Molecular Testing of Thyroid Nodules — Andrew Gianoukakis, MD
Thank you, Peshimam and Dan for the invitation and for this introduction and thanks to all of you for staying till this late hour and we really are out of time So I will stick with the theme of being out of time as I have no choice I have no disclosures This is what I wanted to tell you We were going to talk about the dilemmas, the thyroid nodule dilemma, the thyroid cancer dilemma, discuss the types of well differentiated thyroid cancer and the pathogenesis of thyroid cancer Talk about the diagnostic utility of molecular markers, primarily the gene expression classifier As well as gene mutations and rearrangements that have been making their way to the clinical arena Talk about prognostic utility and close So quickly The thyroid nodule dilemma, there are too many of them as shown in this slide from Dr Mazzaferri from this paper in the early 90s and subsequently shown again in multiple studies Thyroid nodules are very common with approximately 50% of 50 year olds having a thyroid nodule If we look by sensitive means such as an ultrasound And oftentimes the thyroid nodules can come like this as multiple nodules in a gland making it very difficult to decide which to go ahead and biopsy or what to do to distinguish which of these are benign, which are malignant and what to do about them And for the most part 90 to 95% of the pool is benign And so while a fine needle aspiration, sticking a needle in these nodules and using our cytopathology colleagues to help us with the diagnosis is great It’s invasive and ideally we would want to have a noninvasive tool to try to help us with which of these nodules to put a needle in and that would be great for both costs and patient comfort And the ATA has in their upcoming guidelines recommended patterns of thyroid ultrasound, which we should rely on to help us decide initially which lesions we should stick a needle in with high suspicious lesions being most likely to be biopsied And less suspicious lesions to be left alone and potentially monitored But even after we take a nodule that looks suspicious, such as this one with irregular borders and microcalcifications and hypoechogenicity and biopsy it compared to a nodule, a spongiform nodule such as this one that’s thought to be benign And send the cytology off to our pathologist cytopathologist and come up with a diagnosis We do pretty well, but nowhere near perfect When the diagnosis is malignancy, we get it right very, very high percentage of the time This is one of multiple large series where patients who have received an FNA and then come to surgery for whatever reason, the FNA is compared to the surgical followup And so when the call is malignancy, we get it right very often, but there’s this large group of indeterminate lesions where we’re not sure this indeterminate category, which includes suspicious lesions, lesions suspicious for follicular neoplasm or consistent with follicular neoplasm and atypical lesions where a large majority of these patients go to surgery, but only a small minority or a fraction of them actually have malignancy And also we have unsatisfactory results which require re biopsy and even the benign is not a 100% benign So the question is, can we do better? And then we also have the thyroid cancer dilemma The prognostic question, when we diagnose thyroid cancer, is it a bad cancer? Does everybody need surgery, radioactive iodine and thyroid hormone suppression? Or can we do a better job of selecting the nodules and the cancers that are likely to go bad? By and large, we know that the prognosis for patients with thyroid cancer by and large is quite good and it’s proven difficult to select who are going to be the bad players
So can we do better? In terms of the types of thyroid cancer We have your classic papillary thyroid cancer and your follicular thyroid cancer as shown here breaking through the capsule where the cytopathologic architecture drives the diagnosis And we know that both of these well differentiated tumors can de differentiate into more poorly differentiated and then transition even into anaplastic type of lesions And in general, the pathogenesis of thyroid cancer is thought to include genetic factors as well as environmental factors, which coupled together to begin the process of neoplasia And the genetic factors have been broken down into more initiating events and then later mutations which drive the poor differentiation and the aggressiveness of the tumors And looking at it a little bit more schematically, we have a follicular cell which will undergo mutations Ones that are more associated with development of a papillary thyroid cancer as we’ll talk in a moment, or your RET rearrangements and your BRAF mutation versus RAS mutations in the PPAR gamma rearrangement, which will drive the follicular cell more towards that of a follicular adenoma and carcinoma with secondary progressive factors being added That ultimately lead to further de differentiation and aggressiveness of the tumor So when talking about well differentiated thyroid cancer, we have point mutations and gene rearrangements The point mutations primarily of BRAF and RAS and genery arrangements that include the RET PTC gene rearrangement and the PAX8 PPAR gamma These mutations are thought