Episode 8.9 Intrapartum Fentanyl, IUDs and Breast Cancer, Pretrm Test and Bias in Studies
Can IV fentanyl use during labor actually be safe for mothers and newborns beyond six centimeters of dilation? Join us as we unpack a pivotal 1989 study and question the stereotypes surrounding fentanyl use in labor.
Our second segment tackles sensational claims about IUDs and breast cancer, as we dissect a Danish study recently published in JAMA. Listen as we unpack the methodological nuances and confounders that mainstream publications have overlooked, and explore the media’s role in shaping public perception with incomplete or skewed interpretations of data.
This episode closes with an in-depth look at the biases that pervade pharmaceutical research, particularly when funding sources have a vested interest in outcomes. By examining the “sponsorship effect” and its implications on research validity, we highlight the need for rigorous, unbiased scientific inquiry. We examine the evidence for the new Pretrm Test for preterm labor risk. Learn about the challenges of implementing biomarker testing for predicting preterm births and why high-quality evidence must remain the cornerstone of clinical practice. As we celebrate our 100th episode, we discuss the need for truth and transparency in scientific research—creating a future where clinical practices are driven by reliable, evidence-based insights.
00:00:02 Navigating IV Fentanyl Use in Labor
00:07:56 Debunking Misconceptions About IUDs and Breast Cancer
00:22:17 Evaluating Biomarker Testing for Preterm Birth: The PreTRM test
00:38:50 Conflicts of Interest in Medical Research
00:45:15 Biased Influence in Medical Research
Links Discussed
Fentanyl citrate analgesia during labor
Breast Cancer in Users of Levonorgestrel-Releasing Intrauterine Systems
Progestin and Breast Cancer Risk: A Systematic Review
Prediction and Prevention of Preterm Birth: A Prospective, Randomized Intervention Trial
Funding of Clinical Trials and Reported Drug Efficacy
Circulating Angiogenic Factor Levels in Hypertensive Disorders of Pregnancy
The Price of Knowledge: Industry-Sponsored Studies in the Era of Evidence-Based Medicine
Transcript
Announcer: 0:02
This is Thinking About OB-GYN with your hosts Antonia Roberts and Howard Herrell.
Antonia: 0:17
Howard.
Howard: 0:18
Antonia.
Antonia: 0:19
What are we thinking about on today’s episode?
Howard: 0:22
Well, first we should acknowledge that it’s our 100th episode, so, as you can see, we’ve switched things up around here.
Antonia: 0:28
Now you’re going to be the one responsible for telling folks what we’re going to talk about.
Howard: 0:33
Well, we’re going to talk about a new article that’s all over social media about IEDs and breast cancer and, honestly, most of what we’re going to talk about is an answer to a great listener question. That’s going to take us down a few rabbit holes, but first we need to do a thing that we do without evidence, and I guess that’s on you now.
Antonia: 0:52
Oh yeah, this part is turned around too, I guess. Okay, we’ll get used to it. Okay, well, how about avoiding IV fentanyl in labor after some arbitrary threshold, let’s say six centimeters dilation, which we would call active labor?
Howard: 1:09
Sure, all right, this is a good one. So the spirit of this is not to give the mother a narcotic dosage that will be present at some high enough concentration in the newborn’s bloodstream that might tend to cause respiratory depression. For the newborn, the elimination half-life of fentanyl is actually 219 minutes, but we don’t tend to think of it as lasting that long because the clinical effect wears off much sooner than that and it’s accepted practice to dose it every hour or so, as needed during labor.
Antonia: 1:39
Yeah, and we do know that in some special, maybe more rare cases where moms medically cannot get an epidural, they can be given a fentanyl or I should say remifentanil PCA and they can dose that right up until delivery. We know remifentanil has a much shorter half-life but in those cases it wouldn’t be absent from the baby’s circulation. That mother could dose it right up until while she’s pushing the baby out even. But in most cases we have the option of single dose IV narcotic pushes when someone is either maybe intentionally wanting to defer getting an epidural or maybe they are requesting one, but they’re having to wait a little bit for the anesthesia provider and they need something to hold them over. And it’s not a problem in that sense to temporarily expose the fetus to that drug while the fetus is in utero, because that fetus is going to keep getting oxygen through the umbilical cord. What we worry about is suppressing their respiratory drive if they still have a heavy circulating dose of the narcotic in their bloodstream right after they’re born.
Antonia: 2:58
So we try to predict that by doing a cervical exam and then just deciding they might be about an hour away from delivery if they’re maybe eight centimeters or something, and of course this can be very arbitrary and vary by patients quite a bit.
Antonia: 3:15
There’s no well-established standard of care for this at all. So a multiparous patient with history of fast labors now she’s an active labor she may go from five centimeters to complete in less than 30 minutes and then not even push once and have a baby. So I’ve definitely seen the rare nulliparous patient do the same thing and kind of surprise people thing and kind of surprise people. But by those arbitrary cutoffs that MULTIP or that rare primap even could have gotten a dose of fentanyl at five centimeters or maybe even something longer, acting like Nubane or morphine. That would definitely have been in the baby’s system when the baby’s born. And on the other hand maybe a more typical nulliparous patient who is seven or eight centimeters could easily still be several hours away from delivery. But by those kind of arbitrary cutoffs that we’ve decided on she would be denied getting any IV narcotics at all and be in pain for potentially seven or eight hours or more.
