied to estimate the effect of potential unpublished studies. To assess temporal trends, studies were allocated to decades according to the median year of follow-up. Further analysis increased the granularity by allocating to 5-year periods according to the median year of follow-up. All analyses were performed in STATA MP version 14.2 (StataCorp. 2015. Stata Statistical Software: Release 14. College Station, TX: StataCorp LP). Subgroup analyses Analysis of prespecified subgroups was conducted. If there were at least 10 study results, random effects meta-regression was performed to explore potential sources of heterogeneity that were measured at study group level [53,54]. Temporal trends, (decade of study conduct), setting (community; community and inpatient; and inpatient), and world region (North America [USA and Canada]); United Kingdom; Scandinavia [Denmark, Finland and Sweden]; Other Western Europe and Israel; Oceania [Australia and New Zealand]; and Far East) were included as covariates in univariate meta-regressions for each subgroup combination of SMI diagnosis (schizophrenia and related conditions; BD), and each mortality outcome (all circulatory disease; CHD; and CVA). Heterogeneity was assessed for each meta-regression by comparing the I2 from the random effects meta-analysis and the residual I2 after fitting covariates. The amount of between-study variation accounted for by covariates was measured by calculating adjusted R2. p-Values for model fit were used to assess the association between covariates and SMI and outcomes [55,56]. Subgroup analyses of cohort studies were conducted by sex and by age group at baseline and assessed by forest plots. Studies that reported results for ages ≤60 years were plotted separately from results for older groups or all ages. Sensitivity analyses Sensitivity analysis was performed to assess the effect of making different decisions on inclusion of studies in the meta-analyses, where studies were subject to overlapping periods of recruitment and thus may have included the same patients, as described above. The effect of inclusion of results for different age groups, medications, or follow-up periods was also assessed in sensitivity analyses. Sensitivity analysis was conducted to test the effect of replacing included studies with studies that had been excluded due to overlapping recruitment periods. As there was only one case–control study, this was combined with cohort studies.

