Thyroid Nodules: What We Know and What We Don’t - HD

Introduction

Thank you, Jason, and thanks to the members of the program committee for inviting me to give this talk, which, as it happens, my entire radiology career coincides temporally with the evolution of thyroid ultrasound.

I thought I would back up to about a 30,000 foot view and consider what we've come to know and what we still don't know about thyroid nodules.

And I have no disclosures.

What We Know

First off, what we know, thyroid nodules are very common, but thyroid malignancies are rare and aggressive cancers are rarer still.

We've made quite a bit of progress on ultrasound, and we've come to identify that certain features and patterns on ultrasound are suggestive of benignity, whereas other features and patterns are suggestive of malignancy.

But of course, there's a lot of overlap in the ultrasound features.

What We Don't Know

Unfortunately, there's still a lot that we don't know about thyroid nodules, especially the biology of thyroid nodules.

For example, we know that thyroid follicular cells differ in their sensitivity to stimuli, but we don't know why it is that some people get thyroid nodules and other people don't.

We also don't know in great detail yet the biology of malignant nodules.

We don't know why some early stage cancers, in fact, most of them are destined for indolent behavior, and which early stage cancers are destined for aggressive behavior.

And even more profoundly than that, we don't know, in fact if it's all driven by the tumor or if in fact it's likely that there are important interactions from the host that govern the ultimate fate of these cancers.

Epidemiology of Thyroid Nodules

So just to drill down a little bit on these things, thyroid nodules are palpable in about four to 7% of adults.

But since imaging has come along, we know that they're visible in up to 67% of patients on ultrasound.

And they're seen incidentally in about 20 to 25% of imaging studies that encompass the thyroid gland.

And about half of all adults over age 60, have them.

Malignancies though, are rare among these nodules, only about up to 10% or so, and the aggressive cancers are even rarer still.

In fact, we know that many patients, if not most, of papillary cancers are found incidentally on autopsies.

36% of patients in one autopsy series were found to have clinically occult thyroid cancers.

And the five year survival of early cancers is close to a hundred percent.

Why is it that we are just drowning in thyroid nodules now in our practices?

That's because they're unfortunately common and they're a bit difficult still to characterize.

Types of Thyroid Nodules

So what are these things?

Well, nodules derive from the two cell types that make up the thyroid, the follicular cells and para follicular cells.

The follicular cells are the constituents of the vast majority of nodules that we see.

And the benign nodules have a bit of a confusing set of terminologies.

They have colloid nodules, which of course are self-explanatory.

Those are the colloid, I mean, colloid cysts.

Those are fairly self-explanatory.

But within the hyperplastic nodule class, which are proliferations of otherwise fairly normal follicular cells, you have these interchangeable terms, follicular hyperplasia, nodular hyperplasia, adenomatous nodules, they all mean pretty much the same thing.

Adenomas are also benign proliferations of follicular cells.

In some cases they're monoclonal, but not in all cases.

The big difference though, is that these are by definition encapsulated.

And then a small minority of focal nodules in the thyroid are nodules nodular forms of Hashimoto's thyroiditis.

The malignant nodules derived from two cell lines.

The differentiated thyroid cancers, which are the papillary and follicular carcinomas of which her cell carcinoma is a subcategory of follicular are derived from the follicular cells, as are the aggressive anaplastic carcinomas.

And the para follicular C cells, which secrete calcitonin give rise to the medullary thyroid cancers that are seen both sporadically and in the MEN syndromes.

Clinical Workflow for Thyroid Nodules

So in terms of the clinical workflow, when a thyroid nodule is found, what are the next steps?

Well, the first step, if the nodules found clinically, is to assess the patient by the endocrinologist and their guidelines from 2015, or I should say the American Thyroid Association guidelines state that only nodules larger than a centimeter should be evaluated, but that occasionally there are symptoms such as lymphadenopathy that may require nodules less than a centimeter be evaluated.

But ordinarily with a palpable nodule, the clinician will check the TSH and if it's suppressed, they'll order a radionuclide scan to identify a hyperactive focal nodule.

And if it's normal, they'll move on to sonography.

Now, the other route to thyroid ultrasounds has been the incidental nodule detection on CT and MR.

And here we owe a debt of gratitude to Jenny Hong and the a CR Incidentalomas committee because they've worked out this flow chart for us for what to do when thyroid nodules are identified.