Howard: 4:24
Well, remember how long that half-life is. And if you think about the half-life of fentanyl, you might even be stacking these repeated hourly doses, which I realized that now that I’ve said that, I’m giving people something new to worry about. But the reality of it is they shouldn’t be worried about this at all, and we’ve known for a long time that it was safe to give particularly fentanyl even when the mother was completely dilated. I think this is one of those theory versus practical applications where the theory is trumped by empiric evidence. So there was actually a study published in the Gray Journal way back in 1989, where they gave fentanyl as often as every hour all the way up to 10 centimeters, and then the newborns were assessed independently and in a blinded fashion at birth and at a couple of hours later and then 24 hours after that, for any adverse effects, including respiratory depression of course, and they found no difference in the infants whose mothers had received fentanyl compared to patients who had not received fentanyl.
Antonia: 5:22
Right. So let’s say in that example, if a woman was five centimeters, you gave her a dose of fentanyl and then she delivered 30 minutes later, you wouldn’t have to call the NICU team and respiratory therapy stat. You’re probably going to have to innovate this baby, because there was one dose of fentanyl to the mom. That would not happen. Dose of fentanyl to the mom that would not happen.
Antonia: 5:49
It really would likely have to be a very extreme case where the mom gets probably a mistakenly large dose it would have to be so large that it probably sedates the mother as well and then have her immediately give birth massive dose and then have a baby five minutes later. And that might be like those twilight births from generations ago that we’ve talked about on here before, where they would actually make the mother semi-unconscious and then just cut the massive episiotomy and pull the baby out with the forceps. Those babies probably did have a little bit of temporary respiratory depression. But we have also talked previously on this podcast about cesarean under general anesthesia and that common but false belief that newborns that have been exposed to general anesthesia right before they’re born would have any kind of respiratory distress. And so you go so fast so that you can prevent that. That’s not true either, so it’s the same issue here.
Howard: 6:50
Yeah, theory, and then we eventually have the empiric evidence to overcome these theories. And, in fairness, the history of these drugs include things that we’ve used in laboring, include drugs like a paradine and very high doses sometimes of morphine, which were associated with respiratory depression, but our modern lower doses of fentanyl just aren’t. So if she’s 10 centimeters and you think she’s going to push for a little bit, you’ll certainly give it to her, and if you don’t think she’s going to push for a long, very long, you could do a pudendal block. You could do both. It’s just not something that we need to be that afraid of.
Antonia: 7:23
Yeah, a good pudendal block should give relief for at least 20 minutes, maybe even up to two hours, depending on what anesthetic you use, and so there’s always also the possibility of neonatal Narcan if necessary, but in practice I’ve never seen it need to be used. I’ve never seen it used.
Howard: 7:43
Me neither.
Antonia: 7:44
All right. Well, let’s move on to our listener question. That will probably take up most of this episode, but before we get into that, we have time for a couple of quick headlines to sneak in and then the listener question. So recently the New York Times and a few other news outlets covered a study that reported that patients who use levonorgestrel IUDs may see increased risks of breast cancer, and we’ve already been seeing patients be anxious about this because they see this in the news and it just feels like another unnecessary blow against birth control.
Howard: 8:21
Yeah, it seems to be very popular right now to make birth control the boogeyman, which is, I think, really sad because it’s been such a positive thing for so many women in the world. But this is the moment we live in and the media loves these sorts of stories. This particular story comes from a research letter that was published in JAMA on October 16th 2024, and it’s from a group in Denmark. This was a data collection from a registry group that included first-time users of any of the levonorgestrel IUDs, so that’s the 52 milligrams, the 19.5 milligram or the 13.5 milligram devices.
Antonia: 8:57
Yeah, they had a cohort of over 78,000 IUD users compared to the same number of matched patients who are not on hormonal birth control, and they followed the cohorts for almost seven years. And of all of them together, they found about 1,600 patients with breast cancer and the breakdown was 720 of those in the IUD group and 897 in the non-birth control or no hormones group. So they actually found more breast cancers in the patients who were not on birth control. So why would they attribute IUDs as a risk factor?
Howard: 9:40
Yeah, it’s really confusing when you look at it, but they followed the non-birth control users for 7.7 years and the IUD users they followed for 5.9 years. So the number of cases per year was calculated as being slightly higher in the IUD user group because it was, even though it was fewer cancers. It was for fewer years. So the total number of cancers was, as you said, higher in the half of the patients who didn’t have an IUD or any other hormones by a very significant amount. But since they divided it out by the total follow-up years, then they found that higher case per year in the IUD group and their outcome was incidence per person years of use. So it was slightly higher in the IED group.
Antonia: 10:24
Okay Is. Is that the correct way to do a study on breast cancer incidents, by having different follow-up durations between the two exposure groups and then make the outcome be dependent on the duration of follow-up? I don’t understand why. Would they not have had the length of follow-up time just be matched up?
Howard: 10:46
Yeah, I think you’re asking the right question for this. Those groups had also previously used other hormonal birth control for sometimes several years or more, but the really interesting thing is that they did not see more breast cancer with longer duration of use. That was a finding they commented on. Longer duration of IED use was not associated with a trend towards more breast cancer. So despite that fact, they still made the assumption that if the person years had been the same in the two groups, then there would have been even more breast cancers in the IED group, even though there wasn’t an upward association trending.
Howard: 11:23
But if anything, the study shows that dose exposure over time does not contribute to risk, because the groups themselves were exactly matched for years. In other words, the average age of group of the two groups were the same. They just had slightly longer follow-up, which again was an arbitrary assumption.
Antonia: 11:41
These findings are really good information, but it’s still just unfathomable. How did they draw the opposite conclusion of what their data showed? The data is interesting, but then the conclusion is just escapes. The logic escapes me. Okay, let’s be clear about what they. There were 720 breast cancers among IUD users, compared to 897 breast cancers among non-birth control users at the same point in life, by age.
Howard: 12:14
Yeah, the exact same number of women in both groups at the exact same age. Yeah, there were a lot more cancers in the non-IUD user groups.
