Chris Wallis: Hello, and thank you for joining this first URO Today Journal Club discussion. Today we're talking about a recent in press publication entitled Polygenic Risk of Any, Metastatic, and Fatal Prostate Cancer in the Million Veteran Program. I'm Chris Wallis, an Assistant Professor in the Division of Urology at the University of Toronto. And with me today is Zach Klaassen, an Assistant Professor in the Division of Urology at the Medical College of Georgia.

You can see here the citation for this recent publication in the Journal of the National Cancer Institute, which was published online. PSA screening, as most will know, is the preferred approach for early detection of prostate cancer. However, it remains highly controversial due to trade-offs between the benefits in cancer related morbidity and mortality and harms related to overdiagnosis and overtreatment with their associated morbidity. However, we know that despite an inability to identify specific driver mutations, prostate cancer means one of the most heritable cancers, and as such, genetic risk stratification may provide a promising approach to identify individuals most at risk of developing advanced prostate cancer and thus stratify our screening approaches to really focus on those patients at greatest potential to benefit and greatest risk of prostate cancer.

We know that genetic risk scores outperform family history. Now there are clinical risk factors in risk stratifying patients. So the ideal genetic test focuses on identifying patients who have aggressive rather than just any prostate cancer. It estimates an age specific risk given that prostate cancer likelihoods change over the lifespan. Age specific genetic risk could inform individualized decisions regarding PSA testing and screening behaviors. However, in terms of applying these genetic risk scores to our practice today, we have significant limitations because early studies really focused on men of European ancestry and thus trying to roll these out to widespread utilization is at risk of exacerbating existing health disparities and really doesn't inform our practice.

However, more recent studies have included more diverse populations and provides us more generalizable and actionable data. So the polygenic hazard score, PHS, was developed to identify men who are likely to develop aggressive prostate cancer at a younger age. It is derived from a single saliva sample and is strongly associated with the age of the diagnosis of aggressive prostate cancer. Use of the PHS improves the accuracy of conventional screening with PSA, and the model has been both developed and then expanded to optimize the performance across men of all ancestries. So in this study, the authors sought to use the PHS to identify men at risk of metastatic or fatal prostate cancer in the context of the Million Veterans Program.

This Million Veterans Program includes individuals age 19 to more than 100-years-old at 63 Veteran Affairs medical centers across the US. Enrollment began back in 2011 and for the purposes of this analysis, the authors restricted to approximately 600,000 men. The median age of included men was 69 years with an inter quartile range ranging from 59 to 74. This clearly overlaps heavily with a PSA screened population. All participants in the Million Veteran Program provided blood samples for DNA extraction and genotyping. This was collected by phlebotomists and banked at the VA Central Repository. DNA was extracted underwent genotyping. The MVP 1.0 array contains over 700,000 variants and is notably enriched for low frequency variants that are seen in the African and Hispanic populations. This really provides an ability to derive a risk score and to validate a risk score that is widely generalizable. So to compliment their genotype based analyses, the Million Veterans Program includes a clinical data extraction where each participant's electronic health record is integrated into the biorepository. This includes both standard diagnostic codes as well as laboratory values, and then a natural language processing, which is used to review reports to determine the presence of metastatic cancer.

The most recent version of the polygenic hazard score, PHS290 was calculated as a vector product of the participant's genotype across 290 variants with corresponding parameter estimates that were derived from a Cox proportional hazards model. The distribution of PHS290 is presented within the context of the manuscript as histograms as stratified by race or ethnicity. So the authors sought to assess the association of PHS290 and age of diagnosis, age of diagnosis of nodal/distant metastasis, age of death from prostate cancer. For each of these endpoints, the authors derived cause specific cumulative incidence curves and stratified these by risk group. Cox proportional hazardous models were used to identify associations between PHS290 and each of these endpoints both overall in the cohort and within racial and ethnically defined subgroups. PHS290 was operationalized as a continuous variable and then also categorically with thresholds of the 20th quantile, the 30th quantile, the 70th quantile, the 80th quantile, the 95th quantile.

The authors performed a number of comparisons as we highlight here, comparing the highest 20% to the lowest 20%, highest 5% to an average risk group and the lowest 20% to an average risk group. The multi-variable Cox proportional hazardous models included self-reported race, ethnicity, family history, and the PHS290. Race categories were divided into Asian, black or African American, Native American, Pacific Islander, white or other. Ethnicity was defined as Hispanic or not and only considered within the context of white individuals. Family history was defined in a binary fashion as the presence or absence of one or more first degree relatives with prostate cancer.

The Cox proportional hazard models were used to test associations between PHS290 and any prostate cancer diagnosis, metastatic disease or fatal prostate cancer. And additionally, the authors examined the additive value of PHS290 to a base model comprised of race and ethnicity and family history by examining the increase in Harrell's c-index. All models here used a two-sided alpha 0.01, which notably differs from the typical study using 0.05. Confidence intervals were calculated using 1000 bootstrap samples with replacement and the authors performed a number of additional analyses to identify particular subgroup of interest. At this point in time, I'm going to hand it over to Zach to walk us through the results of this study.

Zachary Klaassen: Thanks so much, Chris. So this is the table looking at the participant characteristics for self-reported race and ethnicity groups. And you can see that in terms of all the participants, almost 600,000 with an excellent breakdown by race and ethnicity, you can see over 102,000 black or African-American men and over 420,000 non-Hispanic white men with also a good sampling of Asian, Hispanic, white and Native American men as well as Pacific Islanders. And you can see that this also resulted in a high number of events in terms of prostate cancer diagnoses, metastases from prostate cancer and death from prostate cancer. Not surprisingly, just from a descriptive standpoint, looking at the age of diagnosis, black or African-American men were younger than non-Hispanic white men and other men of minorities including Asians and Native Americans.

