Stacey Missmer, ScD - Critical Issues in Endometriosis Incidence and Risk
Endometriosis Foundation of America
Medical Conference 2019
From Biomarkers to Precision Surgery
March 8-9, 2019 - Lenox Hill Hospital, NYC
I'm Stacey Missmer. I'm a reproductive epidemiologist, so I'm going to be starting our day talking a bit about some of the complexities and reinforcing some of the themes from yesterday, and to keep in mind as you're thinking through most of today, which will be focused on treatment and outcomes and the really taking to heart the diversity of patients with endometriosis and where we have to go in terms of improving treatment, particularly improving diagnosis and improving the lives and long-term health of girls and women with this disease. These are my academic and professional affiliations. None are a conflict for what I'm going to discuss today.
Perfect. Thank you. As we talked about yesterday, the conventional wisdom, and you'll see this quoted in pretty much every publication and every statement about endometriosis is that the prevalence is about 10% of women of reproductive age. There are some strong data to support this, but we really need to explore who actually is getting diagnosed and how we're including those numbers. Dr. Ferland introduced yesterday a bit of the nurses' health study. This is the largest longitudinal population-based data that we have on the prevalence and incidents of endometriosis. There are almost 120,000 U.S. nurses, so they're medical professionals with a high level of knowledge and accuracy in the information they report.
They've been followed now for 30 years, and we've retained more than 90% of those participants. They've been an incredibly altruistic and generous resource for scientific discovery. Among those women, when we enrolled them in 1989, they were 25 to 42 years of age, and at that point, the prevalence of endometriosis was 5%. Over the next 20 years of follow-up, now, the youngest woman was 45. The prevalence became 9%, and so if we think of prevalence or cumulative incidents, this is falling right around that 10% prevalence rate.
Now, interestingly, Germaine Buck Louis and the ENDO study, which was based at three sites across the U.S. it was an intramural NIH project, they included girls and women who had been surgically diagnosed surgical visualization with histologic confirmation, but then they also included this unique group where they had 131 women in the general population who underwent MRI who hadn't presented specifically to a specialist or a specific complaint. They actually observed through this imaging, 11% of the women were found to have endometriosis, which is actually much higher than we would expect because remember that the superficial peritoneal disease, the most prevalent presentation in the general population wouldn't have been captured.
This is starting to give a suggestion that actually the prevalence may be higher. I know a lot of people here believe that it certainly is. Now, if we accept that 10% and we use the international census data for world population, that means we're at 176 million women with endometriosis, so not in any way trivial. I'm just going to take a moment again to pound home this point of funding given the prevalence and impact of this disease. As we reviewed a bit, the funding from the NIH has been on average about nine million a year. This slide shows through 2016. Unfortunately, in 2017, the funding went down to $7 million, so we're not getting better. We're staying pretty static here.
Yesterday, Dr. Bime compared these numbers to diabetes. Dr. Griffith compared them to erectile dysfunction. I'm now going to show you with the blue bars being the prevalence, and the red bars being the funding on the right, you can see a comparison with Crohn's disease, which has less than 1% prevalence and yet received $65 million on average and with COPD, which is again less prevalence endometriosis but received 10 times the funding. Totally unacceptable. There are lots of reasons driving this, but we cannot be hesitant to fight back and be very public and impressing upon our country about how really egregious this imbalance is.
Yesterday, David asked, "Are we thinking of endometriosis now? Should we be thinking about it as different diseases or subtypes of lung disease?" I'm going to present a bunch of data suggesting that I think most of us are thinking of informative subtypes, but we really need to step back and think when we say endometriosis, when we think of animal models of endometriosis, when we're looking at cells from women with endometriosis and comparing them to controls without endometriosis, what are we talking about? We have had a revolution in discovery in the last 15 years in the cancer world. This is also becoming true in other diseases.
Dr. Gregersen mentioned yesterday that now we know that there are four informative subtypes of rheumatoid arthritis, but in the cancer world, subtyping is everything, it revolutionized how we approach every aspect of these diseases. Breast cancer is a very straightforward example. We know for breast cancer, and this is even much more simplified over the details that we know, but we know that depending on the molecular components of the tumors themselves, there are very, very different subtypes, different subtypes in terms of the risk factors for the disease, very different subtypes in terms of which treatments will be successful, very different subtypes in terms of prognosis.
