Labor Analgesia and Cesarean Delivery: An Individual Patient Meta-analysis of Nulliparous Women

The base sample of this individual patient meta-analysis was 4,465 women of mixed parity (2,703 nulliparous and 1,762 multiparous) randomized to either epidural analgesia or intravenous meperidine analgesia in five studies conducted at Parkland Hospital between November 1, 1993, and November 3, 2000. 11–13,16,17The current analysis was limited to nulliparous women only. The study protocols used for the patients described in this meta-analysis were developed by investigators from the Departments of Anesthesiology and Obstetrics and Gynecology and approved by the Institutional Review Board of the University of Texas Southwestern Medical Center at Dallas. All protocols were conducted at Parkland Hospital, which is a tax-supported institution serving the medically indigent of Dallas County. Women giving written consent were randomly assigned using numbered, sealed envelopes to receive either epidural analgesia or intravenous analgesia at their first request for relief of labor pain. The randomization sequences were computer derived in blocks of 20 subjects. Eligible participants included women with uncomplicated pregnancies in labor at term (≥ 36 weeks gestation) as well as women with pregnancy-induced hypertension at 36 weeks’ or more gestation.

All pregnancies were managed by certified nurse-midwives under direct supervision by obstetric faculty and house officers following a written protocol established by medical staff. Routine intrapartum management of all women included intravenous fluid administration and electronic fetal heart rate surveillance for 30 min after commencing epidural or intravenous analgesia. Continuous internal electronic fetal heart rate monitoring was used in those women with meconium-stained amnionic fluid, auscultated fetal heart rate decelerations, or inadequate progress of labor. Pelvic examinations were performed approximately every 2 h to evaluate the progress of labor.

A cervical change of less than 1 cm/h coincidental with hypotonic uterine contractions measured using intrauterine pressure transducers resulted in augmentation of labor with oxytocin. Oxytocin was administered per written protocol, which has been described previously. 18Briefly, oxytocin starting at 6 mU/min was increased by 6 mU/min at 40-min intervals up to a maximum of 42 mU/min. Uterine activity of 200–250 Montevideo units for 2–4 h was considered adequate. Dystocia was diagnosed when adequate uterine activity did not result in progressive cervical dilation or descent of the fetal head. Indications for the use of forceps were limited to inadequate voluntary pushing or fetal heart rate abnormalities. Inadequate voluntary pushing was determined at the bedside when lack of fetal descent due to inadequate maternal expulsive efforts was observed. Umbilical artery blood for analysis of gases was obtained at all births from a doubly clamped cord segment.

Women randomized to epidural analgesia received an intravenous bolus of 500 ml lactated Ringer’s solution immediately preceding analgesia. Analgesia was initiated with either epidural bupivacaine 11,12,16,17or intrathecal sufentanil. 13Epidural analgesia was maintained with various concentrations of bupivacaine and fentanyl (table 1). Analgesia was maintained throughout the first stage of labor. If progress during the second stage of labor was inadequate after 1 h, the infusion was halved or discontinued to restore maternal expulsive efforts. Additional boluses of fentanyl and/or bupivacaine were injected to overcome inadequate analgesia. If required, the epidural catheter was replaced. Left uterine displacement was maintained to avoid aortocaval compression.

Maternal blood pressure was recorded every 5 min for 30 min and then every 30 min until delivery. Intravenous fluids were given to treat hypotension defined as a systolic blood pressure less than 25% of baseline or a systolic blood pressure less than 100 mmHg. Persistent hypotension was treated with 5 mg ephedrine intravenously as needed.

All women randomized to meperidine analgesia in these protocols received an initial bolus of 50 mg meperidine and 25 mg promethazine hydrochloride intravenously. The methods of meperidine administration after the loading dose are shown in table 1. Maternal blood pressure was recorded as already described in women who received epidural analgesia.

The primary outcome was the overall cesarean delivery rate. Also analyzed were the indications for cesarean delivery, forceps delivery rate, duration of the first and second stages of labor, labor complications, and infant outcomes.

An individual patient meta-analysis using the original databases was performed. All tests of significance were performed using two-tailed tests. Initially, data were analyzed with use of SAS statistical software (SAS Institute, Inc., Cary, NC; statistical significance, P < 0.05) using the unpaired Student t  test, Pearson chi-square test, and Mann-Whitney U test as indicated. Data were analyzed according to group assignment at randomization (intent-to-treat analysis) regardless of the eventual analgesia received. Secondarily, patients compliant to their randomization assignment were also compared.

Because data were pooled from five different studies, we also estimated heterogeneity of results using the Breslow-Day statistics. If P  values were less than 0.10, a random effect model was used to estimate the effects; otherwise, the Cochran-Mantel-Haenszel statistic was incorporated to establish statistical significance. This approach differs from typical meta-analyses in which only summary statistics and sample size are known. Having data available from the individual patients allows the incorporation of general statistical methods adjusting for the individual studies. A statistical difference between groups was considered to occur if the combined 95% confidence interval (CI) did not include 1 for the odds ratio (OR).

