Understanding How Clinical HPV Testing Leads to Personalized Medicine At Its Best

MARK STOLER, MD:

Today's webinar is about really to have a conversation about how HPV testing really impacts our patients in the screening for cervical cancer in the triage of patients of who will be treated, and who can have just repeat surveillance, and how we can make better diagnoses.

Our objectives today are really in three parts. We're going to talk about the biology of HPV and carcinogenesis so we're all on the same page. We're going to describe the importance of what we mean by clinically valid HPV testing, and how clinically valid HPV testing is indispensable in cervical cancer screening today, and then we'll finish up with a discussion of how we can just adjunctive testing, immunohistic chemistry, and in Situ hybridization, not only cytologically, but more importantly, histologically to improve our diagnostic accuracy for our patients.

Biology of HPV and carcinogenesis

Papilloma viruses; human papilloma viruses. It says here there are more than 100 types identified; there are actually over 200 types of human papilloma viruses, and somewhere between 30 or 40 or 50 types that infect the mucosal surfaces. The mucosal tropic or mucosal-attracting viruses can be broken into two main groups. The so-called oncogenic types, those types that are associated with cancer, although we should emphasize the majority of oncogenic HPV infections do not lead to cancer. The non-oncogenic types that mostly cause genital warts but can very, very rarely, probably 1,000 times less frequently than the oncogenic types also would be associated with cancer.

This is a diagram of the HPV family tree. The mucosal viruses, the so-called alloplapapilloma viruses are above the dotted line, and the other genuses of papilloma viruses below the dotted line, all vertebrate species on the planted where they've been looked for have a population of papilloma viruses. Let's see if I can get the pointer to work. You can see here the HPV 16 associated group of papilloma viruses. HPV 16 and its closely related viruses, 31, 33, and 35, as compared to when you click here, the HPV 18 associated viruses with its closely-associated cousin HPV 45.

So it's been known for several years now that virtually all cervical cancers are associated with HPV and there were very large studies done in the early 2000s using PCR techniques that showed that 99.7%, just under 1,000 specimens from countries around the world were HPV-positive. This study has been expanded 10-fold in more recent times, over 10,000 specimens with these very, very high fractional associations of HPV in cervical cancers, confirmed cervical cancers from around the world.

The natural history of papilloma virus infections can be divided into two basic populations. The vast majority of HPV infections lead to transient infections where it manifests this mild cytologic abnormalities, and if you time the PAP smears appropriately these transient infections are the productive phase of the HPV life cycle. These are the lesions, the morphologic lesions that correlated with virion production and transmissibility. In contrast, a small fraction of HPV infections, regardless of type, tend to lead to persistent infections. These persistent infections, the longer they persist are associated with the development of histologic and cytologic precancerous lesions that can eventually lead to the development of invasive cervical cancer and other carcinomas.

So we can think of HPV infections as morphologically manifest in a period-type scheme in the United States where they're screening, cancer is relatively rare outcome, although there's still 12-15,000 cancers that occur in the United States each year, more than half of these cancers, of course, occur in people who have never been screened. In contrast, the mild cytologic abnormalities are incredibly common in the population. 1-5 million ASCUS cases or LSIL cases are seen cytologically here in the population of about 70-80 million PAP smears done annually.

In places like the United States, Europe, Canada, Australia, where screen has effected a marked reduction in cervical cancer, we have to note that the majority of HPV infections that put people at risk for the development of cancer and pre-cancer per the unscreened populations of the world, sub-Saharan Africa, South America, Asia, et cetera. And then today we know that it's not just cervical cancer that is associated with HPV, but significant fractions of vaginal squamous cell carcinoma, vulvar squamous cell carcinoma, penile carcinoma in men, anal carcinoma in both sexes, and increasingly of clinical importance, which we'll touch on at the end, oropharyngeal carcinomas are associated with the alpha subgroup of papilloma viruses. The beta papilloma viruses, which we're not going to talk about, are highly associated with cutaneous cancers, particularly in immuno-compromised patients.

And so one of the final concepts I leave you with is that HPV infections are very common. They occur mostly early at the onset of sexual activity with the exposure of one or more sexual partners. Some people consider HPV infections over the lifetime of the patient part of your normal experience of being a human being. In contrast, cancer and pre-cancer are relatively rare. They do develop over time and they take the time courses as one of years to decades for the transit time between an infection and the development of invasive cancer.

