Holly, thank you so much. And it's a pleasure and honor to be here and to also be the first of three great speakers, friends and colleagues. And I will start off speaking on thyroid nodules and, on some level, a bit of a continuation as well on what I spoke on earlier this morning, broadly with regards to thyroid cancer. These are my disclosures. As we think about thyroid nodules in 2024, the theme for my next 20 minutes will be one that kind of looks back at our traditional history and acknowledges that it's been a bit suboptimal. And there's been changes over the last two decades. And the change has really been trying to advance nodule care through better risk assessment. And I'll walk through the different ways we do that, really a holistic approach with demographic, radiographic, cytologic, and molecular risk. And I'll certainly touch on molecular testing as well. Hopefully, you'll leave this portion of the symposium with a little better understanding of how to risk assess in your own clinic. So the starting point is kind of what has been out there for the last 50 years, the traditional approach to a thyroid nodule. And I always put here a thyroid nodule over one centimeter because, for the longest time, that was the standard. And it was very simple. Most of the time, we would do an ultrasound, check serum TSH. Rarely, we would find a functional nodule on the left bottom portion of the screen. But for the vast majority of patients, we would move down to the bottom right and pursue fine needle aspiration of every nodule over one centimeter. And while this has become the suboptimal treatment strategy, the first question is, how did this come about? And what are the data behind it that established this one centimeter standard initially? And I'll just show you some of the old data so we realize it wasn't pulled out of thin air. And the first piece of data here is just looking back in the mid-1990s when many of our former colleagues were actually just studying the original, creating the original staging systems for well-differentiated thyroid cancer. This is the data from Ohio State and Ernie Mazzaferri looking at their experience and the retrospective analysis of 1,300 patients as they were formulating their Ohio State staging system. What you'll note here is that low-risk disease was defined when the tumor was less than 1.5 centimeters. And that was important because there was essentially a 0% mortality risk associated with detecting and treating this small-risk disease. That really led then to trying to analyze a nodule before that size cut off, hence 1 centimeter. Let me give you another piece of data, this one more from the histologic correlate. But you can find the data here in cancer by Machens in 2005. That also looked at a large cohort of well-differentiated thyroid carcinoma. And if you look at the diagram, you'll see on the x-axis is tumor size. And then you'll see on the y-axis, they tried to associate the tumor side with various properties of malignant transformation. In this case, it was extra-thyroidal growth. And if you follow that line from the top right labeled PTC and you follow it down to the bottom left, you'll see that at about 10 millimeters or so, that's when the properties seem to start increasing. Those were the types of rationales that led to everyone recommending generally an approach of fighting delasperation when a nodule was 1 centimeter. But we kind of created a problem and we created a bit of a monster. And you heard me start to speak on it this morning, but I'll show you another example of it. When I first came into the business of wanting to evaluate thyroid nodules about 30 years ago, I started looking at the literature trying to identify what was the standard risk. If a patient came into your office for nodule evaluation, what was the risk that you would quote them that a nodule could be cancerous? And if you go far enough back, it was 5% to 10%. And generally, those data, that number was driven off of what you see here. This was a chart I put together in 2010. And all of these were single institution studies trying to answer that question. What was the proportion of nodules that would prove cancer out of everyone who presented for care? But the studies that I list here were performed with palpation guidance and palpation assessment of the neck. And then in the mid-1990s, you see in blue, those were single institution studies performed with ultrasound and ultrasound guidance for the FNA. And what we witnessed in the mid-1990s as we were aspirating every nodule over one centimeter was that the rate of malignancy in a given population moved from 5% to 10% up to 10% to 15%, clearly simply based on a detection bias that was ongoing. And you saw me show this morning that this was played out as well from US epidemiologic data in the SEER database, that in the mid-1990s, as we improved and increased the use of cross-sectional imaging, we saw a rapid increase in thyroid cancer detection. It was this well-differentiated disease that was local or regional that was often very small that we were detecting. But through detection and treatment of it, the most important finding was that we did not change mortality at all. And so much as I had suggested that the Guidelines Committee this morning is pulling back on how we think about treating thyroid cancer aggressively, in the nodule world, we also have the opportunity to improve this as well. We can do better. And we do better by being more individualistic, by being guided by better thyroid nodule risk assessment of what looks benign and what looks cancerous, and we de-escalate. So let's talk more specific to the nodule world. And in general, there are numerous risk factors that we've come to now delineate that have shown us how to assess the risk in a thyroid nodule of carcinoma. There are demographic risk assessment tools. There certainly is radiographic, there's laboratory data, and then more recently, molecular data. And we have tried to put this all together as we have run an integrated clinic now for about three decades, collecting every patient consecutively and all of