Now, everything, standard of care in a very short amount of time really is completely surrounded around assessing and defining these subtypes and then applying both the current treatments, but also this has revolutionized how we approached discovery. You cannot make advancement in breast cancer and most other cancers now without being very conscious of what subtypes we're discussing and how to advance our even further understanding of subtyping. This is obvious for endometriosis. We have much more diversity and heterogeneity in endometriosis. We know that the phenotypic appearance for ...
A skilled pathologist may be able to have a sense of some of the diversity when they receive a tissue sample from cancer, but from endometriosis, we could pull anyone in off the street and show them pictures across the various patients, and they would believe that there was a difference. We currently lump these all together. Now, we know that there's the ASRM staging system. That system, again as was mentioned yesterday and is important to reinforce, does not correlate with symptoms. It does not correlate with treatment response. It is informative for the bulk of disease and complexity of surgery, so it's certainly informative for those early treatment aspects, but in terms of discovery and subtyping patients, it really is entirely inadequate.
We also know that there's this huge symptom diversity that a large proportion of the patients present with pain, and depending on the type of clinician you are will determine what types of patients you typically will see, whether you're focusing on infertility or focusing on pelvic pain or other aspects, whether they go through GI specialty before they get to the endometriosis specialist. This all has huge variation and has to be both biologically and prognostically informative. We do know that the prevalence of endometriosis varies by indication for laparoscopy, so in this summary of the majority of the prevalence data, it was about 25% in those presenting with pelvic pain, about 20% for infertility, and about 4% for tubal sterilization.
We also know that in the tubal sterilization, in that surgery and process, they're probably missing endometriosis because they're not overtly looking for it, but still, there's diversity here. We also know that if we look at those who have been diagnosed with endometriosis and thinking the reverse, what is the prevalence of different symptoms? Krina Zondervan led the global study of women's health that includes about 1500 women who were scheduled for their first laparoscopic visualization at which they were anticipated to have endometriosis observed. Among that 57 % of the patients were observed to have endometriosis, and for those with and without endometriosis, those with the disease, about two-thirds presented with pelvic pain and about one third had presented with infertility as their primary symptom.
These are almost identical to the distributions we see across the nurses' health study. In our endometriosis patients, about two-thirds are presenting with pain and about one third have been diagnosed due to infertility. Similar to the cancer data, we do see some diversity of risk factor profiles based just on this very macro dichotomy of the symptoms or indications for endometriosis. When we look at cigarette smoking, we actually see a lower risk of endometriosis, and those presenting with pain that may be too hypoestrogenic. Well, you could be influenced by a lot of factors.
When we look at those presenting with infertility, we actually see a higher risk of endometriosis. Keeping in mind that you have to be very conscious of who you're comparing to whom. Dr. Ferland presented yesterday, we do see differences by infertility and pain presenting status in terms of pregnancy outcomes, and that may be because, and the majority of the associations were stronger in those who are pain presenting, and that may be that because in the infertile group, if you're comparing infertile with endometriosis to infertile without endometriosis, we actually might not be seeing an increased risk independent of the infertility. Who you compare to whom matters.
This is very important for a lot of the factors that we look at, where for a diagnostic, you absolutely want to be comparing those who have the indications for an endometriosis evaluation, and those who ultimately are found to have endometriosis versus those who aren't, in which case you're often comparing those with endometriosis to non-endometriosis chronic pelvic pain, those with endometriosis to those with uterine fibroids, but when the pathology, the pathophysiology overlaps, those comparisons are very poor for discovering actual etiologic factors and risk factors because for example, the genetic profile is emerging for endometriosis and for uterine fibroids, and there are overlaps.
If you compare someone with endometriosis to someone with uterine fibroids, they may have a similar WNT4 polymorphism for example, and it will look like WNT4 is not specifically associated with endometriosis, but that's the incorrect conclusion. It's because they actually were associated with both groups, so we have to be very careful about when we're comparing endometriosis to other physiologic disorders and endometriosis to healthy women who are completely asymptomatic. We also see this for many of the diet associations, so this is the healthy eating index, which is a score across all dietary patterns. The higher the score, the healthier the diet.
