Comparing blood pressure measurements between a photoplethysmography-based and a standard cuff-based manometry device

Repeated blood pressure (BP) measurements allow better control of hypertension. Current measurements rely on cuff-based devices. The aim of the present study was to compare BP measurements using a novel cuff-less photoplethysmography-based device to a standard sphygmomanometer device. Males and females were recruited from within the general population who arrived at a public BP screening station. One to two measurements were taken from each using a sphygmomanometer-based and the photoplethysmography-based devices. Devices were considered equal if the mean difference between paired measurements was below 5 mmHg and the Standard Deviation (SD) was no greater than 8 mmHg. Agreement and reliability analyses were also performed. 1057 subjects were included in the study analysis. There were no adverse events during the study. The mean (± SD) difference between paired measurements for all subjects was -0.1 ± 3.6 mmHg for the systolic and 0.0 ± 3.5 mmHg for the diastolic readings. We found 96.31% agreement in identifying hypertension and an Interclass Correlation Coefficient of 0.99 and 0.97 for systolic and diastolic measurements, respectively. The photoplethysmography-based device was found similar to the gold-standard sphygmomanometer-based device with high agreement and reliability levels. The device might enable a reliable, more convenient method for repeated BP monitoring.


Results
Patient characteristics. Out of 1480 subjects recruited, 1057 subjects were included in the final analysis (466 males (44.1%) and 591 females (55.9%)), for which the defined data set was available ( Table 1). The age range for recruited subjects was 8-100 years. 97 subjects (43 males (44.3%) and 54 females (55.7%)) reported a prior diagnosis of hypertension, 88 of them were under anti-hypertensive treatment. In 491 subjects a second BP measurement was taken after a short period of physical activity (5 min of strenuous walking). No adverse events were reported during the study.
Similarity between the measurements of the devices. Table 2 presents the mean and SD measurements of the reference and the BB-613WP devices. The mean ± SD Systolic Blood Pressure (SBP) of the reference device was 117.1 ± 19.5 mmHg and with the BB-613WP 117.2 ± 19.9 mmHg (p < 0.001), the mean ± SD Diastolic Blood Pressure (DBP) of the reference device was 68.2 ± 18.7 mmHg and with the BB-613WP 68.2 ± 11.5 mmHg (p < 0.001). Reliability of the BB-613WP device with regards to the reference sphygmomanometer device was excellent for both systolic and diastolic measurements and in both the first and second measurements, as indicated by an ICC of 0.96 and above ( Table 3).
Evaluation of the delta between the measurements in both devices. The mean deltas for the first (rest) measurement between the reference and the BB-613WP device for systolic and diastolic readings were -0.1 ± 3.6 mmHg and 0 ± 3.5, respectively. The delta for the second (exercise) diastolic measurement was statistically significant but the mean deltas for this second measurement were well below the 5 ± 8 mmHg threshold dictated by the ISO 81060-2:2013 ( Table 4). The percentage of systolic measurements with a delta of less than 5, 5-7, 7-10 or above 10 mmHg between the reference and the BB-613WP devices for the first (rest) measurement were: 95.3%, 0.9%, 3.6% and 0.2%, respectively. Similar distribution was demonstrated for the second (exercise) measurement (Fig. 2). Agreement between the devices. High agreement between the BB-613WP and the reference devices was demonstrated ( www.nature.com/scientificreports/  Figure 2. Percentage of measurements with a delta of less than 5, 5-7, 7-10 or above 10 mmHg between the reference device and the BB-613WP systolic measurements. Sys1 first systolic measurement. Sys2 second systolic measurement. Table 5. Percent agreement between the reference device and the BB-613WP in identifying high and normal blood pressure. Hypertension was defined as a systolic value of > 140 mmHg or a diastolic value of > 90 mmHg. Kappa is given as a value (lower-upper limit). www.nature.com/scientificreports/ ICC showed high reliability for both the systolic and diastolic measurements in subjects with or without reported hypertension. The average difference and the SD between the reference cuff-based device and the BB-613WP in subjects with and without diagnosis of hypertension were within the required limits, according to the ISO 81060-2:2013 (see supplemental file).

