Objective
The purpose of this study was to provide an updated summary of the literature regarding the effects of tocolysis with indomethacin on neonatal outcome by systematically reviewing previously and recently reported data.
Study Design
All previously reported studies pertaining to indomethacin tocolysis and neonatal outcomes along with recently reported data were identified with the use of electronic databases that had been supplemented with references that were cited in original studies and review articles. Observational studies that compared neonatal outcomes among preterm infants who were exposed and not exposed to indomethacin were included in this systematic review. Data were extracted and quantitative analyses were performed on those studies that assessed the neonatal outcomes of patients that received antenatal tocolysis with indomethacin.
Results
Twenty-seven observational studies that met criteria for systematic review and metaanalysis were identified. These studies included 8454 infants, of whom 1731 were exposed to antenatal indomethacin and 6723 were not exposed. Relative risks with 95% confidence intervals were calculated for dichotomous outcomes with the use of random and fixed-effects models. Metaanalysis revealed no statistically significant differences in the rates of respiratory distress syndrome, patent ductus arteriosus, neonatal mortality rate, neonatal sepsis, bronchopulmonary dysplasia, or intraventricular hemorrhage (all grades). However, antenatal exposure to indomethacin was associated with an increased risk of severe intraventricular hemorrhage (grade III-IV based on Papile’s criteria; relative risk, 1.29; 95% confidence interval, 1.06–1.56), necrotizing enterocolitis (relative risk, 1.36; 95% confidence interval, 1.08–1.71), and periventricular leukomalacia (relative risk, 1.59; 95% confidence interval, 1.17–2.17).
Conclusion
The use of indomethacin as a tocolytic agent for preterm labor is associated with an increased risk for severe intraventricular hemorrhage, necrotizing enterocolitis, and periventricular leukomalacia.
Preterm birth represents an important perinatal health problem across the globe, not only in terms of associated mortality rates but also with regard to short- and long-term morbidity and financial costs. In the Unites States, the preterm birth rate reached its peak in 2006 (12.8%). However, despite a gradual decline for the seventh straight year in 2013 (11.4%), the United States ranks as 1 of the top 10 countries in the world with the highest number of preterm births. In the United States, premature birth accounts for nearly 35% of deaths in the first year of life at an estimated annual cost that exceeds $26 billion. In addition, preterm birth contributes to substantial neurobehavioral impairment.
Major recent progress has been made toward the early diagnosis, prediction, and prevention of spontaneous preterm birth. However, the mainstay of therapy for the treatment of acute preterm labor continues to be the employment of pharmacologic agents with the aim of arrest of or decrease of uterine contractility and thereby delaying preterm birth. Although tocolytic agents have been shown to delay delivery for 48 hours to 7 days, their use has not led to an improvement in neonatal outcomes. Nonetheless, this short prolongation of pregnancy allows for the maternal transfer to a tertiary center and the administration of corticosteroids to enhance fetal lung maturity.
A variety of pharmacologic agents that have been used, most off-label, to suppress uterine contractility include beta-mimetics, magnesium sulfate, oxytocin receptor antagonists, calcium channel blockers, cyclooxygenase inhibitors, and nitric oxide donors. Each of these agents has a unique mechanism of action and side-effects. A recent metaanalysis suggested that the calcium channel blocker nifedipine appears to meet several characteristics of an ideal tocolytic agent. A systematic review and network metaanalysis and a metaanalysis with decision analysis concluded that calcium channel blockers and prostaglandin inhibitors had the highest probability of delaying delivery and improving neonatal and maternal outcomes.
There is strong evidence that prostaglandins are involved intimately in the initiation and progression of term and preterm labor in humans. Prostaglandins affect myometrial contractility by direct effect with their own receptors by an increase in the sensitivity of the myometrium to oxytocin and by regulation of myometrial gap junctions. They also stimulate the influx of intracellular calcium that activates the enzyme myosin light chain kinase that results in myometrial contractility.
