| JOURNAL HOME | CME HOME | THIS MONTH | PAST ISSUES | ETOC | COLLECTIONS |
| AUTHORS | REVIEWERS | EDITORIAL BOARD | FEEDBACK | RSS | HELP |
| ||||||||||||||
| ||||||||||||||
|
|
|
|
||||||||||||
|
||||||||||||||
Department of Anaesthesia and Intensive Care, The Chinese University of Hong Kong, Prince of Wales Hospital, Shatin, New Territories, Hong Kong, China
Address correspondence to Anna Lee, MPH, PhD, Department of Anaesthesia and Intensive Care, The Chinese University of Hong Kong, Prince of Wales Hospital, Shatin, NT, Hong Kong, China. Address e-mail to annalee{at}cuhk.edu.hk
 |
Abstract |
|---|
 Top Abstract IntroductionMethodsResultsDiscussionReferences |
|---|
IMPLICATIONS: The authors performed a systematic review of dose-response studies of IV bolus ephedrine for preventing hypotension during spinal anesthesia for cesarean delivery. Prophylactic ephedrine cannot be recommended. The efficacy is poor at smaller doses, whereas at larger doses, the likelihood of causing hypertension is actually more than that of preventing hypotension.
|
Introduction |
|---|
|
TopAbstractIntroductionMethodsResultsDiscussionReferences |
|---|
|
Methods |
|---|
|
TopAbstractIntroductionMethodsResultsDiscussionReferences |
|---|
We searched MEDLINE (from 1966 through October 2002), EMBASE (from 1980 through October 2002), and the Cochrane Central Register of Controlled Trials. The Medical Subject Heading and text terms used in the searches were hypotension, cesarean section, spinal anesthesia, and dose-response relationship. The terms dose or dosage were also used. Finally, references from relevant articles were reviewed to identify additional studies. Although there was no language restriction, all studies included in this systematic review were published in English.
The selection of studies for inclusion in the systematic review was performed independently by the reviewers (AL and WDNK) after using the search strategy described previously. Data were abstracted independently by AL and WDNK by using a standardized data collection form. There was no attempt to blind the reviewers (AL and WDNK) to the authors or results of the relevant trials. Details of anesthetic technique, study population, prehydration, and uterine displacement and definitions of maternal hypotension and hypertension were collected. Where appropriate, the primary author was contacted for clarification of data. Discrepancies were resolved by discussion, or advice was sought from a third party (TG).
The quality of the RCTs was assessed independently. The level of allocation concealment, defined as the process used to prevent the foreknowledge of group assignment in an RCT, was graded as adequate, unclear, or inadequate, as previously described (5). The level of blinding (double, single, or none), losses to follow-up, and whether the authors performed a sample-size calculation before trial commencement were recorded.
For the dose-response meta-analysis (6), the correlation between risk estimates for separate dose levels and the reference group (0 mg) was taken into account. The summary relative risk (RR) was the pooled coefficient ß in the linear logistic regression model lnRR = ßx, where x is the difference in ephedrine dose between each category and the reference category. The individual slopes of each study were combined by weighted average by using the inverse of the sum of the within-trial variance and the residual between-trial variance as weights (6). We used a weighted dose-response model with a 0 intercept (7) because it implied that the risk among parturients taking very small doses would be the same as the risk among those who did not receive ephedrine. A random-effects model was used because it assumes that the studies are estimating different (underlying) effect sizes and includes two sources of variation: the between- and within-study variance (8). When the between-study variance was estimated to be 0, the random-effects model was reduced to the fixed-effects model. We examined residuals to ensure that the normality assumption held. Separate dose-response meta-analysis was estimated for RCTs and cohort studies. To assess whether the timing of the administration of prophylactic ephedrine affected the dose-response relationship for hypotension, subgroup analyses were performed and compared by using a test of interaction (9). Heterogeneity was tested and considered significant if P < 0.10. Analysis was performed by using EPIMETA statistical software Version 1.1 (Centers for Disease Control, Atlanta, GA).
Sensitivity analyses were performed to examine the effects of trial quality and gray literature on estimates of treatment effect. Gray literature is defined as abstracts, unpublished studies, conference proceedings, graduate theses, book chapters, and company reports (10). Exclusion of gray literature from meta-analysis may result in an overestimate of an intervention effect by an average of 12% (10).
A random-effects metaregression (9) was used to examine the association between the change in umbilical arterial pH and the log dosage of ephedrine after a weighted mean difference (8) was calculated. The weighted mean difference was the mean difference in umbilical arterial pH level between ephedrine dose and control (0 mg). The log dosage was used because most physiologic responses are related in this way (11) and because the doses used in most dose-response studies generally occurred in a logarithmic progression. The metaregression was performed with Stata statistical software (Version 7.0; Stata Corp., College Station, TX).
