Journal Club Global Live at MRSi: Intrapatient Variability in Retrieved Oocytes and Ovarian Response Categories Between Cons

Fertility & Sterility is proud to once again partner with MRSi for this live, onsite Journal Club Global. The discussion will focus on the paper, “Intrapatient variability in the number of retrieved oocytes and ovarian response categories between consecutive in vitro fertilization cycles.”

Discussants: Kate Devine, MD; Eric Forman, MD, HCLD; Rachel Weinerman, MD; Allison Eubanks, MD Fertility and Sterility Moderator: Micah Hill, DO

Good morning everyone. Welcome to Fertility and Sterility Journal Club. We are here live at one of my favorite meetings, MRSI 2026 in Chicago.

And I chose this paper because it sort of challenged some of my assumptions, or at least I was a little bit surprised by it. I love seeing papers that confirm my biases or what I think I know, but even more I love things that push my understanding a little bit. So that's what we're going to talk today about.

We have an all-star panel here today that I am very excited to have. We have Dr. Allison Eubanks. She is our very first ever Fertility and Sterility Editorial Scholar.

So she spent two years with us learning how editorship in a journal works, joining our monthly meeting with our four editors in chief and leader team. And so she is going to be presenting the article and joining the discussion. We have Dr. Rachel Wienerman.

She's the RAI Fellowship Director from Case Western and a phenomenal teacher. She actually speaks here at the MRSI board review course with me. I'm sandwiched unfortunately in between two of her talks.

My fellows after the course always say those talks by Dr. Wienerman were amazing and then they forget that I was between or they're like yours is pretty good too. So we're excited to learn from her. We have Dr. Kate Devine.

She is my partner in crime as a co-host of Fertility and Sterility on Air. She is the Chief Medical Officer and Chief Research Officer for U.S. Fertility and has led multiple important clinical trials in our field. And last but never least, we have Dr. Eric Foreman.

He is both the Medical Director and the Laboratory Director as an MD HCLD at Columbia University. Also author and leader on several key clinical trials in our field and a past president of this amazing meeting MRSI. So our paper today is Intrapatient Variability and Retrieved Oocytes in Ovarian Response Categories Between Consecutive IVF Cycles.

Dr. Eubanks, I'll turn it over to you to introduce the paper to those of us who haven't had a chance to read it yet. Sure. Thank you.

So this paper was written by Hatchberg et al. It was published in Fertility and Sterility earlier this year. And the central question this paper is trying to answer is when a patient has a disappointing response, does that mean that the protocol was inadequate? Do we have to change things moving forward? So why does this matter? I think, you know, most of us have a clinical feeling that, hey, if we didn't get where the response we wanted at the initial cycle, you know, we should change something in the next cycle.

And that intrinsically that there's something wrong with the cycle, and that's why we got the outcome we didn't want. So should we increase the gonadotropin dose? Should we switch from Antag to agonist protocol? Should we change the trigger type? And so again, the core question here is, is there underlying variability biologically that we can't influence with protocol or are we able to control more than we think iatrogenically? There are two studies I think that are relevant to understand before this paper was published. And so one was published in 2021, and that showed that switching the protocol on a repeat cycle did not actually improve lab outcomes with any sort of importance or statistical significance.

The second paper published in 2014 also shows that there was no statistically significant change in outcomes when you change the protocol or didn't change the protocol, but in fact that there was some underlying biologic variability, somewhere between 15 and 20 percent in each patient. So this study at a glance was a study that involved over 180 physicians. It was a multicenter retrospective cohort between 2014 and 2024, cross-continental.

It involved…the initial data set involved almost 50,000 total cycles from over 31,000 patients. Ultimately, 801 qualified paired cycles were included in the study. Those 801 patients met the inclusion criteria, which included two ovarian stimulation cycles within six months of each other with the exact same protocol, the same gonadotropin type, same starting and daily dose.

They had to be freeze-only cycles, and each cycle had to have greater than one oocyte retrieved. And the outcomes were the percent difference in oocytes. So again, the primary outcome is the average percent change in oocyte yield, and then each of these outcomes was categorized in the paper into four different categories you can see there in red.

The secondary outcomes included mature oocyte variability. They also included something called an ideal foresight model, which I will explain more a little bit here, but basically attempts to estimate the upper bound of predictability for any model, and then they stratified it, obviously, by age. So again, the data set originally included almost 50,000 cycles.

There were almost 11,000 patients with over two ovarian stimulation cycles, but when you went through all that inclusion criteria, we boiled down to 801 pairs, meeting all the criteria. And you can see the demographics here, and most patients were used in antagonist cycle. The trigger type was the same in about 76 percent of the pairs, and the duration was a little bit different, but the mean duration was really only differed by a day or two.

