Universal cervical-length screening and vaginal progesterone prevents early preterm births, reduces neonatal morbidity and is cost saving: doing nothing is no longer an option

Authors


In a paper in this issue of the Journal, Hassan et al.1 address the question as to whether the incidence of spontaneous preterm birth (PTB), which has been rising inexorably over the past 30 years, can be reduced. The authors demonstrate that, by screening all singleton pregnancies, a group of women at high risk for spontaneous PTB can be identified in the middle trimester by transvaginal measurement of the cervix and that gestation can be prolonged and neonatal morbidity lessened in this high-risk group by daily vaginal progesterone therapy. Furthermore, the treatment is well accepted by the patients and has no known side effects. Also in this issue, Werner et al.2 calculate that this program, and a similar one published previously3, are cost saving. In this Editorial I will give briefly some of the background to this far-reaching plan which, in modern parlance, is a game-changer.

The problem

Preterm delivery, with its collateral effects on neonatal mortality, short- and long-term infant morbidity and astronomical healthcare costs, is the foremost problem in modern obstetrics. In spite of advances in obstetric care, the rate of prematurity has not decreased over the past 40 years and in developed countries there has been an increase, possibly as a result of assisted reproductive technology (ART) programs. The current frequency of PTBs is 12.5% in the USA and 5–9% in European and other developed countries4. Spontaneous PTB, with or without prior rupture of the membranes (PPROM), accounts for two-thirds of PTBs, with the remainder occurring as a result of obstetrically indicated preterm delivery due to conditions such as pre-eclampsia and intrauterine growth restriction (IUGR)4.

Prematurity remains a leading cause of neonatal morbidity and mortality in developed countries, accounting for 60–80% of deaths of infants without congenital anomalies. Greater attention is now being focused on very PTBs < 32 weeks' gestation, because, although this group represents only 1–2% of all deliveries, it accounts for about 60% of perinatal mortality and nearly 50% of all long-term neurological morbidity5. Short-term morbidities associated with early preterm delivery include respiratory distress syndrome, intraventricular hemorrhage, periventricular leukomalacia, necrotizing enterocolitis, bronchopulmonary dysplasia, sepsis, patent ductus arteriosus and retinopathy of prematurity. Long-term morbidities include cerebral palsy, cognitive defects and social and behavioral problems. This recital of the well-known complications of early PTB does not convey the devastating effects on the parents of a severely damaged child, the daily disruption to normal life, the anxieties and stress, the constant visits to hospital, or the sense of isolation and social exclusion from other parents with normal children6. Nor does it convey the long-term pain and suffering and frustrated lives of the children who are affected, usually with multiple disorders7. Unsurprisingly, many marriages do not survive the stresses imposed on them by a severely disabled child8 and, in those that do survive, few couples have the time and resources to attempt to have further children. In purely financial terms, the short- and long-term costs of caring for PTBs are staggering. Short-term hospital costs in the USA, over the first year of life of short gestation/low birth-weight infants, were estimated to be $ 5.6 billion9 and estimates for the annual societal economic burden in the USA was at a minimum of $ 26.2 billion in 200510. A UK study11 using a decision-analytic model estimated the costs to the public purse of children born prematurely up to 18 years of age to be £2.95 billion ($ 4.6 billion). The model calculated that a hypothetical intervention that delayed PTB by 1 week across all gestational-age categories would reduce these costs to £1.95 billion. Judgment on the success or otherwise of any preventative treatment should focus on deliveries occurring before 33 completed weeks of gestation, when the morbidity and mortality rates are high and the incremental cost of care to 18 years of age, compared to a term survivor, is almost $ 100 00011.

