Welcome to Fertility and Sterility On Air, the podcast where you can stay current on the latest global research in the field of reproductive medicine. This podcast brings you an overview of this month's journal, in-depth discussions with authors and other special features. F&S On Air is brought to you by the Fertility and Sterility family of journals, in conjunction with the American Society for Reproductive Medicine, and is hosted by Dr. Kurt Barnhart, Editor-in-Chief, Dr. Eve Feinberg, Editorial Editor, Dr. Micah Hill, Media Editor, Dr. Pietro Bortoletto, Interactive Associate-in-Chief, and Associate Editor, Dr. Kate Devine.
Welcome back to another episode of Fertility and Sterility On Air. I'm Micah Hill, your Media Editor. We are in August 2026, Volume 126, Number 2. Delighted to be joined this morning by my co-host, Kate Devine.
Good morning, Kate. Good morning, Micah. Good to see you, and also our other co-host, our Editor-in-Chief, our fearless leader, Kurt Barnhart.
Good morning, Kurt. Great to see you, sir. Good morning, both of you.
Great to be here, too. So, I'm actually really excited to talk about these papers. Some of them challenged my thinking methodologically, statistically.
I'm really curious to learn from both of you on what you thought about it. Let's start with the front matter really quick. I'll summarize that, and then we're going to dive right in.
So, Dom DeZiegler has an article, or views and reviews, on what to do with ART failures, something that's challenging for all of us. Even with our best prognosis patients, each embryo transfer often results in a failure, and how do we approach that? How do we look at that? And he got together a group of worldwide experts to sort of cover the depth and breadth of that topic. I highly recommend that you look at that.
Our other editorial editor, Eric Widra, along with Allison Eubanks, our first-ever F&S Scholar or Editorial Fellow, looked at fresh versus frozen donor O sites. I know even in our own practices, we've sort of changed over the years, whether we use fresh donor eggs, frozen donor eggs. There's pros and cons on both sides.
I think that's an interesting debate. And then tying back in with what Dom was talking about with ART failures, our final editorial editor this month, Liz Ginsburg, talked about the importance of cycle review meetings, and I really love that. Obviously, I think in our fellowships, we probably all do that to teach our fellows the importance of reviewing IVF cycles, especially those that fail.
We actually have our fellows in our program call those patients. Those are difficult calls. We've had some interesting articles in F&S over the last few months this year that have sort of said that maybe changing protocols isn't always the right thing.
If everything looked good, sometimes transferring the next embryo is the right thing to do. But sometimes there are things we can learn from that, and making those tough calls and talking to patients, I think, is a good thing. So, I really enjoyed that inkling from Liz Ginsburg.
So, as always, look at that front matter. Our editorial editors do a great job of putting that content together. There are also three ASRM practice committee documents.
One is on recurrent implantation failure. Really tough one. I know this was a tough one for the committee to put together.
I suggest that you read that, as this is a common thing for us to discuss with patients. There are two other articles that actually end my time as a practice committee person. I'm almost two years off the committee, but these are my final two documents as a good shepherd that are coming out.
One on transgender care and one on LGBTQ plus family building. Thank you to Molly Moravec for helping lead those documents as well. So, as always, important stuff from ASRM to read.
So, that's the front matter. Really good stuff this month. I highly encourage you to read that.
Kurt, we're going to jump right into you. You have the seminal contribution looking at obesity and is that associated with pregnancy loss. So, something right in your wheelhouse is a pregnancy loss topic.
Early pregnancy expert. I'm really curious about this article and what you thought about it. Thank you, Micah.
Great introduction. That front matter really is important and I'm going to diverge a little bit to give a mild commercial message. I wanted everyone to know, as this new journal comes out, that you are listening to, without question, the premier journal in reproductive medicine.
The impact factor for F&S was just released and I hope you all will recognize, because of your contributions and all of the expertise of our editorial staff, that the impact factor is now 8.1. The first time in the history that the F&S has been above 8. And let's just say it's a privilege to have it increase continuously over the last couple years, which is not the same for many of our journals in the same reproductive space, which have actually declined. So, you're listening to the right podcast and you're reading the right journal and I want to thank all of the authors, the reviewers, and the editorial team for this magnificent work. And it's because of papers like this one that I think will be widely quoted and widely referenced and really in an indirect way, because it's not telling us exactly what to do, will teach you something and might have an impact on your care.
So, the title of this paper, which is the seminal contribution this month, is Obesity is Associated with an Increased Risk of Early Pregnancy Loss After Euploid Frozen Embryo Transfer. It's out of the RMA and the IVRMA group. Christine Whitehead is the primary author and Emri Salih the senior author.
And again, it's a nice use of data. It's written very objectively and it has a nugget of truth to it and I'll go through it. It doesn't try to editorialize too much.
It doesn't try to overdo what it is. It's just really, it's just well done paper. So, the goal of this was to determine whether obesity is associated with increased risk of pregnancy loss.
That's not a new topic. We have all struggled with how obesity affects reproduction in lots of ways. There's a ton of literature on this.
The simplest way to say it, in my opinion, is it's complicated and it's really affected by potentially many variables. So, the contribution to this paper is, while I'm not advocating every patient has to have an euploid embryo transfer, this one narrows it down to it only looked at patients with a euploid embryo transfer, therefore taking a lot of the other potential confounding variables out. Now you're not worried about whether the embryo was euploid or aneuploid and it really has, although it's in multiple practices across the country, it really does have relatively standardized practices and data collection and therefore allowing a pretty good insight into this question.
So, basically, this is a large-scale study. It's almost 15,000 patients across nine fertility centers. It spans five years from 2019 to 2024.
The sites, in case you're keeping score at home, were New Jersey, Philadelphia, Florida, Northern and Southern California, Colorado, Seattle, Houston, and San Diego. The reason I mention that is not to give kudos to those sites, although they're good sites, but mainly to say that it really is across the country and theoretically representative. So, this is, again, as mentioned, a study of all patients who underwent their first autologous single frozen embryo transfer and out of the almost 15,000, we have about 4,000, 3,811 that are classified as obese.
By the way, the primary analysis was just stratifying greater or less than a BMI of 30, although they get into more detail, and approximately 11,000 and change did not. The primary outcome was, again, biochemical loss or clinical loss. Now, as any good study should, it should have a list of covariates and this one includes adjusting for, at least assessing for, the confounding ability of the oocyte age itself, again, because this is a frozen transfer, parity, the IVF center, the protocol used, although the vast majority used one protocol, and also looked at the blastocyst stage, in other words, day five, six, or seven, and also looked at the embryo grade using the SART technology, you know, good, fair, or poor.
Now, it didn't control for underlying diagnosis, and I'll get to that in just a second, which I think is not a flaw, I think is actually a good thing. So, the main findings were that, again, overall, the non-obese group was younger, but the parity in AMH and the blastocyst at the day of transfer was pretty similar. There was a higher portion of patients with obesity diagnosed with PCOS and a slightly lower portion of people that underwent a natural frozen embryo transfer.