Antonia: 12:24
Yes. Yet all over major publications, including the New York Times, this mistaken claim persists that IUDs cause breast cancer. It hasn’t even been established that any systemic progestins increase breast cancer risk and I don’t think we need to rehash this, because we already did that deep dive on those theories in a prior episode about hormones and breast cancer. We’re just extrapolating from one still false imaginary risk factor to another.
Howard: 12:58
Season one, episode three. But yeah, it’s absurd and the same data could have been presented in a different way, with different assumptions by the researchers to argue that IUDs decrease the risk of breast cancer. But yeah, it’s absurd and the same data could have been presented in a different way, with different assumptions by the researchers to argue that IUDs decrease the risk of breast cancer. As I read the paper, it’s also not made clear if the patients kept the devices the whole time or for some unspecified fraction of the follow-up time. The authors just assumed that patients would keep them for the full six years, but we know, of course, some of those IEDs aren’t even indicated for six years of use and it’s just silly to assume that all the patients kept them for the whole time, which is a major assumption, especially when they’re doing this per year thing.
Antonia: 13:37
Yeah, like the Skylaz, for example, they do admit to that specific data. They do admit to that specific data lack, I guess being a problem, and they lack any sign of a dose response and that really does argue against a causal relationship. They also acknowledge that there are many other significant confounders that they didn’t control for, but that is not what was highlighted in the New York Times, or probably TikTok.
Howard: 14:04
That is not what was highlighted in the New York Times, or probably TikTok yeah, and we’ve talked about several great New York Times pieces on here over our seasons where they’ve done a lot of great things about cord blood banking or about the false positive rates of some of the genetic tests. They’ve done a lot of really good stuff over the years and we think of them typically as having a higher caliber of journalistic quality and integrity. But unfortunately we often see the quality of medical research itself and the reporting on it, even in major journals like JAMA or New England, completely missing the mark, and then this becomes dangerous for patients and unfortunately even physicians, who rarely tend to read past the headlines or the abstract conclusion. If the theory is that even very low levels of the progestogen in the IUD is associated with breast cancer risk, then we should be able to see this effect very clearly in birth control industry trying to jump in and somehow counteract this.
Antonia: 15:06
But maybe that’s where the copper IUD really tries to advertise that we’re non hormonal, so we won’t cause breast cancer. But the hormonal pills like Lessina and other generics are some of the best-selling and most used birth control pills of all time. They contain the exact same hormone as the IUD and in dramatically higher doses. Yet studies have failed to find any relationship between breast cancer incidence and oral contraceptive pill use.
Howard: 15:38
And that gets back to the Bradford Hill criteria, which are nine specific criteria that we should consider when we do find a true association between two things, and then whether or not that association represents some causality. Now, to be clear, I don’t believe that they have found an association in this low-quality research letter, let alone causation. They did a poor job of adjusting for confounders, they didn’t design it in a way to make a direct comparison by having this problematic, misadjusted follow-up time, and they actually found more cancers in the control group.
Antonia: 16:11
Yeah, exactly, but let’s say just hypothetically. They did follow the patients for the exact same amount of time and did find more cancers in the IUD group, which they didn’t, but let’s pretend they did. Even then you still could not claim causation just on that basis. You would still have to apply the Bradford Hill criteria first to make that leap from the association to causation.
Howard: 16:36
Yeah, and we can go through those nine criteria quickly. This is a good exercise when you’re reading anything and you’re thinking about associations and causation. So the first one is strength of association. In this case, the effect size they claimed was very small, which often implies that you’re seeing just noise rather than true signal in your data. Next, you have to consider consistency, and this finding is not consistent with other information we know about levonorgestrel and breast cancer, like we just mentioned with birth control pills. The next is specificity. There really isn’t any specificity in this claim. It might be more specific if, for example, they claim that progesterone receptor positive breast cancers are more common in the group who use the IED compared to those who didn’t, or had other types of cancers, something like that. The next criteria is temporality, and the author specifically didn’t find a time relationship to the length of use, suggesting that this criteria is not satisfied. Next is biologic gradient, basically like a dose response relationship, which, again, they didn’t find a dose response relationship. The 51 milligram devices were no more likely than the 13.5. And the longer you had it, the more exposure was no more likely.
Howard: 17:47
The next criteria is plausibility. Now, right now, we don’t have a plausible theory about why levonorgestrel would cause malignant transformation of healthy breast cells into cancer. As we discussed previously, the only plausible link is that an active tumor that’s progesterone et cetera, positive in the breast that’s already there, could be stimulated to perhaps differentiate quicker, grow quicker, divide quicker invade because of the stimulation from progesterone. But this is why we don’t place Mirena’s, for example, in patients who we know have breast cancers, to avoid that possible link, and that’s not the question they’re addressing, though.
Howard: 18:23
Okay, the next criteria is coherence with basic laboratory research, and I’m unaware of any basic laboratory research that supports the hypothesis that you can promote malignant transformation by exposing breast cells to progesterone, for example, or animal models, things like that. And the next criteria relates to the experiment making the claim there is no experiment in this case, meaning this wasn’t a trial, there’s no blinded control trial or anything like that. The final criteria is analogy, and if you understand that we don’t actually believe, for example, that estrogen is causal of breast cancer no more than progesterone, then any potential analogies you might make with estrogen fall away, and I mentioned that because that’s the one that everybody’s going to make. They’re going to say this is like estrogen.
Antonia: 19:06
The really ironic part here is that there’s already hundreds of other low quality retrospective studies like this that don’t support these same conclusions, but then it seems like this one that does have this seemingly scary association, even though it’s false. But it has this kind of sensational claim, it gets more attention, even though the conclusion is, of course, very flawed. And there’s nothing wrong with retrospective studies. We have no reason to think that these people tampered with their data or falsified anything or nothing like that. But there should be stronger editorial oversight and clear statements in the journals that this data does not mean that IUDs are even associated with breast cancer, let alone causative.
