Tocolytic Therapy a Meta-Analysis and Decision Analysis

Home , Ritodrine, Terbutaline, Tocolytic

Tocolytic Therapy A Meta-Analysis and Decision Analysis

David M. Haas, MD, MS, Thomas F. Imperiale, MD, Page R. Kirkpatrick, Robert W. Klein, Terrell W. Zollinger, DrPH, and Alan M. Golichowski, MD, PhD

OBJECTIVE: To determine the optimal first-line tocolytic ing prostaglandin inhibitors, only 80 would deliver within agent for treatment of premature labor. 48 hours, compared with 182 for the next-best treatment. METHODS: We performed a quantitative analysis of ran- CONCLUSION: Although all current tocolytic agents domized controlled trials of tocolysis, extracting data on were superior to no treatment at delaying delivery for maternal and neonatal outcomes, and pooling rates for both 48 hours and 7 days, prostaglandin inhibitors were each outcome across trials by treatment. Outcomes were superior to the other agents and may be considered the delay of delivery for 48 hours, 7 days, and until 37 weeks; optimal first-line agent before 32 weeks of gestation to adverse effects causing discontinuation of therapy; absence delay delivery. of respiratory distress syndrome; and neonatal survival. We (Obstet Gynecol 2009;113:585–94) used weighted proportions from a random-effects meta- analysis in a decision model to determine the optimal first-line tocolytic therapy. Sensitivity analysis was per- reterm birth, defined as any birth before the gesta- formed using the standard errors of the weighted propor- Ptional age of 37 weeks, is responsible for most of the 1–3 tions. neonatal morbidity and mortality in the United States and consumes 35% of all U.S. healthcare spending on RESULTS: Fifty-eight studies satisfied the inclusion crite- 4 ria. A random-effects meta-analysis showed that all to- infants. In the United States, 12.7% of infants are born colytic agents were superior to placebo or control groups preterm, totaling more than a half million births in 2 at delaying delivery both for at least 48 hours (53% for 2005. placebo compared with 75–93% for tocolytics) and 7 days To mitigate both maternal and neonatal risks (39% for placebo compared with 61–78% for tocolytics). resulting from preterm birth, current practice is to No statistically significant differences were found for the delay delivery for as long as possible.5 In extremely other outcomes, including the neonatal outcomes of low birth weight infants, a delay of 1 week decreases respiratory distress and neonatal survival. The decision neonatal mortality by 30%6 and allows opportunity to model demonstrated that prostaglandin inhibitors pro- transfer the mother to a tertiary care facility with a vided the best combination of tolerance and delayed neonatal intensive care unit and to administer ante- delivery. In a hypothetical cohort of 1,000 women receiv- natal corticosteroids.1 Tocolytic agents delay births caused by preterm From the Departments of Obstetrics & Gynecology and Internal Medicine, labor. However, no one tocolytic has been identified Division of Gastroenterology, Indiana University School of Medicine; Regenstrief as the best first-line option.1 Risks and benefits of all Institute, Inc.; and Medical Decision Modeling, Inc., Indianapolis, Indiana. tocolytic options for both the fetus and the mother Supported in part by grant K24DK02756 from the National Institutes of Health 7 to Dr. Imperiale. must be considered. Multiple Cochrane systematic 7–11 The authors thank Patrick O. Monahan, PhD, and George Eckert for their reviews exist for individual tocolytic medications, statistical expertise and Ronald Wielage, MPH, for DA programming. but there has been no rigorous, quantitative synthesis Corresponding author: David M. Haas, MD, MS, Department of OB/GYN, of the data comparing tocolytic drug classes. Wishard Memorial Hospital, 1001 West 10th Street, F-5, Indianapolis, IN The objective of this study was to determine the 46202; e-mail: [email protected]. optimal first-line tocolytic agent for the specific ma- Financial Disclosure ternal and neonatal outcomes based on the existing The authors did not report any potential conflicts of interest. literature. The optimal agent would combine the © 2009 by The American College of Obstetricians and Gynecologists. Published by Lippincott Williams & Wilkins. highest tolerability and the highest proportion of ISSN: 0029-7844/09 delayed delivery.

