Upper arm blood pressure (BP) measurement is the most commonly accepted method of noninvasive BP monitoring. During a cesarean delivery (CD), accurate measurement from an automated device may be impossible in more than half of the attempts, because of arm movement.a
Patients may voluntarily move their arms, or neuraxial anesthesia may induce shivering, causing the patient’s arm to shake.1 An audit of deliveries in our operating room found that shivering caused failure of BP measurement in 38% of cases with epidural anesthesia and 9% with spinal anesthesia.b To reduce the effect of movement of the upper arm, BP has been measured on the leg, which is immobile during neuraxial anesthesia. However, the findings of several studies suggest that measuring BP on the ankle or calf may not accurately detect changes (increase or decrease) in BP in patients undergoing CD.2,3 It seems reasonable that the forearm, with less muscle mass, could be stabilized more easily than the upper arm during shivering. Therefore, positioning a BP cuff on the forearm (near the wrist) may be useful in obtaining BP measurements in shivering patients.
Previous studies have found that forearm BP overestimates BP measured at the upper arm in a variety of clinical settings and patient populations under static (patient at rest, not submitted to an intervention that might alter BP) circumstances.4–7 It is unknown whether this overestimation is clinically relevant when BP changes occur rapidly, such as during a CD. During neuraxial anesthesia for CD, it is important for the anesthesiologist to know both absolute values and clinically significant changes in BP. The detection of dynamic changes in BP, such as during neuraxial anesthesia, has been studied in obstetric patients in alternate sites, such as the finger, the calf, or the thigh. Nevertheless, either because of the location of the BP cuff (e.g., lower limb, which may be influenced by aortocaval compression3) or the technology used to measure BP (e.g., Penaz technology, not comparable with oscillometric technology8), BP was not comparable with the upper arm BP.
This study compared the upper arm (the current gold standard) and the wrist BP measurements on patients undergoing an elective CD under spinal anesthesia. We hypothesized that changes in the wrist systolic blood pressure (sBP) would accurately trend with changes in the upper arm sBP measured at the same time points, with an absolute difference between methods of less than ±10%.
After IRB approval (August 7, 2012, University of British Columbia, Children’s & Women’s Health Centre of BC, Vancouver, British Columbia, Canada) and registration with Clinical Trials (August 29, 2012, clinicaltrials.gov NCT01677234), patients consented in writing to participate in this single-center, prospective, observational trial. Patients were included if they were healthy (ASA physical status I and II) female adults (age >19 years), having a scheduled CD under spinal or combined spinal–epidural (CSE) anesthesia. Exclusion criteria were any contraindication to neuraxial anesthesia, use of general anesthesia, previous history of cardiovascular disease, body mass index >38 kg/m2, inability to read and understand English for the purpose of informed consent, and previous history of tremor disorders.
Patients enrolled but not submitted to statistical analysis included those who had CSE anesthesia but subsequently had an epidural top-up before delivery, in addition to those who had a difference of >10% between upper arm measurements (mean blood pressure [mBP]) or >10% difference between wrist measurements (mBP) before surgery (see below).
To determine eligibility for the study, mBP readings were done simultaneously in both upper arms and in both wrists. This ruled out subjects who might have anatomic abnormalities of the subclavian, brachial, or radial arteries that could produce bias.
Consenting subjects received standard care according to routine hospital protocols. All patients had an IV line inserted in the upper aspect of the forearm (midway between wrist and crease of elbow) on the same side as the wrist BP cuff. If the IV line could not be placed in the upper forearm above the wrist BP cuff, it was placed in the antecubital fossa (crease of the elbow). A normal saline solution infusion was started. All study measurements took place in the operating room, with the patient in the supine position and a 15° wedge placed under her right hip to promote left uterine displacement.
BP cuffs were placed on the arm or forearm (wrist), with the tubing facing the respective artery (brachial or radial, respectively).
Initiation of neuraxial anesthetic was performed with patients in the sitting position. A 25-gauge Whitacre needle was used for spinal anesthesia and a needle-through-needle technique with a 17-gauge Tuohy needle and a 27-gauge Whitacre needle for CSE anesthesia. After clear cerebrospinal fluid was visualized, intrathecal hyperbaric bupivacaine 0.75% (dose ranging from 10.5 to 12.75 mg at the discretion of the anesthesiologist), fentanyl 10 µg, and morphine 100 µg were injected.
