A systematic review, including a meta-analysis, on the timing effects of neuraxial analgesia (NA) on cesarean and instrumental vaginal deliveries in nulliparous women was conducted. Of 20 articles identified, 9 met the inclusion quality criteria (3,320 participants). Cesarean delivery (odds ratio, 1.00; 95% confidence interval, 0.82-1.23) and instrumental vaginal delivery (odds ratio, 1.00; 95% confidence interval, 0.83-1.21) rates were similar in the early NA and control groups. Neonates of women with early NA had a higher umbilical artery pH and received less naloxone. In the early NA group, fewer women were not compliant with assigned treatment and crossed over to the control group. Women receiving early NA for pain relief are not at increased risk of operative delivery, whereas those receiving early parenteral opioid and late epidural analgesia present a higher risk of instrumental vaginal delivery for nonreassuring fetal status, worse indices of neonatal wellness, and a lower quality of maternal analgesia.
THE American Society of Anesthesiologists and the American College of Obstetricians and Gynecologists (ACOG) have produced a joint statement#1asserting that maternal request is a sufficient medical indication for pain relief during labor, and that pain management should be provided whenever medically indicated. Neuraxial analgesia (NA) is the only available consistently effective technique of pain control during labor and delivery.2–4In North America and Western Europe,5,6pregnant women with painful labor request most frequently NA.
American Society of Anesthesiologists and ACOG evidence-based practice guidelines7,8are useful for making decisions to improve obstetric care outcome, but the topic of NA timing is still controversial. The current Practice Guidelines of the American Society of Anesthesiologists7state that cervical dilation is not a reliable means of determining when regional analgesia should be initiated, and that regional analgesia should be administered on an individualized basis. This position is supported by randomized controlled trials (RCTs)9–11in which NA timing was the primary independent variable of analysis and no association between early NA administration and operative delivery was found. However, these RCTs were small and with high crossover rates.9–11The current ACOG Practice Bulletin8states that, when feasible, obstetric practitioners should delay the administration of NA in nulliparous women until cervical dilation reaches 4–5 cm, and other forms of analgesia should be used until that time. This recommendation is tempered by the statement that women in labor should not be required to reach an arbitrary cervical dilation. ACOG clinical guidelines are sustained by observational and randomized controlled studies12–14in which early NA was associated with an increased risk for cesarean delivery (CD). However, this association was an observation12or the result of a secondary analysis of RCT data.13,14Therefore, both practice guidelines on this topic are based primarily on consensus and expert opinion, and in a recent review the lack of consistent evidence has been reaffirmed.15In June 2006, ACOG has updated its position by a new report of consensus and expert opinion,16stating that, when compared with intravenous systemic analgesia, the initiation of early NA does not increase the risk of CD.
The main objective of this investigation is to perform a systematic review of the effects of NA timing in nulliparous women on CD and instrumental vaginal delivery (IVD) as primary outcomes, and on neonatal parameters related to newborn wellness as secondary outcomes.
Materials and Methods
Searching, Selection, and Validity Assessment
We sought RCTs and cohort studies (CSs), either as full text or in abstract form, about the effects of NA timing on obstetric outcomes. We included abstracts in the attempt to search all available data with the aim of reducing publication bias.17Articles were retrieved from the Cochrane Library (January 2006), EMBASE, and MEDLINE (January 1990 to June 2006) databases using the following terms as medical subject headings, text, and key words, both stand-alone or in combination: epidural and combined spinal–epidural analgesia ; labor first stage and labor onset ; instrumental vaginal delivery and cesarean delivery . We identified additional studies in the reference lists of previously published reviews and retrieved articles, and we hand-searched for other data sources in the annual proceedings (2000–2005) of the Society for Obstetric Anesthesia and Perinatology and the ACOG. Publication language was not a search criterion.
We selected studies on the effects of early NA (defined as epidural or combined spinal–epidural analgesia initiated before the cervix is dilated to 4–5 cm) on obstetric outcomes. Studies that defined “early labor” relative to fetal station were excluded. We identified quality studies adopting the three-pronged restriction criteria proposed by Vandenbroucke,18based on research topic, study design, and analysis. Therefore, studies were selected according to the following inclusion criteria: (1) effects of early NA administration on obstetric outcomes as main research topic; (2) nulliparous women requiring labor pain relief as participants of the study design; and (3) comparison of early NA versus early parenteral opioid, or late epidural analgesia (EA), or both early parenteral opioid and late EA (control group) in laboring women as group analysis. Studies in which the analysis was performed by stratifying according to the degree of cervical dilation and/or with adjustment for potential confounders were excluded.