We observed a lower risk of endometriosis with this healthier diet but only in those significantly who are pain presenting. That may be because the diet is, in fact, influencing the endometriosis, or it may actually be that the diet is influencing the symptoms and therefore the women have lower symptoms, less impactful symptoms, less severe symptoms and never make it to that surgical evaluation to be counted as an endometriosis case. We also have heard a lot yesterday through the really wonderful afternoon discussions around all the opportunities and all that's happening across the dynamic discovery with the new and more state of the art Omics platforms across many, many different biologic pathways and milieus.
There's a huge opportunity for discovery here. As Dr. Bime reminded us yesterday, we really need to be approaching this not as one individual pathway but the synergy across the pathways and hopefully, some of the technology that Dr. Griffith and others are developing will help to move that forward in an agnostic discovery fashion. We've done very little so far in terms of the molecular and systematic phenotypes of the ectopic lesion tissue itself. A little bit more has been done in the eutopic endometrium tissue and comparing and contrasting, but when you think of where we are for cancer pathology and subtyping and diagnoses, we have these absolutely identical opportunities in endometriosis that have not yet moved forward in part because of funding issues, in part because of lesion size and diversity and those types of things, but it's definitely something that all of the technology that we need is right there waiting for us to capitalize upon it.
When we think about these dynamic factors, the Omics, epigenomics, transcriptomics, that change over time and are influenced by age, by external factors, as Dr. Burns presented, by exogenous environmental factors, by lifestyle and changes, those things. We really have to think carefully about what snapshot we're taking when we're collecting those samples and defining those parameters. As we've also discussed a very small bit so far at this conference is that 99% of what's been published in endometriosis is focused on the adult, the samples, the symptoms, the collection, the journey of those who have been successfully diagnosed and successfully diagnosed usually years after their symptom onset.
We're getting that window in time. That window in time may not be reflected of the early etiologic factors. It may also not be reflective of the most discernible differentiating diagnostic factors. That's something that we definitely have to take into account, that we may be erroneously missing some significant discoveries because of the time window where we're currently able to collect information. Not news here, already mentioned this morning that we all know that for most patients their symptoms emerge much sooner than when they're actually being diagnosed when samples are being collected.
Again in that global study of women's health from 16 centers across 14 countries, about one-fifth of the women who now had finally received endometriosis surgically visualize diagnosis had their symptoms before the age of 18, and two-thirds of them had their symptoms before the age of 30. We're really missing these early etiologic and the potential, the key potential for early diagnosis and intervention. When we're talking about diagnosis and intervention, and the natural history of disease, we're looking in this window when we're talking about etiology, when we're talking about potential prevention, when we're talking about origins of the disease, we really actually have to be looking sooner than the emergence of symptoms, and thinking about opportunities for longitudinal cohorts, thinking about opportunities for sample collection and existence of resources to start to tease out what is the absolute norm in cancer and cardiovascular disease discovery.
The in utero and early life exposure pathways and points of intervention are very well established and very well researched and is really absent in our field thus far. Again, when we're thinking these dynamic factors, all of those issues come into play, but when we think of the genome, the genome is static, or you're taking genetic measurements, it doesn't matter when you take them. As an epidemiologist who thinks a lot about who we're comparing to whom, confounders, temporal associations, mediators, none of that really matters for genetics. Who you're sampling from matters, but in terms of when you're taking the measurement and how to interpret that, that all becomes quite static.
However, as we're thinking about these informative subtypes, this is much more complex. We're really thrilled that we are now in the fourth wave of meta-analysis from the International Endometriosis Genome Consortium. The original paper in 2010 published with Krina Zondervan and Grant Montgomery and myself included just three populations. It included a study based in Australia. It included Wellcome Trust data based in the UK and included the nurses' health study. We now are in early preliminary analyses with data from 22 sites including the Somatics 23andMe data, and now have more than 60,000 cases and almost three-quarters of a million of controls, so a phenomenally large data set now starting to rival some of the standard cancer discoveries.
However, as we're seeing the initial results, some of which will be presented at SRI this coming week for looking at endometriosis as a whole, treating it as one entity, we've tripled the number of genome-wide significant low PSI that has been observed, so that's very exciting, but again, if we're trying to think about pushing the field forward for informative subtypes, when we start to look at the subtypes, we already have some data that there are differences and looking very again at this macro level of just discerning dichotomizing on the SRM staging between stage one-two disease versus stage three-four disease.
