Measuring blood pressure from Korotkoff sounds as the brachial cuff inflates on average provides higher values than when the cuff deflates

Objectives. In this study, we test the hypothesis that if, as demonstrated in a previous study, brachial arteries exhibit hysteresis as the occluding cuff is deflated and fail to open until cuff pressure (CP) is well below true intra-arterial blood pressure (IAPB), estimating systolic (SBP) and diastolic blood pressure (DBP) from the presence of Korotkoff sounds (KS) as CP increases may eliminate these errors and give more accurate estimates of SBP and DBP relative to IABP readings. Approach. In 62 subjects of varying ages (45.1 ± 19.8, range 20.6–75.8 years), including 44 men (45.3 ± 19.4, range 20.6–75.8 years) and 18 women (44.4 ± 21.4, range 20.9–75.3 years), we sequentially recorded SBP and DBP both during cuff inflation and cuff deflation using KS. Results. There was a significant (p < 0.0001) increase in SBP from 122.8 ± 13.2 to 127.6 ± 13.0 mmHg and a significant (p = 0.0001) increase in DBP from 70.0 ± 9.0 to 77.5 ± 9.7 mmHg. Of the 62 subjects, 51 showed a positive increase in SBP (0–14 mmHg) and 11 subjects showed a reduction (−0.3 to −7 mmHg). The average differences for SBP and DBP estimates derived as the cuff inflates and those derived as the cuff deflates were 4.8 ± 4.6 mmHg and 2.5 ± 4.6 mmHg, not dissimilar to the differences reported between IABP and non-invasive blood pressure measurements. Although we could not develop multiparameter linear or non-linear models to explain this phenomenon we have clearly demonstrated through ANOVA tests that both body mass index (BMI) and pulse wave velocity are implicated, supporting the hypothesis that the phenomenon is associated with age, higher BMI and stiffer arteries. Significance. The implications of this study are that brachial sphygmomanometry carried out during cuff inflation could be more accurate than measurements carried out as the cuff deflates. Further research is required to validate these results with IAPB measurements.