So, that's the baseline. So, now, drum roll, please, you know, what did they find? Well, first of all, the raw numbers, I think, are pragmatic. They're going to sound high, and I read them three times, but the overall biochemical loss in the obese group was 44%, and the clinical loss was about 56%.
In the non-obese, it was similar, but a little bit flipped. The non-obese had a higher biochemical loss at 53%, and a slightly lower clinical loss at 48%. Why do I say that? This doesn't sound like a commercial message for PGTA.
These are real-world loss records. They're not claiming that there's, you know, a 5% loss in all other patients. There's a real loss here in transferring PGTA embryos.
But having said that, in this large population, now you can make a direct comparison. They chose appropriately relative risks rather than odds ratios, because it's not exactly an uncommon outcome, and the adjusted relative risk was 1.19, which was statistically significant, meaning you had obese women had about a 20% increased loss compared to non-obese women. Again, controlling for all these factors.
And I'm going to get to what intrigued me right here and now, which is, instead of controlling for diagnosis, I've looked back at diagnosis. Is there a difference in miscarriage for endometriosis or DOR or something like that? And it's not really clear, but there has been pretty good evidence about PCOS affecting miscarriage. So rather than controlling for PCOS, they did what I recommend many people should do, is they actually stratified on PCOS.
In other words, they took all the PCOS patients out and got essentially the same relative risk. Why do I say that? Because remember, controlling is trying to find statistically some similarity between the groups. In other words, you try to find the ones that are similar enough that have either both, and then look at those, but you're kind of limiting your population.
And you really can't control for PCOS. You either have it or you don't. So I really like in the papers going forward for things like PGTA or PCOS, it should be with it, without those variables, not trying to control for it because you can't find a middle ground of having PGTA or not, or having PCOS or not.
So I think that strengthened the paper. So that being said, let's go through quickly just for educational purposes. It's nice and convincing that they looked at the different classes of obesity as well, and they found really what can only be described as a pretty good dose-response curve.
The risk in class 1 diabetes of increased miscarriage was 13%, in class 2 it was 20%, and in class 3 it was 42%. And because they had enough power, all of those were statistically significant. They also looked at underweight women, and they found what many people believe to be the correct U-shaped curve.
In other words, that underweight women also had increased miscarriage. And it was really just the normal weight women, which ends up to be the comparison group here, that was the quote-unquote lowest or best outcome. So in other words, you had an increased risk of miscarriage if you were really, really underweight, and then a progressively worse outcome as you were getting farther and farther in the obesity group.
So that all sounds good, but it's also interesting to look at something else, and this gets a little bit more speculative. And then they went into, well, in a sense, why? Can we figure this out? And one thing they found that interested me was the implantation rate was actually about the same in all groups. So it was, again, it was 0.98, again, relatively tight comparison group spanning one, suggesting that the primary impact of obesity, whatever they found here, was affecting the pregnancy itself rather than the initial implantation or the quality of the embryo or the quality of the gametes.
And then the rest kind of makes sense. You know, the clinical pregnancy rate, if you look at the different surrogate outcomes and you look at biochemical rate, clinical pregnancy rate, and ongoing pregnancy rate in live birth, you would expect to see, you know, a greater impact in terms of the statistics, you know, going from 0.96 to 0.94 and then ultimately 0.92. But again, that's definitional, that's not biology. So I like the paper.
I think it adds to this debate about obesity. It doesn't tell us what to do about these patients. It doesn't tell us if losing weight is a good idea or a bad idea.
It doesn't tell us how to change your care with patients. It doesn't say you should change your stimulation. But it does pretty convincingly say that in a mild way, obesity does seem to have an impact.
So its main strength was restricting to the euclid embryos, and removing other aspects of potential miscarriage. But, you know, again, it is a retrospective study about the standard of care, about quality of care, rather than an interventional study. It's also some intriguing aspects in the study and their discussion, as well as the reflection by Barbara Lawrence and Ulan Fatami.
But again, this is speculation, which you should enjoy doing, just be a little bit careful. You know, it doesn't necessarily say that it's all in the uterus. It doesn't say that it's metabolic, although it certainly could be insulin resistance, or hyperestrogenism, or progesterone resistance, or chronic inflammation, oxidative stress.
All of these things are being thrown about as possible explanations. I can't really say they are or they're not. And of course, we can't say this is definitive, and someone else should obviously look at the study more careful, because it is retrospective.
And I'll end with one final aspect. It might not be BMI. You know, BMI is a pretty imprecise measurement.
And one of my colleagues, Iris Lee, has done a couple studies now that's published in F&S, and actually one in Human Reproduction, that says that there's probably better markers of obesity, central obesity, or central fat obesity, or weight, or dexa, that might give you more specificity than just BMI. We're stuck with BMI as a relatively crude measure at the moment. So, I like this paper, obviously, which is why I chose it as the seminal contribution.
I think it's nicely done, which is the nice point, direct. You know, it doesn't try to overstate it. And the findings seem to make sense.
But like all studies like this, this is really an observation. It's really not a mechanistic study. So, we've learned that obesity probably does matter.
Unfortunately, we just haven't learned what to do about it or how to change our care. But it is something—I'm not sure I would tell patients about this. I mean, I don't hide the information.
I wouldn't, like, warn everybody. Hey, you know, you've got a greater chance of miscarriage. But it does help with the biology of it.
So, Mike and Kate, what'd you think about this? Hey, you take the first swing at it. I loved Kurt's summary. I loved it, too.
And as someone who enjoys speculating and has the two of you to impose guardrails to make sure I'm being careful, you know, I think that this is such a disciplined analysis, as you say. It does not overreach in terms of what the findings are. It doesn't, you know, the findings are not surprising either, but it confirms something that we already see clinically.
You're 100% right, and this is where the speculation comes in, Kurt, is that we really still don't have compelling data as to whether we should be advising these patients to delay treatment in order to lose weight. And I think that really is kind of like the next frontier in how we counsel these patients because, of course, we're also racing the clock, and there are studies of better and worse quality in both directions as to whether it's the right thing to do to advise patients to wait, to pursue weight loss. You know, obviously those are homophones, and so we need more work to be done there.
We need better studies, and it's a very hard thing, obviously, to randomize, but this is a call to our authors out there, to our readership, that this is a question that really our patients deserve to have answered. Yeah, I'll just follow up on that, and then Beth could say something as well. I agree with that.
I remember the effect is small. It's clinically relevant, but it's small, and there's a lot of people that did have a lot of birth with obesity and without. So, it's not something that you have to be so afraid of that, you know, your outcomes are bad.
You can still treat with these women with confidence. So, if we're going to do an interventional study, please do it ethically and with, you know, really good information and with, you know, the best patients in mind. Just don't start by treating everybody with dramatic weight loss drugs just because you saw a 20% difference.
