Prevalence of postural hypotension in primary, community and institutional care: a systematic review and meta-analysis

Background Postural hypotension (PH), the reduction in blood pressure when rising from sitting or lying 0to standing, is a risk factor for falls, cognitive decline and mortality. However, it is not often tested for in primary care. PH prevalence varies according to definition, population, care setting and measurement method. The aim of this study was to determine the prevalence of PH across different care settings and disease subgroups. Methods Systematic review, meta-analyses and meta-regression. We searched Medline and Embase to October 2019 for studies based in primary, community or institutional care settings reporting PH prevalence. Data and study level demographics were extracted independently by two reviewers. Pooled estimates for mean PH prevalence were compared between care settings and disease subgroups using random effects meta-analyses. Predictors of PH were explored using meta-regression. Quality assessment was undertaken using an adapted Newcastle-Ottawa Scale. Results One thousand eight hundred sixteen studies were identified; 61 contributed to analyses. Pooled prevalences for PH using the consensus definition were 17% (95% CI, 14–20%; I2 = 99%) for 34 community cohorts, 19% (15–25%; I2 = 98%) for 23 primary care cohorts and 31% (15–50%; I2 = 0%) for 3 residential care or nursing homes cohorts (P = 0.16 between groups). By condition, prevalences were 20% (16–23%; I2 = 98%) with hypertension (20 cohorts), 21% (16–26%; I2 = 92%) with diabetes (4 cohorts), 25% (18–33%; I2 = 88%) with Parkinson’s disease (7 cohorts) and 29% (25–33%, I2 = 0%) with dementia (3 cohorts), compared to 14% (12–17%, I2 = 99%) without these conditions (P < 0.01 between groups). Multivariable meta-regression modelling identified increasing age and diabetes as predictors of PH (P < 0.01, P = 0.13, respectively; R2 = 36%). PH prevalence was not affected by blood pressure measurement device (P = 0.65) or sitting or supine resting position (P = 0.24), however, when the definition of PH did not fulfil the consensus description, but fell within its parameters, prevalence was underestimated (P = 0.01) irrespective of study quality (P = 0.04). Conclusions PH prevalence in populations relevant to primary care is substantial and the definition of PH used is important. Our findings emphasise the importance of considering checking for PH, particularly in vulnerable populations, to enable interventions to manage it. These data should contribute to future guidelines relevant to the detection and treatment of PH. PROSPERO:CRD42017075423.


Background
Postural, or orthostatic, hypotension (PH), is the fall in blood pressure (BP) when rising from seated or supine to standing [1]. It is associated with an increased risk of falls, cognitive decline, reduced quality of life and mortality [2][3][4][5].
Current National Institute for Health and Care Excellence (NICE) hypertension guidelines advise testing for PH in the presence of type 2 diabetes, postural symptoms or aged 80 or over [6]; European guidelines also suggest checking in older people and those with diabetes [7]. Whilst PH is routinely tested for in primary care when symptoms are reported, we have found that it is only considered one third of the time for older people and rarely with diabetes, in the absence of symptoms [8].
Since the majority of people with PH are asymptomatic, they are likely to go undetected under current practices, placing them at avoidable risk of sequelae [5,9].
In 2011, a consensus definition for PH: a sustained reduction in systolic BP ≥20 mmHg or diastolic BP ≥10 mmHg within 3 min of rising to a standing position, was proposed [1]. However, many other definitions of PH exist; reported prevalence estimates are likely dependent on the definition used, making this a source of variance and uncertainty around diagnosis of PH. Prevalence may also vary depending on the method of BP measurement, population and care setting under investigation. The prevalence of PH has been reported as ranging from 2 to 57% in community settings, primary care and institutional care cohorts [4,10,11]; increasing prevalences have been associated with older age, diabetes and hypertension [9,[12][13][14].
The large variation of reported prevalences may create uncertainty for clinicians as to who should be assessed for PH [15]. By describing the prevalences of PH in settings and conditions relevant to primary care, and identifying factors associated with greater prevalences, we aim to raise awareness of those patients most likely to have asymptomatic PH. Such evidence could counteract clinical inertia and facilitate rational choices, in the face of rising workload, as to when to invest time in testing for PH [16,17]. Increased recognition of PH would permit appropriate interventions, such as review of medications, to reduce risks of falls and other sequelae [18]. We undertook the following systematic review, metaanalyses and meta-regression to address these questions.