Incidentally, on imaging studies, you triage the patient by, first off, whether there are any suspicious MR or CT findings such as local invasion or lymphadenopathy.

And if there aren't, then from there the algorithm works by assessing the patient's life expectancy, comorbidities and age, and then from there, size cutoffs.

Ultrasound Patterns Indicative of Benignity

Now with greater than 20 years of ultrasound experience under our belt, we've been able to say some things fairly certain in terms of the ultrasound appearance and that there are some patterns that are characteristic of benign disease.

First off is the familiar sponge offor pattern.

This pattern of tiny cysts separated by septations.

To call a nodule sponge form this pattern needs to encompass at least 50% of the nodule volume.

Another pattern indicative of benignity is the colloid cyst, an koic cystic region containing punctate echogenic foci with com tail artifacts when you see these punctate foci with com tail artifacts in a cystic region of a nodule that suggests benignity.

Another characteristic pattern of benignity is this pattern of micro nodules less than a centimeter that are hypo coic diffusely infiltrating the thyroid gland and may become confluent.

And this is the micro nodular pattern of diffuse hashimoto thyroiditis.

The Hashimoto's can also give rise to this characteristic so-called white knight nodule, which is the regenerative nodules may or may not have a halo against the micro nodular background parenchyma.

Benign Nodules Without Characteristic Patterns

Unfortunately, not all the benign nodules have a characteristic benign ultrasound pattern.

So on the left you can see a path slide of a follicular adenoma, which is a sharply circumscribed ovoid kneel nodule with a capsule around it.

And on the right is an ultrasound image of follicular adenoma.

It often has this pseudo testis appearance where it's oval and homogeneous and looks like a testis, but it can be hypo iso or hyper coic.

And it's not possible to make the distinction of whether these are benign or malignant, or indeed whether these even are follicular adenomas.

They often degenerate and have cystic regions containing debris.

Hashimoto thyroiditis, which we said gives rise to that characteristic mic r nodular pattern when diffuse unfortunately also has a focal form, which can be impossible to differentiate from other focal nodules in the thyroid.

Ultrasound Patterns Suggestive of Malignancy

So now we move on to looking at the patterns that we've come to recognize are suggestive of malignancy.

Well, first off, we know that the majority of papillary carcinomas have solid and hypoechoic appearance, but those are both non-specific features.

One powerful predictor of malignancy is this taller than wide appearance.

And in some series it's been shown to have a positive predictive value of 77%, although it's important how you measure the nodule.

And we're fortunate to be followed in this talk by Dr. Tessler, who will drill down a little bit more on how we should measure and assess these nodules.

In the new Rads classification.

In addition to hypoechoic, a subcategory of very hypoechoic is an even stronger predictor of malignancy in that what we mean by hypoechoic is that it should be darker than the overlying strap muscles.

And this sign has a positive predictive value of approximately 80%.

Although it's not especially sensitive, we can see this hypo coic appearance in the so-called micro carcinomas that are less than a centimeter.

And one feature that's especially important to look at is the margin features of a nodule.

And here we have a nodule, which has a very dramatic speculation, which my colleague Brooke Jeffrey has nicknamed the Jagged Edge.

And we found that this is a highly suspicious feature for malignancy.

But we know that irregular and lobulated margins also have a higher association with malignancy than smooth margins.

Extra thyroidal extension may sometimes be difficult to pick up on ultrasound, but when we see the loss of the echogenic capsule, and certainly when we see invasion into the overlying strap muscles, it's likely that extra thyroidal extension is present.

And here we can see a case of thyroid cancer breaking out through the anterior capsule and invading the overlying strap muscles microcalcifications, these punctate echogenic foci that are very bright, but without shadowing, were shown early on to have a high association with thyroid cancer.

And they are thought to represent these Sam bodies that on pathology may represent dead little follicular cells surrounded by concentric calcium deposition.

Microcalcifications are a highly useful finding when you can be sure that they're present.

But the inter observer variability is fairly significant when it comes to assessing these things.

So you can see that the positive predictive values and sensitivities range fairly wide in different studies, but the odds ratio is also from about five to eight times normal.

Now, early on it was thought that eggshell calcifications were indicative of benignity, but it was shown that particularly when the eggshell calcification is discontinuous, that actually that is a concerning sign for malignancy.

Although, here we see a papillary carcinoma on the left with an eggshell calcification and a follicular carcinoma on the right when the calcification is thick and complete, it can of course be extremely difficult to assess the echogenicity of the tissue within macro calcifications, likewise have a higher association or confer a greater risk of malignancy, I should say, with an odds ratio of close to three.