Most patients were—the average age of the patient was almost 39 years old, where over half of the patients were over 40. You can see here the BMI was roughly 24, and just confirming our external validity here. So, the primary results, I think, are important to interpret here.

So, the big headline is that there's about a 62.7 percent mean change in oocyte yield. The thing you have to remember about mean change is that the small denominators are going to impact things more. So, if I originally had two oocytes and then I had three oocytes, that's an average change of one egg, but that's a 50 percent change, whereas if I went from 20 to 21, that's a 5 percent change.

So, I think using all of that information, the median is the better value to look at here, which is still a 40 percent variability difference between cohorts with the exact same protocol. Over 50 percent of patients had a greater than 33 percent difference between cycles. And then, again, remember, they put all those patients in different categories.

So, this chart here shows how patients change from one cycle to another in different categories. So, the blue boxes are patients that stay the same. Red means they went down a category, and the light blue means they went up in a category.

I think, again, one of the important outcomes here is that almost 50 percent of patients did shift a category, but it's bidirectional. So, you can see that a little over 20 to 25 percent of patients went up or down. So, both directions, patients moved.

But overall, patients…nobody went from an excellent cycle to a complete poor outcome. And I think what's important to understand in this paper is the ideal foresight model. So, basically, that was an attempt to confirm with a model what the variability could be and what was intrinsic to biology.

I think the ideal foresight model takes into account the mean from each cycle. So, if one patient had eight oocytes in one cycle and 12 in the next cycle, the error-minimizing protocol model would say that the patient should get 10 in each cycle, and the difference between 8 and 12 and 10 is, therefore, the underlying variability. The concern with that, though, is that when you include the averages in the model, you are, by default, inflating the model.

So, I would interpret it as kind of a high-end variability rather than a true mean. But overall, I think the outcomes of this paper are reassuring. Forty percent of patients…there was a 40 percent change…median inter-cycle change in oocyte yield for patients.

About 50 percent of patients shifted category, again, with no changes in protocol whatsoever. And their ideal foresight model predicts that about 17 percent of variances, that irreducible biologic noise. Great.

Thank you, Dr. Eubanks. So, we're going to start with a discussion with our experts, and then I would like to leave the last 10 or 15 minutes for audience questions. So, please put your thinking caps on.

Give us your insights. Give us your thoughts. Challenge these experts with your questions.

So, just to summarize, these are cycles where the first one, the doctor was happy with. Whether they had high yield, normal yield, low yield, it was within what the doctor expected because they did the same stem, the exact same type of downregulation, the exact same type of gonadotropin dosing. But, yeah, we saw a pretty big biologic variability.

They went from a huge number of cycles to get down to about 800 to find those pairs, which is why this was an international collaboration. It was still more variability than what I expected. So, let's spend a few minutes talking about strengths and weaknesses from a methodologic standpoint.

I'll open it up to our panel. Let's just start with strengths. What did you like from a methodologic or statistical standpoint about what this paper did? I'll open it up.

Rachel, go for it. Yeah. So, first of all, it took me a little bit of time to dig, to find out where this data came from.

So, thank you to Eric Foreman, who was the sleuth there. And it came from a large data set that was used by a company, correct? That's what we found at Determine. An AI company, yeah.

Yeah, an AI company. And from what the paper has described about this data set, it is very robust in that it comes from multiple countries, and they were able to really control for some of the very specific variables, type of gonadotropin, dose of gonadotropin, even down to the physician managing the cycle. So, I did feel reading the paper that there was a good pair for each of the pairs included in this study, that they were actually doing what they said that they were going to do, which was show you cycles that were actually the same in theory, but the outcome was different.

So, I think that was a big strength. Kate, Eric, go ahead. I mean, I agree with Rachel that the fact that they really were very rigorous in terms of defining the population well was the major strength of this study.

I also liked that they looked at multiple primary outcomes. Sometimes we, on the podcast and just in general, looking at papers, have a critique when continuous data are categorized, but they looked at the exposure of the change, so exposure being the change in number of oocytes, a number of different ways. And I really think that the fact that they analyzed the data rigorously with multiple models strengthened their conclusion that there's really a heck of a lot of change from cycle to cycle that can't be explained by the protocol, which was exactly the same.

Yeah, I don't have too much to add on. I mean, the strengths, I think. Yeah.

Large population needed to get this type of matched cohort, but I have some other thoughts we'll get to next. Yeah, let's move on to limitations and do it in reverse order. So, we'll start with you, Eric.