Randomized controlled trials were also evaluated according to the criteria proposed by Jadad et al. 19The Jadad validated scale is based on three items (randomization, blindness, and description of withdrawals and dropouts) and has a maximum score of five points, assigned on the basis of the quality of randomization and blinding method (absent or inappropriate = 0, appropriate but not described = 1, appropriate and described = 2) and of the outcome report of all enrolled subjects (not described = 0, described = 1).
The Vandenbroucke criteria and Jadad scoring system were independently evaluated by two investigators (M.M., G.P.), and when evaluation or score differed, the study was further assessed to reach consensus.
Data Abstraction and Study Characteristics
Data were independently collected by two investigators (M.M, G.P.), with any discrepancy resolved by reinspection of the original article. To avoid transcription errors, the data were input into statistical software and rechecked by different investigators (G.C., N.B.).
We analyzed RCTs and quality CSs on the effects of early on-demand EA or combined spinal–epidural analgesia on mode of delivery in nulliparous women. To better define each study, the following pieces of information were retrieved: sample characteristics (race, prenatal education, age, weight, height, gestational age, premature or spontaneous rupture of membrane, and pregnancy-associated disease), details of obstetric care (induced labor, amniotomy, oxytocin augmentation and dose, and labor management protocol), anesthetic interventions (pre-NA fluid loading, mode and dose of analgesia), and neonatal interventions (whether naloxone was administered to the infant). For each RCT, we searched for information on withdrawals and dropouts of enrolled laboring women, resulting in their exclusion from the analysis. In addition, we noted noncompliance and crossover of the patients who were originally included in the studies.
Primary outcomes were the CD and IVD rates. Secondary outcomes were neonatal weight, incidence of neonatal Apgar score less than 7 at 1 and 5 min, neonatal umbilical arterial pH (UApH) and venous pH (UVpH), and naloxone use. All outcome measurements were clearly and consistently defined.
The analysis was also performed on other maternal outcomes such as oxytocin use and labor duration. The latter was divided into two stages: from time 0 to full cervical dilation (first stage), and from full cervical dilation to delivery (second stage). In the selected studies, time 0 was defined as time of randomization,9,10,20time of admission to the delivery room,11time of diagnosis of “true” labor,21time of initiation of analgesia,22and time corresponding to 4 cm of cervical dilation.23Duration of labor was reported as mean in five studies9–11,20,21and as median in two studies.22,23Methodologic details such as noncompliance with the assigned protocol and crossover to other treatment, used as surrogate outcomes for maternal satisfaction and analgesia effectiveness, were also analyzed.
Quantitative Data Synthesis
To reach a high statistical power while maintaining a low systematic error, we performed a meta-analysis of RCTs and quality CSs. Meta-analytic techniques (analysis software RevMan, version 4.2; Cochrane Collaboration, Oxford, England, United Kingdom) were used to combine studies using odds ratio (OR) and 95% confidence interval (CI) for dichotomous variables, and weighted mean difference and 95% CI 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 weighted mean difference. An OR of less than 1 or a negative weighted mean difference favored early NA when compared with the control group. Two-sided P values were calculated. A random-effects model was used for all analyses.24Heterogeneity was assessed by using the Q and I2tests.25,26When the Q test P value was less than 0.05 and/or the I2was greater than 25%, heterogeneity was considered significant, and the pooled results were recalculated by excluding heterogeneous studies. If only one study was heterogeneous, it was reported when its sample was larger than pooled included studies. If more than one study was heterogeneous, they were pooled together, and their results were reported only if homogeneous.
The definition of duration of first stage of labor varied from study to study. Therefore, to pool data with different time 0, duration ratio (99% CI) was used and calculated as follows27:
Pooled result expresses the fraction (or percentage) of the increase or decrease of the first-stage labor duration of early NA with respect to control.