A total of 4,465 women of mixed parity (nulliparous, n = 2,703; multiparous, n = 1,762) were randomized in five studies (fig. 1). 11–13,16,17Only nulliparous women were included in this meta-analysis. Of these 2,703 nulliparous women, 1,339 women were allocated to receive epidural analgesia, and 1,364 women were allocated to receive intravenous meperidine analgesia. Five hundred fifteen (19%) of all nulliparous women were diagnosed with pregnancy-induced hypertension. 16Two hundred forty women (18%) in the epidural group and 182 women (13%) in the intravenous meperidine group did not receive their allocated analgesia (fig. 1). One hundred sixty-five women (12%) who received intravenous meperidine analgesia as randomized later crossed over to epidural analgesia because of inadequate pain relief. An adjustment for the five studies did not affect any of the results, so we elected to simplify the presentation and present analysis on the combined data set.

As shown in table 2, there were no significant differences in maternal demographic characteristics, cervical dilation at the time of analgesia, or number of women with pregnancy-induced hypertension between the epidural and intravenous meperidine analgesia groups.

Labor events were analyzed in relation to the type of analgesia used, and the results are shown in table 3. Epidural analgesia was significantly associated with prolongation of the first (P < 0.011) and second (P ≤ 0.001) stages of labor, the need for augmentation of labor with oxytocin (P < 0.001), and maternal fever (P < 0.001).

Methods of delivery are shown in table 4. Significantly fewer women randomized to epidural analgesia experienced spontaneous deliveries compared to those randomized to intravenous meperidine analgesia (adjusted OR, 0.69; 95% CI, 0.57–0.83; P < 0.001). This difference resulted from significantly increased forceps deliveries (both low and outlet forceps) rather than cesarean deliveries. Specifically, 13% of women receiving epidural analgesia (172 of 1,339) had forceps deliveries compared to 7% (101 of 1,364) in the intravenous meperidine analgesia group (adjusted OR, 1.86; 95% CI, 1.43–2.40; P < 0.001). The overall cesarean delivery rates for the epidural analgesia and intravenous meperidine analgesia were 10.5% (140 of 1,339) and 10.3% (141 of 1,364), respectively (adjusted OR, 1.04; 95% CI, 0.81–1.34; P = 0.920;fig. 2). There were no significant differences in cesarean delivery for dystocia or fetal heart rate abnormalities. We could not detect heterogeneity in the study results for this outcome.

The cesarean delivery rate was similar among three different methods of epidural analgesia (conventional epidural analgesia, 12%[94 of 777]; combined spinal-epidural analgesia, 9%[30 of 336]; patient-controlled epidural analgesia, 7.1%[16 of 226]; P = 0.54).

Infant outcomes are summarized in table 5. One-minute and 5-min Apgar scores less than 7 were significantly increased in the intravenous meperidine analgesia group. Umbilical artery blood acidemia (pH < 7) was not related to the type of analgesia given to the mother. Six neonates in the epidural analgesia group and six in the intravenous meperidine analgesia group required transfer to the neonatal intensive care unit (P = 0.954). There were no neonatal deaths.

Because approximately 20% of the all randomized nulliparous women did not comply with their analgesia protocol (fig. 1), we separately compared those patients who did comply (table 6). There were no significant differences in demographic characteristics at the time of analgesia between the protocol-compliant groups. Analysis of the protocol-compliant groups also showed no significant difference between the epidural analgesia and intravenous meperidine analgesia groups in the rate of cesarean deliveries (epidural analgesia, 11.5%vs.  parenteral analgesia, 9.6%[OR, 1.19; 95% CI, 0.93–1.53]; P = 0.172). However, epidural analgesia was associated with a higher forceps delivery rate (epidural analgesia, 15%vs.  intravenous meperidine analgesia, 6%; P < 0.001), and prolonged first and second stages of labor. We could not detect heterogeneity in the study results for these outcomes, and an adjustment for the five studies did not affect any of the results

Table 7summarizes the progress of labor in women who crossed over from intravenous meperidine to epidural analgesia because of inadequate pain relief. The cesarean delivery rate was significantly higher in women who crossed over from meperidine analgesia to epidural analgesia compared to those who did not cross over (crossover, 18.8%vs.  no crossover, 9.2%, [OR, 2.05; 95% CI, 1.42–2.95]; P < 0.001). Also, the forceps delivery rate was higher and the first and second stages of labor were prolonged in women who crossed over compared to those who did not.

The preanalgesic visual analog pain scale scores were similar between the two study groups (epidural, 9 ± 1.6 vs.  intravenous meperidine, 9 ± 1.7; P = 0.09). Women who received epidural analgesia reported lower pain scores during the first stage (epidural, 2 ± 3 vs.  meperidine, 4 ± 4; P < 0.0001) and second stage (epidural, 3 ± 3 vs.  meperidine, 5 ± 4; P < 0.001) of labor. When parturients were queried within 24 h after delivery, 95% of women who received epidural analgesia rated their satisfaction as excellent or good compared to 69% of women who received intravenous meperidine analgesia (P < 0.0001).