VOL. 113, NO. 3, MARCH 2009 OBSTETRICS & GYNECOLOGY 585 MATERIALS AND METHODS lide, ketorolac, rofecoxib, celecoxib, and mefenamic This project was approved by the Indiana University- acid. Nitrates included nitroglycerin and glyceryl Purdue University Indianapolis-Clarian Institutional trinitrate. We did not control for clinical heterogene- Review Board. We used the Quality of Reporting of ity in the medication dose and schedule. Meta-analyses (QUOROM) statement as a guideline We assessed articles for randomization allocation for conducting this analysis.12 QUOROM provides a using the Cochrane Collaboration A, B, C criteria. In standardized approach to performing and reporting a an effort to limit selection bias, we included articles meta-analysis. with an allocation score of “A” (the assigned treat- We searched the following computerized databases ment was adequately concealed) or “B” (unclear using the terms “preterm labor,” “tocolytic,” and “ob- whether the treatment assignment was adequately stetric labor, premature”: MEDLINE (1950–present), concealed), and excluded articles with an allocation MEDLINE In-Process (January 2008), EMBASE (1988– score of “C” (assigned treatment was not adequately 13 We did not exclude trials for a lack of 2008), The Cochrane Database of Clinical Trials (4th concealed). investigator-blinding postallocation when comparing quarter 2007), and CINAHL (1982–2008). We limited oral with intravenous tocolytic agents because the the search to articles reporting randomized controlled outcomes were objective in nature and reduced the effect trials in humans. Duplicate trial entries were ex- of bias due to a lack of blinding. To increase the cluded. We performed the search in January 2008. To clinical homogeneity among the study groups, we ensure completeness, we cross-referenced our search excluded studies with a mean gestational age of results with the Cochrane Reviews concerning toco- participants at randomization of less than 28 weeks or lytic medications. We read abstracts of titles that 33 weeks or more. If the mean gestational ages of both appeared relevant and then obtained the full text comparison groups in a study were 28 weeks or more articles of abstracts that appeared to fit the topic. and less than 33 weeks but were statistically signifi- Articles were reviewed by two authors (D.H. and cantly different between the groups, we extracted data P.K.), who read the articles and extracted data from on tocolytic efficacy and adverse effects but not on those that satisfied the study entry criteria. Discor- neonatal outcomes. dance between the two authors was resolved by Two authors independently extracted data, which consensus. Abstracts for articles in non-English lan- included allocation quality, presence of blinding, mean guages were reviewed. If the article seemed relevant gestational ages, interventions compared, use of antena- to the review, the full text was obtained and translated tal corticosteroids, and study entry criteria. Outcomes for possible data extraction. Six articles were trans- data extracted included the numbers of participants who lated (three in Chinese and one each in French, had delivery delayed by 48 hours, by 7 days, and until German, and Spanish). Published abstracts alone 37 weeks of gestation, as well as the number of women were not included because insufficient information who had medication adverse effects severe enough to was provided to conduct the quantitative analysis. discontinue the drug or to switch to another drug. If the We included randomized controlled trials that authors of the trial stated that antenatal corticosteroids reported a comparison between different medications were used, we included neonatal outcomes of the pres- or between a medication and a placebo or usual care. ence of respiratory distress syndrome (RDS) and neona- We included trials comparing tocolytic drugs in the tal death. Because current recommended practice in- same class (ie, two betamimetics like ritodrine com- cludes use of antenatal corticosteroids to accelerate fetal pared with terbutaline) but excluded trials that only lung maturity,14 we believe that neonatal outcomes compared different doses of the same agent. Interven- reported in studies that did not use this therapy tions were grouped into categories of control, betami- would not be applicable to the current standard of metics, calcium-channel blockers, magnesium sulfate, care. To clarify this issue, we attempted e-mail nitrates, oxytocin receptor antagonists, and prosta- contact with the authors of studies that did not glandin inhibitors. Control treatments included pla- contain explicit statements regarding the use of cebo treatments, bed rest, intravenous fluids, and antenatal corticosteroids. If we were unable to usual care. Betamimetic drugs included ritodrine, clarify, neonatal outcomes were not extracted. terbutaline, hexoprenaline, isoxsuprine, nydrilin, sal- Study participants enrolled in the trials were butamol, and fenoterol. Calcium-channel blockers pregnant women diagnosed with preterm labor or included nifedipine and nicardipine. Oxytocin recep- threatened preterm delivery. When results were strat- tor antagonists included only atosiban. Prostaglandin ified based on membrane status or the presence inhibitors included indomethacin, sulindac, nimesu- of multiple gestation, data were extracted only for

586 Haas et al Tocolytic Therapy OBSTETRICS & GYNECOLOGY women with intact membranes and singleton preg- fects, proportion of neonates with RDS, and neonatal nancies. For studies that did not stratify data, compos- death. The first chance node of the decision tree was ite data were extracted. “adverse effects requiring discontinuation of the med- Data were extracted for the outcomes identified ication.” If the patient had to stop or switch medica- and combined by drug category to calculate a tion, we assumed that it would not be considered an weighted mean and standard error for proportions effective first-line choice; the base case analysis con- of successful outcomes using rmeta library software sidered stopping medication to be a failure. After (2.14) for the statistical software R (2.5.1). Because medication tolerance, the next node indicated clinical we aggregated data from individual trials according outcome (eg, delaying delivery for 48 hours, RDS, to treatment group, effectively disassembling the trials, etc). A probabilistic sensitivity analysis was performed we generated weighted proportions based on the num- to determine how frequently each comparator treat- ber of subjects in each study. Using the DerSimonian- ment was most preferred for each outcome. We used Laird random-effects model, we compared each inter- the standard error of each base case proportion in vention to control, computing proportions and 95% distributions of nodal branching probabilities. One confidence intervals for the rates of successful outcomes.15 Because disassembling the trials precluded the thousand samples were generated for each outcome, direct comparisons required for odds ratios and Forest and the proportions where each treatment had the plots, neither was generated. lowest failure rate were determined. After completing the meta-analysis, we con- Figure 1 depicts the decision tree combining the structed a decision tree using TreeAge Pro 2007 chance nodes for tolerability and subsequent delay of software (TreeAge Software, Inc. Williamstown, MA) delivery to at least 48 hours after admission. All to determine whether one tocolytic class of medica- base-case models had the same structure, with varying tion was superior to others. A superior tocolytic agent outcomes but with outcome-specific probabilities. An would have the highest efficacy-to-toxicity ratio. The alternative, intention-to-treat approach placed chance weighted means and confidence intervals of the nodes after intolerance, assuming that intolerance pooled studies from the meta-analysis were used to resulted in the same failure rates as placebo. An represent the probabilities of: delaying delivery by 48 additional analysis was performed in which delaying hours, 7 days, until 37 weeks of gestation, proportion to 7 days was conditional on successfully delaying of women discontinuing therapy due to adverse ef- delivery for 48 hours.

48-hour delay Success Tolerate No 48-hour delay Observation/placebo Failure Intolerant Failure

48-hour delay Success Tolerate Betamimetics No 48-hour delay Failure Intolerant Failure Choose treatment 48-hour delay Success Tolerate No 48-hour delay Calcium channel blockers Failure Intolerant Failure

Magnesium sulfate

Oxytocin antagonists

Prostaglandin inhibitors Fig. 1. Decision tree for a woman between 28 and 33 weeks of gestation presenting with premature labor. The decision node involves choice of tocolytic therapy. The first chance node is intolerability requiring discontinuation of the drug. The second chance node is delay of delivery for 48 hours. For illustrative purposes, only the first three treatment choices are fully branched out. All illustrated steps were followed for each treatment option. Haas. Tocolytic Therapy. Obstet Gynecol 2009.