There's those guardrails. Absolutely. So, I love figure two and figure three in this paper, like the heat map and the curves are wonderful, but to Kate and Kurt's point, we don't know that moving backwards in either of these directions will then resolve that.
And, you know, the relative risk was, you know, 1.2. It was 1.19. It was fairly substantial, but the absolute risk, which is why I really like table two, five to ten percent overall, which means you're talking about a number needed to harm of 1 in 10 to 1 in 20, and we don't know that going backwards in the weight or going up in the weight will reverse that if you're underweight or overweight. So, I think those caveats are important because I think a patient could see this and think, well, I should delay or do those treatments. So, I really appreciate Kristen and her work that she does with RMA.
I love all the points you made about this paper, and I appreciate that they didn't overstate their findings. This was a fantastic seminal contribution to F&S this month. All right.
So, we are moving on to Andrology, and I have the next one, and I'm just going to admit my bias from the start. Dr. DeCherney, Alan DeCherney, is one of my mentors, and he always told me be careful of ratios. Ratios are rarely biologically true.
They're sort of human constructs, but this one is looking at a ratio. It's looking at the testosterone to FSH ratio, and does it predict sperm retrieval, and then with idiopathic non-obstructive azospermia? It's actually a really interesting question. This is from a group of authors who are with Mobius, and Mobius is the Male Infertility Research Consortium, and this is a group of authors throughout the U.S. that are trying to look at male infertility, an important and under-addressed aspect of the infertility evaluation.
So, this is looking at men who have non-obstructive azospermia, and what is the likelihood that we have a successful microtessie? So, successful being defined as we get sperm from them on microtessie, not necessarily looking at the birth outcomes or IVF outcomes. This was a retrospective study, and they used 60 men, and they used half of those 60 men to develop their predictive model, and then the other half of them to develop whether or not that held the prediction of whether they would get sperm or not, and they randomly assigned those men. So, they did not necessarily have any bias in how men were assigned to either developing what was predictive or in confirming what was predictive.
They excluded men that had a genetic abnormality. So, in other words, if these men had cystic fibrosis or Klinefelter's or a Y chromosome microdeletion, those men were excluded from this analysis, and they looked at things that I think most other studies have looked at in the past—testosterone, FSH, LH. Other things other studies have looked at would be testicular size.
Obviously, these men have a VAS that's present because of the way they designed the study. They did univariate regression and multivariate regression to see what was associated with the ability to find sperm on microtessie, but then they went further into looking at things that are actually predictive. So, just a reminder to our fellows, if we're just looking at odds ratios, relative risks, those are looking at associations, not really prediction.
Things like ROC curves, where we start looking at sensitivity, specificity, even getting into positive, negative predictive values are looking more at predictive statistics and trying to see if there was something predictive in these ratios or tests that could help us better counsel men. So, 30 of the 60 men were assigned to developing the model to look at what was predictive, and then another 30 looking at if that held up in those. There were differences between the two cohorts.
The prediction cohort was much more likely to be Sertoli-only syndrome. Just as a reminder for all of us, Sertoli-only syndrome doesn't mean that they don't have Leydig cells, it just means that they have germ cell aplasia. So, they have Leydig cells, they have Sertoli cells, so it's a little bit of a misnomer, but they have germ cell aplasia.
That was 77% in the developmental model and only 53% in the validation model. There were also some other differences. The time of surgery was longer in the validation model, but I think those are overall minor details.
I think what's important is what did they find. If you look at figure one, it's actually really helpful because they give you these box and whisker plots. The box and whiskers are really long and overlapping for FSH, LH, and testosterone.
What that tells you is these aren't normally distributed data. It's not parametric. It doesn't follow a bell-shaped curve.
You're going to have patients that are all over the map on these, but as you start winding them down into ratios, the box and whiskers get a little bit tighter, not as tight as I would love them to be, but a little bit tighter, and the groups between which you got sperm and didn't get sperm begin to separate a little bit. So, I think that's maybe where this paper is helpful. Just a reminder to those of us who are maybe REIs and not the male specialists, in the men, FSH and LH are a little bit more decoupled than it is in women.
In women, they typically move together. Maybe an exception is PCOS, where we can have an LH to FSH ratio that is different, but in men, FSH really is a marker of spermatogenesis. It really is controlled by the Sertoli cells, by feedback, by inhibin, maybe by AMH, whereas LH is really controlled by testosterone, probably mediated at the KNDY neurons, by estradiol, which is why Clomid works.
So, FSH and LH are a little bit more decoupled in men than what we're used to seeing in women. It's actually interesting to me that this paper focuses on the FSH to testosterone ratio, because LH and testosterone should be coupled. FSH and testosterone shouldn't be.
It should be the LH that's controlling that, but the FSH, again, is a marker of spermatogenesis. Testosterone is a marker of T function. Do we have Leydig cells? Do we have LH control? So, when they went through all this univariate and multivariate analysis, interestingly enough, the T to FSH ratio was the only one that sort of came out in the wash as being predictive.
Now, when I say it came out in the wash, its relative risk was 1.02, with an odds ratio of 1.00 to 1.04. So, it's not like it's coming out as massively predictive. When they applied it to predictive statistics, though, it did reasonably well. AUCs between 0.07 to 0.08 for getting sperm on egg retrieval.
I do wish they had given us the confidence intervals for those AUCs. They didn't, and you can easily calculate those. Given that they only had 30 patients in both the model development and the confirmatory arm, I think those 95% confidence intervals of that 0.7 to 0.8 predictiveness are probably going to be pretty wide.
But overall, I think, like the last paper, these authors did a good job in being relatively conservative in how they interpreted their results. They don't say that we should use this as a way to exclude men from having surgery, and I actually don't think they did the analysis that would have been appropriate for that, because sensitivity and specificity don't tell us what we want from the patient. They tell us how the test is doing.
We would actually want positive and negative predictive values, and if you're going to tell a man you shouldn't have surgery, you'd want something that's really high on its negative predictive value. If you're going to encourage a man to go to surgery, you'd want something that's reasonably positive predictive. I do think, though, this probably aids into the counseling from a reproductive urologist in telling a man what we think the chance of success is with getting this surgery.
Knowing testosterone, knowing FSH, knowing LH, these aren't new. These have been published in dozens of other studies, but I think the ratio, and they have, I don't know, seven or eight different supplementary tables where they really drilled down statistically on this analysis, I think is where the value add was. My only criticism would be when you start looking at the positive and negative predictive value, which they didn't really dive into for a test, I think that matters, and can you develop a threshold that is actually clinically meaningful, and they didn't really drill down on that, and I'll just use an example where we have, there have been prior studies looking at premature progesterone elevation and its ratio to eggs or follicle numbers, and that that could represent follicular dysfunction and IVF.