Howard: 19:55
They should just give these disclaimers very boldly to try to avert these kind of opposite headlines of what the papers actually found these kind of opposite headlines of what the papers actually found, and the journalists and lay folk and even physicians are all just too commonly guilty of association equaling causality.
Howard: 20:12
And the truth is, association almost never equals causality, and so we see this problem repeated over and over again.
Howard: 20:19
It could still be possible that there is a true association, and then we should explore all of the possibilities for why that association might exist.
Howard: 20:28
But given the lack of controls between the two groups and the unknown confounders and the lurking variables because it’s not really a trial it could be, for example, that women with stronger family histories of breast cancer preferentially chose IEDs over birth control pills because they believe perhaps falsely that birth control pills would increase their risk of breast cancer because maybe their mom or somebody had an estrogen receptor positive breast cancer, and then that patient would be steered more towards an IED and therefore be a higher risk patient. Or maybe more women that smoke could have been steered away from combined hormonal contraceptives to get the IED because of the contraindication with smoking, and more women with delayed childbearing past age 35 could have had an IED or things like that that are independently associated with an increased risk of cancer. And any of those confounders, or many more that we’re not even thinking of could have skewed the data to make it look like breast cancers were more common in the IED group. So association does not equal causation.
Antonia: 21:26
All right. Well, let’s just get on to our listener question then. I don’t think we should keep. We can get lost in all these different articles. Okay, so our listener writes hello, dear OB. Can you address the new direct to consumer preterm labor test that was in the news this week? This is October 2024. I’m always skeptical of this type of marketing, but would love your advice on how to discuss this when patients ask about it or when they tell me their results and they give the website for this product. It’s spelled it’s pretrm.com p r e t. Signed skeptical in Saskatchewan.
Howard: 22:06
A lot of S’s there.
Antonia: 22:07
Yeah.
Howard: 22:08
Well, this listener is very right to be skeptical of this new test. There’s a lot that we could potentially talk about here, and patent holders is to monetize their business and to make money, and the job of physicians is to rely on scientific literature to decide when we should use some product or intervention or whatever in our clinical practice.
Antonia: 22:34
Yeah, this reminds me a little bit in the last episode of where we talked about direct marketing of the anti-malarian hormone test, and then there’s also many other similar fertility type tests out there that are direct to consumer. So the more that a product requires this type of marketing, the less it’s likely to be actually worth anything.
Howard: 22:56
Well, just think about all the drug commercials you see on TV all the time.
Antonia: 23:00
Yeah.
Howard: 23:00
Are those for doctors or are those for patients, or both Right. But a good drug typically sells itself because high quality evidence and guidelines support its use. And just do think personally that there is enormous potential that we’re going to see in the coming years to decades in the use of proteomic biomarkers. But it’s still really the Wild West out there. We’ve gone down roads before over the last 20 years or so with proteomic screening tests for cancer. There was a maybe we could talk about it sometime a test for ovarian cancer that had all sorts of controversy around it. Maybe it’ll be back and it’ll work. I think we’re going to see this in the future, but so far these things have ended up nowhere and so far I actually really do believe in this stuff. And they have a trial that they referenced on their website that was initially published in 2023, at least online and they linked to a preprint server version of that on their website, and then this ultimately was published more recently in the journal Diagnostics.
Antonia: 24:14
It’s not a journal I had heard of before, but yes, their trial was called AVERT preterm trial preterm trial and they offered this test to a group of women at around 20 weeks gestation, and of about 1,400 and 1,460 patients who took the test, 34% of them screened positive, and they defined that here as having a test result that gives at least 16% risk of spontaneous preterm birth. And so for those 34% of patients, they were offered a group of interventions that included vaginal progesterone and baby aspirin and twice weekly phone calls just to check in on them, I guess, or review their symptoms. So of the patients that screen positive and were offered these interventions, 56% of those accepted, and then the rest declined. So then, of the 56% of the patients who accepted the interventions, so now we’re at 19% of the test population, it’s about 270-ish patients.
Antonia: 25:23
Now. They followed some composite neonatal morbidity and mortality outcome numbers and then compared those to a historic cohort that had no involvement with this test, and they claimed in the paper to have found a shorter neonatal length of stay in the intervention group, and they used a lot of kind of different statistical analyses to get to that point. But overall, the gestational age at birth was the same in the prospective versus the historic arms, and the number of infants actually in the NICU was very small, with what appears to be some outliers making the historic group then have a longer length of stay. But this is the touted benefit, supposedly, that you’ll reduce the length of the neonatal hospitalization if you do this test and then do the interventions.
Howard: 26:17
Yeah, this is a really great paper for a journal club interventions.
Howard: 26:21
Yeah, this is a really great paper for a journal club.
Howard: 26:22
And you actually have to download the supplement to the paper to see the outcomes in a data table, because the text of the paper does, honestly, a lot of just writing that obscures the actual findings of the paper.
Howard: 26:31
It’s not enough to say that this biomarker identifies patients who are at increased risk of preterm birth. It probably does, but in order for it to become something that we adopt as a commonly used test in OBGYN or a universal screen, even for low-risk patients, then we need to have some kind of beneficial intervention for the patients who we’ve identified as being at increased risk. So that’s the point of this trial is to claim that if you screen a population, you’re going to somehow improve neonatal outcomes through some series of interventions. So they use an aggregate outcome, hoping that something is statistically significantly different enough, and it’s essentially this length of stay issue. But even in that they’ve not reported the total data. They’ve chosen to use some data from the 8.5% quantile. It’s all weird when you read it, but basically they’ve sliced this data with some researcher degrees of freedom, some choices that they made in order to find something that looks to be statistically significant and beneficial.