When dose-response meta-analysis was not possible, a Mantel-extension 
2 test for overall trend (12) was used to examine the dose-response relationship in individual studies with dichotomous outcomes. The 95% confidence intervals (CI) were calculated around the summary RR.
A funnel plot (plot of treatment effect against trial precision) was used to detect bias in the dose-response meta-analysis of prophylactic ephedrine trials on preventing hypotension. The funnel plot will be skewed and asymmetrical in the presence of bias, which will usually lead to an overestimate of the treatment effect. The degree of asymmetry was measured by Egger et al.’s method (9) with Stata statistical software.
To judge whether therapy was worthwhile for an individual, the absolute magnitude of benefit and harm was estimated by calculating the number-needed-to-treat (NNT) and the number-needed-to-harm (NNH) (9) when a significant dose-response relationship was found. Because an NNT derived from meta-analysis can be sensitive to factors that change the baseline risk, a more useful NNT was estimated by applying the pooled RR to a relevant baseline risk (9). We assumed that the baseline risks of maternal hypotension and hypertension were 80% (13) and 11% (1), respectively. To determine the threshold at which benefit equals harm, we graphed the ephedrine dose (x axis) and the reciprocals of NNT (y axis) and NNH (y axis) (14). The reciprocals of NNT and NNH are equivalent to absolute risk difference (9).
|
Results |
|---|
|
TopAbstractIntroductionMethodsResultsDiscussionReferences |
|---|
|
The definition of maternal hypotension varied between studies. Hypotension was defined as systolic arterial blood pressure less than 80 mm Hg in one RCT.1 In the original article by Loughrey et al. (16), hypotension was defined as systolic arterial blood pressure that decreased to <90 mm Hg or a reduction more than 30% of baseline pressure. However, we obtained unpublished data for hypotension when it was defined as systolic arterial blood pressure less than 100 mm Hg and a reduction of 20% of baseline pressure. Therefore, two RCTs (2,16) defined hypotension as systolic arterial blood pressure less than 100 mm Hg and a reduction more than 20% of baseline pressure. Carvalho et al. (15) defined hypotension as a reduction more than 20% of baseline systolic blood pressure. In another study, hypotension was defined as a decrease of 
30% less than baseline pressure or systolic blood pressure <100 mm Hg (3).
The meta-analysis of RCTs1 (2,15,16) showed that there was a significant dose-response relationship. The overall slope from the fixed-effects model was -0.0128 (95% CI, -0.0213 to -0.0044) (Fig. 1). There was homogeneity of slopes (23 = 3.05; P = 0.38). There was also a significantly decreased risk of hypotension with increasing ephedrine dose in a cohort study (2 = 9.67; P = 0.002) (3). A sensitivity analysis, excluding an abstract publication,1 showed a similar dose-response relationship. The overall slope from the fixed-effects model was -0.0123 (95% CI, -0.0208 to -0.0038). There was homogeneity of slopes (22 = 1.62; P = 0.44). There was no evidence of bias in the dose-response meta-analysis of the four RCTs1 (2,15,16), as shown by the symmetry in the funnel plot (intercept, -0.42; 95% CI, -2.39 to 1.55; P = 0.64; Fig. 2).
|
|
Hypertension was defined as an increase in systolic arterial blood pressure more than baseline by more than 20% (2,15,16) or 30% (3). There was a significantly increased risk of maternal hypertension with increasing doses of ephedrine in the three RCTs pooled (2,15,16). The overall slope of the fixed-effects model was 0.0563 (95% CI, 0.0235–0.0892) (Fig. 3). There was homogeneity of slopes (22 = 1.32; P = 0.52). However, there was no significantly increased risk of hypertension with increasing ephedrine dose in a cohort study (2 = 0.23; P = 0.63) (3).
|
2 = 2.23; P = 0.14) or vomiting (2 = 0.63; P = 0.43) with increasing doses of ephedrine (15). In the other two RCTs (2,16), the incidence of nausea, vomiting, or both reported by patients was collected. When these data were pooled, the fixed-effects model showed no significant dose-response relationship (slope = -0.0002; 95% CI, -0.0189 to 0.0186), and the slopes were homogeneous (21 = 0.001; P = 0.97).
