Episode 6.10 Magnesium Sulfate for Neuroprotection for Preterm Infants


In this episode, we review the literature regarding the use of magnesium sulfate for neuroprotection to prevent CP in preterm infants.

00:00:02 Magnesium Sulfate for Neuroprotection
00:11:00 The Controversy Surrounding Magnesium Sulfate
00:36:37 Research Study Mistakes and Bias
00:46:33 Magnesium Sulfate’s Effects on Preterm Birth
00:52:24 Magnesium for Fetal Neuroprotection Reevaluated

Links Discussed

Prenatal Intravenous Magnesium at 30-34 Weeks’ Gestation and Neurodevelopmental Outcomes in Offspring: The MAGENTA Randomized Clinical Trial

Can magnesium sulfate reduce the risk of cerebral palsy in very low birthweight infants?

Is tocolytic magnesium sulphate associated with increased total paediatric mortality?

Perinatal death and tocolytic magnesium sulfate

School-age outcomes of very preterm infants after antenatal treatment with magnesium sulfate vs placebo

School-age outcomes following a randomized controlled trial of magnesium sulfate for neuroprotection of preterm infants

Comparison of Cerebral Palsy Severity Between 2 Eras of Antenatal Magnesium Use

Antenatal magnesium sulphate for the prevention of cerebral palsy in infants born preterm: a double-blind, randomised, placebo-controlled, multi-centre trial

Effect of antenatal magnesium sulphate on MRI biomarkers of white matter development at term equivalent age: The MagNUM Study

ACTOMgSO4Crowther2003https://pubmed.ncbi.nlm.nih.gov/14645308/4g bolus then 1 g/hr for 24n=1062 <30 wkCP, death, and CP+deathCP 6.8%Death 13.8%CP 8.2%Death 17.1% Not SS
PREMAGMarret2006https://pubmed.ncbi.nlm.nih.gov/17169012/4g over 30 minn=688 < 33 wkWMI, death or WMI+deathWMI=10D=9.4Comb=16.5WMI=11.7D=10.4Comb=17.9
Not SS
MAGNETMittendorf2002https://pubmed.ncbi.nlm.nih.gov/12032519/4 g then 2-3 g/hr or 4 g bolus onlyn=14425-33 wkMag levels, IVHMore mag, more IVH
4 g bolus then 1g/h for 24n=1544 <37 wkMortality, CP at 18 months
OR 1.07
BEAMRouse2008https://www.nejm.org/doi/full/10.1056/nejmoa08011876 g bolus then 2 g/hr for 12n=224124-32 wk
Wolf2020https://pubmed.ncbi.nlm.nih.gov/32237024/5g bolus then 1 g/hr for 24hn=560>32 wkMod Sev CPDeatghNSS
MAGENTACrowther2023https://pubmed.ncbi.nlm.nih.gov/37581672/4g IV over 30 minutesn=143330-34 wk Death or CP23/6913.3%18/6742.7%More RDS and PPH in Mag group
Doyle2014https://pubmed.ncbi.nlm.nih.gov/25226476/4g bolus then 1 g/hr for 24n=867School ageCP, cog., beh., funct., outcomesNo differences
MagNUMPoppe2023https://pubmed.ncbi.nlm.nih.gov/35331677/4g IV over 30 minutesn=10930-34 weeks Brain white matter neurostructureLess fractional anisotropySuggests if anything an adverse affect
Chollat2014https://pubmed.ncbi.nlm.nih.gov/24837863/n=4327-14 years oldDeath, CP, cog, motor, etcNo differences
Stetson2019https://pubmed.ncbi.nlm.nih.gov/30508016/BEAM protocoln=3736 infantsNo reduction in CP attributable to Mag


Howard 00:18


Antonia 00:19


Howard 00:20

What are we thinking about on today’s episode? 

Antonia 00:22

Well, we’re going to talk about my museum sulfate for prevention of cerebral policy in preterm infants, which I think might be your favorite topic ever. 

Howard 00:33

My favorite topic?

Antonia 00:34

Well, you certainly talk about it a lot and write about it a lot. 

Howard 00:38

Isn’t this another one of those drum up the ratings by picking a controversial topic? . 

Antonia 00:44

Maybe, but also there was a pretty recent study relating to this. I know you’ve been excited to talk about it too. So the bottom line up front here is this can also be our thing. We do for no reason, except there’s so much to talk about. It’ll just be our whole episode today, and then the listener can decide at the end if they agree and whether they feel strongly about continuing to do this practice, because I know it’s still very widespread and probably institutionalized. So why don’t you start us off by telling us about the new study and what they found? 

Howard 01:21

Well, I’ll admit, this new study peaked my interest to go back and look at this again because I hadn’t really thought too much about it in about five years and there has been a lot of new stuff developed, and I’ll also say that I came into it with an open mind in terms of where is the science at now and where would it take us. So let’s go through this with an open mind and see where we get. So, as you said, there’s a new paper published this summer that has some results published from the Magenta trial, and this is the sixth trial now, or seventh depending on how you look at it. To look at the topic of magnesium sulfate for neuroprotection, and this was written by Carolyn Crowther and colleagues, and they’re the group that performed the first such trial back in 2003. And they’ve also subsequently co-authored or she has co-authored a couple of systematic reviews on the topic, and so she’s definitely been an advocate for using magnesium to prevent cerebral palsy for a long time. 

Antonia 02:16

I know this topic about magnesium and running cerebral palsy and something you’ve been interested in for many years and you’ve been quite critical of it for probably two decades or so. 

Howard 02:29

Maybe yeah. 

Antonia 02:30

Yeah, sometimes in a very public way, regarding the evidence that supports the practice, so I have been patiently waiting for us to get into it on this podcast. 

Howard 02:41

Well, you’re being diplomatic, but yes, I’ve attacked the practice in the past, so you’re the diplomat. 

Antonia 02:46

Yin and yang, I guess. Or good cop, bad cop maybe Everyone needs a little balance, including you sometimes. 

Howard 02:52

All right, fair enough. 

Antonia 02:53

Well, I’ll say, before we go any further, that the official position of ACOG on this isn’t actually what most people would think that, especially people that have a protocol for it and do it routinely. So I’ll just read from the committee opinion, which is number 455. This was written in 2010 and reaffirmed in 2020. So they actually do a fairly balanced job of going through the literature up until that point, up until 2020, specifically focusing on three major trials. So here’s the last paragraph, which summarizes their review of literature up to that point. 


The Committee on Obstetric Practice and the Society of Maternal Fetal Medicine recognize that none of the individual studies found a benefit with regard to their primary outcome. 


However, the available evidence suggests that magnesium sulfate given before anticipated, early preterm birth reduces the risk of cerebral palsy in surviving infants. Physicians electing to use magnesium sulfate for fetal neuroprotection should develop specific guidelines regarding inclusion criteria, treatment regimens, concurrent tocolysis and monitoring in accordance with one of the larger trials and they say with one of them, because there are three large trials that they’re referring to and each of them had completely different protocols about how they administered mag and also their patient populations and other parameters of their studies. So in one of them, a one-time four gram load with no maintenance dosing, while in another one there was a six gram load, solid by two grams an hour for 12 hours and then stopping it if the labor didn’t progress and resuming it if it did progress. So very wide variation in the dosing regimens and definitely in the total dose that the fetuses were exposed to and, like I said, they all included very different populations. 

