Article Text

Download PDFPDF
Orthostatic hypotension and mortality risk: a meta-analysis of cohort studies
  1. Wei Xin1,
  2. Zhiqin Lin2,
  3. Shuhua Mi1
  1. 1Department of Cardiology, Beijing Anzhen Hospital, Capital Medical University, Beijing, China
  2. 2Department of Cardiovascular Medicine, Affiliated Hospital of Guiyang Medical College, Guizhou, China
  1. Correspondence to Professor Shuhua Mi, No 2 Anzhen Road, Chaoyang District, Department of Cardiology, Beijing Anzhen Hospital, Capital Medical University, Beijing 100029, China; mishuhua{at}hotmail.com

Abstract

Context Orthostatic hypotension (OH) has been found to be related to increased risk of mortality, although results of previous cohort studies were not always consistent.

Objective The aim of this meta-analysis was to investigate the association between OH and mortality risk.

Data source Medline, Embase and references cited in related reviews and studies.

Study selection Cohort studies evaluating the association between OH and mortality risk were identified.

Data extraction Two investigators read all papers and extracted all relevant information.

Results A total of 56 125 subjects with 11 580 mortality cases from nine cohorts were included in the meta-analysis. Pooled results indicated that the presence of OH at baseline was significantly associated with an increased risk of all-cause mortality (adjusted risk ratio (RR) = 1.40, p<0.001). Subgroup analysis suggested that the association between OH and all-cause mortality was less strong for the studies in which classic risk factors were adequately adjusted compared with those in which they were not adequately controlled. In addition, although our meta-analysis failed to reveal a significant association between OH and cardiovascular (CV) mortality (RR=1.20, p=0.47), we did find that subjects with OH had an increased risk of non-CV mortality (RR=1.18, p=0.05).

Conclusions The presence of OH is associated with a significantly increased risk of all-cause mortality, which may partially be mediated by classic risk factors. Further research is needed to determine whether the association between OH and mortality risk is causal.

Statistics from Altmetric.com

Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.

Introduction

Orthostatic hypotension (OH) is a multifactorial clinical syndrome which occurs when the cardiovascular (CV) reflexes fail to maintain blood pressure (BP) on standing from a supine position.1 ,2 The prevalence of OH varies significantly depending on the population studied. Although OH is more prevalent in old people, epidemiological studies have suggested that OH can also be detected in middle-aged populations.3 ,4 Moreover, OH is quite common in hospitalised patients.5 A previous study has shown that, in the USA, the prevalence of OH in hospitalised patients is 36 per 100 000 patients, with a higher rate of 233 per 100 000 in those aged 75 years or above.6 Of note, although some patients with OH may present with symptoms of dizziness or syncope, most are asymptomatic, which may make OH underdiagnosed in the usual clinical settings.1 ,7 The clinical relevance and implications of OH have been intensively studied during the past decade. OH has not only been linked to the incidence of many CV diseases, such as coronary artery disease (CAD),8 stroke,9 congestive heart failure10 and atrial fibrillation,11 the presence of OH has also been related to increased risk of many non-CV-related conditions, including dementia,12 ,13 falls14 and chronic kidney disease.15 More importantly, it has increasingly been shown that OH is significantly associated with a higher risk of future mortality events,3 ,4 ,16–19 although the results of these studies are not always consistent.20–22 Therefore, in this meta-analysis, we summarise the relevant studies and provide an overall quantitative estimate of predictive ability of OH for all-cause mortality. Furthermore, we also explore whether OH is associated with increased risk of both CV and non-CV mortality.

Methods

We tried to follow the Meta-Analysis of Observational Studies in Epidemiology protocol23 and Cochrane Handbook guidelines24 throughout the design, implementation, analysis and reporting for this study.

Literature searching

Pubmed and Embase databases were searched for relevant records, using the terms ‘OH’, ‘postural hypotension’ in combination with ‘death’ and ‘mortality’. The search was limited to studies in humans without restriction of languages. We also analysed reference lists of original and review articles using a manual approach. The final literature search was performed on 25 December 2012.

Study selection

Studies were included for analysis if they met the following criteria: (1) published as full-length article in any language; (2) reported as cohort studies in humans (either of prospective or retrospective design, regardless of sample size and follow-up duration); (3) included adult population (≥18 years of age) at baseline; (4) OH was identified at baseline, and defined as a decline in systolic BP ≥20 mm Hg or a decline in diastolic BP ≥10 mm Hg within 3 min of standing from a supine position according to the international consensus criteria25; (5) documented mortality outcomes during follow-up (all-cause mortality, CV death or non-CV death), and reported the multivariable-adjusted risk ratio (RR) or HR, and their corresponding 95% CIs for mortality outcomes comparing individuals with OH at baseline with those without OH.

