Effect of Epidural vs Parenteral Opioid Analgesia on the Progress of Labor: Title and subTitle BreakA Meta-analysis FREE

EPIDURAL ANALGESIA effectively alleviates labor pain; however, controversy exists about the effect of epidural analgesia on labor outcome. In particular, several observational studies^{1 - 3} and 1 randomized controlled trial⁴ have shown an increased incidence of cesarean delivery in women who receive epidural labor analgesia compared with women who receive parenteral opioids. These results have led some health care plans to deny reimbursement for epidural labor analgesia in an attempt to minimize costs.⁵ Other randomized controlled trials, however, report no increase in the cesarean delivery rate in parturients who receive epidural analgesia.^{6 - 7} If epidural analgesia is merely associated with dysfunctional labor,⁸ restricting access to epidural analgesia unnecessarily denies parturients appropriate pain relief. Alternatively, if epidural analgesia can increase cesarean delivery rates, laboring women should be informed of this risk.

Parenteral opioids are the most frequently prescribed alternative to epidural labor analgesia. However, in routinely used doses, parenteral opioids have little effect on labor pain.^{4 ,9} Severe, unrelieved labor pain causes patient dissatisfaction and is associated with postpartum depression and posttraumatic stress disorder.^{10 - 11}

Several well-executed randomized controlled trials have been published since the most recent systematic reviews of the effect of epidural analgesia on labor outcome.^{12 - 13} The purpose of this meta-analysis is to combine data from the available randomized controlled trials to determine the effects of epidural vs parenteral opioid analgesia on the cesarean delivery rate. In addition, we examined and combined, when possible, other maternal and neonatal outcomes.

Randomized controlled trials comparing epidural anesthesia with parenteral opioids for labor pain relief were analyzed. Articles were retrieved from MEDLINE (January 1966-January 1998) using the following search terms as MeSH headings and text words: analgesia, obstetrical; analgesia, epidural; cesarean delivery; analgesics, opioid. Additional references were sought in the Cochrane Library (September 1997), the personal files of the authors, and the reference lists of previously published reviews. In addition, the tables of contents of the International Journal of Obstetrical Anesthesia (inception to December 1997), which is not abstracted by MEDLINE, were searched by hand. Finally, published abstracts (1993-1997) from the following meetings were reviewed: Canadian Anaesthetists Society, American Society of Anesthesiologists, Society of Obstetric Anesthesia and Perinatology, and Society of Perinatal Obstetricians. Thus, we attempted to use all available data, published and unpublished, in an effort to reduce publication bias.¹⁴

Each trial was scored independently by 2 of us (S.H.H., A.O.) using a quality index developed by Jadad et al.¹⁵ This scale has a maximum score of 5 points, with 0 to 2 points assigned for the quality of the methods of randomization and blinding (0, inappropriate; 1, not described; 2, appropriate method) and 1 point given if the study described the outcome of all enrolled subjects. We rereviewed the articles and arrived at a final score by consensus when initial scores differed.

The primary outcome was the incidence of cesarean delivery for any indication. Secondary maternal outcomes included cesarean delivery rate for dystocia, total instrumented delivery rate, instrumented delivery rate for dystocia, use of oxytocin after initiation of analgesia, the length of the first and second stages of labor, fever (temperature >38.0°C), hypotension, nausea, pain during the first and second stages of labor, and satisfaction. Perinatal outcomes included the incidence of 1- and 5-minute Apgar scores less than 7, umbilical artery pH of less than 7.15 or 7.20, presence of meconium at delivery, fetal heart rate abnormalities during labor, and early (2-4 hours) and late (24 hours) infant neuroadaptive capacity scores.¹⁶

To avoid transcription errors, data were recorded independently by 2 of us (S.H.H., B.L.L.), with any discrepancies resolved by reinspection of the original articles. The data were then entered into the statistical program (by S.H.H.) and rechecked (by B.L.L.).

Meta-analytic techniques (Metaview software, Revman 3.0, Cochrane Library, Oxford, England) were used to combine the results of the randomized controlled trials using a random effects model. Odd ratios (ORs) and 95% confidence intervals (CIs) were calculated for dichotomous variables using the methods of DerSimonian and Laird.¹⁷ We tested for heterogeneity using the Breslow-Day method.¹⁸ The weighted mean difference (WMD) (random-effects model) and 95% CIs were calculated for continuous variables. A statistical difference between groups was considered to occur if the pooled 95% CI did not include 1 for the OR or 0 for the WMD. An OR of less than 1 or a negative WMD favored epidural over control.

