Tocolytic agents and the failure to improve neonatal outcomes

Abstract

Premature delivery has long been a major problem in obstetrics. Increased perinatal morbidity and mortality as well as substantial health care costs, both short and long term, have driven the efforts to combat this complication. While multiple agents have been studied over the last 70 years leading to hundreds of publications and involving tens of thousands of patients, no benefit to neonates has been demonstrated to date for any of these agents. Nonetheless, clinical use of reputed “tocolytic agents” has been incorporated in both the medical and lay literature as the standard of care. Unfortunately, use of medications without evidence of efficacy has only added to health care costs, caused harm in some cases, and rendered necessary placebo‐controlled trials difficult if not impossible. This clinical opinion will consider the stories of various attempts to identify an effective medication in order to demonstrate the shortcomings of the current approach to the study of possible tocolytic agents and suggest changes that may lead to a more rational approach to drug development.

Preterm birth has long been one of the major problems facing the practice of obstetrics. The cost, both human and financial, is highly significant and has led to an enormous literature on prophylaxis and treatment of this medical problem. Efforts to develop an effective intervention are challenging due to the variety of risk factors and generally poor understanding of the physiologic mechanisms for the onset of labor. Specific conditions leading to preterm labor include infection, premature rupture of membranes, multiple gestation, and abruptio placentae, but in most cases, the precipitating factor(s) are unknown. The pathophysiologic mechanisms are likewise poorly understood. To date, there has been no comprehensive picture of the pathophysiology triggering the onset of labor that would inform the pharmaceutical approach to the prevention of preterm birth.

Studies of putative tocolytic agents have filled the obstetrics literature for at least the last 80 years. This review will address the recent history of the various attempts to stop labor. It does not presume to be a systematic review of all studies of tocolytics as multiple such studies have been published with remarkably similar conclusions. Rather, this paper will attempt to trace the evolution of the care for women in preterm labor and the science, or lack thereof, that lead to these interventions incorporating conclusions of textbooks, published reviews, and meta‐analyses, as well as specific examples of relevant clinical trials that illustrate the problems with the current state of practice.

The American College of Obstetrics and Gynecology (ACOG) has taken the position that “because tocolytic therapy has generally been effective for up to 48 hours, only women with fetuses that would benefit from a 48 hour delay in delivery should receive tocolytic treatment,” and further states that “the evidence supports the first line tocolytic treatment with beta‐adrenergic agonist therapy, calcium channel blockers, non‐steroidal anti‐inflammatory drugs for short term prolongation of pregnancy” [1]. Based largely on these recommendations, tocolytic treatment is widely considered to be the standard of care [2].

Progesterone has long been considered significant in the maintenance of pregnancy. It is not surprising that progesterone was among the first agents to be suggested as a treatment to forestall delivery. Based on work in rabbits [3, 4], small uncontrolled studies suggested the need for further investigation. However, follow‐up trials did not confirm benefit and suffered from poor study design [5, 6]. Discussion of progesterone as a tocolytic disappeared from Williams Obstetrics by 1979 [7]. A Cochrane review of more recent data concluded that “there is insufficient evidence to advocate progestational agents as a tocolytic for women presenting with premature labor” [8].

Again, based on early studies in rabbits that showed that alcohol inhibited milk ejection and postponed delivery, clinical trials of I.V. alcohol were conducted in the 1960s [9]. The alleged mechanism of action was suppression of posterior pituitary oxytocin, similar to the known suppression of vasopressin. As seen with other compounds, small observational studies without benefit of controls appeared to show benefit [10, 11]. However, a review in 1979 stated that “there is now disagreement as to the role, if any, of endogenous oxytocin in human labor” [12]. Lack of neonatal benefit and significant maternal adverse events lead to abandoning ethanol as other agents came to the fore [12].

For some time, epinephrine in low doses had been shown to have a depressant effect on the myometrium [13]. Isoxsuprine was an early beta‐adrenergic agonist marketed at the time as a vasodilator (Vasodilan). Early uncontrolled studies appeared to confirm preclinical experience showing delayed delivery in treated women [14]. Nevertheless, its use resulted in only transient effects accompanied by significant maternal adverse events, and a small controlled trial could not confirm benefit [15]. Experience with this agent did lead to subsequent work with later more selective beta‐2 agonists, including salbutamol, ritodrine (Yutopar), and terbutaline [16, 17].