Indomethacin, a cyclooxygenase inhibitor that decreases uterine contractility by blocking the conversion of arachidonic acid to prostaglandin, has been used as a tocolytic agent since 1974. Many obstetricians continue to use indomethacin as a first-line tocolytic agent, which is an indication that is supported by the American College of Obstetricians and Gynecologists. A number of studies have raised concerns about the safety of indomethacin because it crosses the placenta and inhibits prostaglandin synthesis in fetal organs. Case reports and observational studies have suggested that indomethacin may cause adverse neonatal outcomes that include necrotizing enterocolitis (NEC), intraventricular hemorrhage, periventricular leukomalacia, and other cardiac, pulmonary, and renal abnormalities. Accordingly, indomethacin’s role as a current option by obstetricians in the treatment for possible preterm labor is controversial. Since 2005, 2 systematic reviews with metaanalyses that assessed neonatal outcomes after indomethacin tocolysis, while using similar sources, have reported conflicting results. The first of these reports included both observational studies and randomized trials; the subsequent report included only observational studies. Subsequent to these publications, more recent observational and prospective studies that assessed indomethacin tocolysis have been published. The goal of this current study was to review these new studies, to reanalyze studies that were included in the 2 previous reviews, and to pool the data to determine more accurately the neonatal effects of indomethacin exposure, thus providing needed guidance regarding the use of this medication for tocolysis.
Materials and Methods
This systematic review and metaanalysis was conducted according to the Metaanalysis of Observational Studies in Epidemiology (MOOSE) guidelines. Searches were conducted for published literature from January 1966 to March 2014. The key words indomethacin and tocolysis were used in the search independently and then in conjunction with the following other key words: bronchopulmonary dysplasia , intraventricular hemorrhage , patent ductus arteriosus , necrotizing enterocolitis , and neonatal mortality . Prospective and retrospective observational studies and clinical trials that evaluated tocolysis with indomethacin as an exclusive agent or in combination with other tocolytics for preterm labor vs a comparison group without tocolytics or with a different tocolytic agent were identified. In addition to including those studies that were assessed in the previous systematic reviews, we identified more recent observational studies and clinical trials that evaluated neonatal outcomes with the use of indomethacin for tocolysis in patients with preterm labor using the following computerized databases: PubMed, MEDLINE, and Cochrane. All of the included studies assessed ≥1 of the following neonatal outcomes: intraventricular hemorrhage (IVH) that was graded by Papile’s classification based on head ultrasound scanning, NEC that was based on Bell’s staging criteria or x-ray findings of pneumatosis intestinalis and/or intestinal perforation, patent ductus arteriosus (PDA) that was diagnosed by echocardiography, bronchopulmonary dysplasia (BPD) that was based on oxygen requirements of the neonate at 36 weeks postmenstrual age, respiratory distress syndrome (RDS) that was based on clinical and/or chest x-ray findings, periventricular leukomalacia (PVL) that was diagnosed on head ultrasound scanning or other imaging, neonatal sepsis that was evidenced by positive cultures and/or clinical symptoms, or neonatal death that represented death during initial hospitalization after birth. Because each report did not assess all outcomes of interest, specific outcome metaanalyses were performed that were based on a variable number of studies that were related to that outcome. Studies were excluded if they lacked a comparison group, lacked assessment of neonatal outcomes of interest, or lacked sufficient quantitative data for extraction.
Each study was scored for quality by the primary author and a second obstetrician with the use of the Newcastle-Ottawa Quality Assessment scale. Data that were collected from each study included first author, study design, publication year, control group definition, gestational age at birth of subjects, neonatal outcomes that were measured, number of subjects in the study and control group, other tocolytics that were used, neonatal birthweight and gender, and the administration of steroids for fetal lung maturity. Raw data were extracted by the primary author using 2 × 2 tables for each neonatal outcome that was measured in the antenatal indomethacin exposure group and in the comparison group. Two other independent researchers then confirmed these data.
Metaanalyses were performed for each neonatal outcome with a Stata statistical software package (version 11.0; Stata Corp, College Station, TX). Relative risk (RR) for each outcome and 95% confidence intervals (CI) were calculated for the antenatal indomethacin-exposed group compared with the no-exposure comparison group for dichotomous outcomes. Estimates of RR were calculated with fixed-effects (Mantel-Haenszel) and random-effects (DerSimonian and Laird) models. Random effects models were used whenever there was evidence of heterogeneity ( P < .10). The null hypothesis underlying the overall test of association was that the overall RR was equal to 1. Publication bias was evaluated with the Egger test and by inspection of funnel plots, which plot RR against study sample size. When biases and heterogeneity are absent, the variation in the estimated effect decreases with increasing sample size, and the plot resembles a symmetric funnel. Asymmetric funnel plots are caused by small trials that report greater effects than larger trials, which suggests publication or other biases.