Introduction
Brachial cuff sphygmomanometry involves the placement of an appropriately sized cuff on the upper arm which is inflated until the underlying brachial artery is fully occluded and blood ceases to flow (Korotkoff 1905).As the cuff pressure (CP) is slowly reduced, conventional wisdom states that the CP at which the first Korotkoff sounds (KS) are heard by a trained operator using a sensitive stethoscope corresponds closely to systolic blood pressure (SBP) and the termination of the KS aligns closely to diastolic blood pressure (DBP).This is the basis for the calibration of all non-invasive blood pressure (NIBP) monitoring devices according to the recently released universal standard (ISO 81060-2:2018) (Stergiou et al 2018) for the validation of NIBP monitors.This standard states that 'A device will be considered acceptable if its estimated probability of a tolerable error (10 mmHg) is 85%'.Given that public health consequences (Psaty et al 1997) of even errors as small ±5 mmHg can be significant, the relatively large errors of 10 mmHg considered acceptable suggest some inherent problems with the method of classical sphygmomanometry.A related explanation may be that these broad limits reflect the long known and well documented (Kallioinen et al 2017, Picone et al 2017, Dankel et al 2019, Seidlerová et al 2019) reality that NIBP measurements generally underestimate SBP and overestimate DBP when compared to simultaneous intra-arterial measurement.In the study conducted by Kallioinen et al (2017), significant effects of individual sources of error ranging from −23.6 to +33 mmHg for SBP and −14 to +23 mmHg for DBP were recorded.
KS are fundamental to the calibration of all NIBP monitors.Blank et al (1988) used wideband (0.1-2000 Hz) external pulse recording during cuff deflation to identify three distinct phases of the KS (K1, K2, K3).K1 is a lowamplitude, low frequency signal <20 Hz, that is present with cuff pressures above SBP, K2 is a triphasic signal appearing at SBP and disappearing at DBP, which approximately corresponds to the audible KS and has frequency spectra in the range of 20-80 Hz and K3 appears with CP between SBP and DBP and continues to be present below DBP.Using Millar catheters for greater accuracy, Blanks also noted that he onset and disappearance of K2 was closely correlated to SBP and DBP derived from auscultation, with mean values of SBP generally higher than SBP and lower than DBP derived from auscultation.Blank et al (1988) showed also that the visual technique detects the onset (disappearance) of K2 a few beats before (after) the Korotkoff sound becomes audible (inaudible) at SBP (DBP) and as a result gives closer values to intra-arterial BP determinations than the conventional auscultatory technique.Furthermore, numerous studies including ours (Celler et al 2017) have demonstrated that there are significant inter-operator differences in estimating BP using sphygmomanometry, especially with the determination of DBP.
Our previous studies (Celler et al 2015(Celler et al , 2017) ) suggest that the accuracy of auscultatory sphygmomanometry is dependent on (i) the sensitivity of the stethoscope, (ii) the hearing acuity of the operator, and (iii) the amplitude and particular waveform morphometry of the KS.It was also shown that complete silence occurs on average in less than 50% of cases, making the determination of DBP particularly unreliable when KS are listened to with stethoscopes.Other studies in the literature of automatic or semi-automatic auscultatory NIBP estimation methods (Wu et al 2016, Chu et al 2017, Alpert 2018, Zhang et al 2019) also showed that visual auscultatory technique is significantly correlated with the manual auscultation.
In a recently published paper (Celler et al 2021) reporting on invasive experiments where the CP, the KS and the intra-arterial blood pressure (IABP) were simultaneously recorded, the authors demonstrated conclusively that the K2 KS did indeed correlate accurately to the very first blood flow from the occluded artery and ceased when the CP was approximately at DBP.The most important result of this study however was the observation that as the CP was reduced, first blood flow and hence K2 KS, were in many cases delayed until well below true SBP, up to 24 mmHg.We concluded that following occlusion of the brachial artery, there was a variable delay in the re-opening of the artery probably associates with low arterial compliance.We noted in one subject with particularly low BP, and in subjects following the infusion of glyceryl trinitrate (GTN) (100-200 mg), that the brachial artery could indeed also re-open prematurely.These data confirm and to large measure explain the individual subject to subject variability in NIBP measurements reported in many studies (Picone et al 2017, Saherwala et al 2018).
If indeed there is a delay in the reopening of the brachial artery following occlusion by an inflated cuff a logical hypothesis follows, that if the cuff is inflated whilst recording K2 KS, the first sounds heard should be associated with DBP and the very last sounds heard should be heard just before occlusion of the artery as the CP exceeds SBP, thus obviating the delayed re-opening observed when the CP falls post occlusion of the artery.

Aims and objectives
The aims of this study are to investigate whether SBP and DBP estimates based on the presence of K2 KS are different when the CP is linearly deflated as per conventional methods, to when the CP is linearly inflated.
Based on our hypothesis we would expect that on average the SBP on inflation will be higher than SBP measured on deflation and that DBP should be relatively similar.We will further explore whether non-invasive estimates of arterial compliance can explain some of the expected variability in SBP estimation.

Methods
We have developed a battery powered instrument to achieve in the non-invasive setting, a similar study to that reported in Celler et al (2021).The instrument was designed to record CP, high fidelity broadband (<1-500 Hz) KS using a piezo-resistive transducer, a single lead electrocardiogram (ECG) and a photoplethysmogram (PPG) signal.The data acquisition and the servo-control of the CP was achieved using a National Instruments USB6002 multifunction I/O module connected to a battery powered laptop computer via a USB cable.We developed a proportional-integral (PI) controller using LabView to control the Oken Seiko air pump (P54A02R) and a release valve to permit the CP to be increased rapidly 30-40 mmHg past estimated SBP, and then to fall linearly at approximately 3.0 mmHg s −1 to approximately 20 mmHg, before beginning a linear inflation to the same peak CP.