to activate to intracellular signaling pathways that are very important and those include the map kinase and AKT pathway and these mutations are thought to be mutually exclusive of each other Looking at it schematically, we have commonly a growth receptor, a tyrosine kinase growth receptor on the surface of the cell where mutations can occur either at the intracellular kinase domain of the receptor or in either of the MEK-ERK or AKT pathways where you have these mutations that we just mentioned that lead to a turning on A constituent of activation of the pathways that lead to cellular signaling, nuclear signaling, which results in cell proliferation that’s uncontrolled I will skip over the details other than to say that the BRAF mutation is the most common alteration in PTC and it’s specific to PTC And the other rearrangement The RET PTC that’s associated with papillary thyroid cancer There are multiple different subtypes, type one and type two represent 90% of your RET PTCs and they’re more commonly found in patients with radiation exposure and childhood PTC For your follicular cancers, the mutations of RAS, which could be N, K or HRAS Again, we have constituent of activation of the pathway and this is most commonly found in follicular but also papillary thyroid cancer and is commonly found in the follicular variant of PTC But also in follicular adenomas and the PAX8 PPAR gamma rearrangement is a rearrangement that leads to cell transformation by as yet unclear mechanism but has been associated primarily with follicular thyroid cancer and follicular adenomas Then we have our secondary mutations that are common to multiple different tumor types and they can include mutations in your tumor suppressor genes such as P10 and P53 or other genes that are important to cell differentiation and replications such as the beta-catenin in the wind pathway and PIK3CA in the AKT pathway So what is the prevalence of these mutations in thyroid cancer?
Well, that has been a progressive target, a moving target over the last 20 years We started in the early 90s knowing and being able to identify a mutation in only about 25% of patients with thyroid cancer And by 2000 we are up to 35% When BRAF was discovered by 2005 we were up to 70% and today in 2014 we can identify a mutation in over 90% of thyroid cancers So then the question becomes how can we use these markers to help us with the dilemmas that we mentioned earlier, the thyroid nodule dilemma, the thyroid cancer dilemma And while initially when testing and evaluations occurred, all the diagnostic categories of FNA were being evaluated The focus fell mostly ultimately on the indeterminate category that we mentioned earlier, which is utilized 15 to 30% of the time with most patients into surgery But the histologic malignancy rate could be as low as 10% And so the first large study to look at this was the study by Yuri Nikiforov where they took a large number of indeterminate cytology samples and they looked at the histology compared to the mutations And looking at this seven mutation panel back then and back then we’re talking about 2011 What they found was that if they took this initial diagnostic category and this indeterminate category, either the atypical lesions, the follicular plastic lesions or the suspicious for malignancy When they looked at only pathology, their risk of malignancy was as shown here, 14%, 27%, 54% If they added this panel of testing If you are positive for any one of these seven mutations that increased your cancer risk to 88%, 87% and 95% And therefore was found to be complimentary to FNA So if you added the molecular testing to the FNA, there was a high specificity and positive predictive value So you could be more sure about sending your patients to surgery However, if the testing was negative, you still in the AUS/FLUS category, you got down to close to 5% which is the gold standard, low end of acceptability that’s provided by a benign FNA But for these other categories the risk was still quite high to make any recommendations other than maybe to go for a lobectomy as opposed to a total thyroidectomy So complimentary to FNA and a rule and test with a high positive predictive value and a high specificity At the same time, a gene expression classifier was being developed where a group of approximately 150 genes and their expression was being analyzed and found to be predictive of predicting a benign lesion And so this classifier was published in 2012 with a high negative predictive value on the order of 95 and 94% and a high sensitivity So the way this was supposed to work out is that we would take this indeterminate category, perform the classifier, and if the classifier was negative It would decrease the risk because of the 95% negative predictive value down to 5% and you could comfortably watch these patients and not send them to surgery If this classifier was suspicious, there was approximately a 40% risk of malignancy, but all these patients were going to surgery anyway So this was found to also be complimentary to FNA And this was a rule out test, a test with negative predictive value And so algorithms began to be developed and published in the literature about taking this indeterminate group and performing either one or both of these tests in series And that became quite complicated You need a separate sample Most of the time you can wash out the needle, but most of the time you need a separate sample from each of these And this group is only diagnosed a small percentage of the time So you would have to save a sample, get your diagnosis, and then send a sample off
