Validation of blood pressure monitoring using pulse transit time in heart failure patients with Cheyne–Stokes respiration undergoing adaptive servoventilation therapy

Abstract

Purpose

Using pulse transit time (PTT) and an ECG appears to be a promising alternative for invasive or noninvasive monitoring of blood pressure (BP). This study assessed the validity of PTT for BP monitoring in clinical practice.

Methods

Twenty-nine patients with chronic heart failure (HF; 27 male, 70.5 ± 9.9 years) and nocturnal Cheyne–Stokes respiration were noninvasively ventilated for one hour using adaptive servoventilation (ASV) therapy (PaceWave™, ResMed). BP was measured using two devices (oscillometrically via Task Force Monitor™, CNSystems and PTT via SOMNOscreen™, Somnomedics) at least every 7 min for 30 min before, during, and after ASV.

Results

Mean systolic BP was 118.1 ± 14.4 mmHg vs. 115.9 ± 14.1 mmHg for oscillometric method vs PTT, respectively. Corresponding values for diastolic BP were 72.3 ± 10.3 mmHg and 69.4 ± 11.1 mmHg. While clinically comparable, differences between the two methods were statistically significant (p < 0.05). The difference between the two methods showed an increasing trend over time. A total of 18.5 % of PTT-based measurements could not be analyzed. The direction of a change in BP was opposite for PTT vs oscillometry for 17.0 % and 32.8 % of systolic and diastolic BP measurements, respectively.

Conclusions

When monitoring BP in HF patients, overall BP monitoring using PTT is comparable to oscillometry for a period of 2 h (including a 1-h ASV phase). However, PTT shows a tendency to underestimate BP over time and during ASV.

Introduction

Discovery of pulse wave velocity (PWV) goes back to the early twentieth century when Crighton Bramwell first found PWV to be dependent on arterial wall tension and blood pressure (BP), thus allowing indirect measurement of arterial wall elasticity [1, 2]. Since Schmitt and colleagues’ invention of the plethysmograph in the mid-twentieth century it was only a matter of time before it was suggested that PWV could be used to noninvasively monitor BP [3]. The concept evolved to where an ECG and plethsymograph could be used to calculate PWV [3]. PWV is inversely correlated with BP meaning that absolute values and changes in BP could be measured indirectly [4, 5]. However, the limitations of such methods have long been recognized, including the presence of many artifacts in PWV recordings [6].

More recently, positive results have been obtained using a device to automatically determine PWV [7]. That study, and others, specifically focused on using PWV to determine aortic stiffness as a surrogate marker for predicting cardiovascular mortality [7, 8]. Even currently and in the field of research in sleep apnea PWV is employed as a reliable tool to determine aortic stiffness [9, 10]. The invention of the cuff sphygmomanometer shifted the focus of BP monitoring to the peripheral, specifically the brachial artery [11, 12].

In current clinical practice there is high demand for simple, cost-effective and convenient BP-monitoring devices. This is particularly true in a sleep laboratory setting, where overnight oscillometric BP monitoring would be associated with arousals and sleep fragmentation. However, the ability to monitor BP in this setting is important, particularly during positive airway pressure therapy, which may have hypotensive effects, especially in chronic heart failure (HF) patients [13]. This importance becomes even clearer bearing in mind that obstructive sleep apnea (OSA) was shown to be highly prevalent in patients with HF being associated with inverse effects on cardiac function [14–16]. This study was designed to validate PTT-based BP measurement in patients with chronic HF over a 2-h monitoring period, including 1 h of adaptive servoventilation (ASV) intervention.

Methods

Subjects

From August to October 2012 a total of 29 consecutive patients with stable HF (left ventricular ejection fraction [LVEF] <50 %; New York Heart Association [NYHA] class ≥ II) and nocturnal Cheyne-Stokes respiration (CSR; apnoea–hypopnoea index [AHI] ≥15/h) treated with adaptive servoventilation (ASV) therapy were enrolled. Patients were excluded if they had a >10 mmHg variation in systolic or diastolic BP between measurements on their two arms.