We published a paper in F&S about five years ago drilling down on that. When you take two variables, FSH and testosterone, and add them into a model, if you just look at relative risk, they're going to do better, right, because it's two variables. You're adding two variables into one, but is it better than the two variables separately? The other thing is when you add FSH into the denominator, you've now transformed FSH, like logarithmically, you've made it one over FSH, and so really what I would have liked to have seen is the positive and negative predictive value of testosterone versus one over FSH compared to the ratio of the two of these.
I actually guess that it probably wouldn't be that additive, but that's my own bias going back to where I started, that Alan told me biology rarely functions in ratios. It tends to be more a construct that we as humans make, so I think this was an interesting study. I liked it.
It was statistically very rigorous. I wish it had pushed it just one level further to actually tell us if it's additive to making the ratio versus having these two as separate variables, and whether there's a threshold that is clinically meaningful, because I think most of these men are going to elect to undergo surgery anyway, which is what the authors sort of say with donor sperm is the backup. So those were my thoughts.
Really interesting paper. Curious what Kate and Kurt thought. Yeah, I echo your analysis, Micah.
It's super helpful to have you break it down, and I too remember Alan's points of caution about over-interpretation of ratios, so I was happy to assign this paper to you. In terms of what we need here for our patients that are making assessments and decisions around whether to undergo this relatively invasive procedure, we do need something that has nearly 100% negative predictive value if we're going to tell them not to do it. So it's helpful.
Our patients want to go into surgery with reasonable expectations, but as you mentioned, they're all still going to do it. So there are some really interesting diagnostics coming down the pike as well, like super high-resolution testicular ultrasound that our colleague Paul Shin is helping to evaluate. I am really hopeful and optimistic that we will someday have a tool that will really guide our patients better, because so many do go through this very invasive procedure only to be disappointed.
But for the time being, this paper, as good as it is, doesn't really move the needle for me in terms of, you know, counseling more patients against giving this a try. I think that negative predictive value is really where I come down on too. If we could get to some place where we're close to 100% on it, you know, depending upon your urologist, we've probably all seen testicular micro testes samples that come back where you think half the testes was removed, and if there's no sperm that's coming out of that and we're reducing their testosterone production, we've probably done that guy a disservice.
So if we can get better prediction beforehand. So I agree with you. I think this moves the needle a little bit.
I think there's still more that research that needs to be done that's going to build upon this story. So we are now moving on to assisted reproduction. Kate, we have an interesting study on telomere markers.
I know this is something that we've been talking about for over a decade now, but this is a pilot study looking at a novel embryo evaluation assay and does it add anything to our prediction of what's going to happen with these embryos? Yeah, I'll leave it to you two gentlemen to assess at the end whether it moves our prediction. It's certainly a paper well worth the read, especially for our fellows just thinking through how assay and metric validation does and ought to work. We could spend this entire podcast talking about the supplemental figures and the validation that these authors performed on the TELO score, which they are implementing here in trophectoderm cells of PGTA-tested blastocysts.
So first author Lo and senior author Li out of Taiwan present this validation and clinical correlation data. They define the metric as TELO score, which is to assess telomere length, and it is telomeric reads per million total reads. So that breaks down into absolute units.
And they essentially take these data and then correlate it with clinical pregnancy primarily and live birth secondarily following PGTA-euploid single embryo transfers. So lots and lots of analysis here. They purport that the novelty of the study lies mostly in that the patients went through a PGTA platform with NGS that is a standard platform with routine depth and read length versus prior publications looking at telomere length or TELO score in blastocysts relied upon higher sequencing depth and longer read lengths than standard PGTA platforms that we typically would be using clinically.
So again, we can talk a little bit through whether this is novel and builds upon prior studies as well. What they looked at was 5,251 blastocysts from 1,451 patients who underwent IVF over an 18-month period from 2023 to 2025. So quite recent data.
The validation studies they did, which again, we could spend quite a bit of time just talking about the validation of the TELO score as a metric. They were done in cumulus cell masses. And then they also compared the NGS-based TELO score to qPCR-based assessment of telomere length.
And really interesting discussion as well on how they even chose the primers that they used for their qPCR in terms of balancing kind of sensitivity and specificity there. The mean TELO score was 8.4 with a standard deviation of 4. And they also evaluated the development cohort versus a validation cohort. We can talk too about whether this is a true validation cohort because I think we can think about that in different ways.
The development cohort had 234 transfers and the validation cohort 101 transfers. And ultimately what they found for their primary outcome on both univariate and multiple logistic regression, that there was a statistically significant positive association between TELO score and clinical pregnancy. They go back and forth between calling it implantation and clinical pregnancy here, given that they're single embryo transfers, it's the same thing.
It's a gestational sac following the transfer. And they report an odds ratio on the multiple logistic regression for TELO score of 1.096 per unit with a confidence interval of 1.011 to 1.188. So fairly modest association, but one that's there statistically nonetheless. They start out in terms of incorporating TELO score by creating strata.
And another thing I remember being taught over and over in fellowship is to divide the data is to change the data. So one of the things I would have loved to see was an association or a curve of the TELO score as a continuous variable. They looked at the strata as 0 to 4, 4 to 7, 7 to 10, 10 to 13, and greater than or equal to 13, finding that those in the lowest strata had significantly lower clinical pregnancy than the other groups.
So 37% in the lowest strata versus 63 to 66% in all the other groups, which were not statistically significantly different from one another in terms of clinical pregnancy based on TELO score. They then incorporated the score into a predictive model. So again, lots of analyses going on here that included day of biopsy, BMI, and morphology.
And they found an area under the curve for this model of 0.668 in the development cohort and 0.634 in the validation cohort. So again, the authors throughout the paper very assiduously, cautiously, and appropriately note this is meant to be hypothesis generating. I would agree with that 100%.
I have a couple of, you know, I would say quibbles with the analysis that I'd love to hear both Micah and Kurt's thoughts on as well. For one thing, the way that they stratified the data was somewhat problematic. The group that was in the zero to four group for TELO score in the development data set, so the primary data set, was only 19 transfers, of which seven implanted.
And so it's hard to draw really strong conclusions from that. And they go on to rely very heavily on that finding. They don't show how they arrived upon that threshold or how they created those strata.
So this could be somewhat artificial and spurious. The other thing is that they show very clearly that TELO score is associated with embryo quality, both the rate of progression and the morphology of the embryos, and yet both are incorporated into the model where they report their AUC. We don't see a model of what the AUC would be absent the TELO score or whether it actually adds anything at all to the predictive capacity of their model, which is modest anyway in the 0.6 to 0.7 range.
The other thing is I think about a development data set and a validation data set a little bit differently than this. Typically, I would think that you use your development data set to create a model that optimizes your diagnostic parameters or your predictive parameters, and then you apply that same model to a second data set. What the authors really showed here was to compare that they had similar distributions and outcomes kind of across the board, which is helpful, but it's to me kind of just dividing it into two-thirds and one-third and doing more or less the same analysis on both of them.