Antonia: 27:33
Well, how about the interventions? What do we actually know about baby aspirin for preventing preterm labor or, for that matter, the 200 milligram vaginal progesterone treatment? Because they’re assuming those medicines might reduce preterm birth, even though the average gestational age at birth was not changed by them, so that, I think, makes their claim about better NICU outcomes or shorter lengths of stay a bit shaky. It’s like are they suggesting some kind of novel mechanism about aspirin or progesterone improving fetal lung maturity, maybe without actually prolonging the pregnancy? But of course they don’t explore that and they can’t because there’s nothing there. There’s no basis for thinking that.
Howard: 28:22
Yeah, and it’s also different data from different years. Remember they compared this to a historic cohort, which is unnecessary. They could have done a contemporary cohort and this historic cohort of data existed before they designed the study right. So maybe there were new neonatologists with different protocols for how long they kept certain newborns in those two different time periods. Maybe there was more opioid addicted mothers in one group. Maybe there were just a couple of random bad outcomes that shifted the data due to really long outlying lengths of stay. But this was not an interventional trial and also pragmatically, why did they pick this old cohort? They could have easily found, even in the same data set, the same patients. They could have made a comparison and had real-time prospective data with similar NICUs and similar employees and similar doctors taking care of them. So it all seems very sneaky and the results were not published in an important journal and I think that this is the reason why.
Antonia: 29:25
And this isn’t even the first study that the company did. Right To look at this test.
Howard: 29:29
Yeah, and I’ll say too. Just the skeptical part of me wonders how many studies they have done, because that’s one of the problems sometimes, too, with industry sponsored studies is we don’t see all the data published. But yes, there was a very similar paper published in 2021 where the screened positive patients were offered a group of interventions to decrease their risk of spontaneous preterm birth and they looked at the rate of preterm birth, along with the gestational ages at delivery and the NICU length of stay, and in this case, they compared the screened patients to a cohort of unscreened patients. And they talk about this being an adaptive study design, where the number of enrolled patients would vary. That’s an interesting problem too, because some studies like this are designed to stop the study early once you see data that looks significant, but then it may still be underpowered.
Howard: 30:18
Many of those studies go on to wash out significance if you ultimately continue your enrollment. They did stop that study early and they claimed that it was because they ran out of funding, even though this study was sponsored and paid for by the company that designed the study and implemented it. But in any event, there was not a statistically significant difference in this study in the rate of spontaneous preterm birth, and there was no difference in the gestational age at delivery, nor was there a difference in the total length of neonatal stay or the NICU length of stay, but in a small subset again of just literally a few infants, there was a difference in the NICU length of stay for infants admitted for spontaneous preterm birth, and that was the conclusion they took away from it.
Antonia: 31:02
There are some interesting things about this study to point out. So they intentionally chose a low-risk group to do this test on, so they excluded people with other known risk factors for preterm birth and they gave a list things like twin gestation or prior preterm deliveries and they were pretty good at excluding, because the overall rates of preterm birth then were 3.5% in the unscreened group and 2.7% in the screened groups. Those differences were not statistically significant. So in this study, 16 women in the screened group ended up having spontaneous preterm birth, and I couldn’t find the breakdown in this paper of how many of those 16 that were screened screened positive or screened negative. But likely, it was fairly negligible because the median gestational ages at delivery were still the same.
Antonia: 31:56
We don’t have a directly applicable study of intervention versus non-intervention for this population of women that are low risk but screened positive on the preterm test, but we do have indirect historical studies that show, for example, that the intramuscular progesterone does not work, and that is in higher risk populations. The other interventions for the screen positive women besides progesterone included the baby aspirin, additional cervical length monitoring and education about preterm labor symptoms. My question is then where is the data that any of those interventions will prevent spontaneous preterm birth, even in high-risk patient populations, let alone these low-risk ones. These two studies that we’ve just talked about have shown that, really, the preterm screening test does not change the gestational age at which the patients deliver, nor do any of the interventions.
Howard: 32:58
Yeah, the gestational ages are the same and so you have to tease out of that, if anything, why the NICU lengths of stay were different. And when you talk about that last study, only having 16 total patients in it, you can see where a couple of outlying patients would really skew the lengths of stay. Now, the author of the newer study, the one we talked about first was also a co-author of a randomized controlled trial published in the Lancet in 2020 that was called ASPRIN. Again, I wish I had the job naming studies that randomized about 12,000 patients in low resource countries to a baby aspirin versus placebo, and in that study they found a slightly lower rate of preterm birth.
Howard: 33:36
In the group that received baby aspirin, there was an absolute difference of about 1.5% in the rate of preterm birth in the two arms. Baby aspirin there was an absolute difference of about 1.5% in the rate of preterm birth in the two arms. Some of this may be related to hypertensive disorders, because in the placebo group, more pregnancies were delivered early due to hypertensive disorders and, interestingly, the rate of hypertensive disorders was the same in the two groups, but that barely reached the level of statistical significance in terms of earlier deliveries. One immediate concern about the study is that it was over-enrolled. Given the large number of patients that were present in the study, there were almost no outcomes that were statistically significantly different in the two groups and for some reason they never differentiated between spontaneous versus iatrogenic preterm birth. This would be really important if the increased risk of preterm birth was due to hypertension, which aspirin was mitigating. They said they did this because the majority of their births occurred in facilities which would not have the ability to differentiate between iatrogenic and spontaneous preterm birth.
Antonia: 34:37
That’s a little bizarre to say that someone can’t tell the difference between someone who comes in and labor or someone who is induced. Of course you can tell the difference. I think maybe they’re saying they’re not able to reliably document the difference.