VOL. 113, NO. 3, MARCH 2009 Haas et al Tocolytic Therapy 587 Potentially relevant RCTs steps to the meta-analysis are summarized in Figure 2. identified at initial search and Table 1 lists the studies in the final analysis. Among screened for retrieval N=2,715 the included studies, 10 contained data on a placebo or control arm,16–25 39 reported results for betamimet- Excluded based on title ics,16–18,22,24,26–59 20 reported results for calcium-chan- or abstract, which were 28,32,35,39–42,44,46,47,51,52,55,58–64 not relevant nel blockers, 19 reported re- n=2,579 sults for magnesium sulfate,16,25,29,38,48,54,56,57,60,61,63–71 8 reported results for oxytocin receptor antago- RCTs retrieved for more 20,26,27,36,39,50,53,62 detailed evaluation nists, 12 reported results for prosta- n=136 glandin inhibitors,19,23,30,43,45,49,65,68,70–73 and 3 reported results using nitrates.21,31,66 A total of 16 different Studies excluded from analysis: n=78 head-to-head comparisons were made among the Not RCT: 9 trials. Aggregated trial group characteristics are Wrong patient Inclusion criteria: population: 7 shown in Table 1. We were unable to confirm ante- Pregnant women with Wrong treatments: 10 natal corticosteroid use for 20 trials. Data from sub- preterm labor Wrong outcome RCT comparing two different measures: 10 jects using nitrates were not included due to a paucity drugs or a drug to Combination therapies: 7 of studies with all data points. placebo or No listed mean EGA: 7 Aggregated proportions and 95% confidence in- standard care Mean EGA of 33 Allocation assessment weeks or more: 5 tervals for each outcome are shown in Table 2 and “A” or “B” Mean EGA of less Figure 3. All tocolytic agents were superior to placebo Mean EGA between than 28 weeks: 2 28 and 33 weeks Allocation score “C”: 6 or control groups at delaying delivery for at least 48 Mean group EGAs not Published abstracts or hours and for at least 7 days. However, none of them significantly different unable to locate full article: 12 was superior statistically to placebo or controls for Cochrane Review cited delay of delivery to 37 weeks of gestation. The 95% personal communications: 3 confidence intervals for rates of RDS overlapped with the placebo or control group for all tocolytic drugs, RCTs with useable information although the overlap for betamimetics and prosta- included in the meta-analysis glandin inhibitors was minimal. Rates of neonatal n=58 death were low and were not significantly different Fig. 2. Summary of stages of study inclusion and exclusion among treatment groups. The proportion of women for the meta-analysis. RCT, randomized controlled trial; experiencing adverse effects that required discontinu- EGA, estimated gestational age. ing the medication was similar for all groups except Haas. Tocolytic Therapy. Obstet Gynecol 2009. for betamimetics, which had a significantly higher rate of discontinuation. RESULTS Table 3 displays the results of the decision anal- We retrieved 136 full-text articles, of which 58 satis- ysis. The individual treatment options are compared fied the study inclusion and exclusion criteria. The for each outcome to determine which agent might be

Table 1. Aggregated Trial Characteristics Listed by Treatment Trials With Total No. of Mean Gestational Trials Reporting on at Trials With Cochrane Treatment Data Participants Age (wk) Least 5 of 6 Outcomes “A” Allocation Score Placebo/control 10 904 30.6 3 (30) 7 (70) Betamimetics 39 2,567 29.6 15 (38) 26 (66) Calcium-channel 20 868 30.6 10 (50) 12 (60) blocker Magnesium sulfate 19 935 30.9 2 (10) 14 (74) Oxytocin receptor 8 1,249 30.1 5 (62) 7 (88) antagonists Prostaglandin 12 442 30.2 0 (0)* 10 (83) inhibitors Nitrates 3 211 30.3 1 (33) 3 (100) Data are n (%). A trial comparing two drugs will be represented in two lines above and the number in each arm will contribute to the total number of participants. * Two trials (17%) using prostaglandin inhibitors reported on four of the extracted outcomes.

588 Haas et al Tocolytic Therapy OBSTETRICS & GYNECOLOGY Table 2. Weighted Percentages of Tocolytic Agents for Both Efficacy and Toxicity Delay of Delivery Neonates Neonatal Adverse Drug 48 h 7 d After 37 wk With RDS Death Effects Placebo/control 53 (45–61) [9] 39 (28–49) [8] 36 (20–52) [3] 21 (17–26) [3] 2 (0–5) [3] 1 (0–2) [6] Betamimetics 75 (65–85) [29] 65 (59–71) [26] 46 (36–56) [15] 13 (8–18) [17] 2 (1–3) [20] 14 (9–18) [32] Calcium-channel blocker 76 (57–95) [17] 62 (56–69) [10] 47 (32–62) [12] 19 (4–33) [11] 1 (0–3) [12] 1 (0–3) [16] Magnesium sulfate 89 (85–93) [11] 61 (39–84) [5] 42 (31–53) [7] 16 (11–20) [9] 1 (0–2) [9] 3 (1–6) [16] Oxytocin receptor antagonists 86 (80–91) [8] 78 (68–88) [6] No data 14 (8–21) [5] 1 (0–2) [6] 2 (0–5) [6] Prostaglandin inhibitors 93 (90–95) [8] 76 (67–85) [3] 43 (6–79) [4] 11 (4–18) [4] 2 (0–4) [4] 0 (0–2) [6] RDS, respiratory distress syndrome. Data are % (95% confidence interval) of women experiencing the outcome and [number of studies reporting the outcome]. Adverse effects are those that required discontinuation of the medication. considered the optimal first-line treatment. The deci- blockers were the superior agent. To enhance the sion model shows that prostaglandin inhibitors pro- clinical relevance of the analysis, a hypothetical co- vide superior results for all outcomes except delaying hort of 1,000 women were simulated, and the number delivery until 37 weeks, where calcium-channel of failures for the individual therapies was calculated

Fig. 3. Weighted percentages and 95% confidence intervals of success rates for tocolytic agents for both efficacy and toxicity outcomes from meta-analysis. RDS, respiratory distress syndrome. Haas. Tocolytic Therapy. Obstet Gynecol 2009.