In addition, we performed two subgroup analyses on the basis of the study design (RCTs and quality CSs) for the main outcomes (CD and IVD), and two sensitivity analyses for CD and IVD indications (nonreassuring fetal status [NRFS] and dystocia).
Trial Flow and Study Characteristics
We identified a pool of 20 published articles, 18 obtained by electronic database search and 2 abstracts found by means of manual searches (fig. 1). Three articles28–30were eliminated because the NA administration timing was related to fetal station. Another 7 articles12–14,31–34were excluded because they did not comply with one or more Vandenbroucke criteria.18Another article35was rejected because no detail on NA intervention was reported. Reasons for exclusion are reported in table 1. Therefore, 9 articles9–11,20–23,36,37meeting all of the inclusion criteria were selected: 5 RCTs, 1 impact CS, and 3 retrospective CSs (including 1 abstract). These studies involved a grand total of 3,320 patients and were all published in English. The studies were performed either in the United States9,10,21–23,37or in Israel,11,20,36from 1994 to 2006. Regarding the retrospective CS available only in abstract form,37only the comparison between early EA versus early parenteral opioid control groups (cervical dilation < 3 cm) was included (table 1); some print errors were clarified by checking the 2003 conference report of the Society for Obstetric Anesthesia and Perinatology,**and details of the population characteristics and interventions were obtained from another article38enrolling the same participants. Dr. Wong sent us additional unpublished data about her RCT,22regarding the mean and the SD of first- and second-stage labor duration, and the number of neonates receiving naloxone (Cynthia A. Wong, M.D., Associate Professor, Department of Anesthesiology, Northwestern University, Feinberg School of Medicine, Chicago, Illinois, written communication, January 17 and April 6, 2006).
All selected articles were concerned mainly with the effects of NA timing on obstetric outcomes and had a population sample of nulliparous women at term (> 36 weeks) with singleton fetus; only one CS had a mixed-parity sample, but reported the outcome of nulliparous women separately.36All studies enrolled women with uncomplicated pregnancies, except for the abstract,37which included a subgroup of women with pregnancy-induced hypertension. Two RCTs included women with induced labor.9,20All of the studies compared laboring women with early NA administration versus those who received early parenteral opioid, or late EA, or both early parenteral opioid and late EA (control group). Early NA groups had patients requesting initiation of EA in latent and early active phase of labor, defined as cervical dilation of 4 cm or less.39In one RCT,11the patients randomly assigned to the control group did not receive any intervention for pain relief before EA administration; moreover, no patient from either group received any intervention for pain relief after full cervical dilation.
All studies reported main obstetric outcomes. All studies but one36had data on CD rate, and all studies but one37reported data on IVD rate. Indications for CD and IVD were presented only in RCTs, and included dystocia and NRFS. Definitions of dystocia and NRFS were in accordance with the specific criteria of obstetric management and the interpretation of fetal heart rate (FHR) tracing adopted by each medical team (table 2).
Quantitative Data Synthesis
Overall rates of CD (OR, 1.00; 95% CI, 0.82–1.23; 8 studies, 2,980 participants) and IVD (OR, 1.00; 95% CI, 0.83–1.21; 8 studies, 2,816 participants) were similar in the early NA and control groups (figs. 2 and 3). No statistical heterogeneity among studies was detected (CD: Q statistic P = 0.78, I2= 0%; IVD: Q statistic P = 0.56, I2= 0%). In the RCT and quality CS subgroup analyses, no significant difference in CD and IVD between early NA and control groups was found (table 3). Sensitivity analyses for CD and IVD indications showed a significantly higher risk of IVDs for NRFS in parturient women with early parenteral opioid and late EA administration.
No significant difference between early NA and control groups was found in neonatal weight and Apgar scores less than 7 at both 1 and 5 min. Because the heterogeneity of Apgar score less than 7 at 1 min was significant (Q statistic P = 0.11, I2= 55.2%), the pooled result was recalculated by excluding the heterogeneous study.22Excluding the heterogeneous trial (OR, 0.63; 95% CI, 0.44–0.91; 728 participants; P = 0.01),22no significant difference was observed (table 3). A significant difference favoring the early NA group was observed in neonatal UApH. No difference was found for UVpH, although heterogeneity was significant (Q statistic P = 0.02, I2= 73.3%). Excluding the heterogeneous trial (weighted mean difference, 0.00; 95% CI, −0.01 to 0.01; 655 participants; P = 0.99),22a significant difference favoring the early NA group was detected for UVpH (table 3). Naloxone administration was significantly more frequent in newborns of women not receiving early NA (table 3).
