The duration of the orthostatic blood pressure drop is predictive of death

Abstract

Background : Orthostatic hypotension (OH) affects 6% of community-dwelling older people. This increases to 60% when non-invasive, continuous blood pressure (BP) monitoring is used, due to identification of transient drops in BP which recover rapidly.

Aim : To determine the clinical relevance of these transient orthostatic BP drops.

Design : Five-year clinical observational study.

Methods : One hundred three consecutive new patients attending a Falls and Syncope Clinic in the UK from 1 February 2009 underwent continuous BP monitoring during an active stand. BP profiles were analysed to quantify all reductions in BP, measuring the duration of any drop below diagnostic criteria. Five-year follow-up data were extracted from hospital clinical records to assess clinical outcomes.

Results : Systolic BP (sBP) dropped ≥20 mmHg in 76 (74%) individuals, with 65 (63%) having ≥10 mmHg drop in diastolic BP. However, only 22 (21%) cases were diagnosed clinically with OH. A sustained reduction in BP (≥30 s) had a sensitivity of 0.91 and specificity of 0.88 for a clinical diagnosis of OH, being more accurate than absolute BP reduction alone. A sustained reduction in sBP was associated with greater use of vasopressors (36%, P 0.001) and an independent, significantly greater risk of death (45% at 5 years, P 0.009).

Conclusion : An orthostatic reduction in sBP lasting ≥30 s improves accuracy of diagnosis. Moreover, given the significant adverse outcomes with a sustained reduction, clinicians should consider this when diagnosing and treating patients, as transient OH does not appear to be clinically significant.

Introduction

The prevalence of orthostatic hypotension (OH) has previously been found to be as high as 6% in the community and 70% in long-term care. ^1,2 However, since the advent of non-invasive, continuous blood pressure (BP) monitoring, prevalence studies have found OH in as many as 60% of healthy, community-dwelling older people and 30% of young healthy people. ^3–5 This highlights one of the dilemmas of continuous BP monitoring. Both clinicians and academics using beat-to-beat BP monitoring are regularly faced with orthostatic BP profiles that demonstrate very brief drops in BP that recover spontaneously; the significance of these being unknown. ⁶ Previously, international consensus produced a diagnostic criterion, setting a diagnostic threshold of a systolic drop reaching 20 mmHg and diastolic drop of 10 mmHg upon standing. ⁷ While this consensus did much to standardize the diagnosis of OH, the advent of continuous BP measurement demonstrated that it is common to reach this diagnostic threshold, without having the sustained decline in BP typically seen in those with OH. ^3,8

In 2011, an update to the consensus criteria was produced. ² The most notable update was the addition of the word ‘sustained’ to the diagnostic criteria. However, no attempt was made to define what constitutes ‘sustained’. One of the difficulties in addressing the issue of transient orthostatic BP drops is that their clinical significance is unknown. A further addition was made to the 2011 update concerning initial OH. This is defined as a symptomatic, exaggerated and transient orthostatic drop in BP (40 mmHg systolic or 20 mmHg) occurring within the first 15 s of standing upright. ² Because it occurs rapidly it is unlikely to be identified without the use of continuous BP monitoring.

There have been no follow-up or observational studies to describe the natural history of transient OH or the longer term sequelae. ⁶ It is unknown whether it results in similar outcomes to ‘classical’ OH such as falls, syncope or an increase in mortality, or whether it is a precursor to a sustained OH.

It would be of practical use for clinicians and academics to understand whether the duration of the BP drop is of additional significance to the absolute BP drop, when diagnosing OH. Moreover, the clinical outcomes based on the duration of the BP drop would be of importance while making decisions regarding diagnosis and management. This study aims to address these two clinically important, unanswered questions.

Methods and Materials

Setting

The Newcastle upon Tyne Hospitals NHS Foundation Trust’s Falls and Syncope Service in the UK is a specialist clinical service for people who fall, have syncope or have dizziness, gait or balance abnormalities. The service covers a wide geographical area over much of Northern England, covering urban and rural communities.

Participants

All patients who attend the service undergo a comprehensive medical assessment, including an Active Stand to detect OH. The Active Stand is performed in accordance with international consensus methodology. ² This test is performed in all new patients attending the clinic, regardless of reason for referral. All new patients attending the service over a 6-week period (beginning 1 February 2009) who were aged over 18 years and who consented to allow their clinical data to be used were included.