their data. First, we put it together in cancer, Lila Yassa in 2007, and then more recently, Trevor Angel's publication, JCENM, looking at just over 20,000 consecutive nodules we had aspirated. And let me walk you through kind of the process, the findings that have come forth, some of which from our data, some from many other colleagues throughout the world. And some of this debunks kind of conceptions that were standards long ago, and others just have looked remotely at separate risk factors. One of the initial standards, if you will, 20, 30 years ago, was that even if you had biopsied a nodule and it proved benign, that growth of that nodule meant something bad. And so we began to look at that in detail, publishing this in the annals. What was the natural history of benign nodules? That has been followed up as well by Cosimo Durante's work in JAMA, published just seven years ago. And what you do find is that nodules, even when they're benign, will commonly grow a little bit. Some of them will be stable. Most will grow a little bit over time. But growth in and of itself does not mean anything necessarily harmful, so long as it's kind of at this standard rate. More recently, we have been able to understand, however, that differential growth rates can be useful, noting that most benign nodules grow about one millimeter a year. If you do see solid growth of over four to six millimeters in a year, there is a higher likelihood that you've seen a malignant process or perhaps a falsely benign cytology in the past. And so we do watch that. I mentioned this morning that multinodularity is certainly very common as a function of age. But the second process that we wanted to dig into early in our clinic's career was trying to better understand what it meant to have one nodule or multiple nodules in the gland. So there certainly was belief that a solitary nodule many years ago was somehow at a higher risk of being cancer than was a multinodular goiter. And many people would say that a multinodular gland was protective against malignancy. We were actually following up on initial work by Baroweta that you see at the bottom. Mary Fradies and myself put together our data, looking consecutively at all of the patients coming forth and asking the question on a per-patient basis, whether you had one nodule in the gland or whether you harbored two or more, a multinodular goiter. And the answer was no. The risk of cancer in the patient was exactly the same. What was notable in the patients with multiple nodules, however, was that if you only decided to aspirate the largest nodule, you would miss the cancer about half the time. Also, if you had three or more nodules each over one centimeter, you would find the cancer in only one of those nodules, again, about half the time. And this really was the tipping point to acknowledge the importance, in particular of ultrasound, to risk assess each nodule for its own individual cancer risk and then prioritize aspiration from there. Then the field began looking, though, at many other risk factors. And I think we now have much greater clarity as to how to think about risk factor analysis for a nodule. In particular, take patient age. It's now, I think, quite clear that this is a very important actual factor, that younger age, in a linear and continuous fashion, significantly increases the risk any nodule will prove cancerous. We showed it here in the data you see published, but then more recently Ari Wassner also extending it into the pediatric population. Having a nodule when you are young, in your teenage years or in your 20s, almost puts it at a one in four risk that nodule will prove cancerous. The paradox to age, however, is that while the risk is higher, a nodule will prove cancer when you're younger, the paradox is when you're older, there's a much higher risk that you will have a more advanced cancer, a high risk finding if indeed malignancy is detected. Some other standard demographic risk factors, how about patient sex? I think it's now well clarified that if you present as a male with a thyroid nodule, you run about a 1.5 to 1.8 increased risk that nodule will prove cancerous. When we think of other laboratory data, some of the early work by Megan Haymart and Jackie Jonkles has clearly shown us that serum TSH, even within the normal reference range, as TSH increases, the risk that nodule will prove cancerous seems to also increase. More recently, Natalie Moray and our group had looked at the question of background Hashimoto's disease and the concept of whether inflammatory processes in the gland somehow influenced whether that growing nodule was benign or malignant. And indeed, we found that it does, that if you have background Hashimoto's disease and you have a discrete nodule, that risk has gone up about 1.6 fold that you have a chance of cancer. Nodule size has been long a question about whether that influences the risk of cancer. And in fact, most of the data, I would argue, suggests that increasing size within the typical range in which a nodule sits, roughly one to four centimeters, does not seem to modify the risk that that nodule's cancer. It's somewhat independent. But the only exception to that is the large nodule. For most of us who've been in the business long enough, there was an old standard as well, that once a nodule was over four centimeters, you really should think about whether it should be removed surgically, even if it's benign. And why was that? What was going on? Well, sometimes with a large nodule, there's certainly a structural component, and that's a reason to consider removal. But the other issue that was noted is that there was this higher risk of falsely benign cytology. And what happens when nodules are over four centimeters is this increased risk of invasion, capsular vascular invasion that can't be identified with an aspiration. On the bottom of the slide, I just point to some novel risk factors. We recently looked, Theodora Papa in our group, at nodule shape as an independent risk factor for malignancy, finding that the closer a nodule is to a true sphere, the higher the risk of