Howard 04:45

Yeah, In fact in that committee opinion they state that the studies weren’t really comparable because of how different the inclusion and exclusion criteria were and obviously how different the therapeutic approaches were in terms of doses and length of dose and all those things. And that’ll be an interesting thing to remember later, because maybe the great sin that’s occurred with all of this is that folks have tried to compare these studies through meta-analysis and drawn some broader conclusion that’s not necessarily supported. 

Antonia 05:16

Okay, yeah, so that authors of this committee opinion were acknowledging that there wasn’t a statistically significant benefit to the mag, but that there appeared to be a trend towards towards production and cerebral palsy. 


And I think this was paired with the attitude of how harmful is it then to give the mag if it might help? And we had already, as a profession, been used to using it, first for preterm labor and also for preeclampsia, and overall we had already been aware of the side effects using it for those other indications and deemed them to be manageable and acceptable in comparison to what we think that the benefits were. So, for decades, I think the most common interpretation of this committee opinion has been if it might help the baby, go ahead and use it. We’re not telling you you have to use it because the studies don’t prove that it actually works. But if you want to use it, just decide how you’re going to use it, or you can carefase that how you’d like to. But I think over time that has really morphed into people thinking that they said you have to use it because now it’s our pastoral protocol, because told us you have to make a protocol, so we have to use it, which is not what they said. 

Howard 06:32

Yeah, I think that’s a fair assessment. You said it backwards that we first used it for prevention or treatment of seizure or eclampsia, and then prevention of preeclampsia. 


And then people noticed, hey, didn’t it take longer to induce that lady who was on magnesium for hypertension, so maybe it works to slow labor. And then we thought, well, maybe we can use it to slow labor, for preterm labor. And it’s evolved since the 1950s into where we are today. Keep looking for a use beyond the thing it definitely works for, which is treatment of and prevention of seizures in women with preeclampsia with severe features. 


But I think that the original, more muted message from ACOG has been forgotten by many of the folks who do use it, because it’s become institutionalized really at every major academic center in the United States and spilled over from there. And in most cases it was the higher dose protocols that were continued because that’s what they had been used to doing when they had been using magnesium as a tocolytic or even for preeclampsia. So they were comfortable with it and validated the continued use of magnesium, maybe even as a tocolytic. Maybe you thought, maybe it worked as that, so we can kill two birds with one stone, maybe. And the problem is that 15 years later now, younger physicians and training nurses staff like that, they think it’s this amazingly effective and super important, universally recommended thing that prevents cerebral palsy in surviving infants, and that it needs to always, always be done. And they’ve lost sight of the original positioning of the idea. 

Antonia 08:04

Yeah, and they probably don’t think that it’s amazing because they’ve seen it work wonders or that they’ve followed babies in the NICU and after they leave the NICU and grow up and compare them to the same babies, that did get me. I doubt that that’s it. I think it’s almost purely from habit, from the fact that it’s protocol. I was told when I first started training or first started working here that this is what we do and we have to do it and it’s very important and don’t really question further than that. And then maybe sometimes they think about the theory of how they think it should work to help prevent cerebral palsy. 


But by now I think we can all agree that magnesium definitely does not work as a tocolytic. That finding has been confirmed by all the studies we’re going to talk about today, because it had no impact on the gestational age or how long after presentation the children were born. And of course I think we’ve seen that in our own practice too. You definitely can’t stop real preterm labor just with MAG or really anything else. So I think at least magnesium not working as a tocolytic is finally accepted. But in fairness, even that was still kind of a place until at least up until about 2010. 

Howard 09:28

Yeah, it was, and maybe in some places even after that. But that, of course, was despite us knowing from very high level systematic reviews and multiple trials dating back many, many years. I remember the initial systematic review that said this does not work, stop using. It was in 1991. Yet it took 20 years or whatever for people to finally pony up and say okay, yeah, it’s not a tocolytic, maybe we’ll do something different. 

Antonia 09:57

Yeah, so there’s your axiom. I’ve heard you say before that it takes an average of 17 years for confirmed scientific evidence to be integrated into practice. 

Howard 10:07

Yeah, well, not my axiom, Somebody else came up with it. But, yeah, it’s another example of that, whether it’s this or PapSrner guidelines or anything else, and thankfully not everything the COVID vaccine we did a little bit better on, but remember nearly half the people didn’t get the COVID vaccine and lots of folks don’t advocate for it, but historically that’s been the trend and basically a whole generation has to pass, where people who are invested in an idea have to retire if you will, and younger people say, hey, why were we doing that? 


And here we go again. 

Antonia 10:38

Perhaps Well, ok. So then, instead of starting with this new article, why don’t we start at the beginning quickly and bring us up to speed about the studies and then get to that new article? 

Howard 10:52

All right. Well, at least let’s get up to the time that this committee opinion was written in 2010. And then we can talk about what’s changed since then. So I always think you do have to ask the question. First of all, why magnesium sulfate? Of all the things, of all the medicines or chemicals or salts or compounds, or herbs or spices or whatever you could give someone, why would anyone think that magnesium sulfate would be effective for preventing cerebral palsy and unborn fetuses? And that question is very relevant because we could be talking about 1,000 other drugs or minerals or something else right now, but we’re not. We’re talking about magnesium and from a Bayesian perspective, you have to wonder why would anyone think of this to begin with, because you could do the same sorts of studies and develop the same type one and type two error rates about any different compounds, so pretest probability of the hypothesis matters. So why magnesium? 

Antonia 11:46

Well, you have talked about that and I think it’s because it was already being used to treat and prevent eclampsia. So people were familiar with using it on pregnant women on the obstetric ward and then people would notice secondary effects and maybe they would theorize that if it works to prevent seizures, maybe it can work similarly on the neonates as well. And when this all started it was also almost universally being used as a treatment for preterm labor, so it was even being used in that same gestational age group for a different reason. So there were already a lot of patients who were receiving it, so lots of data that could be generated about those patients and their outcomes. So probably not the best option if you just want to start from scratch and think what possibly could reduce the rewrapossi. But this was already available as far as being in use and then also being studied. 

Howard 12:47

Right. Well, that’s exactly right. But recognize that as the availability bias which diminishes its a priori probability of being an effective agent compared to a million other potential therapeutic ideas. But again, I raise this point because you do have to consider from a Bayesian perspective what the pretest probability that that particular salt of magnesium sulfate, at particular dose, even the preeclampsia doses, would prevent cerebral palsy in infants? And again, of all things in the world, what’s the chance that this one thing is able to prevent maternal seizure and hypertensive woman and stop labor and prevent cerebral palsy, all in one great thing. And it’s available over the counter and obviously the probability of that is very, very low. But due to the accessibility or availability bias we pursue this and to the detriment of ignoring other things that we could have been spending time with. That’s the great sin of tocolytics is that we spent 30 or 40 years ignoring other pathways to prevent or treat preterm labor. 