Data extraction and quality assessment

Two authors (WX and ZL) performed the literature search, data extraction and quality assessment independently according to the inclusion criteria. Discrepancies were resolved by consensus. The primary outcome of interest for the current meta-analysis is all-cause mortality, and the secondary outcomes include CV death and non-CV death. Extracted data include: (1) general information: year of publication and country where the study was conducted; (2) baseline characteristics of the study population: source of the population, numbers of the participants included, age, gender, mean body mass index (BMI), proportions of participants with comorbidities, such as hypertension, diabetes mellitus (DM), CAD and stroke; (3) definitions and prevalence of OH at baseline; (4) follow-up data: follow-up duration, numbers of cases for each outcome, and variables adjusted for analysis; (5) outcome data: association of OH and mortality risks reported as RRs or HRs and 95% CIs.

The quality of each study was assessed using the Newcastle-Ottawa Scale.26 This scale varies from 1 to 9 stars and judges each study on three broad categories: selection of the study groups; comparability of the groups; and ascertainment of the outcome of interest.

Statistical analysis and data synthesis

We used RR as a common measure of the associations between OH and the risk of mortality outcomes across studies. The HRs were directly considered as RRs. RRs and corresponding SEs, which were derived from 95% CIs or p values, were logarithmically transformed to stabilise variance and normalise the distribution.24 Interstudy heterogeneity was formally tested using Cochrane's Q test, and significant heterogeneity was considered to exist if the p value was <0.10. The I2 statistic, which describes the percentage of total variation across studies that is due to heterogeneity rather than chance,27 was also examined, and a value of I2>50% indicated significant heterogeneity. If Cochrane's Q test showed significant heterogeneity, a random effects model was used to calculate RR and its 95% CI. Otherwise, a fixed effects model was applied.

Meta-regression and predefined subgroup analysis were performed to explore the possible source of heterogeneity. Median values of continuous variables were used as cut-off values for grouping studies. Potential publication bias was assessed by funnel plots with the Egger regression asymmetry test.28 We also performed the non-parametric ‘trim-and-fill’ procedure24 to further assess the possible effect of publication bias in our meta-analysis. This method considers the possibility of hypothetical ‘missing’ studies that might exist, imputes their HRs, and recalculates a pooled HR that incorporates the hypothetical missing studies as though they actually existed. RevMan (V.5.1) and Stata (V.12.0) software were used for the meta-analysis and statistical analysis.

Results

Search results

The study selection process is shown in figure 1. Overall, the database search identified 458 citations from Pubmed and Embase, of which 10 published articles3 ,4 ,16–22 ,29 based on nine cohort studies, with 56 125 participants and 11 580 mortality cases, were finally included in the meta-analysis. Two of the papers were based on the same cohort study and reported all-cause mortality and CV mortality outcomes separately.4 ,29

Figure 1

Search and selection of studies included in the meta-analysis.

Study characteristics

The baseline and follow-up characteristics of included studies are shown in tables 1 and 2. Overall, eight of the included studies were of prospective design,3 ,4 ,16–21 ,29 while the other one was of a retrospective design.22 As for the source of the participants, five of the studies included community-based populations,3 ,4 ,16 ,18 ,19 ,29 while the other four studies were performed in specified patient groups, including patients with chronic renal failure,17 older patients from geriatric wards21 and fall clinics,22 and older residents of long-term healthcare facilities.20 The sample sizes of cohorts ranged from 179 to 32 797, and the mean ages of the included participants varied from 45.6 to 85.5 years. The baseline prevalence of OH varied from 5% to 54% among the included populations. The follow-up durations of the included cohort studies ranged from 2 to23 years. When presenting the associations between OH and mortality risks, we adjusted different groups of potentially confounding variables among the included studies as listed in table 2.

Table 1

Baseline characteristics of included cohort studies

Table 2

Outcomes and follow-up data of the cohort studies included in the meta-analysis

Quality assessment

The overall quality of studies included in the meta-analysis was good, with four studies3 ,4 ,18 ,21 ,29 scoring 9 stars on the Newcastle-Ottawa Scale, three studies16 ,19 ,20 scoring 8 stars, and the other two17 ,22 scoring 7 stars.

OH and the risk of all-cause mortality

All of the nine included cohort studies investigated the association between baseline presence of OH and the risk of all-cause mortality, among which, six studies3 ,4 ,16–19 showed a significant association between OH and increased mortality risk, while the other three20–22 did not. By combining the results of these studies with a random effects model, our meta-analysis indicated that participants with OH had significantly increased risk of all-cause death (RR=1.40, 95% CI 1.20 to 1.63, p<0.001; figure 2A), although the heterogeneity among these studies was significant (p<0.001, I2=75%).

