The Relationship Between a Priori Defined Prognostic Risk Groups and Overall Survival in Men with mCRPC - Susan Halabi

June 23, 2023

Andrea Miyahira and Susan Halabi discuss Dr Halabi's prognostic model for overall survival in patients with metastatic castration-resistant prostate cancer. Originating in the early 2000s, the model sought to provide patients with more specific information about their prognoses and categorize them into different risk groups. Dr. Halabi's model comprises eight easily obtainable clinical variables, such as ECOG performance status, hemoglobin levels, and patterns of metastasis. The latest validation study utilized data from seven clinical trials and sought to examine whether the model could still be effectively used on a contemporary patient group, including patients of diverse ages and races. Despite changes in patient characteristics over time, the model maintained its predictive accuracy. Looking ahead, Dr. Halabi envisions the model could be enhanced by integrating more genomic data and applying it to improve the efficiency of patient recruitment for clinical trials.


Susan Halabi, Ph.D., Professor of Biostatistics & Bioinformatics Chief, Division of Biostatistics, Duke Cancer Institute, Durham, NC

Andrea K. Miyahira, Ph.D., Director of Global Research & Scientific Communications, The Prostate Cancer Foundation

Read the Full Video Transcript

Andrea Miyahira: Hi everyone, I'm Dr. Andrea Miyahira here and I'm the senior director of Global Research and Scientific Communications at the Prostate Cancer Foundation. Joining me today is Dr. Susan Halabi, a professor of biostatistics and bioinformatics at Duke University.

Susan Halabi: Hi, good afternoon. Thank you so much, Andrea, for having me today.

Andrea Miyahira: Yeah, so wonderful to have you here. So today we're going to talk about your recent publication in Journal of Clinical Oncology, where your group published a prognostic model on overall survival in patients with metastatic castration resistant prostate cancer and this is a validation study. So you actually previously performed a smaller study where you created this model. What was the rationale for that original study?

Susan Halabi: So this dates back over maybe two decades, when I built the first model which was in the early 2002, patients who had castration-resistant prostate cancer had really terrible disease. The median overall survival was around 14 months, and there were really no models out there. So the idea was can we build a model that will inform patients about their outcomes, specifically about their survival? And more specifically, can we try to categorize patients into different groups based on prognosis? Because obviously not every patient is the same. Prostate cancer is a heterogeneous disease, and at the time the thought was, it'll be good to identify patients who are at higher risk of death, and perhaps those patients need to be treated differently than someone who had a lower risk of death.

Andrea Miyahira: And this is very important because now in the current era, we have so many different medications for patients in this disease state. So what are the components of the model?

Susan Halabi: This is really an excellent question because the model we had previously developed include patient characteristics that's related either to them as patients, like how they're functioning, such as ECOG performance status, hemoglobin. We're looking at patterns of spread in terms of their metastases, but it also included some factors that's known as tumor related factors such as we're looking at alkaline phosphatase, whether the LDH level was higher than the upper limit of normal. So when you look at PSA, alkaline phosphatase and LDH, they're characterizing the tumor itself. So when we look at the model, it really contains eight variables, and those variables are easily obtained from clinical trials or from the clinic and that includes ECOG performance status, whether the patient had pain that required them to have opioid analgesic cues and then we had hemoglobin, alkaline phosphatase, PSA, we had albumin pattern of spread. Basically, these are the variables that were in the model.

Andrea Miyahira: Okay, wonderful that they're easily clinically obtainable also. So in the current study, you're aiming to validate this model, what were the cohorts that you used and what were the larger questions that you were asking?

Susan Halabi: Right, this is an excellent question, Andrea. So what happened is in 2014, we looked at the model, we updated the model, we added a few other variables, but again, the model was validated in patients who were treated with docetaxel. And we had a group of patients from another clinical trial that we've used. However, as you know that the patient have changed over time so we see a shift in terms of patients who are the patients right now with castrate resistant prostate cancer? They're totally different than 10 years ago. And specifically while the model was useful, it was validated in patients who had perhaps a higher volume of disease in 2014.

So what we did is we obtained data from seven clinical trials of which five of those clinical trials used docetaxel as their primary treatment. And then the other two trials included more contemporary type of patients who are treated with abiraterone acetate. And basically we tested whether the model still works in this group of patients. And not only did we do that, we looked at patient accuracy, the predictive accuracy by age, by racial composition, because the previous model was, I would say primarily validated in men of White descent. And we looked at that and overall the model did predict well in terms of predictive accuracy as measured by the area under the curve.

Andrea Miyahira: Okay, wonderful. And what were the main take home messages, the findings from your study?

Susan Halabi: What was really interesting to me is even though the model did very well, you could see there is a shift in terms of patient characteristics. So for instance, in the trials where patients were randomized to docetaxel as a control, those patients had higher volume disease, higher levels of alkaline phosphatase, higher levels of LDH, lower level of hemoglobin. So the patient characteristics were different than the contemporary patients were randomized to later studies. And despite that, the model did well and we were able to categorize the patients into two or three risk grouping that we hope to use that for trial enrollment because as you might know, there are a lot of competing trials and we need to be efficient in terms of designing future clinical trials and identifying the patient to be enrolled on those trials.

Andrea Miyahira: Okay, thank you. And how do you see this model being used by clinicians going forward?

Susan Halabi: I think the model has been primarily used for identifying patients and to balance patient characteristics at baseline. So you want to make sure that when we look at pattern of spread in terms of liver metastases or other lab variables such as higher hemoglobin level, higher LDH and so forth, you want to make sure that the randomization works so we balance that, but we are now trying to improve on the model. We are hoping to include more genomic data and bring in the biology of the disease more on that. So we have an RO1 that's funded by the National Cancer Institute, and that's what we're working on integrating some of the androgen receptor genetic structural rearrangements. And then we're looking at other genetic data that we have from a large randomized phase three trial that was conducted by an NCI, National Cancer Institute, National Clinical Trials Network.

Andrea Miyahira: Okay, wonderful. And what would be your final take home messages for our listeners today?

Susan Halabi: The first one that if a patient is interested in looking at their clinical outcome in terms of overall survival, we can use this model to predict their overall survival. And then the second message would be that you can use the data from their individual predicted survival and based on their patient characteristic, a patient could know or an investigator would know whether the patient is high risk or low risk. Third take home message I would say is it is so critical to have data shared with researchers and investigators. I've waited almost eight years to have access to that data, and the data are sitting there doing nothing.

So by sharing data, scientists like me and others will find this a very useful way of being current in terms of what to do, how to prioritize patient recruitment and enrollment in trials. So I thank all our sponsor, whether it's the NCI or the industry sponsor for sharing that data with us. And I also want to thank the Prostate Cancer Foundation because this work was funded partly by the Prostate Cancer Foundation and the NCI. As a matter of fact, my first model that I've built, I received a young investigator award from the Prostate Cancer Foundation so that goes back over 20 years. So again, I really appreciate the Prostate Cancer Foundation for investing in a junior investigator, investing again in a senior investigator and giving us the opportunity to really do important work so that families and the patients could benefit from our research findings.

Andrea Miyahira: Okay, wonderful. Thank you for joining me today to share this study. And for all of our listeners who are interested in learning more about this study, the link should be below. Thank you.

Susan Halabi: Thank you very much.