Immediately after the drugs were injected, the patient was placed in the supine position with left uterine displacement. Baseline BP was recorded, and an IV vasopressor infusion of either phenylephrine (100 µg/mL) or a combination of phenylephrine and ephedrine (ratio of 50 µg: 2 mg/mL) was started—the rate varying at the discretion of the attending anesthesiologist.
Two calibrated monitors were used to simultaneously measure BP on the upper arm and the wrist. The additional BP monitor used was the Datex-Ohmeda/GE Healthcare Carescape V100 (Monitor: S5 Compact, Module: M-PRESTN; GE Healthcare, Helsinki, Finland). This monitor is equivalent to (same module) the monitor currently used in our operating rooms, the Datex-Ohmeda/GE Healthcare (Monitor: S5, Module: M-NETPR). The monitors were calibrated by the biomedical department before the start of the study.
The cuff size was appropriate to the size of the arm; the American Heart Association’s recommendations9 for BP measurement in humans and experimental animals were followed. Because patients with obesity were excluded, all subjects had a “regular adult” cuff size for the upper arm (16 × 30 cm). For the wrist, we chose the “small adult/older child” cuff size as a standard (9.5 × 18 cm).4
The 2 BP cuffs remained in the same place from the initiation of anesthesia until delivery. Although the upper arm BP was programmed to cycle every minute, some measurements took longer than 1 minute. To ensure that wrist BP measurement was done at the same time, the wrist BP measurement was started manually by the investigator each time the arm BP started to cycle. The study ended with the delivery of the baby. BP was recorded on the data collection sheet after each measurement.
Data collected included age, gestational age, height, weight, body mass index, parity, indication for CD, single or multiple gestation, heart rate, BP measurements (systolic, diastolic, mean BP), time of initiation of anesthesia, dose of intrathecal drugs, time of delivery, ASA physical status, and intraoperative treatment for changes in BP. We considered a 20% decrease or increase in sBP to be clinically significant, because this change normally triggers treatment.
For this study, we arbitrarily decided that a difference of ≥10% between the wrist sBP and the upper arm sBP would be considered clinically important. The accuracy of the wrist sBP compared with the criterion standard upper arm sBP was analyzed during the entire period of the study, including the times when the subject was normotensive, hypotensive, and hypertensive, and when she received medication to treat abnormalities in vital signs. Mean BP and diastolic blood pressure (dBP) were also compared between the 2 measurement sites.
The primary outcome was the level of agreement between the percentage change in sBP measured at the wrist and at the upper arm over the course of surgery.
The secondary outcome was the agreement between BP measured at the wrist compared with upper arm BP at the baseline measurement and over time.
We used the Bland-Altman method comparison procedure using limits of agreement (LOA) with the extension for repeated measures on a quantity that varies over time.10 This method requires estimation of 2 variances: (1) the variance of the repeated differences between the 2 methods on the same subject, and (2) the variance of the differences between the averages of the 2 methods among subjects. The within-subject variance is estimated as the residual mean square from a 1-way analysis of variance (ANOVA) with the difference in matched pairs as the response, and the subject ID as the predictor. The variance of the average difference among subjects is the difference between the mean square for subjects and the residual mean square divided by Equation 1, which depends on the number of observations per subject:
where mi is the number of observations on subject i, and n is the number of subjects. The result from Equation 1 is then added to the residual mean square (variance 1 above) to give the total variance for single differences between the methods on different subjects (Equation 2).
The square root of Vartotal is used in place of the SD in the Bland-Altman equation for the LOA:
is the mean difference calculated as the mean of the individual differences, and df is n − 1.
For the primary outcome (comparison of changes in sBP measurements between the upper arm and the wrist), we performed the Bland-Altman analysis for the percentage change in sBP, dBP, and mBP from baseline. For the secondary outcome, we performed 2 Bland-Altman analyses comparing the baseline data using the standard Bland-Altman procedure and data obtained over multiple measurements for sBP, dBP, and mBP using the modified Bland-Altman procedure.
Bland-Altman comparison methods are problems of estimation and not hypothesis testing; we therefore estimated the sample size required to give reasonable precision around our estimates of the LOA. For the baseline comparisons, the SE of the LOA is approximately
, where SD is the standard deviation of the differences between measurements and n is the sample size.11 The 95% confidence intervals of the LOA are then approximate limit ± 1.96 × SE, which can be rewritten as
. A sample of 50 subjects would therefore give precision estimates of approximately 0.5 × SD, or half the SD, which is a reasonable level of precision.
