Prostate Cancer: Using 1yr and 2yr PSA Metrics to Improve Outcomes and Limit ADT
If we are considering de-escalation of ADT in unfavorable disease, we need to critically assess short term outcome metrics beyond just the Phoenix definition.
In my post last week on Prostate Cancer - Can we de-escalate “better”?, I walked through our history of dose escalation and the data supporting the “preferred” hypofractionated approaches. I argued in that review that, back in the day, our ability to dose escalate was limited by technology. But today, with improvements in technology and a better understanding of risks, we can push higher. With this historical context, I argued simply that we should:
Push dose higher and get to 90% or higher control rates at 5-7 yrs for as many men as possible using today’s technology and avoid ADT in those men where that level of local regional control can be achieved. That, to me, is a better path forward.
And today this is even MORE important with EMBARK trial data released March 16th: Just as a reminder, SoC for local failures is moving towards ADT + XTANDI. Pre-trial XTANDI was a 2.6B US drug. Does it help prostate outcomes in recurrent disease - appears to yes but also recall no drug is without toxicity: Per Pfizer press-release 1.7% additional seizure risk, 10% risk of Ischemic Heart Disease, 6% higher risk of fractures and in the AFFIRM trial 16% discontinued this drug due to acute toxicity.
To me: Today it is even more important to cure more men the first chance we get. We should be using safe proven dose escalation approaches to optimize our initial treatment courses.
So that thought leads to the premise of this article :
Can we use PSA to more quickly sort out the pretty good from the really good options? What dose or PSA metric is needed to hit 90% or 95% LRC for a subset of patients?
NCCN radiation guidelines for unfavorable intermediate risk disease lists about 16 different fractionation regimens. Seriously?
For us and for our patients, we need to do better and at least eliminate some of these - narrow the field. No place in life are 16 options truly equivalent. Imagine if there were 16 different advertised ways to give concurrent cisplatin or a review of 16 cars said they were all essentially the same. It doesn’t pass a common sense test, yet that is what we promote.
And even worse, from a technical standpoint, the ones listed as “preferred” in the NCCN guidelines are significantly lower BED than other listed options - I can only assume that is because they are shorter / cheaper - as if we, ourselves, don’t think radiation dose is important. (And for prostate cancer we have decades of data showing that dose escalation is beneficial - this is NOT lung cancer).
And to support my argument further, look simply to PACE B where they are comparing 78 / 39 fractions to 62 / 20 fractions to 36.25 / 5 fractions. The trial treats low risk and intermediate risk men - specifically excluding 4+3 disease. So it will cover some lower risk UIR disease - mostly FIR disease. They do not allow ADT.
Why 62 in 20 from the same group that standardized 60 / 20? No ADT. And even in this pretty darn favorable group (excluding many UIR patients), when they amended the protocol in 2016, they KNEW that 60 / 20 isn’t enough dose for many of these patients. These are really smart people. As I pointed out in that earlier discussion, 60 / 20 is the LOWEST BED of the hypofractionated arms. Why it is still a “preferred” approach in the US is beyond me. Experts across the pond understand it is a compromise whose benefit is shorter / cheaper - not best for curing the cancer.
And it is in these types of patients, more specifically the unfavorable intermediate risk prostate cancers, I argued that we should be able to significantly reduce ADT usage IF we can get DFS high enough. So rather than quick treatments, we need the BEST approaches to cure the disease. And with the recent ProtecT trial data (ref 1), we need to work on reducing treatment morbidity more aggressively today as it is clear that active monitoring is a viable solution as I discussed in my prior post here.
To do this safely and efficiently, solid early metrics are required if we are going to change our approach. To discuss the risks, you need to know your outcomes and your expectations and short term metrics become critical to assessing your results and few physicians sit in one place for a decade to compile their prostate results.
So we have a end goal we need to achieve and now the question becomes…
Simply stated: how can we clean up the plethora of options that we have created. :)
The Phoenix Definition:
A second Consensus Conference was sponsored by ASTRO and the Radiation Therapy Oncology Group in Phoenix, Arizona, on January 21, 2005, to revise the ASTRO definition. The panel recommended: (1) a rise by 2 ng/mL or more above the nadir PSA be considered the standard definition for biochemical failure after EBRT with or without HT; (2) the date of failure be determined “at call” (not backdated).
First off, its from 2005 and based on old thoughts and old data. (ref 2) Now nearly 20 years old and much has changed. In ways, congrats to the original designers! It has held up remarkably well. It still does a lot of things right but in ways, from my perspective, today we need to support it with additional PSA metrics.
Four main issues have changed since the development of the Phoenix Definition:
1) We have good salvage options beyond simply ADT - think SBRT for oligometastatic disease.