Historical perspective

The legendary Danish/American obstetrician Fritz Fuchs carried out the first randomized double-blind study of progesterone in the prevention of preterm delivery12. He was influenced by the work of Arpad Csapo which showed that the uterine myometrium was held in a state of quiescence by a ‘progesterone block’ and that normal labor was initiated by a fall in progesterone and withdrawal of this block13. Csapo's work was carried out on rabbits and laboratory rodents, where the source of progesterone depends on the corpus luteum and the fall in progesterone occurs following the involution of the corpus luteum. Progesterone withdrawal also occurs in larger placental sub-primates, such as sheep, as a consequence of a shift of steroidogenesis that favors estrogen production at the expense of progesterone. However, a fall in serum progesterone levels does not occur in primates and Fuchs's attempts to treat women who had experienced premature contractions with ‘high-dose’ daily oral progesterone was ineffective in delaying delivery. Nevertheless, it was widely recognized that progesterone plays a major role in maintaining uterine tranquillity in human pregnancy and several studies of low power and sometimes poor design, and with conflicting results, were subsequently carried out to assess the efficacy of a variety of mainly synthetic progestins in the prevention of recurrent miscarriage and preterm delivery. A meta-analysis published in 1989 by Goldstein et al.14 seemed to put the issue to rest; they found no benefit of progesterone in either the prevention of miscarriage or of preterm delivery in high-risk patients and recommended that progestogens should not be used outside of randomized trials. The following year, Keirse15 published a further meta-analysis, including only studies which used 17α-hydroxyprogesterone caproate (17OHP-C) administered by weekly intramuscular (IM) injection and found that, although ineffective in the prevention of miscarriage, this particular formulation, when given to women with a history of preterm delivery, was associated with a significant reduction in the incidence of PTB in the current pregnancy. However, it was too late; the damage to the reputation of progestins was apparently fatal and the world moved on to finding the perfect tocolytic, on the assumption that preterm labor was due to a primary activation of myometrial contractility, probably through an interaction between oxytocin and prostaglandin F2α16. Prevention gave way to the suppression of premature contractions and, over the next 20 years, there was a plethora of randomized trials studying a large number of tocolytics, including beta-2 receptor agonists, calcium channel blockers, prostaglandin synthetase inhibitors, oxytocin receptor blockers, magnesium sulphate and nitric oxide donors. The seemingly endless search for the perfect tocolytic ended in failure. A meta-analysis17 including all robust RCTs that compared tocolytics either with controls or with no treatment concluded that while most would delay delivery for up to 1 week, tocolytics were not associated with improved neonatal outcomes and many were associated with significant side effects. Indeed, tocolytics were shown not to be associated with a significant reduction in deliveries before 30, 32 or 37 weeks' gestation.

By the end of the 20th century, researchers had turned again to the idea of prevention, a move stimulated by three developments: 1) a greater understanding of the etiological factors leading to spontaneous PTB; 2) improved prediction of spontaneous PTB by measurement of cervical length; 3) the rediscovery of progesterone.

Changing concepts of causation

The initiation of labor involves two interdependent processes, i.e. remodeling, effacement and eventual dilatation of the cervix (with or without membrane rupture) and the triggering of rhythmic contractions of increasing amplitude and frequency. The principal mechanism whereby the fetus maintains its immunological privileges in the uterus is the tight regulation of cytokine levels at the maternal–fetal interface. Progesterone has an important role in the maintenance of pregnancy by modulating antibody production and reducing the production of proinflammatory cytokines by direct and indirect action on immune cells. It is now believed that, at term, spontaneous uterine activity occurs secondary to modulation of the progesterone receptors in the decidua, amnion and cervix, creating a ‘functional’ progesterone withdrawal which allows the release of proinflammatory cytokines18. This process of cervical ‘ripening’ precedes the myometrial contractions of labor by weeks. The most likely pathway for the initiation of preterm labor involves premature decidual activation occurring in the context of an occult upper genital tract infection, with release of cascades of proinflammatory cytokines in the cervix and the choriodecidual interface19–21. Cytokines such as interleukin (IL)-1β, tumor necrosis factor-alpha (TNFα) and IL-6 increase the production of matrix metalloproteinases (MMP-1, MMP-8 and MMP-9), which digest collagen type 1 in the extracellular matrix to induce cervical softening and effacement22. This process is also thought to weaken the cervical mucus plug, further reducing the ability of the cervix to resist infection, with possible weakening and rupture of the membranes. Increasing myometrial contractility in the presence of subclinical infection probably involves the macrophage system and its cytokines, such as TNFα and IL-1, which stimulate prostaglandin production in the amnion and decidua23, 24. These proinflammatory cytokines have also been immunolocalized to leukocytes in the myometrium, resulting in an increase in the production of prostaglandins and oxytocin25. This process of cervical effacement commences weeks before preterm labor occurs and can be recognized by shortening of the cervix on transvaginal ultrasound examination. Confirmatory evidence of this process is the finding that a cervical length < 15 mm is associated with a 22% incidence of significant elevation in amniotic fluid MMP-8 levels26.