Between August 31, 2012, and May 12, 2013, 105 subjects were evaluated for participation. Sixty-nine subjects provided written informed consent. Fifty-four subjects had complete data. Among those, 5 were excluded from analysis (Fig. 1). In the first 3, the wrist BP was measured with an “infant” cuff size (6 × 16 cm). This resulted in an extreme overestimation of BP, and therefore, we used a “small adult/older child” cuff size for all subsequent wrist BP measurements. These first 3 subjects were excluded from analysis. Another 2 subjects were excluded because of shivering that interfered with several BP measurement attempts. Demographic characteristics of the study population are reported in Table 1.
Bland-Altman Plots for Agreement over Multiple Measurements on Percentage Change from Baseline
For this series of Bland-Altman analyses, we calculated the percentage change from baseline for each measurement on each individual for both methods (n = 913 pairs of measurements). One subject (no. 49) was removed from this analysis because of missing baseline data. Graphical examination of the differences suggested no relationship between the bias and the average BP measured for percentage change in sBP, dBP, and mBP. The Shapiro-Wilk tests strongly rejected the null hypothesis of a normal distribution for the differences between methods in the 3 BP measures (all P < 0.001; Supplemental Digital Content 1, http://links.lww.com/AA/B155), and D’Agostino tests also rejected no skewness or kurtosis (all P < 0.01). However, tests of normality are very sensitive to small deviations from an ideal normal distribution when sample sizes are large, and graphical examination of Q-Q plots and distributions did not indicate major deviations from normality for any of the 3 difference measures, although there was evidence of some kurtosis. Therefore, we were confident in continuing with the 1-way ANOVA to estimate the variance components needed for the LOA calculations. However, to be conservative, we altered the Bland-Altman LOA to use 99% prediction limits instead of 95% in Equation 3. The results of the analysis are described in Table 2.
When multiple measurements were considered across individuals, the 99% LOA are approximately ±25 percentage points for sBP, ±30 percentage points for dBP, and ±22 percentage points for mBP (Fig. 2).
When the time series for each subject was examined for percentage change from baseline, the 2 methods tracked each other well most of the time, but there were fairly large discrepancies in the absolute measurements for some individuals. This is reflected in the intraclass correlation coefficients for each subject (Fig. 3).
Bland-Altman Plots for Baseline Data
There was no indication that the difference between methods was related to the average BP (Fig. 4). There was no indication of deviation from normality for the differences between methods by the Shapiro-Wilk test (all P > 0.05) or by graphical examination. The results of the analysis for baseline data are described in Table 2.
Bland-Altman Plots for Agreement of BP over Multiple Measurements
Graphical examination of the differences suggested no relationship between the bias and the average BP (Fig. 5). The Shapiro-Wilk tests rejected the hypothesis that the data (n = 932 pairs of measurements) came from normal distributions (all P < 0.001), and D’Agostino tests also rejected no skewness or kurtosis (all P < 0.01). However, graphical examinations showed no indications of major deviations from normality except for some kurtosis (Supplemental Digital Content 2, http://links.lww.com/AA/B156), and 1-way ANOVA was used to estimate the variance components. To be conservative, we calculated the 99% LOA for the percentage change from baseline analysis. The analysis for the agreement obtained over multiple measurements is described in Table 2.
When the time series for each subject was examined individually, it appeared that for some subjects there was good agreement in BP measurements across all time points. However, there were a substantial number of subjects for whom the bias was very large (Fig. 6). For sBP, there were only 2 subjects whose measurements never differed by >10 mm Hg in either direction.
Upper arm BP measurement has been the standard of practice during CD. Because of movement artifact interference with upper arm BP measurements, alternative sites have been tested, including the thigh and the ankle. However, none has proven accurate,2,3,12 and there was a failure to detect hypotension at those sites.2,3 Previous studies have shown that aortic compression causes leg BP readings to be lower than those in the arm,3,13,14 and Bieniarz et al.14 showed that supine aortic compression increased with increasing systemic hypotension.14 Therefore, a clinically important decrease in BP might not be detected if the lower limb is used as the measurement site.
We speculated that the wrist BP measurements could provide an alternative BP monitoring when the upper arm BP fails because of shivering. Therefore, we aimed to verify the accuracy of the oscillometric BP measurements at the wrist compared with the upper arm during dynamic changes in elective CDs under spinal anesthesia. To our knowledge, no previous studies have correlated acute changes in BP over time on the wrist compared with the upper arm.