Literature searches
A systematic review was undertaken to determine the prevalence of postural hypotension across care settings. This study was prospectively registered with PROS-PERO: CRD42017075423. We searched Medline (including Medline in Process and Old Medline) and Embase from their respective commencement dates until 1st October 2019, using a broad search strategy based on key search terms (Appendix 1). Further studies were identified from the authors' archives and from reference lists of included studies and review articles. Study titles and abstracts were screened independently by two authors. Disagreements were discussed to reach consensus, with provision for adjudication by a third author, if needed. Two authors assessed and agreed full texts for inclusion, undertook data extraction and assessed study quality; the review process was managed using Covidence (Veritas Health Innovation, Melbourne, Australia).

Inclusion and exclusion criteria
Studies were eligible for inclusion if BP was measured in a lying or seated position followed by standing and using either a manual or automated sphygmomanometer. Eligible study settings were primary care, community or residential/nursing home populations. We identified 78 distinct definitions of PH in scoping studies for this review. To minimise heterogeneity of findings due to definitions, we restricted inclusion to studies which either reported using the consensus definition or adopted a definition encompassed within the consensus definition [1]. Exclusion criteria are summarised in Table 1.

Data extraction
Study level demographics were extracted for care setting, mean age, BP measurement device, resting position (seated or supine) and medical history of hypertension, diabetes, Parkinson's disease or dementia. Where a range of health status existed within a study population, if more than 50% of the total cohort included individuals with a particular condition, hypertension, for example, we applied the appropriate disease classification, i.e. the cohort would be classed as a hypertensive cohort. Populations were included within the community category, unless specifically selected from a primary care or institutional care setting. The Newcastle-Ottawa Scale (NOS), with questions adapted to PH specific context, was used to assess study quality (Appendix 2). Where multiple reports for a cohort were retrieved, extraction was primarily taken from the main publication, with addition of detail from subsidiary reports where needed.

Statistical analysis
Pooled estimates of mean prevalences for PH were calculated and compared between settings and populations using meta-analysis of proportions, undertaken in Stata v16 (Statacorp, Texas, USA) [19]. Random effects models were used throughout due to anticipated heterogeneity between included studies. Statistical heterogeneity was assessed using the I 2 statistic, and explored with sensitivity analyses, using meta-analysis, based on care setting, disease status or BP measurement method. We also conducted sensitivity analyses according to the definition of PH, i.e. whether PH was reported using the consensus definition or a definition that fell within the consensus definition parameters but did not fully meet them (e.g. by measuring BP over less than three minutes of standing). Univariable meta-regression analyses were undertaken to examine association between study level factors [mean age, percentage of females, mean absolute resting systolic BP, care setting, BP measurement method (auscultatory or oscillometric) or position (seated or supine), disease status (hypertension, diabetes, Parkinson's disease or dementia)] and prevalence of PH [20]. Factors suggesting univariable associations with PH (using P < 0.1) were entered into multivariable models, with a priori inclusion of age, care setting and presence of diabetes and hypertension. Publication bias was assessed visually using funnel plots and quantified with the Egger test [21].

Results
Searches identified 1816 unique citations; 356 full texts were reviewed; 92 studies met inclusion criteria, but only 61 fell within the consensus definition of PH, thus contributing to the meta-analyses. Reasons for the exclusion of studies are summarised in Fig. 1.

Description of studies
All included studies were cross-sectional or cohort studies, with cohort size ranging from 40 to 32,797 participants (Table 2). On quality assessment, areas of low quality (defined as falling below the median NOS total      score of 8; range: 3-10) were notable in categories relating to the response rates of participants and in comparability between respondents and non-respondents (usually due to lack of information provided), and the use of non-validated methods for BP measurement (Table 3.).
Univariable meta-regression showed three study level factors to be associated with mean prevalence of PH: age (P < 0.01), history of falls and disease status (all P < 0.05, see Table 4). For multivariable analysis, age (Fig. 7.) and presence of diabetes remained as predictors of PH (P < 0.01, P = 0.13, respectively; R 2 = 36%).