Discussion This large-scale systematic review has been conducted in a methodologically robust way and has yielded up-to-date comprehensive evidence on the association between SMI- and CVD-related mortality and incidence. It has found that the CVD mortality rate for people with SMI is around twice that of the general population. People with schizophrenia are at greater risk than those with BD; however, the disparity exists across all types of SMI and both cerebrovascular and cardiac mortality. The review also found a consistent relationship between SMI and increased incidence of CVA, CVD events, and heart failure. The relationship between SMI and CHD was less clear; a positive association only became apparent for high-quality studies with low risk of bias. Some evidence was found of changes in the risk of CVD over time, and results from some studies suggest that the gap between SMI and the general population appears to be widening. The size of the association between SMI and both CHD and all circulatory mortality increased until the 2000s, following this it appeared to decline. Estimated risks of the associations between schizophrenia and CVA mortality were also highest in the 1990s, whereas for BD, they were highest for CVA mortality in the 2000s. Studies reporting incidence of CVD since the 1990s had larger associations on most measures between SMI and diagnosis of CVA and CHD compared with earlier studies. ORs appeared to show the reverse trend; however, only 5 studies contributed results that could be used in this trend analysis. The apparent increases in CVD risk for people with SMI relative to control participants that we have found in recent decades are in marked contrast to the reducing rates of CVD experienced by the general population. Since 1990, there has been a steady decline in the rates of CVD mortality and incidence in the populations of high-income countries [17]. Most European countries have also seen a halving in overall population CHD mortality since the 1980s [75]. Increasing trends in CVD mortality for people with SMI relative to control participants have been noted by other authors [30,44,76,77]. An increase in CHD hospital admissions for people with schizophrenia compared with the general population since the 1990s has also been reported [44]. The increased relative risk of CVD diagnosis in more recent decades may be a result of disparity in smoking prevalence between people with SMI and the general population [78,79] or increased use of antipsychotics. The changes since the 1990s approximately coincide with the release of newer, second-generation antipsychotics (SGAs), which are known to have worse metabolic effects [80–82]. The use of these drugs has increased over the last 20 years for treatment of both schizophrenia and BD [83–86]. There was significant heterogeneity between studies, which was partly explained by decade of results and world region, the latter due to differences in CVA incidence between the Far East and the UK. Although some evidence of differences between decades and between regions were found, the large number of CIs calculated from multiple meta-regressions mean that findings should be interpreted cautiously. Some evidence of small study effects was found by the review, affecting the association between schizophrenia and CVD mortality. However, publication bias is unlikely to account for the increased risk of CVD mortality found in the review as the pooled effect remained elevated, even after application of the trim and fill method. Small study bias could only be assessed for incidence studies when both types of SMI exposure and all CVD outcomes were combined. Although significant bias was found, analysis found no additional missing studies. This analysis may be flawed as it was not possible to exclude double counting of populations at risk or of outcomes. Thus, results may not have been independent, and the random effects assumption may have been violated. However, the analysis is of value in demonstrating that the scale of any potential publication bias is likely to be small. Comparison with previous literature Mortality. Schizophrenia Cerebrovascular accident. In contrast to all circulatory disease and CHD, the pooled CVA mortality ratio for schizophrenia, SMR: 1.93 (95% CI: 1.63 to 2.28, p < 0.001), is in the opposite direction to that found by other authors [24,28]. It is unclear which studies were included in these reviews, but the reviews were relatively small: One review only included 3 studies (median SMR: 0.69, IQR: 0.61 to 1.30) [24], and the other reported an SMR based on only 490 deaths (SMR: 81, 95% CI: 73 to 87) [28]. The current review, however, included 14 studies in the pooled SMR; it was based on 2,800 deaths, and all studies showed a higher risk of CVA mortality for people with schizophrenia compared to the general population. Pooled HRs and rate ratios were also elevated and were not reduced by the substitution of excluded studies in sensitivity analyses. Coronary heart disease. The pooled result for the 14 studies reporting CHD SMRs for schizophrenia, SMR: 1.94 (95% CI: 1.58–2.37, p < 0.001), was higher, but in the same direction as other smaller reviews [24,28]. Further, sensitivity analysis showed that substituting some of the excluded studies could change the pooled SMR and HR/rate ratio. Small study effects or publication bias could account for some of this discrepancy; an analysis accounting for small study bias reduced the estimated risk to SMR: 1.61 (95% CI: 1.11 to 2.33, p = 0.012). All circulatory disease. The pooled results for circulatory mortality in schizophrenia, from 28 studies (HR/rate ratio: 1.91, 95% CI: 1.52 to 2.41, p < 0.001; SMR: 1.96, 95% CI: 1.61 to 2.39, p < 0.001), were slightly lower than, but in broad agreement with, those of a recent review based on 9 studies [5]. Sensitivity analysis found that rate ratios were increased by the inclusion of one study that had high HRs [87] that was also included in the previous review [5]. Bipolar disorder All circulatory disease. The pooled SMR of 1.65, 95% CI: 1.53 to 1.77, p < 0.001 and combined HR and rate ratio of 1.52, 95% CI: 0.84 to 2.75, p = 0.162 for all circulatory mortality in BD are similar to ratios reported by other authors [5,27]. Incidence. Schizophrenia Cerebrovascular accident. Previous reviews have also reported increased risks of stroke with schizophrenia and with bipolar disease [5,22,23,25]. Coronary heart disease. The current review found that people with schizophrenia are less likely to be diagnosed with CHD than the general population, in contrast to previous reviews [5,25]. The disparity may be explained by variation in included primary studies. Heterogeneity in incidence of CHD could reflect differences in ascertainment. Five schizophrenia studies only included inpatient diagnosis of CHD [65,88–91], which could have missed both milder cases treated in the community and deaths from undiagnosed CHD outside of hospital. Bipolar disorder Coronary heart disease. The nonsignificant positive associations between CHD and BD were similar to those reported elsewhere [5,22]. Schizophrenia and bipolar disorder All circulatory disease. The increased likelihood of total CVD events and SMI (both schizophrenia and BD) correspond with the findings of previous reviews [5,25]. Change over time. We have found evidence that the gap in CVD mortality between people with SMI and the general population appears to be widening. This finding supports the observations in some of the included studies that CVD mortality has increased in recent years for people with SMI relative to the general population [21,44,76,77]. This is likely to be due to a number of factors. In developed countries, adult smoking rates have fallen for the general population in recent decades: from average rates of 33.1% in 1980 to 23.5% in 2012 [92]. However, people with long-term mental health problems have not seen a similar decline [78,79]. SGA medication has been in use since the 1990s [81,82]. However, there is a possible trade-off in their use between the benefits of treating psychotic symptoms against the metabolic problems that can result which increase the risk of CVD. Use of some SGAs, especially clozapine and olanzapine, is likely to result in weight gain, particularly with long-term use [93]. Certain SGAs (particularly clozapine, olanzapine, and quetiapine) are associated with a higher risk of diabetes than first-generation antipsychotics (FGAs) [94]. There is also some evidence of elevated lipid levels