Importance of clinically valid HPV testing

So I want to move now to the consideration of the cervical cancer screening process, which for those of you who are involved in this process, that appreciate that there are several steps. One screens the population of basically healthy individuals looking for abnormal cytology or HPV positivity today, or both. We have to decide what to do with the patients who are positive while better reassuring the patients who are negative. Positive patients most often in a colposcopy where their biopsied in an attempt to try and confirm the presence or absence of precancerous lesions that could be treated and thereby lead to the prevention of the basic cancer.

In each one of these steps we're going to touch briefly upon the role of HPV testing and cytology in both screening and triage, and how some of the newer concept of looking at the risks associated with the different HPV genotypes, as well potentially triaging patients with immunocytology will be an increasing consideration. Then finally we'll look at the applications of immunohistic chemistry, particular p16 immunohistic chemistry, an insight to hybridization methods for improving biopsy diagnosis. So we consider cervical cancer screening. We have to remember that preventing all cancer by screening is unrealistic. We can have a dramatic effect in the United States; 50-60 years of PAP smears have effected almost a 90% reduction in the incidence of cervical cancer in the population, but it's never going to take it to zero by itself. PAP smear cytology has been an incredibly powerful tool in our battle against cervical cancer, but it's the prevalence, as we'll talk about, as the prevalence of cancer and pre-cancer drops with repeated rounds of screening, or now with HPV vaccination, we absolutely have to have better, more sensitive techniques than cytology alone.

If you talk to anybody who works in the field of making guidelines for cervical cancer management, the concept that is frequently talked about is the idea that patients who are at equal risk should be managed the same way, something that seems intuitively obvious in retrospect, but if you look at older guidelines, was not always the case.

So indeed, if you look at the most recent cervical cancer screening guidelines from the American Cancer Society, the fundamental goal for cancer screening is we talk about the concept of balance, balancing the benefit of screening with the potential harms of screening and over-treatment if we over-screen individuals and over-diagnose people who don't really have something that needs aggressive treatment. Of course, our goal is to do what's best for the patient, ideally regardless of cost. So the risk of cancer is always present. Screening means nothing if we don't find pre-cancer, detect it, treat it, to prevent the development of cancer; that's really what the cervical cancer screening process is all about.

But there are risks to treatment. The treatment involves surgery, surgery involves removal of a significant portion of the cervix, and this has been shown at least in some studies to put patients at risk for premature delivery and other potential obstetric complications, not to mention the cost and discomfort associated with treatment.

So we can think of the cervical cancer screening process as a tradeoff between detection of cancer and pre-cancer, thereby identifying the PF patients who need treatment specifically, while sensitively identifying all those patients who were at solo risk for having cancer, pre-cancer, that they can be reassured and not have to be screened, except at whatever the routine screening interval is. In screening we want to be maximally sensitive to reassure the patients who don't have disease, but in triage we want to be specific. We want to send as few patients as necessary to colposcopy, to biopsy, to excisional therapy. So we talked about balancing the number of cancers and pre-cancers prevented, and used the number of colposcopies as a surrogate measure for the potential harms associated with treatment.

The idea that I want to get across is that today it's not just the cytology that we can take the patient's cytology, we can take their HPV test results, we can look at the colposcopic results, we can look at their biopsy results, and we can stratify their risk. So we can risk-stratify patients by looking at these factors together with their age, and realize that we fundamentally only have three or four things we can do for the patients. We can continue to screen them because they have a very low risk of pre-cancer. We can look at them more closely. We can repeat the interval; instead of every three-year or every five-year cervical cancer screening, we can do an every-year screening or every six months screening. Those at sufficient risk in the United States, somewhere between 5 and 10% risk of a pre-cancerous lesion being present in the cervix is the threshold for taking the patient for colposcopy. Then there are some patients, as we'll talk about, where any one factor may put them at such high risk that you can skip some of this stuff, so potentially just treat the patient. Somewhere around 40, 50, 60% risk of pre-cancer, many patients may just choose to be treated in the appropriate clinical setting.