The Breslow-Day method was used to test the homogeneity treatment effect across the studies to determine the combinability of the individual studies. In addition, L’Abbe plots were inspected visually to assess homogeneity across studies. Sensitivity analyses were performed by sequential omission of each study and analysis of the overall impact of that particular study on the pooled results. Metaregression analysis was performed to identify causes of heterogeneity among any statistically significant neonatal outcomes by examination of the study variables of neonatal birthweight, antenatal steroid use, sex, gestational age at delivery, delivery within 48 hours of last indomethacin dose, other tocolytic use in control infants, and mean dosage of indomethacin that was administered. The number needed to treat or harm with 95% CIs was calculated for any outcome that revealed statistical significance.
Results
In addition to the overall total of 21 observational studies that previously were reported in the 2005 metaanalysis by Loe et al and in 2007 by Amin et al, 7 new observational studies that evaluated antenatal indomethacin tocolysis and its impact on neonatal outcomes and that met the inclusion criteria were identified. Randomized clinical trials were not included in this updated analysis. The study selection process is detailed in Figure 1 . One previously reported study was excluded because it contained data on postnatal indomethacin treatment only and was included inadvertently in the study by Loe. There were 2 new observational studies by Soraisham et al that included overlapping populations of infants; each of the studies looked at different primary outcomes but also reported on the secondary neonatal outcomes that were relevant to this study. Therefore, only the study that included the largest number of subjects was included. The characteristics of the studies that were included are detailed in Table 1 . The 27 observational studies that were included assessed 8454 infants overall, which is an increase of 2375 from the previously reported metaanalyses. Of these infants, 1731 were exposed to antenatal indomethacin, and 6723 were not exposed. The studies varied with respect to size, cause of preterm delivery, use of other tocolytics in the study groups and in the comparison groups, the dose and duration of indomethacin therapy, and the time from last dose of indomethacin to delivery.
| Study | Year | Infant, n | Study group tocolytic | Comparison group tocolytic | Gestational age, wk |
|---|---|---|---|---|---|
| Norton et al | 1993 | 114 | Indo ± Mg, β-mimetic | None, Mg, or β-mimetic | ≤30 |
| Souter et al | 1998 | 79 | Indo ± salbutamol or nifedipine | None, salbutamol, or nifedipine | ≤30 |
| Gardner et al | 1996 | 124 | Indo ± Mg, β-mimetic | None, Mg, or β-mimetic | <32 |
| Vermillion and Newman | 1999 | 225 | Indo + Mg | Mg | ≤32 |
| Van Overmeire et al | 1998 | 76 | Indo + β-mimetic | β-mimetic | ≤33 |
| Ojala et al | 2000 | 176 | Indo | Not Specified | <33 |
| Weintraub et al | 2001 | 2794 | Indo | None, Mg, or β-mimetic | ≤32 |
| Gerson et al | 1990 | 57 | Indo + Terb and Mg or ritodrine | Terbutaline + Mg or ritodrine | <33 |
| Abbasi et al | 2003 | 248 | Indo ± Mg, terbutaline | None, Mg or terbutaline | <34 |
| Niebyl and Witter | 1986 | 135 | Indo | None or “other” | ≤34 |
| Iannucci et al | 1996 | 56 | Indo + Mg | Mg | <30 |
| Suarez et al | 2001 | 70 | Indo | Mg | <33 |
| Major et al | 1994 | 759 | Indo ± Mg | Mg ± β-mimetic | <30 |
| Hammerman et al | 1998 | 105 | Indo | Not specified | <33 |
| Al-Alaiyan et al | 1996 | 30 | Indo | None | ≤33 |
| Parilla et al | 2000 | 110 | Indo | None or Mg | <37 |
| Baerts et al | 1990 | 159 | Indo + fenoterol | None or fenoterol | <30 |
| Pietrantoni et al | 1995 | 280 | Indo | Not specified | ≤32 |
| Friedman et al | 2005 | 236 | Indo | None | <32 |
| Murata et al | 2005 | 201 | Indo + ritodrine | Mg ± ritodrine | <33 |
| Doyle et al | 2005 | 549 | Indo + Mg | Mg | <34 |
| Baerts et al | 2013 | 36 | Indo | None or “other” | <30 |
| Cordero et al | 2007 | 116 | Indo ± Mg, terbutaline | Mg ± terbutaline | ≤28 |
| Amin et al | 2008 | 248 | Indo ± Mg | None or Mg | ≤29 |
| Soraisham et al | 2011 | 462 | Indo | Not specified | ≤28 |
| Sood et al | 2011 | 628 | Indo ± Mg | None or Mg | <32 |
| Sharma et al | 2010 | 381 | Indo | Not specified | <36 |
Data on neonatal outcomes were pooled ( Table 2 ). Significant heterogeneity among studies was noted for all neonatal outcomes, except PVL and neonatal sepsis. For all other outcomes, a random-effects model was used because of the presence of heterogeneity. The risk of severe IVH, which was defined with Papile’s classification stage III and IV, was significantly higher among infants who received tocolysis with indomethacin (RR, 1.29; 95% CI, 1.06–1.56), compared with those who received no tocolysis or tocolysis with other agents ( Figure 2 ). The number needed to harm for this outcome was 26. A significant increased risk for NEC was also noted among infants who were exposed to antenatal indomethacin (RR, 1.36; 95% CI, 1.08–1.71; Figure 3 ). The number needed to harm for this outcome was 30. In addition, the incidence of PVL was increased significantly among those infants who were exposed to antenatal indomethacin (RR, 1.59; 95% CI, 1.17–2.17; Figure 4 ). The number needed to harm for this outcome was 28. No significant differences were found for the risk of RDS, PDA, BPD, IVH (all grades), neonatal sepsis, or neonatal death.