malignancy. Why would that be? That kind of fits with the angiogenesis hypothesis that the easiest way to obtain nutrients and oxygen to all growing cells within any given nodule would be a perfect sphere that affords it a better opportunity to grow and expand. But the biggest step forward in nodule risk assessment has clearly been these two entities. The first is our ability to sonographically risk assess, whether you're using the American Thyroid Association assessment tool or TYRADS, and then separately, the Bethesda Classification System for Cytology Assessment, first identified and created by Ed Sevis and Zubair Baloch Saeed Ali. As I've walked through all of these risk factors and our understanding of nodules, I also want to share you the story that I think equally important in this process is to understand the limitations of any given risk factor. While we understand that they're useful, the reality also is that no single variable is perfectly predictive of thyroid cancer or benign disease. It is the process of bringing these variables together that is most valuable for individualized risk assessment. And I'll just show you here on the left, as it relates to sonography, to ultrasound assessment of a nodule, take here the data by Perschietti, a blinded prospective analysis of 100 nodules, simply asking if I took the same exact image and I showed it to two separate radiologists, would they call out the same components the same both times? And the answer is no, not always. In fact, there's only kind of moderate agreement. That's very important because what one radiologist thinks is a microcalcification, another blinded expert could say that probably is colloid, which is a much lower risk scenario. And we've seen the same play out in the cytology world as well. This was a study we performed about a decade ago, Ed Sivas and Zubair Baloch were the two lead authors. And this was really looking at the comparability, the reproducibility, of the indeterminate Bethesda 3, 4, or 5 cytologic classifications. And essentially, if one expert cytologist called a nodule as Bethesda 3, and then you sent the exact same slide to a separate expert cytologist, as you can see here, it turns out only one in three times would the second expert call it Bethesda 3. Two out of three times, they would change the diagnosis. That simply is the limitation of what we're dealing with. But just for fun, I will share with you the one other trick we played. In that study, we saw the signal that this was quite shocking that two different experts would call something different two out of three times. So we realized it would be very interesting to send the exact same slide to the exact same cytologist just 30 days later and see if they could reproduce their own reading. And it turns out that they don't. But 20 to 40 percent of the time, they too will actually change their analysis of what Bethesda category it is. Well, with that, obviously, the most recent step in thyroid nodule risk assessment is molecular analysis. And it really started with the BRAF analysis on the left, the mutation 100 percent predictive of thyroid cancer responsible as a driver for over half of our well-differentiated carcinoma. But we've all been cautioned with this assumption that a mutation in a DNA-based gene or a fusion, the assumption always was that when you have a mutation, it must mean something bad. And that has proven simply not to be the case. And RAS was actually our insight into that. That RAS positivity is picked up in benign nodules about half the time. Same thing with the RET, PTC, Pax8, PPAR, Gamma translocations. That is what has made molecular risk assessment a little more tricky. I showed this slide this morning. There are numerous actual molecular targets that are looked at now and widely used as it relates to thyroid nodule assessment. I still believe, as I do for thyroid cancer prognosis, that RNA-based analysis is proving to be the most robust. And why is that? That is the case because you can quantitatively express all 22,000 expressed units in genes in the body to analyze pathways to understand the ability to iterate and improve constantly until you find the pattern. We have three available molecular tests currently in clinical use in the United States. And I certainly realize, country by country across the world, everyone's opportunities to use tests are different based on what's developed and what's supported and paid. But the three tests we have in the United States will be VeroCyte's Afirma, the ThyroSeq version 3, and a third product that goes by the trade name of Thigenics or Thiramir. All three of them now currently use RNA-based expression as critical components of their analysis. And while, again, I realize not every test is available everywhere, I will show you what I think are the leading tests and the leading data out of the United States. I think we have two tests that actually have met high-quality validation and are actually quite similar. I'll refer to the Afirma test here as the RNA-based gene sequencing classifier. And then ThyroSeq is the DNA version 3, as you see on the slide. I think what you will find and what we are seeing in the U.S. is that that third test that I had mentioned called Thigenics and Thiramir simply is not supported in part because of its poor validation and then the limited real-world use is not necessarily demonstrating consistency. But if you look for how molecular tests are playing out in the U.S., we now have many studies, and so I like to look here at the systematic reviews and meta-analyses. The data by Lee and Nasser, I think, are very compelling. And essentially what they show, as they've looked at all of these studies together, is that these two tests are both quite good. If you get a negative test, it actually is quite confident that that nodule is benign. But I will say there still are patterns that I think are apparent. If anyone was in this business 10, 15 years ago, you'll remember that the Afirma test was felt to be designed to be a rule-out test. It was there to maximize its negative predictive value. Many people felt that the