Data

All consecutive new patient encounters were recorded and reviewed. Continuous, beat-to-beat BP was measured and recorded using digital photoplethysmography, with the CNSystems Taskforce Monitor. The beat-to-beat BP data, which were stored electronically, were exported into individual data files so each BP profile could be visually inspected to extract the degree of BP drop and the duration of such drops. Patients wore the equipment while resting in the supine position for 10 min and were then assisted into the standing position where they remained for 3 min. Five-year follow-up data were reviewed in March 2014 for the 103 cases who had had an active stand in the study period. This follow-up data were retrieved from the hospital’s electronic record, focussing on hospital admission due to falls/syncope/dizziness, requiring vasopressor treatment (fludrocortisone/midodrine/pyridostigmine/clonidine) and all cause mortality.

Diagnosis of OH

A diagnosis of OH is made by a specialist physician within the Falls and Syncope Service. This is based on the international consensus statement on the definition of OH (a drop of 20 mmHg systolic, or 10 mmHg diastolic within 3 min of standing, or a drop of 40 mmHg systolic or 20 mmHg diastolic within 15 s of standing). ² The use of continuous BP monitoring is associated with false positive drops in BP; ^3,5,6 the physician identifies a false positive as the presence of a drop in BP (as described above) in the absence of symptoms (fall, syncope, dizziness, weakness, cognitive disturbance, blurred vision, tremulousness, vertigo, pallor, anxiety, palpitations, clammy feeling and nausea ⁹ ).

Ethical permissions

All patients whose data are included in this study, provided written consent for their clinical data to be used for audit and research. The NHS Foundation Trust reviewed and approved the methods of gaining written consent and performing the data collection.

Data analysis

Normally distributed data are presented as mean with standard deviation, with non-parametric being summarized as the median with the range. Student’s t-test was used to compare parametric data between two groups and the Kolmogorov–Smirnov Z test was used in preference to the Mann–Whitney U when comparing two non-parametric data sets with numbers <25 per group. Relationships between two categorical variables were assessed by performing the chi-square test; on review of expected frequencies, Fisher’s exact test was not required. Where logistic regression was performed, assumptions which were verified were multicollinearity of the independent variables and linearity of the continuous independent variables with the natural log of the dependent variable. Independent variables violating these assumptions were excluded from the model. The goodness of fit of the model is described with Cox’s R² , the associated χ2 statistic and significance value. The regression coefficient ( B ), standard error the significance level ( P ), the odds ratio and the 95% confidence intervals are tabulated. A predefined significance level was set at 0.05.

Results

Of 114 eligible patient data files, one patient declined consent, four were repeat tests and six contained a poor unusable data signal. One hundred and three eligible continuous BP profiles were included in the analysis. The resulting cohort had a median age of 73 [range 22–94, interquartile range (IQR) 60–80], with a median of 3 comorbidities (range 0–10, IQR 1–5), with 43% taking vasoactive (including anti-hypertensive and anti-anginal) medication.

Diagnosis

An orthostatic drop in systolic BP (sBP) occurred, to some degree, in 100 of the 103 cases. Of those 100 cases with a systolic drop, 76 cases met the diagnostic threshold, decreasing by 20 mmHg or more on standing. However, despite meeting this threshold, only 22 were given a clinical diagnosis of OH (29% of those who crossed the diagnostic threshold); meaning 54 orthostatic BP drops were not considered as clinically significant. Similarly, 94 cases demonstrated a diastolic BP (dBP) drop on standing, with 65 reaching the diagnostic threshold. Table 1 compares the characteristics of those who were diagnosed with OH with those who reached the diagnosed threshold, but who were not considered to have OH. Those with a clinical diagnosis of OH had significantly greater number of comorbidities and had significantly greater drop in sBP and dBP, with the drops below diagnostic threshold lasting for significantly longer.

Table 1.

Demographic and BP characteristics of those with a clinical diagnosis of OH, and those who had BP drops that reached the diagnostic threshold but were not considered to be clinically significant (IQR, interquartile range)

	Orthostatic drop in systolic or dBP meeting diagnostic criteria		P
	Not diagnosed with OH (59)	Clinical diagnosis of OH (22)
Age	71 (range 23–94, IQR 56–81)	75 (range 34–88, IQR 73–78)	0.065
Comorbidities	2 (range 0–9, IQR 1–5)	4 (range 0–7, IQR 3–5)	0.024
Prescribed vasodepressor	24 (41%)	9 (41%)	1.0
Resting sBP	130 (105–229)	138 (106–198)	0.264
Resting dBP	81 (13)	83 (18)	0.459
Nadir sBP during standing	103 (23)	78 (21)	<0.001
Nadir dBP during standing	66 (14)	53 (14)	0.001
Systolic BP drop	30 (11–77)	55 (27–113)	<0.001
dBP drop	14 (1–39)	29 (11–66)	0.001
Time below systolic threshold	5 (0–150)	153 (6–180)	<0.001
Time below diastolic threshold	3 (0–180)	68 (0–180)	0.001