| Outcome | References | Study group, n/N | Comparison group, n/N | Relative risk (95% confidence interval) | Heterogeneity ( P value) |
|---|---|---|---|---|---|
| Periventricular leukomalacia | 66/595 | 83/1432 | 1.59 (1.17–2.17) a | .974 | |
| Respiratory distress syndrome | 366/708 | 668/1581 | 0.92 (0.77–1.08) b | .000 | |
| Intraventricular hemorrhage | |||||
| All grades | 181/751 | 194/1018 | 1.17 (0.89–1.56) b | .006 | |
| Grade III-IV | 147/1179 | 245/2794 | 1.29 (1.06–1.56) b | .011 | |
| Sepsis | 134/720 | 319/2172 | 1.12 (0.94–1.34) a | .140 | |
| Death | 168/1013 | 379/2337 | 1.04 (0.77–1.41) b | .001 | |
| Bronchopulmonary dysplasia | 119/372 | 321/1141 | 1.12 (0.79–1.59) b | .015 | |
| Patent ductus arteriosus | 376/1031 | 871/2895 | 1.14 (0.97–1.35) b | .001 | |
| Necrotizing enterocolitis | 112/1090 | 256/3183 | 1.36 (1.08–1.71) b | .031 |
a Mantel-Haenzsel pooled relative risk, fixed effects model
b DerSimonian and Laird pooled relative risk, random effects model.
The results of the metaregression analysis did not alter the findings of the statistically significant increases in severe grade IVH, NEC, and PVL ( Table 3 ). Visual inspection of funnel plots did not identify evidence of publication bias.
| Neonatal outcome | Covariate | Coefficient (95% confidence interval) | P value | τ 2 a |
|---|---|---|---|---|
| Intraventricular hemorrhage grade III-IV | Gestational age at delivery | –0.71 (–1.57 to 0.14) | .09 | 1.87 |
| Birthweight | 0.004 (–0.003 to 0.01) | .21 | 1.35 | |
| Infant sex | 0.49 (–0.06 to 0.16) | .35 | 0.99 | |
| Antenatal steroid exposure | 0.02 (–0.04 to 0.08) | .52 | 0.68 | |
| Absence of other tocolytics in control subjects | 0.52 (–1.50 to 2.54) | .59 | 0.56 | |
| Delivery within 48 hr of last indomethacin dose | 0.76 (–1.66 to 3.20) | .51 | 0.68 | |
| Necrotizing enterocolitis | Gestational age at delivery | –0.41 (–2.04 to 1.22) | .56 | 0.62 |
| Birthweight | 0.002 (–0.01 to 0.02) | .72 | 0.37 | |
| Infant sex | 0.03 (–0.11 to 0.18) | .58 | 0.58 | |
| Antenatal steroid exposure | 0.01 (–0.05 to 0.08) | .63 | 0.50 | |
| Low mean dose of indomethacin b | 0.66 (–2.83 to 4.16) | .71 | 0.37 | |
| Periventricular leukomalacia | Gestational age at delivery | 4.73 (–16.15 to 25.62) | .43 | 0.98 |
| Birthweight | –0.03 (–0.17 to 0.11) | .43 | 0.98 | |
| Infant sex | –0.01 (–0.44 to 0.43) | .95 | 0.08 | |
| Delivery within 48 hr of last indomethacin dose | –0.13 (–1.93 to 1.67) | .89 | –0.14 |
a Smaller value of τ 2 indicates less between-study variability
Results
In addition to the overall total of 21 observational studies that previously were reported in the 2005 metaanalysis by Loe et al and in 2007 by Amin et al, 7 new observational studies that evaluated antenatal indomethacin tocolysis and its impact on neonatal outcomes and that met the inclusion criteria were identified. Randomized clinical trials were not included in this updated analysis. The study selection process is detailed in Figure 1 . One previously reported study was excluded because it