Ritodrine, a beta‐2 adrenergic receptor agonist, rose to prominence in the late 1970s with the publication of a study appearing to show benefit in delaying labor [18]. Follow‐up studies were unconventional by current standards. One publication [19] included 150 patients from three centers presenting with threatened premature labor who were randomized by sealed envelope to ritodrine or ethanol. However, at the conclusion of the trial, 15 patients were eliminated for various reasons. It was not specified as to which group these patients had been assigned, leaving 67 patients, including nine with twins, in the ethanol group and 68 patients, six of whom had twins, in the ritodrine group. Delivery was postponed for more than 72 h in 49/67 patients in the ethanol group, and 61/68 patients treated with ritodrine. Twins were initially included in the study, but omitted from the analysis of outcomes, with 52/58 singletons surviving in the ethanol group and 58/62 singletons treated with ritodrine (Chi square = 0.022, p = 0.83).

A subsequent study in 1977, but with only 29 patients, found no difference in outcomes with ritodrine [20], as did a slightly larger study (N = 61) in 1980 [21]. In 1980, a larger study showed a different outcome [22], but this publication was unusual in that results were presented for a “series of prospective randomized double blind controlled studies comparing ritodrine with either ethanol or placebo…”. Of the 366 patients entered into the study, a disproportionately higher number of patients was “randomized” to oral ritodrine only after they responded to IM ritodrine. Furthermore, a primary outcome variable was not specified. A total of 366 patients entered the study, with analysis based on 309 patients remaining after 57 were dropped for various reasons, including twin gestation, premature rupture of membranes, fetal loss, and protocol violations. The incidence of neonatal death, respiratory distress, and achieving 36 weeks’ gestation was reported for the remaining subjects, most favoring the ritodrine group.

While this study is unlikely to pass current standards, it nevertheless provided the first apparent beneficial compound for the management of preterm labor and established beta‐2 agonists in general as the standard for future comparison. Nonetheless, subsequent experience with these compounds was at best mixed, with some studies finding no significant improvement in outcome. A meta‐analysis including 16 studies of various betamimetics in 1988 showed an “unequivocal effect” in delaying delivery, but without any improvement in perinatal mortality or severe neonatal respiratory disorders [23]. A subsequent randomized trial including 708 women found “no significant beneficial effect on perinatal mortality, frequency of prolongation of pregnancy to term or birth weight” [24]. A study of national trends in preterm birth found no impact of these agents on the incidence of low birth weight [25].

The search for an agent to improve outcomes continued with several more candidates studied, including indomethacin, magnesium sulfate, calcium channel blockers, and oxytocin receptor blockers. Indomethacin was evaluated in a small, randomized, placebo‐controlled study [26]. Treatment failure was defined as progression of cervical dilation beyond 4 cm within 24 h, which occurred in 9/15 in the placebo group and 1/15 receiving indomethacin. Yet no difference in neonatal outcome could be detected. A subsequent study comparing indomethacin with ritodrine (without a placebo control group) showed them to be “equally successful,” but again without demonstration of improvement in neonatal outcome [27]. Subsequent concerns about serious neonatal adverse effects have significantly decreased the use of this agent [28].

Magnesium sulfate use became more common through the 1980s with uncontrolled trials claiming benefit [29, 30] and studies comparing outcomes between magnesium and ritodrine showing no apparent differences [31]. Another trial including magnesium, terbutaline, and a placebo showed “no significant differences between the groups” for any of the outcomes specified [32]. In contrast, a randomized prospective study with an untreated control group found “no significant effect on duration of pregnancy, birth weight, neonatal morbidity or perinatal mortality” [33]. The role of magnesium sulfate as a tocolytic effectively ended with the publication of a review in 2006, whose authors bluntly state that it is ineffective in stopping premature labor and its continued use “reflects inadequate progress toward rational therapeutics in obstetrics” [34].

Nifedipine became more attractive as a potential tocolytic after trials comparing nifedipine with ritodrine, but without a placebo group, appeared to show similar outcomes with significantly reduced incidence of maternal adverse events [35, 36, 37]. However, other reports comparing nifedipine to placebo failed to show success in prolonging pregnancy, even by 48 h [38, 39].