And these are the calcifications we see that have acoustic shadowing associated with them.

Vascularity

Now, what about vascularity?

Well, initial reports suggested that per nodular vascularity greater than intrinsic vascularity was suggestive of benignity, whereas intrinsic greater than per nodular flow was indicative of malignancy.

So with that in mind, I ask, which of these nodules is malignant?

And this being a trick question, it turns out that here's a benign nodule with extensive intravascular flow and a malignant nodule with more peripheral flow.

So it turns out that because there's so much variability and sensitivity of doppler instruments, the vascular pattern is really nonspecific and we'll see that it's not currently incorporated into several of the scoring systems that have come to be used.

On the other hand, it's extremely useful and in fact should be deployed in assessments of all nodules, including cystic nodules.

In a series that we did in 2003, we found that approximately 6% of the cancers we looked at were cystic, but they all had vascularity in the solid appearing portions.

And then more recently, a larger series of 360 malignant nodules from Mayo found that all the cystic nodules had other suspicious findings.

So when a cystic nodules identified interrogation of the solid portion with doppler is a must.

Sono Elastography

In more recent years, we've seen the introduction of sono elastography, both with strain imaging that looks at the deformation of the nodule, as well as ShearWave imaging that looks at the propagation speed of a shear wave developed within the nodule.

And color overlays are deployed, as we saw in Dr. Barr's talk this morning, to depict stiffness.

And here we have a dual screen image that shows a nodule with the stiffness shown in red.

So stiffer than background is felt to be concerning for malignancy.

And some of the meta-analyses incorporating sono elastography papers suggest that sono elastography is a highly useful predictor of malignancy with a positive likelihood ratio second only to taller than wide for predicting cancer.

On the other hand, elastography has some pretty significant limitations, especially in nodules that have significant fibrosis because they can be, they will appear stiff and fibrosis can occur in both malignant and benign nodules.

Coarse calcifications present in benign nodules can result in increased stiffness and cystic nodules.

The cyst, the sono elastography measurement will reflect the fluid in the nodule and not the stiffness of the solid component, which is the area of interest.

So as of right now, we don't incorporate elastography routinely in our practice.

And this is the kind of problem we run into though, where we see nodules or cystic nodule with some flow, a hypo coic nodule with some peripheral flow.

These were a atypical carcinomas.

Likewise, here's an atypical carcinoma and that this is an iso coic papillary cancer, and here is a papillary cancer that had what look like a sponge form appearance.

So as much as we would like ultrasound to be specific, in fact, what we need to determine is for a nodule with a given ultrasound appearance is fine needle aspiration indicated.

And for that, we'll hear much more from Dr. Tessler who will speak next about the new a CR tyros scoring system that was developed.

Fine Needle Aspiration and Bethesda Classification

But in practice, the next step, once you've got a thyroid nodule that you can't definitively characterize, is to perform fine needle aspiration and biopsy is cytology is feasible because there is a characteristic set of nuclear features that can be used by pathologists to make a diagnosis of papillary cancer.

The cytology results, as we heard this morning from Dr. Langer, are reported using the Bethesda classification, and that's then used to dictate the type of management.

So the categorization ranges from one to six as we heard this morning.

And the categories of suspicious for malignancy and frank malignant clearly indicate that surgery is necessary, but there is an indeterminate category of atypical or follicular lesion of uncertain significance, which gives rise to ambiguity as to what management is needed.

Over-Diagnosis of Thyroid Cancer

Now, since fine needle aspiration was introduced, what's happened is there's been a tremendous rise in thyroid cancer diagnosis.

In fact, here we can see the time point of introduction of fine needle aspiration at around 1990.

And since then, incidents has tripled, but there's been no change in mortality, which is the dashed line shown here at the bottom.

And when this is seen on epidemiologic chart, what's happening is that malignancies, histologic malignancies that would have otherwise remained clinically occult are now being diagnosed and treated.

And this is the situation that we believe is happening now with thyroid cancer over-diagnosis.

And it's resulting in aggressive treatment of an indolent disease, which is, and a treatment is not without its morbidity, and it results in a significant expenditure of medical costs at a time where we're seeing more and more constraints necessary in costs.

Distinguishing Aggressive Malignancies

So here's what we don't know.

How do we distinguish potentially aggressive malignancies?