Howard: 34:52
Yeah, perhaps it’s very questionable and the study hasn’t had a large impact, I think for that reason and a few other reasons, I wonder how this stuff gets published. But there are a lot of people who think baby aspirin is a panacea for everything, and perhaps this is through some anti-inflammatory mechanism that affects the preterm labor cascade, which has a lot of inflammatory action in it no-transcript to be able to see the benefit sooner, rather than just giving the aspirin to everyone and potentially diluting evidence of benefit in that way.
Antonia: 36:20
But this study was published in February 2022 from the Netherlands and they found that low-dose aspirin did not significantly reduce the spontaneous preterm birth rate.
Howard: 36:30
Yeah, and that study is honestly more applicable because it was in a country with a highly developed healthcare system, among higher-risk patients, which, if you buy the premise of the preterm test, you’re identifying higher risk patients with that test in a first world country with a highly developed healthcare system, and they just didn’t find that aspirin was beneficial for this at all.
Antonia: 36:51
Okay, and we already know that the intramuscular progesterone doesn’t work. We know that baby aspirin does not have any compelling data for this indication. So what about the 200 milligram vaginal progesterone?
Howard: 37:04
Well, a lot of researchers seized upon the idea of vaginal progesterone as a potential treatment for the prevention of recurrent preterm birth After it became clear that 17-hydroxyprogesterone wasn’t the panacea that everyone thought it was, or I should say, after it became clear that it was going to be withdrawn from the market and people wanted to still give progesterone because they believed it worked. And so then you start to see trials looking at vaginal progesterone for the prevention of recurrent preterm birth. So there’s been some data about this in patients with shortened cervixes that are discovered around the time of the mid-trimester, and data that shows it’s effective and probably safer than cerclage. But that’s not the population we’re talking about here. We’re talking about patients with a normal cervical length and it’s important to make that distinction.
Howard: 37:50
There is a sort of progesterone cult in OB that believes sincerely that progesterone is the cure-all for everything, from infertility to miscarriage, recurrent miscarriage to preterm birth, to all these things. But as we’ve moved away from 7-handroxyprogesterone, we’ve seen more small, low-quality studies in most cases advocate for the use of vaginal progesterone to take its place. It’s as if we just have to fill that hole for progesterone that was created when we took away the weekly IM injection of McKenna, but I’ll put a link to a 2022 systematic review from the Gray Journal about vaginal progesterone for prevention of recurrent preterm birth in patients with prior spontaneous preterm birth. So, again, a high-risk population would be similar to the risk of the patients identified with the preterm test, and the conclusion of that paper is that there’s no convincing evidence to support the use of vaginal progesterone to prevent recurrent preterm birth or to improve any perinatal outcome in pregnancies with a history of spontaneous preterm birth.
Antonia: 38:50
Okay, well, surely someone would say it. Just to counteract this. Maybe those things by themselves won’t work. But what if you gave him a bundle? Things by themselves won’t work. But what if you gave him a bundle baby aspirin plus the progesterone, plus calling them twice a week?
Howard: 39:06
Yeah, it’s the phone calls. That’s what we’re missing, apparently.
Antonia: 39:08
Yeah.
Howard: 39:10
Well, yeah, so of course that’s what folks would claim, and but the burden of proof is on the person making the claim to show that these things as a bundle or individually or in some combination, are effective. And the two studies that we’ve just been discussing show that patients did not deliver at a later gestational age. They weren’t more likely to make it to full term and have a different rate of spontaneous preterm birth, despite the author’s absolute best efforts at finding a significant p-value for something in there that would make the company that paid for the studies happy.
Antonia: 39:42
Okay, well, let’s talk about the effect of bias in these studies. You mentioned that it looks like they manipulated the study design and the researcher degrees of freedom to find some compelling evidence or some p-value that they could hang their hat on. They really want to be able to say that using this biomarker test, along with this bundle of interventions, will shorten NICU stays for babies born preterm, even if the same number of babies are born to the same degree of preterm, which doesn’t make sense. But why are they so in denial that this doesn’t work? Why are they trying to force this conclusion and force it into mainstream practice anyway, aren’t they? Aren’t there some scientists involved in here that are trying to disprove their theories?
Howard: 40:38
You make me laugh a lot. You’re very, very funny, but your questions, I know, are rhetorical because you already know the answer to the questions you’ve asked. The first study published in 2021 is in a paid to publish open access journal with a very poor reputation and after, I’m sure, it was likely declined by several other more prestigious and more respected journals, and it was written by authors who were employed and or held stock in CIRA Prognostics, which is the company that makes this test. Cira Prognostics funded the study and it was carried out by their employees and stockholders.
Antonia: 41:13
Yeah, and that was the one we said. Had just a little subset data group about the length of stay in NICU. That probably was skewed by one or two kind of outlier patients.
Howard: 41:25
So that study didn’t quite do what the company needed, and then, of course, it included 7-handroxyprogesterone, which is being withdrawn from the market while they’re around the time they’re working on the study results. So they needed something else and they did another study at least that we know of one more and that’s this newer study, the AVERT preterm trial. That’s on the website that we first discussed. It appeared online in a manuscript form in 2023, and then it was finally published in July of 2024. It was again funded by Sierra Prognostics and the authors were paid consultants of the company. They also declared that the study plan was agreed upon by the funder and that the manuscript itself was written by an employee of the funder.
Antonia: 42:07
Yeah, and that’s revealed in the conflicts of interest section, and the study design that was approved is key, because this is not how you would normally design this study If you were actually seeking a truthful answer of does this work or not?
Howard: 42:23
Yeah, and I always encourage people like in journal club, think how would just take the question and how is the ideal way of designing that study? And then of course, there’s going to be pragmatic limitations to that. But there was no reason for them to use that historic cohort just to provoke conversation. Lots of other things too. But these sorts of conflicts of interest are unfortunately rampant in medicine and again, I think this is why these papers were published in low quality, open access, pay to publish journals.
