Secondary Logo

Journal Logo

Using Heart Rate Variability to Stratify Risk of Obstetric Patients Undergoing Spinal Anesthesia

Section Editor(s): Birnbach, David J.Chamchad, Dmitri MD*; Arkoosh, Valerie A. MD*; Horrow, Jay C. MD, MSstat*; Buxbaum, Jodie L. MD*; Izrailtyan, Igor MD; Nakhamchik, Lev MS; Hoyer, Dirk PhD; Kresh, J Yasha PhD

doi: 10.1213/01.ANE.0000140953.40059.E6
Technology, Computing, and Simulation: Research Report

In this study, we evaluated whether point correlation dimension (PD2), a measure of heart rate variability, can predict hypotension accompanying spinal anesthesia for cesarean delivery. After the administration of spinal anesthesia with bupivacaine, hypotension was defined as systolic blood pressure ≤75% of baseline within 20 min of intrathecal injection. Using the median prespinal PD2 (3.90) to form 2 groups, LO and HI, all 11 hypotensive patients were in the LO group, and all 11 patients without hypotension were in the HI group. Baseline heart rate in the LO group was 95 bpm (10.2 sd), versus 81 bpm (9.6 sd) in the HI group. PD2 shows promise as a predictor of hypotension in pregnant women receiving spinal anesthesia.

IMPLICATIONS: The heart rate variability statistic peak point correlation dimension (PD2) successfully predicted hypotension within 20 min of spinal anesthesia in all 22 pregnant women undergoing cesarean delivery. If further research supports this finding, real-time electrocardiograph analysis of peak PD2 might prevent adverse circulatory events in these women.

Departments of *Anesthesiology and †Cardiovascular Medicine and Surgery, Drexel University College of Medicine, Philadelphia, Pennsylvania

Accepted for publication July 16, 2004.

Address correspondence to Jay C. Horrow, MD, MSstat, Mail Stop 310, 215 N. 15th St., Philadelphia, PA 19102-1192. Address e-mail to Reprints will not be available.

Hypotension accompanying spinal anesthesia can occur because of hypovolemia or autonomic nervous system instability, a condition reflected by decreased heart rate variability (HRV) (1–3). This study evaluated whether point correlation dimension (PD2), a nonlinear metric of HRV, can predict hypotension in pregnant women receiving spinal anesthesia for cesarean delivery.

Back to Top | Article Outline


With IRB approval, 22 full-term pregnant nonlaboring women scheduled for cesarean delivery provided written, informed consent. No woman had atrial fibrillation, sinus node dysfunction, or hypertension (either current or treated).

The study participants had a baseline 10-min electrocardiogram (ECG) in left uterine displacement, as well as arterial blood pressure measurement by automated arm cuff (Hewlett-Packard, Cupertino, CA). They then received IV Isolyte® (Braun, Ontario, Canada) 1.5–2 L. Injection of intrathecal 0.75% bupivacaine, 12.5–15 mg (4), plus 150 μg of preservative-free morphine sulfate occurred with the patient sitting, after which she was turned supine with a 1-L saline bag placed under the right hip to achieve left uterine displacement. The dose of bupivacaine was determined by the individual clinician and was based on the expected duration of the cesarean delivery. An alcohol-soaked pad, touched to the skin 5 and 10 min after injection, was used to determine the anesthetic level. Patients received IV ephedrine to treat hypotension, the dose and timing of which was determined by the individual clinician. Standard noninvasive monitors measured vital signs; arterial blood pressure was measured every minute for 20 min. Systolic blood pressure ≤75% of baseline within 20 min of intrathecal injection was defined as hypotension (5).

DataQ hardware and software (Akron, OH) digitized and recorded the lead II ECG signal at 1000 Hz per channel. Subsequent offline data analysis proceeded in two ways. First, spectral analysis (6) provided the distribution of each component of the frequency spectrum by integration of the low-frequency (LF; 0.04–0.15 Hz) and high-frequency (HF; 0.15–0.40 Hz) components. Second, separately, the 10-min recording was divided into hundreds of small epochs for PD2 calculation (7). A log-log histogram of these PD2 values identifies the peak PD2 (pPD2), which is the mode of these values, for each subject.