Discussion
The 2017 guidelines published by The American College of Cardiology/American Heart Association (ACC/ AHA) encourage changes in the management of hypertension. They recommend a holistic approach including the use of team-based care, routine use of home and ambulatory blood pressure monitoring (out-of-office BP measurements), and they stress the importance of consistent, accurate, and standardized measurements of BP [13][14][15] . The change in the guideline's definitions of hypertension resulted in an abrupt surge in the proportion of hypertensive adults in the United States 17,18 . Moreover, the new target of treatment for BP is accordingly lower. Known discrepancies between office and home BP values, defined as white-coat hypertension and masked hypertension, together with the narrower BP goals now recommended, require closer BP monitoring and a need to frequent the use of out-of-office BP measurements. Recent estimates are that more than 800 million adults worldwide have an SBP of 140 mmHg or higher. Hypertension's association with stroke, cardiovascular disease, heart failure, and chronic kidney disease places it with other leading preventable causes of death in the United States, and its consequences are projected to increase 14,15,19 .
Expanding the use of home blood pressure measurements is important yet challenging. The quality of measurements among hypertensive patients varies, and teaching people to perform good-quality and reliable blood pressure measurements necessitates dedicated resources and is challenging, particularly in demanding clinical  www.nature.com/scientificreports/ settings 18 . Ambulatory Blood Pressure (ABP) measurement is now acknowledged as the preferred method for predicting the risk of cardiovascular events relative to an individual's BP level 20 . Following these perceptions, questions have been raised on the accuracy, and consequently the role, of manual BP measurement in routine clinical practice 14,15,21 (see also https ://www.nice.org.uk/guida nce/cg127 ). Routine office BP measurements (both oscillometric and manual) are not only more susceptible to a "white coat effect", but are also less accurate, correlate quite poorly with ABP, and are associated with approximating readings to the nearest zero value [22][23][24] .
Reducing human involvement in the measurement of BP in the office, especially if considering that proper measurement techniques are rarely followed, is suggested to improve the quality of readings 20,23 .
The above-mentioned insights and findings emphasize the importance of having a device that allows better practice and compliance by users by being non-invasive, cuffless, and wireless, enabling multiple measurements without disrupting the user's daily routine.
The subject population in this study was heterogenous and represented both genders. 26.7% had background chronic diseases (Table 1) including medical conditions and characteristics that increase the risk of cardiovascular diseases, and 16% were taking medications on a chronic basis. 8.6% were smokers. This subject population represents the general Israeli population that visits the local hypertension clinics. 9.1% of the included subjects reported being pre-diagnosed with hypertension, with an average age of 66 years and with an average BMI of normal-high (mildly overweight).
A first blood pressure measurement was recorded in 1057 subjects. The average difference between the cuffbased reference device and the BB-613WP was 0.1 mmHg (SD of 3.6) for systolic BP measurements and 0 mmHg (SD of 3.5) for diastolic BP measurements, i.e. statistically insignificant and below the threshold levels defined by different regulatory protocols such as the ISO 81060-2:2013.
The Intraclass Correlation Coefficient (ICC) presents high reliability of the BB-613WP device with a 0.992 value for systolic BP and a 0.975 value for diastolic BP.
In the 491 subjects for whom a second measurement was taken, the average systolic BP in the reference device increased from 117.1 to 135.0 mmHg, without any change in the average diastolic BP. The BB-613WP device detected these changes with high reliability-ICC = 0.989 for systolic BP and 0.957 for diastolic BP. This means that the novel PPG-based device can detect and follow changes in BP along time and not only when the BP is stable and without any changes. Consistently, the average changes and the SD are well under the required by the ISO 81060-2:2013.
