Read the Full Video Transcript

Christopher Logothesis: I would also like to thank the organizers for giving me an opportunity to educate myself on the topic. I didn't feel expert and I was wondering why I was asked. But there's a certain amount of clarity when you look at topics without being an expert at them that provides a unique look at this that I found very useful. And you'll find that many of the biases that I had regarding the topic of racial and ethnic distribution and geographic distribution really were unsupported and I think that I'm among the many who were intellectually lazy and didn't examine our assumptions very close. 

So the challenges in prostate cancer, the worldwide challenge, is an emergent problem in as much part because the illness, the prevalence of the illness, the impact of the diagnostics and the costs as the illness itself. And I've divided the topics up into three areas of concern. Are there international differences in reported incidence in mortality? And are the reported differences subject to ascertainment bias influenced by pattern or practice, access to diagnostic resources and treatment resources? Are geographic detected prostate cancers biologically distinct by region or by race? And should the differences in ethnicity be determinants of care in individual patients?

There's a huge bias that has exist in assumptions that somehow we should approach these patients differently with the intent, because of the risk to development of disease or the risk of development of a lethal disease and the question is, is that assumption supported? So, on this first slide, I just want to show you the age-standardized distribution of prostate cancer incidents above, red being the highest and the dark blue being the lowest and then the mortality rate below.

And what you notice that's pretty unique to this disease, but common among diseases with low penetrance is that there's a marked difference between incidence and mortality and that difference is non-randomly distributed by sites if you look at the United States, Canada, and compare incidents to mortality, that difference is much more striking than in perhaps areas such as Southeast Asia where that difference is much less prevalent.

The second thing is when you look is this specific to prostate cancer or is this a broad malignant disease phenomenon? And what you can notice again in the purple above the prostate cancer, there's marked differences in the incidence of prostate cancer prevalence, but if you actually look at other disease types, the differences are much less. So, there appears to be this exaggerated difference in incidence in mortality that seems to be more prevalent in prostate cancer relative to other neoplastic diseases.

Now, there is this well-known distribution of the incidence of the disease by disease type across here and you can see there's a very, very wide range in distribution of the disease, so it is a fact that diagnosed, detected prostate cancer has a marked difference in incidence across geographies. There is less of a difference, it appears, in mortality and the question is that biological or is it not?

Now, you see on this slide is a different phenomenon and this is basically entered with the deployment of PSA over time, and if you look at the differences between incidence and mortality in the era through PSA, this is change over incidence in mortality. The biggest changes occur in the patients in incidence and much less in mortality. Again, this would be in favor that there's a detection bias that accounts for this difference and there's less of an effect on either our therapeutic interventions or biology. But there's a clear greater influence over time in the change of the prevalence of the disease.

Now, I'm using this slide as a placeholder just to describe the large European and the PLCO studies, had extensive screening effect studies and basically the findings are in line with those observations indicating that if you screen large populations, you will have a modest effect on outcomes of survival in prostate cancer by pure PSA screening. These are large studies, so there is a series of data in line that prevalence is much more vulnerable to change over time and there's less of an issue of mortality change over time and that screening in line with this, doesn't seem to have a proportional impact if you compare it for example to uterine carcinomas, if you compare it to lung cancer, if you compare it to colon cancer. So, we have a unique phenomenon in prostate cancer, a vulnerability to detection bias that appears to be present.

Now, the incidence in mortality again, there's greater differences. I'd like you to focus on the differences between the incidence and mortality and this is not now by country. Countries have relatively arbitrary borders, as you know, and not ethnicity and even by regions, geographic regions not confined by the borders of the country. You see this differences in mortality and incidence that is striking and again, relatively unique to prostate cancer.

So, the conclusion to the first question is there are genetic differences in reported incidences compared to mortality by geography and change over time, but this we don't know if it's related to an inherent bias in ascertainment or a real, biological basis. So incidence and mortality are two very different diseases in this tumor, which overall has a low penetrance and marked heterogeneity in its penetrance.

Now, looking at this slide as we're looking at the impact of survival over time versus incidence over time in these patients. And what you can see by country, again, there's greater influence over time and this is here that you can see by year, this goes on early, in mortality the lower number than incidence, which seems to increase. There's one notable exception here and that exception seems to be the Republic of Korea where the mortality seems to be increasing proportional to the incidence, and it's a unique standout among countries. All the other ones seem to be going with a disproportionate increase in incidence without a proportional increase or a slight reduction in mortality, again raising several issues on the significance of the findings.