There was a significant dose-response relationship in the change in umbilical arterial pH from baseline (2,15,16) (Fig. 4). The estimated slope was -0.03 (95% CI, -0.05 to 0.00) (P = 0.02), and the intercept was 0.06 (95% CI, 0.01–0.12). Fetal acidosis was defined as umbilical arterial pH <7.2 in all studies. There was no evidence with the random-effects model of an increased risk of fetal acidosis with increasing doses of ephedrine (Fig. 5). The overall slope was 0.0078 (95% CI, -0.1710 to 0.1864). There was homogeneity of slopes (22 = 3.81; P = 0.15). There was no association between dose and risk of fetal acidosis in a cohort study (2 = 3.06; P = 0.08) (3).
|
|
2 = 0.88; P = 0.35). Because no significant dose-response relationship was found with fetal acidosis, the absolute risk differences for maternal hypotension and hypertension were plotted (Fig. 6) against an increasing dose of ephedrine. The threshold at which the potential for benefit equaled the risk of harm was at 14 mg. At this dose, the NNT was 7.6 (95% CI, 4.8–21.1) and the NNH was 7.6 (95% CI, 3.7–23.4). Therefore, the largest dose for preventing hypotension at which a small benefit outweighs the risk of reactive hypertension is 12 mg. At 12 mg, the NNT was 8.8 (95% CI, 5.5–24.5) and the NNH was 9.4 (95% CI, 4.7–27.9). At 30 mg, the ratio of harm to benefit was almost 2:1.
|
|
Discussion |
|---|
|
TopAbstractIntroductionMethodsResultsDiscussionReferences |
|---|
We used several outcome measures to examine the efficacy of prophylactic IV ephedrine. In general, the dose-response relationship from the four RCTs1 (2,15,16) was similar to the findings of the cohort study (3) for most outcomes considered, despite differences in the study designs and the reference groups used. Therefore, this consistency in results gives further evidence for the presence of dose-response relationships for the specific outcomes examined.
Significant dose-response relationships were found for hypotension, hypertension, and umbilical arterial pH. The dose-response relationship for hypotension was robust; there were no changes in slope estimates by trial quality, timing of ephedrine administration, or exclusion of gray literature. On the basis of the assumption that the baseline risks of hypotension and reactive hypertension were 80% (13) and 11% (1), respectively, the dose at which the likelihood of benefit marginally outweighed the risk of harm was 12 mg. At this dose, there was minimal change in the umbilical artery pH (0.00; 95% CI, -0.06 to 0.06). However, the threshold at which benefit and harm were equal was based on the assumption that baseline risks of hypotension and hypertension were of equal importance. This threshold would increase or decrease depending on how individual anesthesiologists both define and value benefit (prevention of hypotension) and harm (reactive hypertension). Accordingly, the optimal dose of ephedrine would be adjusted up or down.
The association estimated with the dose-response meta-analysis was stronger for hypertension than for hypotension. These findings suggest that the use of larger doses of ephedrine does not completely eliminate hypotension but causes reactive hypertension and a minor decrease in umbilical arterial pH.
We found no evidence of a dose-response relationship for nausea or vomiting, fetal acidosis, or Apgar scores. Although theoretically it might be expected that increasing doses of ephedrine would be associated with a decreased incidence of fetal acidosis by reducing the incidence of maternal hypotension and consequent reduction in uteroplacental perfusion, this is unlikely given the inverse relationship between ephedrine dose and umbilical arterial pH in the data. Fetal tachycardia appeared to be dose-related, because the mean total ephedrine dose was larger in patients with abnormal cardiotocography tracings compared with patients with normal tracings (2).
The significant nonzero intercept from the metaregression of change in umbilical arterial pH and log dosage of ephedrine suggests the presence of bias or of a treatment effect that was not mediated via ephedrine prophylaxis (8). Other factors, such as uterine incision to delivery time, maximum decrease in systolic arterial blood pressure (18), and total ephedrine dose, also affect umbilical arterial pH. Prolonged uterine incision to delivery times were associated with fetal acidosis (19,20). Marked maternal hypotension may decrease uteroplacental perfusion, reducing the gas exchange across the placenta and resulting first in acute fetal respiratory acidosis and then in fetal metabolic acidosis, both of which are reflected by the umbilical artery pH (21). Uterine incision to delivery time and maximum decrease in systolic arterial blood pressure were factors not considered in our metaregression because the data were not published in the studies included. It was not possible to assess the association between the total ephedrine dose used and changes in umbilical arterial pH without access to individual patient data.
Our meta-analysis focused on the use of prophylactic ephedrine. There is widespread use of ephedrine for the prevention and treatment of hypotension during spinal anesthesia (22). Historically, ephedrine has been used most often to prevent or treat maternal hypotension during spinal anesthesia for cesarean delivery. This practice is based largely on observations in pregnant sheep that showed that ephedrine was more effective for increasing arterial blood pressure with better preservation of uteroplacental blood flow compared with other vasopressors (23,24). This was explained by ephedrine’s predominant ß-effect, which caused an increase in arterial blood pressure by increasing cardiac output rather than by vasoconstriction. However, more recent studies in humans have shown that ephedrine is associated with lower values for umbilical arterial pH and a more frequent incidence of fetal acidosis compared with 
-agonists, such as phenylephrine and metaraminol (18,25–27). Thus, there is controversy as to what the vasopressor of choice is in obstetric patients.