VOL. 113, NO. 3, MARCH 2009 Haas et al Tocolytic Therapy 589 Table 3. Results of Decision Analysis Proportion Tolerating No. Failure Proportion of Simulations Group Treatment and Achieving Outcome Events/1,000 Women Where Treatment Is Best 48-h delay Placebo 0.584 416 0.000 Betamimetics 0.662 338 0.000 Calcium-channel blockers 0.718 282 0.004 Magnesium 0.818 182 0.000 Oxytocin antagonists 0.798 202 0.007 Prostaglandin inhibitors 0.920 80 0.989 7-d delay Placebo 0.485 515 0.000 Betamimetics 0.585 415 0.000 Calcium-channel blockers 0.611 389 0.000 Magnesium 0.667 333 0.163 Oxytocin antagonists 0.703 297 0.198 Prostaglandin inhibitors 0.750 250 0.639 Contingent* Placebo 0.485 515 0.000 Betamimetics 0.543 457 0.000 Calcium-channel blockers 0.603 397 0.027 Magnesium 0.630 370 0.041 Oxytocin antagonists 0.734 266 0.488 Prostaglandin inhibitors 0.736 264 0.444 Delay to 37 wk Placebo 0.337 663 0.005 Betamimetics 0.421 579 0.123 Calcium-channel blockers 0.466 534 0.519 Magnesium 0.404 596 0.078 Oxytocin antagonists — NA NA Prostaglandin inhibitors 0.370 630 0.275 Absence of RDS Placebo 0.782 218 0.003 Betamimetics 0.722 278 0.000 Calcium-channel blockers 0.815 185 0.141 Magnesium 0.780 220 0.012 Oxytocin antagonists 0.788 212 0.043 Prostaglandin inhibitors 0.870 130 0.801 Neonatal survival Placebo 0.965 35 0.190 Betamimetics 0.834 166 0.000 Calcium blockers 0.946 54 0.060 Magnesium 0.920 80 0.013 Oxytocin antagonists 0.936 64 0.088 Prostaglandin inhibitors 0.977 23 0.649 NA, not applicable; RDS, respiratory distress syndrome. Successful outcome is defined as the proportion of women treated who both tolerated the treatment (did not have adverse effects requiring stopping the medication) and achieved the desired outcome. Proportion of 1,000 samples with the fewest failures was determined by probabilistic sensitivity analysis. * Seven-day delay from studies that also reported a 48-hour delay. for each outcome. Only 80 of 1,000 women treated delaying delivery to 37 weeks of gestation, where initially with prostaglandin inhibitors would deliver calcium-channel blockers were superior. For the within 48 hours, as compared with 182 to 416 for 7-day contingent outcome, oxytocin antagonists and other treatments. An intent-to-treat sensitivity analysis prostaglandin inhibitors were essentially equivalent. did not substantially change the results. The probabilistic sensitivity analysis showed that the treatment DISCUSSION rankings of the alternatives were robust, with prosta- Deciding which tocolytic agent to use as the first-line glandin inhibitors most frequently yielding the lowest drug is a difficult decision for clinicians. This quanti- number of failures for each outcome except for tative analysis demonstrated that all tocolytic drugs