That was another way in which I thought we could see some additional validation of these data, maybe in an external data set for the next set of authors that use this. I also, and this is maybe my ignorance, but I was really not able to tell whether this is something that could be performed without going back to the stored DNA. They did have, they purport that this is novel insofar as it's using standard PGT platforms, but they do go back to the stored DNA in order to obtain their results.
Again, it was unclear to me whether this was something that they could have done just based on the bioinformatic data that already existed and the going back to collect the data from the stored DNA in terms of performing the qPCR, et cetera, was just for validation purposes or whether to do this in clinical practice, we would need to go back to the stored DNA from the trophectoderm cells. I think really interesting and exciting paper we all know well and don't have time to cover in this podcast, the biologic plausibility behind telomere length as a marker of reproductive aging and of reproductive competence in various reproductive samples across the board. I look forward to learning more about this.
Unfortunately, and no time for this today either, they evaluated MITOS score, which has been really kind of a flop and disappointing to all of us in that it does really just predict morphology and PGT outcomes, but is not additive beyond that. They showed that in this analysis as well. But I think there's promise for teloscore.
I think this is a pilot study and we need to see more. Micah and Kurt, what did you think? I think the value of this paper is in its novelty and by definition pilot. I mean, you did a nice job saying why just because it's published in Fertility and Serility doesn't mean that it's definitive and you should adopt this and it should be put into clinical care.
There's always a balance between something that's novel and new, and I think we should be publishing things that can move forward. But you did a very nice job articulating why this can't be definitive. It's becoming a very soft standard to divide your dataset to validate, but I'd still agree with you, it's nowhere near as definitive as doing it on another dataset.
And one should always be concerned, as you mentioned, regarding how you manipulate the data. So again, we've toggled for you on this podcast between large datasets that have clinical implications to really pilot dataset, because I think you should read both. But you should also know the difference of when something is really ready to use and when something is published in science for the purpose of the necessary replication before it actually can be considered true.
I can't summarize it any better than what you and Kate just said. I completely agree. And I applaud the authors for at least saying that this is pilot study, hypothesis generating, warranting further research, and being careful to say this doesn't represent clinical validation.
Kate, I completely agree with you on figure one. If they hadn't divided those 19 pregnancies by these sort of arbitrary telomere scores, I think they'd have a flat graph. But because it was divided that way, they found one thing that was statistically significant.
Because once you get over zero to four, it's completely flat. It adds absolutely no additive value. And just to put it in perspective, the relative risk of the telomere score was 1.09. Just day five biopsy alone was two.
So like the difference between what this would add incrementally, even if it's true, over just the day that it gets to a blastocyst is relatively small. So we'll see how this story plays out. Definitely hypothesis generating.
Exciting to see what next research comes out as this gets further validated. Kurt, really curious to learn. We're staying in the area of ART.
I mean, you are the expert on early pregnancies, HCG trends, HCG values. Is it ectopic? Is it a viable pregnancy? What did you think of this study that came out of the RMA group? Yeah, I got another good paper that not necessarily earth shattering novel information, but a nice clean way of looking at an old question that allows us to get, again, a clinical nugget. And it's a good paper because it's so confined and concise, not concise in terms of the amount of data.
They have a lot of data. So I'm talking about the paper prognostic utility of serum beta HCG level following a single euploid embryo transfer for live birth. So I seem to be the single euploid embryo transfer guru today.
This is out of the Department of OB-GYN, Aiken Medical School, as well as Reproductive Medicine Associates in New York, as well as some help with the U.S. Fertility Group, if that's your well aware of in Rockville, Maryland. The first author, Emily Clark, did a very nice job, and the senior author, Philip Romanowski. So this is an age old question, right? And like, you know, something that's been around for a long time to evaluate the live birth outcomes on the basis of a single HCG following a single euploid embryo transfer.
Similar to what I said before, no one's recommending that every embryo has to be a euploid embryo and everyone has to have PGTA, but this is, again, taking advantage of a common practice, isolating the data to remove a lot of confounding factors and try to isolate the question you're asking. So a lot of the literature on this, and there's lots of literature on the prognostic value of a single HCG, is really confounded by when did you get it? What was the quality of the embryo? What was the patient? What was underwent? So this is a way of kind of normalizing that. So it's a large study with a decent amount of patients, more than 6,000 transfers that resulted in live births, and the methodology is relatively simple.
Let's divide them by their initial HCG and then let's find out what their ultimate outcome is. So just to ground you all, because I'm sure we're going to talk about this in a second, in this particular practice, the HCG is measured nine days after a frozen embryo transfer. I'm going to get back to whether everybody does that or if that's correct or not in just a second.
So based on that, they took that large group of people and divided it by their single HCG into six groups. They're perfectly fine groups, intuitive. The first one is an HCG of 2.5 to 11.
Now that happens to be the smallest group that only has about 375 patients in it. Then they incrementally increased it from 11 to 25. That has another 350 patients in it.
Then 25 to 50, 550 patients in it. 50 to 75, another 550. Another group, 75 to 100, that has another 575 patients in it.
And the largest group is greater than 100, which has 4,000 patients in it. So the devil's in the detail, of course. The smaller groups are smaller, but it still gives you a nice answer.
What they found was that the chance of having a live birth increased as one had a higher single HCG. It went from as low as 2% in the smallest group, their HCG was really quite low, less than 11, up to as high as roughly 88%, 87.9%, if your initial HCG was above 100. Now, in Figure 1, there's a nice figure that shows you graphically how that it's not a linear curve, but the curve is increasing dramatically, minor quibble.
You shouldn't be connecting the lines in these curves. It's not a line, it's just different groups. But nonetheless, you can see that the groups are increasing with the HCG value.
Now, if you want to play with numbers, it's not just you increase your chance of pregnancy rate from 2 to 11 to 35 to 64 to 76 to 80. You can play with relative risks. And if you compare group 2 to group 1, you have a sixfold increase of having a live birth.
If you compare group 3 to group 2, you have another threefold increase. If you compare group 4 to group 3, you have another 1.8 increase. You get the idea.
And they did a ROC curve to say basically that the break or at least the value that they say that their most confidence in for maximizing sensitivity and specificity is around 76. I don't know if that's exactly right, but I think that's good enough for the moment until I find it on my notes. So the finding from the study provides really a counseling tool is the way they're couching it for both the patients and physicians.
And it is pretty clean that you're talking about a single transfer, a euploid embryo, same day of HC measurement. They did look at blastocysts day 5, 6, and 7 again to see if there's any difference. And they found some sporadic differences, especially in the middle group where counterintuitive, but in the group who were around 70, HCG of 75, if you had a day 5 embryo, you actually had a lower chance of having a live birth than if you had a 6 or a 7 day embryo.