First-stage labor duration (evaluated by the surrogate outcome duration ratio) and second-stage duration were not different. Regarding first-stage labor duration ratio, after exclusion of two studies21,22because of heterogeneity, the absence of difference was confirmed. Moreover, neither oxytocin use (overall and after randomization/analgesia) nor prenatal education was significantly different between groups (table 3).
Women receiving early NA were more compliant with the assigned treatment than those receiving control analgesia. Because heterogeneity was significant (Q statistic P = 0.01, I2= 71.8%), the analysis was performed after excluding heterogeneous RCTs9,10and RCT with not estimable OR (all women compliant).11The pooled analysis of remaining studies (as well as the analysis of excluded studies) confirmed a significant difference between groups (table 3). Much fewer laboring women crossed over from the early NA to the control group than vice versa . In two RCTs,11,20OR was not estimable because no patient crossed over (table 3). In the three remaining RCTs9,10,22the 95% CI of each study did not include 1 for the OR. In a reanalysis for noncompliance and crossover outcomes, adding a nominal value of 0.5 in all 2 × 2 cells to enable calculation of ORs and retesting for heterogeneity, no significant difference was found when comparing with the initial analysis.
The main pooled results of the present meta-analysis on the effects of NA timing show the following:
In nulliparous women, on-demand early administration of NA for labor and delivery is not associated with an increased risk of CD and IVD compared with the control group. Laboring women receiving early systemic opioids before late EA have a higher incidence of IVD for NRFS than women receiving early NA.
Early NA is associated with better neonatal outcomes. When early NA is provided, neonatal UApH values are higher, and less naloxone is used than in the control group. Neonates of women receiving early EA present a higher UVpH than the control group. No difference in neonatal weight was found, and Apgar scores less than 7 at both 1 and 5 min are observed equally often in both early NA and control groups.
Early NA is a more effective maternal pain relief method, as indicated by lower noncompliance and crossover rates of laboring women. There are no statistically significant differences in prenatal education, first- and second-stage labor duration, or oxytocin use between early NA and control groups.
Cesarean and Instrumental Vaginal Delivery
The question of whether the timing of NA related to cervical dilation affects maternal outcomes is still the subject of controversy. Some studies have reported that early EA is associated with an increased rate of CD and, consequently, the authors of these studies have proposed delaying the initiation of EA until women in labor reach 4–5 cm of cervical dilation.12–14However, in these studies, EA timing was analyzed by a retrospective analysis with adjustment for confounding factors based on a propensity score,12by a retrospective subgroup approach,13or by a post hoc analysis.14Therefore, the reported “harmful” side effects of early NA initiation on mode of delivery may derive from design and conduct of studies (e.g. , inability to prevent systematic errors and analytic bias). All of these studies were excluded from the current meta-analysis because they did not meet quality criteria.18
The pooled results of the current meta-analysis show that nulliparous women receiving early NA (≤ 4 cm of cervical dilation) are not at increased risk of CD and IVD. Similar results have been obtained in a perspective CS comparing women (50% nulliparous) receiving early versus late EA with timing related to fetal station.30A recent meta-analysis by Liu and Sia3compared nulliparous women receiving EA versus parenteral opioid analgesia, focusing on the effects of low concentration epidural infusions on maternal outcome. No difference in CD rate was observed. However, EA was associated with an increased risk of IVD, even though an investigator bias could not be excluded. As a matter of fact, when studies with elective indications for IVD were excluded from the analysis, Liu and Sia did not find any difference in IVD rate.3In the current meta-analysis, it is noteworthy that all of the excluded and included studies—except two Israeli studies11,36in which traditional concentration epidural infusions were suspended or replaced by parenteral opioid during labor second stage to minimize potential side effects4(table 2)—adopt epidural infusions with a low drug concentration. Apart from this common feature, the results by Liu and Sia3are not strictly relevant to our main focus because they do not address the question of the effects of NA timing. However, it is remarkable that different investigations, following separate routes in the field of NA for labor, reach analogous conclusions.