But yes, people were also just interested in the safety profile in the 1990s in particular of magnesium sulfate and how that affected newborns because it was so widely used. So people like me in the early nineties were arguing that folks should stop using it as a tocolytic. You remember 1991 systematic review says don’t use it. But the retort is always well, what could it possibly hurt? If it even helps one woman, why not use it? And these are the same sort of arguments. By the way, they’re a type of the same argument that was made around the drug DES, diethylstrabescerol, in the 1950s and 60s. Even after we had evidence that it didn’t work in systematic trials in the 50s people said well, but if it prevents one miscarriage yeah, I know your trial said it didn’t work, but even one miscarriage. 


So people on both sides of the argument in the nineties for magnesium were trying to prove that magnesium was safe or it was unsafe or something like that for the children who were exposed to it. 


Because if you could prove a harm, if you could prove that it was hurting children, that might give more kindle to the fire to stop using it. And if you could prove it was safe, maybe you could argue well, I know it probably doesn’t work, but I’m going to use it, just in case it helps one patient. So there were some preliminary studies that were done and a lot of these preliminary studies of course start out with retrospective chart reviews or things like that. And in 1995, nelson and Gethers did a retrospective case-controlled study in California of 155,636 children followed to age three and they found that 7% of the very low birth weight babies who developed cerebral palsy had been exposed to magnesium in utero, while 36% so a much higher number of the very low birth weight babies who did not develop cerebral palsy had been exposed to magnesium. In other words, very low birth weight babies who did not develop cerebral palsy were much more likely to have been exposed to magnesium in utero. 

Antonia 15:46

Right. So this was a potential morsel and hope for magnesium to have a benefit in this regard. I think it appropriately raised the question of does this prevent cerebral palsy? Based on this case control retrospective review. Obviously this resulted not demonstrate causation, but at least a potential association, and you’ve written about that study before and pointed out that only 117 infants of that whole group were actually studied for this question. As you said, it was a review of over 155,000 children, but only 117 met the inclusion criteria for using it for cerebral palsy prevention or that even had the cerebral palsy outcome. So from that perspective it already seems a little weak and a little bit of a stretch and kind of disingenuous to put so much hope on that one finding. But I guess you could spin it and make it sound more impressive. If you say they found this reduction in cerebral palsy in a study of 155,000 children, then if you say what they really did, which was they plucked out less than 0.1% of children from a retrospective, uncontrolled record review to make this conclusion. 

Howard 17:09

Yeah, definitely, and this is a lesson to always read beyond the headline and try to understand what the authors actually did rather than be impressed by the first or last sentence of the abstract. But I will say that when folks write about this paper they rarely say 155 children or whatever the number was. They always talk about a study of 117 children. They always talk about a study of 155,000. Children showed a reduction in CP from 36% to 7% with magnesium exposure. So definitely, how you present data tends to bias people and with rare data and rare outcomes it’s really easy. The data set is really influenced by lurking variables and minor differences in enrollment in the two sides. Differences in enrollment that themselves may not be statistically significant to create differences. So this is part of the problem with all of these papers. In any event, that was 1995. 


Then in 1997, Mittendorf and colleagues reported on a study that they were currently doing on magnesium as a tocolytic in the Lancet. Now this is a study, yet another study after 1991. That was a trial of magnesium as a tocolytic that just never died. But in any event they were forced to stop this study and they publish an interim analysis of the magnet trial is what they called it and this was published later. The whole trial was published in 2002, but in the interim analysis in 1997, they published this because they were concerned that they were seeing an increase in total pediatric mortality among the children exposed to magnesium at these higher tocolytic doses and this finding was statistically significant after controlling for every other variable that they were at least collecting in that trial. In fact, as I said, the IRB ended their trial about halfway through, at least in that form. They continued to collect samples and process and do what they could with some of the data and materials they collected and later published that. But they ended the trial due to this significantly increased risk of death that they were seeing in a magnesium group. So they published this in the Lancet in 97 as a warning to other physicians and researchers. 


Now Scudaro and company in 2000 followed up on magnet with a case control study at the Chicago Lying Inn Hospital in which they found that exposure to more than 48 grams of magnesium was associated with a 4.7 odds ratio of field death. And these findings were consistent with magnet, which indicated essentially that the higher dose of magnesium exposure, the more likely the adverse event like inner ventricular hemorrhage. Now note that the typical tocolytic magnesium doses that a kid would be exposed to on average was over 50 grams for 48 hour course. So you might get a load of four or six grams and then you might get two grams an hour for up to 48 hours. So it was common in those days for children to be exposed to 100 grams and this finding led advocates of magnesium for whatever opportunity to start investigating lower dose magnesium protocols to try to get whatever benefit it might afford, but avoid the inner ventricular hemorrhage or death that we’re seeing in those two studies. 

Antonia 20:33

Yeah. So for a time, after this interim analysis, people did become more animated about stopping the use of magnesium as a tocolytic, all the way up until David Grimes published an editorial in the Green Journal in 2005 where he said, frankly, it was unethical to use it. 

Howard 20:52

Yeah, it was a great time to be alive and I gained a lot of respect for David Grimes and love that editorial, but Grimes, I think, caught a lot of flak for that, and so this battle went back and forth. David Grimes, by the way, is probably the most respected statistician that’s worked in OBGYN and did a lot of very important work in the 80s and 90s and just even just documents about how to use statistics correctly, but when he did that, I think a lot of people were very upset with him. So, any event, people changed tack and started researching magnesium, specifically in relation to neonatal outcomes like mortality and death, to try to see is this really causing this or is it safe? And so then now we get into these six or perhaps seven big trials that have informed the neuroprotection debate. 


So the first one was published in 2003 by Carolyn Crowther and colleagues who also did this most recent study we’re going to talk about. 


This was called the ACTOMgSo4 sulfate trial, and they enrolled 1,062 women who were less than 30 weeks gestation. Her new trial goes from 30 to 34 weeks and they looked at infant death or cerebral palsy or the combination of the two by the age of two years old, and they found no statistically significant reductions in the occurrence of death or cerebral palsy, or both combined, when the two arms of the study were compared. So on the one hand this was reassuring that it wasn’t killing children, but on the other hand you might correctly conclude that it was just doing nothing at all, because it also didn’t reduce the rates of cerebral palsy. But in a subset analysis they found that children exposed to magnesium had lower rates of an inability to walk without assistance and the raw numbers of cerebral palsy were just 6.8% in the magnesium group but they were 8.2% in the placebo group. And although that wasn’t statistically significant, folks essentially claimed that that paper was a success and that magnesium improved outcomes because the raw number was lower. 

Antonia 23:00

By 1.4% absolute risk. That wasn’t even powered. 

Howard 23:06

Yeah, and low and not statistically significant Right right. 