The oscillometric method measures BP through the relationship between the low-amplitude pressure pulses induced in the pressurized cuff and the cuff pressure. The pulse amplitude is maximal when the cuff pressure equals the mean arterial BP. The pressure applied across the arterial wall balances the mean arterial BP, resulting in peak pulsations.15,16 The mean BP, which is directly measured with the oscillometric method, was used to determine the eligibility for the study. We chose the systolic BP as the primary outcome because this variable is the most commonly used.
Overall, the wrist BP tended to overestimate the upper arm BP both for baseline data (sBP bias = 13.4 mm Hg; LOA = −8.3 to +36.9) and for data obtained over multiple measurements (sBP bias = 12.8 mm Hg; 99% LOA = −19.0 to +44.6 mm Hg).
Our results were in accordance with previously published studies of upper arm versus forearm measurements in nonobstetric patients.4–7 The higher BP readings obtained at the wrist compared with the ones obtained at the upper arm most likely reflected the progressive proximal to distal pulse amplification related to the stiffening of the arterial tree. However, when percentage change from baseline over multiple measurements was analyzed, the wrist BP values correlated well with the upper arm measurements in most cases. However, although the mean difference was only −0.2%, the LOA were between −24.9 and +24.5 percentage points. Thus, we reject our hypothesis that the absolute difference between the 2 BP measurement methods is less than ±10%. Although not statistically analyzed, the BP measurement changes appeared to occur in the same direction in most subjects, which might be a clinically relevant result.
Our study was not designed to examine the incidence of shivering or the frequency with which it interfered with BP measurement. Nevertheless, 5 subjects had shivering that interfered with measurements and 2 subjects had to be excluded from the study because shivering impeded measurement in so many events that the data could not be analyzed. No interventions were made to reduce shivering. The overall incidence of failure to measure BP because of shivering was 7.25% (5 of 69). Eleven readings were missed because of shivering in the upper arm, and only 2 readings were missed at the wrist. Although these data are only circumstantial, it suggests that measuring noninvasive BP at the level of the wrist may be more successful than at the upper arm in the event of shivering.
There are several limitations to our study design and conclusions. A weakness of this study is that the allocation of the 2 BP modules to the wrist or the upper arm was not randomized. Another limitation was the routine use of a vasopressor. Although it is the standard of practice at our institution to initiate a prophylactic infusion, this practice does limit BP variability. This could limit the generalizability of our results.
One of the difficulties with this study was estimating the required sample size to arrive at a meaningful conclusion. Because no previous study had compared these 2 methods of measuring BP, we were unable to find estimates of the SD of the difference between them. Fortunately, when multiple measurements are taken from the same individual, the level of precision is increased (width of the confidence interval is decreased), although not to the same extent as if all measurements were considered independent.10 Conservatively, if we expect that multiple measurements (approximately 20 per subject in this study) would increase the effective sample size to 100, then the precision of the LOA would be increased to 0.34 × SD, which is fairly narrow. The results of this study will assist in sample size calculations for future studies, and 50 subjects were considered both adequate and feasible for this purpose.
In conclusion, our results suggest that the wrist BP measurements do not provide accurate values compared with the gold standard upper arm BP measurements in healthy parturients scheduled for CD under spinal anesthesia. However, the wrist BP monitoring provided reliable tracking of the direction of the dynamic changes in most cases. Thus, if the upper arm BP measurements fail, wrist BP monitoring would be expected to provide clinically relevant, but not accurate, trending of BP measurements.
Future research is still needed to confirm that wrist BP is suitable for obtaining readings in shivering patients and patients with cardiovascular disease, especially those with hypertensive disorders of pregnancy.
Name: Ilana Sebbag, MD.
Contribution: This author helped design the study, conduct the study, collect the data, analyze the data, and prepare the manuscript.
Attestation: Ilana Sebbag approved the final manuscript and attests to the integrity of the original data and the analysis reported in this manuscript.
Name: Simon R. Massey, MB, BCh, MRCP, FRCA, FRCPC.
Contribution: This author helped design the study, conduct the study, analyze the data, and review the manuscript.
Attestation: Simon R. Massey approved the final manuscript and attests to the integrity of the original data and the analysis reported in this manuscript.
Name: Arianne Y. K. Albert, BSc, PhD.
Contribution: This author helped design the study, analyze the data, and review the manuscript.
Attestation: Arianne Y. K. Albert approved the final manuscript and attests to the integrity of the original data and the analysis reported in this manuscript.
Name: Alison Dube, BSc.
Contribution: This author helped conduct the study and collect the data.
Attestation: Alison Dube approved the final manuscript and is the archival author.
Name: Vit Gunka, MD, FRCPC.