590 Haas et al Tocolytic Therapy OBSTETRICS & GYNECOLOGY were superior to placebo at delaying delivery for 48 testing by Myers et al79 found that treating with hours and 7 days, although not at delaying delivery tocolytic medication (the model assumed betamimet- until 37 weeks. No significant therapeutic differences ics) was preferred over fetal lung maturity testing were seen in the outcomes of RDS or neonatal death. under 34 weeks of gestation. While these analyses Our analysis suggests that prostaglandin inhibitors attempted to answer a question about the preferred may be the superior first-line tocolytic agent because treatment strategy, our analysis went further by con- of high tolerability and effectiveness at delaying de- sidering all commonly used tocolytic drug options. livery by at least 7 days. Delaying delivery long Additionally, our analysis included many recently enough to administer antenatal corticosteroids is piv- reported trials and several foreign language trials not otal to improving neonatal outcomes.74 included in older reviews. Prostaglandin inhibitors have been used safely in Our analysis is limited by the data presented in the mid trimester for many years. However, there is the studies obtained. We were unable to use the concern about their use after 32 weeks of gestation neonatal outcome data for several studies that either due to the risk of premature closure of the fetal ductus did not state the use of or did not use antenatal arteriosus.5 A retrospective study of 57 infants whose corticosteroids. Although we attempted to obtain this mothers were treated with indomethacin at or before information, we were unable to do so for several trials. 30 weeks showed a higher rate of necrotizing entero- This limitation may affect the validity of our findings colitis, intracranial hemorrhage, and patent ductus for RDS and neonatal death. The proportion of arteriosus.75 However, the Cochrane Review for this occurrence of these outcomes, however, is relatively class of drugs failed to demonstrate a statistically consistent among studies, suggesting that the data we significant increase in any adverse neonatal out- have for RDS and neonatal death are representative comes.8 Because our analysis was limited to studies of this literature. These neonatal outcomes are the with fetuses of mean gestational ages between 28 desired endpoints. However, no tocolytic improved weeks and 32 weeks, the combination of tolerability these outcomes compared with controls. Perhaps if and efficacy makes prostaglandin inhibitors seem to the meta-analysis were performed for studies report- be the superior first-line tocolytic therapy. One ing outcomes for pregnancies less than 28 weeks, reason why prostaglandin inhibitors may be supe- when these neonatal outcomes are more prevalent, rior is the large proportion of cases of preterm labor differences in individual tocolytic classes might be that are associated with inflammation and subclin- present. We did not stratify the trials by medication ical infection.76 dosage used. Although there is variation in treatment We are unaware of another combined meta- regimen among the trials, drug dose and schedules analysis and decision analysis designed to determine were similar to commonly used doses and schedules. the optimal first-line tocolytic drug. A decision anal- Using weighted proportions helped minimize the ysis by Macones et al77 discussed preterm labor contribution of smaller trials that used less common management strategies at different gestational ages, dosing strategies. Our decision analysis was a simple starting at 32 weeks. These investigators found that at model of tolerability and outcome. Tocolytic thera- 32 weeks, tocolysis was superior to no tocolysis or pies vary in their costs. Our analysis did not consider amniocentesis for fetal lung maturity; at 34 weeks, cost of the medications or the cost of administration of tocolysis and no tocolysis yielded equal outcomes; the medications. A future analysis may include the and at 36 weeks, no tocolysis was the preferred costs of the therapeutic options and adverse events in strategy. Their analysis focused on ritodrine for toco- the decision model. Standard utility estimates for lysis. As demonstrated in our analysis, betamimetics various obstetric and neonatal outcomes are lacking were found to have the highest rate of adverse effects in the literature. Ascertaining utilities for the out- requiring discontinuation, which may limit their de- comes of preterm delivery would also allow for a sirability as a first-line agent. Similar to Macones et al, richer decision tree. we found tocolysis superior to no tocolysis in a Our analysis deconstructed the individual trials and gestational age range from 28–32 weeks, but we also aggregated the data by treatment arm. This methodol- assessed a variety of tocolytic medications. A cost- ogy has been reported for other conditions with multiple effectiveness analysis performed by Ferriols Lisart treatment options80–82 and is a practical approach to and colleagues78 found that using ritodrine as the pooling data across trials comparing different interven- first-line agent with atosiban as a rescue agent was the tions. Because generating individual odds ratios for each more cost-effective option. A cost-effectiveness anal- of the 16 different paired comparisons was impractical, ysis of tocolysis compared with fetal lung maturity this disassembling of trials was necessary. Thus, there