So the step-wise concept of cervical cancer screening, let's talk a little bit about how HPV testing integrates into this concept. Before we do that, I want to make this point about clinical validity of HPV testing. Because there's no question that study after study, and there have been numerous studies worldwide that have demonstrated that the sensitivity of HPV testing compared to cytologies, 20-40% more sensitive. What we mean by that is if you have a CIN3 lesion, the typical sensitivity of a PAP smear detecting that CIN3 is somewhere between 50 and 70%, whereas for HPV testing, clinically valid HPV testing, it's more than 90%. So it's important to note that what we're doing with HPV testing in the context of screening is we're not looking for HPV, we're trying to establish who has CIN3 or CIN2-3, and who does not. We want to do that without destroying the specificity of that test, and that's the concept of clinical validity.

There have been guidelines written by us and others in the United States, as well as in Europe, that fundamentally talk about what it means to be clinically valid, that the tests have to have a certain spectrum of papilloma viruses, that its sensitivity has to be at least 90% for detecting CIN3+ with relatively preserved specificity, and labs should be able to do this in a reproducible and robust manner.

As you may know, there are only four FDA-approved, validated HPV tests in the United States. This'll probably change in the next year with the potential approval of a fifth test. BD Diagnostics has a test before the FDA that performs very similar to the other PCR-based tests. Some of these tests, historically, the first test on the market, hybrid capture 2, was a test for 13 high-risk types that didn't provide genotyping. The Cobas test is probably the most widely used test currently in the U.S. It tests for 14 high-risk types. It gives you HPV 16, 18 genotyping integrated into the assay. The Aptima RNA test is also quite popular in the United States. Again, 14 high-risk types, and a second test is applied for partial genotyping for types 16, 18, and 45.

I also want to say that in my opinion and the opinion of many others, given what's happening around the world, primary screening is a coming thing. Though some of you may be aware that the United States Preventative Services Task Force is revising their guidelines, which also came out in 2012. In the guidelines that are out for commentary, they single out primary HPV testing every five years as a method that they're potentially going to recommend for cervical cancer screening.

The reason for moving to HPV testing, you can think of it as simple mathematics. This is a graph that talks about the sensitivity of a PAP smear at 50% sensitivity, and the effect of prevalence on the performance of this test, and you can see, we go back 30, 40 years in the United States. The prevalence of CIN2-3 in cancer in the United States was above 10%. At that point you can see what the negative and positive predictive values were in the population. But today the prevalence of CIN3 in populations, about 1%, and with HPV vaccination it's going to fall to less than 1%. As we fall from 1% to some fraction of a percent, with ongoing screening and with vaccination penetrating the population, what you see is that the positive predictive value in particular starts to become very, very low in the population and requires the use of higher sensitivity methods, until even the most sensitive thing we know, which is HPV testing, will cease to function. So change, in my opinion, is a mathematical necessity.

So we look at the concept of HPV primary screening. As I alluded to, it's being implemented in a number of countries around the world. The Netherlands has already started primary HPV testing, replacing their well-recognized and very successful cytology system. Australia starts in two weeks, replacing their cytology system with a system of HPV primary screening. Most countries are using HPV primary screening, meaning they test for HPV first and then the triage patients with cytology and/or genotyping.

In the United States we've had since 2014 an FDA-approved algorithm for HPV primary screening. The algorithm looks like this one, tested with a clinically validated HPV test. In this case, the only one approved for this indication. As far as the Cobas HPV test, if you're HPV-negative, one goes through routine screening, which based on the data from the ATHENA trial is assumed to be three years, but the United States Preventative Services Task Force draft, which you can look up on the web, they're talking about five-year HPV primary screening, which is the standard as it's being implemented in Europe, Australia, et cetera. Patients who are HPV 16, 18 positive, because the data I'll show you in a minute are such high risk of developing or having CIN3 that they need to go to colposcopy, and the 12 other high-risk HPV posits, which is about 7% of the entire screening population get cytology. If their cytology is abnormal at any level they also go to colposcopy, whereas normal cytology would get a much closer interval follow-up compared to every three- or five-year screening.