Despite the methodologic heterogeneity among the studies included in the meta-analysis due to the different interventions within the early NA (EA, combined spinal–epidural analgesia, variability in dose of local anesthetic administered alone or with neuraxial opioids) and control groups (early parenteral opioids, late EA, or both) (table 2), we observed a strong statistical homogeneity and consistency for the main outcomes CD and IVD. Similar results were found in the RCT and quality CS subgroup analyses, suggesting that when topics, designs, and analyses are well defined, randomized and observational studies may produce equivalent conclusions.40,41Therefore, the obstetric practice of delaying NA is not expected to reduce CD and IVD rates.
By sensitivity analysis, we found that laboring women receiving early parenteral opioid with late EA are at higher risk of IVD for NRFS than those receiving early NA. This finding might be related to the effects of parenteral opioids. Their efficacy on maternal analgesia is largely dose dependent rather than drug dependent,42and opioids may indirectly affect the fetus by altering maternal ventilation or uteroplacental flow. Moreover, because an amount of drug readily crosses the placenta, depending on the dose and the route of administration, opioids may directly affect the fetus by decreasing baseline FHR and FHR variability.43Hill et al. 44found a greater incidence of abnormal FHR tracing patterns, usually associated with NRFS, in nulliparous women receiving parenteral opioids than in women receiving EA. Therefore, the strategy of delaying EA and providing early analgesia with parenteral opioids may result in a higher risk of IVD for NRFS and may not be advantageous. The two studies included in this sensitivity analysis are from the same institution, and the finding should be confirmed by other groups of investigators, because this result is dependent on adhering to predefined indications for IVD.
Finally, the inferences of our results in clinical practice are that patient factors (women with complicated labor more likely request early NA) and obstetric decisions (labor protocol criteria for dystocia and NRFS not well defined), and not the NA technique itself, are the most likely causative factors of operative delivery.
The neonates of laboring mothers receiving early NA are more alert, have less need for naloxone, and have a higher UApH value. Fetal and neonatal status is usually assessed by UApH, umbilical artery base excess/deficit, and arterial partial pressure of carbon dioxide. In a meta-analysis comparing epidural with systemic opioid analgesia, Reynolds et al. 45found that EA is associated with improved neonatal acid–base status, as demonstrated by higher UApH (8 RCTs, 2,322 participants) and umbilical artery base excess (4 RCTs, 1,927 participants).
Analyzing the UVpH, we observed a significant heterogeneity in one trial. Excluding this study, in which early analgesia was provided by using neuraxial opioid (combined spinal–epidural analgesia),22heterogeneity vanished and a higher UVpH was found in the early NA group. Because UVpH reflects maternal acid–base status and placental function, this finding could be a result of improved uteroplacental blood flow in the early NA group,46,47or of maternal hypoventilation in the group using parenteral opioids.42
In summary, these results suggest that fetal–neonatal status is improved with early NA compared with parenteral opioid analgesia.
Neuraxial Analgesia Effectiveness and Maternal Satisfaction
Noncompliance with the assigned protocol and crossover to other treatments are surrogate outcomes for maternal satisfaction and analgesia effectiveness. The lower noncompliance and crossover rates of early NA laboring women suggest that early NA provides superior maternal pain relief than early parenteral opioids and/or late EA. However, significant key differences in participants and interventions among RCTs require a more detailed analysis. The American RCTs with women (Caucasian rates 87%,980%,10and 76%22) receiving early parenteral opioid plus late EA presented significant crossover rates (ranging from 3%22to 35%9). Moreover, a higher number of women in the control group described poor-quality analgesia9,10,22and dissatisfaction,9,10as evaluated by verbal rating scores. On the contrary, the Israeli RCTs showed that women undergoing late EA considered their childbirth experience satisfactory20and did not cross over to receive early EA.11,20
Labor pain is a complex phenomenon with sensory, emotional, and perceptive components. The pain threshold may be modified by multiple factors, such as heterogeneity inherent in labor, anxiety, previous suffering experience, and social and cultural influences.48Moreover, labor pain is associated with autonomic, psychological, emotional, and behavioral responses.48A neuroendocrine response elicited by prolonged and severe pain during labor and delivery causes a decrease in placental perfusion and uncoordinate uterine activity, and may be a cofactor for dysfunctional labor in women at risk.49Laboring women requesting NA are at risk for complicated labor because they are often nulliparous, at term, with cephalic presentation in occiput–posterior position, have heavier neonates, are admitted to the hospital with less cervical dilation and higher fetal head station, have slower rate of labor progression, and need oxytocin augmentation.50Laboring women requiring more local anesthetic51,52or opioid53for analgesia more frequently are affected by dystocia. Yet, opioids are inadequate to control the progression of labor pain.54,55Whatever the main determining factors of labor pain intensity and their relation with dysfunctional labor (associative or causative), a more effective and safer analgesia, as produced by early NA, should be beneficial and should not be delayed while waiting for further cervical dilation.