Summary
This is, to our knowledge, the first systematic review and meta-analysis to present estimates of PH prevalence in populations regularly encountered in primary care, including general practices and related outpatient clinics, care or nursing homes and community settings. Our findings confirm that PH, when tested for, is a common finding across care settings and disease subgroups, with the highest prevalences observed in people residing in care/nursing homes (and primary care, when low quality studies were omitted from analyses), and in those with dementia; age itself appears to be the key predictor of prevalence. The definition of PH used can impact prevalence estimates and therefore must be considered carefully in clinical practice. The type of BP measurement device and

Strengths and limitations
This study provides insight into PH prevalences across a variety of care settings and disease cohorts. Our search terms were intentionally broad, thus it is unlikely that substantial numbers of relevant publications were overlooked. Data extraction was limited to English language papers and published records, although non English language and Grey literature data generally have been shown to make limited impact on review findings where a substantial body of published evidence exists [81]. We found some evidence for publication bias against low prevalence small studies; overall, there was considerable heterogeneity of PH prevalence estimates across Fig. 3 Prevalence of postural hypotension according to the consensus definition across disease subgroups. 'Control' group represents those individuals with no co-morbidity different care settings, disease cohorts, PH definitions and BP measurement methods that was not accounted for in our sensitivity analyses. The utility of the NOS for assessing study quality has previously shown poor agreement between reviewers, with calls for more specific guidance in its use [82]. We adapted the generic guidance to give context specific to this PH review (Appendix 2), however, we did not find any substantial impact on heterogeneity in subgroup analyses according to quality assessment of studies. High residual levels of heterogeneity limit our ability to draw firm conclusions from the data. PH prevalence varied widely across studies (2.0-56.8%) and residual heterogeneity probably reflects cumulative effects of nonsystematic variations in population size and health status, limitations in classifying cohorts by condition at study level, and the discrepancy in PH definitions and measurement methods employed across studies.
Our univariable meta-regression showed that the presence of disease was associated with increasing PH prevalence, according to condition. This association did not persist when multivariable regression was undertaken, but there was co-linearity of disease status with diabetes, which was included a priori in the multivariable model. Increasing BP per se, a known risk factor for orthostatic hypotension [83], was not associated with increasing PH incidence; however, data for baseline BP were, surprisingly, only reported in 13 studies, limiting our ability to explore this association. The relationship of PH with hypertension is complex; PH is associated with both uncontrolled hypertension and the number of antihypertensive drugs used in managing high BP [25,38,84,85], but effective treatment of high BP in elderly persons is associated with reduced PH prevalence [41,86]. Consequently, a non-linear or 'U' shaped relationship of prevalence to absolute BP might be expected, with interaction in analyses between a diagnosis of hypertension  Current guidelines for postural hypotension management recommend clinicians undertake a comprehensive medication review if systolic BP falls by 20 mmHg on standing [6]. The de-escalation of antihypertensive medication is a common treatment method and may increase the probability of recovery from postural hypotension with no increased risk of adverse cardiovascular events [87], however, further work in this area is required.

Comparison with existing literature
This review builds on existing reviews that have summarised prevalence of PH in specific cohorts, such as those with diabetes or Parkinson's disease and individuals over 60 years of age [13,[88][89][90][91]. Here, we report that PH affects 18% of individuals across care settings and disease cohorts. Our data show that PH incidence rises from community care settings to those attending primary care and residing in institutions. These findings reflect the likelihood that multimorbidity, and the subsequent risk of PH, is more common in care/nursing home settings than general practices or in the community [92]. We also found that individuals with chronic disease have increased prevalence estimates of PH compared to groups without such diseases present. This may be due to a number of factors, including medication (e.g. diuretics, antihypertensives), development of peripheral and/or autonomic neuropathy (associated with diabetes mellitus and dementia) or physical deconditioning (due to age-related changes or continued bed rest) [4].
There appears to have been an exponential rise in interest in PH, with~70% of the studies reported in this review published in the last decade and~50% in the last 5 years. This may reflect interest in rising longevity, multimorbidity and rates of diabetes (risk markers for PH) [93,94]. Recent reporting of improved cardiovascular outcomes with intensive lowering of BP is also relevant [95,96], given the risks of adverse events such as PH and falls, associated with lower BP targets [97].
Our findings are consistent with studies that have reported high prevalences of PH in individuals with diabetes (type 1, 19% and type 2, 20%) [88], and in the aged [98]. Prevalences approaching 50% have been reported in Parkinson's disease with low prevalence of orthostatic symptoms, making the case for routine postural BP testing when reviewing all sufferers [99].
On subgroup analyses, we found no significant difference in PH prevalence when measuring BP in the sitting position rather than supine, prior to standing, and this finding remained on exclusion of low quality studies. This approach may therefore be justified as Fig. 7 Bubble plot of study level association between mean age and prevalence of postural hypotension according to the consensus definition (systolic ≥20 mmHg or diastolic ≥10 mmHg within three minutes of standing). Circles represent estimates from each study, sized according to precision of estimate an alternative to the gold-standard supine-to-stand approach, if undertaken with rigid methodology. Shaw et al. have previously suggested that the sit-to-stand method is a good alternative for busy clinicians when the supine-to-standing method cannot be achieved; they proposed reducing diagnostic thresholds for PH to a systolic drop ≥15 mmHg or a diastolic drop ≥7 mmHg to maximise the sensitivity and specificity of the test and to reflect the reduced orthostatic stress of moving from sitting to standing, compared with lying supine [100]. We found no evidence to support a change in diagnostic threshold in this review, but suggest future studies should directly compare supine versus seated followed by standing PH measurement methods. We also found that adopting auscultatory or oscillometric methods of measuring BP did not impact prevalence estimates. Further work is required across larger cohorts to determine the most appropriate diagnostic criteria for PH in primary care if the pragmatic sit-tostand method is to be adopted.
When the definition of PH did not fulfil the consensus description, but fell within its parameters, we found that prevalence was underestimated irrespective of study quality. This highlights the importance of adopting the consensus definition to minimise under-detection of PH whenever possible [1].