ThyroSeq, looking for DNA-based mutations, was more of a rule-in test. I still think you do see the signal, and that's why I've outlined these in the circles, that you see a slightly higher NPV with Afirma, you see a slightly higher PPV with the ThyroSeq test. But again, I think two very well-validated and useful tests. While we're on the topic of molecular analysis, it's extremely important also to acknowledge the importance of understanding hurtle cell neoplasia and its molecular drivers. The important data now published about five years ago by Raj Gopal and Ian Ganley was quite informative, that the understanding suddenly is that hurtle cell carcinoma is just totally different. It's driven by two other molecular processes we don't see in well-differentiated carcinoma. The first is big losses of chromosomes, and the second, then, is a mutation pattern uniquely specific to mitochondrial DNA. Those will be very important to assess in any molecular diagnostic test that is applied. So I walk you through all these variables because I think this is where we're at right now. We had begun to put together this diagram first in JCENM back now in 2008, and then have refined it here on the right in a publication in the Lancet Diabetes and Endocrinology just two years ago. This concept of multimodal risk assessment has become, I think, the norm and also quite valuable. In 2024, for most patients, I think this is how we should look at a thyroid nodule, acknowledging that we have many different tools to risk assess, both initially, the demographic and radiologic variables, and then, importantly, the risk refinement with cytology, with molecular assessment, all for the concept here of optimal treatment strategies, and equally important, acknowledging that there is no perfect test. You have to integrate them together because of the uncertainties and limitations that each single variable has. But if you ask, are we making a difference, the answer, I think, is yes. These are our internal data that we published in 2019, and it's looking at this very important variable called surgical yield. The concept, of course, is if we're in the business of aspirating thyroid nodules because we want to detect cancer and then send it to surgery for appropriate treatment and removal, the hope would be that you're sending everyone to surgery who has cancer. So you would like the surgical yield to be 100%. We're far from that. We're probably hitting close to 50%, but this ability to utilize these risk factors has at least clearly shown that we're doing better, albeit we're also aspirating smaller nodules over the course of time. So I like to tell a story, and as I get to the end, this is certainly the story I've walked through over the last 30 years or so, which is acknowledging we all walked into the business understanding that ultrasound and TSH should lead us just to aspirate any nodule over one centimeter, but that's really too much, and we've really been over-diagnosing and not necessarily helping many patients. And in 2024, we should have a much different mindset, utilizing the extensive diagnostic tools we have, but acknowledging that there are steps along the way and considerations equally as important. I like to say in the current era, we are guided not only by what we can do, but equally as much by what we should do. So this is how I think about a modern mindset to thyroid nodule care in 2024. It still utilizes ultrasound and TSH. It definitely takes a much broader awareness that a singular size cutoff is not perfect, but we're using FNA for both cytology and molecular analysis more intelligently, often thinking in a proactive way, why are we doing it? And along the way, as you kind of have sensed, there are times when I pull back quite a bit, asking myself, should I limit intervention or at times increase active observation? Or how am I bringing together all of these risk factors together, acknowledging that some have limitations? When I'm using molecular testing, I'm only doing it if I think it will help, and I'm intelligently asking myself which one is the optimal one and test to use for the patient that I'm thinking about. More than just a diagnostic process, I'm thinking prognostically, and probably the biggest change that's going to hit all of us over the next decade will be the use of artificial intelligence. I'm not an expert. This is worthy of another talk down the road, but no doubt two facets of care will impact the thyroid world in the coming years. One is the use of AI to assist us with interpreting ultrasound, and the second, equally important, will be the ability of AI to assist with cytologic interpretation. These are areas commonly already being looked at in other illnesses as well. So with that, I will just bring together my summary with, again, appreciation for the chance to launch this symposium here this afternoon. As I think about management and evaluation of nodules in 2024, I would start by just saying nodules are increasingly common. They're largely influenced by our aging population and the expanded use also of cross-sectional imaging. The vast majority of nodules will be benign, and even when thyroid cancer is detected, please remember it's generally low risk. And so we want to use this multimodal risk assessment, all of the different tools that I walk through, to really individualize care to a degree that we never were able to do so before. And when we expand that risk assessment, I think we're becoming able to be a bit more conservative, to de-escalate, to become more deliberate with how we think about who needs a nodule biopsy or what to do with that information if we do. We're driven not only by better cancer risk assessment, but also the realization that unnecessary intervention leads to harm. So with that, I can imagine you're thinking like I am. There still remains a lot of progress needed and a lot of questions left to answer, but I really do appreciate your time and the chance to come here and speak to you today. Thank you so much.

thyroid nodules

molecular testing

risk assessment

artificial intelligence

personalized care

RNA-based tests