Here are those data I alluded to. These are data, the long-term follow-up data from the ATHENA trial. If you have a normal PAP smear, independent of your PAP cytology, if you have a normal PAP smear and you're HPV 16+, your cross-sectional risk of having CIN2 or CIN3 in this case, it varies from 17% at baseline to over three years, 25%. So one in four HPV 16+ individuals have CIN3 if you take them to colposcopy and do an adequate number of biopsies. HPV 18 similarly varies from just under 10% to just over 10% over three years. It actually goes up quite a bit more in other studies if you look at longer-term follow-up. 12 others vary from 4 to 5.4%, but as we'll talk about in the middle, this is an average of the prevalence of the 12 types and their relatively variance, and as we'll see there are some newer data to suggest that there's clinically useful information potentially hidden within this line here. Of course, the important point is that the HPV-negative population stays right close at baseline with a very, very low risk compared to an unscreened population.

Here is data from Denmark that emphasized that persistent HPV infection, as I said, persistent HPV infection, this is with people who have normal cytology. Persistently HPV-negative infected people have almost a 50% chance of having a CIN3 compared to a single infection. These are different populations, different HPV tests, and very similar results. But you can see here the interesting point that's not evident in the graph I just showed you previously. HPV is 31 and 33, which are genetically the closest relatives of HPV 16, have risks similar to HPV 18 in terms of the risk of developing CIN3 over a decade in the population.

So these are data from the Clarity, the BD assay that's before the FDA trial that was presented at several meetings so I feel comfortable presenting these data here. The Clarity test allows you to get genotypic information. It's broken down on the side here. A few of the types are grouped because of limitations of the PCR assay. I want you to just pay attention to the absolute risk numbers. So your HPV 16+ cross-sectional in this trial, the risk of having CIN3 is 20%, and 31 and 18 are both above a 10% natural break in the population here. The next group, which is below 5% to 2%, HPVs 33, 52, 45, between the 2 and 5% risk of having CIN3. And then there's a population of very low-risk viruses, including this cluster of 59, 56, and 66. So the idea is that the genotypes have different levels of variance, and it actually allows you to risk-stratify your patients in terms of who needs to go to colposcopy and who doesn't.

So one possible prediction, I'm not saying this is going to happen but people are talking about extensively at conferences where they talk about screening modifications, is that the genotyping information inherent in clinically valid HPV tests could potentially risk-stratify patients without doing a lot of additional testing into groups that might need either colposcopy, or groups that could be submitted to increase surveillance, or groups that could probably just be treating with regular-interval surveillance.

Well, one of the pushbacks one often gets about HPV testing in terms of the concept of primary screening is, well, does HPV testing really work? Does it really get everybody we want to identify in terms of CIN3 in pre-cancer risk? As I said, perfection is not possible, but in the ATHENA trial which went before the FDA, we asked the over 300 cases of CIN3, or adenocarcinoma - - CIN3+ that occurred, 282, or 92.5% of them were identified by the Cobas test, but 7.5% weren't. The question is, are those HPV-negative CIN3s, are they actually falsely negative or is there a biological reason for the negative result? It's important to remember that we can't necessarily test for every possible HPV. We only have 14 types in the clinically valid assay. We don't have 40 types. The reason we don't have 40 types is that it would destroy the specificity of the test, and you still wouldn't have perfect sensitivity. But if you follow this flow diagram, what you see is that at least half of the so-called HPV-negative CIN3s, or adenocarcinoma and SITE-2s are due to HPV types that we don't test for, for the reasons I just mentioned. The other half, half of them were p16 positive, which we'll talk about in the next few minutes, is a good surrogate marker for HPV-negative-associated transformation. A couple of these were for types we do test for, but the majority of them were for other types as well. Then we had a few cases that were p16- and cobas-. Three of them on review were ruled by an adjudication panel to just be immature metaplasia.

So what you see here is that basically there are no HPV-negative CIN3s, right? They're either HPV-positive for the types you test for or the types you didn't, and there are a few technical reasons for imperfection, but compared to the 50-70% sensitivity of the PAP smear, there's no question that HPV testing is superior.