Yeah. So, I think it is interesting. I mean, I think we all talk to patients every day about cycle to cycle variability, both in quantity and quality, that there's limitations to what we could predict.

And I think when you summarized, Mike, you hit on what I kept thinking is I'm not sure that this answers the question of whether changing the protocol improves things. This tells us in a subset of patients that they do well because they didn't change the protocol. So, either they're like some just inertia, like sometimes you just put the same protocol in or patient shows up a couple months later, or like you said, in people who do well because you didn't feel like you needed to tweak it, there's some variability.

But that doesn't tell us if they had changed it, would the variability have been even more? So, that would kind of be the next step, which they might be able to call out from their data, like take similar types of patients where they increased the dose or they were expected to be in one category and they underperformed, they increased the dose, like did that change it more? So, I mean, I think this gets to some of the intrinsic biology that there is this inherent variability cycle to cycle, but that next clinical question, I'm not sure this means that we should never change protocols. Yeah, I completely agree with you. I think this study better answers that question of intrinsic variability that's innate to the patient.

Kate, limitations from your perspective. I think we still can't avoid bias in this population. And in fact, I think that this particular population is particularly prone to bias, even though they defined the population analyzed well, it's still a very small subset of a very large dataset.

And there may be characteristics, as we've touched upon already, that they had a good cycle relative to expectations the first time. They're also certain, while they chose same protocol type and same starting dose, they didn't control for every factor in the stimulation. For example, the trigger type may have changed, the rate of change of the dose of medications over time, the length of the stimulation may have changed.

And it would have been interesting to me to see did that affect, did those cluster along with the bidirectionality of the change in number of oocytes? That said, I think the flip side of the coin to everything I just said in terms of limitations is that it is very reassuring that it was bidirectional. Because if we thought that this bias were influencing the results tremendously, we probably would have seen the change more so in only one direction and probably down. Rachel.

Yeah, I agree with everything that's been said before. And I think that any retrospective study has inherent limitations. You don't know why a patient chose to do a second cycle.

You know, if the doctor was happy and didn't change the protocol, why did the patient choose to do another cycle? Was it because that they were not meeting their reproductive goals? Were these banking cycles, you know, that those patients might be different than patients who have infertility? I thought it was also interesting to note that patients at the older age of the spectrum and with decreased ovarian reserve tended to have more variability. So it's something to think about who those patients are and what's happening to them. One other thing that I noticed, which is in some ways reassuring, because it seems like it was bidirectional, but if about a third of the patient were using only urinary gonadotropins, I think about two thirds were either using or competent gonadotropins or a combination.

And obviously, there's, you know, quality control that goes into purifying urinary gonadotropins, but there also can be cycle to cycle variability in the potency of urinary gonadotropins. So that's just something that I was interested in to see. They didn't really break it down.

And there was probably, I would assume it was distributed, you know, symmetrically throughout the data set so that you would, you know, probably see that variability throughout, but I would have been interested to see that specifically called out. I think Allison touched on one of the statistical weaknesses of this. Oocyte yield is never normally distributed.

So if you're going to use mean, if you're going to use average, you're always going to distort the data, right? You can't have negative oocytes depending on your patient population. Your mean might be around 10 to 12, 15, maybe lower if it's poor responders. You're always going to have a longer right tail because you can't have negative oocytes.

And so you saw that here when they used mean change, which is what like shocked me when I first read the abstract, over 60%, that's huge. When you look at it with median, which is more appropriate, just that in and of itself, the way you look at population distribution, you get to something that's not quite as big. The other thing that Allison touched on is they chose to go with percent change instead of just the difference in yield.

So that means a poor responder is really contributing much greater. She gave the example, someone going from two to three eggs versus 20 to 21 eggs. They're each in and of one, they're each one in the denominator.

One's contributing 100% to the change, the other 5% to the change, but they're each getting one egg different, which clinically is the same change. So I would have liked to see a deeper dive into maybe egg yield change as the primary outcome rather than percentage change. Because I think they ended up overstating the effect.

So the more I've marinated on this paper, this is now the third journal club I've done on this paper. The more I've come to say, yeah, there is some variability. I think they maybe overstated it with how they chose to do their analysis.

I think the other thing that also inflates that number is that they did independent t-tests for these patients instead of paired. They're missing the fact that these cycles are paired. So if you don't correlate the cycles together, you're also missing that value, which is going to inflate that number as well.

So we've got a couple of factors that are going to inflate that number. Again, I think the direction is appropriate, but the actual number of 63% roughly is much too high. Another factor we know, I mean, egg yield is a surrogate marker.

