Introduction

Continuous blood pressure (BP) monitoring is extremely important for adequate intraoperative management of patients. Fast detection and aggressive correction of intraoperative hypotension reduce the incidence of ischemic damage to various organs, especially kidneys and myocardium, severity of postoperative multiple organ failure and mortality [1, 2]. These data were confirmed in a recent randomized study. The authors reported decrease of postoperative multiple organ failure severity by the 7^th (OR = 0.73, 95% CI [0.56; 0.94], p = 0, 02) and the 30^th (OR = 0.66, 95% CI [0.52; 0.84], p = 0.001) postoperative days in case of intraoperative maintaining of reference systolic blood pressure with 10% accuracy [3]. Continuous BP monitoring may be invasive or non-invasive. Continuous invasive BP monitoring is performed by insertion of an intra-arterial catheter and its connection with BP monitoring system. This highly accurate technique is a "gold standard" in modern anesthesiology. At the same time, this method is associated with a risk of various complications including trauma, bleeding, infections, thrombosis, embolism, distal ischemia and false aneurysms [4]. Oscillometric BP measurement devices are used in most surgical procedures due to their non-invasiveness and simplicity. However, this technique ensures only intermittent measurement (for example, every 3—5 minutes) and does not always reflect BP changes in timely and effective manner. Intermittent BP monitoring can skip up to 20% of hypotensive episodes during intraoperative anesthetic care [5]. For the last decade, the Infinity® CNAP™ SmartPod® monitor (Draeger Medical Systems Inc., USA) has been actively used for intraoperative and emergency continuous non-invasive BP measurement in adults. The CNAP™ system (Continuous Noninvasive Arterial Pressure monitoring) is based on the volume clamp method developed by the Czech physiologist J. Penaz [6]. CNAP™ monitors blood flow within the cuff on the finger and transmits its fluctuations as a constant pulse wave of blood pressure. Technically, this device consists of two cuffs attached to the patient's fingers and a pressure sensor attached to the forearm. CNAP™ calibration is periodically performed via non-invasive BP measurement with a cuff on the upper arm. Various authors compared the results of CNAP™ and invasive BP measurement in adults and obtained contradictory data [7—10]. The English-language PubMed database of medical and biological publications created by the National Center for Biotechnology Information comprises only 3 studies devoted to accuracy of intraoperative CNAP™ in pediatric surgery. Tobias J.D. et al. performed all these studies. One study recruited 20 children weighing ≥ 40 kg and reported an acceptable comparability of mean BP obtained in non-invasive and invasive monitoring [11]. There is another study comprising 20 patients weighing 20 — 40 kg. The researchers found clinically useful accuracy of non-invasive BP monitoring although the need to improve absolute accuracy associated with ineffective finger-to-cuff contact was emphasized [12]. In the third study comprising 30 children younger 10 years with a weight of less than 20 kg, the authors obtained low accuracy of continuous non-invasive BP monitoring if the cuff was placed on the lower limb [13]. Moreover, BP detection efficiency and accuracy of CNAP™ in children with hypotensive episodes are poorly understood.

The purpose of the study was to analyze accuracy of continuous intraoperative non-invasive BP measurement with CNAP™ monitor compared to invasive BP monitoring in children during and outside hypotension episodes.

Material and methods

A prospective observational study was carried out in the Dmitry Rogachev National Medical Research Center of Pediatric Hematology, Oncology and Immunology for the period from February to June 2020. All patients and their legal representatives signed an informed consent for inclusion in the study, storage, collection and processing of medical data. The published data are impersonal, ethical principles are observed.

The study included patients younger 18 years who underwent elective surgery for cancer in our center.

Exclusion criteria:

— no need for continuous intraoperative invasive BP monitoring;

— preoperative between-forearm difference in BP > 5 mm Hg obtained during oscillometric measurement in supine position;

— congenital or acquired anatomical differences between the wrists (vascular pathology of the upper limbs including vascular implants in the area of BP measurement);

— previous or current edema and ischemia of the upper limbs and fingers;

— impossible insertion of arterial catheter into radial artery (type C and D in Barbeau test), as well as complications related to this vascular access (hematoma, multiple punctures, arterial catheter malfunction);

— unavailable non-invasive continuous BP monitoring on the opposite limb (surgical field within one of the upper limbs, discrepancy between CNAP™ cuff dimensions and the patient's fingers);

— pacemaker, atrial fibrillation, paroxysmal tachycardia, no informed consent.