Experimental protocol and data acquisition
In traditional brachial sphygmomanometry the CP is inflated to approximately 30 mmHg above the estimated SBP to ensure complete occlusion of both the brachial artery and vein.As the CP is reduced, blood begins to flow below the cuff and a reactive hyperemia (RH) is observed as an overshoot in the SBP.RH arises to ensure that enough oxygen flows to the occluded area and that dead cells and metabolites are rapidly flushed from the area to reduce possible tissue damage.In figure 1 we note that RH is completed by the time the CP is reduced to below DBP.
Blood flow restriction (BFR) via the application of external pressure to occlude venous return and restrict arterial inflow, has been shown to increase muscular size and strength when combined with low-load resistance exercise (Mouser et al 2018).Following BFR blood flow decreased in a nonlinear, stepped fashion.Blood flow decreased at 10% of occlusion and remained constant until decreasing again at 40%, where it remained until 90% of occlusion.From roughly 30% to 60% arterial occlusion pressure (AOP), tissue pressure is increased underneath the cuff but remains insufficient to have a direct impact on occluding the arteries of the arm.Finally, from roughly 60% to 100% AOP, the pressure applied by the cuffs is impacting arterial flow directly leading at 100% to the complete occlusion of the brachial artery.
Superficial venous pressure is low in the arms (Ochsner et al 1951, Todini et al 2012) typically 10-15 mmHg, and the pressures applied in this study at 10% AOP ranged from 12 to 25 mmHg.This would be sufficient in most cases to block all superficial venous return.As pressure increases, the deep veins of the arm would also become occluded, as pressure transfer from the cuff to the tissue is to a small degree attenuated with increased tissue depth (Graham et al 1993).Even though this attenuation only amounts to about 20 mmHg, that appears sufficient to require increased pressure in order to completely occlude all venous return from the deep veins.
The brachial veins are deep veins which share the same name of the arteries they accompany.Pressure within the named veins is usually between 8 and 10 mmHg.Veins have thinner walls and larger diameters than arteries with less muscle and elastic tissue.This means that they have high vascular compliance so that the rate of change in volume with changing pressure is high and, therefore, changes in venous blood volume produce relatively small changes in venous distending pressure.In fact, veins have a compliance that is 30 times that of arteries and can expand easily to accommodate large volumes of blood.
These data clearly suggest blood flow in the deep veins continues to flow until cuff pressures increase from roughly 60%-100% AOP.Venous engorgement is therefore unlikely to be a significant factor given the short duration (∼20 s) of complete arterial occlusion and more than 45 s of only partial or no occlusion as shown in figure 1.An alternative method of ensuring rapid venous drainage would be to elevate the subject's arm for a few seconds, but this procedure interfered with the operation of the servo-control system and was therefore not adopted.

Data acquisition
Following informed consent and the recording of basic demographics including age, gender, height, weight, arm circumference and the distance between the mid sternum and the tip of the index finger, subjects were required to relax comfortably seated with both feet on the ground for a minimum of five minutes.A brachial cuff of appropriate dimensions for the patient's arm diameter was placed on the right upper arm and a sensitive piezoelectric transducer with flat frequency response from <1-500 Hz, was placed at the border of the cuff over the brachial artery to record the KS.ECG leads were placed at the LA, LL and RL position for a Lead III configuration.A Nellcor Compatible SpO2 Sensor (DS-100A) was placed on the right index finger.The cuff was rapidly inflated to 30-40 mmHg above the patient's SBP, and then deflated linearly at a rate of ∼3 mmHg s −1 to approximately 20 mmHg.The cuff was then linearly inflated at the same rate under servo-control to the selected peak CP.During this time, CP, KS, PPG and ECG signals were recorded at 1000 samples s −1 .This procedure was repeated three times, with a rest interval of five minutes between recordings.
It was later observed that a physical filter which was added after the differential pressure transducer to reduce pump noise during inflation was causing a pressure drop during rapid inflation or deflation.However, because of the low rate of inflation the bias introduced was measured as being very low, in the order of a few mmHg.Nonetheless a digital filter was designed to compensate for this pressure drop and was applied to all data recorded.
The monitoring instrument and the research protocol were approved by the University of New South Wales Human Research Ethics Committee (HC200066) on the 27th July 2020.This research was conducted in accordance with the principles embodied in the declaration of Helsinki and in accordance with local statutory requirements.All participants gave written informed consent to participate in the study.