Assuming a two-sided significance level of 5 (alpha) and 90 (beta) percent of power a sample size of 29 patients was calculated to be required to detect a difference of 5 ± 8 mmHg [17]. As documented in previous studies, BP determination by PTT is limited by missing values. To take this into account and to realize a standard minimum power of 5 % (alpha) and 80 % (beta), a cohort of at least 23 patients was calculated to be necessary [18, 19].

The study protocol was approved by the local ethics committee, all patients gave written informed consent to participate in the trial, and the study was carried out in accordance with the Declaration of Helsinki.

Study design

The aim was to validate a new continuous noninvasive BP-monitoring device (SOMNOscreen, Somnomedics, Germany), which is based on PTT measurements. The system uses a 3-channel ECG and pulse oximetry (photoplethysmography) to determine the start and arrival of the pulse wave (Fig. 1). It requires calibration with a single reference BP value, plus the inclusion of some basic patient data (gender, age, weight and height). A detailed description of the device can be found elsewhere [18, 19].

Fig. 1

Blood pressure measurements through pulse transit time (PTT; Somnoscreen®) and oscillometry (TFM, Task Force Monitor®). This figure illustrates how and where BP measurements were performed. PTT based BP monitoring required a three-channel electrocardiogram (2) to detect R waves as the start of a pulse wave and an oxygen probe on the digital finger (1) to detect the arrival of the pulse wave. On the contralateral arm, an upper arm BP was measured by oscillometry (OM) (3). Continuous OMs were performed based on a finger-press-system on the second and the third finger (4)

BP values obtained using the new device were compared with oscillometric measurements. Standard measurements were obtained using a standard BP cuff on the upper part of one arm and a BP-recording device attached to one finger on the contralateral arm (Task Force Monitor, CNSystems, Graz); the finger used for measurements was alternated every 20 min (Fig. 1) [20].

After resting in a supine position for at least 30 min, BP was monitored without interruption using the two different devices for a total period of 2 h, which included one hour of ASV intervention. Oscillometric BP was determined at least every 7 min on the upper arm, and finger BP and BP determined by PTT were recorded continuously. Patients rested quietly in a dark room for the first 30 min of BP assessment. Over that time PTT-derived BP monitoring was set up to note the exact time of the first three oscillometric BP measurements. Patients then received one hour of noninvasive ventilation with ASV (PaceWace™, ResMed, Australia). Ventilator pressure levels were set individually, with minimum end-expiratory positive airway pressure (EPAP) and minimum inspiratory positive airway pressure (IPAP) support levels set at the same as those used by the patient the previous night. After ASV there was another 30 min of BP monitoring.

DOMINO software (Version 2.5.0, Somnomedics, Germany) was used to display measurements of PTT, and the 3 markers set up during the first 30 min were used to enable PTT recordings to be converted to BP measurements at the precise time that oscillometric measurements were taken. Measurements were excluded when there was no PTT signal available at this time or if the conversion process resulted in more than 50 % artifacts. Firstly, all available oscillometric measurements (upper arm BP) were compared with the corresponding PTT-derived BP values available at exactly the same time (marker). Secondly, mean BP values for continuously measured BP on one finger were compared with mean BP values measured via PTT over the corresponding period of time.

Statistics

All data are expressed as mean ± standard deviation (SD). A paired t test was used to compare differences between absolute BP using PTT versus oscillometric method. Mann–Whitney Rank Sum test was used if data did not have a normal distribution. Bland Altman method of comparison was used to compare BP levels obtained using the two different methods [21]. Patients were divided into subgroups based on the overall difference in BP measurements between the two methods. These data were then analyzed using ANOVA. However, if data did not show normality or equal variance then Kruskal–Wallis One-Way Analysis of Variance on Ranks or Friedman Repeated Measures Analysis of Variance on Ranks was used. In case of categorical data, chi-square testing was used. A p value < 0.05 was considered to be statistically significant.