Sonography of the cervix

Shortening and funneling of the cervix was first described in relation to the diagnosis of cervical incompetence, but it was the paper of Andersen et al.27 that first drew attention to the possibility that the transvaginal sonographic measurement of cervical length could be used to predict a risk for preterm delivery for a significant percentage of all PTBs. Since this study, there have been detailed descriptions of the nature of the morphological changes detectable on transvaginal ultrasound, including changes in the width and depth of funneling at the internal os (with or without abdominal pressure), herniation of membranes, loss of the cervical mucus plug, alteration in the appearance of the cervical mucus glands and altered cervical index. While all of these features are helpful in predicting spontaneous PTB, recent studies have demonstrated that the functional cervical length is the single best predictor and that all other parameters add little to the predictive value of this test28, 29. However, it is usually recommended to observe the cervix for about 30 s to detect spontaneously occurring cervical shortening, a process which can be speeded up by 10 s of gentle fundal pressure.

Unscheduled shortening and effacement of the cervix is therefore a ‘final common pathway’ for many, but by no means all, of the causes of preterm delivery. To estimate what proportion of early PTBs would be predicted by a short cervix is difficult because many of the studies were performed on high-risk gestations28. Even those studies carried out on normal populations give conflicting results. Iams et al.30, in a multicenter study of selected ‘normal’ women examined at 24 and 28 weeks' gestation, found that a cervical length ≤ 20 mm detected 23% of PTBs occurring before 35 weeks. Heath et al.31 studied an unselected UK population between 22 and 24 weeks' gestation and found that a cervical length ≤ 15 mm occurred in 1.7% of the population and included 58% of births occurring before 32 weeks. The positive predictive values also varied considerably. In the Iams study, the PPV at 24 weeks for deliveries before 35 weeks was 26%, while in the Heath study, the PPV for births before 32 weeks was 52%. To et al.32, from The Fetal Medicine Foundation (FMF), developed a model combining maternal risk factors with cervical length and found that the inclusion of maternal age and previous history of PTB improved the prediction. The detection rate of screening for preterm delivery before 32 weeks at a fixed false-positive rate of 10% was 38% for maternal factors, 55% for cervical length and 69% for combined testing. This concept is important when considering which patients to include in spontaneous PTB prevention programs.

Rediscovery of progesterone

Although progestins had been shown to be ineffectual in preventing miscarriage15, it was recognized from an early stage that in women undergoing IVF treatment, luteal support with progesterone was required to improve implantation and pregnancy rates, as down-regulation with luteinizing-hormone-releasing hormone agonists and the removal of granulosa cells during follicular aspiration can interfere with both corpus luteum formation and progesterone production. Various formulations for natural progesterone therapy, which could be taken daily by IM injection, orally or transvaginally, were developed. While IM and oral treatment generated higher blood levels of progesterone, several studies have shown higher endometrial progesterone concentrations in women who received vaginal progesterone despite lower serum concentrations, indicating some mechanism of direct transport between the vagina and uterus33, 34. As a result, vaginal progesterone treatment is associated with lower serum levels of metabolites and fewer side effects than is oral therapy35 and patients prefer it to IM injection, finding it less inconvenient and associated with less discomfort36. Vaginal progesterone can be taken in various formulations, but micronized progesterone in oil capsules (200 mg tid) and micronized progesterone as a bioadhesive gel (90 mg qd) are popular. The dose for luteal support can be as high as 800 mg/day for the capsules and twice daily for the gel. A meta-analysis of seven randomized trials37 showed that there was no difference in clinical effectiveness for luteal support between the different vaginal formulations at accepted doses and that side effects were negligible. As will be discussed below, vaginal progesterone is now being used in the prevention of spontaneous PTB. The doses used for the capsules for this indication are approximately one third those that are currently employed in luteal support, although the doses used for the gel are similar.