We see that the low PSI that has been observed so far, there has been 15 robust low PSI really are most strongly associated and some of them only associated with stage three-four disease. What we have to think about is that that staging system is not informative in terms of symptoms, in terms of prognosis. What is it that could be driving this? Could this be that we're finding low PSI that is strongly associated with endometrioma? Could this be that we're finding low PSI strongly associated with the propensity for scoring and adhesion and risk for a frozen pelvis in that progression?
We really have to discern what that means in terms of actual biologic translation, and as we're starting to do this, despite having almost 60,000 cases of data, when we try to pull the information, even just with the ASRM staging or with infertility history, yes or no, we find that the numbers suddenly become very, very small because very few groups and very few databases and data sets and very few medical records even include this very simplistic level of detail for girls and women with endometriosis, and that's a big problem as we're thinking about the potential for big data and broader discovery.
Big data isn't just having these larger numbers. It also matters what types of data you have, so some evidence that we're struggling in this field when we looked at the nurses' health study and we had these very accurately diagnosed cases, we thought we'd have great success in going back to all their medical records and pulling out these details that really would only be in the surgical and the clinical reports. We intended to pull more than 4,000 records. We pulled the first thousand records from across the country and found that any detail other than endometriosis, yes-no, was only present in less than 10% of the records.
Of those that it was present, there was so little overlap in how that information was recorded and what it was recorded that it really became a waste of time and funds to pursue this form of data collection. Even in a clinical setting, we have the same struggles. So many of you, you all being very endometriosis-focus, skilled, elite clinicians, you have a great deal of detail that you collect data on. I will argue that having looked at much of the data across the country, that you're all collecting different things, but at least the people in this room are collecting detailed information but most actually are not.
In a clinical hospital setting, we wanted to look at endometriosis sub-phenotyping just again at this macro level of superficial peritoneal disease, deep endometriosis, endometrioma yes or no. We reviewed more than 17,000 IVF and ICSI cycle data from 8,000 women and found that in the end, we only had sufficient detail in those with endometriosis for just over 400 cycles and less than 200 women out of all of those. This ended up being a sample size out of this subset. There was only 2% of those in this infertility cohort. Clearly biased and uninformative and mates made it impossible to move forward with specific discovery.
For endometriosis, we don't just want yes-no. We don't just want the ASRM staging. We want to have lots of details, details like they have for most of cancer and cardiovascular discovery. We need this detail from many women. It rarely exists in biobanks and electronic medical records for endometriosis, even less so for chronic pelvic pain, and it does exist in the large longterm cancer cohorts, which just to remind everyone, this isn't an academic exercise. When we look at the predictive genetics of breast cancer, all of these factors end up influencing which genes are associated and informing us more definitively about which genes are driving the different subtypes of cancer.
These factors are not just part of a patient report that's nice to have. They're actually biologically informative, but the cancer cohorts, by and large, do not have endometriosis data, and so, again, we've gone back for this large DUS study to pulling in all of the data that exists across the world from these cancer cohorts with that very small proportion that has these details. This is why we launched the WHERF EPHect project trying to harmonize and collects data uniformly. We're now in year five of EPHect just to remind everyone that this provides and the Endometriosis Foundation of American, both Dr. Sectin and Dr. Gregersen both contributed to this process.
It suggests a standard data collection from patients, a standard recording for surgical findings, and then standard processing and storage for samples so that when we're combining samples from different studies and different sites, we know that those protocols are similar. It was really eyeopening to see among the experts across the globe, there were more than 52 experts who joined in this process and how different their quote standard protocols for tissue and fluid collection and storage actually was. Everyone thought they were doing in the same way and basically, no one was. When we have trouble replicating and finding robust findings, some of this is because of differences in these study design and bench science factors that actually are probably missing really important findings.
We now have more than 24 centers across the globe including the ROSE study, so there's a huge body of biobanks and data now available for anyone who is thinking of hypotheses and want to take them out of their immediate clinic sample or immediate laboratory sample to a broader sample size to replicate and validate their findings. Now, I really don't want to conclude without reinforcing who we're missing when we're looking at endometriosis. Again, going back to this 10% prevalence, so everything we know about endometriosis, no matter what type of science, what type of clinical care, 100% of what we know is only among those who are successfully diagnosed.