Results

Demographics

Demographic and clinical data for all patients are summarized in Table 1.

Table 1 Patient demographic and clinical characteristics at baseline

Blood pressure measurements

In four out of 29 patients (13.8 %) artifacts in PTT measurements did not allow the determination of PTT-derived BP measurements at any of the three time points. Data on BP measurements for the remaining 25 patients provided comparisons of 658 points of measurements by the upper arm BP cuff with 536 corresponding PTT-derived BP values. Overall, lost PTT-derived BP values accounted for 18.5 % of the total, with systolic and diastolic BP equally affected. PTT-derived measurements failed in 13.7 % of baseline measurements (first 30 min), in 21.9 % of measurements during the 1 hour of ASV intervention and in 15.6 % within the final 30 min. For further analysis of differences between the two methods missing values were disregarded and calculations were based upon “pairs of BP values,” namely 536 simultaneously measured BP values through PTT and oscillometry.

Systolic oscillometry-based BP for the entire 2 h was 118.1 ± 14.4 mmHg versus 115.9 ± 14.1 mmHg for PTT-derived systolic BP (p < 0.001). Corresponding values for diastolic BP measurements were 72.3 ± 10.3 mmHg and 69.4 ± 11.1 mmHg (p < 0.001). Overall mean differences between the two methods were 2.2 ± 9.1 mmHg for diastolic BP and 2.9 ± 8.1 mmHg for systolic BP, which were clinically comparable despite the statistically significant difference (Fig. 2).

Fig. 2

Differences in systolic (a) and diastolic (b) blood pressure (BP) measurements (OM oscillometric measurement, PTT pulse transit time. Bland Altman plots for OM- and PTT-based BP [mmHg] measurements over the entire study period of 2 h: averaged systolic values (n = 536; a), diastolic values (n = 536; b), respectively, in contrast to the difference of both methods (PTT-OM). The lower/upper line correspond to the mean difference ± 1.96 SD (=Limits of Agreement)

Differences between oscillometry- and PTT-based measurements were neither clinically nor statistically significant different within the first 30 min and without any intervention. Systolic BP was 115.5 ± 14.8 mmHg (oscillometry) versus 116.3 ± 13.8 mmHg (PTT; p = 0.138), and corresponding diastolic BP was 69.9 ± 8.9 mmHg versus 70.4 ± 9.9 mmHg (p = 0.125). During 60 min of ASV intervention, oscillometry versus PTT-based systolic BP was 117.8 ± 14.8 mmHg versus 114.9 ± 14.2 mmHg, and diastolic BP was 72.8 ± 10.8 mmHg versus 68.6 ± 11.7 mmHg (p < 0.001). These differences between groups persisted during the last 30 min of monitoring when oscillometry and PTT-based systolic BP values were 122.1 ± 12.3 mmHg and 116.8 ± 14.5 mmHg, respectively (p < 0.001; Fig. 3) and diastolic BP values were 74.6 ± 10.4 mmHg versus 69.7 ± 11.5 mmHg (p < 0.001), respectively (Fig. 4).

Fig. 3

Differences in systolic blood pressure (BP) before (a), during (b), and after (c) ASV treatment (OM oscillometric measurement, PTT pulse transit time). Bland Altman plots: averaged values (n = 168 before, 242 during, 126 after ASV) of BP values through OM (upper arm BP cuff) and PTT in contrast to the difference of both methods (PTT-OM). Difference in systolic BP and for 30 min before (a), after (c), and during 60 min of ASV intervention (b); for differences in diastolic BP respectively (see Fig. 4a–c ). The lower/upper line correspond to the mean difference ± 1.96 SD (=Limits of Agreement)

Fig. 4

Differences in diastolic blood pressure (BP) before (a), during (b), and after (c) ASV treatment (OM oscillometric measurement, PTT pulse transit time)