Data analysis
All data processing and analysis were carried out using MATLAB (2020b).The broadband Korotkoff signal was high pass filtered at 20 Hz according to the K2 algorithm proposed by Blank et al (1988) to accentuate the high frequency (HF) components of the KS (HF KS) that are in the audible range and are related to blood flow in the artery.To improve the signal to noise ratio and facilitate the automated detection of peaks, the root mean square (rms) energy of the K2 Korotkoff signal was calculated and a moving average zero phase digital filter using a Hamming window of 80 ms width, was applied by processing the input data in both the forward and reverse directions.
Peak detection of the Korotkoff energy was carried out using the findpeaks Matlab command with a threshold of detection set at 10% of the peak energy signal.A similar command was used to detect the QRS peaks of the ECG and the peaks of the PPG signal.From the R peak of the ECG, the S peak was identified using gradient methods.The PPG data were filtered using the bandpass command (0.5-10 Hz), which performs zero phase filtering on the input using a bandpass filter with a stopband attenuation of 60 dB.The foot of the PPG was determined as the maximum value of the second derivative of the PPG wave.
Pulse transit time (PTT) is the sum of the pre-ejection period (PEP) and the vascular transit time (VTT) and is often incorrectly used to calculate the pulse wave velocity (PWV).In our study the brachial PWV study was calculated according to the method of Kortekaas et al (2018) where the VTT was estimated as the interval from the S wave of the ECG to the foot of the PPG signal.The S wave closely coincides with the opening of the aortic valve and the ejection of blood as evidenced using Doppler mode echocardiography.Dividing the distance from mid-sternum to the tip of the index finger in meters, by the VTT in seconds provides an estimate of PWV in meters/sec.The PWV was calculated and averaged from between three and five consecutive cardiac beats.

Determination of SBP and DBP
The determination of SBP and DBP points both during cuff deflation and cuff deflation were semi-automated, by detecting all peaks of the Korotkoff energy that were larger than 10% of the highest energy recorded.Noise and artefact could on occasion generate energy signals greater than the 10% threshold, but these could be easily ignored if not part of a consistent ensemble of signals one cardiac period apart.The SBP and DBP points were then manually selected using some simple rules.These include no adjacent peaks within two adjacent cardiac beats and participation in an ensemble of signals separated by one cardiac period, generally increasing and then decreasing in amplitude.In due course this process can be fully automated using deep learning algorithms

Statistical analysis
All data were tested for normality using the single sample Kolmogorov-Smirnov goodness of fit hypothesis test and the Lilliefors' composite goodness-of-fit test.Continuous variables such as % changes were tested with a one sample t-test with a null hypothesis of 'mean is zero'.For data that was not normally distributed the Wilcoxon rank sum test for equal medians was used.Before and after analysis of the same variables was carried out using the paired t-test.One-way analysis of variance (ANOVA) was carried out using the anova1 command when comparing the means of more than two groups of data.Multi-way (n-way) ANOVA (anovan) was used to determine whether particular categorical variables could explain the variance in an output variable.