2) We have a new technology in PSMA that will detect failures earlier and more precisely than CTs and bone scans alone.
3) Our options for primary radiation treatment are all over the map from 8 wks to SBRT/implant approaches with different risk / benefit profiles.
4) Early salvage options are improving in ways but are becoming more costly and more intense / toxic. (XTANDI now considered much earlier)
So can we just tighten up Phoenix now that we know more and treatments are generally to higher doses?
There is a great SBRT dataset that looks specifically at this question.
Refining the definition of biochemical failure in the era of stereotactic body radiation therapy for prostate cancer: The Phoenix definition and beyond (Ref 3, top tier reference residents should know).
They answer the question elegantly in a 2000+ patient multi-institutional dataset with over 7 yrs follow-up. I think the answer will hold up here in longer fractionation due to PSA bounce data being quite similar (more thorough review performed but not shown here today):
The Answer is: No.
From their work:
False positivity of nPSA + 2 was 30.2%, compared to 40.9%, 57.8%, and 71.0% for nPSA + 1.5, nPSA + 1.0, and nPSA + 0.5, respectively.
So in simple terms, if you tighten up Phoenix to catch more failures the false positive rate jumps up significantly (at least pre-PSMA scans). The paper then goes on to describe additional metrics that can be used to tighten up your confidence in a Phoenix based definition. It is very well done, but it isn’t simple to remember and I’m not sure it will broadly apply to all treatment approaches - it might work in all situations, it might not. (it is a great SBRT dataset and it is written up very well). But then, that same month PSMA imaging was broadly approved in the US.
Therein lies the issue with journals and publications - they are stagnant once pushed and just as it was printed, PSMA was released and now the context has changed. That shift changes the questions and potentially the answers. (One reason why I like this format online - addendum and notes are possible every day to keep things on point or closer to current).
And clinically, PSMA scans change a TON of clinical decisions. As a physician, the bar used to be you could order a bone scan or CT but the odds of finding failures with really low PSAs bumps was quite small - the tests just were not that sensitive. Your other answer was to declare failure and start them on ADT which brought toxicity and major quality of life chances for something that “likely was” a recurrence. But today, the option is far easier. PSA is up a bit and you do a PSMA scan - negative and follow, find something - take action. And if your patients PSA increases 2 over nadir, many physicians will recommend ADT + XTANDI. The world of follow-up is now different.
(Whether these extra scans and more aggressive early treatments really help is a different question, but I would argue that is why it is even MORE important today that we push for the most clarity / best results with our primary treatment course.)
Today, I believe we need to consider new metrics in addition to the Phoenix definition to help assess patient cohort outcomes at a shorter time interval:
Three main reasons:
To more quickly compare and contrast the 16 options we have. Options are ok, but we need to figure a few “best” practices and refine those. At this point, we’ll never clearly move forward. There will be comparisons to different approaches. in effect, a blurred wall of data that hopefully moves in the correct direction, but that path will be far less clear than progress along a single sequence of comparisons. We are replicating this blurred data wall in SBRT due to shear number of approaches. To me it is poor quality science.
And we will need to push for better outcomes because…
Well first off, that is what we owe patients. We need to clarify the best outcomes and pair those with minimal use of ADT.
PSMA will soon dominate our failure patterns. Well before Phoenix officially calls the failure, PSMA will. Eventually that will undercut our definition and we’ll be looking back to earlier metrics regardless - so we should move first.
ProtecT data / EMBARK data: quite clearly modern day prostate cancer has limited long term survival risks and my argument is that we should reassess our old habits before routinely adding ADT toxicity given our options to dose escalate via improvements in technology. And at the same time, Phoenix failure means more and more systemic treatment with greater costs and greater toxicity risks.
The Path There:
Step 1 in my mind is to attempt to look at traditional fractionation down to say 20 fractions and find early metrics that correlate with long term outcomes. There are a few outliers in the implant / SBRT world that will take more time to review so I need to split these up into a separate discussion.
By no means is this a new idea. In fact, here are three older studies looking at earlier time points - I think this is back when we were trying to figure out the dose question more precisely. These three datasets are all “clean”. There is no ADT use which lets us accurately assess the radiation kinetics.