I mean, ultimately, if they're doing a second cycle, either they fail the transfer, we don't have that, or they're banking embryos. But I know you mentioned in the podcast, which I listened to and also have marinated on this about, you know, how you were surprised that it seemed like their past performance wasn't the best predictor, and that's always what you thought. And I'm still not sure we could talk about whether we can conclude that.

I'm not sure that we can, but even that, you know, I think from a lab perspective, at least I still believe like performance in the lab is very predictive. So it would be interesting to see like fertilization rate, blastocyst development rate, and is there as much variability there as there is in absolute egg number? Because I think that might be more, you know, intrinsic to like less variability within the patient. But I'd be curious to see that.

And I'm sure that kind of data is going to come out from these types of analyses in the future. So, Eric, you just touched on the next question I was going to ask. You know, Kate Devine and I trained at the same place at the NIA Challenge Attorney.

Jim Seegers were our mentors, and they always taught us, you know, in your first cycle, it's going to be your age, your AFC, your AMH. Those are going to be your biggest predictors of yield. When you have a second cycle, your best predictor is going to be how you did in that first cycle.

I don't know that they addressed that question. I've actually asked these authors if they can run that analysis, and they're going to work on it. But for any of you on the panel, does this change that assumption? Do you think based on this paper, you still feel that way, or do you not feel that way having looked at these data? It's a good question.

I think it definitely did challenge my assumption to some extent because I was surprised by the extent of the variability. You always know they're going to find variability. I think we always talk to patients.

I tell them, your body didn't read the textbook. It doesn't know what it's supposed to be doing in response to these medications. But I have seen in my own practice that past prediction is the best predictor of future performance or past performance.

So, I was a little bit surprised by that. But I would also say is that this is within six months, but you also don't know what's happening in that patient's life within those six months. Is the patient starting a birth control pill? They said they weren't able to control on the study for esterase lead-in, birth control lead-in that we think might affect what's happening or what's happening in their life.

There's been interesting data. I don't know. It's so convincing about seasonal variation in human performance and what else might be happening in that person's life to influence their success.

So, I would say overall, I think that this study is humbling for me as a physician. And I think that although we just talked about some of the limitations of the counseling, I think I would use this study to tell a patient, I'm happy with how you did. We're going to do the same cycle again, but I'm going to strengthen my counseling that I can't guarantee that she's going to have that same outcome.

Yeah, I found it also along the same lines. Can we still rely on AMH, AFC? All of these things are called into question here, but it calls into question what's already a huge debate in our field, which is, is it appropriate to categorize patients a priori or after one cycle as DOR or POR? And looking at bologna and Poseidon through the lens of these data, it's a fresh eye, right? Because patients are affected by having these labels. And so are we, in a way, in terms of do we move them on to a totally different treatment algorithm to pigeonhole folks based on one cycle? I think this paper does a really nice job in discouraging.

And for me, there's a little bit, as much as it's humbling as a clinician, for sure, I think that it gives reason for hope for some patients who may think that they're just poor prognosis across the board. Can I just add one thing to that? Because I think that's a very important point. For anybody who was here yesterday for the growth hormone debate, raise your hand.

I thought that was a really interesting talk. And I think one of the big, there's a temptation as a physician to say, oh, we're going to add growth hormone, as I was mentioning yesterday, because in some DOR patients that has shown to be successful. So let's add it to a 35-year-old with poor oocyte yields and see if that helps.

Hopefully we all realize that's probably not the right thing to do after the talk yesterday. But I think that that goes along with what you're saying is that if you see somebody who has a bad response, the impulse is, oh, let's change something. Let's add something that might help.

Let's change something. And I think that what this paper is demonstrating is that you might have that variability no matter what you do. Now, we don't have the answer what actually happens if you do change, like Eric was mentioning, but I think that that impulse to change is very strong.

And I really like the fact that this paper says, okay, yeah, don't say exactly what's going to happen in the next cycle based on this performance. Maybe you need to see that patient in multiple cycles to really get a sense of what their response is. You will see that variability, right? Not just you might.

Right. You will. So I'm assuming we think this intrinsic as opposed to extrinsic variability from these cycle to cycle is innate to their ovarian reserve pool that becomes gonadotropin responsive that month.

I think we would all probably postulate that. They didn't really, unfortunately, they looked at their baseline AFC, not with each cycle, but, you know, in their screening AFC, maybe some of them had a higher AFC that second cycle and did better. Maybe they had a lower AFC that second cycle at that baseline scan for stem start.

Is this actionable? Do you think you should maybe like someone had 10 eggs last time and this time they come in and there's three antral follicle counts on their AFC? Would you maybe consider delaying that? Has anyone done that? Does anyone clinically do that? Should we consider that? Should study that? I think we should study that. I definitely don't think that we should do that. Patients ask about it all the time.