And so in addition to it being complicated and difficult in practice to run both these tests, at the same time We also began to consider the importance of the prevalence of disease when trying to understand what these tests mean And so the prevalence of disease in our population impacts the positive and negative predictive value and why the Bethesda system for reporting cytopathology expects that in the AUS/FLUS category, the malignancy rate will be 5 to 15% and in the follicular neoplasm, 15 to 30% When you look at studies such as this in different centers that report their malignancy rates for the atypia and the follicular neoplasm categories We see that there can be a great variability with a six percent rate of malignancy For example, at Northwestern compared to a 48% at Brigham and Women’s Same thing, your follicular neoplasm category Very low rates of malignancy with very highly rates of malignancy elsewhere And so depending on the center and your cytopathologist and your rate of malignancy, the outcome of these tests, the negative and positive predictive value can vary greatly And so it’s very important to know what it is at your own center And at our center, at Harbor, at Cedars and at Ronald Reagan Medical Center, the prevalence of malignancy in the AUS/FLUS and the follicular neoplasm categories is about 30 to 35% And so if you do the math at our centers and you look at the GEC classifier and you take the sensitivity and specificity of the test and you do the math with a 35% prevalence of malignancy, the negative predictive value drops to 91% So less than the 95% and less than that gold standard 5% cutoff that we’re shooting for So it’s important to know that if at our centers you use this test, you’re up to now 9, 10% of patients who you could be sending away with disease So as all this was happening, next generation sequencing came around and it revolutionized our ability to sequence genes Going from increasing logarithmically multiple logs, the number of bases that could be sequenced and minimizing the costs from thousands of dollars to a few cents And so in addition to our understanding and discovery of new molecular markers, our ability to sequence at a very low cost multiple genes has now revolutionized the field over the last couple of years This was that the first expanded panel from that 15 gene panel, from that seven gene panel that I mentioned earlier, up to 15 genes that was published in 2013 And the anticipation was that the increased number of genes would now improve the sensitivity and the negative predictive value of the molecular testing And we can see on the panel here multiple mutations being analyzed with most tumors having one mutation but some tumors having multiple mutations And the idea again was that we could now with the improved number of markers further decrease this negative predictive value of this test down to 4% or below and meet that gold standard target of a benign FNA And that led this year to the publication of this thyro-sequencing two panel, which includes 60 genes And there were two components to the study, a retrospective and a prospective group which performed similarly in terms of sensitivity, specificity, and high negative and positive predictive values And this is the overall test performance And so now when you look at this expanded panel of 60 genes and you consider the cancer prevalence along with the positive and negative predictive value And you plug in the 35% malignancy rate at our centers, you come up with a positive predictive value of 88% and a negative predictive value of 95% So one test that can provide both a rule in and a rule out are both good positive and negative predictive value In terms of prognosis and molecular markers briefly say that the presence of over one driver mutation may lead to aggressiveness and that these markers here have been associated with aggressiveness of tumors Still very much in the early stages and the BRAF mutation may have some prognostic ability,
but there’s lots of controversy The ATA and their new proposed guidelines will be recommending with some caveats the use of molecular testing And to summarize and conclude, ultrasonography should be used to direct selection of thyroid nodules to be aspirated Molecular markers appear to be complimentary and they can improve the diagnostic accuracy of FNA, particularly in the indeterminate a cytology group The gene expression classifier is highly sensitive and has a good negative predictive value, but we need to be aware of the prevalence of disease at our individual institutions to be able to interpret the tests for our patients The thyro-sequencing two panel is highly sensitive and specific with a high negative and positive predictive value We again, still need to be aware of the disease prevalence in our population, but it appears to now be a test that can provide both negative and positive predictive value And the prognostic significance of molecular markers remains controversial Thank you for your attention