Figure 2

Forest plots of pooled analyses for the associations between orthostatic hypotension and the risk of all-cause mortality (A), cardiovascular mortality (B) and non-cardiovascular mortality (C). IV, inverse of variance.

In view of the fact that significant heterogeneity was detected across the included cohort studies, we performed predefined meta-regression and subgroup analyses to explore the potential source of heterogeneity. Studies were defined as adequate adjustment for classic risk factors if the associations between OH and all-cause mortality were presented after adjustment of at least six of the following risk factors: age, sex, hypertension (or systolic BP), DM, total cholesterol (or low-density lipoprotein cholesterol), BMI and smoking status. Using this criterion, the results of our meta-regression analysis suggested that whether these risk factors were adequately adjusted seemed to be an important modifier of the association between OH and risk of all-cause mortality (regression coefficient = −0.34, 95% CI −0.71 to 0.03, p=0.07; table 3), which may partially explain the heterogeneity, whereas other study characteristics, such as sample size, general health status, mean age and gender of the participants, mean baseline BMI, proportion of participants with hypertension, DM, CAD and stroke, prevalence of OH, follow-up duration, events rate and study quality scores, seemed to have no obvious impact on the association between OH and all-cause mortality (table 3). Subsequent subgroup analyses produced similar results. The predictive value of OH for all-cause mortality was less strong in studies that adequately adjusted the classic risk factors (RR=1.24, 95% CI 1.05 to 1.48, p=0.01), compared with those in which these risk factors were not adequately adjusted (RR=1.74, 95% CI 1.42 to 2.12, p<0.001; p value for subgroup interaction, 0.01; table 4), suggesting that the prognostic effect of OH for all-cause death may partially depend on its effect on these classic risk factors. Moreover, subgroup analyses also found significant associations between OH and all-cause death in studies of prospective design, and in studies performed in community-based populations.

Table 3

Impact of study characteristics on the association between OH and the risk of all-cause mortality: univariate meta-regression analysis

Table 4

Impact of study characteristics on the association between OH and the risk of all-cause mortality: subgroup analyses

OH and the risk of CV mortality and non-CV mortality

Three of the included studies3 ,18 ,29 investigated the predictive value of OH for the risks of CV and non-CV mortality. By pooling the results of these three cohort studies, our meta-analysis failed to find a significant association between baseline presence of OH and increased risk of CV death (RR=1.20, 95% CI 0.72 to 2.00, p=0.47; figure 2B), although significant heterogeneity was detected (p<0.001, I2=92%). Of note, by pooling the same cohort studies, our meta-analysis indicated that participants with OH had an increased risk of non-CV death (RR=1.18, 95% CI 1.00 to 1.38, p=0.05; figure 2C), and no significant heterogeneity was found (p=0.20, I2=38%).

Publication bias

For the association between OH and all-cause mortality, the funnel plot seemed to be asymmetrical on visual inspection (figure 3), although the result of the Egger regression test seemed to suggest no significant publication bias (Egger test, p=0.119). To further analyse the influence of potential publication bias on the estimated result, we performed ‘trim-and-fill’ analysis, which conservatively imputes hypothetical negative unpublished studies in order to produce a symmetrical funnel plot. After incorporation of two hypothetical studies, the association between OH and all-cause mortality is somewhat attenuated, but remains significant (RR=1.31, 95% CI 1.13 to 1.52, p<0.01; figure 3). The publication bias for the associations between OH and CV or non-CV was difficult to estimate because only three studies were included.

Figure 3

Funnel plots with trim-and-fill analysis for the association between orthostatic hypotension and all-cause mortality. The white dots present the identified studies included in the meta-analysis; the black dots present the estimated missing studies after adjustment for publication bias. RR, risk ratio.

Discussion

In this study, by pooling all available cohort studies together, our meta-analysis indicated that subjects with OH were associated with 40% higher risk of future all-cause mortality events. Moreover, the predictive value of the baseline presence of OH for future all-cause mortality seemed to be less strong for studies in which classic risk factors were adequately adjusted compared with those in which these risk factors were not adequately adjusted. In addition, although our meta-analysis failed to reveal a significant association between OH and CV mortality, we did find that subjects with OH have an increased risk of non-CV mortality.