We have actually one site in our network where the physicians have decided very firmly that they will not tell the patient their AFC at the pre-med visit of their cycles because of how little predictive capacity they think it has for what the ultimate oocyte yield will be and the immense amount of anxiety and need for additional counseling that it causes. You know, I don't think we know the answer to that question, but, you know, as the reflection that you and George Patnakas wrote that I thought was excellent, there's the intrinsic variability. Then there's also all of the laboratory factors that you go into, the retrieval factors.

And, you know, when we were talking about antral follicle count at the baseline scan for an IVF cycle or an egg freezing cycle, which by the way, this is a mix, there's so much inter-observer variability on the part of the sonographer who is biased by knowing the age of that patient and so on and so forth. So I would say hard no on that one. And I think it's something worth studying because we do get asked that a lot.

And again, we have to spend time explaining that it has limited predictability there's no guarantee it's going to be better the next month. And time is, you know, in our field, the enemy. So that's how I counsel patients.

Like, you don't know, you could have a cyst next month, you could get sick before you know it. One month delay becomes three months delay. So, you know, very rarely, if ever, recommend delaying.

I do think it's interesting, though, to think about telling the patient the AFC. One of the things we talked about in this article before with Dr. Seegers was that message you're sending to the patient of, well, we didn't do as well as we hoped we would, so we're going to change things. And that underlying, you know, you're implying to the patient that, well, I could have done better the first time, but I'm going to do it this time.

And so making sure that we communicate, if we do change something, that we communicate in a way that doesn't suggest that I had this trick up my sleeve, I was saving it for the second time just to see how things went. So I do think that there's probably some underlying factor in this paper, too, that will never, you know, fully be assessed, but how much of the change in variability was from the confidence the patient received from, hey, we're going to keep you exactly the same, we did the best we could for you, we're going to do it again, versus, you know, that message that we were unintentionally sending the patient of, I've got some more tricks up my sleeve, but I didn't get them to you the first time, I can do better. Just a show of hands from the audience, does anyone delay a cycle to the next month if you see a low AFC in someone that's different from what you saw at their screening? Okay, not seeing any hands raised from our expert audience.

So should we incorporate, for those programs that have predictive models for oocyte yield and live birth, obviously in the first cycle we don't have this data, but should we incorporate a variable for what that response was in that first cycle into our predictive algorithms that are being developed by several companies, several networks? Any thoughts on that from the experts? Again, I just don't think we can answer it from this, because like we talked about earlier, you know, we don't know how much more or less variability there would be with changes, so I still, and this like we've heard is a small subset, less than 2% of the population of cycles, so I still think that the prior response has predictive value, just maybe not as much as we thought before, and we have to accept there's this variability. I think it would be interesting to see the time factor that we came up, I think that came up in your podcast as well, like we're doing more back-to-back cycles, there's this concept of priming. This went bi-directional though, so I mean, I think we sometimes, some of us have been counseling patients, there may be this priming, you might even do better just from sort of the prior stimulation, so this was interesting that it went both directions, but it'd be interesting to know if it was, you know, one or two months, is that more likely to go up versus six months? They didn't really tease that out.

I'll just say, as the past president of SART, one of the things I'm working on is a predictor model that'll be available for all patients and all SART member clinics. We're working with Enrique Shisterman and Sonny Mumford and the expert team at Penn, and we just reviewed some of the data yesterday, and actually in the second cycle, the second largest predictor of live birth is their response in the first cycle. The most one is age, obviously, age is still queen as always, so we'll need to delve into that data more to see if maybe we can answer that question with SART data, but at least the direction we're seeing initially is showing what we were taught maybe is still true, but maybe it's just not as big of a factor as what we initially thought.

Let us open this up to the audience. I'm sure we have a lot of experts here. Please, let's pass a mic around.

I see Seagal in the back has some comments or questions. I would love to hear what the audience thinks about these data. Okay, so I think this is a really fascinating study, and it's reassuring in the sense of counseling patients, although I do agree with the comment that, you know, we don't want to let patients feel like the first cycle was a trial cycle, and a lot of times patients are like, well, I know most people don't get pregnant with the first cycle, so you learn from this, right? That's obviously not right.