The difference between BP values obtained using the two methods increased over time (Figs. 3, 4, and 5). The mean difference was +0.8 ± 6.7 mmHg (systolic) and +0.6 ± 7.0 mmHg (diastolic) for the first 30 min, −2.9 ± 10.0 mmHg and −4.2 ± 7.8 mmHg during ASV and −5.3 ± 8.9 mmHg and −5.0 ± 8.6 mmHg over the last 30 min. This suggests that ASV intervention and/or time from calibration are associated with an increasing difference between the two methods, with a tendency for BP to be underestimated using the PTT method compared with oscillometry.

Fig. 5

Differences in systolic blood pressure and diastolic (BP) before, during and after (c) ASV treatment (OM oscillometric measurement, PTT pulse transit time). This figure summarizes the differences in systolic (filled circle)/diastolic BP (square) for the distinct periods before, during and after ASV and between PTT based BP measurements and OMs. The data is expressed as mean ± standard deviation

In 24 of 25 patients (in one patient the finger BP cuff did not work adequately) continuously measured (and periodically recalibrated finger cuff BP measurements) were compared with continuously-measured PTT-derived BP. Although it was technically impossible to compare the two methods on a beat-to-beat basis, we were able to calculate each patient’s individual difference between both methods for the resting 30-min baseline period, one hour of the ASV intervention and the final 30 min.

Overall, there was a difference of −2.5 ± 10.5 mmHg for systolic BP and +0.7 ± 6.3 mmHg for diastolic BP, respectively. Over the different time periods, systolic values were +1.6 ± 8.3 mmHg for baseline, −3.4 ± 3.4 mmHg for ASV and −5.7 ± 12.0 mmHg for the 30-min period after ventilation. Corresponding values for diastolic BP were +2.7 ± 5.3 mmHg, −0.2 ± 6.8 mmHg, and −0.5 ± 6.2 mmHg. In 21 of 24 patients, the tendency of PTT-derived values to underestimate BP coincided with initiation of ASV (Fig. 6).

Fig. 6

Differences in systolic (a) and diastolic (b) blood pressure (BP) measured using oscillometry versus pulse transit time during versus before adaptive servoventilation (ASV). This figure shows the difference in systolic (a)/diastolic (b) BP between the two methods before (left box plot) versus during ASV (right box plot). These box plots include the mean of the bias between continuous OM BP and PTT based BP measurement before and during ASV of 24 patients. p < 0.05 for comparison of boxplots in both a and b. The boundary of the box closest to zero indicates the 25th percentile, a line within the box shows the median, and the boundary of the box farthest from zero indicates the 75th percentile. Whiskers (error bars) above and below the box indicate the 90th and 10th percentile

Given the small number of patients studied, multivariate analysis of all potential confounding factors did not seem reasonable. Therefore, the following data are intended to be hypothesis-generating only, and should be interpreted with caution. Patients were divided into six (1–6) different groups according to the overall difference between both BP-measurement methods; a seventh (7) group included patients in whom PTT did not allow conversion into PTT-derived BP values (Table 2). By analyzing demographic data we found differences among subgroups of patients with respect to mean LVEF, baseline systolic and diastolic BP, prevalence of renal insufficiency and diabetes, result of a 6-min walk test, prevalence of a pacemaker ECG/frequent ventricular extrasystoles (VES) on ECG and intake of diuretics. Patients with poor signal quality (group 7) had lower baseline systolic BP compared with all other groups (except group 5), the lowest LVEF of all groups and more patients presenting with left- or right-bundle branch block (LBBB or RBBB) on the ECG. The prevalence of diabetes in groups 2 and 3 (systolic BP measured >5 mmHg too high/too low with PTT) was significantly higher than in group 1 (acceptable difference for systolic values; p = 0.04 for 1 vs. 3). Mean 6MWT was by far the worst in group 3 compared to group 1 (p = 0.02) and group 2 (p = 0.002).