The agreement percentage between the reference device and the BB-613WP in identifying high and normal blood pressure was high (94.85%, κ = 0.881 and 97.19%, κ = 0.783, respectively). In 15.2% of the non-hypertensive subjects the reference cuff-based device showed a high BP value, while the BB-613WP device showed similar high BP values in 92.8% of these subjects. Bland-Altman plots demonstrate that the agreement between the BB-613WP and the reference device is within the limits of the ISO 81060-2:2013 definitions, suggesting good/ acceptable agreement in all subjects. A subgroup analysis of subjects with or without reported hypertension demonstrated similar results.
In conclusion, when comparing BP values measured by the standard cuff-based device to the novel BB-613WP device in subjects with both normal and high BP levels, we show that the mean value of the differences in systolic and diastolic measurements for all subjects is less than 5.0 mmHg and the SD is less than 8.0 mmHg, as required by different regulatory protocols such as the ISO 81060-2:2013. This indicates that the PPG-based BB-613WP device can be used as an accurate and reliable BP measurement device.
Beyond its accuracy and reliability, a PPG-based device is simpler to place and use than the cuff-based device, reducing the involvement needed by health personnel. These findings may change the way these parameters will be monitored in the near future, both in pre-hospital and in-hospital settings, as well as in home care scenarios over prolonged periods of time. Further studies should be undertaken to strengthen this data. www.nature.com/scientificreports/ This was a short-term study with only one or two measurements collected and being compared between the two devices, minutes after the reference baseline value (calibration) took place in the BB-613WP device. As such, a prolonged protocol with more measurements per subject is needed to substantiate our findings. That said, already in this setting we were able to show the accuracy and the usability of such a wearable, wireless and noninvasive device. We are now in the process of concluding several other clinical studies addressing this limitation.
Based on the ISO 81060-2:2013, measurements of BP should be taken by two investigators, with the reference value being the average of these measurements. This was not performed in the study as we could not interfere and change the way the designated hypertension screening stations operate. This also is being addressed in several other clinical studies using the BB-613WP.
When EMS personnel take measurements with the cuff-based device, they round the numbers to the nearest 5 mmHg value. This has an influence on the comparison and is seen in the Bland-Altman plots. The research team decided not to change this pattern, with the intent to prevent any influence and allow the EMS personnel to continue with their tasks as usual.
When using any sort of physical exercise to increase the BP, it will decay back to normal rapidly upon cessation of the exercise, therefore methods for standardization should take no more than few seconds of measurement. The research team did not want to change and interfere too much with the routine measurement of BP. Thus, despite this challenging environment, subjects that performed the 5 min exercise had the BB-613WP on them during the session, and upon completion of the exercise session BP measurement using the cuff-based device was immediately taken and in parallel the value in the BB-613WP device was recorded, as detailed in the methods.

Methods
This comparative prospective study was approved by the institutional review board (IRB) of the Tel-Aviv Medical Center, Tel-Aviv, Israel (0032-15-TLV).

Study population.
Male and female volunteers were recruited with no age restrictions. Each came for a BP checkup provided by the Israeli Emergency Medical Services (EMS) in designated BP measurement public screening stations deployed across the country. Each signed an informed consent form as defined by the Institutional Review Board (IRB). Informed consent was provided by the parents of subjects under the age of 18 years. The research team received coded numbers of the participants without any personal identifiers. Recruitment of subjects lasted for a period of two months, aiming to recruit more than the minimal sample size defined by the ISO 81060-2:2013 (85 subjects and 255 valid paired measurements).