The converse conversation, though, is those patients who, you know, you do what you think is right, and you get four eggs, and you're disappointed, and so you increase the dose, and you still get four eggs, and so you change the protocol, four eggs. I mean, how many of you have had those cycles? You add growth hormone, you take it away, right? I think we've all had those cycles, right? And that's frustrating, too, and so then we spend a lot of time thinking about these cycles and perseverating and thinking we really know best what to trigger, what dose to use, you know, what protocol to use, and I just feel like we probably can do this better with machine learning, with AI, like, you know, as much as I really love my job, and I really like the process of, you know, running cycles and all this, I do think that there is a better way to do this with machine learning, and we have this amazing database with SART, right? And so I was just wondering what you guys think about this sort of this converse issue and the fact that maybe the issue is we just don't know, right? We somehow are not predicting this well, and these first patients, this very small percentage of first patients were happy with the first cycle, and so the physician did the same thing. Most of us change our minds after the first cycle, right? We don't switch, and that's what the data showed, because when patients want to do something different, and maybe we thought that it was the very best cycle, we really should do the same cycle, but, you know, for the psychology of it, we changed the cycle.

Maybe this justifies that we shouldn't, but again, we don't know very much about these patients. So I guess the question is, what do you think about this sort of the converse of this, where, like, you do everything, and the patient's just going to make however many eggs her ovarian reserve predicts, right? And two, what do you think about the human factor versus maybe an unbiased machine learning factor that might do this better? Well, I don't know if I can fully answer those questions. I'm very excited about the SART predictor.

I would love to see what that shows. Like you were saying, SART data really draws from, you know, national databases across the United States, and I think that gives us a lot of opportunity to research that question, especially if we combine it with AI modeling to see, you know, does AI do a better job predicting than, you know, we can? And I think that this study might actually say that maybe AI can't, because maybe there's a limit to prediction in terms of how human biology works that, you know, even AI might not be able to help. So I actually think that's a kind of an interesting question that the study raises, whether a prediction model actually can reach any better prediction than what we can do just based on looking at the patient's past performance.

Because ultimately, like you were saying, it is still a predictor. It may not be the absolute best predictor, but it's probably the best predictor that we have. The other thing I would say, Seagal, to your experience, is how many times does a patient say to you, well, I would like to get more eggs next cycle, so please increase the dose, right? And then, you know, how do you explain to the patient, I could increase the dose, but that's actually not a guarantee you're going to get more eggs, and you might get fewer eggs.

So I think that having these types of models, and I hope that the SART model will also maybe help to answer that question. Maybe this research group would want to answer that question, what happens if you increase the dose in the next cycle? Because I think that's very important for patient counseling, and also speaks to the limits of what we can do medically that, you know, there is not necessarily a one-to-one correlation in each patient between dose and response. And it's something we've been thinking about a lot recently, and how can we use the large datasets that we have at our disposal to answer that question of what is the ROI per unit of gonadotropin in a specific patient, right? Because there are some patients, I think most of us come to the conclusion, and that's been a theme in some of the talks over these past few days, some of the excellent talks, frankly, on ovarian stimulation, is that more eggs is better, all else equal, right? I mean, obviously, we're wanting to avoid OHSS risk, but thankfully, you know, stimulation has gotten a lot safer.

But we still don't know, and, you know, Alan Penzias had a show of hands, and it was across the board, right? If you've got a young patient with low reserve, do you start low, or do you start high? Because some of them, you max dose them, and they'll get more eggs, and some of them, you max dose them, and they're going to get four, four, four, four, just like you say. So, unfortunately, I don't think that machine learning and AI have been able to help us with this yet. I agree with you that it seems like it should improve over time, but never will it be perfect because of all of the unpredictable biologic factors that we can't put into the model, probably never will be able to.

And then there's the human component that you mentioned, right? That in order to encourage patients who ultimately likely will achieve their goal of parenthood, we need to be able to keep them in treatment. And so, how do we do that? There are some patients that won't be convinced by, nope, we should just do the same thing again. Eventually, your ship will come in, you know, you'll regress to the mean.

You know, it's not a particularly, you know, warm and fuzzy conversation to have. Yeah, I agree. I think that's, you know, another theme that staying engaged in treatment, some patients have the one and done, one really good cycle, and they have their whole family, but because of this variability, it's not that uncommon that people don't meet their goals.

And if you never do that second cycle, you won't see that group that bumped up a category or two. So, I think we can use this, you know, to hopefully encourage patients that they can do better. And again, I think, you know, choosing amongst, we don't know which protocol or what dose is better.

So, I don't think it's unreasonable to have that discussion. Like, I think you did well, and we can do the same thing. If you want to change, you know, the difference in outcome may not be due to the change, may just be due to the variability.

But for some patients, they need that change. What other questions do we have from our audience? So, I have a question and a comment. So, question in this study, because intrinsically, I mean, I tell patients routinely, you know, plus or minus two in terms of what the expectation is.