with use of clozapine, olanzapine, and risperidone compared with FGAs [95]. Differential access to physical health treatments for people with SMI is also an issue. Rates of cardiac revascularisation procedures have increased over the last 20 years [96]. However, people with schizophrenia who have acute coronary syndrome are much less likely than others to receive this type of treatment [8]. Strengths and limitations Strengths. The strengths of this review are its size and scope. Both schizophrenia and BD and different cardiovascular diagnoses were included. A wide search strategy was used, not limited by language or date of publication. Studies reporting different measures of relative risk were included, and changes over time were investigated. Robust methodology including duplicate screening, and data extraction was used. Possible sources of heterogeneity between studies were investigated. Most studies had an average follow-up duration of at least 5 years, some considerably longer, and long enough for CVD outcomes to occur. The adequacy of reporting follow-up was generally good as many of the studies were carried out in Scandinavia, which has near complete recording of population health records. Limitations. There are also several limitations. It was difficult to exclude studies that had overlapping populations. Although attempts were made to minimise this, it is possible that there may be some double counting of patient populations. This means that the CIs may be slightly narrower than they would have been if all duplicated patients had been excluded. Many of the studies were at high risk of bias as they drew a comparison cohort from general population rates. The majority of studies did not adjust for cardiovascular risk factors. Furthermore, it was not possible to use reported results that were adjusted for important confounders as there was inconsistency in the confounders reported. Despite including conference abstracts, there was some evidence of small study bias among studies reporting CHD and circulatory mortality for schizophrenia. These were studies with a higher risk of bias that used general population mortality rates to report SMRs. However, even after adjustment for possible unpublished studies, the risk of excess mortality remained elevated. Although the literature searches found many papers measuring the association between SMI and the risk of cardiovascular mortality, searching one database for the additional search term “cause of death” revealed some additional studies. Thus, a small number of relevant studies may have been missed by failing to include this term in all database searches. However, any effect on the results is likely to be small due to the large number of included studies. SMRs were calculated from available information if not reported directly (S24 File). If only whole numbers of expected deaths were available, this may have led to rounding errors. This is unlikely to have had much effect on the results. Change over time. Temporal changes were assessed using decade of median year of follow-up. By comparing estimates from different studies from different time periods, we are performing an ecological analysis. Studies could have had outcomes spanning multiple decades but would only have been counted against one. This methodology may have reduced the apparent difference in relative risk between decades highlighted in the review. More insight on temporal changes may be gained by considering large, long-term cohort studies, and the use of individual patient data. SMI definitions. There was some inconsistency in definitions of SMI diagnosis, which could have accounted for some of the heterogeneity between studies. Most studies used DSM or ICD coding to define SMI, but some defined SMI less robustly as confirmed by psychiatrist or otherwise recorded in notes [77]. Around half the included studies reporting CVD outcomes for schizophrenia grouped schizoaffective disorder together with schizophrenia. Another study grouped schizoaffective disorder with BD [97]. Schizophrenia causes more significant disability than BD [98]; schizoaffective disorder has symptoms of both conditions, so its inclusion could have reduced the average severity of schizophrenia exposure or increased it for BD [99]. Some definitions of schizophrenia included schizotypal disorder [89,100–106] (a type of nonpsychotic personality disorder below the clinical threshold for schizophrenia) [36]. Others did not specify either the diagnostic system or the codes used [45,107–113], so severity was not able to be determined. The severity of schizophrenia cases could also have been reduced by the exclusion of schizophrenia with substance use disorder [101]. Although studies that included psychotic depression were eligible for inclusion, other forms of major depressive disorder were not, due to the difficulty in distinguishing major depression from other types and severity of depression. Studies where all participants were treated with lithium included those diagnosed with BD and schizoaffective disorder. However, unipolar depression and other SMI diagnoses that did not meet inclusion criteria were also present [97,114,115]. It was not possible to exclude these people, so some misclassified patients were included, although numbers were small. Categorisation of SMI in other studies may also be subject to misclassification due to variation in psychiatric diagnostic labelling [116]. CVD ascertainment. Most studies did not demonstrate adequately that CVD was not present at the start of the study. Although studies were selected such that SMI was diagnosed before cardiovascular death, presence of diagnosed or undiagnosed CVD prior to death was unknown for most studies. In incidence studies, although it is likely that SMI diagnosis was made at a younger age than incidence of CVD, it was not possible to be confident that CVD did occur after the start of SMI diagnosis. Age. Studies were included regardless of the participants’ age at cardiovascular outcome, which was not known for most studies. However, larger associations between SMI and CVD were noted among mortality studies that did report relative risks for younger age groups. This could be due to a healthy survivor effect: Those with less SMI are more likely to survive to older ages. Confounding factors. Not many mortality studies adjusted for confounding factors, other than age and sex, so exploration of the effects of important covariates, such as smoking, ethnicity, deprivation, and other cardiovascular risk factors was not possible. However, some studies noted that antipsychotics were associated with higher risks of cardiovascular mortality [87,117,118]. More research is needed to assess the effects of confounders on CVD mortality. Most of the incidence studies did not adjust for important confounders with inconsistency in the confounders reported for different studies. Using results that were adjusted for confounders made little difference to the results, but it is likely that adjusting for factors such as smoking, comorbidities, obesity, or antipsychotic use may explain the increased risks. Setting. Included studies were generally representative of persons with SMI. Those that included only inpatients were likely to have missed less severe cases, potentially overestimating risks, although no significant heterogeneity due to setting was detected. Community settings were underrepresented in the review: Many of the recruited exposed populations were inpatients. Consequently, the review includes an overrepresentation of participants with more severe symptoms of exposure and outcome. A lack of patients with relatively mild symptoms from primary care settings may explain some of the apparent underdiagnosis of CHD seen in the review. Risk of bias. There was a risk of selection or information bias affecting many of the included studies. The risk of CVA diagnosis and total CVD events found for all studies remained elevated among the high-quality studies. However, risk of CHD diagnosis was elevated only for the high-quality studies. Reporting by sex. Results for all persons were included where possible, but for some studies, data were only available separately for males and females. In these cases, results from each sex were included separately in the meta-analyses. However, results for males and females within the same study are not independent, so the random effects assumption of the meta-analysis could be violated. In practice, this is unlikely to have much effect on the conclusions.