VOL. 113, NO. 3, MARCH 2009 Haas et al Tocolytic Therapy 591 were no “paired” groups with which to generate odds 3. McCormick MC. The contribution of low birth weight to infant mortality and childhood morbidity. N Engl J Med ratios or Forest plots for the outcomes. This disassembly 1985;312:82–90. of trials, however, did limit the available diagnostic 4. Lewit EM, Baker LS, Corman H, Shiono PH. The direct cost capabilities in the software. An indirect comparison of low birth weight. Future Child 1995;5:35–56. meta-analysis (also known as multiple treatment meta- 5. Goldenberg RL. The management of preterm labor. Obstet analysis or network meta-analysis) would be a method Gynecol 2002;100:1020–37. to attempt meta-analysis while not deconstructing the 6. Finnstrom O, Olausson PO, Sedin G, Serenius F, Svenningsen 83 N, Thiringer K, et al. The Swedish national prospective study trials. An indirect comparison analysis has the poten- on extremely low birthweight (ELBW) infants. Incidence, tial to generate more precise estimates of effect. This mortality, morbidity and survival in relation to level of care. type of analysis carries with it other sets of assumptions, Acta Paediatr 1997;86:503–11. however, and is beyond the scope of the current analy- 7. Anotayanonth S, Subhedar NV, Garner P, Neilson JP, Hari- gopal S. Betamimetics for inhibiting preterm labour. The sis. Although a clinical trial comparing six treatments Cochrane Database of Systematic Reviews 2004, Issue 4. Art. would be a more rigorous approach to answer the No.: CD004352. DOI: 10.1002/14651858.CD004352.pub2. research question, there are logistical limitations to con- 8. King J, Flenady V, Cole S, Thornton S. Cyclo-oxygenase ducting such a trial, not the least of which is the sample (COX) inhibitors for treating preterm labour. The Cochrane Database of Systematic Reviews 2005, Issue 2. Art. No.: size requirement. Accounting for multiple comparisons, CD001992. DOI: 10.1002/14651858.CD001992.pub2. a six-armed trial would need nearly 2,000 subjects in 9. King JF, Flenady VJ, Papatsonis DN, Dekker GA, Carbonne B. each arm to achieve adequate power to determine a Calcium channel blockers for inhibiting preterm labour. The statistically significant difference in delayed delivery Cochrane Database of Systematic Reviews 2003, Issue 1. Art. No.: CD002255. DOI: 10.1002/14651858.CD002255. until 37 weeks of the magnitude observed in our analy- 10. Crowther CA, Hiller JE, Doyle LW. Magnesium sulphate for sis. The random-effects model analyzes variance within preventing preterm birth in threatened preterm labour. the individual treatment arms, not by individual study Cochrane Database of Systematic Reviews 2006, Issue 4. Art. and accounts for some of the individual trial variation. No.: CD001060. DOI: 10.1002/14651858.CD001060. Table 1 demonstrates that the treatment arms were of 11. Papatsonis D, Flenady V, Cole S, Liley H. Oxytocin receptor antagonists for inhibiting preterm labour. Cochrane Database similar mean gestational ages and had similar propor- of Systematic Reviews 2005, Issue 3. Art. No.: CD004452. tions of trials of the highest quality. Thus, a meta- DOI: 10.1002/14651858.CD004452.pub2. regression controlling for these factors was not per- 12. Moher D, Cook DJ, Eastwood S, Olkin I, Rennie D, Stroup DF. Improving the quality of reports of meta-analyses of formed. A meta-regression would not analyze the direct randomised controlled trials: the QUOROM statement. Qual- effect of a covariate on an individual subject’s outcome ity of Reporting of Meta-analyses. Lancet 1999;354:1896–900. and would add little to the random-effects model used to 13. Higgins JP, Green S, editors. Assessment of study quality. compare the aggregated data. Cochrane handbook for systematic reviews of interventions 4.2.5 [updated May 2005]. The Cochrane Library, Issue 3. In conclusion, this analysis suggests that tocolytic Section 6 ed. Chichester (UK): John Wiley & Sons, Ltd; 2005. drugs are superior to placebo or control at delaying 14. American College of Obstetricians and Gynecologists Com- delivery by 48 hours and 7 days. There is little difference mittee on Obstetric Practice. ACOG Committee Opinion No. among treatments in RDS or neonatal death. The 402: Antenatal corticosteroid therapy for fetal maturation. decision analysis demonstrated that prostaglandin inhib- Obstet Gynecol 2008;111:805–7. 15. DerSimonian R, Laird N. Meta-analysis in clinical trials. itors may be the superior first-line tocolytic agents before Control Clin Trials 1986;7:177–88. 32 weeks of gestation to delay delivery for 48 hours and 16. Cotton DB, Strassner HT, Hill LM, Schifrin BS, Paul RH. 7 days, whereas calcium-channel blockers may be supe- Comparison of magnesium sulfate, terbutaline and a placebo rior first-line agents to delay delivery until 37 weeks of for inhibition of preterm labor. A randomized study. J Reprod gestation. These agents have the best combination of Med 1984;29:92–7. tolerability and efficacy and should be considered the 17. Ingemarsson I. Effect of terbutaline on premature labor. A double-blind placebo-controlled study. Am J Obstet Gynecol best choices for first-line tocolysis, taking into account 1976;125:520–4. maternal and fetal factors that might influence the 18. Leveno KJ, Klein VR, Guzick DS, Young DC, Hankins GD, choice of tocolytic agent. Williams ML. Single-centre randomised trial of ritodrine hydrochloride for preterm labour. Lancet 1986;1:1293–6. 19. Niebyl JR, Blake DA, White RD, Kumor KM, Dubin NH, REFERENCES Robinson JC, et al. The inhibition of premature labor with 1. ACOG Committee on Practice Bulletins–Obstetrics. ACOG indomethacin. Am J Obstet Gynecol 1980;136:1014–9. practice bulletin. Management of preterm labor. Number 43, 20. Romero R, Sibai BM, Sanchez-Ramos L, Valenzuela GJ, May 2003. Int J Gynaecol Obstet 2003;82:127–35. Veille JC, Tabor B, et al. An oxytocin receptor antagonist 2. Martin JA, Hamilton BE, Sutton PD, Ventura SJ, Menacker F, (atosiban) in the treatment of preterm labor: a randomized, Kirmeyer S, et al. Births: final data for 2005. Natl Vital Stat double-blind, placebo-controlled trial with tocolytic rescue. Rep 2007;56:1–103. Am J Obstet Gynecol 2000;182:1173–83.