Antonia 23:09

So I think that that where there’s a non-significant finding that still gets lauded and somehow embellished, is going to be a trend with all of these papers, because none of them have found statistically significant improvements in outcomes, but all of them claim to have been positive studies regardless in favor of magnesium. So Just want to underscore that point here and we’ll keep underscoring it. But would you say there’s anything wrong with that conclusion about substantial gross motor dysfunction or the ability or slash inability to walk for these kids? 

Howard 23:49

Well, the only criticism I would have is that that single finding comes from an underpowered subset analysis, and what findings from subset analysis are supposed to do is trigger a larger replicating trial with that as a primary outcome. We use subset analysis to mine data essentially and find something to investigate, not to make claims of efficacy. So you shouldn’t change your practice based upon some subset analysis, particularly when it’s an unexpected finding that hasn’t been seen before, and we hold ourselves to some due diligence with statistical significance. For a reason. It’s not unusual that the rates of cerebral palsy or death are numerically different in the two groups, and one group’s always going to look a little better than the other if the two digits are not the exact same number. But that’s why we have an assessment of whether they’re statistically significant. 


And in real life studies those numbers go back and forth as you continue to enroll patients or do different studies. So the hard part about science is distinguishing noise from data and avoiding the traps of pareidolia, where people see trends in data that aren’t real. And the mistake is when people look at non-statistically significant trends or things like that and then make bold claims about them. And of course, this is a rare outcome and it is hard to enroll adequately for, and so you do worry about type two errors as much as you worry about type one errors. In this case there have been no positive findings to talk about for a type one error. But sure, if you’re under enrolled you have a good chance of a type two error. But you know type two errors are less harmful than type one errors are, and it’s up to the burden of the researcher to design studies adequately to avoid type two errors associated with under enrollment. 

Antonia 25:33

It seems like people are more afraid of type two errors, though, like they’re afraid of missing something that was actually there. 

Howard 25:41

And that might be the case. But we have to balance that with the fact that the majority of published findings have type one errors in them. So if we change our sensitivity too much to avoid type two errors, then we cause a bunch of type one errors. And frankly, if something is profoundly beneficial say the COVID vaccine for preventing COVID or something like that, there’s little chance of type two error. And that gets back to pre-study probability. How likely is it that this works? Well, if it’s not very likely, you’re not going to find something, and then you can talk about type two errors all you want. But if something is, has a good magnitude of effect and it’s highly probable, the rate of type two errors goes down tremendously. And I say all that because essentially the crowd who supports magnesium has just argued that all these papers had type two error. 

Antonia 26:29

Well, so far, just after this study, with the slightly increased benefit of walking without assistance, no one in the mainstream OBGYN community was really arguing that they should start using magnesium for neuroprotection yet. So that was still to come, with some, I guess, subsequent studies. 

Howard 26:49

That was 2003. And around that time Mittendorf did publish the 2002, the complete data from the magnet trial at the same time that he referenced. So we referenced five years before. 

Antonia 26:59

Yeah, so again, that’s the one where they had stopped it early because of the concerns that magnesium was harming the babies. 

Howard 27:07

Yeah, and although this trial wasn’t done specifically to look at prevention through Bopals remember it was a tocolytic study so sometimes it’s excluded by folks who do systematic reviews they did enroll 144 children and they did a lot of analyses with that data that had things comparing umbilical cord blood levels of magnesium and things like that, and they found essentially that in addition to the increased risk of death that led to the trial being stopped, that higher levels of magnesium sulfate concentration were associated with more rates of intraventricular hemorrhage. 


So I think around that time, given the available data, a general idea emerged that maybe magnesium might actually be associated with less surviving children with cerebral palsy because it came at the expense of a mortality risk for very vulnerable fetuses who would be at risk for intraventricular hemorrhage and make their intraventricular hemorrhage worse, which likely would result in death, but eliminate a severe intraventricular hemorrhage who might have survived with cerebral palsy. So it became very important then to follow a combined outcome of death plus cerebral palsy, because you might be trading one for the other if that’s actually true. And there was also a lot of interest again in the most appropriate dose, avoiding that greater than 50 grams exposure. So now most of the studies going forward. After Mittendorf and Magnet, use smaller doses to avoid that 50 gram threshold that was identified with an increased rate of intraventricular hemorrhage. 

Antonia 28:41

Yeah, and that mechanistically could make sense too, because we sometimes think of magnesium as a hemorrhage risk for the mother, and it is listed as a risk factor in some of the little algorithms for calculating postpartum hemorrhage risk. So what dose exactly? Is it even a true association? And then, if so, what dose? 

Howard 29:03

And the common mechanism. There is calcium channel blockage, something that might impede the uterus from relaxing, but also something that may be associated with the stability of the brain matrix. 

Antonia 29:16

Most magnesium protocols at that point and that I’ve seen still beyond that point did include a bolus with a continued maintenance dose. But then these new studies started looking at just giving a bolus just for neuroprotection and not for or a one gram maintenance for a shorter period of time, yeah, and so there are a lot of hospitals today that are doing it that way as well, but I really haven’t seen that in my practice. 

Howard 29:46

Well, the thing is, if you’re still seeing four or six gram boluses followed by two grams an hour, that doesn’t come from any of the trials. So that’s actually in violation of the ACOG committee opinion too, because none of the trials did that. 

Antonia 29:58

Yeah, yeah. 

Howard 29:59

Okay. Well, the next study came out in 2006 and this was called the pre-mag study, and this was by a merit and that’s a French name, so if the T is silent, please email me. But they enrolled 688 infants well, fetuses under 33 weeks gestation with just a four gram bolus over 30 minutes, no chaser, and outcomes that they looked at included white matter injury, death or combination of white matter injury and death. Ultimately, in their final analysis, they found that none of those things were statistically significantly different and, to their credit, rather than just claim a benefit that didn’t exist, they at least said that their results were inconclusive. Now, even that’s a little interesting when you do a negative study and then you just say that the results are inconclusive but, in fairness, there is a possibility of type two error. 


Of course that exists in any study. So this was a negative study. It found no benefit and they found no improvements in the things they looked at in their study. 

Antonia 30:57

And this is another study where, if you look at the raw rates of white matter injury, for example, there was 10% in the magnesium group versus 11.7% in the placebo group, and then also the outcome of death was 9.4 versus 10.4, respectively, magnesium versus placebo. So again, some people could still look at that and say there was less white matter injury and less death in the magnesium groups. So even though that wasn’t statistically significant, maybe it’s a trend. 

Howard 31:30

Right and that’s exactly what people have done. But again, I would encourage folks to do some basic reading about null hypothesis testing and understand that not statistically significant is not statistically significant. If you just take random data and sort it in various ways, well, it’s never going to produce the exact same number, the actual ordinal number, in two different columns of data. I mean it could by chance, but it’s very rare. So it’s just merely a question of chance as to which group has a little more or a little less, and that’s why we do testing to see if the data is significant. It also could be related to some lurking variable we don’t understand or didn’t control for. Remember, association doesn’t equal causation and we could do a whole other hour on if this meets the nine criteria of approving association and causation, because it best meets one In any event. 