Table 2 Comparison between different patient groups according to the overall difference between both BP measured by pulse transit time versus oscilliometric methods

Patients in groups 5 and 6 (diastolic BP measured >5 mmHg too high/too low with PTT) were characterized by lower systolic and diastolic BP, a higher prevalence of renal insufficiency and more patients with VES compared to group 4 (acceptable difference in diastolic values).

IPAP levels differed significantly between patients in groups 5 and 6 (p = 0.038). Prevalence of a pacemaker ECG was highest in groups 2 and 5 (overestimating systolic/diastolic BP >5 mmHg). Finally there were more patients taking diuretics in groups 2, 3, 5, and 6 (unsatisfactory difference in systolic/diastolic BP) compared to groups 1 and 4 (satisfactory difference in systolic/diastolic BP).

Detecting changes in BP

To determine the sensitivity of the PTT-based approach for assessing BP changes, the number of changes in oscillatory measurements (upper arm BP cuff) of ≥5 mmHg for both systolic and diastolic BP were first identified. Next, the number of times that PTT-based BP measurement successfully detected that change as the right direction, change of more/less than 5 mmHg or even as an change in the opposite during the corresponding periods (Table 3). For detection of increases and decreases in systolic BP of >5 mmHg it appears that sensitivity was worst during ASV intervention. In contrast, sensitivity to detect increases in BP of >5 mmHg in diastolic BP was best during ASV intervention. Detection of decreases in diastolic BP of >5 mmHg was even best during the last 30 min of assessment. Overall sensitivity to detect both increases and decreases in BP was better for changes in systolic, rather than diastolic, BP.

Table 3 Sensitivity of PTT based BP monitoring for detection of changes in BP

Discussion

This study shows that PTT-based BP monitoring provides clinically acceptable BP readings, but with a tendency to underestimate BP over time and during ASV.

Noninvasive continuous BP monitoring is useful in a variety of applications in cardiovascular, intensive care and sleep medicine. Increasingly these disciplines are starting to overlap. For example, sleep and cardiovascular medicine are closely linked in HF patients with sleep-disordered breathing (SDB). In patients with HF, treatment with positive airway pressure (PAP) ventilation can cause unexpected falls in BP [13]. PTT-based BP measurement devices would appear to have practical applications in a sleep laboratory setting because they do not disturb the patient or cause arousals from sleep [22, 23]. With more and more HF patients being treated with a PAP device, the reliability of PTT-derived BP measurements is of increasing importance.

In a previous study, we were able to demonstrate that PTT-based BP measurements had good reliability compared with invasive BP determination over a monitoring period of 60 min and without any additional interventions [19]. These results are similar to those of the current study in that no statistically significant differences between the oscillatory and PTT methods were detected during the first 30 min of monitoring prior to the initiation of ASV. Also in agreement with previous data, the current results showed that application of PAP therapy using an ASV device was associated with a statistically (but not clinically) significant difference in BP measurements obtained by our two devices [18]. There was better agreement between monitoring methods for systolic BP than for diastolic BP, which is a finding that has been reported previously [24]. This observed increase in inaccuracy for monitoring of diastolic BP particularly attracted our attention since diastolic BP is generally known to be closely linked with coronary artery disease and stroke [25]. In detail, MacMahon and colleagues found a decrease of 5mmHG in diastolic BP to be associated with at least 34 % less stroke and 21 % less coronary artery disease [25].

For incoming devices European Union standards for BP measurements accept a deviation limit of ≤5 ± 8 mmHg for systolic and diastolic pressures. For validation of BP measuring devices European Society of Hypertension demands a precision of 5 mmHg to standard auscultatory measurements to characterize new devices measurements as “very accurate”; devices are described as “slightly inaccurate” for pressure deviations of 6 to 10 mmHg [17, 26]. In our study PTT-based BP measurements struggled to meet these standards during ASV intervention and as the duration of monitoring increased. As has been documented, PTT-derived BP measuring tended to be less effective and accurate in patients with right- or left-bundle branch block, a lower LVEF and a lower systolic BP at baseline [19].