Our study confirmed the previous hypothesis that subjects with OH have an increased risk of all-cause mortality. Results of subgroup analyses suggested that the association between the baseline presence of OH and increased all-cause mortality risk remained significant when the study population was limited to a community-based population or those who were at lower risk of future mortality (≤ 40 per 1000 person-years). Moreover, the predictive value of OH for mortality events seemed not to be affected by prevalent comorbidities of the population, including hypertension, DM, CAD and stroke. Furthermore, the predictive value also seemed not be influenced by follow-up duration. Although these results seemed to be robust, at this stage, we cannot decide whether or not OH was causally related to all-cause mortality and whether or not OH was only a marker of a generally increased risk of mortality events. In addition, the mechanisms contributing to the association between OH and increased mortality risk are not fully understood. Several suggested hypotheses may be of value in explaining this association. First, evidence from cross-sectional studies and other epidemiological studies indicates that OH may be related to a few classic risk factors for CV diseases and mortality, such as aging, hypertension, DM, metabolic disorders and, possibly, higher BMI.1 ,2 Therefore, OH may confer an increased risk of mortality by interacting with these classic risk factors. Indeed, our meta-regression and subgroup analyses indicated that the predictive value of OH for all-cause mortality seemed to be less strong for the studies in which these classic risk factors were adequately adjusted compared with those in which they were not adequately controlled, suggesting that OH may lead to a higher mortality risk partially through the impact of these classic risk factors. Second, subjects with OH are likely to have increased BP variability related to body posture, and a large proportion of thoracic blood volume may be displaced to lower limbs due to gravity during orthostasis, which may lead to reduced filling of the right atrium and subsequently decreased cardiac output.30 Thus, intermittent ischaemia of important organs, such as heart, brain and kidney, may often occur during OH, and the related acute change in haemodynamic and organ perfusion status may lead to permanent damage of these organs. Third, baroreflex dysfunction is well recognised as one of the most important mechanisms of pathogenesis of OH, probably caused by baroreceptor impairment due to aging or atherosclerosis.31 ,32 However, impairment of the baroreflex as a marker of autonomic dysfunction has been related to an increased risk of mortality in patients with type 2 DM33 or chronic renal failure34 or after myocardial infarction35 ,36 perhaps through the induction of systematic inflammation.37 These facts suggest that the association between OH and increased mortality risk may at least partially be explained by potential autonomic dysfunction, and OH may be a marker of impaired autonomic CV regulation. Further studies are needed to explore the potential role of autonomic CV deficiency in the association of OH and increased mortality risk.

Owing to the small number of the studies and the considerable heterogeneity across the included studies, our study failed to show a significant association between OH and CV mortality. As previous cohort studies have shown that OH was associated with increased risk of incident CAD8 and stroke,9 we believe OH would be associated with an increased risk of CV mortality if more large-scale cohort studies with a long enough follow-up were included. Therefore, future studies are warranted. As for non-CV mortality, we did observe that OH increased its risk. This is not surprising in view of the fact that OH has been related to an increased incidence of several non-CV conditions, such as dementia12 ,13 and falling,14 which may contribute to injury-related death.

There are a few limitations in our study which should be considered when interpreting the results. First, the number of included studies for estimation of some outcomes is small, so the results for these outcomes (associations between OH and CV and non-CV death) should be interpreted with caution. Second, we did not have individual patient data, and therefore we were unable to perform subgroup analysis based on individual patient information. So, this subgroup analysis could not exclude the possibility that certain subgroups of participants with OH are at particularly high risk of all-cause mortality, and future studies are warranted to explore the potential source of heterogeneity. Third, our meta-analysis is based on observational studies. Hence, there may be residual confounding factors that were not adjusted for when the association between OH and mortality risk were estimated. However, as there seems to be no evidence-based effective treatment of OH,38 it is currently difficult to confirm our results in a large randomised trial. Nevertheless, our study also has numerous strengths, including the relatively high quality of the studies included (quality score ranging from 7 to 9), a large pooled sample size, and use of trim-and-fill analysis to handle potential publication bias.

In conclusion, our study indicates that the presence of OH is associated with a significantly increased risk of all-cause mortality, which may be partially mediated by classic risk factors. The association between OH and CV mortality was not significant, although we did find a significant association between OH and increased risk of non-CV mortality. Further studies are needed to explore the potential mechanisms underlying the association between OH and increased mortality risk and to determine whether the associations between OH and mortality risk are causal.

Acknowledgments

We thank Dr Marc Budge, Dr Avraham Weiss and Dr Alexandra Yalov for generously sharing valuable unpublished data from their studies. We also thank Ms Xuan Zhang for her excellent help in translating the paper published in Spanish.

References

View Abstract

Footnotes

  • Contributors All authors are in agreement with the content of the manuscript, are aware of the submission, and guarantee the work. WX and SM conceived and designed the experiments. WX and ZL performed the experiments. All authors analysed the data, and WX wrote the paper.

  • Competing interests None.

  • Provenance and peer review Not commissioned; externally peer reviewed.