Two hundred and seventy-eight patients were analyzed regarding inclusion in the study; 255 patients were excluded before invasive and non-invasive BP measurement. Three patients were excluded after BP measurement (violations of invasive BP measurement technique in 2 patients and no full-fledged trend record in electronic anesthetic chart in another patient). Patient selection scheme is shown in Fig. 1. Characteristics of our cohort are summarized in Table 1.

Fig. 1. Flowchart diagram.

Table 1. Demographic and clinical data

Note. Data are presented as median and percentiles (Me (P25; P75)), mean and standard deviation (M ± SD), absolute numbers and percentages (n (%)).

Anesthetic care was carried out according to a single protocol adopted in our center. Anesthesia induction was achieved using sevoflurane, fast inhalation induction or intravenous injection of propofol 1% 1.5—2.5 mg/kg if vascular access was available. After that, we administered muscle relaxant (rocuronium bromide 0.6 mg/kg) and narcotic analgesics (fentanyl 2—5 μg/kg) and performed tracheal intubation. Intraoperative normothermia was maintained using a thermal mattress and convection air systems for heating the patient with control of central and peripheral body temperature. Epidural catheterization for analgesia with morphine (0.025—0.05 mg/kg) or ropivacaine (0.2—0.4 mg/kg/hour) was applied according to indications. Combined anesthesia was maintained using inhalation of sevoflurane (1–1.2 MAC), intravenous infusion of fentanyl (5–10 μg/kg/hour). Muscle relaxation was ensured by infusion of a relaxant. Conventional ventilation was carried out under control of capnometry and arterial (venous) blood gases. The last ones were intraoperatively controlled every hour. Infusion therapy was performed using crystalloid solutions (4—2—1 rule) in accordance with the Holliday, Segar or Oh scheme. Blood pressure was controlled using pharmacological vasoconstriction (norepinephrine 0.03—0.5 µg/kg/min). After surgery, patients could be weaned from ventilator, followed-up and transferred to the specialized department. In another case, they were transferred to the intensive care unit under mechanical ventilation.

Blood pressure measurement and control. A 20—22G catheter (Vasofix Certo "B. Braun", Germany) was inserted into radial artery under sterile conditions and ultrasonic navigation (Logiq E "GE Health Care", USA, 12L-RS 4.2—13.0 MHz linear transducer). This catheter was connected with a disposable sterile transducer-equipped line filled with heparinized saline solution (heparin 0.5 U per 1 ml of 0.9% NaCl). Transducer positioned on the patient's midline was calibrated in relation to atmospheric pressure. Standard flushing test was performed to determine natural frequency and damping factor of invasive catheter. Intraoperatively, transducer position control and calibration were carried out after each arterial blood sampling for routine clinical testing.

Continuous non-invasive BP measurement using a cuff on the shoulder and two cuffs on the index and middle fingers was performed on the side opposite to arterial line. The appropriate CNAP™ cuff size was selected according to the manufacturer's instructions. Finger cuff size was selected considering the size of the proximal phalanx of the patient's index finger and the image on the CNAP™ controller. If the size of the proximal phalanx was less than the smallest indicator by more than 10%, the patient was excluded from the study. Forearm cuff width was chosen so that it made up 37 — 47% of the forearm circumference at the midpoint. Calibration and finger cuff change time was set at 30 minutes.

Systolic, diastolic and mean blood pressures obtained from the CNAP™ system and arterial line, as well as all other indicators, were automatically recorded in electronic file of clinical information collection system every minute throughout the entire surgery. Prior to data analysis, artifacts of blood pressure measurement such as arterial blood sampling and patient position changes were detected and removed using Microsoft Office Excel 2007 software.

Severe intraoperative hypotension was recorded during invasive monitoring. According to the American Heart Association guidelines, systolic blood pressure < 90 mm Hg determines critical hypotension in children over 10 years old [14]. Mean blood pressure < 55 mm Hg in children over 10 years old and adults is associated with acute renal and myocardial injury [15, 16].