Now, as we talked about the algorithm for HPV primary screening in the United States, it looks like this. HPV primary screening helps you risk-stratify patients into the various risk groups. This is a breakout of that diagram that I showed you before. HPV-negative individuals have very, very low risk, way lower than the background prevalence, and be re-screened at long intervals. Again, in the United States, typical screening intervals that are recommended in guidelines are 3-5 years, and in Europe there's a pretty much uniform five-year screening guideline as will be the result in Australia and probably Canada in the not-too-distant future.

How can we improve our diagnostic accuracy?

But can we even refine this algorithm more? Can we use something better than cytology, given the knowledge that we have about its relative limitations in terms of sensitivity? Well, here's where I think the near future, and by near future I mean an assay that's now on the market in Europe and is available in Canada, is available potentially in Australia, and is being talked about as substituting for cytology in the primary screening algorithm, this concept of the dual-stain cytology where you immunohistochemically stain a thin-prep slide. p16, as you know, is a protein that puts the brakes on the mycotic mitosis trying to affect cell cycle arrest in cells that are proliferating. Ki-67 is a widely-recognized immunohistochemical marker that's correlated with mitotic activity, and the idea here is that coexpression of p16 and Ki-67 is evidence of a cell that really has lost its way, it's deregulated or unregulated cell where the two factors are in direct conflict and normally doesn't occur in the normal situation, but it's highly correlated with HPV-negative-associated-induced neoplasia.

These assays are very easy to interpret. The dual-stain with the dark red T-67 and the brown and red staining of both the nucleus and cytoplasm is what a positive looks like, and in the data I'm about to talk about, positive just means finding one of these cells on the slide.

And so we look at a study that was published early this year, last January, in Gynecologic Oncology, where we retrospectively stained a residual ThinPrep file, so collected in the ATHENA trial, and asked, what's more sensitive for the prevalent CIN3 in the population? You can see that dual-stain cytology was about 25% more sensitive compared to PAP cytology in a highly statistically significant manner with virtually no cost in terms of specificity. So higher positive predictive values, but more importantly, 50% reduction in the residual risk in terms of reassuring the patient with a negative result that they don't have CIN3 and can safely be followed by normal longer interval screening.

So we can look at this in terms of the potential different algorithms that we would have for detecting CIN3, and you can see that compared to a PAP smear, dual-stain cytology is 25% more sensitive. Adding genotyping decreases specificity but adds a little bit of sensitivity to a primary screening algorithm and dual-stain plus genotyping gives you the best combination as we go up in this ROC-like curve in terms of sensitivity and specificity.

To look at these data, another way to look at these four different possible algorithms, I just want to call your attention to this slide here where we detect the most CIN3s earliest in the screening system. Yes, we do more colposcopies, but we find much more disease compared to the other algorithms. So the balance, if you will, between sensitivity and specificity is a net of about seven colposcopies performed to detect a single case of CIN3 in this situation with very, very high sensitivity.

So in the near future, I think what you're going to see in the area of HPV primary screenings is substitution of dual-stain cytology as a test that, as I say, is already on the market in Europe with abundant data supporting its potential use, a test that's currently about to undergo formal FDA evaluation in the United States.

So let's turn now-well, before we do that, the data clearly show that a dual-stain result splits the population into a population that's at very low risk, and they have close follow-up or regular rescreening, whereas dual-stain positive individuals all need to go to colposcopy, and the HPV 16+ dual-stain positive individuals are entering the realm where serious consideration could be given to doing primary therapy.

Well, we know cytology, for those of you who practice cytopathology, is hard, but biopsies? Every surgical pathologist reads biopsies and they think they do it well. But when one looks critically at the ability of pathologists to read biopsies, particularly cervical biopsies, we find that the interpretation is often subjective and there is variability that can lead to under- and over-reporting of significant cervical disease, can put your patient into the wrong bin, if you will, in terms of how they should be managed.

This was first in modern literature brought forward when we looked at the - - cell. So triage study at the comparison between local study diagnosis, in this case groups of academic pathologists at four major academic centers, and another group of academic pathologists, and we looked at thousands of cervical biopsies and we ask, how often do pathologists agree? Only 71% of the diagnoses were on the diagonal. You can look, for instance, in CIN1 diagnoses, which remember, when we get a biopsy we always know the patient has an abnormal PAP. The most commonly over-called cervical biopsy in this study was the CIN1, where just about 50%, 48, 50% of CIN1 diagnoses by one group were called normal by the other group. CIN2, as you all know, is also highly variable; about a third of CIN2s will get called a CIN3, and another third will get called a CIN1 or normal in a population, whereas normal biopsies we're very, very good at when it's obvious and easy and normal.