We found that epidural analgesia was associated with prolonged labor, an increased need for oxytocin stimulation, an increased incidence of maternal fever, and an increased number of forceps deliveries. However, the number of cesarean deliveries was not increased. Using the observed overall cesarean delivery rate in the meperidine analgesia group as the baseline, our sample size has 80% power with less than 0.05 significance to detect an absolute increase due to epidural analgesia of 3% (from 7% to 10%) in the overall cesarean delivery rate.

The results of this individual patient meta-analysis of randomized epidural versus  intravenous meperidine analgesia during labor in nulliparous women represent the culmination of 7 yr of clinical trials focused on measuring the effects of such analgesia on childbirth in a general obstetric population at a single hospital. 11–13,16,17It is our opinion that the combined data of over five trials including 2,703 randomized nulliparous women provides a unique opportunity to analyze the effects of epidural analgesia on childbirth.

Although meta-analysis is an often-used method of aggregating data published in the literature to enhance sample size, individual patient meta-analyses as we have conducted using actual data from each studied subject is comparatively uncommon. Our meta-analysis also has the additional advantages of including only patients from a single center using standardized obstetric management as well as consistent definitions of the outcomes of interest. Moreover, individual patient data permit statistical adjustment because each subject’s specific characteristics can be linked to an outcome of interest. Such adjustment is not possible in a standard meta-analysis.

Several retrospective and prospective studies conducted over the past decade have attempted to evaluate the effect of epidural labor analgesia on cesarean delivery. Retrospective studies 7–9have selection bias because women choosing epidural analgesia may have more intense pain due to more difficult labor and therefore may be intrinsically at greater risk for dystocia necessitating cesarean delivery. 19–21In this study, we found that those women who crossed over to epidural analgesia after meperidine failed had more difficult labor and a higher cesarean delivery rate (table 7). This supports the premise that selection of women for study, based on their need for epidural analgesia, falsely increases the associated cesarean delivery rate. Prospective studies have also had methodologic limitations. 22Many women have firm views on the type of analgesia they prefer in labor and are reluctant to consent to receiving a predetermined method of analgesia, or they may consent and then withdraw from the study, thereby contributing to protocol failures. 11–13Another important consideration is unsatisfactory pain relief in women randomized to the control group that ethically mandates permitting these women to cross over to epidural analgesia. 11,13–15This in turn significantly distorts both study arms. Methodologic problems such as these have prevented definitive conclusions from several prospective trials as to whether labor epidural analgesia causes cesarean births.

Problems with study execution, such as crossovers from meperidine to epidural analgesia, obscured the effects of epidural analgesia in our two earliest trials, in which the crossover rates from meperidine to epidural analgesia were very high (16% and 17%). 11,13This problem was minimized in our three subsequent trials, largely because of patient satisfaction with patient-controlled intravenous meperidine, which resulted in few crossovers (2%, 1%, and 6%). 12,16,17In these subsequent trials, there was no difference in the rate of cesarean deliveries between the two analgesia groups based on intent-to-treat analysis according to original randomization as well as by analysis of protocol-compliant groups.

Recent reports indicate that epidural analgesia is not associated with excess cesarean births. Leighton and Halpern 23performed a meta-analysis in 2002 of 14 studies including 2,161 women who received epidural analgesia versus  2,163 who received opioids for labor pain. They reported no increase in the cesarean delivery rate attributable to epidural analgesia. Zhang et al.  24recently reported on the effects of the introduction of an on-demand labor epidural analgesia service at Tripler Army Hospital, Honolulu, Hawaii. In late 1993, a policy change within the Department of Defense required the availability of on-demand labor epidural analgesia in military medical centers. As a result, the incidence of epidural analgesia increased from 1% before the policy to 84% at 1 year after the policy had been implemented. The primary cesarean delivery rates were 14.4% before and 12.1% after this dramatic change in epidural use.

It is important to emphasize that there has been continual refinement of techniques used for epidural analgesia during labor. It is likely that early methods of epidural analgesia using high doses of analgesic agents had greater effects on labor than contemporary low-dose techniques such as those used in our trials.

Although epidural analgesia was not associated with increased cesarean deliveries in our meta-analysis, such analgesia had effects on labor. Specifically, epidural analgesia increased the mean length of labor by approximately a half-hour, which may in part explain the increased used of oxytocin for correction of prolonged labor. The second stage of labor was also prolonged when epidural analgesia was used. Clinically, we believe epidural-related poor maternal expulsive efforts could be responsible for the increase in forceps deliveries that we observed. Further, epidural analgesia was associated with an increased incidence of maternal fever. However, the underlying mechanism for fever among parturients receiving epidural analgesia is unclear.

In summary, epidural analgesia can lengthen the duration of labor and result in more frequent use of oxytocin to stimulate labor. However, these effects did not result in increased cesarean delivery. It is our opinion that the fear of unnecessary cesarean delivery should not prevent women from choosing an effective method of pain relief during labor.