Conclusions This systematic review and meta-analysis found that people with SMI are at increased risk of mortality from CVD compared with the general population and that the risk may be increasing over time. It also found that the risk of diagnosis with stroke, heart failure, and major cardiovascular events was elevated for people with SMI compared with control participants. It was not possible to explore the effects of confounders, such as smoking and obesity, so at least some of the excess risk may be attributable to these residual factors. There was considerable heterogeneity between studies, partly explained by decade of results, world region, and age group. Prediction intervals included unity for some combinations of SMI and CVD mortality; thus, a future study may not find an increased risk of CVD mortality for people with SMI. However, this review included results from 82 studies, only 7 of which reported any estimated risks smaller than one. Any future study is unlikely to materially change the review findings. Deaths from CVD make a large contribution to the inequalities in life expectancy experienced by those with SMI [6]. This result highlights the need for targeted interventions to improve the CVD health of people with SMI. A better mechanistic understanding of the reasons for the higher mortality rates would enable identification of where inequalities could be reduced. More research is needed to explore the role of smoking, obesity, antipsychotic medication, and access as well as adherence to CVD treatments for the excess cardiovascular mortality. More research is needed to understand the reasons for the higher morbidity risk and to assess why it may have been worsening in recent decades. Such research should include the investigation of the relationship to cardiovascular risk factors, such as obesity, diabetes, blood pressure, smoking, and antipsychotics, and the role of access and adherence to treatments for these risk factors. Cohort studies accounting for these, and other important confounders, could help further investigate these factors. Author contributions AML conducted the literature searches, carried out the analysis, and drafted the review. TM, HMP, and RR critically appraised and provided guidance during the development of the protocol and the review process. TM and HMP edited and redrafted the manuscript. AML, EP, HMP, TM, MO, MR, NT, TSA, GM, AL, and KA screened titles and abstracts or reviewed full texts. AML and EP conducted data extraction and risk of bias assessment. MJP provided advice on statistical analysis. CUC and MS contributed data from a previous review and the data extraction form. All authors read, edited, and approved the final version of the final manuscript.

Acknowledgments The authors would like to thank Danai Bem for screening titles and abstracts and providing guidance and critical appraisal during the development of the protocol and the review process and Arwa Abdel-Aal for help with piloting the title and abstract screening used in this systematic review. Disclaimers The views expressed in this publication are those of the authors and not necessarily those of the NIHR, NHS, the UK Department of Health and Social Care or other funding institutions.

[END]

[1] Url: https://journals.plos.org/plosmedicine/article?id=10.1371/journal.pmed.1003960

(C) Plos One. "Accelerating the publication of peer-reviewed science."
Licensed under Creative Commons Attribution (CC BY 4.0)
URL: https://creativecommons.org/licenses/by/4.0/


via Magical.Fish Gopher News Feeds:
gopher://magical.fish/1/feeds/news/plosone/

You are viewing proxied material from magical.fish. The copyright of proxied material belongs to its original authors. Any comments or complaints in relation to proxied material should be directed to the original authors of the content concerned. Please see the disclaimer for more details.