Validity Limitations and Research Agenda
The studies included in this meta-analysis, although fulfilling quality criteria,18have some limitations due to the peculiarity of the topic. For example, one of the main biases is the lack of blinding of the RCTs,19due mainly to the marked differences in the quality of analgesia, because both patients and caregivers can readily differentiate the two analgesic techniques. The lack of appropriate blindness is the reason why no RCT of the current meta-analysis had a Jadad score higher than 3.
Furthermore, only in two RCTs,20,22the analysis was performed on an intention-to-treat basis that is more appropriate, although not the perfect solution, when crossover occurs.27Labor treatment protocols are crucial to minimize performance bias. The lack of description of FHR interpretation criteria in two trials9,10may have introduced an investigator bias. To reduce selection bias, all CSs were selected for predefined homogeneous labor characteristics of comparison cohorts (early NA and control), but the presence of other potential confounders cannot be excluded. Some neonatal outcomes derive only from three RCTs9,10,22of two American research groups that published their findings in 1994 and 2005. When their data were pooled, a statistical heterogeneity for effect size, but not for direction, was found. This may have reduced the statistical power of the analysis.
First-stage labor duration ratio has been suggested as an indicator when time 0 is not standardized.27However, analysis by using duration ratio is not a validated statistical method in meta-analytic technique; moreover, the use of means instead of medians for describing labor duration may be inappropriate,27and therefore, computing a ratio of means, in this context, may not be valid. These limitations, and the statistical heterogeneity of the pooled analyzed studies, suggest that no definite conclusion can be drawn about first-stage labor duration.
The quality of research design has improved in recent times. However, we suggest that future trials on labor analgesia incorporate a patient-oriented approach56by adopting accurate, precise, and repeatable criteria to better define those maternal and fetal/neonatal conditions that activate therapeutic interventions, i.e. , dystocia and NRFS as indications for operative deliveries.
The current meta-analysis supports no association between patient-requested early NA and overall rate of IVD and CD and strongly corroborates the current practice guidelines of American Society of Anesthesiologists7and recent ACOG Committee Opinion.16Early parenteral opioids with late EA are associated with a higher risk of IVD for NRFS, a lower quality of maternal analgesia, and worse indices of neonatal wellness compared with early NA. Therefore, we recommend providing parenteral opioids only when on-demand NA is medically contraindicated.
The obstetric practice of delaying NA is not a sound healthcare policy. Under this perspective, the Obstetric Anesthetists’ Association and the Association of Anesthetists of Great Britain and Ireland††jointly recommended that for giving quality care, where a 24-h epidural service is offered, the time elapsing between the call to the anesthetist and his or her contact with the parturient should not exceed 30 min. Organizational factors such as availability, accessibility, and continuity of obstetric analgesia services, as well as individual and social acceptability, are components of healthcare quality.4Nowadays, effective labor pain management is frequently a patient expectation.
The authors thank Rocco Giuliani, M.D. (Full Professor, Department of Emergency and Organ Transplantation, University of Bari, Italy), for his review of the manuscript; Cynthia A. Wong, M.D. (Associate Professor, Department of Anesthesiology, Northwestern University, Feinberg School of Medicine, Chicago, Illinois), for supplying unpublished personal data; and Pasquale Cinnella, Ph.D. (Full Professor, Department of Aerospace Engineering, Mississippi State University, Starkville, Mississippi), for his help with manuscript preparation.