The applicability of the results from this systematic review is limited to healthy, nonlaboring women with term fetuses. We identified 5 studies involving nearly 400 patients with ephedrine doses ranging from 5 to 30 mg. The calculated dose-response slopes were based on up to four RCTs. However, only one (2) used doses larger than 20 mg. Therefore, the results at the larger end of the dose-response relation must be considered with this limitation in mind. Another limitation was that, because there was no standard definition of hypotension or hypertension in this systematic review, we chose to rely on the definitions given by the authors of each trial. Despite our attempt to standardize the definition of hypotension in the trial of Loughrey et al. (16), there is some degree of clinical heterogeneity among trials. We acknowledge that the incidence of hypotension varied greatly among studies, and this may have been due to fluid administration, spinal local anesthetic dose, and other factors. However, there were insufficient trials in this systematic review to examine these issues in a meaningful way.
The quality of the trials included in this systematic review was fair, with two trials that had both adequate allocation concealment and double-blinding. Compared with trials that have adequate allocation concealment or are double-blinded, those with unclear allocation concealment or that are not double-blinded are associated with a larger treatment effect (41% and 17%, respectively) (5). However, our sensitivity analyses showed that the quality of the trials did not influence the overall dose-response relationship. Overall, the dose-response results from this systematic review appear robust, because there was no evidence of bias, as highlighted by the symmetry in the funnel plot. Sources of asymmetry in funnel plots can be due to selection bias (publication bias, English language bias, citation bias, and multiple publication bias), true heterogeneity, data irregularities, choice of effect measure, and chance (9).
Another potential limitation of our systematic review is the indirect evaluation of the dose-response effect of ephedrine on umbilical arterial pH by the use of metaregression. We compared each dose group with its placebo group and compared RR estimates across studies. Therefore, although the original studies may be RCTs, the metaregression was across trials, which did not have the benefit of randomization. Ideally, one would pool individual patient data from studies to overcome this limitation and explore other covariates that may affect the outcome in the process. It is unclear what the potential gains are, in terms of precision and bias, for pooled results with individual patient data over summary data (8). Nevertheless, we believe that meta-analysis of dose-response RCTs provides more precise estimates and has an important role in causal inference, especially when there is clinical and statistical homogeneity in the meta-analyses.
In conclusion, significant dose-response relationships were found for hypotension, hypertension, and umbilical arterial pH. These results suggest that the use of larger doses of ephedrine (>14 mg) does not eliminate hypotension but causes reactive hypertension and a minor decrease in umbilical arterial pH. Because the absolute risk reduction for preventing hypotension is small, prophylactic IV ephedrine cannot be recommended in nonlaboring women undergoing spinal anesthesia for cesarean delivery.
|
Acknowledgments |
|---|
|
Footnotes |
|---|
|
References |
|---|
|
TopAbstractIntroductionMethodsResultsDiscussionReferences |
|---|
This article has been cited by other articles:
|
|
 |
|
  P. J. Balestrieri Combined Spinal-Epidural Anesthesia for Cesarean Delivery: The Dose-Dependent Effects of Hyperbaric Bupivacaine on Maternal Hemodynamics Anesth. Analg., February 1, 2007; 104(2): 453 - 453. [Full Text] [PDF]
|
|
|
|
|
|
M. Van de Velde, D. Van Schoubroeck, J. Jani, A. Teunkens, C. Missant, and J. Deprest Combined Spinal-Epidural Anesthesia for Cesarean Delivery: The Dose-Dependent Effects of Hyperbaric Bupivacaine on Maternal Hemodynamics Anesth. Analg., February 1, 2007; 104(2): 453 - 454. [Full Text] [PDF]
|
|
|
|
|
|
A. G. M. Aya, N. Vialles, and J. Ripart Is Ephedrine the Best Vasopressor for Treating Spinal Anesthesia-Induced Hypotension in Patients with Pre-Eclampsia? Anesth. Analg., December 1, 2006; 103(6): 1584 - 1584. [Full Text] [PDF]
|
|
|
|
|
|
M. Van de Velde, D. Van Schoubroeck, J. Jani, A. Teunkens, C. Missant, and J. Deprest Combined spinal-epidural anesthesia for cesarean delivery: dose-dependent effects of hyperbaric bupivacaine on maternal hemodynamics. Anesth. Analg., July 1, 2006; 103(1): 187 - 190. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 T. Erkinaro, T. Kavasmaa, M. Pakkila, G. Acharya, K. Makikallio, S. Alahuhta, and J. Rasanen Ephedrine and phenylephrine for the treatment of maternal hypotension in a chronic sheep model of increased placental vascular resistance Br. J. Anaesth., February 1, 2006; 96(2): 231 - 237. [Abstract] [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
|