Study 1: Postradiotherapy 2-Year Prostate-Specific Antigen Nadir as a Predictor of Long-Term Prostate Cancer Mortality
Dr. Zelefsky with data out of MSKCC in 2009 (ref 4). 844 patients. Median follow-up of 9.1 years. The prescribed dose levels ranged from 64.8 Gy to 81 Gy.
that nadir PSA ≤1.5 ng/mL at the landmark was an independent predictor of progression-free survival after adjusting for T stage, Gleason score, pre-RT PSA value, and RT dose (p = 0.03). The 5- and 10-year cumulative incidences of DM were 2.4% and 7.9%, respectively, in those with nadir PSA levels ≤1.5 ng/mL at the 2-year landmark, and were 10.3% and 17.5%, respectively, in patients with higher nadir values.
The 5-and 10-year PSA relapse-free survival rates from initiation of RT for patients with nadir PSA ≤1.5 ng/mL and >1.5 ng/mL were 85% and 63% vs. 72% and 52%, respectively (p < 0.001).
This trial used a wide range of doses and 1.5 ng/ml at 2 yrs was strongest - but in today’s world, I don’t believe 5 and 10 yr results of 85% and 63% are strong enough. So trial one says lower than 1.5. (Note: 85% is exactly the Canadian / Australian / French 5 yr results with 60 / 20. At 10 yrs, I’d expect similar decay towards at least 70%).
Study 2: Early prostate-specific antigen (PSA) kinetics following prostate carcinoma radiotherapy: prognostic value of a time-and-PSA threshold model
This series is from 2004 and looks at the Cleveland clinical experience (ref 5). The treatments are OLD school - 1987 - 2000. 839 patients. Most common was 1.8 to mean of 68.4 Gy but also included the early 70 Gy / 28 fraction approach (a “preferred” regimen today).
The median PSA nadir … was 0.60 ng/mL (range, 0.1–159 ng/mL). The median time to nadir was 23.8 months (range, 2.0–124.9 months). Patients with sustained bRFS had a median PSA nadir of 0.4 ng/mL (mean, 0.81 ng/mL) and a median time to nadir of 28.9 months, whereas patients who experienced biochemical recurrence had a median PSA nadir of 1.3 ng/mL (mean, 3.9 ng/mL) and a median time to nadir of 15 months (P < 0.0001 for both nadir level and time to nadir).
During follow-up, 769 (91.7%), 730 (87.0%), 601 (71.6%), 412 (49.1%), and 211 (25.1%) patients did not reach PSA threshold values of 3.0, 2.0, 1.0, 0.5, and 0.2 ng/mL, respectively; PSA levels below these thresholds at any time during follow-up were associated with observed bRFS probabilities of 69.4%, 72.1%, 79.4%, 84.7%, and 91.0%, respectively.
Crossing a PSA threshold of 3.0, 2.0, or 1.0 ng/mL by 3 months predicted long-term bRFS rates > 85%, > 90%, and > 95%, respectively.
This study gives a very short 3 month nadir of 2 linking to 90%. It also describes long term nadir values but those can come beyond 2 yrs. And according to this data, ultimate nadir values need to be in 0.5 to 0.2 level to be in the ballpark of 90% (closer to 0.2 ng/ml).
With our very short SBRT approaches and even 20 fraction approaches, I think moving towards such a short intermediary time frame of only 3 months does not make sense. One would think that with more time, the variance would settle down and ultimately be more consistent across more current approaches - at least that is my thinking. That said, it points to pretty low nadirs for long-term control and argues again that a nadir over 1 results in poor control (mean nPSA was 1.3 and 65% remained bRFS - older ASTRO definition < was stricter than Phoenix).
Study 3: Prostate-Specific Antigen Nadir Within 12 Months of Prostate Cancer Radiotherapy Predicts Metastasis and Death
This is a study of 1000 men out of Fox Chase with patients again treated a LONG time ago - 1989 - 1999 but to 76 Gy (ref 6). The median nPSA12 achieved was 1.2 ng/mL. With a median follow-up of 58 months, 26% of patients had BF (n = 260).
Median nadir is quite low but failure is quite high. I think this is a reminder that when we look back this far, techniques were very different. We see this in high dose MSKCC data and again here. Not apples to apples comparisons with today’s approaches. Likely the posterior prostate was purposefully under dosed to protect the rectum.
Still three big series say above 1 isn’t good and long-term control needs median nPSA of at least 0.5 ng/ml (and likely closer to 0.2-0.3 ng/ml) to hit 90% DFS at 5 years - often this happens at 3-4 yrs post treatment.
So we just need to get back to what we once did:
Report and evaluate PSA kinetics.
These all show “earlier to lower” and “lower overall” is better. And when failure rates get high enough and the nadirs don’t get low enough, it can even result in higher DM rates. This is exactly what we want to avoid as we de-escalate the ADT in UIR.