The typical path for someone presenting with pain would be a pain to a primary care physician, referral to a specialist and ultimately to a surgical diagnosis, but we know that many, many women who experience pelvic pain don't ever report it to any medical professional. We also know that many, many and the EFA has been really phenomenal in highlighting this particular circumstance where we know that many of these patients, part of the diagnostic delay is because they have talked to their primary care physician, to their pediatrician, to their family doctor and are told that what they're experiencing is normal.
Often, they also are prescribed first line OCs, and may not have remediation of symptoms or may have remediation of symptoms, but there's no further follow up. Those patients will never be surgically visualized and diagnosed with endometriosis. Whether they don't have pain remediation or they have success, the key here is we're only capturing those who move on for surgical visualization. We also know, and all of you see this in your practices, the patients often go to the wrong specialist. They're diagnosed with irritable bowel. They're diagnosed with many other conditions before getting to the accurate diagnosis of endometriosis.
The key here is we really have no idea how large this population is. We also really have no idea if the characteristics, if the symptoms, if the potential treatment response, if the genomics if the Omics differ in this group from the group that we're actually capturing. This is also important for infertility presenting. I think that for those who specialize in this field, it can be lost that the proportion of women who actually get to an infertility specialist for care is really small. From US estimates, there are about seven million women who have a longer time to pregnancy than what would be expected, about 2 million who meet the criteria of 12 months of attempting to conceive without success.
Of those, half, 50% ultimately mention anything to a healthcare provider. Among those, only 10% and depending on where you are in the country, even less, actually have a full infertility workup and have an IVF evaluation. Dr. Ferland led a study in the nurses who again are medical professionals. In that cohort, about 60% who meet the criteria for infertility actually move forward and present, which is higher than the general population, still only 60% out of all those who meet the criteria, but about 90%, the care that they received was ovarian stimulation. They don't get to access IVF.
Also from that cohort, even among these medical professionals, we know that the women who do get to an infertility evaluation are healthier, wealthier, and more likely to live in urban settings, which influences access to care. It also influences environmental exposures. Here is the map of ART clinics across the United States. As you can see, within most of the country, there is no opportunity for access to care. I would like you to also think about this map with the lens of who can access endometriosis surgeons, certainly skilled endometriosis surgeons who are able to do thorough excision surgeries.
Bringing back to the breast cancer example, the important thing for endometriosis is we don't know the distribution of these subtypes. For breast cancer, back when we were lumping everything together, about 60% of breast cancers are ER-positive breast cancers. That defined all of what we knew about risk factor profiles that were swamping the data. We now know that the other subtypes have very different profiles. We don't know what this distribution is for endometriosis. For endometriosis, the future is actually extremely bright. In just the 15 years since the discovery of triple negative breast cancer, the breast cancer field and survival and discovery has completely revolutionized.
There's no question that this will be true for endometriosis. We see much more heterogeneity and diversity here to explore. It will happen with events like this, with knowledge, with increased collaboration, larger sample sizes and an agnostic humble approach to complex biology. We know it's a prevalent disease. We know that the number of undiagnosed women is completely opaque to us at the moment. We don't know how this diagnostic bias may be influencing what we believe about endometriosis now. It's a critical window. We don't know what those timeframes are. We know that there are risk factor differences and we have to discover those as well, and, again, a large collaboration that's ongoing now and is poised to be even larger will be successful.
I always like to conclude also, I know the patient day is tomorrow, but in terms of patients and those who care about women and girls with endometriosis, what can we be doing? We have to be consistently loudly reinforcing that distressing life impacting pelvic pain is not normal, but that discussing female reproductive health is normal. If we can have erectile dysfunction commercials and ads everywhere, we certainly can discuss women's reproductive health. We have to demand greater funding for research in general, but specifically for reproductive health, and also a focus on science and math and the complexity of embracing what we need for this discovery.
We also need everyone to be contributing to research, whether it's what information you're recording and recognizing that each of your patients, each of the girls and women are contributing unique characteristics and experiences that may help us to inform where the field needs to go in the future. I'm going to conclude with reminding everyone that the next World Congress is only a short bit away. It's 13 months away. Three years goes very quickly, and I know there have been great contributions from the EFA there. Remind everyone that you can look at the WHERF EPHect website and see some of the great work we're doing there and download all of the tools. I thank you.