Those are the ones we look for the needle in the haystack from those that are destined for indolent behavior.

Now, once the nodule comes out, histologic classifications can suggest that there, that certain variants, namely the tall cell variant columer and diffuse sclerosing variants, these are the variants that tend to have a more aggressive behavior clinically, whereas follicular variants tend to behave in a more indolent fashion.

So some groups have looked at what the ultrasound correlates of these histologic variants are, and they described some of the features that are associated with them, et cetera.

Unfortunately, though most of the aggressive thyroid cancers that are encountered in clinical practice are still conventional papillary thyroid cancers.

So that is going to continue to be fraught with problems.

Molecular Diagnostics

Nevertheless, as we saw this morning, a whole new category of diagnosis has now been enabled by working out of signal transduction pathways.

This field of molecular diagnosis, which is currently deployed to assess cytologically indeterminate nodules.

We know that the mutations that drive the development of thyroid cancer and thyroid neoplasms occur in many signaling pathways, but particularly the signaling pathways in the so-called MAP K pathway, including these familiar mutations in RAF and RAs and other proteins.

And so molecular diagnostics has now been used in the realm of these indeterminate cytology cases, the ayia and follicular lesions of uncertain significance.

As we heard this morning, the AFFIRMA test, the gene expression classifier test of mRNAs can be used to rule out cancer because it has a high negative predictive value in that when these genes are not expressed, you can say that cancer is not present, but it's not sufficiently sensitive to rule it in.

Whereas this next generation gene mutation infusion panel test called thse Developer Pittsburgh, has both a high negative and positive predictive value.

So it's felt to be able to rule it in and rule it out.

Unfortunately, these tests are quite expensive and it would be nice to be able to come up with other methods to assess these indeterminate cases.

Now, in the future, molecular diagnostics is likely to be more widely deployed, not just to determine whether patients will have a lobectomy versus a total thyroidectomy or to treat the iodine refractory metastatic thyroid cancers with molecular targeted therapies, but what we're hoping is that this may prove to be a means by which we can distinguish the behavior of the cancers as to whether they'll have an aggressive clinical behavior or not.

So some work suggests that the B-R-A-F-V 600 e mutation present in conventional papillary thyroid cancer is associated with a worse prognosis, especially when it's present with this TERT promoter mutation that has an even more aggressive course.

And then likewise, the RAs mutation is thought to be the one driving follicular thyroid cancer.

Unfortunately, though, these studies are difficult to do because so few thyroid cancers behave aggressively at all.

So it's likely that it will take a while to work all this out.

Future Directions

Now, in the meantime, what would be useful is to be able to stratify the risk of aggressive behavior in early stage thyroid cancers before making a decision to biopsy.

Now, intravascular microbubbles, we heard a little about that this morning just looking at the vascular pattern, it's not clear that that's going to be of additional yield beyond what we already have.

On the other hand, newer generations of targeted microbubbles that can act like silver bullets and zero into the surfaces of cells harboring abnormal surface proteins that could indeed prove to be not just a useful marker to identify these cancers, but potentially to treat them.

And one of the areas that I've been interested in is looking at machine learning algorithms to look at the features of thyroid nodules on ultrasound features that may not be visible or apparent to the human observer, but which may be apparent to a machine such as texture and other factors called kurtosis.

And what you do with machine learning is you take a set of labeled inputs.

In this case it would be labeled images.

You would extract the features either by a human or the machine.

They would be developed into a so-called feature vector and used to develop a machine learning algorithm, which you would then test by feeding the algorithm many new images and seeing how the algorithm labels these.

And as you've probably heard in many other domains, machine learning is a huge area of investigation.

Many groups are working on this, as you can see, including ours.

So we're hoping that we'll come up with some additional interesting ways of analyzing thyroid nodules by looking at some of the imaging features that might not be apparent to a human.

Conclusion

So in conclusion, what we know is that the adult population harbors a very large number of both clinical and subclinical thyroid nodules.

And that histologic malignancy does not equate to clinically aggressive behavior.

The sonographic features associated with malignancy have, for the most part been described and used to develop scoring systems.

But in the future, we need to work toward better preferably non-invasive ways of distinguishing those rare, aggressive forms of thyroid cancer from those destined for indolent behavior.

And so with that, I'd like to acknowledge my colleagues, AYA Kamaya and Brooke Jeffrey, who have been invaluable inspirations in this field of thyroid nodules and our Stanford sonographers.