So, patients changed category, but that could have meant they went from four to six to seven to ten. I mean, it's a change. So, how much of a change was it? I apologize, I haven't read the paper.

So, that's one question. Is it still within that, you know, for most people, plus or minus two? And secondly, there were older papers, at least one, that looked at trying to find the right cycle to start in based on the baseline AFC and showed no difference. That if somebody's intrinsic capacity is X, I mean, in the old, old days, they did it with FSH, which clearly didn't matter because FSH is all over the map.

But you did commiserate with your highest FSH, even if your FSH was now lower. And then when AFCs came in, they did it with AFC, and it didn't really matter. So, I think hiding that number from your patients, hiding when they come in for their baseline scan is an appropriate thing to do, which is what I think Kate said they do.

I can put this chart back up here. I think the other thing to remember, too, is that over 50% of the patients were over 40 and had a oocyte yield of under nine. So, these are small numbers to begin with.

But this is, I think, the best way to visualize like how much change there actually was. So, the poor response category all the way to the high response category of over 15 in terms of the actual... But, Marcel, your point gets to what I was saying is I think one of the main limitations is they're using percentage change and not giving us the raw data. It's not in the paper.

I've tried to reverse calculate it based upon the means and what these categories are, and it's about four eggs based upon a rough estimation. But it may not be because we don't know. They may have just gone up one egg and moved categories, and we don't know.

To me, that's one of the main criticisms. I just wish they had done it by absolute number because as clinicians, that's what we care about. We don't care about that percent change, which biases it for being larger direction.

So, yeah, I 100% agree with your point. I wish it was in there. I just have a question.

So, my question is about the fact that so many patients were above the age of 40 and with poor ovarian response, and the paper did comment that the most variability was seen in that specific patient population. So, the question that I have is, why do we think that patients with DOR have such intrinsic variability, and does that mean that as, you see, I was intimating in your comment that no matter what we do, we're going to get few eggs in these patients, or do we think that if we study that population, changing protocol may actually have an impact on them? I think the larger variability in that group was completely a statistical construct because they did it by percent change. So, if they have few eggs and you get one different in either direction, statistically, it's a huge change, whereas in a high responder, statistically, it's a low change, but the actual absolute difference is the exact same.

So, I think it was a complete statistical construct, in my opinion, but we don't have the data to know that for sure because of how it's presented in this paper. I have a couple of comments, a good paper to talk about. I think this does not answer the question, just priming work, because it's up to six months, so they're not back-to-back.

They could kind of stratify them by the six-month, the five-month, the four-month. Maybe there's some data in there. Looking at antral follicles, I get fooled by antral follicles.

I say, oh, there's only four there, and all of a sudden, there's eight follicles growing. So, the super small antrals, I think we all miss. Not too accurate.

There used to be, a few years ago, a doctor on the other side of my state who would get an AMH every month and tailor it according to the flexibility and variability, and we all know the problems with the assay, so I think that's interesting. And the other thing is, you do want your patients to think that you did pay attention and make a tweak, right? So, I think we can kind of give the illusion of a tweak. You say, oh, I'll do a little change here.

When we know the inherent variability, month-to-month probably trumps that, but a little tweak here and there just for rapport and let them know that you're paying attention to them certainly doesn't hurt. And another thing I mentioned, if you go to a higher dose, sometimes you'll have fewer days of stim. So, you think you're going to get more eggs, but if you go from an 11-day stim to a 9-day stim, have you really gotten more eggs, better eggs? So, I'd like to mention that, but good paper.

Thank you. Thank you for those comments. We have time for one more audience question or comment before our experts have their closing words.

Hi there. I just had some thoughts as someone from the laboratory. I've noticed that some of the docs that I work with seem to get more eggs than others from one cycle to the next, and sometimes I hear them say that they're draining all the follicles, even the small ones.

So, let's say they get five eggs one cycle, the next cycle, a different doc performs it, drains everything, and gets even with the small follicles. They may have 10, but only five are mature. So, I don't know.

That's just something that I've seen and just wanted to kind of note and see what your thoughts were. I think, I mean, in this paper, Rachel and I were talking. They state that in these practices, the primary doctor does the retrievals, but, I mean, in the practices we work in and are familiar with, I think it's hard, and I don't think they actually verify that that was the case every time.

So, I mean, I think there are practices where you try, but I think that is another variable, and I think that has been looked at. There's some by regional, but at least here, it's in the same practice. So, I think within practices, I mean, if you're seeing that big of a difference, that's interesting, but I think most practices have similar approaches, like we drain everything, or we leave tiny follicles.