So if all cervical biopsies look like this, normal, CIN1, CIN2, CIN3, we wouldn't be having this conversation, would we? But the reality is, is that the data shows that we need some help. We need to be able to refine our diagnostic acumen so that when one pathologist reads a biopsy, another pathologist comes to the same conclusion. Clinicians hate that kind of variability, I can tell you for a fact.

So how can we do that? Well, the thesis here is that we can use the in Situ hybridization or immunohistochemistry to help adjudicate biopsies. Immunohistochemistry, as you know, is a well established methodology available at virtually every pathology lab. It may or may not address the pathology of interest. The in Situ hybridization, historically, has been technically more demanding, but for the context of HPV-negative-associated neoplasia the in Situ hybridization has a lot of appeal because normal and reactive processes are not HPV-positive at the level of in Situ hybridization, whereas neoplasia in CIN1, CIN2, CIN3, cancer, adenocarcinoma and in Situ all should be HPV-positive.

Situ hybridization and immunohistic chemistry allow you to correlate the morphology exclusively. A term I often like to use is "seeing is believing."

The in Situ hybridization, whether it be for DNA, RNA, you have to be aware that historically the methodologies were quite variable. The sensitivities were variable, the types of HPVs included in the assays were variable, and methodologies have evolved over time. So I've been doing in Situ hybridization for three decades, but these types of radioactive in Situ hybridizations using our radiography were never clinically useful because of the time it took to get the result out of the laboratory.

Morphologic non-radioactive in Situ hybridizations using immunohistochemical-like detection systems have been around for decades, but they've been somewhat variable in their diagnosis, and the HPV and DNA in Situ hybridizations that were most widely used because of limitations of types factor, because of variability in sensitivity, and quite frankly because of some abuse of these assays in various laboratories, are no longer readily available in some markets.

Yet these assays did correlate in a fraction of cases and were relatively easy to read. Here you can see a low-grade lesion, you can see the transit amplification of HPV DNA from the basal cells to the coitalcitic cells as it goes up towards the surface. These cells have even less than 10 or 100 copies of HPV DNA per nucleus, these cells having thousands and thousands of copes of HPV DNA per nucleus.

So these assays were useful, but as I said, clinical trials leading to the real clinical validation of HPV DNA in Situ hybridization so they could be widely recommended for clinical use were somewhat problematic in my opinion.

In contrast, up until recently there haven't been formal clinical trial, but I should make you aware that the FDA recently formally approved the clinical data from a large clinical trial for the use of p16 as a diagnostic adjunct for biopsy interpretation. Positive p16 immunohistochemistry is so called block-positive where runs of at least 6 basal cells and then extending up in the epithelium at least a quarter of way as defined as this dark, strong, nuclear cytoplasmic positivity, so-called block-positivity.

The biomarker p16 is useful in differential diagnosis, so some may look at this and say, well look, there's a parabasal proliferation extending up into the epithelium, and there's maybe some atypia here and some vacuolation, and this is an HPV-positive lesion, whereas others may say the nuclear atypia is not significant enough. I think this is reactive squamous metaplasia. The more one sees a p16 reaction like this in the context of a squamous proliferation, one can generally be reassured that this is an HPV-associated lesion.

Here is a similar lesion, metaplasia versus dysplasia - - in endocervical gland, in this case a completely negative, clean p16 immunohistochemistry would suggest that this is a reacted metaplastic process and not CIN involved in the gland. There are cases where, unfortunately for us, pathologists, we have to admit that we miss things, and this is a biopsy signed on at our laboratory as negative. The patient had a significant risk of having CIN2 or CIN3 in that they had an eight-cell cytology. You can see here that it's high-power. There is this little area of metaplasia and hyperchromatic cells that was missed at low power, but at low power the p16 immunohistochemistry highlights these lesions, and this is particularly useful in the context of a patient who has a high risk of having CIN2-3 and a negative biopsy.