A regression analysis was performed to identify which BP parameters were independently associated with a clinical diagnosis of OH, whilst controlling for the number of comorbidities ( Table 2 ). All 103 cases were included in the model to capture all sizes of BP drops. The regression model demonstrates that it is the time below the systolic diagnostic threshold that is the most predictive parameter for a clinical diagnosis of OH. The systolic nadir is also significantly associated with the diagnosis, although less so, but perhaps more interestingly, the degree to which the sBP and dBP decreases during orthostasis is not independently associated with the diagnosis.

Table 2.

The regression model demonstrates which BP characteristics are associated with a clinical diagnosis of OH

	B	SE	OR	95% confidence interval		P
				Lower	Upper
Comorbidities	0.338	0.242	1.402	0.873	2.252	0.162
Systolic nadir	−0.059	0.029	0.943	0.891	0.997	0.039
Systolic drop from baseline	−0.037	0.048	0.964	0.878	1.058	0.441
Time below systolic threshold	0.093	0.027	1.097	1.04	1.157	0.001
Diastolic nadir	0.058	0.046	1.06	0.968	1.159	0.207
Diastolic drop from baseline	0.043	0.085	1.044	0.884	1.233	0.613
Time below diastolic threshold	−0.051	0.025	0.95	0.904	0.998	0.042
Constant	−2.716	3.39	0.066			0.423

Time below the diagnostic sBP threshold is independently associated with the clinical diagnosis, more so than the absolute drop in sBP itself (SE, standard error, OR, odds ratio). Cox R² 0.506, Homer and Lemeshow 4.938 (8), P 0.783.

With the results of this regression in mind, a receiver operating characteristic (ROC) with curve coordinates was created to compare the sensitivity and specificity of the BP drop compared with the duration of the BP drop below the diagnostic threshold. For sBP, a duration of 30 s resulted in the greatest sensitivity and specificity, while it was 10 s for the dBP. The area under the curve of the ROC is presented in Figure 1 alongside the specificity and sensitivity. The consensus criteria were used as diagnostic criteria (systolic drop >20 mmHg and dBP drop >10 mmHg) and are compared with the length of time below the diagnostic threshold derived from the ROC analysis. The figure demonstrates that by incorporating the duration of the BP drop the specificity increases greatly and sensitivity reduces slightly. However, using a diagnostic criterion of sBP drop >20 mmHg with a sustained duration of at least 30 s yields a sensitivity of 0.91, specificity of 0.88 and an area under the curve of 0.956.

Figure 1.

The sensitivity and specificity of each of the four criteria to diagnose OH. The size of the bubble represents the area under the curve of the receiver-operating characteristic, with the value displayed. A = dBP drop below diagnostic threshold lasting over 10 s, B = sBP drop below diagnostic threshold for over 30 s, C = any dBP drop below diagnostic threshold, D = any sBP drop below diagnostic threshold.

Clinical outcomes

The 5-year outcomes for a transient, a sustained or no drop in orthostatic BP are summarized in Figure 2 . A significantly higher proportion of those with a sustained drop in BP was on vasopressor medication at 5 years or had died ( P  < 0.001 for each outcome). No significant difference was noted in hospital admissions over the 5-year follow up.

Figure 2.

A comparison of the clinical outcomes of the three diagnostic categories (no significant BP drop, sBP drop lasting <30 s and a sustained sBP drop lasting over 30 s. The combined outcomes do not include symptoms, as these were recorded at time of Active Stand rather than during follow-up).

Mortality

Given that 45% of those with a sustained diagnostic drop in BP had died by 5 years follow up, a logistic regression model was created to control for age and comorbidity. The regression model, which is displayed in Table 3 demonstrates that a systolic drop in BP that lasts over 30 s below the 20 mmHg threshold, is independently associated with death, even when controlling for age and comorbidity ( Table 3 ). In contrast, a drop in sBP lasting <30 s, is not independently associated with death, nor is age or comorbidity.

Table 3.