There are other ways of using data like this that’s not statistically significant and that’s where there’s actually a lot of interest in the Bayesian approach to doing analysis, because you can sometimes see benefit in non-statistically significant data with Bayes. Or you can do systematic reviews or meta-analyses and try to combine data from different, smaller, underpowered studies to protect against the type two error. But this study was a negative study, just like Crowther’s original study, and the only positive study we’ve had so far of the three or four we’ve talked about is the Mitten Door study, which showed more death and inner ventricular hemorrhage with higher doses of magnesium. So it was positive, but in the wrong way. 

Antonia 33:06

Okay, so it gets us through 2006, but we’ve still got four more years between then and when this ACOG committee opinion came out. So what’s next? 

Howard 33:15

Well, there’s another study from 2007 that is almost never included by folks who were arguing for magnesium as a neuroprotective agent which will also be a trend and that’s the 2007 Magpie study, a large international study. 

Antonia 33:30

Okay, yeah, that was actually looking at whether magnesium sulfate was effective for preventing seizures in women with what we used to call mild preeclampsia or preeclampsia without severe features. 

Howard 33:45

That’s right and it was a large international study that involved many thousands of women I think over 6,000. And it’s the main reason today that we don’t give magnesium to women who have preeclampsia without severe features. But in that many thousands of women, there were 1,544 preterm infants who were exposed to magnesium for that reason, and the authors did collect mortality rates at 18 months along with cerebral palsy rates at 18 months for those exposed children, and it did analysis of this, and so this actually ends up being the largest study up until this point of people, of pregnancies that were randomized for maybe a different reason, but nevertheless were randomized to see if preterm infants have less death or cerebral palsy, in this case at 18 months of age, when exposed to magnesium sulfate versus placebo. Now this would be at doses used to prevent maternal seizure. So not just a bolus, but the bolus and maintenance rate that you’re probably familiar with for preeclampsia with severe features or even what we were using for tocolysis. 

Antonia 34:50

Okay, well, and I’m sure that the MAG supporters would immediately argue that the preeclampsia is a confounding variable here that could affect cerebral palsy in some other way. But regardless of that, what did they find in the MAG PI study? 

Howard 35:06

Well, they found that there was no statistically significant difference in death of cerebral palsy at 18 months in the two subgroups and just so you know, the trend was actually towards more death and cerebral palsy in the magnesium group than in the placebo group. Now I’m not going to claim and do some meta-analysis including that data to use that trend away or negatively for magnesium, to claim that magnesium is harmful in that sense. But this is what happens in trials is that you take noise in trials and just random data. That’s not different. But it’s going to fall back one way or the other, flips back and forth. So in this case it goes against magnesium, just by chance. In the other case it may go slightly for magnesium, just by chance. That’s what non-significant data does. One study shows the difference going one way, the next one does the other way. It’s random. 

Antonia 36:00

Okay, and as you said, the study is often excluded from systematic reviews on magnesium for fetal neuroprotection, but it does appear to be the best evidence, up until that point at least, about both fetal neonatal mortality and cerebral palsy. So now it’s looking pretty grim for the magnesium neuroprotection. 

Howard 36:23

Well, hope brings eternal. So that brings us up to the BEAM trial in 2008 by Rouse and colleagues, and I’ve written extensively about this paper in many different formats over the years, including in my book. As an example of some methodologic mistakes that are often made in papers I will just set out. Say, from the beginning there is this idea of prevailing bias. So for many years we had a prevailing bias that progesterone prevented miscarriages. So we made mistakes like DES, which I know wasn’t exactly a progesterone, but it’s the same bias and people still use it because of the prevailing bias, despite the scientific evidence. We had a prevailing bias at tocolytic. We have a prevailing bias that tocolytics are effective, despite no trials supporting that idea, and that magnesium was an effective tocolytic. We had a prevailing bias that 17-hydroxyprogesterone caproate would prevent recurrent preterm birth. And prevailing bias we know affects how studies are done. So one of the things that we need to do in studies is measure bias. And I say all that tangent because just the name of this trial BEAM beneficial effects of antenatal magnesium that’s not the scientific method. If you believe that it’s beneficial, try to disprove it. That’s the scientific method. If you’ve already started a trial called the Beneficial Effects of Antenatal Magnesium. There’s some bias involved in how you design the trial, your inclusion-exclusion criteria, so-called researcher degrees of freedom. 


I could go on for a long time about this. But in this trial they randomized and Tinniter randomized 2,000 women. Based on their power analysis, they actually over-enrolled 2,241 women and despite that, admittedly refused to use a correction for that in doing their statistical analysis for their p-value. I’ve written about this elsewhere. But these women were randomized to either 6-gram bolus then 2-grams an hour for 12 hours, which again is only 30-grams, so it avoids that 50-gram threshold versus a placebo. And these patients were all between 24 and 32 weeks. Most of them had ruptured membranes not all of them and their primary outcome that was selected was death plus severe cerebral palsy, which again made the most sense at the time to study that, given the things we’ve talked about. And so they did that and they found no difference in the two groups. 

Antonia 38:46

Okay, so that was 2008. So that pretty much gets us up to the time that this 2010 committee opinion was written, and this is also the study that people often cite as the reason why today they get magnesium to prevent cerebral palsy, and it aligns pretty closely to protocols that I’ve seen in my training and practice, that 32 weeks and below 6-grams followed by 2-grams an hour. But if they found no difference, as you said, then I think you have to explain a little bit more about why it was still positive. 

Howard 39:24

Yeah Well, when you name your paper Beneficial Effects of AntenalMagnesium and you find no difference, you’ve got to do some work here. So I’ll put a link to an article where I talk about a few of these things. I will say a lot of the stuff in my book is, I think is more on point, but also some of the statistical stuff that they did that we talk about there and in my book. There was a lot of things. There’s a lot of things to talk about. We could spend a whole hour just on the BEAM trial, misallocation of twins. There’s a lot of issues here that affect this data. 


But essentially what happened was they did a subset analysis here we go again and where they teased out moderate or severe cerebral palsy alone and they found 20 cases out of 1,041 allocated patients in the magnesium group compared to 38 cases out of 2,095 in the placebo group, and that difference was statistically significant. Maybe that whole other issue here Now when they corrected that, that’s the raw numbers. But they also provided data corrected to exclude pregnancies that had major congenital anomalies, which makes sense, and the subsequent p-value for that data point was 0.0491, just the chin’s hair under 0.05. So that’s the p-value that gave us the ACOG bulletin and the use of magnesium sulfate, just 9,000ths, 9,000th of a point below 0.05. And that’s where elements like bias and any statistical manipulation or misallocation of subjects or inappropriate exclusion criteria become really important. 


And that was again without them having done a power analysis correction for over-enrolling right, right or appropriately well, yeah, but even in this subset analysis, the problem with the subset analysis is they’re not appropriately enrolled, right? 


So, one phenomenon that happens in an inappropriately enrolled study is that one group it goes ahead a bit. The other group head goes ahead again. Imagine this If I did just a simple study you can do we take a quarter and we flip it, and we flip it 10 times and I get seven flips out of 10 or heads. Well, we all know it would be inappropriate for me to conclude that the head, that the quarter, is biased towards heads. For example, somebody else does a study and they flip it and they get six out of 10 or heads, and maybe somebody else does it and they get eight out of 10 or heads. 


