- Top of page
- Supporting Information
Accumulating evidence over the last 15 years has shown that screening of pregnant women for preterm birth, based mainly on obstetric history and measurement of the length of the uterine cervix, can identify more than 50% of those who will deliver before 34 weeks' gestation1. At almost the same time, two large randomized studies2, 3, soon followed by more, showed that administration of progesterone in women with a singleton pregnancy at risk can significantly reduce the rates of preterm birth both in cases of previous history and in cases with a short cervix.
Therefore, having both a reliable screening tool and a treatment option for women identified as being at highest risk, it may be that ‘doing nothing is no longer an option’ in singleton pregnancies4. Progesterone has been used for decades in reproductive medicine without any documented effect on the risk for congenital abnormalities5, and the allegedly increased risk for fetal hypospadias does not appear to be confirmed in currently used progestogens6; the United States Food and Drug Administration has recently approved the use of hydroxyprogesterone caproate for the prevention of prematurity in women with a previous history of preterm birth7.
However, although progesterone prophylaxis clearly reduces the risk of preterm birth in women at risk, its effects on the actual perinatal and long-term consequences of prematurity are more difficult to assess. This is because these events are relatively uncommon even in preterm births, with the exception of extremely preterm births, and therefore individual studies are usually underpowered for these outcomes because of small sample size. As a result, existing randomized controlled trials (RCTs), and to some extent systematic reviews as well, have so far focused mainly on the primary outcome of reduction of preterm birth rates.
The aim of this meta-analysis was to systematically review published evidence and pool data on the perinatal outcome in women treated with progesterone for the prevention of preterm birth.
- Top of page
- Supporting Information
Preterm birth has always been a major cause of perinatal mortality and morbidity with long-term consequences27–31. However, it is only in the last decade that progesterone has been successfully shown to prevent preterm birth in singleton (but not in multiple) pregnancies at risk, either because of a short cervix32 or previous history of preterm birth2.
As the impact of progesterone treatment in the reduction of prematurity rates has been extensively quantified (level-1 evidence), in this meta-analysis we focused on the effects of progesterone on the actual perinatal outcomes of treated pregnancies. We found that prophylactic progesterone administration in singleton pregnancies at risk reduces rates of neonatal mortality, RDS, admission to the NICU and composite adverse outcome. In contrast, current data fail to support the presence of a beneficial effect of progesterone in multiple pregnancies and, in fact, the rates of perinatal death, RDS and composite adverse outcome may even be increased in twins.
In singleton pregnancies, we found that progesterone decreased the risk for neonatal mortality and common morbidity, when all data were pooled. We furthermore planned to run subgroup analyses based on the indication for progesterone treatment and the type of progestogen used. The design of available studies allowed for only two groupings, i.e. local (natural) progesterone for women with a short cervix and systemic (essentially 17 alpha-hydroxyprogesterone caproate (17-OHPC)) progestogen for women with previous preterm birth. For the latter pair, data could be pooled only for neonatal mortality and NICU admission; progesterone prevented one death per 24 women treated and one NICU admission per three women treated. For local progesterone in women with a short cervix, 17 women needed to be treated in order to prevent one perinatal death or complication. However, statistical significance was not reached for most of the other outcomes, although the direction of effect was usually in favor of progesterone. For the outcome of neonatal mortality, 829 infants per arm would be needed based on our pooled results, in order for the effect to reach significance. In addition, we were not able to reach any conclusions about the most clinically relevant question, i.e. whether local or systemic progesterone is better for women with a short cervix.
Studies and systematic reviews on twins have shown that progesterone fails to reduce the rate of preterm birth33. Two commonly cited hypotheses for this are (1) that the doses used in singleton pregnancies may be insufficient in twin pregnancies and (2) that the mechanisms underlying spontaneous preterm delivery in twins are different from those in singletons18, 21, 33. It is difficult to prove or discard these hypotheses but, at least for the first hypothesis, maternal body mass did not have significant impact on the effectiveness of 17-OHPC in a study of singletons34, and its half-life and plasma levels were not found to be affected by parity and the number of fetuses35. Therefore, a threshold effect for 17-OHPC has not been proposed so far, and we would expect it to have some effect on preterm labor rather than none. As for the second hypothesis, accepting that infection would be the main cause for preterm birth in singletons vs uterine stretch in twins, the effectiveness of progesterone would depend on its ability to reduce uterine excitability. Many36, 37, but not all38, studies indicate that progesterone may indeed induce uterine quiescence through a multitude of mechanisms, from alteration of the electrical activity of the uterine muscle to alteration of gene expression and peripheral leukocyte activation.
Nevertheless, if the only problem was lack of effectiveness, increasing the dose or selectively treating women with twin pregnancy and a short cervix might be a viable alternative; however we found that progesterone may actually increase the risk of adverse outcome in twin pregnancies. According to our pooled results, there may be one additional case of perinatal death per 71 fetuses, one additional case of RDS per 39 fetuses and one additional case of composite adverse outcome (perinatal death or complication) per 31 fetuses treated. It is not clear why this happens, but it appears to be independent of the rates of preterm birth, which did not differ between treated women and controls for the cut-off of 34 weeks (pooled rates)18, 20, 22, 32 weeks (pooled rates)18–20, 23 or 28 weeks (pooled rates)18–20, 23. The relatively small number of triplets prevented our achieving statistically significant results in this group; however one should still be cautious as the study of Combs et al.25 indicated an increased risk for fetal death in treated women.
A weakness of meta-analyses in general is that pooled studies cannot have identical inclusion criteria, as well as treatment and reporting protocols. This problem is addressed by subgroup and sensitivity analysis, at the cost of reducing the sample size. Indeed, when analyzing pre-defined groups, statistical significance was lost for many of the perinatal outcomes. However, in our case there was marked consistency in the results across studies for most outcomes, even before subgroup analysis. This was true even for the composite adverse outcome, despite its being defined differently in different studies. The lack of heterogeneity increases the strength of the results and may indicate that the number of fetuses is a more significant predictor of the effectiveness of progesterone treatment than the type of progesterone used (and the indication for treatment in singleton pregnancies). The outcome for which the highest variation was recorded was NICU admission, which may reflect the diversity in defining neonatal special care needs, together with different admission criteria across studies.
The next step after testing the effects of progesterone treatment on the rates of preterm birth and immediate perinatal complications is to examine its impact on the longer-term neurodevelopment of treated children. At the moment there has only been one study assessing neurodevelopment in twins up to 18 months of age; the authors report almost identical results in terms of Ages and Stages Questionnaire mean scores and rates of low scores20.
Based on current data, it appears that progesterone treatment in women with a singleton pregnancy at risk succeeds in reducing the rates of neonatal mortality, some of the common perinatal complications and composite adverse outcome (perinatal death or complications), in contrast to the case with twins, in which it may even exert an adverse effect.
Among ongoing studies, one will address the impact of cervical length screening and progesterone treatment for length ⩽ 30 mm on perinatal outcomes and cost39, an individual patient data meta-analysis in twin pregnancies will focus on the effect of progesterone treatment on perinatal mortality and morbidity40, while another prospective follow-up study will evaluate the differences in developmental outcomes of children aged 23–25 months, born to mothers who participated in a trial of the efficacy of 17-OHPC41.
SUPPORTING INFORMATION ON THE INTERNET
Appendix S1 and Tables S1–S5 may be found in the online version of this article.