So, I think that probably corrects for it, but it's another one of the variables. Yeah, absolutely. We have one recently retired physician who does like flushing on everybody, even if it's 30 eggs, and so he might get more, but they might be more mature, and we have other people that are very fast and are going to go for the ones that are over 10 millimeters, and they might get fewer eggs, but they both get the same amount of M2s at the end of the day.

So, that's certainly variability. As Eric mentioned, they tried to account for that by hoping it was the same physician, but they couldn't verify that from the record. So, even within the same laboratory, the same practice, you may have variation in how people practice.

Very good comment. I certainly think that would increase the variability that they found rather than help tighten it up and be more precise. All right, let's give our experts the final word.

Eric, we'll start with you, and we'll work to the other side. Just your high-level take-home point for us from this paper. Yeah, I mean, I think that there's variables we can't control, and, you know, what we've been taught in our training, there's still a lot of ability to that.

The cycle-to-cycle variability and predictive value of prior performance, don't throw that out yet, but there's more to learn. I agree. I think that this paper really gives food for thought, and trying to be an optimist about it, I think it gives us reason to reassure our patients when they don't do well that that doesn't mean that they will never do well.

I also think that as much as we've talked about, from an emotional perspective, it's helpful to tweak the protocol or to tell the patients that we're changing that. I so wish that we didn't have to do that, and that we can someday get to a point where we can, you know, convince patients using the actual data. I think there's true benefit to standardization where standardization is appropriate.

I think it decreases errors. I think it decreases the overall cost of treatment and the efficiency of centers, and, you know, in a situation where not everyone can afford treatment, things like that at scale can go a long way. So I think as a field, we should try to be more disciplined about not doing unnecessary things.

Probably changing the protocol is less harmful than some other things, but it's something that I think we should aim towards. Actually, I have one other point that I think I just thought of is useful for counseling, is when you have a patient that may be traveling for the summer or getting new insurance in three months, I think that this is actually useful for that patient to say, like, am I compromising my chances? I mean, that came up during COVID with delays, and it was pretty reassuring, but I think, you know, you could say, like, I would have done this protocol now. You know, if you're getting new insurance in three months, that's what I would do.

If they underperform, you know, it doesn't mean it was a delay. There's that variability. I think I always tell people it's safe to take a vacation.

You don't have to put your life on pause a month or two here or there. It isn't going to ultimately change the outcome. Rachel? Yeah, I think that's very helpful for me, for my patients, to reassure them that just because we're going to do something different doesn't mean that the outcome is going to be different, and just because we don't do anything different doesn't mean that the outcome is not going to be different.

So I think that's very helpful for patients and reassuring. I would love to see this group do the follow-up study and find pairs where the dose was increased, where the protocol was changed. They have the data, clearly, the large data set.

They have the methodology. So that's my request. I would love to see that study, and then we can really have some good counseling that we can then share with our patients about whether dose change matters.

Allison, you spent a lot of time digging in. You have about a 45-slide deck here. We just did a couple of them.

Tell us, you spent a lot of time. What's your final take-home from this paper? Yeah, I think both my take-home from this paper, as well as just for the people in the crowd who are trying to really get good at dissecting a paper, is really pay attention to the statistics and the underlying statistics. Again, focusing on median versus mean.

Everybody in this room knows what the difference between median and mean is, and thinking about the impact of those two variables in different situations. Independent t-tests versus paired statistical calculations, right, and the impact that those can have on outcomes before you just take the face value of a paper, I think, is really important to understanding what this paper is communicating effectively and what it's not communicating effectively. And those are all concepts that everybody in this room can understand.

So don't be intimidated by the statistics and not be able to dig through the paper. But I think specifically for this paper, my takeaway is don't put somebody in the DOR category too early because 45% of patients who were initially assigned to that category in the poor category moved out of it. So I think for everybody's sake, let's ease up on that diagnosis a little bit.

Well, thank you for all of our listeners around the world for listening to all we do with FNS, the Fertility and Sterility Journal Clubs, the FNS on-air podcasts, all of our social media platforms. To all of the researchers, this is why you should publish in Fertility and Sterility. We'll get your article broad exposure globally and help raise the awareness of the important research that you are doing for us.

I want to thank MRSI for inviting us once again and partnering with FNS. At the start of this meeting, Dr. Beltzos challenged us to take one or two things home. I've heard at least 10 or so things that I'm thinking about that I can do differently in my practice.

And we heard an amazing talk from Dr. Cedars yesterday that was both inspiring and challenging for the future of our field. So for all the young residents and fellows, embryologists, nurses in the room, carry that torch forward and carry on these kind of research and this learning that you had here. Thank you, MRSI.

We appreciate it.