Nulliparous and multiparous patients were analyzed as subgroups of the total.

We performed 3 sensitivity analyses on the primary outcome (total cesarean delivery rate): high-quality trials only (Jadad score ≥3), peer-reviewed articles only, and trials grouping patients by intent to treat only.

We searched MEDLINE on February 1, 1998. Our search identified 155 published articles that discussed labor analgesia. We eliminated 147 articles for the following reasons: 72 were retrospective studies, 29 were randomized studies that did not compare epidural with parenteral opioid analgesia, 17 were reviews, and 29 were prospective cohort studies. Two trials, presented only in abstract form, met inclusion criteria and were included. We obtained additional data from the authors of these abstracts (Ann L. Clark, MD, written communication, April 3, 1998; Holly Muir, MD, FRCPC, written communication, May 5, 1998) and from 1 of the published trials (Shiv Sharma, MD, FRCA, written communication, November 3, 1997). All retrieved studies were published in English between January 1, 1980, and December 31, 1997. All studies enrolled only healthy women with uncomplicated pregnancies. One reference included 2 studies,¹⁹ and the results of 1 study were split into 2 publications.^{20 - 21} Thus, 10 studies met the inclusion criteria, describing a total of 1614 nulliparous and 755 multiparous randomized patients.^{4 ,6 - 7 ,19 - 25} All trials reported cesarean delivery data; in 3 small studies, no cesarean deliveries occurred.^{19 ,23}

All but 1 of the trials reported outcomes with patients grouped by the treatment to which they were originally randomized (intent-to-treat analysis). We used the intent-to-treat data from these 9 studies. One study reported outcomes only on patients who remained in the groups to which they were randomized ("protocol compliant").²² We analyzed the protocol-compliant data from this study. Of note, in the 2 studies that reported both intent-to-treat and protocol-compliant analyses, the results are similar in magnitude and direction to the 2 analytic methods (Ann L. Clark, MD, written communication, April 3, 1998).⁶

Data on patient parity, countries of origin, and quality scores are shown in Table 1. The caregivers were aware of group assignment in all studies; however, in 1 study a perinatologist unaware of group assignment was consulted before cesarean delivery was performed.⁷ Nine studies enrolled only women with spontaneous labor; 1 study also included patients undergoing labor induction.^{20 - 21} Seven studies included only nulliparous patients, and 3 trials included both nulliparous and multiparous patients.^{6 ,20 - 22}

Criteria for operative intervention and protocols for labor management and labor analgesia are shown in Table 2. Treatment crossovers occurred in 8 of the studies. In 7 trials, some patients in the parenteral opioid group found their analgesia inadequate and also received epidural analgesia. No analgesia was received by some patients in the epidural group in 5 studies and by some parenteral opioid patients in 1 study. In 2 trials, a few patients randomized to receive epidural analgesia received parenteral opioids instead.

We found statistical evidence of heterogeneity among the trials (P<.05) for the primary outcome of total cesarean delivery rate. Much of the heterogeneity is attributable to 1 of the smaller trials, which showed a marked cesarean delivery rate increase in the epidural group.⁴

The total rate of cesarean delivery did not differ significantly between patients receiving epidural analgesia (8.2% [97/1183]) vs parenteral opioids (5.6% [67/1186]) for labor (OR, 1.50; 95% CI, 0.81-2.76) (Figure 1). There was no difference between treatment groups in the rate of cesarean delivery for dystocia (OR, 1.63; 95% CI, 0.79-3.36). Data on 1025 nulliparous patients were available for analysis (Shiv Sharma, MD, FRCA, written communication, November 3, 1997).^{6 - 7 ,19 ,22 - 25} There was no statistically significant difference in the overall cesarean delivery rate for nulliparous women: 44 (8.5%) of 516 in the epidural group vs 39 (7.7%) of 509 in the parenteral opioid group (OR, 1.28; 95% CI, 0.55-2.93). Data from 364 multiparous patients were available for analysis.^{6 ,19} The cesarean delivery rate was 4 (2.2%) of 178 in the epidural group vs 5 (2.7%) of 186 in the parenteral opioid group (OR, 0.83; 95% CI, 0.22-3.15).

There were 2001 patients in published studies; the cesarean delivery rate was 79 (7.9%) of 999 in the epidural group and 43 (4.3%) of 1002 in the parenteral opioid group (OR, 1.90; 95% CI, 0.93-3.86). There were 1317 patients in studies with Jadad quality of 3 or better; the cesarean delivery rate was 55 (8.2%) of 668 in the epidural group and 48 (7.2%) of 669 in the parenteral opioid group (OR, 1.33; 95% CI, 0.63-2.81). There were 1500 patients in studies grouping patients by intent to treat; the cesarean delivery rate was 58 (7.7%) of 751 in the epidural group and 50 (6.7%) of 749 in the parenteral opioid group (OR, 1.27; 95% CI, 0.66-2.46).

