Progesterone for short cervix
Ripe-tomato.org often criticises triallists for not following a pre-specified analysis plan. Today we examine one trial that registered late, and another that did all the right things, but tripped up on a secondary data-driven analysis.
Background
Pre-term birth is an important health problem. A short cervix (neck of the womb) detected on ultrasound scan may predict it, and perhaps progesterone hormone treatment can prevent it. There have been two trials.
- Fonseca and colleagues (click here) randomly allocated 250 women with a short cervix to treatment with progesterone (n=125) or placebo (n=125). The rate of spontaneous pre-term birth before 34 weeks was significantly reduced in the vaginal progesterone group. But the trial had not been registered at clinicaltrials.gov (click here) until Jan 15 2007, despite ending in May 2006, nearly a year earlier. This means we cannot be certain that the primary endpoint, sample size and analysis plan were all pre-specified. Hitting the target of exactly 125 participants per group, with zero losses to follow up, in eight hospitals spread over three countries and two continents, also seemed a bit too good to be true!
- The next trial by Hassan and colleagues in 2011 (click here) had no such problems. It had been registered here in 2008. It was double-blind. The planned sample size was 450 (achieved sample size 465), and the primary endpoint of delivery before 33 weeks was the same in the trial registration and the published paper. Only seven women were lost to follow up. “Women allocated to receive vaginal progesterone had a lower rate of pre-term birth before 33 weeks than did those allocated to placebo (8.9% (n = 21) vs 16.1% (n = 36); relative risk (RR), 0.55; 95% CI, 0.33–0.92; P = 0.02″ Ripe-tomato.org was impressed.
But in Jan 2012 the US Federal Drugs Administration (FDA) turned down the manufacturer’s request for progesterone to be licensed for this indication. Here are the publicly available FDA pre-meeting briefing documents progesterone fda background doc. Unsurprisingly the share price of the manufacturer, Columbia Laboratories, fell off a cliff the day they were released.
For those who are interested, study 300 is a negative trial (O’Brien et al. 2007 – click here for the published report) of the same drug used to prevent pre-term labour in high risk women in general. In a secondary analysis the applicants had tried to show that it worked in women with a short cervix. The FDA were unimpressed.
But nor were they impressed by Hassan (study 302 in the briefing document). The problem was that pretty much all the benefit came from a half dozen or so small non-US sites, which between them recruited only 50 participants, and had 8 pre-term labours in the treatment group and only 1 in controls. The FDA statistician redid the analysis, including these worrying cases, but adjusting in a statistically more conservative way. The difference in the primary endpoint (bold row) was no longer statistically significant.
They also re-plotted the primary endpoint results for US and non US sites separately.
Of course this is a post hoc analysis, and the triallists must have felt hard done by. But the FDA was considering a hormone to be given to thousands, perhaps tens of thousands of pregnant women. They must have remembered another hormone used to prevent miscarriage, diethylstilboestrol (DES), which later turned out to be not only ineffective, but to cause vaginal cancer and other problems in the baby (click here). Unless a single trial’s data are very convincing, it’s best to demand another trial.
The FDA appears not to have commented on another issue. Although the mean gestational age at first dose was the same in both groups (Table 1 main trial report), the appendix to this main report (click here) reports an imbalance in recruitment outside the planned gestational age window of 20 and 24 weeks gestation. 29 women were recruited before 20 weeks (20 placebo and 9 progesterone) and 26 women after 24 weeks (7 placebo, 19 progesterone). The appendix does not say where these erroneous recruits came from, but the imbalance is unlikely to have occurred by chance and in both cases would tend to favour the active treatment.
Fortunately other trials are ongoing. In Nottingham we are recruiting similar patients to vaginal progesterone or placebo in the UK-wide OPPTIMUM trial led from Edinburgh – click here for details. Let’s get that one completed quick.
Jim Thornton
Ripe-tomato.org 
We are also recruiting for OPPTIMUM in Exeter. As you can imagine there are a number of women asking for progesterone based on the already reported trials. Let’s hope OPPTIMUM reports soon.
It is interesting how these things go round in circles and how much research is paid for by those who might benefit, though now covertly. We have been trials on local anaesthetic infusions post laparoscopic surgery, all funded locally, and guess what it does not work, but that is nor what you would believe using the available literature. i do now believe that all procedures should be subjected to randomised trials but the problem is that they they are almost impossible to organise unless the power calculation shows that it will work or it is paid for by some vast anonymous funding body.