Obviously, the next question is are these specific to ages or age-specific vulnerabilities? And here what you can see is the age distribution on incidence and it increases similarly across all. So there's no geographic specific age distribution difference that we can see. Some people have inferred that young prostate cancers may be different than patients who get older prostate cancers. This data would suggest that difference, if it exists, does not apply geographically. And then, there's the issues of mortality, again and what you can see is that you have to be very careful about assigning mortality, because once you're diagnosed with prostate cancer, whether it's cause of death or not, the death certificate often states prostate cancer. So, there is a slight increase in mortality, but the changes in mortality are not as vulnerable, not affected as much in the PSA screening or influenced by age. Again, a very profound effect.

So, there's observed in distinct trends specific to incidence in mortality, the difference between the two, across geography or race within geographic areas as well that suggest that ascertainment biases and available therapy may account for some of these findings and these findings need to be understood and I would argue pose a major problem for many of the communities because they suggest a great risk for widely deploying these tests of excess interventions, excess detection, excess concern without proportional benefit.

Now, this is just taken again as a placeholder. This is Collin Pritchard's work in the New England Journal, which basically tells us all patients who get prostate cancer are not the same. Some of them have alterations that are deleterious that can affect probability of responding to specific therapies and have outcomes. So, one of the questions can say do we have evidence of biological difference in any of these, do we have evidence of a biological difference reflected in mark or difference in any of these disease types? And one of the things that comes across all these studies is their remarkably limited data to actually tell you and probably powered, uniform genetic studies with proper coverage to actually support the notion that there are differences that are geographic-specific or ethnic-specific.

Now, I listened to Ros Eeles' this morning and I know that's going to be revised. So, the issues are do we have an impact and how broad are these studies in what they can make? But from the available data ... so, these studies each have their limitations, but I would like to point out one study that I found very intriguing as I looked at this. The other ones tended to be underpowered or the platforms used had limited coverage. And this study is the study ... it's a pan tumor study and it basically looked for mitochondrial metabolism and changes in metabolic across disease space normalized to ethnic backgrounds.

And this is many different disease types, prostate cancer, laryngeal cancer, and oral cancer, and what it shows is across African-Americans, it is different than Caucasians. So, there may be a basic biological difference in metabolic pathways that crosses these disease types, that's vulnerable to what I'm going to call the exoenvironment that gets changed that can really be influenced that can affect this, but this appears to be consistent with the notion that there are general modifiers of disease behavior that can happen across this.

So, are there geographically, ethnically, nationally linked differences? The answer is yes, they exist. Do they interact with the specific environment or the specific ethnic background in a way to cause a specific disease? Unknown. There's huge limitations to those studies published to date and I don't think they reach the level of evidence that you can act on this.

Now, perhaps the biggest concern is the following: what do we tell countries and people responsible for health and what do we tell patients and physicians on how to behave with this data? So, when I do the factual analysis, incidence in mortality by race seems to change. The specific biologies, we don't know which ones are implicated that drive this. The geographic-specific vulnerabilities are unclear. The thing gave me an extra minute. Somatic alterations and reasons by race don't exist, but if you do the counterfactual analysis, you see a patient, should you change your behavior beyond the staging, the geographic, the molecular changes? And the answer to that is you expose the patient to great risk of mistreatment based on biases that I had before this analysis that you would watch an Asian patient in a different way who was at risk for prostate cancer than you would an African-American patient. It turns out stage, grade, outcomes, normalize all that and I don't think it's justified and you expose patients to greater risk.

So, the conclusion is know there's no evidence to support ethnicity as an independent determinant of clinical phenotype who should influence therapy. I think that that's key. So, how do we go forward? And I put this caution sign here. Unless we benchmark it into disease states that we understand the clinical phenotype, highly AR responsive, highly AR indifferent, aggressive, variant, or maybe this bone-forming phenotype. 

I don't think we benchmark against things and make a difference. We have to benchmark against known things to make a difference. So, the future we have to look at the gene environment interaction. Consider the exoenvironment, microbiome diet, other things that are happening and benchmark to known biological states and control for health reporting and outcomes. So, thank you very, very much.