Recent studies

In 2003/4, three RCTs based on high-risk populations with a previous history of PTB were published, with conflicting results. Meis et al.38 published a randomized controlled study of weekly injections of 17OHP-C and demonstrated a 33% reduction in the rate of PTB in the treatment group significant at < 37, < 35 and < 32 weeks. No improvement in neonatal morbidity was shown, but a follow-up study demonstrated no long-term adverse effects of the synthetic progestin39. In a study from Brazil, da Fonseca et al.40 randomized 142 high-risk pregnancies to a daily vaginal suppository of 100 mg micronized natural progesterone or placebo and found a large reduction in deliveries before 34 weeks in the treatment arm (2.8% vs. 18.6%). O'Brien et al.41, in the largest of these studies, recruited 659 women with a history of PTB who were randomized at 18–23 weeks to a vaginal bioadhesive gel containing 90 mg micronized natural progesterone daily or placebo. There was no reduction in the frequency of recurrent PTB ≤ 32 weeks or neonatal morbidity. It could be argued that these conflicting findings were the result of the different formulations, but as the bioavailability of the bioadhesive gel formulation is equal to or possibly superior to that of the vaginal suppository, this is unlikely. A more probable explanation is the unreliability of previous history as the basis for randomization. DeFranco et al.42 highlighted this in their subanalysis of the O'Brien41 study. The original protocol included measurement of the cervix at the time of enrolment. In the subanalysis, it was found that in women with a cervical length < 28 mm, the rate of PTB ≤ 32 weeks was significantly lower in the progesterone group, with a significant reduction in respiratory distress syndrome in the treated group. This study highlights the problem that had bedevilled all previous therapeutic intervention studies where randomization is based on a history of PTB, as this is likely to have a small impact on the overall rate of spontaneous PTB because fewer than 10% of spontaneous early PTB occur in women with a previous history. The problem with the DeFranco subanalysis was that the results of only 46 patients were analyzed (19 in the progesterone group, 27 in the placebo group) and that randomization was based on obstetric history and not on cervical length. The implication from this study was that randomization should be based on cervical length measurement.

The first study of progesterone prophylaxis based on randomization on cervical length was provided by Fonseca et al. and The FMF Second Trimester Screening Group3 who, in a multicenter trial, enrolled 24 620 unselected women to have a transvaginal ultrasound scan between 22 and 25 weeks' gestation; 413 (1.7%) women had a cervical length of 15 mm or less and 300 of these patients were randomized to receive 200 mg vaginal progesterone capsules or placebo daily. Analysis was on the intention-to-treat principle. There was a significant 40% reduction in spontaneous deliveries before 34 weeks' gestation in the progesterone group (19.2% vs 34.4%) and a non-significant reduction in neonatal mortality and morbidity.

The second study randomized on the basis of cervical length is from Hassan et al.1 and is presented in this issue of the Journal. This is an international multicenter trial of women recruited from a screened population of 32 091 who had cervical length measurement at 19–24 weeks' gestation; 733 women (2.3%) had a cervical length between 10 and 20 mm, of whom 465 were randomized to receive 90 mg progesterone in bioadhesive gel or a placebo gel transvaginally daily.

Analysis by intention to treat demonstrated a significant 45% reduction in deliveries < 33 weeks (8.9% vs 13.1%), a 50% reduction in births < 28 weeks (5.1% vs 10.3%) and a 47% reduction in very low-birth weight infants (6.4% vs 13.6%) in the progesterone treated group. In the treatment group there was also a significant reduction in respiratory distress syndrome (3.0% vs 7.6%) and a composite of neonatal morbidity indices.

Despite minor differences in the protocols, the findings of the two studies were similar. The Hassan study1 demonstrated a significant difference in indices of neonatal morbidity which were not demonstrated by the Fonseca study3, although in the latter study there was a non-significant trend towards reduced morbidity. This could be explained by the fact that the Fonseca study was not specifically powered for this purpose and, unlike the Hassan study, included twin gestations which have not been shown to respond to progesterone prophylaxis. Also, the studies used different cut-offs of cervical length for inclusion in the study. Hassan et al. excluded cervical lengths < 10 mm on the premise that these patients were less likely to respond to progesterone treatment. This may explain the 50% higher incidence of PTBs in the placebo group in the Fonseca study as the highest incidence of early preterm delivery occurs with cervical lengths of 0–10 mm.

Roberto Romero and Kypros Nicolaides are presently conducting an individual patient meta-analysis of the Fonseca3 and Hassan1 studies (R. Romero, pers. comm.) which should strengthen the message from these important trials, but meanwhile a brief summary of the findings and conclusions of these key papers and the papers of Cahill et al.43 and Werner et al.2 is as follows.