Results
All subjects were volunteers from staff and students at the University of New South Wales or family and friends of the authors.Sixty two subjects with an average age of 45.1 ± 19.8 years (range 20.6-75.8years) were tested, including 44 men (45.3 ± 19.4, range 20.6-75.8years) and 18 women (44.4 ± 21.4, range 20.9-75.3years).Fifty-eight subjects had three recordings and four had only two recordings.No recordings were rejected for any reason other than technical problems such as excessive noise or artefacts.Each recording was analysed individually.However as there were no significant differences between recordings (p ?0.05), they were averaged and analysed as a single record. Figure 2 is an example of a single experimental record and table 1 summarises all significant variables for all subjects, male subjects, and female subjects respectively.All data were tested for normality.In 62 subjects the mean difference between SBP recorded as CP decreases, and SBP as CP increases was 4.8 ± 46 mmHg, of which fifty one were positive (range 0-14 mmHg) and eleven were negative (range −0.3 to 70 mmHg).These data effectively confirm the core hypothesis motivating this study.Of those who were negative four were in the age bracket <30, two were between 30 and 40 years, and five were older than 60.
It is instructive to look at figure 2 in some detail.The slope of the CP traces down and up is servo-controlled at approximately 3.5 mmHg s −1 and were not significantly different (p = 0.5977) across all subjects.The first Korotkoff energy pulse detected as the cuff deflates is always significantly (p < 0.0001) larger as a fraction of the peak (0.22 ± 0.08), than for the rising CP phase (0.19 ± 0.12).This suggests that as the cuff deflates the breakthrough pulse of blood flow is more vigorous than for the rising phase where the occlusion of the artery is progressively more complete and blood pressure is pumping against an increasingly higher pressure as the brachial circulation is increasingly occluded by the increasing CP.
PWV (p = 0.0029) also increase significantly between these two groups.PWV increases significantly across all five sub-categories.These results are summarised as box plots in figure 3 and show that subjects aged >50 and with a BMI >25 consistently have SBP diff that are positive.The younger cohort and those with low BMI, in contrast do on occasion demonstrate negative values of SBP diff.For SBP, the Bland-Altman plot shown in figure 4, gives a mean difference of 4.8 mmHg, an r 2 value of 0.8823 and a sum of squares error(SSE) of 6.7 mmHg.
For DBP, Bland-Altman plots (not shown) give a mean difference of 2.5 mmHg, an r 2 value of 0.7788 and a sum of squares error (SSE) of 5.3 mmHg.

Multiparameter linear and non-linear modelling
Multiparameter linear regression (MLR) is used to model the linear relationship between the explanatory (independent) variables and response (dependent) variable.In this study we used the fitlm function in MATLAB to explore which independent variables could predict the SBP diff.Attempts were made to develop MLR models to describe SBP diff as a function of numerous variables including BMI, PWV and heart rate (HR) but none gave p-values <0.05 and acceptable root mean squared error.Unsupervised learning analysis based on k-means, kmedoids (Arora et al 2016) and Gaussian mixture (Maugis et al 2009) clustering was similarly unsuccessful.3.2.Three-way analysis of variance using anovan Three-way ANOVA (anovan) is used to determine if there is an interaction effect between three independent variables (age, PWV and BMI) on a continuous dependent variable SBP diff.That is, we test if a three-way interaction exists.We use all the available data (N = 62) and segment each independent variable into two factors, being < or a selected break point.We then test the interaction for a number of breakpoints.Table 2 shows the results for age, BMI and PWV with breakpoints set at 50 years of age, a BMI of 25 and mean PWV (7.86 m s −1 ) respectively.
The low p-values for BMICat25 * PWVCatMean suggest that SBP diff, the dependent variable is significantly influenced or interacts primarily with the only two independent variables BMI and PWV at the selected breakpoints.

Discussion
In this study we propose a novel new method for brachial sphygmomanometry whereby we record the K2 KS as the brachial cuff is inflated linearly under servo-control.An extensive literature review revealed only a small number of publications where consideration was given to using either the rising or falling phase of CP inflation or deflation for estimating NIBP.Alpert (2007) undertook a clinical evaluation of the Welch Allyn SureBP algorithm for automated blood pressure measurement.Nukita et al (2020) reported that the repeated estimation of BP on the inflation phase can significantly reduce the risk of subcutaneous hemorrhage.Yamashita and Irikoma (2018) conclude that NIBP detection on cuff inflation detected hypotension faster than conventional NIBP without compromising the reliability of measurement, thus leading to early treatment of maternal hypotension and the prevention of adverse events related to the mother and the fetus.The efficacy of a new BP monitor (Takahashi et al 2020), based on detection of BP during inflation was evaluated in a randomised controlled study.
However, in all the quoted studies (Alpert 2007, Yamashita and Irikoma 2018, Nukita et al 2020, Takahashi et al 2020), the algorithm for detecting SBP and DBP is based on the oscillometric method and the rationale for choosing the rising phase of CP was in large part, to shorten the time required to take a measurement, to reduce the maximal inflation pressure and improve patient comfort and outcomes.None discuss the fundamental questions raised in this study.A technical paper (Vazquez et al 2021) describes a sensor fused BP measuring device capable of recording KS in inflationary curves, with the authors noting that 'The pump noise makes it difficult to get a reliable automatic auscultatory reading'.Zheng et al (2012) and Zheng et al (2013) derived the oscillometric waveform envelope (OWE) (Forouzanfar et al 2015) for both the rising phase of CP and then the falling phase.They found that during the rising phase the OWE was shifted upwards and resulted in a higher estimation of mean arterial pressure (MAP).These results are indirectly consistent with the conclusions of this study.
It is well recognized that BP is variable over time and changes subject to emotional and environmental factors (Adams and Leverland 1985, Soueidan et al 2010, Soueidan et al 2012).Intrinsic physiological oscillations in BP can also cause shifts of up to 20 mmHg within a few heartbeats (Hansen and Staber 2006).In this study we took great care to maintain all environmental variables comfortable and constant and followed the recommendations of the international standard body (Stergiou et al 2018) by waiting for five minutes between sequential experiments.Although differences between the three measurements were not significant, we did observe in some cases significant changes in SBP and DBP between the first and the last measurement, as the subject relaxed and became less mentally stressed.We minimize this effect by averaging the results of the three experiments, but the issue of maintaining a constant state of mental alertness during the recording of NIBP warrants further consideration.