Statistical analysis. According to the recommendations of the Association for the Advancement of Medical Instrumentation (AAMI), the minimum number of patients in these studies should be over 15 [17, 18]. Lu M.J. et al. [19] published the manuscript in the International Journal of Biostatistics in 2016. The authors substantiated the equation for sample size for assessing agreement between two methods of measurement by Bland-Altman method. It was impossible to determine the sample using the data obtained by Tobias J.D. [11—13]. Indeed, these authors applied an absolute non-directional difference, i.e. positive and negative difference between two BP measurements was recorded only as a positive value. The International Organization for Standardization (ISO) and AAMI specify the difference between two methods of BP measurement within ± 5 mm Hg, while limits of agreement should be within ± 8 mm Hg [17, 18]. Thus, at least 846 paired measurements were required to achieve a power of 80% at statistical significance level of 95%. We decided to increase sample size by 30%, and the study was conducted until the required number of patients or paired measurements was reached.

Demographic characteristics are described by descriptive statistics. Continuous variables are presented as means and standard deviations or median and interquartile ranges, qualitative variables as frequencies and percentages. The Kolmogorov — Smirnov test (with Lilliefors correction) or Shapiro — Wilk test (in case of small sample) were applied to test the hypothesis of normal distribution of quantitative variables. The Levene’s test was used to assess the equality of variances. We applied a logarithmic transformation of blood pressure values to obtain normally distributed data. The Wilcoxon test for paired samples was used to assess the differences in BP measured by 2 methods, since data had abnormal distribution. Spearman's rank correlation coefficient was applied to establish the relationship between quantitative variables. Scatter diagrams with a line of best fit presented the results of correlation analysis and blood pressure comparison. The Bland-Altman method was used to assess agreement between measurements. We determined the difference and mean for each pair of BP values measured by 2 methods. Median, 2.5 and 97.5 percentiles were calculated for the difference. We used percentiles as the upper and lower limits of agreement instead of the interval ± 1.96 × SD (standard deviation), since the differences had abnormal distribution. The results are presented in Bland-Altman diagram. We also analyzed the percentage error as a ratio of limits of agreement to the median of the reference method. The percentage of agreement between the indicators in the obtained and recommended limits was determined. "Agreement/tolerance interval" ratio was also calculated. We assessed specificity, sensitivity, positive and negative likelihood ratios, predictive value of positive and negative results to analyze diagnostic efficiency of detecting critical hypotension through non-invasive blood pressure monitoring. Characteristic curves (ROC curves) were applied to determine split point. ROC value with the highest sum of sensitivity and specificity was defined as a split point of certain parameter. Differences were significant at p-value <0.05. Statistical analysis was performed using MedCalc version 11.3.3 (MedCalc Software) and Microsoft Office Excel 2007 software packages.