Mean, sd, and median summarize the pPD2 values of each cohort. Unpaired unequal-variance Student’s t-tests were used to compare continuous variables; Fisher’s exact test was used to compare categorical data. To construct a test to predict hypotension by using pPD2, the median of baseline pPD2 values, an unbiased discriminant divided the cohort into LO and HI groups of 11 patients each. Then Fisher’s exact test was used to compare the frequencies of hypotension in these groups. Repeated-measures analysis of variance was used to analyze the effects of time (prespinal or postspinal) and group (LO or HI) on pPD2. All comparisons used two-tailed significance at P < 0.05.

Back to Top | Article Outline


Of the 22 study participants (aged 22–40 yr), 19 had had 2 or 3 prior pregnancies. Before anesthesia, R-R intervals [mean (sd)] were 694 ms (93.4 ms), corresponding to 88.1 bpm (12.0 bpm). The median pPD2 before anesthesia (3.90) formed 2 groups of 11 each (LO and HI). Groups LO and HI did not differ in demographic profile, anesthetic dose, initial arterial blood pressure, volume of IV prehydration, or highest dermatome level of anesthesia (Table 1), although patients in the LO group had more rapid baseline heart rates: 95 bpm (10 bpm) versus 81 bpm (9.7 bpm) (P = 0.00397; Table 1). Traditional HRV spectral analysis revealed no differences between groups in LF or HF spectral power (Table 2).

Table 1

Table 1

Table 2

Table 2

After anesthesia, pPD2 decreased (Table 3; P = 0.00034). No patient in the HI group developed hypotension (systolic range, 92–130 mm Hg), whereas all 11 LO group patients did (systolic range, 64–96 mm Hg; P = 0.0000028). Two patients from the LO group also had bradycardia (heart rate <60 bpm), but none in the HI group had bradycardia. As a result, study participants in the LO group received more ephedrine than those in the HI group (Table 1).

Table 3

Table 3

Back to Top | Article Outline


Hypotension, with or without bradycardia, occurs in one third of spinal anesthetics (8). This study has demonstrated that a baseline pPD2 of >3.90, an unbiased discriminant obtained from the median value, predicts hypotension after spinal anesthesia for cesarean delivery. If validated, this preliminary finding suggests that a real-time baseline pPD2 measurement could predict and possibly prevent postspinal hypotension.

Pregnant women receiving spinal anesthesia experience autonomic nervous system changes (9–11). The work of Hopf et al. (12) and Gratadour et al. (13) suggests three distinct patterns of response: 1) hypotension and tachycardia, from abolition of sympathetic tone below T4; 2) hypotension and bradycardia, due to parasympathetic tone increased relative to sympathetic tone; and 3) little or no change in hemodynamics. The ability to anticipate those pregnant women likely to experience hypotension would allow clinicians to intervene selectively and early.

HRV analysis can stratify patients with acute cardiac events (3,14). Nonlinear system analysis of HRV (15,16) has now replaced traditional time- and frequency-domain techniques (17,18). Nonlinear analysis measures the functional order and temporal unfolding of heart rate (19) by treating heart rate as an information-generating source. It detects high-risk, abrupt autonomic changes better than conventional techniques (3,20,21). Traditional spectral analysis requires stationarity of the R-R interval data, a condition valid only for short segments (22). Nonlinear methods, such as PD2, better quantify the chaotic dynamics of HRV.

The most important observation of this study is that pPD2 predicts postspinal hypotension. Although this pilot study did not test the pPD2 discriminant of 3.90 in a separate, independent population, its selection as the median prespinal pPD2 value adds credibility to the utility of pPD2. Nevertheless, the median value was not prespecified, and only independent confirmation of the predictive value of a pPD2 of 3.90 in another prospective independent cohort can validate its predictive value. Cohorts other than pregnant term women might have different discriminants than 3.90. Alternatively, 3.90 might be robust across many populations.