  • A transvaginal scan in the second trimester in singleton pregnancies carried out between 19 and 24 weeks to measure cervical length is the best method with which to identify a group of women (approximately 2% of the pregnant population) who would benefit from prophylactic progesterone treatment to prevent spontaneous PTB.

  • Progesterone prophylaxis in women with a short cervix, preferably using natural progesterone by the vaginal route, reduces the incidence of PTB before 33 weeks by 45%.

  • Progesterone prophylaxis will significantly reduce the incidence of respiratory distress syndrome and composite indices of neonatal morbidity.

  • Vaginal progesterone prophylaxis has no reported side effects attributable to the drug and treatment compliance is high (> 90%).

  • Progesterone prophylaxis for women with a short cervix appears to be a cost-effective strategy with the potential to provide net financial benefit to the healthcare system compared with a no screening/no treatment policy.

Issues for discussion

How does progesterone work?

It is now clear that although progesterone has a uterine quiescent effect, the mechanism of action in preventing preterm labor involves primarily the cervix. Both synthetic progestin and natural progesterone treatment have the effect of lengthening the cervix44, 45 and clinical studies have demonstrated that antiprogestins induce cervical ripening both in the first trimester and at term46, 47. Progesterone can therefore be regarded as a cervical strengthening hormone and local withdrawal of progesterone at the cervical level will result in the release of proinflammatory cytokines, with resulting softening and effacement of the cervix but not necessarily uterine contractions47, 48. Progesterone treatment acting locally at the cervical level appears to control the release of proinflammatory cytokines and prevent extracellular matrix degradation of the cervix. Research is also underway to determine if progesterone improves the antimicrobial activity of the cervical mucus plug (R. Romero, pers. comm.). One of the issues to be determined is whether the dose of progesterone currently being employed is adequate for women with a very short cervix. The vaginal progesterone doses used for luteal support in ART programs with vaginal capsules (although not the gel) can be three times those used in PTB prophylaxis and have minimal side effects or long-term risks for the fetus.

Which cervical cut-off is optimal?

Spontaneous PTB has many causes but a short cervical length is the final common pathway for many of these clinical conditions. In the study of Fonseca et al.3 the cervical length was 15 mm or less in 1.7% of pregnancies and this group included 25.8% of all spontaneous PTBs occurring before 34 weeks of gestation. A program of progesterone prevention based on this cut-off has been shown to be cost effective2, so this is a considerable advance on treatment based on history alone which addresses fewer than 10% of the total number of births before 34 weeks' gestation. The Fonseca study found that the cervical length was 16–25 mm in 8.3% of pregnancies and this group included 20.4% of spontaneous PTBs occurring before 34 weeks of gestation. An argument could therefore be made for extending the screening cut-off to cervical lengths of up to 25 mm as this would allow almost 50% of women who would give birth before 34 weeks to receive progesterone therapy. In most countries, both developed and developing, a mid-trimester scan is carried out routinely and is recommended by the relevant national clinical advisory boards. For example, in the UK it is specifically recommended by the National Institute for Clinical Excellence (NICE) and the Royal College of Obstetricians and Gynaecologists (RCOG). Extension of this ultrasound examination to include a 5-min transvaginal scan to measure cervical length would not involve a significant increase in expenditure. Furthermore, if a significant reduction in PTBs before 34 weeks was demonstrated by extending the screening cut-off to include cervical lengths of 25 mm or less (and the fact that Hassan et al. showed a significant reduction between 10 and 20 mm is encouraging), the impact on the totality of spontaneous PTBs may make this extension worthwhile.

A further question is how we manage cervical lengths less than 10 mm. Hassan et al.1 excluded these cases from their study as they26 had previously demonstrated a high incidence of proinflammatory cytokines in the amniotic fluid of such cases. They believe when the cervical length is ≤ 10 mm, there is a high chance of an unstoppable progression to delivery that is unlikely to be affected by progesterone. They also quote a personal communication from Kypros Nicolaides that in the Fonseca study3, women with cervical lengths < 5 mm were refractory to progesterone treatment. It would seem reasonable to revisit at least the group of patients with cervical lengths of 6–10 mm, as they constitute over 30% of women with cervical lengths < 20 mm and are associated with an 87% incidence of delivery before 32 weeks' gestation. In the study of Vaisbuch et al.26, although 23.4% of women with a cervical length < 10 mm had increased levels of MM-8, only 4.8% had positive amniotic fluid cultures. It may be that increasing the vaginal dose of progesterone and perhaps the addition of antibiotics in these cases49 would result in prolongation of pregnancy sufficient to provide significant clinical benefit.