Results

Invasive systolic, mean and diastolic blood pressures measured in radial artery were 96 (89; 105), 67 (63; 74) and 54 (49; 59) mm Hg, respectively. Non-invasive systolic, mean and diastolic blood pressures measured with CNAP™ device were 94 (86; 104), 63 (56; 72) and 45 (40; 54) mm Hg, respectively. The data on invasive and non-invasive blood pressure and their logarithmic values had abnormal distribution (Kolmogorov-Smirnov test with Lilliefors correction — p <0.01; skewness coefficient — p <0.001; kurtosis coefficient — p <0.001). Considering these data, we used the Wilcoxon test for paired samples to assess between-group differences in blood pressure. This test confirmed significant differences for paired measurements of systolic (z = 12.96; p <0.0001), mean (z = 28.78; p <0.0001) and diastolic blood pressure (z = 41.17; p < 0.0001). Considering data distribution, we could not use intraclass correlation coefficient and Lin's concordance correlation coefficient for assessment of agreement. Therefore, Spearman's rank correlation coefficient was applied. Correlation analysis and between-group comparison of BP measurements are presented in scatter plots with a line of best fit (Fig. 2—4). According to the recommendations by Y.H. Chan [20], correlation strength between systolic blood pressures corresponds to trivial reliability of agreement between two measurement methods, between mean and diastolic blood pressures — moderate reliability of agreement. The Bland-Altman method was used to assess agreement between measurements. We calculated bias and mean for each pair of BP measurements. Differences in systolic, mean and diastolic blood pressures had abnormal distribution (Kolmogorov-Smirnov test with Lilliefors correction — p <0.01; skewness coefficient — p <0.03; kurtosis coefficient — p <0.001). Logarithmic transformation of data and calculation of difference of new values did not change the nature of distribution. In 1999, J.M. Bland and D.G. Altman described a non-parametric approach to comparison of two measurement methods in the journal “Statistical Methods in Medical Research” [21]. This approach was approved by P.J. Twomey in recommendations published in the journal “Annals of Clinical Biochemistry” in 2006 [22]. According to these recommendations, median of bias instead of mean is required in this situation, as well as 2.5 and 97.5 percentiles as the upper and lower 95% agreement limits instead of interval ± 1.96 × SD (standard deviation). Repeatability test will be irrelevant in determining the reliability of agreement in this situation. Median of bias, limits of agreement and percentage error are presented in Bland-Altman diagrams (Fig. 5—7). A new measurement technique is considered acceptable if at least 85% of measurements with the device under test agree with measurements obtained with a reference device [23]. The agreement limits comprise 96.5% of biases for systolic and mean blood pressure, as well as 98.3% of biases for diastolic blood pressure. At the same time, percentage errors significantly exceed the maximum recommended limit of 30% [24, 25]. ISO and AAMI specify that difference between two measurements of blood pressure should be ≤5 mm Hg, limits of agreement — within ± 8 mm Hg [17, 18]. Only difference between systolic and mean blood pressure corresponds to these recommendations. Only 86.2% of mean blood pressures correspond to the recommended maximum range of limits of agreement (± 13 mm Hg). In 2008, Columb M.O. [26] described an alternative method of assessing agreement (agreement / tolerance interval) in the journal “Current Anesthesia and Critical Care”. ATI ratio was 3.8 for systolic blood pressure and 2.5 for mean and diastolic blood pressure, i.e. this technique of measurement is unacceptable for clinical use. In 1999, Bland J.M. and Altman D.G. described a non-parametric approach to comparison of methods in the journal “Statistical Methods in Medical Research” [21]. Considering this approach and scientific data on non-invasive and invasive blood pressure measurements, limits of agreement ± 5 mm Hg are recommended for clinical practice [11—13, 17, 23, 27]. Only 33.6% of systolic blood pressures, 41% of mean blood pressures and 35.4% of diastolic blood pressures fall into this range of agreement. Thus, CNAP™ monitoring receives the lowest possible grade "C", and reasonable clinical use of this device only for mean blood pressure measurement is emphasized. The device is characterized by grade "D" for systolic and diastolic blood pressures. Therefore, this method is inappropriate for clinical use [21]. Measurement accuracy of CNAP ™ monitoring does not change for limits of agreement ± 10 and ± 15 mm Hg.

Fig. 2. Scatter plot of systolic blood pressure measurements with the line of best fit.

R_ho — Spearman’s correlation coefficient with 95% CI; p — the level of statistical significance; IBP-S — systolic blood pressure measured in the radial artery; CNAP-S — systolic blood pressure measured noninvasively.

Fig. 3. Scatter plot of mean blood pressure measurements with the line of best fit.

R_ho — Spearman’s correlation coefficient with 95% CI; p — the level of statistical significance; IBP-M — mean blood pressure measured in the radial artery; CNAP-M — mean blood pressure measured noninvasively.

Fig. 4. Scatter plot of diastolic blood pressure measurements with the line of best fit.

R_ho — Spearman’s correlation coefficient with 95% CI; p — the level of statistical significance; IBP-D — diastolic blood pressure measured in the radial artery; CNAP-D — diastolic blood pressure measured noninvasively.

Fig. 5. Bland-Altman diagram for systolic blood pressure values.

IBP-S — systolic blood pressure in the radial artery; CNAP-S — noninvasive systolic blood pressure (measured by the CNAP system); Bias — median of difference between systolic blood pressure measured invasively and noninvasively with 95% CI; Upper limit and Lower limit — upper and lower limits of agreement (2.5^th and 97.5^th percentiles with 95%CI); PE — percentage error (%).

Fig. 6. Bland-Altman diagram for mean blood pressure values.

IBP-M — mean blood pressure in the radial artery; CNAP-M — noninvasive mean blood pressure (measured by the CNAP system); Bias — median of difference between mean blood pressure measured invasively and noninvasively with 95% CI; Upper limit and Lower limit — upper and lower limits of agreement (2.5^th and 97.5^th percentiles with 95%CI); PE — percentage error (%).