Well, you cannot conclude from those underpowered studies that the coin is unfair. Nor would it be fair for you to meta-analyze that data and say, well, gosh, we think that this coin has 70% of the flips or heads. It’s essentially what we’re talking about here. But it’s a well-documented fact that data, in working your way up through ideal enrollment, that data will go in one trend and it’ll flip back. The other stuff that looks statistically significant will turn out not to be. Later on, things come in spurts and that’s why appropriate enrollment not under enrollment, not over enrollment is appropriate to make these sorts of conclusions, and so that’s the problem with subset analysis. 

Antonia 42:38

Okay, well, we don’t have to get into splitting too many hairs on the P-value, because you do that quite extensively elsewhere. But I think the key point is that everything in practice changed based upon that incredibly marginal, questionable P-value of a subset analysis of the bean trial, and we normally generally should not make clinical changes based upon subset analyses or single studies due to how frequently we see type one errors. 

Howard 43:12

Yeah, that’s right. So lots to potentially talk about. But we’ll take it for what it is right now and I think it’s efficient to say that the best the BEAM study really did was raise the question a little more and demand more studies. If you were being fair about it you would say, well, we need a better study. But in practical terms, I think people said and that’s what’s in the ACOG committee opinion, I think they said, well, what’s the harm? Maybe it does work, Maybe it works and it’s not hurting anybody. We thought for a few years that it might be causing death or hemorrhage and it maybe it isn’t. So maybe it works, Maybe we should try it. And the BEAM trial, probably more so than finding a positive benefit for prevention of CP, elade some of the concerns about death and harm from magnesium exposure that people had just a few years earlier. 

Antonia 44:02

Although the trend was more death in the magnesium group and not study. 

Howard 44:06

Well right, that’s exactly right. That’s why the combined outcome which they were powered to study in fact the over enrolled of death plus cerebral palsy was not statistically significant because there were more deaths in the magnesium group, essentially balanced out by less cerebral palsy. But when they looked at the deaths alone, because they were underpowered, they didn’t find a statistically significant difference. In an honesty, if you’re counseling patients based on the BEAM study, what you should say is there’s no difference in the probability that your child lives survives with cerebral palsy because it’s the combined outcome that was powered and wasn’t statistically significant. 

Antonia 44:46

Yeah, but isn’t that just the type two error due to how rare it is the death? Yeah? 

Howard 44:52

It certainly could be right If they didn’t find that the death alone was more statistically significant. It’s clear that one of those is a type one or a type two error, or both even are possible because the combined outcome was powered appropriately. 

Antonia 45:09

The other thing I noticed is in the ACOT Committee opinion they list the odds ratio from the BEAM trial as a reduction of cerebral palsy, being 0.5, and that’s commonly cited in other review articles as well. But I think that’s the rate that’s unadjusted for congenital anomalies. 

Howard 45:30

Yeah, it doesn’t change that much when you do the adjusted numbers, but it’s a sign that the data for magnesium for cerebral palsy prevention is always presented in the most favorable way and sometimes in a bit disingenuous way. 


I guarantee that there are residents right now learning that when you give magnesium to children it reduces the rate of cerebral palsy by 45%, which isn’t even a fair statement to make about odds ratios, because it’s not a relative risk and certainly not a fair statement to make in the context of everything we’re talking about. 


The newborns with congenital anomalies are going to have other issues going on besides just prematurity. In fact, we’re talking about prevention of cerebral palsy and hopefully we all recognize that something like 95% of cerebral palsy are due to in utero events well beyond our control during the hours leading up to birth. That’s one of our biggest problems in OBGYN, at least with attorneys, is clarifying that fact and this kind of implies that we have something immediately under our control, as if the cerebral palsy was happening at that moment. I’ll also say that the trial looked at differences in objective outcomes. The BEAM trial did in children called Bailey scores and these were no different in the two groups. Even in their subset analysis, these were no different. If anything, this would imply that cerebral palsy was just overdiagnosed in some of the children, because the more objective scores were not different in either group. 

Antonia 46:57

Okay, so again we’re at 2010. Now, with this ACOT Committee opinion, the world completely changes its practice. Everyone starts using magnesium per the BEAM trial criteria. So return labor 24 to 32 weeks, six grand bolus to a grand maintenance for 12 hours. And of course, that committee opinion has not been updated since then, just reaffirmed. But it still doesn’t recommend that you use it. It only talks about what you should do if you’re considering using it. So what has happened since the BEAM trial? 

Howard 47:33

Well, several important things In 2014,. Doyle and colleagues published in JAMA a school age outcome analysis of the infants from Crowther’s original 2003 study and they looked at things like mortality, cerebral palsy, motor function, IQ, academic skills and a bunch of other functional and cognitive and behavioral things in the children and they found that the children exposed to magnesium were not different in any of those categories, testing multiple hypotheses and all the things they looked at compared to the children in the placebo group. 

Antonia 48:05

Okay, so that’s a tiebreaker for the original Crowther study that set all of this off, because that original study had these non-statistically significant trends. But following these children up in a wide variety of outcomes later in life shows that they are not different. So that tends to confirm that this was a completely negative study. 

Howard 48:27

Yeah, that’s right and that’s an important thing. When you have a study with, you’re wondering if there’s a type 2 error, this is a great way to add evidence to that study and continue following the outcomes in the same subjects for many years to answer that question. And it looks like there wasn’t a type 2 error. So that study shouldn’t be used to claim positive benefits for magnesium and similar to that in 2014,. And again, if anyone who speaks French is listening, please feel free to yell at me, but Cholet and colleagues sure I butchered that did a follow up of the Pre-Mag study from 2006 by Merritt and they looked at a similar battery of outcomes in school-aged children who had been enrolled in that trial and again, they found no difference in any of a variety of outcomes. 

Antonia 49:13

Okay, so we’ve got the Crowther and the Merritt and really the Rouse trials, the being all negative studies. 

Howard 49:21

Well, at least add to the follow up. I mean, rouse hasn’t had, to my knowledge, a school-aged children’s study done. Maybe they’re doing one, but in terms of that, these studies that were questionable. Maybe you’re worried about a type 2 error legitimate thing to do, not disagreeing but now we have evidence that they’re both negative studies. That tends to exclude type 2 errors. And then, of course, we have Middendorf study that showed harms for magnesium, so you can’t count that one. And then MagPie was a negative study with, if anything, trends the other way, so no one’s using that one in their systematic reviews. So that essentially deals with all the studies prior to the BEAM trial that, by the time we get to 2014, and the BEAM trials, all that little thread of hope may pinned on. 

Antonia 50:06

Okay and yeah. At that point it’s only the thinnest and most questionable suggestion that Mag before early preterm birth could benefit the babies. So Well if it Going. 

Howard 50:19

So the systematic reviews that came around that came out in 2009, included those two trials we just discussed. 


The ACOG committee opinion includes two trials and they use all that data together, but I think that’s unfair in light of the fact that we know that the school-aged children in those studies have no differences in outcomes. 