Contribution: This author helped analyze the data and review the manuscript.
Attestation: Vit Gunka approved the final manuscript.
Name: M. Joanne Douglas, MD, FRCPC.
Contribution: This author helped design the study and prepare the manuscript.
Attestation: M. Joanne Douglas approved the final manuscript.
This manuscript was handled by: Cynthia A. Wong, MD
We acknowledge the members of the anesthesia department at BC Women’s Hospital for compliance with the study protocol.
a Hamad M, Freeman R, Wrench IJ. Incidence of failure of upper limb automated blood pressure measurement during caesarean section. Int J Obstet Anesth 2001;10:218
b Sebbag I, McCurdy S, Bright S, Murray D, Douglas, J. The incidence of noninvasive blood pressure measurement failure, as a result of shivering, during delivery in the operating room: an audit. SOAP Annual Meeting 2013, S1
1. Crowley LJ, Buggy DJ. Shivering and neuraxial anesthesia. Reg Anesth Pain Med. 2008;33:241–52
2. Moore C, Dobson A, Kinagi M, Dillon B. Comparison of blood pressure measured at the arm, ankle and calf. Anaesthesia. 2008;63:1327–31
3. Sanghera S, North A, Abernethy S, Wrench I. Arm and ankle blood pressure during caesarean section. Int J Obstet Anesth. 2006;15:24–7
4. Palatini P, Longo D, Toffanin G, Bertolo O, Zaetta V, Pessina AC. Wrist blood pressure overestimates blood pressure measured at the upper arm. Blood Press Monit. 2004;9:77–81
5. Pierin AM, Alavarce DC, Gusmão JL, Halpern A, Mion D Jr. Blood pressure measurement in obese patients: comparison between upper arm and forearm measurements. Blood Press Monit. 2004;9:101–5
6. Singer AJ, Kahn SR, Thode HC Jr, Hollander JE. Comparison of forearm and upper arm blood pressures. Prehosp Emerg Care. 1999;3:123–6
7. Schell K, Morse K, Waterhouse JK. Forearm and upper-arm oscillometric blood pressure comparison in acutely ill adults. West J Nurs Res. 2010;32:322–40
8. Epstein RH, Kaplan S, Leighton BL, Norris MC, DeSimone CA. Evaluation of a continuous noninvasive blood pressure monitor in obstetric patients undergoing spinal anesthesia. J Clin Monit. 1989;5:157–63
9. Pickering TG, Hall JE, Appel LJ, Falkner BE, Graves J, Hill MN, Jones DW, Kurtz T, Sheps SG, Roccella EJSubcommittee of Professional and Public Education of the American Heart Association Council on High Blood Pressure Research. Subcommittee of Professional and Public Education of the American Heart Association Council on High Blood Pressure Research. . Recommendations for blood pressure measurement in humans and experimental animals: Part 1: blood pressure measurement in humans: a statement for professionals from the Subcommittee of Professional and Public Education of the American Heart Association Council on High Blood Pressure Research. Hypertension. 2005;45:142–61
10. Bland JM, Altman DG. Agreement between methods of measurement with multiple observations per individual. J Biopharm Stat. 2007;17:571–82
11. Bland JM, Altman DG. Statistical methods for assessing agreement between two methods of clinical measurement. Lancet. 1986;1:307–10
12. Drake MJ, Hill JS. Observational study comparing non-invasive blood pressure measurement at the arm and ankle during caesarean section. Anaesthesia. 2013;68:461–6
13. Rees SG, Thurlow JA, Gardner IC, Scrutton MJ, Kinsella SM. Maternal cardiovascular consequences of positioning after spinal anaesthesia for caesarean section: left 15 degree table tilt vs. left lateral. Anaesthesia. 2002;57:15–20
14. Bieniarz J, Maqueda E, Caldeyro-Barcia R. Compression of aorta by the uterus in late human pregnancy. I. Variations between femoral and brachial artery pressure with changes from hypertension to hypotension. Am J Obstet Gynecol. 1966;95:795–808
15. Ramsey M III. Noninvasive automatic determination of mean arterial blood pressure. Med Biol Eng Comput. 1979;17:11–8
16. Amoore JN. Oscillometric sphygmomanometers: a critical appraisal of current technology. Blood Press Monit. 2012;17:80–8
17. Shrout PE, Fleiss JL. Intraclass correlations: uses in assessing rater reliability. Psychol Bull. 1979;86:420–8
Supplemental Digital Content
© 2015 International Anesthesia Research Society