A much simpler hemodynamic variable, such as baseline heart rate or systolic blood pressure, or a change in arterial blood pressure with position might have predicted postspinal hypotension equally well (23). Table 4 compares the predictive abilities of measured hemodynamic variables in this trial. However, this investigation did not aim to compare various predictors or to find independent predictors—only to examine pPD2.

Table 4

Table 4

Participants in this study received 12.5–15 mg of hyperbaric bupivacaine in their spinal anesthetic. In a study comparing spinal bupivacaine 12 and 15 mg, DeSimone et al. (4) found that pregnant women who received the 15-mg dose had, on average, a level of sensory anesthesia 2.2 spinal segments higher than those in the 12 mg group. This study did not assess differences, if any, in the resulting hypotension. However, once the level of sensory block reaches T3-4, sympathetic block is complete. Therefore, we would not expect, nor did we see, a correlation between the dose and incidence of hypotension with the spinal bupivacaine dose range used in this study.

In conclusion, HRV analyzed via pPD2 shows promise as a novel predictor of postspinal hypotension. It stratified patients undergoing cesarean delivery into risk cohorts with 100% sensitivity and 100% specificity. Additional studies are required to validate a pPD2 of 3.9 as a risk discriminant in pregnant women and to identify pPD2 discriminants in other populations. In addition, real-time calculation of this measurement must be developed for this technology to have clinical utility.

Back to Top | Article Outline

Appendix 1: Nonlinear Dimension Analysis

In general, a system that is characterized by n independent variables can be thought of as “residing” in an n-dimensional space (24). The computed dimension would effectively represent the number of active degrees of freedom, a region in phase-space to which all deviating trajectories or perturbations will tend to converge (25). This phase-space “attractor” can be nonintegral, i.e., fractal. Hence, the R-R interval time series can be viewed as a projection line of a trajectory of a newly organized system that is confined by the dimensionality of the attractor. The system dimension can range from 0 to infinity; the lower the number, the simpler the dynamics. In the normal range of physiological response, a dimension of 10 (serving as an upper limit) would resemble white noise. In the intact innervated heart, the HRV dimension was measured to be ∼4, whereas in the Langendorff-perfused isolated heart, the dimension had a metronome-like (∼1) property (19). A decrease in the computed dimension would imply a loss in the number of active degrees of freedom and a decline in the complexity of the R-R series. PD2 offers the advantage of requiring a small data set for analysis of nonstationary signals (15) compared with the classic Grasberger-Procaccia determination of correlation dimension (D2). The point D2 estimate of the correlation dimension was developed by Skinner et al. (7). Like D2, each PD2 reference vector (i.e., the vector for a given R-R interval for a given epoch size of m RR intervals) remains fixed, whereas each of the j vectors (the vector from a given beat to subsequent nonneighboring beats) runs through the entire data series. However, the PD2, j vectors that contribute small log-r values, arises from subepochs with scaling characteristics similar to the surrounding the i vector. Previous studies of HRV have shown that the number of PD2 variables is 4–5 in the general population (19).