contained data on postnatal indomethacin treatment only and was included inadvertently in the study by Loe. There were 2 new observational studies by Soraisham et al that included overlapping populations of infants; each of the studies looked at different primary outcomes but also reported on the secondary neonatal outcomes that were relevant to this study. Therefore, only the study that included the largest number of subjects was included. The characteristics of the studies that were included are detailed in Table 1 . The 27 observational studies that were included assessed 8454 infants overall, which is an increase of 2375 from the previously reported metaanalyses. Of these infants, 1731 were exposed to antenatal indomethacin, and 6723 were not exposed. The studies varied with respect to size, cause of preterm delivery, use of other tocolytics in the study groups and in the comparison groups, the dose and duration of indomethacin therapy, and the time from last dose of indomethacin to delivery.
| Study | Year | Infant, n | Study group tocolytic | Comparison group tocolytic | Gestational age, wk |
|---|---|---|---|---|---|
| Norton et al | 1993 | 114 | Indo ± Mg, β-mimetic | None, Mg, or β-mimetic | ≤30 |
| Souter et al | 1998 | 79 | Indo ± salbutamol or nifedipine | None, salbutamol, or nifedipine | ≤30 |
| Gardner et al | 1996 | 124 | Indo ± Mg, β-mimetic | None, Mg, or β-mimetic | <32 |
| Vermillion and Newman | 1999 | 225 | Indo + Mg | Mg | ≤32 |
| Van Overmeire et al | 1998 | 76 | Indo + β-mimetic | β-mimetic | ≤33 |
| Ojala et al | 2000 | 176 | Indo | Not Specified | <33 |
| Weintraub et al | 2001 | 2794 | Indo | None, Mg, or β-mimetic | ≤32 |
| Gerson et al | 1990 | 57 | Indo + Terb and Mg or ritodrine | Terbutaline + Mg or ritodrine | <33 |
| Abbasi et al | 2003 | 248 | Indo ± Mg, terbutaline | None, Mg or terbutaline | <34 |
| Niebyl and Witter | 1986 | 135 | Indo | None or “other” | ≤34 |
| Iannucci et al | 1996 | 56 | Indo + Mg | Mg | <30 |
| Suarez et al | 2001 | 70 | Indo | Mg | <33 |
| Major et al | 1994 | 759 | Indo ± Mg | Mg ± β-mimetic | <30 |
| Hammerman et al | 1998 | 105 | Indo | Not specified | <33 |
| Al-Alaiyan et al | 1996 | 30 | Indo | None | ≤33 |
| Parilla et al | 2000 | 110 | Indo | None or Mg | <37 |
| Baerts et al | 1990 | 159 | Indo + fenoterol | None or fenoterol | <30 |
| Pietrantoni et al | 1995 | 280 | Indo | Not specified | ≤32 |
| Friedman et al | 2005 | 236 | Indo | None | <32 |
| Murata et al | 2005 | 201 | Indo + ritodrine | Mg ± ritodrine | <33 |
| Doyle et al | 2005 | 549 | Indo + Mg | Mg | <34 |
| Baerts et al | 2013 | 36 | Indo | None or “other” | <30 |
| Cordero et al | 2007 | 116 | Indo ± Mg, terbutaline | Mg ± terbutaline | ≤28 |
| Amin et al | 2008 | 248 | Indo ± Mg | None or Mg | ≤29 |
| Soraisham et al | 2011 | 462 | Indo | Not specified | ≤28 |
| Sood et al | 2011 | 628 | Indo ± Mg | None or Mg | <32 |
| Sharma et al | 2010 | 381 | Indo | Not specified | <36 |
Stay updated, free articles. Join our Telegram channel
Full access? Get Clinical Tree