It is also possible that the presence of diabetes and the patient’s overall status (including renal failure, VES and diabetes) could play a role in the differences in BP determined via the two methods. Some previous studies have also demonstrated that this may be the case. Woolam and colleagues were the first to document the influence of diabetes on PWV [27]. In the current study groups 2 and 3, in whom PTT-measured BP was >5 mmHg above or below the oscillometry reading, had a higher incidence of pre-existing diabetes compared with groups who had less difference in BP readings. In addition, it has been reported that the relationship between PTT and BP is impaired in patients with chronic HF compared with healthy subjects [28]. This provides an explanation for the current results, where those in group 5 had the most impaired LVEF and those in group 6 had the worst 6MWD. Another co-morbidity that might affect reliability of PTT-based BP measurements is renal failure. Patients with renal failure and/or receiving diuretics were overrepresented in groups 2, 3, 5, and 6, in whom PTT-BP often was read as too high or too low. Schmalgemeier confirmed this finding in 72 CPAP-ventilated patients, showing that the prevalence of renal insufficiency was one of the best predictors of the difference in systolic values; in addition CPAP itself influenced the difference between the two methods [18]. In the present study ASV intervention led to more artifacts in PTT recordings and, therefore, BP determination. In addition, antihypertensive drugs, which were being used by the majority of patients, might delay PTT itself [29].

A new observation in this study is that pacemakers and VES appear to influence the difference in BP measurements between the oscillatory and PTT-based methods. This makes sense, given that calculation of BP is heavily dependent on the correct detection of R waves in ECG recordings, but requires further investigation.

Based upon the above mentioned our result that overall difference between the two devices is statistically different (118.1 ± 14.4 mmHg versus 115.9 ± 14.1 mmHg (p < 0.001) for systolic and 72.3 ± 10.3 mmHg and 69.4 ± 11.1 mmHg (p < 0.001) for diastolic values) needs to be interpreted cautiously. As shown such difference obviously shows an inter- and intra-individual dimension meaning that there are certainly patients (inter-individual) where PTT based BP differs only slightly from OM especially with few time from calibration (intra-individual).

Paradoxically, the difference between finger cuff and PTT-based measurements was less for diastolic BP than for systolic BP. A possible explanation for this is that higher PTT-based diastolic BP values at baseline prevented too much of a difference between diastolic BP measurements—due to PTT based BP values underestimating BP with time—at the end of the recording period, resulting in an apparent improvement in sensitivity (Fig. 6, Table 3). Two other studies have noted low sensitivity for detecting changes in mean arterial BP compared with invasive monitoring [30, 31]. Based on the available information, it would appear that the overall capability to detect changes in diastolic BP is poorer than that to detect changes in systolic BP. This applies to both increases and decreases in BP. In this context, periodical recalibration of PTT-based BP measurements might lead to improved accuracy [32, 33].

Limitations

We were not able to validate PTT-based BP recordings against invasively-measured values because the study was conducted in a sleep laboratory, but the validity of oscillometry has been widely confirmed [20, 34, 35]. Another limitation is the small sample size, which did not allow appropriate multivariate analyses. However, based on the initial data obtained, future larger studies can be planned.

Overall, PTT-based BP measurements are clinically acceptable over a period of two hours, including a one hour ASV intervention. However, in patients with chronic HF, PTT has a tendency to underestimate BP over time and during ASV intervention. Therefore, the system might need recalibration if it is used for longer periods. In addition, PTT-derived values for systolic BP appear to be more robust and reliable than those for diastolic BP. Further studies, including more and better characterized patients, are warranted to appropriately identify factors contributing to the accuracy of PTT-based BP measurements.