They speculated to say the day 5 embryo is quote unquote more advanced. So you would have expected a higher HCG and therefore the fact that it was lower was actually a bad prognostic sign. You know, I get the logic that they're trying to fit a square peg into a round hole, but I'm not sure you can make too much about that.
Again, guardrails as Kate said. So we'll look at that. So the area under the curve is 75.7 that they felt was the most to maximize your sensitivity at 86.7 and your specificity of 66 with predictive values around 86 for a positive birth rate value and 67.3. Now this is where I want to ask people to have caution.
This is a nice clean data set. You can really see intuitive levels here. It really mimics what we think we know is true, and I guess you can be a little bit more enthusiastic or pessimistic with your patients in the first call, but these are not thresholds that I would act on.
These are actually relatively poor sensitivity and specificity and predictive value, and I would never give a prognosis value on one value. They did have one pregnancy that went to term with an ACG that started at four, so you can't even say zero, and I am not a fan of what they call the Yowden Index, which maximizes the sensitivity and specificity in the area under the curve because it's lazy. It's basically just saying, well, what's maximizing the true number? What you really are interested in something like this is one or the other.
You really want to know that you've got everybody that's normal or you've got everybody that's abnormal. You don't want to find something in the middle that has a bad prediction of both. So I think that you're, you know, this is again a very well-done paper.
I compliment the authors. That's why we published it in F&S, but it really is just a clean counseling method. Now there are some other nuggets that make sense.
The lower the ACG, the higher the risk of a copy pregnancy. That makes sense. The higher the ACG, the higher the risk of twins.
Now, yes, these were single embryo transfers, but we do know that some of them do split, and it is with the higher ACG. And, you know, I was hoping, again, I'm complimenting this paper. I don't want to give you a bad statement here, but I was hoping they could have done more.
I would have really loved to see what the second ACG was and what the change was because that's really what we look on is not the first ACG, but perhaps the change in the first two. I would bet gives you a lot more prognostic information than just a single one. So it's as said by Esram Eskan and Leah Bernardi at Northwestern and the Reflections, this is really another example of cleaning up what we think is a really important question with a good data set, but it still really doesn't change the biology or changes our overall ability.
It just gives us more information for that really difficult first call that we make for somebody that wants to know how their pregnancy is. But this is not, again, I repeat again, should be used for stopping care or telling people about the prognosis because there's examples of both. There's examples of losses with a high ACG.
There's examples of pregnancies and low ACGs. And, you know, I'm not really sure what to make of the area under the curve with an ACG of 76 because, again, that's just an average number that you can feel good about. So I like the paper.
I think it does help us with our intuition and solidifying what we think is true. It does help us with the clinical call, and I'm glad to see it was done in such a narrow data set. Buried in the discussion is a question, it's a discussion I want to have here, which is not every lab is the same, not every practice is the same, and not everyone does their ACG at the same day.
So you have to be careful about posting this on your wall with your nurses to say use this information, because an ACG a day earlier or a day late is probably going to have very different prognostic values, and so you can't use these absolute numbers. So the first step is please make sure that you find out what you're doing in your practice before you try to use this, even for a counseling tool. But, again, congratulations.
The authors did a nice job with a nice data set. It's a large data set. It's very precise, and we will, again, as Michael loves when I say this, let it flow over us to understand better how our patient had to treat everybody, but it doesn't give me a threshold.
One of my biggest concerns in our field is that we're so busy we tend to practice threshold medicine, if this, do that, and I just don't want this to become a new standard where, you know, if your ACG is less than 25 on your first one, you say, you know, stop the progesterone. You've got a failed pregnancy. Let's move on quickly.
Yeah, that might make the practice better, but that's just not doing the patient a service. So what do you guys think about this? I love that discussion. I mean, this is fireside stats with Kurt Barnhart, and this is, I think, the thresholds you've developed and published before about where you can rule in a non-viable pregnancy with 99%, you know, plus prediction, and you can't do that from a single value, and you have to see those trends, I think is really, really important.
I think this adds to the science. I'm letting it wash over me. As you say, one of my prior fellows did this exact same study five years ago, and we decided not to publish it because I just questioned the clinical utility.
Like, if we tell a patient you have 11% chance or 20% chance or 50% chance that it's going to be viable, is that helpful for them or not helpful? I don't know the answer. Maybe for some patients, it is, but, you know, ultimately, you're going to have to see what the value is in two more days, and you're going to have to keep them on medicine. You're probably going to have to see what it is in four more days, and you're probably going to have to see what the ultrasound shows.
They're going to have to wait. So for some patients, is this information helpful? Potentially. Certainly, the worst thing we could do would be to change clinical management, as you say, and withdraw medication and then cause that pregnancy to fail because we acted upon that one early single value.
I was hoping this was where your discussion and summary would go. It was... Kate, what do you think? Yeah, you guys said it all. You took the words out of my mouth with the thing I'm not going to do is put this up on the wall for my nurses to do their positive beta calls and answer the questions with these exact percentages because it just doesn't map on one-to-one in that way, and it could really give patients the wrong impression if not delivered skillfully and with the appropriate nuance.
I think that where it is most useful, and kudos to RMA New York for these really outstanding success rates, frankly, but is in these patients, which it was the overwhelming majority of their dataset, 4,000 of the patients who have a high HCG at that first level from a PGTA-euploid transfer, that their chances of a baby are outstanding. So 88% in this clinic, obviously going to be different from clinic to clinic, but to be able to bolster the excitement of a patient who has a positive result in this setting when it is high is also of value. All right.
In the interest of time, I'm going to move us along and I'm going to try to summarize an incredibly complex trial quickly so that we can leave time for Kate on the last one. We're back to emulating target trials. This one is, again, from our friends at the University of Pennsylvania, first author DiTosto, Enrique Schusterman, Sonny Mumford, who are awesome epidemiologists and have really been pioneering how we can emulate target trials.
This one is looking at a target trial of surgical removal of uterine fibroids and atherosclerotic cardiovascular disease. The first thing is, I didn't even know that we needed to have this question. I literally did not know that fibroids had been associated with an increase in cardiovascular disease.
They talk about the biologic plausibility of it. Essentially, there's at least six studies that have showed that having fibroids is associated with a significant risk of increased cardiovascular disease in women. The proposed mechanism is smooth muscle cell proliferation, hormonal alterations of uterine fibroids that could accelerate plaque formation and systemic cardiovascular disease.
I emailed two of my mentors, who are both experts in fibroids. One was excited about this hypothesis. The other was very skeptical.
Bill Catherino, editor of F&S Sciences, in typical Bill fashion, just doubted that maybe there's a true causality here, but this is what they're investigating. It's actually a really interesting use of retrospective data. They're looking at almost a million women with fibroids in this Optum database, so this big database of women from an insurance standpoint, almost a thousand of them.