Now, in 2019, I think probably the most important study you’ve never heard of was published. It’s actually just a research letter and there’s definitely a bias right now at prevailing bias in the obstetric community, and that includes editors of our journals and especially in the MFM community and people who might referee journals and review articles about using magnesium as a neuroprotective, just as there was a bias towards using it in tocolytic for many years after evidence fell to show its benefit. But in 2019, in JAMA pediatrics leave it to the pediatricians, although there’s an obstetrician involved here too. A paper was published by Bethany Stetson and colleagues that is a before-and-after cohort analysis of children exposed to antenatal magnesium. So they identified 3,736 kids who had been matched cohorts delivered before and after 2009 when the BEAM protocol was implemented in their institutions and they found no difference in neurologic outcomes. They found that the protocol did not affect the neurological development or rates of sensory problems or anything in their patients in a real-life implementation of the BEAM protocol. 

Antonia 51:46

Yeah, you’re right, I had not heard of that. I don’t recall ever hearing about that in any OBGYN meeting or reading about it anywhere. But with those numbers that’s actually the largest study to date about how implementation of the protocol would affect outcomes. 

Howard 52:03

Which is important. If you’re worried about how to account for the possibility of a type 2 error, this is the kind of study you need to do that. So they actually concluded and I’ll just read it. I love this author and this paper so much after reading this paper. But their last paragraph is Administration of magnesium sulfate is a resource intensive intervention that is not without risks. 


Given our findings, the role of magnesium for fetal neuroprotection may need re-evaluation at the national level. Physicians should evaluate the long term impact of clinical interventions and mandate not unique to magnesium trials for prevention of CP. Many industries have developed lean thinking tools, such as the plan-do-check adjust cycle that can drive evaluation of healthcare recommendations adopted following large randomized trials, and I just think that’s wonderful on so many levels. The real impact of data like this and the other two follow-ups of school-aged children is the sort of evidence that we need to assess the potential benefits of something that might be there, that we might have seen in a trial but didn’t rise to the level of statistical significance, or that might not have been replicated in other trials. And that’s exactly what we’ve needed for this issue and their data, which involved more children than the BEAM trial did in fact, more data, I think, than all three trials added together In a real world setting of implementation of this very carefully done, showed no difference in the kids being exposed to it. 

Antonia 53:26

All right. Well, I think we do speed it up because we still have two more studies to discuss in the right amount of time. 

Howard 53:32

Okay, well, sure. So Wolf and colleagues in 2020 did their own study in Denmark, where they randomized 560 pregnant women before 32 weeks’ gestation to receive a 5 gram bolus, followed by 1 gram an hour, and they looked at a primary outcome of moderate to severe CP, along with secondary outcomes of mortality and other morbidities like blindness and stuff like that, and they found that the outcomes were not statistically significantly different. 

Antonia 53:58

But, like some of us other studies, they still claim that magnesium was effective for reducing cerebral palsy and they even wrote a follow up systematic review, claiming that, after publication of their non-significant data, that we finally have compelling evidence that magnesium was effective. I feel like at this point, if it really was effective, don’t you think that neonatologists would even be giving after birth? 

Howard 54:25

That’s a wonderful question. 

Antonia 54:27

But they don’t. 

Howard 54:28

Yeah, it’s a really amazing thing. It’s just weird and bizarre that we sit here and talked about negative study after negative study after negative study and in this case, the author writes a paper and finds no finding and then claims he did and writes immediately, writes at the same time, concurrently, publishes a systematic review claiming that he solved the problem finally. So again, they were just saying that their trend of reduction, which wasn’t statistically significant, was evidence that it was effective. So it’s this new age where we take non statistically significant data and count it as significant, and where trends matter in the world again, where 80% of the papers published with statistics significant data end up being falsified in felling replication trials. Guess what it looks like when you start to include all the studies that merely trended towards something. 


So this is a dangerous experimentation by folks trying to parlay negative studies into something significant. Now, in this particular case, they use something called link trial, sequential analysis, because they did a little pilot study to try to determine their power, essentially that they would need that they would need for the study, and a lot about this was written at the time it was published and they use this technique to claim significance that they didn’t find. And for the sake of sparing the listener and because we’re about out of time, I’ll just summarize this by saying that the Cochrane group says that this statistical technique is not reliable and shouldn’t be used. 

Antonia 55:56

I think people probably realized that was all a bit controversial because we haven’t heard a ton about that study or that systematic review. I vaguely remember, just maybe, some headlines in my little medical journal news feeds that I get in my email junk email box. But in their abstract conclusion they say, quote magnesium sulfate before 32 weeks’ gestation decreases likelihood of moderate to the earth cerebral palsy. So they say that very definitively as if they actually found a finding which they did not. 

Howard 56:30

Yeah, it’s a brave new world where the study doesn’t matter as long as you know how to paint it in some way with some statistical tool. So it’s, I don’t know. It’s all just bizarre when you actually talk about it. 

Antonia 56:42

Okay, well, let’s get to this past summer’s article by Crowther. 

Howard 56:46

I will say. Before I talk about that, I will say if you’re a resident or a young faculty member, who a lot of our listeners are, just wonder to yourself did you know what the magnitude of effect was and what the data was? And ask that question about everything that you do Like it’s your responsibility to know the magnitude of effect of an intervention and how certain we are about the data and not teach it with a higher level of certainty or a higher benefit or magnitude of effect than is actually known. And even when people do meta-analyses and twisted and contorted ways, it’s always high risk of bias, low certainty, and so just appreciate that that probably means it’s not a real finding and that people don’t read the whole paper. They read the headline. 


Okay, well, the new Crowther study. So her original study again, which was not statistically significant and which now has a long term follow up of the infants showing no difference in a battery of cognitive, functional and other tests. They only included pregnancy except to 30 weeks. So she set out to do another trial that included 30 to 34 week pregnancies and that’s what was published now this summer in JAMA. So they enrolled 1,433 pregnant patients and they found no statistically significant difference in the two groups. The only thing that was significant was more respiratory distress syndrome in the magnesium exposed children and, as you said earlier, more postpartum hemorrhage in the magnesium exposed pregnancies, and their conclusion is that it was ineffective, and the accompanying editorial in JAMA calls for an end to using magnesium sulfate at least over 30 weeks, and also some, some forlorn about all the opportunities for reducing cerebral palsy that we’re missing by focusing so much on this one intervention. 

Antonia 58:30

Interesting, so maybe this could be the start of the end of magnesium for CP neuroprophylaxis. 

Howard 58:38

Well, we’ll see. The rate of CP plus death in the magnesium group in this current study was 3.3% compared to 2.7% in the placebo group. So in this study the numbers fell out on the other side, but of course they’re not statistically significant and one has to wonder if it went the other way, if those numbers were reversed just by chance, if folks would be claiming again that this was the evidence that it worked, because that’s essentially what’s happened, and when it goes the other way we just bury it a little bit. So this is the danger of people latching on to non-significant trends. 