Jeremy
A very interesting analysis but what is really important is outcome rather than gestation at delivery. A few extra days gestation at birth is of no consequence if the outcomes are unchanged. Also did all those babies receive optimal management of a preterm birth or did it vary ? Specifically of course I am referring to cord management at birth. ACOG will be making a statement later this year supporting the idea (based on the evidence) that clamping off the only supply of oxygen to the baby is harmful and also that depriving the baby of a significant volume of blood (around 30%) together with the stem cells in the placental transfusion is harmful; in other words they are supporting delayed cord clamping. Hopefully we will follow with this advice here in the UK..
I’m not sure the evidence for delayed cord clamping is as strong as David suggests. In pre-terms there appear to be reductions in important outcomes, but the quality of the trials is low.
In term babies the beneficial effects are only in things like Hb and iron stores which, as David notes above for gestation, will not necessarily translate into improved substantive outcomes.
Let’s not forget that in utero the ideal Hb level is about 18-20 g/dl, but once you move to air the ideal is 12-14 g/dl. It is at least plausible that the baby is glad to not have that extra 30% blood volume shoved in.
I hope the various colleges will hold off making ex-cathedra recommendations until the evidence is in. Trials are possible in this area and ongoing.
OK, it is fair enough criticize what is an appropriate outcome measure for obstetric/neonatal interventions. To take it to the extreme one could look at quality adults life and age of death but we will never achieve that ! Jim admits there are reductions in important outcomes from the studies of preterm babies but the “quality” of the trials is not low it is only the numbers of babies involved and that is changing rapidly. Everyone can learn and discuss a lot more by coming to the conference in Birmingham on transition at birth. Judith Mercer will be reporting more of her study showing improved brain function in babies with delayed cord clamping.
http://www.birmingham.ac.uk/transitional-care
In term babies the arguments regarding cord clamping starts from a different position. Healthy term babies do not need any medical intervention. The mother simply needs to be supported to deliver her baby and the baby needs support to transition from placental to lung respiration. Cord clamping is a medical intervention if it is carried out while there is still a placental circulation. If the cord is clamped and cut when this circulation has ceased then it is reasonable to consider it is a minor and non-medical intervention.
So Jim accepts, and the evidence for this is based on large numbers of babies from numerous trials, that Hb and iron stores are reduced by early cord clamping. Clearly the intervention is having a significant effect. Other smaller studies have shown that the blood volume of the neonate will be reduced by up to 30% by early clamping. In a child or adult reducing the blood volume by 30% would be associated with severe circulatory compromise so why do we not see it in babies who have early cord clamping ? Virtually every baby born in hospital has early clamping. Most of these babies will have had an apparently normal labour. Some of these babies are found to have have severe circulatory compromise at birth. Every obstetrician knows of babies who are born in unexpected poor condition. Other explanations are looked for. Early cord clamping could be responsible but is it ever considered ? We cannot be sure but it is a reasonable possibility. I is impossible to investigate retrospectively for a number of reasons. Firstly it is impossible to measure the blood volume of a neonate and distinguishing between hypoxia induced circulatory collapse and hypovolaemic circulatory collapse is very difficult. Just look at neonatal resuscitation advice; if the baby does not respond to ventilation and cardiac massage consider the possibility of a inadequate placental transfusion and give an emergency transfusion of O-neg blood. (Obstetrics by TEN TEACHERS, 19th edition Neonatology, page 286) Then if resuscitation is unsuccessful the pathologist cannot determine the babies blood volume. Lastly the clinical notes will not document the timing of cord clamping.
So Jim wonders if the baby might be glad not to receive the extra 30% of blood “shoved in”. In any of the studies of delayed cord clamping was there evidence that the babies who had the extra blood “shoved in” were less glad than those who were denied the blood ? There is no doubt that there is a redistribution of blood between the placental and the baby immediately after birth but this can hardly be described as “shoving in ” blood to the baby. Indeed the term placental transfusion is inaccurate since the blood is the babies own blood, it has never been out of its own circulation and some blood returning via the “transfusion” in the umbilical vein immediately after birth may well be back in the placenta again before the circulation finally closes down. I am puzzled that Jim is prepared to consider that because the fetal haemoglobin is about 19g/dl and neonatal Hb eventually falls to 13g/dl the obstetrician is at liberty to assist this reduction by reducing the blood volume ! We all know that it is a lot more complicated than simply a fall in Hb; there is a gradual change in the nature of the Hb from fetal to adult type to take into account the different oxygen dissociation curves needed by fetal placental breathing and neonatal pulmonary breathing.