This figure looks at the PHS290 score density plots in this Million Veteran Program. And a couple of points here, there was significant differences between all groups with regards to probability, density of the individuals Hispanic and ancestry overlapped with European men, as you can see in the middle. And there was lower density among Asian men in this curve right here and higher density among black men in this curve to the right.

This looks at the association from a Cox proportional hazard model, looking at several of these outcomes and what we can see here, if we look at starting at the bottom of this table, this is for prostate cancer diagnosis. So in all men, in terms of this breakdown that Chris previously described, so this is looking at 80 versus 20 in the PHS 90 cutoff. 20 versus 50 at 80 versus 50 and 95 versus 50. We see that in this high versus low. So 80 versus 20 has a ratio of 5.2. We don't see a difference in the 20 versus 50, but again, we see a significant hazard ratio for 80 versus 50 and 95 versus 50 in terms of prostate cancer diagnosis. And this pretty much held throughout all of the race and ethnicity background breakdowns here. Again, we see significance for metastatic prostate cancer among all except for the hazard ratio of 20 versus 50. And again for fatal prostate cancer, significant hazard ratios for all patients. Again, also seeing this in the breakdown of non-Hispanic white and to a lesser degree, black or African-American men when looking at high PHS290 cutoff versus lower.

This figure looks at the cause-specific cumulative incidents in the Million Veteran Program stratified by PHS290. And so on the left here we see in the legend that the red line is PHS290, 95 to a hundred. So the highest cutoff in the orange is 80 to a hundred, and gray is 30 to 70 and in blue is zero to 20. If we start over here on the right side of the screen looking at prostate cancers, we see a nice delineation with higher PHS290 cutoff, having a higher cumulative incidence of prostate cancer diagnosis as compared to lower cutoffs. Again, a nice delineation for metastatic cancer breakdown and once again, a little less dramatic in terms of separation of the curves, but still a breakdown between the lower, mid and higher cutoffs of PH290 for fatal prostate cancer.

This is a similar looking curve, but essentially looking at more of a race and ethnicity breakdown in the sort of brown color line. This is the white PH290, sort of in the middle of the 30 to 70th cutoff. And then we have breakdown by black or African-American. So dark green is 95 to a hundred, so the highest, light green is 80 to a hundred, black is 30 to 70 and purple is zero to 20. And again, working from right to left. This is a prostate cancer diagnosis. We actually see that the white patients in the 30 to 70th cutoff is about the same as the black patients in the zero to 20 cutoff. And then we see delineation with higher cumulative incidents of prostate cancer for black men with increasing PH290 score. This again holds for metastatic prostate cancer diagnosis. Again, highlighting that the average white risk is about the same as the lowest black risk and a little less clear for fatal prostate cancer. But a similar pattern to what we see for metastatic prostate cancer and just prostate cancer diagnosis on its own.

This looks at metastatic prostate cancer with the same delineation as the previous slide, but looking at age and so age 50, 60, 70, 80, and 90. And again, this looks very similar to the previous slides with an overlap between the average white man and lower PSH290 risk, black men with increasing PHS290 score among black men with higher incidents increasing over increasing age as well.

This is the multi-variable model combining self-reported race and ethnicity, family history, and PHS290. So among these three, you can see the PHS290 breakdown, family history, and then race and ethnicity. I've highlighted the boxes looking at the hazard ratios for PHS290 after controlling through family history and race and ethnicity. And the take home point from this slide is that PHS290 remained an independent predictor of prostate cancer risk, including prostate cancer death when accounting for race and ethnicity as well as family history.

This final slide in the results looks at the Herrell's Concordance Index for self-reported race and ethnicity plus family history plus/minus PHS290 score. So in the middle column here, this is the Herrell's Concordance Index for these clinical endpoints with just including race, ethnicity, and family history in the model. You can see that it's just under 0.6, but when we add the PHS290 to the model, we see it increase in the Herrell's Concordance Index for all of these clinical endpoints, including for 0.7 among fatal prostate cancer and nearly 0.7 for metastatic and prostate cancer diagnosis. So increasing the model Concordance Index when adding PHS290 to the model.

So this study represents the largest and most racially, ethnically diverse independent validation of the Association of a polygenic score with lifetime risk of metastatic and fatal prostate cancer. After accounting for current guideline recommended risk factors including family history as well as race and ethnicity, PHS290 remained a strong independent predictor of dying from prostate cancer. PSH290 also was associated with age of diagnosis of metastasis from prostate cancer. Importantly, men at highest risk of metastatic or fatal prostate cancer are potentially those most likely to benefit from PSA screening, thus adding PSH290 to guideline recommended risk factors improved stratification for these endpoints.

So in conclusion, this study showed that PHS290 stratified US men for lifetime risk of any metastatic and fatal prostate cancer. And this genetic risk stratification was successful within racial and ethnic subgroups in this diverse dataset. Secondly, PHS290 was higher on average among black men who were also at higher risk for metastatic prostate cancer.

The combination of race, ethnicity, family history, and PHS290 performed better than any single risk factor in identifying men at highest risk of prostate cancer metastasis and death.

And finally, predicting genetic risk of lethal prostate cancer with PHS290 may inform individualized decisions about screening and early cancer detection.

Thank you very much for your attention, and we hope you enjoyed this UroToday Journal Club discussion.