Implications for research and/or practice
Our univariable regression analyses confirmed that an increasing PH prevalence is strongly associated with increasing age, with age-related chronic diseases and with previous falls. Multivariable analyses revealed that increasing age and presence of diabetes were particularly associated with increased PH prevalence; such individuals may benefit from routine checking for postural hypotension. The population is aging [101], and people are living for longer periods in older age with levels of dependency, or in care settings [102]. European hypertension guidelines, recommend checking for PH in older people, and this will include greater numbers, with attendant workload pressures, over time [7,16]. By describing the commonly encountered disease states and care settings associated with higher than background prevalences of PH, we provide evidence to encourage improved recognition of this condition through targeted testing. Ideally, BP should ideally be measured from supine to standing using auscultatory methods and our results support the use of the consensus definition [1,80]. Pragmatically, however, the sit-to-stand method may also be employed as an alternative to the gold standard if the methods are rigorous [100]. However, further work comparing supine versus seated followed by standing measurement methods should be undertaken to clarify the most approach resting positions and thresholds for accurate PH diagnosis.

Conclusion
Overall, these findings demonstrate the substantial prevalence of PH across a range of populations and care settings relevant to primary care. Our prevalence findings suggest that checking for the presence of PH should be routinely considered when treating chronic conditions, such as diabetes, particularly in older persons. Failure to follow the consensus definition of PH appears to underestimate prevalence, therefore we advocate adoption of the consensus as a standard whenever checking for PH. Further work is needed to confirm the diagnostic thresholds for postural hypotension when BP is measured in the seated rather than supine position.

Appendix 1
Search strategy (Medline and Embase)  a) Comparability between respondents and nonrespondents characteristics is established, and the response rate is satisfactory. * b) The response rate is unsatisfactory, or the comparability between respondents and nonrespondents is unsatisfactory. c) No description of the response rate or the characteristics of the responders and the nonresponders. 4) Ascertainment of the exposure (i.e. measurement of sitting/lying and standing blood pressure): a) Validated measurement tool. ** b) Non-validated measurement tool, but the tool is available or described.* c) No description of the measurement tool.
Comparability: (Maximum 2 points) 1) The subjects in different outcome groups are comparable, based on the study design or analysis. Confounding factors are controlled. In this context it is controlling for co-variates.
Reported prevalence may be reported as unadjusted (usual) with logistic regressions or adjusted. Can indicate which in extraction. Go for unadjusted or adjusted?
a) The study controls for the most important factor (systolic BP or Age probably both important). * b) The study control for any additional factor (candidates include BMI, BP, age, gender use of antihypertensive medication). *

Outcome: (Maximum 3 stars)
The outcome for this review is the prevalence of postural hypotension: 1) Assessment of the outcome: a) Independent blind assessment. ** b) Record linkage. ** c) Self report. * d) No description.
These descriptors are unhelpful so suggest re-classify as: a) presented as n/N or proportion for each relevant group in results. ** b) Self report i.e. in some way reported by unblinded investigators. * c) No description 2) Statistical test: In this reviewthe calculation of proportion(s) with postural hypotension, so should be clearly derived from n/N without unexpected or unexplained omissions from numerator or denominator.
a) The statistical test used to analyze the data is clearly described and appropriate, and the measurement of the association is presented, including confidence intervals and the probability level (p value). * b) The statistical test is not appropriate, not described or incomplete.