Fig. 7. Bland-Altman diagram for diastolic blood pressure values.

IBP-D — diastolic blood pressure in the radial artery; CNAP-D — noninvasive diastolic blood pressure (measured by the CNAP system); Bias — median of difference between diastolic blood pressure measured invasively and noninvasively with 95% CI; Upper limit and Lower limit — upper and lower limits of agreement (2.5^th and 97.5^th percentiles with 95%CI); PE — percentage error (%).

Diagnostic efficacy of continuous non-invasive systolic and mean blood pressure measurement for detecting hypotension is presented in Table 2. ROC-curves indicating the optimal split point as the greatest sum of sensitivity and specificity are shown in Fig. 8—9. Analysis of area under ROC curve showed low accuracy of diagnosis of critical hypotension for systolic blood pressure measurements and good accuracy for mean blood pressure measurements with CNAP ™ monitor. For mean blood pressure only, optimal split point is the closest to diagnostic one within the recommended range of agreements limits ± 5 mm Hg.

Table 2. Parameters of diagnostic efficiency of detecting hypotension with a CNAP monitor

Note. Data are presented as value and 95% CI; BP — blood pressure.

Fig. 8. Non-invasive detection of critical hypotension based on systolic blood pressure (CNAP-S).

The data are presented in the form of a ROC curve; for the cut-off value ≤98 mmHg. AUC_ROC (95% CI) = 0.7 (0.69; 0.72), SE=0.009, p<0.0001.

Fig. 9. Non-invasive detection of critical hypotension based on mean blood pressure (CNAP-M).

The data are presented in the form of a ROC curve; for the cut-off value ≤57 mmHg. AUC_ROC (95% CI) = 0.84 (0.83; 0.85), SE=0.0135, p<0.0001.

Discussion

Continuous intraoperative blood pressure monitoring is an important part of anesthetic care due to the need for reducing the time of undesirable blood pressure oscillations. Modern ideal perioperative blood pressure monitoring should be non-invasive, continuous, free from the risk of complications, characterized by excellent consistency and interchangeability with intravascular blood pressure monitoring. Currently, technologies of intraoperative non-invasive blood pressure monitoring cannot completely replace continuous invasive monitoring in children. Ilies C. et al. [24] believe that this fact is determined by low efficiency of these devices for detection of sharp oscillations of blood pressure and episodic hypotension. According to our study, intraoperative systolic and diastolic blood pressure measurements with a CNAP ™ monitor in children does not meet the ISO and AAMI acceptance criteria. Only mean BP values obtained with a CNAP™ monitor meet the recommended maximum range of agreement limits [23]. Importantly, standards have not been developed for patients under inotropic support and those with severe hypotension. However, their advantage is determined by avoiding subjective interpretations of what is or is not an acceptable indicator [28]. Considering this fact, we applied alternative methods for assessing agreement in this study in addition to the above-mentioned standards [26]. Moreover, non-parametric approach to comparing methods for BP measurement was used [21]. CNAP™ monitoring receives the lowest possible grade "C", and reasonable clinical use of this device only for mean blood pressure measurement is emphasized. Other methods of assessing agreement have also shown trivial reliability of mean BP measurements by CNAP™ monitor. K. Lakhal et al. explained these results by inaccurate calibration of CNAP™ finger cuffs with its own automated shoulder cuff system. CNAP™ monitor recalibration is required after 4 minutes of continuous measurements [29]. Frequent calibration can damage to nerve plexuses and skin of the shoulder, as well as potentiate local thrombotic processes and perfusion impairment in distal limb used for blood pressure measurements. This is especially true in patients with intraoperative hemostatic disorders throughout the periods of hypotension. Z. Wang et al. [7] explained low efficiency of CNAP™ monitor by low arterial pulse amplitude following constant cuff-related pressure on the finger and venous outflow obstruction. These facts explain the ineffectiveness and significant oscillations of CNAP™ monitor data in relation to real blood pressure in patients with poor tissue perfusion following hypotension, shock, inotropic therapy, rapid arterial vascular tone changes during anesthesia induction, motor activity in insufficient muscle relaxation, hypothermia, and arrhythmias.