Antonia 59:13

Well, we’ll just have to see if or when major academic medical centers will stop using it for this indication. But I have a suspicion that ACOG or the AAP would actually have to come out with a new revised committee opinion and say clearly it does not work for cerebral policy prevention and is not recommended. And until they do that I think it’s probably not going to budge. 

Howard 59:39

Yeah, even though the current opinion doesn’t say it is effective, or? 

Antonia 59:42

that you should use it Right. 

Howard 59:43

Yeah. So I think once you start doing something without a firm evidentiary basis, then it’s easy to slide further and further into all sorts of things. Now there was also a secondary study of some of the children in this study in a magenta trial, and so this group did brain scans of 109 children from that study and they found no favorable differences. The differences they found, if anything, showed less fractional isotropy and don’t worry, I read about this for you, that’s not a good thing in the infants exposed to magnesium. So, if anything, their study found evidence of harm. 

Antonia 01:00:18

It really is fascinating, after going through all this, that there’s so many people that seriously believe that even as little as 20 or 30 minutes of exposure to magnesium will somehow have such an enormous impact on the Gable brain and really on their whole life. 

Howard 01:00:35

In fact, yeah, in fact, the way it’s being taught is like we need to infuse it right before we go back for the C-section and get that in there Right before, like it’s a prophylactic antibiotic. That doesn’t come from any of the studies and and just the miracle that this Magical ion would somehow, in 20 minutes of exposure, would do all this. It’s just really incredulous. But all this reminds you, or should remind you, of the story of 17-hydroxyprogesterone  caproate after the last several years, or of the three decade use of magnesium as a tocolytic, where people grasp any straws or justification to continue using it without any Evidence of benefit and just like in magnesium, where now we’re seeing these school-aged children differences and things like that, or real  17 hydroxyprogesterone caproate level implemented studies in . We’re finally seeing follow-ups of kids and we’re, more importantly, seeing population implementation studies that show it doesn’t work. It didn’t decrease the rate of preterm birth. 

Antonia 01:01:31

Okay, well, we’re definitely over time. Can we still briefly mention the systematic reviews and meta-analyses on these, because they all tend to conclude that magnesium sulfate is still beneficial against cerebral palsy. 

Howard 01:01:44

We should do an episode on systematic reviews, but I’ll say a few things. Firstly, all the meta-analyses that that are out there were essentially the ones that people cite were written by some of the authors of these studies who were advocates for magnesium. Again, Some of them were written contemporaneously to the publication of the studies. So again, the question is can a meta-analysis take a series of negative studies and turn them into a positive result? Well, not when the studies themselves were underpowered remember the example of the coin toss. The studies were so Disparately done, with different inclusion and exclusion criteria in different doses. 


It is possible to do this, and one of the things that people will shoot back is Cerebral palsy is a rare outcome and it’s hard to design a large enough trial to capture meaningful differences in in different outcomes. And In preterm birth is a rare outcome. So you got a rare outcome and a rare outcome. How do you do this? And that’s all justified, but, just as the ACOG committee opinion stated, you can’t do a meta-analysis of studies with such different inclusion and exclusion criteria, using different doses of the medicines on different patient populations. That’s not how meta-analysis was intended to work. The results of that are not going to be reliable. They’re highly subject to any amount of bias and any just one of the studies. So we could spend again a whole hour talking about Choices made and how you homogenize heterogeneous studies and things like that. But unfortunately there are almost more systematic reviews and meta-analysis nowadays being published in real research because it’s an easy thing to do if you need a publication and it’s become an area rife for statistical manipulation. But again, the coin example I think is a good example. If you’re meta-analyzing well done studies that met enrollment criteria and had a point zero five percent error rate for type one errors and an eighty to eighty five percent rate for type two errors, avoidance and you take all those and you put them all together With similar study, similar inclusion, exclusion criteria is you can do something fantastic with meta-analysis. That’s what it was intended for. But you can’t do that when the studies you’re cobbling together are Underpowered, insufficient and done differently garbage and garbage out. 


I will say that if there is a Beneficial effective magnesium for children in prevention of cerebral palsy or limiting other neurologic outcomes, then we should see that as a consistent Effect in these studies, in all the studies we mentioned, not just the ones folks choose to include, and we would see that in the follow-ups as well. 


In fact, you might even see it magnified in the follow-ups when it’s easier to test IQs and academic and Academic achievements and things like that in school-age children. We should even detect it at a heightened level, but we don’t. And Finally, in that 2019 study that’s the wonderful thing about that longer term follow-up lots of different neurologic outcomes and we just don’t see it. Instead, we see this pattern that we’ve talked about of data that just hovers above and below average, above and below not all not significant. Above and below average. Sometimes it favors mag, sometimes it doesn’t. But if you selectively Take the ones that just favor mag and exclude all the papers that don’t, and then you have a biased author of the paper using questionable techniques to heterogenize the data, then you can come up with a Ben analysis that tends to support it. So read those with utmost caution. 

Antonia 01:05:09

So this is the thing we do for no reason, then I know your answer and I agree with you and I think everyone out there at least should concede that, based on that newest study in JAMA, that using magnesium to prevent cerebral palsy after 30 weeks is Definitely a thing we do for no reason and you should just go ahead and stop it, it’s the thing you do for no reason. But based on everything else that we just talked about that came before that study, really using it for anyone for any earlier preterm births, for neuro protection, is also a thing we do for no reason. 

Howard 01:05:43

The editorial in jam and actually calls for us to stop using it. And again they point out the idea that we’re ignoring other things that we could be doing once we get a treatment. That’s what we do with tocolytics. We just ignored research on things. It could be effective at preventing and treating preterm labor because we thought we had a solution, and the editorial points out a lot of areas where we could be doing more to enhance that. 


And the other thing I’ll caution people is we’ve prevented a lot of CP that’s due exclusively to prematurity due to amazing advances in NICU care over the last 20 years. 


So, just like the problem we have with a lot of issues, the NICU side of thing has so improved that don’t be careful not to give the benefit of improvements in CP and premature babies and give the credit to the wrong place. 


It’s important to have controlled trials. I say that because your own hospital might say blank percentage of kids had CP before 2009 and now, but that’s not due to the magnesium and in this importance of that 2019 paper, because they do a really good job of controlling for all those things so well. It’s going to persist at this point, obviously, and maybe the bravest folks will stop doing it after 30 weeks, as the JAMA editorial says you should and start to question it, but at least understand what the confidence is in the data and what the benefit of effect magnitude of effect might really be. Unfortunately, if people view it as harmless, they’ll keep doing it, and that’s why I like the comments about resource wasting, the potential for harm, mistakes made with magnesium dosing and thing like that in the 2019 study. If it doesn’t help, only one person in a thousand has to be significantly harm for it to be a bad thing, and that’s the lesson of DES. 

Antonia 01:07:29

Well, we really should wrap up now. We’ve gone a little bit over, so thanks to the listeners for hanging in there, but the thinking about OBGYN website will include links to all of these studies, and I think you made a pretty nice or you picked a pretty nice little summary the chart of these studies too, so you can post that as well, and we will be back in a couple weeks with something new.