592 Haas et al Tocolytic Therapy OBSTETRICS & GYNECOLOGY 21. Smith GN, Walker MC, Ohlsson A, O’Brien K, Windrim R. 39. Husslein P, Cabero Roura L, Dudenhausen JW, Helmer H, Randomized double-blind placebo-controlled trial of transder- Frydman R, Rizzo N, et al. Atosiban versus usual care for the mal nitroglycerin for preterm labor. Am J Obstet Gynecol management of preterm labor. J Perinat Med 2007;35:305–13. 2007;196:37 e1–8. 40. Janky E, Leng JJ, Cormier PH, Salamon R, Meynard J. A 22. Spellacy WN, Cruz AC, Birk SA, Buhi WC. Treatment of randomized study of the treatment of threatened premature premature labor with ritodrine: a randomized controlled study. labor. Nifedipine versus ritodrine [in French]. J Gynecol Obstet Gynecol 1979;54:220–3. Obstet Biol Reprod (Paris) 1990;19:478–82. 23. Zuckerman H, Shalev E, Gilad G, Katzuni E. Further study of 41. Jannet D, Abankwa A, Guyard B, Carbonne B, Marpeau L, the inhibition of premature labor by indomethacin. Part II Milliez J. Nicardipine versus salbutamol in the treatment of double-blind study. J Perinat Med 1984;12:25–9. premature labor. A prospective randomized study. Eur J 24. Treatment of preterm labor with the beta-adrenergic agonist Obstet Gynecol Reprod Biol 1997;73:11–6. ritodrine. The Canadian Preterm Labor Investigators Group. 42. Koks CA, Brolmann HA, de Kleine MJ, Manger PA. A N Engl J Med 1992;327:308–12. randomized comparison of nifedipine and ritodrine for sup- 25. Cox SM, Sherman ML, Leveno KJ. Randomized investigation pression of preterm labor. Eur J Obstet Gynecol Reprod Biol of magnesium sulfate for prevention of preterm birth. Am J 1998;77:171–6. Obstet Gynecol 1990;163:767–72. 43. Kramer WB, Saade GR, Belfort M, Dorman K, Mayes M, 26. European Atosiban Study Group. The oxytocin antagonist Moise KJ Jr. A randomized double-blind study comparing the atosiban versus the beta-agonist terbutaline in the treatment of fetal effects of sulindac to terbutaline during the management preterm labor. A randomized, double-blind, controlled study. of preterm labor. Am J Obstet Gynecol 1999;180:396–401. Acta Obstet Gynecol Scand 2001;80:413–22. 44. Kupferminc M, Lessing JB, Yaron Y, Peyser MR. Nifedipine 27. French/Australian Atosiban Investigators Group. Treatment of versus ritodrine for suppression of preterm labour. Br J Obstet preterm labor with the oxytocin antagonist atosiban: a double- Gynaecol 1993;100:1090–4. blind, randomized, controlled comparison with salbutamol. Eur J Obstet Gynecol Reprod Biol 2001;98:177–85. 45. Kurki T, Eronen M, Lumme R, Ylikorkala O. A randomized double-dummy comparison between indomethacin and nylid- 28. Al-Qattan F, Omu AE, Labeeb N. A prospective randomized rin in threatened preterm labor. Obstet Gynecol 1991;78: study comparing nifedipine versus ritodrine for the suppres- 1093–7. sion of preterm labour. Med Principles Pract 2000;9:164–73. 46. Laohapojanart N, Soorapan S, Wacharaprechanont T, Ratana- 29. Beall MH, Edgar BW, Paul RH, Smith-Wallace T. A compar- jamit C. Safety and efficacy of oral nifedipine versus terbutal- ison of ritodrine, terbutaline, and magnesium sulfate for the ine injection in preterm labor. J Med Assoc Thai 2007;90: suppression of preterm labor. Am J Obstet Gynecol 1985;153: 2461–9. 854–9. 30. Besinger RE, Niebyl JR, Keyes WG, Johnson TR. Randomized 47. Mawaldi L, Duminy P, Tamim H. Terbutaline versus nifedi- comparative trial of indomethacin and ritodrine for the long- pine for prolongation of pregnancy in patients with preterm term treatment of preterm labor. Am J Obstet Gynecol 1991; labor. Int J Gynaecol Obstet 2008;100:65–8. 164:981–6. 48. Miller JM Jr, Keane MW, Horger EO 3rd. A comparison of 31. Bisits A, Madsen G, Knox M, Gill A, Smith R, Yeo G, et al. magnesium sulfate and terbutaline for the arrest of premature The Randomized Nitric Oxide Tocolysis Trial (RNOTT) for labor. A preliminary report. J Reprod Med 1982;27:348–51. the treatment of preterm labor. Am J Obstet Gynecol 2004; 49. Morales WJ, Smith SG, Angel JL, O’Brien WF, Knuppel RA. 191:683–90. Efficacy and safety of indomethacin versus ritodrine in the 32. Cararach V, Palacio M, Martinez S, Deulofeu P, Sanchez M, management of preterm labor: a randomized study. Obstet Cobo T, et al. Nifedipine versus ritodrine for suppression of Gynecol 1989;74:567–72. preterm labor. Comparison of their efficacy and secondary 50. Moutquin JM, Sherman D, Cohen H, Mohide PT, Hochner- effects. Eur J Obstet Gynecol Reprod Biol 2006;127:204–8. Celnikier D, Fejgin M, et al. Double-blind, randomized, con- 33. Caritis SN, Toig G, Heddinger LA, Ashmead G. A double- trolled trial of atosiban and ritodrine in the treatment of blind study comparing ritodrine and terbutaline in the treat- preterm labor: a multicenter effectiveness and safety study. ment of preterm labor. Am J Obstet Gynecol 1984;150:7–14. Am J Obstet Gynecol 2000;182:1191–9. 34. Essed GG, Eskes TK, Jongsma HW. A randomized trial of two 51. Papatsonis DN, Van Geijn HP, Ader HJ, Lange FM, Bleker beta-mimetic drugs for the treatment of threatening early OP, Dekker GA. Nifedipine and ritodrine in the management labor: clinical results in a prospective comparative study with of preterm labor: a randomized multicenter trial. Obstet ritodrine and fenoterol. Eur J Obstet Gynecol Reprod Biol Gynecol 1997;90:230–4. 1978;8:341–8. 52. Raymajhi R, Pratap K. A comparative study between nifedi- 35. Garcia-Velasco JA, Gonzalez Gonzalez A. A prospective, pine and isoxsuprine in the suppression of preterm labour. randomized trial of nifedipine vs. ritodrine in threatened Kathmandu Univ Med J (KUMJ) 2003;1:85–90. preterm labor. Int J Gynaecol Obstet 1998;61:239–44. 53. Shim JY, Park YW, Yoon BH, Cho YK, Yang JH, Lee Y, et al. 36. Goodwin TM, Valenzuela GJ, Silver H, Creasy G. Dose Multicentre, parallel group, randomised, single-blind study of ranging study of the oxytocin antagonist atosiban in the the safety and efficacy of atosiban versus ritodrine in the treatment of preterm labor. Atosiban Study Group. Obstet treatment of acute preterm labour in Korean women. BJOG Gynecol 1996;88:331–6. 2006;113:1228–34. 37. Gummerus M. Tocolysis with hexoprenalin and salbutamol in 54. Surichamorn P. The efficacy of terbutaline and magnesium a clinical comparison [in German]. Geburtshilfe Frauenheilkd sulfate in the management of preterm labor. J Med Assoc Thai 1983;43:151–5. 2001;84:98–104. 38. Hollander DI, Nagey DA, Pupkin MJ. Magnesium sulfate and 55. Weerakul W, Chittacharoen A, Suthutvoravut S. Nifedipine ritodrine hydrochloride: a randomized comparison. Am J versus terbutaline in management of preterm labor. Int J Obstet Gynecol 1987;156:631–7. Gynaecol Obstet 2002;76:311–3.