The most recent entrant into the tocolytic field was Atosiban, an oxytocin receptor antagonist. The concept of an oxytocin receptor antagonist was introduced in 1985 [40], based on the theory that oxytocin receptors (but not oxytocin) increase prior to the onset of labor. The clinical trial submitted to the FDA suffered from significant design flaws and safety concerns, with higher perinatal mortality reported in the smallest exposed neonates [41]. The study employed a placebo arm at initiation of treatment, but then allowed use of a “rescue tocolytic” in cases of “therapeutic failure,” such as persistent contractions after one hour of treatment. The incorporation of this feature likely improved recruitment, but also compromised assessment of outcomes by including a highly subjective endpoint. Nonetheless, no clear benefit of Atosiban could be demonstrated on neonatal outcomes [42]. More recently, a large placebo‐controlled trial again failed to show improvement in neonatal outcomes [43].

An early systematic review included studies of betamimetics, prostaglandin synthesis inhibitors, calcium channel blockers, oxytocin receptor blockers, nitroglyceride, and magnesium sulfate [44]. The authors concluded that, while tocolytics may briefly prolong pregnancy, they have not been shown to improve perinatal or neonatal outcomes and have adverse effects on women in preterm labor. Nothing has changed over the subsequent 25 years.

Systematic analyses of individual agents continued with the focus on magnesium sulfate, the most commonly used tocolytic in the United States at the time. Citing a Cochrane review in 2002 [45] among other data that showed magnesium sulfate tocolysis was not only ineffective, but was also harmful to infants, Grimes and Nanda stated flatly that it was “time to quit” [34].

Additional Cochrane Library reviews were published in 2014 on magnesium sulfate [46], betamimetics [47], calcium channel blockers [48], oxytocin receptor blockers [49], and in 2015 on cyclooxygenase inhibitors [50]. The results of the magnesium sulfate review confirmed the previous reviews and the conclusion of Grimes and Nanda. While there was some evidence of a delay in delivery with betamimetic agents, no improvement in perinatal or neonatal outcome was evident and significant maternal adverse events were noted.

The reviews of calcium channel blockers concluded that these agents may prolong pregnancy, but without benefit to perinatal survival. Some improvement in NICU admission and preterm birth was reported, but only in comparison with betamimetics. Given the adverse event profile with betamimetics, this apparent improvement with calcium channel blockers may have been the result of higher‐than‐expected rates of complications with betamimetics. Again, this points to the absolute need for placebo, rather than supposedly active controls. To this point, the authors stated “the lack of blinding of the intervention diminishes the strength of this evidence” [48]. Importantly, this review did not incorporate results from the most recent placebo‐controlled trial that failed to show any improvement with nifedipine [39].

Similar results were reported in a meta‐analysis of oxytocin receptor antagonists, with no demonstrated improvement in prolongation of pregnancy or neonatal outcomes. The authors cite concern about the increased number of fetal deaths in the study, although they acknowledge that the number of women enrolled at very early gestations was small [49].

Finally, a Cochrane review of cyclooxygenase (COX) inhibitors reported “…no clear benefit for COX inhibitors was shown over placebo or any other tocolytic agent” [50, 51]. The authors note that, as with the other reviews cited above, significant methodologic shortcomings in study design compromise the ability to draw conclusions of efficacy for any of these agents.

Finally, a recent systematic review and meta‐analysis evaluating tocolytics in cases of extremely preterm labor concluded that there were “…no clear differences in perinatal death nor delay in births” with the use of tocolytics [51].

It must be emphasized that the only purpose of prolonging pregnancy is the improvement in neonatal outcome. The widely used surrogate endpoint of a 48‐h delay in delivery is based on the theory that this delay will allow administration of betamethasone [1]. Unfortunately, multiple systematic reviews of the various agents have been published, all with similar conclusions: none of these agents have been shown to improve perinatal mortality or morbidity. As noted at the start, a systematic review of the studies of various tocolytics was not the aim of this paper. Various individual studies were presented to attempt to provide a historical overview of the challenges presented in evaluating the literature. The accurate diagnosis of preterm labor is notoriously difficult, with 50% [1] or more [23, 32, 43] of patients fitting various criteria for that diagnosis not delivering within 48 h without treatment. The percentage remaining undelivered varies with the stringency of the inclusion criteria and their application, which always leaves the potential for subjective decisions about inclusion. This simple fact demonstrates the absolute need for a blinded study with a placebo arm in any investigation of efficacy.