Is there a role for cerclage?

Cervical cerclage has been used since the 1950s50, 51 to prevent PTBs in women with a history indicating a risk of classical cervical insufficiency (e.g. previous painless mid-trimester miscarriages or cervical surgery) or with mid-trimester dilatation of the internal os. With the advent of ultrasound, women with such a history frequently had cerclage performed on the basis of shortening and funneling of the cervix. However, two randomized trials52, 53 comparing cerclage with no cerclage in women with a cervical length < 25 mm showed contradictory results. In the trial of Althuisius et al.52, delivery was significantly less frequent before 34 weeks in the cerclage group and all but one of the patients had risk factors for cervical insufficiency. In the study by Rust et al.53, in which women were included on the basis of an incidental finding of a short cervix, there was no difference between the cerclage and control groups in the incidence of delivery before 34 weeks. The initial selection of patients appears to have been crucial to determining the outcome in these studies. In the trial of Althuisius et al., cerclage was performed in patients who had risk factors for cervical insufficiency, while in the study of Rust et al., the majority of patients who were included were at high risk of preterm delivery and were, Rust hypothesised, on a final common pathway of multiple pathological processes, such as infection, immunologically mediated inflammatory stimuli and subclinical abruption. The failure of cerclage in women identified on the basis of a short cervical length was confirmed by To et al.54; 253 women were randomized to cerclage or no cerclage at the time of cervical length screening at 20–22 weeks' gestation. The proportion of women delivering < 33 weeks was similar in both groups (22% cerclage vs 26% control) without any significant differences in perinatal morbidity or mortality. The implication is that women identified on the basis of a short cervix are more likely to have inflammatory cervical cytokines and less likely to respond to cerclage. This concept is strengthened by the study of Sakai et al.55, who studied 246 women with a short cervix (< 25 mm) in a non-randomized study. The 165 women who had cerclage had cervical mucus aspirated for the measurement of IL-8 concentrations. In the overall group, cerclage did not reduce the rate of PTB or lengthen the procedure-to-delivery interval. However, women who had low concentrations of IL-8 in the cervical mucus had a lower rate of PTB < 34 weeks and a longer procedure-to-delivery interval than did those without cerclage. Women with elevated IL-8 in the cervical mucus had a higher rate of PTB and a shorter procedure-to-delivery interval than did controls. The implication of this is that cerclage may worsen the outcome in patients with endocervical inflammation, which may contraindicate its use in women with cervical lengths ≤ 10 mm. Two recent papers suggest an indication for cerclage in a specific cohort of high-risk pregnancies. In the study of Owen et al.56, 302 women with a history of a prior delivery < 34 weeks were assigned randomly to cerclage or no cerclage on the basis of a cervical length < 25 mm. There was a non-significant reduction in PTB < 35 weeks in the cerclage group, but this reduction became significant for births occurring before 24 weeks, suggesting that the effect was strongest for women with true cervical insufficiency. A significant reduction in births < 35 weeks was also found in the subgroup of women with cervical lengths < 15 mm. Berghella et al.57 performed a meta-analysis of five randomized trials of cerclage vs no cerclage that met strict inclusion criteria, i.e. singleton pregnancy, short cervical length on transvaginal ultrasound before 24 weeks' gestation and previous PTB before 35 weeks' gestation. In this select group, cerclage was associated with a significantly lower delivery rate < 35 weeks (28% vs 41%), < 32 weeks (19.2 vs 29.5%) and < 28 weeks' gestation (12.8 vs 20.1%). Furthermore, there was a significant reduction of 36% in the composite perinatal mortality and morbidity rate. Berghella et al. postulate that it is possible that these women selected for cerclage, with a previous PTB and painless cervical shortening to < 25 mm before 24 weeks in a subsequent pregnancy, have a clinically significant component of cervical insufficiency. They caution that cerclage should only be offered to the 8% of women with a history of previous PTB and within this the subgroup of patients with a cervical length < 25 mm. This is a small cohort of the women destined to have a PTB, but Berghella et al. estimate that in the USA more than 6500 newborns would be saved each year from perinatal death if this policy was introduced.

Can spontaneous PTB be prevented in twin gestations?