Antonia: 42:49
So I guess we’re saying that right now this preterm biomarker is not ready for prime time.
Howard: 42:53
Well, it might actually identify patients. Again, I’m soft on proteomics it might actually identify patients who are at increased risk of preterm birth. I think we’re going to see a bunch of proteomic testing in the very near future and we’re going to transform a lot of areas of medicine. But the company’s interested in selling the product now and having a product that not just that is a risk marker identifier, but something that they can commercialize. And in order to do that, they have to show that by identifying these otherwise unidentifiable high-risk patients, we can implement some intervention and reduce the risk of preterm birth for newborns.
Antonia: 43:28
Remember, if we had meaningful interventions, we would already be using them on all the other patients that we know are increased risk of preterm birth just based on their history. So that might be the patients that they excluded from this study. Risk of preterm birth just based on their history. So that might be the patients that they excluded from this study, like prior preterm birth or twins, or uterine anomalies or maybe short interval pregnancy or smoking or all kinds of other factors poor dentition, recurrent BV but they would be put at that same risk of 16% risk of preterm birth or 14%, at that same risk of 16% risk of preterm birth or 14%. But we’re not using these interventions on them because the evidence doesn’t exist for that and we have no way right now of meaningfully reducing their risk of preterm birth.
Howard: 44:11
I think that’s exactly the problem, and if we did, this might be a great thing, right? It’s not that the biomarker itself may not successfully identify patients with elevated preterm birth risk, it’s just that we biomarker itself may not successfully identify patients with elevated preterm birth risk. It’s just that we don’t have any interventions and you’re going to see more and more of these things come out. In fact, this company is developing a series of proteomic, a whole suite of proteomic tests for obstetrics to look at diabetes and hypertensive risks and a whole bunch of things. But the question is, if we develop those things, do we have a meaningful intervention for those patients to make a difference in their outcomes? Knowing someone has risk and having a meaningful intervention to help them are two different things.
Antonia: 44:50
Well, but does this mean that we can just never, ever trust a sponsored trial? Because almost every new drug or test or product that enters the marketplace will have some company behind it with some economic interest. So how do we mitigate the competing interests?
Howard: 45:06
here. Yeah, it’s very difficult. It’s not just economic. This is why we have the world’s most successful OBGYN podcast is because there’s so many perverse conflicts of interest out there. Link to a paper from the University of Chicago that reports that drugs are typically reported as being 49% more effective when the trial is sponsored by the drug’s manufacturer or marketing firm, compared to studies where the same drug is evaluated to the same comparators but without the drug manufacturer or marketers involved. Sponsored drugs are also 43% more likely to report statistically significant improvements and 73% more likely to be the most effective drug in a trial compared to the same molecule tested against the same comparator but without funding from the drug manufacturer.
Howard: 45:50
And that’s a quote in case someone thought I was plagiarizing it. I just read from that. The author calls this the sponsorship effect and if you’re interested in this topic you should definitely read the paper and I’ll put a link to it. But this is a massive problem. In research we can read papers and we can estimate the bias in them and even if you take a fraction of the biased outcomes that I just cited from the paper, when you’re reading it you can see that very minimal effects, like in these papers about length of stay or things like that that the authors were claiming they fall away. They lose their significance when you realize that 50 to 70% of the effect size is through study design and manipulation of researcher degrees or freedoms by the companies designing these studies. There’s almost no chance that the two papers we’ve been discussing found anything significant at all that would reduce neonatal morbidity and mortality. In fact, you could even obscure harm from interventions when you have this sort of manipulation.
Antonia: 46:43
I know you’ve answered this question in your book, but how is it possible for a study to find an effect that’s not real without falsifying data, while another study does not find that effect?
Howard: 46:55
Yeah, and it’s actually quite easy. It’s surprisingly easy, and we just need to be aware of that. And again, I talk about manipulation of researcher degrees of freedom. So these are things like how many patients you choose to enroll, what aggregate outcomes you select when you add up these outcomes, when you choose to stop enrolling, how many subjects you plan to enroll in the study and what comorbidities or confounders you plan to adjust for and what methodology you use for adjusting it. Or whether or not you use a historic data set and which historic data set you choose to use. The comparators are really important and when you have control over those comparators, you can do all sorts of things, like that historic data set.
Howard: 47:31
We got off to the wrong foot with 7-hydroxyprogesterone years ago because the comparator arm in that study that received placebo had an unusually high rate, an unexpectedly high rate of preterm birth in it, and then the progesterone looked good by comparison because it had the normal rate of preterm birth in it. Now, I don’t think that was deliberate. I think that was purely by chance alone. But that’s what happens sometimes, even by chance, and that’s why we need to do replicated studies, especially when you see something like that.
Howard: 47:58
That’s a little off. But when you’re setting out already with a decision to use a historic cohort, with a known rate of preterm birth and other outcomes that maybe, nick, you link the stay and the company’s decided to use that cohort and has control over the study design, it’s hard to overcome. Imagine the boardroom meeting where they’re set around saying okay, how do we, what can we do to make this look good? After that first study was published, what was the meeting that led to the second study and the design changes, like the choice to use a historic cohort? They didn’t do that in the first study.
Antonia: 48:28
I think people just get so drawn into their different sort of agendas and biases they forget what the scientific method is, and that is to try to vigorously disprove what you believe to be true.
Antonia: 48:43
And then maybe, if you do a real trial at trying to disprove it and you fail, then that might support what you believe, maybe.