One of the LAST guidelines. So many of you are aware of the lower anogenital squamous terminology project that looked at the literature supporting the use of p16 adjunctive immunohistochemistry in the context of cervical neoplasia. In that paper we estimated that approximately 20% of cervical biopsies would need p16 immunohistochemistry and subsequent analyses, that number is somewhat higher, although it's still less than half of biopsies; using the LAST guideline would require p16.

The LAST guideline suggests that the major uses of p16 are to aid in the differential diagnosis of CIN2+ with a mimic of precancer anytime a morphologic diagnosis of CIN2 is considered, because you don't want to overcall CIN2 and therefore commit somebody to treatment to adjudicate professional disagreements. Importantly, in the context of the patient who had a 20 or 30 or 40, or greater percent chance of having CIN2 or CIN3 and the biopsy looks negative, particularly if it's fragmented or metaplastic or highly inflamed.

So in the last five minutes I want to leave some time for questions. I want to talk about some newer forms if in Situ hybridization, HPV RNA in Situ hybridization. There is a kit on the market developed by a set of probes developed by a [inaudible] cell diagnostics which I helped to engineer. This was strategically structured to help decrease the discordance between your morphologic impression and the assay positivity and negativity, and that we include 18 of the most common high-risk HPV types. So with adequate sensitivity and detection we're going to have much fewer cases where the lesion is due to a high-risk virus. It really is lesional and you get a falsely negative HPV test, one of the limitations with some of the prior HPV DNA diagnostics, particularly those with many fewer HPV types in the probe mix. There's a companion probe set which has six low-risk HPV types, because remember even for CIN1, 15-20% of CIN1s are due to low-risk HPVs, not high-risk HPVs.

We've recently published in the American Journal of Surgical Pathology our first experience with this assay on an automated platform. It's very good to know that these assays are now clinically doable with automation to make inter-laboratory comparability comparisons much more viable. In this analysis we had very good sensitivity for the HPV RNA risk compared to other prior assays, HPV DNA in Situ hybridization. P16 immunohistochemistry and PCR are all done at a very high-level sensitivity and very good specificity for that reactive problematic lesions in normal biopsies don't stain at all.

I can show you some of these data. You can see how clean the normal and metaplastic cervix the assay is; there's virtually no signal that these very sign stringency in Situ hybridizations, whereas low-grade lesions have this very nice speckled pattern of both nuclear post-cytoplasmic RNA, both low-grade lesions and high-grade lesions are easily interpreted with a very low background. There are cases that come up clinically, I'm sure all of you have seen, where you think the lesion is high-grade. Here you can see something that one might call CIN2 based on the degree of nuclear atypia in parabasal proliferation. The p16 is surprisingly negative except for scattered cells, probably some lymphocytes in the base, yet the DNA and RNA in Situ hybridizations are positive. This phenomenon of a high-grade lesion that's p16-negative is pretty rare. We think from other studies that it's probably due to shut off of the p16 mutation and the p16 due to methylation, but in certain contexts it's important to get this right, because more and more patients with HPV-assocated tumors are being treated differently than patients with HPV-negative tumors.

Now, obviously that's not the case in the cervix, but one such example is in the head and neck. Oropharyngeal squamous cell carcinomas are frequently HPV-positive, a very high fraction, and the time in p16 in the oropharynx, the tonsil, the tonsilar crypts, the base of the tongue, et cetera, is very, very well correlated with HPV positivity and overexpression of the E67 octogens as evidenced by this RNA in Situ hybridization. But outside the oropharynx there are cases of squamous cell carcinoma that are p16 positive but are not due to HPV because there are other reasons for p16 overexpression or inappropriate expression. So in certain clinical situations it may be important to test or further test p16-positive or p16-negative results as we gather more and more data about the biology of squamous neoplasia in a variety of sites.