VOL. 113, NO. 3, MARCH 2009 Haas et al Tocolytic Therapy 593 56. Wilkins IA, Lynch L, Mehalek KE, Berkowitz GS, Berkowitz 70. Morales WJ, Madhav H. Efficacy and safety of indomethacin RL. Efficacy and side effects of magnesium sulfate and rito- compared with magnesium sulfate in the management of drine as tocolytic agents. Am J Obstet Gynecol 1988;159: preterm labor: a randomized study. Am J Obstet Gynecol 685–9. 1993;169:97–102. 57. Zhu B, Fu Y. Treatment of preterm labor with ritodrine [in 71. Schorr SJ, Ascarelli MH, Rust OA, Ross EL, Calfee EL, Perry Chinese]. Zhonghua Fu Chan Ke Za Zhi 1996;31:721–3. KG Jr, et al A comparative study of ketorolac (Toradol) and 58. Ferguson JE 2nd, Dyson DC, Schutz T, Stevenson DK. A magnesium sulfate for arrest of preterm labor. South Med J comparison of tocolysis with nifedipine or ritodrine: analysis of 1998;91:1028–32. efficacy and maternal, fetal, and neonatal outcome. Am J 72. Rasanen J, Jouppila P. Fetal cardiac function and ductus Obstet Gynecol 1990;163:105–11. arteriosus during indomethacin and sulindac therapy for 59. Fan L, Wu L, Huang X. The effect of calcium entry blocker on threatened preterm labor: a randomized study. Am J Obstet the management of preterm labor: a randomized controlled Gynecol 1995;173:20–5. study [in Chinese]. Chin J Pract Gynecol Obstet 2003;19:87–9. 73. Sawdy RJ, Lye S, Fisk NM, Bennett PR. A double-blind 60. Floyd RC, McLaughlin BN, Perry KG, Martin RW, Sullivan randomized study of fetal side effects during and after the CA, Morrison JC. Magnesium sulfate or nifedipine hydrochlo- short-term maternal administration of indomethacin, sulindac, ride for acute tocolysis of preterm labor: efficacy and side and nimesulide for the treatment of preterm labor. Am J effects. J Matern Fetal Invest 1995;5:25–9. Obstet Gynecol 2003;188:1046–51. 61. Glock JL, Morales WJ. Efficacy and safety of nifedipine versus 74. Crowley P. Prophylactic corticosteroids for preterm birth. The magnesium sulfate in the management of preterm labor: a Cochrane Database of Systematic Reviews 2000, Issue 2 Art. randomized study. Am J Obstet Gynecol 1993;169:960–4. No.: CD000065. DOI: 10.1002/14651858.CD000065.pub2. 62. Kashanian M, Akbarian AR, Soltanzadeh M. Atosiban and 75. Norton ME, Merrill J, Cooper BA, Kuller JA, Clyman RI. nifedipine for the treatment of preterm labor. Int J Gynaecol Neonatal complications after the administration of indometh- Obstet 2005;91:10–4. acin for preterm labor. N Engl J Med 1993;329:1602–7. 63. Larmon JE, Ross BS, May WL, Dickerson GA, Fischer RG, 76. Goldenberg RL, Hauth JC, Andrews WW. Intrauterine infec- Morrison JC. Oral nicardipine versus intravenous magnesium tion and preterm delivery. N Engl J Med 2000;342:1500–7. sulfate for the treatment of preterm labor. Am J Obstet 77. Macones GA, Bader TJ, Asch DA. Optimising maternal-fetal Gynecol 1999;181:1432–7. outcomes in preterm labour: a decision analysis. Br J Obstet 64. Lyell DJ, Pullen K, Campbell L, Ching S, Druzin ML, Chitkara Gynaecol 1998;105:541–50. U, et al. Magnesium sulfate compared with nifedipine for acute 78. Ferriols Lisart R, Nicolas Pico J, Alos Alminana M. Pharmaco- tocolysis of preterm labor: a randomized controlled trial. economic assessment of two tocolysis protocols for the inhibi- Obstet Gynecol 2007;110:61–7. tion of premature delivery [in Spanish]. Farm Hosp 2005;29: 65. Borna S, Saeidi FM. Celecoxib versus magnesium sulfate to 18–25. arrest preterm labor: randomized trial. J Obstet Gynaecol Res 79. Myers ER, Alvarez JG, Richardson DK, Ludmir J. Cost- 2007;33:631–4. effectiveness of fetal lung maturity testing in preterm labor. 66. El-Sayed YY, Riley ET, Holbrook RH Jr, Cohen SE, Chitkara Obstet Gynecol 1997;90:824–9. U, Druzin ML. Randomized comparison of intravenous nitro- 80. Imperiale TF, Speroff T. A meta-analysis of methods to glycerin and magnesium sulfate for treatment of preterm labor. prevent venous thromboembolism following total hip replace- Obstet Gynecol 1999;93:79–83. ment [published erratum appears in JAMA 1995;273:288]. 67. Lorzadeh N, Kazemirad S, Lorzadrh M, Dehnori A. A com- JAMA 1994;271:1780–5. parison of human chorionic gonadotropin with magnesium 81. Felson DT, Anderson JJ, Meenan RF. The comparative effi- sulphate in inhibition of preterm labor. J Med Sci 2007;7: cacy and toxicity of second-line drugs in rheumatoid arthritis. 640–4. Results of two metaanalyses. Arthritis Rheum 1990;33: 68. McWhorter J, Carlan SJ, OLeary TD, Richichi K, OBrien WF. 1449–61. Rofecoxib versus magnesium sulfate to arrest preterm labor: a 82. Bravata DM, Sanders L, Huang J, Krumholz HM, Olkin I, randomized trial [published erratum appears in Obstet Gardner CD, et al. Efficacy and safety of low-carbohydrate Gynecol 2004;104:200]. Obstet Gynecol 2004;103:923–30. diets: a systematic review. JAMA 2003;289:1837–50. 69. Mittendorf R, Covert R, Boman J, Khoshnood B, Lee KS, 83. Caldwell DM, Ades AE, Higgins JP. Simultaneous comparison Siegler M. Is tocolytic magnesium sulphate associated with of multiple treatments: combining direct and indirect evi- increased total paediatric mortality? Lancet 1997;350:1517–8. dence. BMJ 2005;331:897–900.

594 Haas et al Tocolytic Therapy OBSTETRICS & GYNECOLOGY