Antonia: 48:51
But you have to be completely neutral and completely unattached to a specific outcome and just be committed to finding the truth. And that kind of attitude produces very different types of studies and potentially with very different outcomes than setting off to start a study with an attitude that you’re trying to prove something because you’re a stockholder in the company or a patent holder of the product and your whole wealth depends on it. Or maybe personal pride, or maybe it’s your life’s work and legacy, and maybe it just makes so much sense in your head that you’re just you’re yearning for it to be true and for everyone else to see it as well. We’ve seen that in things like magnesium or progesterone or aspirin, where it doesn’t really make anybody rich to see these succeed in a study, but it maybe makes somebody’s reputation shine. So those are just some of the other ways that bias seeps into studies that really almost rigs the outcome from the beginning, rather than just starting off and maintaining a neutrality throughout.
Howard: 50:01
Yeah, we see it all the time and we talked about the same ideas with the fetal fibronectin test and the influence of the patent holder on that test. Or there’s another new test that we recently discussed, the SFL-T1 marker for preeclampsia. Same sorts of issues.
Antonia: 50:16
Yeah, we talked about that one back in season six, episode four, because that had just become FDA approved. But it was really only useful for predicting over the next week after the test is done. So again, that didn’t really help give any practical advantage to using that test.
Howard: 50:32
Right, and it’s still being heavily marketed, though, and people are looking like inventing for reasons to use it. It does apparently have a good negative predictive value. So there was a study that looked at aspirin discontinuation at 28 weeks in patients with a negative marker and found no difference in outcomes related to development of preeclampsia at term. But then again, there’s problems with the aspirin studies, although that study used 150 milligrams of aspirin. But anyway, the major publication about that marker that’s typically hype comes from the New England Journal of Medicine evidence. There are sub-journal evidence in 2022. And the point I wanted to make about that was that study, too, was funded by the company that sells it and that paper was authored by the two patent holders. So you know this is in the New England Journal of Medicine, too, and you just have to. You have to know that.
Antonia: 51:20
Yeah, and that’s just a huge conflict of interest, especially for someone that trusts the New England Journal. They think this is a good journal. But even in that paper, just like many others, you have to go to the website and download the supplementary materials to then find that conflict of interest form disclosed. And for some reason in this journal they’ve decided to tuck that way at the bottom rather than having it on the first page, like I’ve seen in some other types of papers. So it’s like the authors and maybe even the publishers are playing the game. They’re staying within the lines, but they’re being very cunning or maybe very underhanded, depending on how you look at it. And I know you mentioned one time that you will look up authors on the Open Payments website to see if they’ve received payments from companies or products that they’ve been studying.
Howard: 52:10
Yeah, and I’ll do that, in fact, after the publication, because a lot of times the money doesn’t come until a few months after the paper’s published. That way they don’t have to disclose to the journal that there was any upfront conflict, although, again, that only works for US-based authors.
Antonia: 52:25
Yeah, that’s practicing situational awareness. That’s something we were always taught in training, in the operating room, for example, but this is doing it in the scientific literature world. We really have to recognize that games are being played and rules are being dodged, and we have to see when that’s happening to have a full picture of what’s being presented.
Howard: 52:45
And I think that’s the point. These studies are going to be published. They should be published. It’s fine. It’s a cat and mouse game and we just need to be actively playing it and we need to know this stuff. There was a paper published in 2018 in the British Medical Journal that examined collaboration between academics and industry in clinical trials. Through a survey of lead academic authors, they found that employees of industry funders co-authored 87% of publications and that 92% of trials reported involvement of funders in the design of the trial. Data analysis involved the funders in 73% of cases and only a third of the respondents said that the academic authors had a final say about the design of their study. Several reported that there were unnamed funders or third parties involved in the design data analysis of the studies and participated in even writing it. Most of the authors said that they thought the collaboration was beneficial, but 11% did report having significant disagreement with the funders over study design or data interpretation, or especially the way the conclusions were reported, like they felt a little bad about it.
Antonia: 53:51
Well, there you go. We have to assume that industry-funded papers are presenting information in the very best possible light, or drug medical technology companies stand to make billions of dollars by finding a p-value under 0.05. Yet industry research does also fund a lot of new, very beneficial innovations. So, as busy doctors, we have to figure out how to protect our patients and ourselves from just being sold a bill of goods. Yeah.
Howard: 54:23
Well, I think for me personally, some things. I don’t meet with drug or product reps in my office. I try to insulate myself from making decisions based on sales pitches. But as far as the literature is concerned, I think what we’ve said before we have to remember that what we’ve been preaching on the podcast is you need two. You need multiple randomized controlled trials comparing the intervention to placebo, and then you need a trial comparing the intervention to the best current comparators and then you need a cost analysis and then you can start to consider using it in your practice.
Howard: 54:53
In the case of this new biomarker that we’ve been discussing, if you started using that in your practice right now, based upon company marketing, you would be doing so without any of those trials I just mentioned. You would do so based only on preliminary and exploratory low quality data that came straight from the company and the patent holders, and you wouldn’t even know about it if you wouldn’t read it in a major journal. You only know about it because the company rep showed you those articles. Now how does the end work? Do I have to do the end now?
Antonia: 55:22
Oh well, we can try that out. Just remember to say something about thanks for the question. Oh yeah, Okay, Okay.
Howard: 55:31
Thanks for listening and send us all your questions. We do have some more questions queued up and we’ll be back in a couple of weeks with some new episode, and I’ll try to get the website updated more commonly so you can read these articles for yourself. And that’s thinkingaboutobgyn.com. You’re better at this than I am. It’s my first time.
Antonia: 55:49
Yeah, it just takes a little practice, you’ll get there, yeah.
Howard: 55:52
All right, we’ll see you in two weeks.
Announcer: 55:58
Thanks for listening. Find us online at thinkingaboutobgyn.com. Be sure to subscribe. Look for new episodes every two weeks.