Twin gestation carries one of the highest risks of preterm delivery. In singleton pregnancies, the rate of spontaneous premature birth before 34 weeks is about 1%, while in twins it is 13%58. To et al.58 carried out a study of cervical length measurement at 22–24 weeks' gestation in 1163 twin gestations. Monochorionic twins were included but cases showing evidence of twin-to-twin transfusion syndrome were excluded. The rate of delivery before 32 weeks' gestation was strongly associated with cervical length. Using cut-offs of < 25 mm, < 20 mm and < 15 mm, the respective detection rates of spontaneous PTB before 32 weeks' gestation were 35%, 49% and 67%. Monochorionic and dichorionic twins had a similar incidence of early PTB. To date, all randomized clinical trials investigating the prevention of spontaneous PTB in twin gestations with progestins have yielded negative results. The earliest study59 was carried out on 77 pregnant women between 28 and 33 weeks' gestation who were randomized to weekly injections of 17OHP-C or placebo. There was no reduction in the rate of PTB or perinatal morbidity in the treatment arm. More recently, Rouse et al.60 carried out a similar placebo-controlled study except with larger numbers (655 women) and with the 17OHP-C administration being started between 16 and 20 weeks and continued until 35 weeks' gestation or delivery. Delivery before 35 weeks occurred in 41.5% of women in the progestin group and 37.3% in the control group. In a third double-blind placebo-controlled study61, 500 women with twin gestations were randomized to either 90 mg vaginal progesterone gel or placebo gel for 10 weeks from 24 weeks' gestation. The combined proportion of intrauterine death and delivery before 34 weeks was 24.7% in the progesterone group and 19.4% in the placebo group. No study has addressed the prevention of spontaneous PTB in twin gestations with randomization based on cervical length as a primary outcome. A subanalysis of the Fonseca trial3 showed that randomization of the 24 twin gestations with cervical lengths ≤ 15 mm to vaginal progesterone or placebo resulted in a non-significant reduction in births before 34 weeks in the progesterone group. It is probably too soon to write off progestins in the prevention of preterm delivery in twin gestations with selection based on cervical length, and large trials of vaginal progesterone in women with twins are needed to answer the question as to whether progesterone is beneficial in these cases. It could be that there is a lower incidence of subclinical infection in the cervix of twin gestations. Some of these pregnancies may benefit from cerclage but there are no randomized studies of cerclage for a short cervix in twin pregnancies. In the study of To et al., the rate of preterm delivery in the women with a short cervix who were treated with cerclage was 36% rather than the expected 66%. This is not substantive evidence, but the need for further studies is clear as twin gestations now represent a significant subset of PTBs as a result of the increasing use of ART throughout the world.

Are progestin-based prevention strategies cost effective?

Considering the huge long-term costs to society of PTB11, one would assume intuitively that any strategy which could reduce the incidence of PTB would be cost effective. Three cost-effectiveness studies using decision analysis have been published from the USA. Odibo et al.62 calculated the effect of 17OHP-C administration compared with no treatment in high-risk women. The model favored the use of 17OHP-C in women with a history of both preterm (< 37 weeks) and early preterm (< 32 weeks) birth as being cost saving. In the example of women with a previous early PTB, 17OHP-C saved $ 17 000 and resulted in 5.5 additional quality of life years (QALY) per child. The authors' assumption was that approximately 50% of women would attend for prenatal care before 16 weeks' gestation, 8.7% of these being multiparous with a prior PTB and 1.3% with a prior early PTB (< 32 weeks' gestation). While the calculated savings are impressive, the small number of women eligible for treatment would reduce the impact on the total financial burden of PTBs.