Conclusion
In this study we conclusively demonstrate that SBP estimated from K2 KS and brachial cuff sphygmomanometry as the CP is increased is generally higher by −7 to 14 mmHg than SBP estimated when CP is decreasing.In a previous study (Celler et al 2021) where we simultaneously recorded both invasive and NIBP in sedated subjects, we discovered that following occlusion of the brachial artery at CPs higher than SBP, as the CP is reduced, the artery fails to re-open until CP is well below SBP.Typical errors were found to be from −4 to 24 mmHg, not dissimilar to the range (−7 to 14 mmHg) observed in this study, and in broad agreement with the conclusion of 62 separate studies (Dankel et al 2019) that indirect measures of SBP underestimated true SBP by an average of 4.55 mmHg and overestimated DBP by 6.20 mmHg (95% CI = 5.09-7.31).Although we could not develop multiparameter linear or non-linear models to explain this phenomenon we have clearly demonstrated through ANOVA that both BMI and PWV are implicated, supporting the hypothesis that the phenomenon is associated with age, higher BMI and stiffer arteries, as evidenced by higher PWV.
Whilst further work needs to be done duplicating this study in an invasive experiment with sequential deflation and then inflation of the cuff, the data presented provide a strong suggestion that by recording K2 KS as the CP is inflated may allow more accurate estimates of true intra-arterial SBP independent of age, gender, BMI or arterial compliance.
The implications of this study are potentially quite profound as current international standards (ISO 81060-2:2018) (Stergiou et al 2018) for the calibration of all NIBP monitor require two expert operators to listen to the KS as the CP deflates.Clearly this new method for brachial cuff sphygmomanometry would require a radical amendment to the international standards and a change in the operating modality of every NIBP monitor in the market.

Figure 1 .
Figure 1.Typical recording of IABP (blue trace) and cuff pressure (black trace), during a complete rapid inflation and slow deflation cycle.Red lines mark SBP (134 mmHg) and dotted red lines mark DBP (67 mmHg).The blue dotted line marks the beginning of blood flow through the brachial artery as evidenced by an initial increase in IABP.

Figure 2 .
Figure 2. Example of one experimental record showing results when CP is first reduced linearly and is then increased to the same peak.Peak detection of Korotkoff energy is shown by black circles.The Horizontal dotted line shows the 10% threshold below which local peaks are ignored.SBP and DBP points on both falling and rising phases of CP are identified by vertical lines and dotted lines respectively.For this subject BP recorded was 105/67 mmHg as cuff was deflated and 124/70 mmHg as cuff was inflated, a difference in SBP of 19 mmHg.

Figure 3 .
Figure 3. Boxplot of eight categories based on gender, age and BMI levels.

Figure 4 .
Figure 4. Bland-Altman plots showing differences between SBP recorded as the cuff inflates (SBP UP) and as the cuff deflates (SBP DOWN).

Table 2 .
Three-way analysis of variance with age, BMI and PWV as independent categorical variables.