The first and second stages of labor were longer in patients who had epidural analgesia compared with those who did not (Table 3). Hypotension, fever, and oxytocin use after analgesia are more frequent in patients receiving epidural analgesia. While the total incidence of instrumented delivery was significantly higher in the epidural group, there was no difference in the incidence of instrumented delivery for dystocia. Unfortunately, only 2 studies, with a total of 200 patients, reported instrumented delivery for dystocia.^{7 ,21}

Opioid analgesia provided poor pain relief in both the first and second stages of labor. Patients receiving epidural analgesia had significantly lower visual analog pain scores in both the first and second stages of labor. Similarly, fewer patients were dissatisfied with the analgesic method when they received epidural anesthesia (Table 3).

Fetal and neonatal outcomes are shown in Table 4. There were no differences in the incidence of fetal distress or the intrapartum passage of meconium between the 2 groups. Significantly fewer infants were born with 1- and 5-minute Apgar scores of less than 7 in the epidural group compared with the parenteral opioid group. In addition, naloxone was used less often in the newborns of patients receiving epidural analgesia. There was no difference between groups in either the early or 24-hour measurement of the neuroadaptive capacity score. However, only 87 infants were tested (Holly Muir, MD, FRCPC, written communication, May 5, 1998).^{4 ,23} A low umbilical artery pH (<7.15 or 7.20) was recorded less commonly among neonates born after epidural analgesia than after parenteral opioids. Only 4 infants (2 from each group) had severe asphyxia, indicated by an umbilical artery pH of less than 6.99.²⁶ There were no reports of serious neonatal complications related to either analgesic method.

Epidural labor analgesia was not associated with an increased incidence of cesarean delivery in the articles we analyzed. Advantages of epidural labor analgesia included improved maternal pain relief and satisfaction, less neonatal naloxone use, higher umbilical pH values, and fewer 1- and 5-minute Apgar scores of less than 7. Disadvantages of epidural analgesia included prolongation of the first and second stages of labor and increased incidences of oxytocin use, instrumented vaginal delivery, maternal fever, and maternal hypotension.

Our results differ from those of 2 previous meta-analyses, which reviewed only 2¹² and 4¹³ of the 10 randomized clinical trials available to us. The quality of clinical trials has improved with time; all trials after 1995 reported outcomes with patients grouped by intent to treat.

Grouping patients by intent to treat is vital, for parturients choosing epidural analgesia differ demographically from patients choosing parenteral opioid analgesia. Patients requesting epidural analgesia are more likely to be nulliparous, are admitted to the hospital with less dilated cervices and higher fetal head stations, have slower initial rates of cervical dilation, bear heavier neonates, and need oxytocin augmentation more frequently than patients requesting parenteral opioids.^{1 ,3} All of these factors independently predict the need for dystocia cesarean delivery.^{27 - 29}

Observational studies show that sudden large changes in epidural utilization are not associated with alterations in hospital-wide cesarean delivery rates. This has been demonstrated in urban³⁰ and rural^{31 - 32} hospitals, in academic³⁰ and community^{31 - 32} practices, and with military,³³ lower-class,^{30 ,32} and middle-class³¹ patient populations. Despite the stability of the hospitals' cesarean delivery rates, if one examines the patient subgroups with free access to epidural analgesia, the dystocia cesarean delivery rate is 2 to 8 times higher among patients who choose epidural analgesia.^{30 - 31 ,33} Factors leading to dystocia increase labor pain and epidural utilization.^{3 ,22} Therefore, studies not grouping patients by intent to treat will be biased toward an increased cesarean delivery rate among patients requesting epidural analgesia.