Again, just really quickly to review emulated target trials, this has been shown to approach randomized clinical trials and how it estimates risks or associations or even treatments, although in this case they weren't actually looking at treatments per se. What are they doing? They're looking at women who had fibroids diagnosed in their insurance record and they either didn't have surgery or they had a myomectomy or they had a hysterectomy. Did they have a diagnosis of anything that was associated with cardiovascular disease that followed that? This is where it gets really complex in an emulated target trial.
They essentially emulated the trial every month for over a decade. Every month, did the women have the diagnosis? Did they not? Did they have one of these adverse events or diagnoses or did they not? They end up with about 217 target trials per month looking at this and women could move in and out of these categories because you might have fibroids and have no surgery and then at some point you might have a myomectomy and then at some point you might have a hysterectomy. If you have a hysterectomy, you can't really go backwards from that, obviously, so there's some censoring of the data based upon that.
Now, for an emulated target trial, and Kurt, maybe you can help all of us understand it better as we get there, I think this is what we all try to do with our observational data, but the emulated target trials are just taking it to another level as far as how they censor the data, how they do their DAGs, how they do their inclusion-exclusion, how they wait for what these risk factors are. So there's a ton of things that we know are going to be weighted risk factors for what will cause you to have atherosclerotic disease. I'm just going to summarize them into big categories.
The age at which you enter the trial, the age at which your diagnosis of fibroid starts, the age at which your diagnosis of atherosclerotic disease starts. Are you obese? Do you have high lipids? Do you have diabetes? Do you have hypertension? Do you smoke? Mental health and depression are associated with them. Your reproductive history, parity, infertility, endometriosis.
Your fibroid symptoms. Are you using hormones? Are you not using hormones? All of these things they accounted for and gave weights to to try to emulate this target trial, almost as if you could randomize patients. Now, they were very specific to say they're not trying to emulate a target trial in which you randomize patients to no surgery, myomectomy, or hysterectomy.
They're just trying to emulate the association of those interventions with their outcomes. Kurt, maybe you can help me understand that. They then did this 217 times on a month-by-month basis simulating this target trial.
You end up with so much data that they then put each of those 217 data points into almost like a regression analysis to give you a curve of what that would look like, and then looking over time with a Cox proportional hazard. Some of this methodology is almost beyond my ability to comprehend, but it seems very exhaustive. We're looking at almost a million patients.
Out of those million patients, though, you end up with relatively small numbers of adverse events. I think this makes sense because they're looking at reproductive age women. Mostly 150 had adverse events following myomectomy, 506 following hysterectomy, 606 with no surgery.
The big caveat to the no surgery group, they had a million patients and they're looking at this month-by-month. This took so much computing power that they randomized 5% per month into their no intervention group. Like literally, this is, he's the editor of the American Journal of Epidemiology.
He's a dean of a bunch of schools at Penn. I'm sure he has massive computing power, and even with that, they have to limit it to 5% because it's taking so much computational power to estimate the risk if you had no intervention. All right, enough of the methods.
What we really care about is what did they find. I'm going to summarize it very, very simply. If they had myomectomy, statistically, there was no effect.
If you look at a trend or a direction of it, it was actually to protective, but by less than 1% and it crosses the null, so no difference with myomectomy. What if you had a hysterectomy? It actually showed a trend, I hate using that word, but something in the opposite direction where there was increased risk, but again, it was about by 1% of having these adverse cardiovascular events and in most of the analysis, it hit the null or it crossed the null just barely. So maybe something there, maybe not.
When they broke it down by what type of adverse cardiovascular events you had, there really wasn't anything clear to me that was a signal. So overall, I'm going to give my take on the data. I didn't know this was a biologically plausible hypothesis to start with.
Maybe it is, maybe it isn't. I asked two fibroid experts, I got two different answers, but I think this is a really interesting way to gather over a million patients and look at data and look at it incredibly rigorously and what we essentially take home is that myomectomy, not really an increase or decrease risk. Hysterectomy, maybe a slight increased risk.
Can we postulate why? They excluded patients that had oophorectomies because we know estrogen is cardioprotective, but are we compromising the blood flow to the ovaries and maybe our production of estrogen? And what I love about Enrique is he loves, and Sunny, they love looking at things that are low risk, low cost that affect the masses. And I think that's probably why they asked this question. So is there something more that we could study from this to see is there a risk of having hysterectomies because of fibroids at a young age? Are fibroids actually causative of adverse cardiovascular events? I don't know.
I would consider this hypothesis generating. I hope I didn't take too long on that. This is like probably the most complex study I read this month and I'm still trying to wrap my brain around it.
Kurt, you can correct me on anything methodologically that you think is important for our audience to understand because I think we're going to have more and more of these emulated target trials coming out. So I think as editors, as reviewers, and as readers, we need to understand what are they asking? Are they asking the right question? Can their data actually answer the question? And can we trust the data more than what we historically have thought of from our observational datasets? That was a nice summary, Mike. I think the take-home message was, or at least the hypothesis, my understanding of looking at this paper was that there's a lot of confounding in fibroids because of who gets them and who doesn't race obesity.
And it's hard to figure out whether that confounding is cardiovascular disease or whether disease contributes. But they went beyond that a little bit to say, is the treatment an issue? And I think the findings in a nutshell, as you said, is that a hysterectomy might be over-treatment and may actually be worse than, not worse than the disease, but not the best treatment compared to some others. So it does have public health implications.
Now, I agree emulating a trial is really taking observational data to a new level. And there are papers on this, including about a year ago, we published in F&S how to do a very good emulated trial. And it really has a couple of different foundations, many of which they did.
One is that they have an a-priority hypothesis that you set up the study question before you do the study. That's why you use DAGs to figure out like what should be your analysis and what should be a confounding and what should be an outcome. And very rarely is this done in our literature.
They just kind of throw all the variables in and control for things. So it's a much more informed way of answering the question. So you're not getting misinformation based on interaction of the data that you didn't think were there.
The other really important aspect of an emulated trial is inclusion and exclusion. So you're trying to get rid of the bias from the tails, if you will, like in the hysterectomy group were sicker or older, or the myomectomy group was inherently younger or thinner or something like that. So if you can use criteria to make sure you're really trying to study the same person theoretically, then you can understand the associations better.
It's not the same thing. They're correct in a pseudo-randomization because you didn't randomize them, but you're basically making sure your population is more pure and therefore your associations are more true. Now part that gets me a little confused is the time dependent variables that they use, and that's what takes all the computer power.
It's not just taking information at the beginning, at the end, but they're really doing, as you know, a cox proportion has got time in it, so they're really doing it incrementally all along the time. And that's what the computing power is. They're basically analyzing it like every five years in a 30-year time span.