Most published studies have focused on comparisons of different agents, or even combinations of agents. These studies were not constructed as equivalence trials, which would have required greater numbers; did not include a power calculation, further raising the likelihood of a type 2 error; and did not prove that either arm was actually effective, as most patients would be expected to continue their pregnancy past 48 h or to term as noted above.

The most significant issue with the tocolytic literature is that despite over 70 years of study, hundreds of reports, and tens of thousands of subjects, there has been no evidence of improvement in neonatal outcome, which, as stated, is the only appropriate goal of treatment.

1. CONCLUSION

Given the preponderance of currently available data on the value of tocolytics, the only possible response is to echo the words of Grimes and Nanda: “Time to Quit.” Published literature to date has failed to show neonatal benefit, but demonstrated risk to pregnant patients, and in some reports to the neonate, leading to a risk/benefit ratio that decidedly favors risk. Consideration of the use of tocolytics as the standard of care must stop, and just as importantly, patients must be educated that such interventions have so far proven ineffective, and in some cases unsafe, and will not be offered.

Acknowledging the use of tocolytics as standard of care has adverse impacts beyond that on the individual mother–baby dyad. Concern by patients as well as by IRBs about the inclusion of a placebo arm in trials of putative tocolytics has severely limited the ability to provide compelling evidence of benefit. Concern among clinicians about legal consequences of “failure to treat” affects patient management decisions and widespread use of unproven therapies adds to the already prohibitive costs of American medicine.

The focus of research on preterm labor must change. A more accurate means of determining true preterm labor, or at least of standardization for study purposes, is imperative. Published studies vary widely on the definition of preterm labor, resulting in most included subjects continuing their pregnancies regardless of treatment or lack thereof. Large numbers of such patients in “false labor” dilute the pool of subjects who could show benefit from treatment. Precisely because of this difficulty in predicting who is actually in preterm labor, inclusion of a placebo group in all studies is mandatory.

The defined endpoints in published studies have been inconsistent. Some studies have reported on time to change to a second or different agent, which is not useful as no additional agents have been proven effective. Most studies have used the endpoint of delay in delivery for 48 h, purportedly allowing time for transfer to a higher level of care and/or administration of steroids. This parameter is a surrogate endpoint, the definition of which requires that it significantly predicts a final outcome of real patient benefit. Decades of studies and meta‐analyses refute this contention, as no neonatal benefit has been demonstrated despite claims of a 48‐h delay. Excuses have been made for this failure to show benefit, but the failure remains. A focus on the requirement put forth by the FDA to show benefit to the neonate must motivate any future tocolytic development plan. Surrogate endpoints, if used at all, should be restricted to proof‐of‐concept studies, which themselves must not affect standard clinical care.

As the Food and Drug Administration (FDA) requires for approval that a candidate tocolytic demonstrate improvement in neonatal outcome, clinical trials should focus on subjects most likely to benefit from treatment, specifically the extremely premature patients at 28 weeks or less. While this cohort constitutes only about 0.5% of deliveries in the United States [52], the frequencies of perinatal death and major complications associated with prematurity are significantly higher and improvement in outcomes can be seen with small increases in the length of the pregnancy [53]. Ideally, future studies should involve new compounds and the requisite IND. Repurposing of existing drugs has not proven successful and has led to premature adoption of ineffective treatments. Emphasis to date has been placed on interrupting the mechanics of labor, that is, the final common pathway of uterine contractions. The assumption is clearly that the fetus is better off in utero regardless of the etiology of preterm labor. Given that we do not understand the mechanism(s) by which risk factors initiate preterm labor, the possibility that extreme prematurity may have different mechanism(s), not to mention that the cost of drug development will run into the hundreds of millions of dollars and take several years during which other improvements in care will occur, initial efforts may be better spent on the study of the basic mechanisms of the physiology of labor rather than further clinical trials.

While these changes will make development plans more difficult and expensive, the continued publication of papers with poor study design and inadequate power clutters the literature and creates a false hope of progress. Failure to make significant changes in our approach to the study of treatment of preterm labor condemns our specialty to more decades of inadequate, ineffective, costly, and potentially dangerous interventions.

CONFLICT OF INTEREST STATEMENT

The authors declare no conflicts of interest.

FUNDING INFORMATION

The authors received no specific funding for this work.

DATA AVAILABILITY STATEMENT

Data sharing not applicable to this article as no datasets were generated or analyzed during the current study.