The intergroup difference in cesarean delivery rates in our analysis is 2.5%. This difference is small compared with the above-mentioned variables and other demographic and physician factors that affect mode of delivery. Women older than 35 years have 6 times the cesarean delivery risk of women younger than 20 years.²⁷ Patients who are short or obese have high cesarean delivery rates.^{3 ,27} Oxytocin use is associated with twice the cesarean delivery rate of unstimulated labors.²⁷ Physicians practicing in the same setting may have primary cesarean delivery rates that vary by 22%.³⁴

Epidural analgesia provided clearly superior pain relief. There is no study in which patient comfort in the opioid group equaled or surpassed the comfort of patients in the epidural group. Intermittent nurse-administered opioids provided no measurable analgesia, as assessed by before-and-after visual analog pain scores.^{4 ,22} Patient-controlled administration of intravenous opioids provides some comfort; however, patients receiving epidural analgesia were still more comfortable than patients receiving patient-controlled analgesia with opioids.^{6 ,25}

Inadequate labor pain relief may leave lasting psychic scars. Pain and physical injury increase the incidence of posttraumatic stress disorder.³⁵ Clinically significant postpartum depression is more frequent in patients who have had difficult or painful deliveries.¹⁰ Thirty percent of women demanding elective cesarean deliveries that are not medically indicated do so because of excruciating, unrelieved pain during a previous birth.³⁶ Posttraumatic stress disorder related to the recent delivery was diagnosed in 1.7% of a sample of 1640 postpartum women.¹¹ The popular belief that childbirth pain lacks consequences is incorrect.

The neonates of mothers receiving epidural analgesia were more alert, had a lower incidence of mild acidosis, and required naloxone less frequently than the neonates of mothers receiving parenteral opioids. None of the studies report the need for skilled neonatal resuscitation. Poor Apgar scores, acidosis, and naloxone use correlate with the need for skilled resuscitation, however, and can be used as surrogate outcomes.^{37 - 38} Neonatal resuscitation requires skilled personnel, causes mother-infant separation and parental anxiety, and is costly. Long-term infant effects of labor analgesia cannot be assessed, as none of the studies reported psychomotor outcomes after the perinatal period.

Epidural analgesia does not increase the incidence of instrumented vaginal delivery for dystocia^{7 ,20} (Table 3). However, high overall instrumented delivery rates accompany epidural labor analgesia because epidural analgesia can both prolong the second stage of labor and provide surgical perineal anesthesia. Instrumentation may be applied for resident training purposes or to shorten the second stage of labor. In fact, the authors of 1 study note that the presence of epidural analgesia was used as an indication for forceps delivery for resident training.⁷

The prolongation of the first stage of labor in epidural group patients (42 minutes on average) is probably caused by the temporary labor-slowing effects of intravenous fluid. Intravenous balanced salt solution, 500 to 1000 mL, is usually infused rapidly before epidural drug administration to prevent hypotension secondary to sympathetic neural block.³⁹ Uterine activity does not slow after the initiation of epidural anesthesia in patients receiving up to 500 mL but slows for 20 minutes in patients receiving 1000 mL of intravenous fluid.⁴⁰

Fever associated with epidural labor analgesia has also been reported in prospective observational studies. The increase in maternal core temperature usually begins after 2 to 5 hours of epidural analgesia, continues until delivery, and resolves within 6 hours postpartum.^{41 - 43} Shivering accelerates the temperature increase.⁴³ Maximum maternal temperatures rarely exceed 38.5°C, even with very long labors. While the mechanism of this hyperthermia has not been proven, the most likely explanation is that epidural labor analgesia increases temperature by decreasing maternal hyperventilation and sweating. Hyperthermia is not a general attribute of epidural anesthesia; in fact, hypothermia generally accompanies epidural anesthesia for cesarean delivery or for nonobstetric surgery.^{44 - 45} Further well-controlled prospective trials are needed to define the mechanisms, magnitude, and clinical relevance of epidural-associated maternal temperature disturbances.

Mild hypotension frequently accompanies epidural labor analgesia but is of little consequence to the mother or her neonate if it is detected and treated promptly. Effective epidural analgesia blocks most of the thoracic sympathetic chain, producing vasodilation in the lower half of the body.⁴⁶ This hypotension is prevented and treated by intravenous fluid infusion and ephedrine administration. During labor, common manifestations of clinically significant hypotension are nonreassuring fetal heart rate changes and neonatal acidosis. Despite the risk of hypotension identified in this analysis, epidural analgesia was associated with less neonatal acidosis and no more fetal heart rate abnormalities than parenteral opioids.

The currently available evidence shows no association between epidural analgesia and the overall cesarean delivery rate or the rate of instrumented vaginal or cesarean delivery for dystocia. Thus, a policy of withholding epidural analgesia will not reduce cesarean delivery rates. The mechanism of the effect of epidural analgesia on maternal temperature regulation and the clinical consequences of this effect need to be defined.

Parenteral opioid labor analgesia is not innocuous. Parenteral opioids are associated with poor maternal pain relief and less vigorous neonates. This information should be available to women so that they can make informed choices about labor pain relief.