Back to Top | Article Outline


1. Costa M, Goldberger AL, Peng CK. Multiscale entropy analysis of complex physiologic time series. Phys Rev Lett 2002;89:068102.
2. Liu S, Paul GE, Carpenter RL, et al. Prolonged PR interval is a risk factor for bradycardia during spinal anesthesia. Reg Anesth 1995;20:41–4.
3. Vybiral T, Glaeser DH, Goldberger AL, et al. Conventional heart rate variability analysis of ambulatory electrocardiographic recordings fails to predict imminent ventricular fibrillation. J Am Coll Cardiol 1993;22:557–65.
4. DeSimone CA, Leighton BL, Norris MC. Spinal anesthesia for cesarean delivery: a comparison of two doses of hyperbaric bupivacaine. Reg Anesth 1995;20:90–4.
5. Hughes SC, Levinson G, Rosen MA. Anesthesia for cesarean section. In: Hughes SC, Levinson G, Rosen MA, eds. Anesthesia for obstetrics. 4th ed. Philadelphia: Lippincott Williams & Wilkins, 2002:204.
6. Task Force of the European Society of Cardiology and the North American Society of Pacing and Electrophysiology. Heart rate variability: standards of measurement, physiological interpretation and clinical use. Circulation 1996;93:1043–65.
7. Skinner JE, Molnar M, Tomberg C. The point correlation dimension: performance with nonstationary surrogate data and noise. Integr Physiol Behav Sci 1994;29:217–34.
8. Carpenter RL, Caplan RA, Brown DL, et al. Incidence and risk factors for side effects of spinal anesthesia. Anesthesiology 1992;76:906–16.
9. Eisenach JC, Tuttle R, Stein A. Is ST segment depression of the electrocardiogram during cesarean section merely due to cardiac sympathetic block? Anesth Analg 1994;78:287–92.
10. Landry DP, Bennett FM, Oriol NE. Analysis of heart rate dynamics as a measure of autonomic tone in obstetrical patients undergoing epidural or spinal anesthesia. Reg Anesth 1994;19:189–95.
11. Baron JF, Coriat P, Mundler O, et al. Left ventricular global and regional function during lumbar epidural anesthesia in patients with and without angina pectoris: influence of volume loading. Anesthesiology 1987;66:621–7.
12. Hopf HB, Skyschally A, Heusch G, Peters J. Low-frequency spectral power of heart rate variability is not a specific marker of cardiac sympathetic modulation. Anesthesiology 1995;82:609–19.
13. Gratadour P, Viale JP, Parlow J, et al. Sympathovagal effects of spinal anesthesia assessed by the spontaneous cardiac baroreflex. Anesthesiology 1997;87:1359–67.
14. Malik M, Camm AJ. Heart rate variability. Clin Cardiol 1990;13:570–6.
15. Elbert T, Ray WJ, Kowalik ZJ, et al. Chaos and physiology: deterministic chaos in excitable cell assemblies. Physiol Rev 1994;74:1–47.
16. Mansier P, Clairambault J, Charlotte N, et al. Linear and non-linear analyses of heart rate variability: a minireview. Cardiovasc Res 1996;31:371–9.
17. Smith ML, Ellenbogen KA, Eckberg DL, et al. Subnormal parasympathetic activity after cardiac transplantation. Am J Cardiol 1990;66:1243–6.
18. Yang MW, Kuo TB, Lin SM, Chan SH. Power spectral analysis of arterial blood pressure after spinal anesthesia. Acta Anaesthesiol Sin 1998;36:3–10.
19. Kresh JY, Izrailtyan I. Evolution in functional complexity of heart rate dynamics: a measure of cardiac allograft adaptability. Am J Physiol 1998;275:R720–7.
20. Goldberger AL, West BJ. Applications of nonlinear dynamics to clinical cardiology. Ann N Y Acad Sci 1987;504:195–213.
21. Pollard JB. Common mechanisms and strategies for prevention and treatment of cardiac arrest during epidural anesthesia. J Clin Anesth 2002;14:52–6.
22. Braun C, Kowallik P, Freking A, et al. Demonstration of nonlinear components in heart rate variability of healthy persons. Am J Physiol 1998;275:H1577–84.
23. Frolich MA, Caton D. Baseline heart rate may predict hypotension after spinal anesthesia in prehydrated obstetrical patients. Can J Anaesth 2002;49:185–94.
24. Noble D, Bett G. Reconstructing the heart: a challenge for integrative physiology. Cardiovasc Res 1993;27:1701–12.
25. Hoyer D, Schmidt K, Zwiener U, Bauer R. Characterization of complex heart rate dynamics and their pharmacological disorders by non-linear prediction and special data transformations. Cardiovasc Res 1996;31:434–40.
© 2004 International Anesthesia Research Society