The BB-613WP device. The BB-613WP is a wrist-worn or skin attached device indicated for use in measuring and displaying functional oxygen saturation of arterial hemoglobin (%SpO2) and pulse rate using photoplethysmography (PPG) technology. Most commercially available devices transmit light in specific red (~ 650 nm) and infrared (~ 880 nm) wavelengths through the tested tissue, and a detector on the other side absorbs the transmitted light. These wavelengths have a unique absorbance pattern upon interaction with oxyand deoxyhemoglobin. The detector measures the changing absorbance at each of the wavelengths allowing it to determine the absorbance resulting from the pulsating arterial blood alone excluding venous blood, skin, bone, muscle, and fat. The BB-613WP uses a unique reflective PPG technology, where, unlike most devices, the light source (Light Emitting Diodes, LEDs) and sensor array are placed on the same side (backside) of the device. As the LEDs transmit light into the subject's skin, part of this light is reflected from the tissue and is detected by a photodiode detector. The high temporal and quantitative resolution of the device allows it to capture minute changes in tissue reflectance, calculating numerous vital signs derived from pulse contours, including advanced hemodynamic parameters such as tracking changes in blood pressure. This is based on Pulse Wave Transit Time (PWTT) which is obtained utilizing pulse measurements from the integrated SpO2 sensor, following a calibration process using an oscillometric blood pressure monitor. Once a calibration measurement is taken using the cuff-based BP device, the values are entered into a user's application. The calibration is valid for three months, after which a new calibration measurement should be taken and introduced into the application using the same method.
The results are displayed on the LCD screen of the wristwatch device as well as on the mobile application installed on the user's smartphone. From there, the data is transmitted in real-time to a web application used by health care providers. Study protocol. Comparison between the devices was performed based on the International Organization for Standardization (ISO) guideline for non-invasive sphygmomanometers (ISO 81060-2:2013 25 ), which is considered by the FDA as a relevant protocol for regulatory use. BP measurements were taken by skilled EMS personnel. In each volunteer, a reference BP measurement using a cuff-based device (Welch Allyn DuraShock DS65 hand sphygmomanometer, Skaneateles Falls, NY 13153, USA) was taken and used as the baseline calibration value for the PPG-based wearable device (Fig. 6). Next, the wearable device was attached to each of the subjects by the EMS personnel and left for 30 s to acquire a reading of the PPG signal while the volunteers were sitting. This was repeated on the other hand to show there are no differences. After obtaining a signal, BP was measured concomitantly in all subjects using both devices-the standard manometry device mentioned earlier on one hand and the BB-613WP on the opposite arm. Since the manometry device occludes blood flow to the arm, PPG measurements cannot be obtained if the device is located on that same arm. When comparing the BB-613WP to a cuff-based BP device, and since it takes time between www.nature.com/scientificreports/ the systolic measurement and the diastolic measurement using the cuff-based device, the systolic value of the BB-613WP is recorded once systolic value is reached in the cuff-based device, and once the diastolic value is reached in the cuff-based device, the diastolic measurement in the BB-613WP device is recorded. Healthy subjects with no known background diseases were asked whether they would agree to perform a short session of exercise (5 min of walking fast near the measurement site). Those who agreed and signed were included in this sub-group. The aim was to change the BP and to show that the BB-613WP can detect this change between the two measurements. Adverse events were recorded by the research team, such as local wrist irritation (where the wristwatch was placed), or any other complaint by the user.

Statistical analysis.
According to ISO 81060-2:2013 the PPG device can be considered as equal to the reference when the mean value of the differences in paired systolic and diastolic readings between the reference and the PPG device is equal to or less than 5 mmHg and the Standard Deviation (SD) is no greater than 8 mmHg. The PPG device reliability was determined by the Interclass Correlation Coefficient (ICC) in which an ICC ≥ 0.9 indicates "excellent reliability" 26 . Correlation analysis was also performed using Pearson correlation, and agreement was evaluated based on the Bland-Altman method using 95% limits of agreement (LOA). The agreement in the classification (detection) of hypertensive cases (defined by the European Hypertension Society to be > 140 mmHg systolic pressure or > 90 mmHg diastolic pressure) was evaluated using Kappa agreement. Hypertension (SBP > 140 mmHg or DBP > 90 mmHg) detection ability was evaluated using ROC curves analysis. Pre-specified subgroup analysis of patients with and without previously diagnosed HTN was also performed. P values were set at 0.05. Data analysis was performed using the IBM SPSS Statistics for Windows, Version 23.