The regression model, with death as the dependent variable

	B	SE	OR	95% confidence interval		P
				Lower	Upper
Constant	−6.301	2.064	0.002			0.002
Age	0.038	0.025	1.039	0.989	1.091	0.126
Comorbidities	0.105	0.136	1.11	0.85	1.451	0.442
Systolic BP group						0.005
Transient sBP drop	1.491	1.117	4.442	0.497	39.68	0.182
Sustained sBP drop	2.904	1.106	18.244	2.087	159.468	0.009

It demonstrates that sustained (≥30 s) drops in sBP, which remain below a 20 mmHg reduction in BP on standing, are independently and significantly associated with death at 5 years (where transient is considered as duration below 30 s SE, standard error, OR, odds ratio). Cox R² 0.188, Homer and Lemeshow 5.732 (8), P 0.677.

Discussion

Since the advent of non-invasive BP measurement in the 1970s the relevance of transient drops in BP upon standing upright have been uncertain. ⁶ This study is the first to report longitudinal, follow up data of these transient BP drops. The findings reported here demonstrate that these short-lived orthostatic drops in BP are common in a specialist falls and syncope clinic setting, and at 5 years follow-up, it is the duration of the BP drop which is associated with a clinical diagnosis of OH, with a duration of at least 30 s being associated with adverse clinical outcomes.

These findings will be of interest to clinicians and academics alike. While continuous BP measurement produces a high rate of false positives, it has not been known if these transient drops are false positives, a natural precursor to OH or a normal response to standing. Indeed in 2011, the consensus criteria were updated and the addition of the word ‘sustained’ was added to their diagnostic criteria to remove some of the uncertainty of these brief drops. ² However, no attempt was made to quantify what was meant by the term sustained. The results presented here suggest that a duration of at least 30 s for the sBP below 20 mmHg from baseline would be an appropriate length of time, excluding those with a transient drop in BP.

Incorporating the duration of the BP drop, with a cut-off value of 30 s is reasonable, given the high value sensitivity and specificity presented here. It may not be too surprising that the length of time below the diagnostic threshold is independently associated with a clinical diagnosis of OH, because time below threshold takes into account the consensus diagnostic threshold itself. However, incorporating a cut off value of 30 s greatly improves upon the specificity of the consensus cut-off values alone, while reducing the sensitivity only slightly.

One of the most important findings demonstrated in this 5-year follow-up study is the significantly higher prevalence of adverse outcomes in those with a sustained diagnostic BP drop. The hospital-derived follow-up data demonstrate that progression to pharmacotherapy, death and combined outcomes are all significantly more prevalent in those who have a sBP drop lasting over 30 s. This has important implications for clinicians who are uncertain about how to diagnose and treat people who have transient drops in their orthostatic BP. Indeed, on regression analysis, for diagnostic drop in BP lasting for >30 s there is an independent and significant increased risk of death, even while controlling for age and comorbidity. The lack of an association between age, comorbidity and mortality in the regression model may be explained by demographics of the cohort. The regression model is derived from all patients who attended the specialist unit. The cohort is relatively old and frail and by nature of a specialist clinic does not include the more straightforward cases. The result is that the association between age, comorbidity and mortality may have been diluted out of the model.

Clinical outcomes in this study were limited to hospital data only. This may skew the results, such that only the most serious outcomes are reported. For example, it is not known how common unreported falls were during the follow-up, nor do we know whether quality of life is affected by the duration of the orthostatic BP drop. Because the 5-year mortality data were derived from hospital records rather than death certificates and registers it does limit the overall impact of mortality data.

All patients who participated in this study were attending a secondary care Falls and Syncope Service and therefore may represent the more complex or severe cases than are found in the general population. This may explain why there were no significant differences in the 5-year hospital admission outcomes, as all cases, whether they suffered from OH or not, were at risk of falls or syncope. Ideally, the significant association with mortality would have been explored further with Cox regression analysis, but due to the cohort size, a time-based analysis was precluded. In order to confirm these results, it will be necessary to validate them in a prospective study, which would ideally be placed in a non-specialist setting and capture a wider range of outcomes.

It is expected that these results will have an increasing relevance in the context of the rapidly ageing population. Given that there remains a great deal of uncertainty and lack of evidence for the treatment of OH in older people, these results may help clinicians make decisions regarding treatments based on the duration of orthostatic BP drop. ^6,10,11 In fact, this study provides evidence which compliments the international consensus, using a duration of 30 s to quantify the recent addition of ‘sustained’ to the diagnostic criteria.

Acknowledgements

We are grateful to the staff in Newcastle Hospitals NHS Foundation Trust’s Falls and Syncope Service for assisting with data collection.

Conflict of interest : None declared.