The various colleges have previously based there advice about the timing of cord clamping on tradition but no evidence. Now that the evidence is emerging the least they can do is withdraw their previous unfounded statements. The RCM has looked at the evidence and will be publishing specific advice to midwives to avoid early clamping in normal births in the near future. If obstetricians are to avoid the impression that they are harbouring a guilty secret they must ensure that if they continue to clamp the cord early after birth that they time and document their action.
Dear David, I can’t let you get away with “the “quality” of the trials is not low”. The trials were not just small, they were of at best of modest quality.
In the Cochrane review. Rabe H, Diaz-Rossello JL, Duley L, Dowswell T. Effect of timing of umbilical cord clamping and other strategies to influence placental transfusion at preterm birth on maternal and infant outcomes. Cochrane Database of Systematic Reviews 2012, Issue 8. Art. No.: CD003248. DOI: 10.1002/14651858.CD003248.pub3
Not a single trial was judged to have a low risk of bias on more than 3 out of the 6 criteria the Cochrane authors assessed. As far as I can judge not one of the 15 trials was registered.
Jim
The WHO do not agree with you tha tthe evidence is weak. The WHO Neonatal Resuscitation guideline is just out and it states
In newly-born term or preterm babies who do not require positive-pressure ventilation, the cord should not be clamped earlier than one minute after birth. This is a strong recommendation based on high to moderate evidence.
Jim,
Can I presume therefore that you accept the other points I made ? The Cochrane review for preterm babies had small numbers in each trial with moderate quality but the whole point of a metanalysis is to combine the data from small trials. At least the results were all consistent. Since then data is coming from Judith Mercers trial, which is a registered RCT showing some beneficial long term outcomes from delayed clamping. In term babies the intervention is much more clearly early clamping and it is only necessary to show that it is quite unnecessary. I do not know about the registration of the trials which is certainly important but of little relevance when there were actually no trials to show that early clamping was needed either for the mother or for the baby. I have just been watching Richard Hammond’s “Miracles of Nature which have taken millions of years to evolve and how can we think we can improve on it with no real evidence at all.
This is obviously also the view of the WHO who have just published their recommendation for neonatal resuscitation. They use all the evidence including the latest large trial by Andersson et al (which I am sure will have been registered) and state “Not earlier than one minute” should be understood as the lower limit supported by published evidence. WHO recommendations for the prevention of postpartum haemorrhage recommend that the cord should not be clamped earlier than is necessary for applying cord traction, which the GDG clarified would normally take around 3 minutes. ”
The WHO recommendation also includes “If there is experience in providing effective PPV without cutting the cord, ventilation can be initiated before cutting the cord.” Most of us realise that if a baby has been denied 40% of its normal it is not going to respond well to resuscitation. But Jim wonders if the baby might be glad not to receive the extra 30% of blood “shoved in”. None of Anderssons baby objected. I suspect they might not have been quite so happy with 30% of their blood volume “pulled out” ! Any volunteers prepared to donate 30% of their blood volume just in case you “might be glad” without this extra blood ? As Hancock said over 50yrs ago to the doctor “Now look chum everyone to his own trade but If I’ve got eight pints obviously I need eight pints and not seven by the time you’ve finished with me.”
I am concerned about the use of progesterone in pregnancy which I think may be contributing to the increased risk of thromboembolism in pregnancies following IVF. Increased thromboembolism and progesterone use in IVF pregnancies. I published a response at http://www.bmj.com on 28 January 2013
Re: Incidence of pulmonary and venous thromboembolism in pregnancies after in vitro fertilisation: cross sectional study. Peter Henriksson, Eli Westerlund, Håkan Wallén, Lena Brandt, Outi Hovatta, et al.346:doi:10.1136/bmj.e8632
I would be much happier if a trial was carried out to show the effect of monitored repletion of intracellular magnesium levels. However, that is probably unethical as any magnesium deficiency, which can be diagnosed by measuring red cell magnesium levels, should be repleted for normal cell function. A muscle heat test can show abnormal contraction patterns in muscles if magnesium levels are low.
Interesting point. ivf pregnancies do of course pretty much all get progesterone. we must make sure PROMISE trial reports thrombo-embolism.