So in our opinion HPV RNA in Situ hybridization is going to be very useful for problematic survival biopsies, particularly at the low end of the spectrum where reactive is going to be HPV-negative and p16 is not useful. In head and neck squamous cell carcinoma, vulvar squamous cell carcinomas, anywhere where the specificity of p16 may need to be improved, we think HPV RNA in Situ hybridization is going to be a very good complimentary test to p16. If not, potentially a replacement test. When investigating carcinomas among certain origin, there are some contexts where p16 just doesn’t work very well. So, for instance, in an in vitro cancer p16 could be highly expressed, particularly in the serious carcinomas of the endometrium, and p16, of course, is also expressed in cervical carcinomas, but there's no HPV in any of the endometrial cancers. Virtually all the endocervical adenocarcinomas that are--a large majority of them, even the unusual types, are HPV-positive.

So with that, I remind you once again about the cervical cancer screening process. What we've tried to do in this hour together is talk about how HPV testing in the setting of primary screening allows for the accurate identification of who needs treatment, who needs to go to colposcopy, and who does not. Focusing the patient on who needs treatment using improved genotypic methods or dual-stain, I think, is definitely in the future. Of the algorithm and p16 immunohistochemistry and RNA in Situ hybridization for HPV, I think are going to be increasingly used as diagnostic adjuncts to help us get the right answer on those tissue biopsies for our patients, and decrease in terms of our variability, which was the basis for the FDA approval recently for p16 as an adjunctive test.

Q&A

First, why don't the clinically valid HPV tests include all the known HPV types that we now know are associated with cancer?

DR. STOLER:

So as I said, the reason we don't include all the types is that compared to the prevalence of these viruses, so those other types that aren't included, let's talk about HPV 72, for instance, or HPV 83. They're probably prevalent at about a half-a-percent in the population, and yet the incidence of cancer due to that is 1,000, 10,000 times less. So we included them, we would drive more and more patients to colposcopy and you wouldn't find any lesions. So you can't--it's this issue of balancing sensitivity and specificity. Epidemiologists and pathologists have worked long and hard over the last two decades to try and get the right balance and recommend those HPVs that should be tested for because they lead to clinically actual information without destroying the specificity of what you're doing.

A biopsy of a head and neck squamous cancer is strongly p16-positive. Why would I need in Situ hybridization?

DR. STOLER:

So again, it depends upon where you are in the head and neck. So it is true that many p16--there are HPV-associated cancers that occur, for instance, in the larynx or the posterior pharynx, or even in the oral cavity. But they are a very small minority of the cancers, and what we found in studies in these other sites is that the p16 positivity is not as highly correlated with HPV-associated neoplasia as was once thought. So the recommendations vary amongst laboratories. So, for instance, in our lab we don't do p16 unless it's an oropharyngeal cancer unless it's specifically requested on a protocol for a specific clinical question because of that false positivity. So if you're in the consult and your p16 is positive, you're right, you probably don't need HPV in Situ hybridization, but if you're in the larynx or in the posterior pharynx or in an unusual site, p16, the specificity of the result is not what you think, and HPV in Situ hybridization would give you that specificity that's going to be required.

Patients have been known to clear their HPV infection; what rate of false positive have you seen where the HPV RNA ISH is positive but no morphologic lesion is seen.

DR. STOLER:

So that's a great question, and that's when the real power, I think, of RNA in Situ hybridization, because the RNA in Situ hybridization is evidence of transcriptionally active HPV. We actually did a study a long time ago about the time course of our RNA transcription relative to morphologic abnormality. Sufficed to say that morphologically normal virtually never has HPV RNA identifiable, even with these high sensitivity techniques. So fundamentally if you see the RNA in Situ hybridization positive, you should be able to find areas that people will agree are morphologically abnormal. The false positive rate in reactive or negative lesions is very, very low.

Last question. What was the indicator that prompted the retest of the sample once it was misdiagnosed using the H&E slide to test for a p16 marker?

DR. STOLER:

So, I'm trying to understand what the question is. In the originals studies looking at the utility of p16 in adjunctive diagnosis, the correlations were done independently. That was one of the real power of the study to go back to, for instance, the Gongano [phonetic] study in the American Journal of [inaudible] which we did, and the subsequent studies that went to the FDA, the finding is that if you're going across a threshold of what you think is an HPV-associated high-grade lesion versus non-HPV associated high-grade lesion, the p16 breaks with adjudicated pathology makes the individual pathologist read like a panel of experts. That's really the simplest way of thinking about it. Not sure if I answered the question, but I think that's the relevant concept.