So they're doing 30 trials or observational studies in one to smooth out the curves to look over time, and that's where the modeling and the computer power comes in. So this one was a little bit sophisticated, but the point that I wanted to make was we should be much more rigorous in our observational trials. I'm not sure any of us can be as rigorous as Sunny and Enrique are in their studies, but it's not just get a database, set your exposure and your outcome, or actually look at the data, depending on what you find, then set your exposure and your outcome, and then assume just with one method of controlling you, you've taken all the bias out.
There are really ways to do this better, and it's becoming an art, and you know, this is the group that's kind of leading the way in it. And I'm glad that they chose Fertilinus early, so I hope that it'll enhance our readership. Maybe you don't understand the methods in this paper the first time you read it, that's okay, but the idea that you're seeing what they're trying to do as opposed to just taking a huge database and torturing it until you get the answer you want is the message we're trying to get across here.
Yeah, I think that's well summarized, Kurt. As a past president of SART, and Kate runs our quality assurance committee for SART, and I know you've had critiques of SART observational data and how we analyze it, I really hope that we can have some sort of meeting of the minds to sort of elevate how we look at observational data so that we make sure we're getting to the most correct answer, which is what all of us want as readers, as practitioners, as scientists on this. Right, we recognize there's bias in observational studies, some of which we can address, some of which we can't, but you want to really be sure that you're doing the best you can.
You don't want to be, I think I used the word sloppy before, or lazy. You really want to be rigorous about it. Maybe you don't get to work with Enrique and Sonja on every paper you do, but, you know, working with a statistician that understands this will make your paper that much more believable and powerful.
Awesome. Kate, we have time for the quick summary of our last paper. We're moving on to research letters, actually my favorite section of the journal.
What was our research letter on perceived weight gain with OCPs in women with and without PCOS this month? Yes, this is a study by Catherine Voss et al. out of Rochester. The authors use data from the 2017 to 2019 National Survey for Family Growth, so this is a really rich data set because it surveys greater than 72 million reproductive age women, has a very high response rate, 69%.
They wanted to assess the hypothesis that patients with PCOS at the time of this writing, now obviously undergone a name change to PMOS, had a higher rate of OCP discontinuation due to perceived weight gain than patients without PMOS. And so the way that they looked at this was they evaluated the sample of women who had been exposed to OCPs, reported having used OCPs at any point, and they looked then at the subgroup who self-reported PMOS. So this is an improvement in the survey that in 2017, which is why they chose this time period, the NSFG added this question of did the patients experience PMOS or PCOS is the way it would have been phrased at the time.
And there were 330, so a relatively small sample of patients who reported this, basically having been diagnosed with PCOS, and the authors went on to validate the diagnosis because the patients were also queried as to whether they'd had an ultrasound, whether they had clinical evidence of hyperandrogenism, and whether their periods were regular. And so to me, actually, one of the most interesting and important findings in this research letter was that about one-third of the patients who self-reported PCOS, in fact, did not validate based on the criteria that were available in this data set of the survey. So 237 actually validated as PCOS.
So then they went on and they looked at the characteristics of the patients validated with PCOS versus not among these OCP ever users. What they found was that the group with PCOS were more likely to be white and have a higher BMI, and that they made up 6.6% of the total population of reproductive age women who'd been exposed to OCPs. And this 6.6% of note is lower than the generally reported greater than 10% prevalence.
So is the validation accurate or not, or is this just a low prevalence in this population, or do women generally report PCOS, and in fact, not all of them have it. In terms of their study question as to were the PCOS patients more likely to discontinue OCPs specifically because of weight gain, a very specific question, there was no difference. So 37.6% of women with PCOS and 35.1% of women without PCOS discontinued OCPs.
Of these, 30.3% with PCOS and 19.5% without PCOS cited weight gain as a reason for discontinuation. Neither the rate of discontinuation nor the weight gain as a reason for discontinuation were statistically significant on univariate or multiple logistic regression analysis between the women with or without PCOS. So the multiple logistic regression adjusted for age, BMI, race, ethnicity, parity, marital status, education, income, and smoking status, and again, no difference.
So the hypothesis that women with PCOS were more likely to discontinue OCPs because of weight gain was not borne out by these data. I thought this was a really nice research letter, makes very good use of a robust dataset. There's a nice reflection as well by authors Johnson and Ajun, and it suggests that the groups should have been further stratified by BMI.
I would push back against that a little bit. It's a pretty small dataset already. And if you look only at those patients who were overweight or obese, it would further attenuate the possibility of finding a difference here.
Of course, they may have been underpowered to find this difference as it was. There was a 35.1 versus 19.5% rationale of weight gain stated, and it was not statistically significantly different. The authors also note that, and I think this is astute, the authors of the reflection, that the metabolic aspects of PMOS may counterbalance perceived downsides of OCP.
So because OCPs mitigate some of the metabolic aspects of PMOS potentially, and also the hyperandrogenic symptoms, even though women with PMOS may struggle with weight gain and that part of what they perceive about OCPs may be a reason to discontinue, they also may be more likely to experience some of the non-contraceptive benefits of OCPs in that they mitigate PCOS symptoms. So nice research letter. Interesting to see that even though we worry that our patients, especially those with PCOS, will stop beneficial OCPs because of weight gain, it seems as though in this population, at any rate, it was not different, or at least not statistically different among patients with PMOS versus those without.
So kudos to the Rochester group and a great example of use of an existing high-quality data set. Thank you, Kate, for that summary. Again, as always, to our authors and researchers out there, research letters, you know, some of the studies we talked about today had dozens and dozens of supplementary figures.
They're very complex and need a full article, but sometimes there are tighter questions that have smaller data sets and can be answered in something simple like a research letter. I love reading these because of how informative they are. Kurt, any last thoughts from you as our editor-in-chief? I think we handled some really complex stuff on this podcast.
I can see that our listeners are already smarter, so thank you for listening to us, and we look forward to doing it again. Well, thank you to our listeners from around the world. Thank you to Kurt, who always sends us out to these meetings internationally.
Several of us just got back from ESHRE in London. I want to thank Kathy from Ireland, who came by and said hello. Yousef from Australia, we appreciate you listening, and Miguel from Portugal, who stopped me in the bathroom because I told him he dropped the card to his hotel room, and he said, hey, I know your voice.
You're Micah on the podcast. Yes, I am, and it's very nice that you listen to us. For all of you around the globe, we appreciate that you listen to us.
Our job here is to just continue to advance and share the science that is happening to infertility and sterility. To our authors, thank you for making us such a successful journal, as Kurt mentioned earlier on, and we can't wait to talk to you again next month. This concludes our episode of Fertility and Sterility On Air, brought to you by Fertility and Sterility in conjunction with the American Society for Reproductive Medicine.
This podcast is produced by Dr. Molly Kornfield, Dr. Adriana Wong, Dr. Elena HogenEsch, Dr. Selena Park, Dr. Carissa Pekny, and Dr. Nicholas Raja.