Analysis of diagnostic efficiency for detecting critical hypotension with continuous non-invasive mean blood pressure measurement showed higher sensitivity, specificity, positive likelihood ratio and accuracy compared to critical hypotension detected via systolic blood pressure measurement. Nevertheless, accuracy of detecting critical hypotension based on mean blood pressure measurement with CNAP™ monitor is less than the recommended agreement limit (85%) [23]. Optimal split point for mean blood pressure is the closest to the diagnostic one and lies within the recommended agreement limit of ± 5 mm Hg. Nevertheless, detection of critical hypotension based on non-invasive systolic blood pressure measurement with CNAP™ monitor is characterized by higher positive predictive value. This is due to higher prevalence of critical hypotension diagnosis based on systolic blood pressure measurement compared to mean blood pressure. Considering this fact, physicians should intraoperatively consider systolic blood pressure in addition to mean values. These measures will ensure earlier prevention and treatment of hypotension and subsequent tissue perfusion impairment. According to our data, CNAP-based systolic blood pressure < 99 mm Hg and mean blood pressure < 58 mm Hg require therapy of hypotension.

Study limitations. The limitation of our study is small sample size. According to the European Society of Hypertension (ESH) guidelines, at least 33 patients and 99 paired measurements are proposed to evaluate the new device for blood pressure measurement [30]. Nevertheless, the study was carried out in strict accordance with the sample size required. The AAMI recommendations were followed. The number of paired measurements was over 99 in each patient while the subject of our study was accuracy of blood pressure measurement within and outside hypotensive episodes. CNAP™ monitoring in pediatric population is primarily limited by finger cuff dimension, since they are not manufactured specifically for children. Another aspect is difficult selection of the cuffs due to different sizes and shapes of the fingers. In this study, we excluded patients who were unable to fit finger cuffs according to the manufacturer's recommendations. As a result, a cohort comprising exclusively adolescents was recruited. Additional factors that could potentially bias the results are CNAP™ device placement (contralateral to invasive blood pressure measurement) and radial artery as a component of reference measurement. In our study, this limitation was partially considered. Indeed, patients with preoperative between-forearm difference in oscillometrically measured blood pressure > 5 mm Hg were excluded. We also excluded patients with baseline heart rhythm disturbances, complicated radial artery catheterization and other defects in invasive and non-invasive blood pressure measurement. Importantly, possible measurement errors in CNAP™ monitoring associated with patient movements were excluded because all patients were under general anesthesia. We did not estimate the effect of finger temperature on CNAP™ monitoring. Nevertheless, minimum intraoperative cutaneous temperature was 35.7 ± 0.3°C. A 30-minute calibration interval could contribute to inaccuracies in CNAP ™ monitoring over time.

Conclusion

Intraoperative mean blood pressure measurement with CNAP™ monitor in children meets the ISO and AAMI acceptance criteria. Thus, this approach may be clinically used, especially in emergency cases where there is no arterial access for invasive blood pressure measurement. Systolic blood pressure < 99 mm Hg and mean blood pressure < 58 mm Hg obtained during CNAP ™ monitoring may be triggers for therapy of hypotension. Non-invasive systolic blood pressure measurements can ensure early initiation of therapy for critical hypotension and prevention of tissue perfusion impairment. There are limitations for continuous non-invasive blood pressure measurement with CNAP ™ monitor in pediatric population. Therefore, continuous invasive blood pressure monitoring should be currently used during surgical interventions potentially associated with prolonged episodes of unstable hemodynamics requiring infusion therapy and inotropic support. Despite a large number of paired blood pressure measurements, the study included a small sample and there is a heterogeneity factor in this pediatric cohort. Therefore, further research and meta-analysis of data are required to formulate final conclusions.

Author contribution:

Concept and design of the study — Leonov N.P.

Collection and analysis of data — Shchukin V.V., Gasparyan K.R., Ivanova I.V., Zinchenko A.S.

Statistical analysis — Leonov N.P.

Writing the text — Leonov N.P.

Editing — Shchukin V.V., Maschan M.A., Novichkova G.A., Spiridonova E.A.

Acknowledgment. The authors express their deep gratitude to the nursing staff of intensive care unit and surgical unit of the Dmitry Rogachev National Medical Research Center of Pediatric Hematology, Oncology and Immunology.

The authors declare no conflicts of interest.