Cahill et al.43 used a decision-analytic model to compare the cost effectiveness of universal cervical screening/vaginal progesterone against no screening/no treatment on an estimated annual delivery rate of 4 million (the approximate annual birth rate in the USA). The calculated annual cost, based on spontaneous PTBs < 34 weeks and severe morbidity prevented, fell from $ 46.2 billion to $ 33.3 billion per annum, an annual saving of $ 12.9 billion, which was greater than calculated savings from 17OHP-C treatment on the basis of maternal history ($ 7.9 billion) or cervical-length screening in high-risk pregnancies ($ 3.9 billion). Cahill et al. estimated a more than 50% decrease in PTBs in the universal screened group, which was much higher than the reduction estimated in the second study, by Werner et al.2, on the cost effectiveness of universal cervical-length screening/vaginal progesterone, published in this issue of the Journal. The Werner study used a decision-analytic model, with estimates based on the study of Fonseca et al.3 and assumptions on neonatal mortality and morbidity based on those found in the recent published literature9, 63, 64. For example, the probability of severe disability was estimated to be 10.6% for babies born before 28 weeks and 5% for babies born from 28 to 34 weeks64. The cost estimates of the model are detailed and include a mean estimate of $ 70 for the second-trimester cervical-length scan and $ 206 for vaginal progesterone treatment. The model calculated a saving of just over $ 12 million for every 100 000 women screened with 423.9 QALY being gained in the universal screened/vaginal progesterone group. This is striking when the model assumes a rather conservative reduction of only 11.8% of births before 34 weeks gestation and a 1.8% reduction in neurological deficit. The study is impressive in the detail of the analysis and the assumptions which, commendably in some instances, are biased against screening and progesterone treatment. For example, in Europe and many countries of the world, where the second-trimester scan is part of standard antenatal care, the cost of the ultrasound scan in the Werner model would mostly be absorbed into the existing costs of this examination. While the savings calculated in this model are convincing, it should be stressed that they are hypothetical, as in the study of Fonseca et al.3 a reduction in neonatal morbidity was not demonstrated. However, in an addendum, Werner et al. reanalyze their model using instead the data of Hassan et al.1, which demonstrate a reduction in neonatal morbidity (Table 1). The model now demonstrates improved savings of over $ 19.5 million/100 000 women screened and a net health improvement of 735 QALY compared with no screening.

Table 1. Summary of results of Werner et al.'s decision-analytic model2: effect (per 100 000 women) of universal cervical-length screening/vaginal progesterone administration, compared with no screening in 2010
 Original analysis (Fonseca et al.3 data)Re-analysis* (Hassan et al.1 data)
  • Table provided by Dr Erica Werner.

  • *

    Assumes a reduction in preterm birth with progesterone in women with cervical length 1.5–2.5 cm. QALY, quality of life years.

Cost savings$ 12 119 947$ 19 603 380
QALY gained424735

In summary, these studies indicate massive savings to the healthcare system, from universal cervical-length screening and vaginal progesterone treatment, which are unlikely to be significantly different in any healthcare system or in any country in the world.

Conclusion

Over the past 40 years, improvements in neonatal care by pediatricians have contributed greatly to the reduction in neonatal morbidity and mortality, especially for low birth-weight newborns. Prenatal ultrasound has also made major contributions to the reduction of infant disability, not only through the detection of fetal anomalies and selective termination of pregnancy but also through improved prediction and monitoring of IUGR. Yet, the elephant in the room has always been spontaneous PTB, for this is by far the biggest cause of neonatal death, morbidity and long-term and irreversible damage to the child. Few believed that ultrasound could make a contribution to the prevention of a complex condition which rivals pre-eclampsia as a disease of theories. However, following the realization that shortening and effacement of the cervix is the final pathway common to many if not most cases of spontaneous PTB, ultrasound is now about to make a major contribution towards reducing this intractable problem by the identification of pregnancies at high risk for early preterm delivery, following routine measurement of cervical length from 19 to 24 weeks' gestation and administration of progesterone to women with a short cervix. The impressive study of Hassan et al.1 provides convincing evidence of the effectiveness of this regimen in terms of reducing the incidence of early preterm delivery, and for the first time providing evidence of an improvement in neonatal indices of morbidity such as respiratory distress syndrome. This study, together with the earlier study from The FMF3, will hopefully cause an urgent rethink regarding how antenatal care is planned, facilitating implementation of a program which should save lives and prevent a significant amount of serious and irreversible disability in children. The screening method is simple and in most counties can be piggy-backed on to an existing routine mid-trimester ultrasound scan. The detailed calculations of Werner et al.2 demonstrate that such a program is cost effective. The treatment is vaginal progesterone, which has been shown to be safe and to have virtually no side effects, even at the considerably larger doses given in early pregnancy for luteal support in ART programs. Future studies will define the optimal cervical lengths for inclusion in the treatment program and whether twin gestations should be included. However, this should not delay the implementation of a program based on the paper of Hassan et al. Our patients and their babies deserve no less.

Ancillary