Unfortunately, in my opinion, these metrics fell away from use in the large non-inferiority trials so those trials just don’t present this data. (that’s why I presented three very old datasets.) I understand why they did at the time (kind of) - you just wanted to demonstrate good enough, Phoenix continued to work well defining failure risk, no good next tier options other than ADT. (If I missed something, please leave a note - author of 1).
But things have changed significantly on many levels and we’ve discussed and now is the time to reassess. To me, I think it would be of great service to our field to publish this type of PSA kinetic information in these more recent “standard of care” level publications. Even in the COMPARRE trial I do not see language regarding a planned PSA early metric evaluation analysis at a given pre-determined time point. (that would NOT justify protons mind you, just a planned early marker)
I understand that BED dose comparisons keep getting fine tuned may end up being a pretty reasonable substitute, but PSA is far closer to the patient and accounts for nuance in technique. And ultimately, it is PSA levels and patterns that determines a large portion of a prostate patient’s cancer journey. We need PSA kinetic data.
Back to the SBRT 2000 patient database:
In that series - 2000+ patients that achieve a median nPSA of 0.16 (mean at 4 yrs is just over 0.2), here are kinetics - weighted average (non-failures / failures) - weighted is something I calc’d so not a number in the paper, but readily calculable due to high level presentation of data. Well done authors!!
weighted average (non-failures / failures)
1yr: 0.95 (0.90 / 1.75)
2yr: 0.37 (0.33 / 1.00)
3yr: 0.27 (0.24 / 0.86)
For the series, Median follow-up was 71.9 months. BCF occurred in 6.9% of patient. It included Low (46%) and intermediate risk (FIR 32% / UIR 22%). And it achieved 93% DFS - excellent! This is a clear standard option.
At 2yr, PSA mean was .037. It is another reference point for us to utilize and it demonstrates that, similar to the above data, 0.2-0.3 for a long term average nadir is an excellent result consistent with where I believe we should be.
My back of the envelope calcs based on these studies and several others:
We should expect mean PSA<1 at one yr and <0.5 at 2 yrs for the treated cohort to exceed 90% bDFS WITHOUT ADT.
This should result in an ultimate cohort nadir around 0.3 or 0.2.
I waivered between the two year metric being 0.6 or 0.5. There just aren’t many current data points. And then at year 1, there is some data that supports SBRT maybe having a slower slope - this large multi-institutional SBRT reference does not clearly demonstrate that but the single institution Fox Chase data does (ref 7). I integrated implant kinetics into this estimate but haven’t included them here for brevity and because I need to return to implant / SBRT later.
But I think this is one path. Establish a set of PSA kinetics, that when met via some treatment approach, gives a high level of confidence that long-term cure rates are really strong. Today I think that floor should be at least 90% at 5 yrs for UIR. In most of the current series, if you achieve 90%+ control at 5 yrs, the subsequent failure rate annually is very low and many series remain over 90% or fall to 85% at extended periods of 8 or 10 yrs.
FLAME achieves 92% 5 yr DFS in the SIB arm including many high risk patients without uniform ADT use (ref 8). I dug through the trial publication - including the supplemental data and don’t see PSA nadir or kinetic data
(HINT FOR ANOTHER PUBLICATION INCLUDING FLAME PSA KINETIC DATA).
There are other viable options for early kinetic endpoints:
I think the other metric would be percent of men who achieve 0.5 nPSA by year 1. For example, the Cleveland Clinic data above used this type of approach. I think that works in ways but I found less data. Same for very low nadirs such as 0.2. Some series have used 0.2 as a definition of cure so we could evaluate a percentage of men reaching that value at an early time point. But each of these approaches measures the “best responders” and not an evaluation of the entire cohort.
In the end, I opted to measure the results of the entire cohort, not just the best responders. And along those lines, in my proposal, I opted for mean because I purposefully wanted the worse responders to have more impact. We are aiming for 90% control so we need the failures to be limited. Using mean over median helps to highlight and limit the number of those poor responders. At least - that was the thinking.
Based on my review of the data, I predict that hypofractionated “preferred” regimens should not meet these metrics for unfavorable disease without ADT use. I agree with the NCCN guidelines in that respect. If you use moderate radiation dose schedules for UIR, you will need generous ADT usage to approach 85%-90% 5 yr DFS. I do not think it is wise to de-escalate the ADT if that is your path. I’m just illustrating the data for a second approach.
If we can simply published PSA kinetics from the large non-inferiority trials I reviewed previously or from the FLAME trial, this entire piece could be refuted or confirmed and that, in and of itself, would move prostate cancer treatment forward. (Hint to any residents / young facility with access to these databases). Without more data points, I have simply presented an estimate largely based on older data and older techniques. It is much like argument for ADT in UIR patients today in that regard - really no different. But we should strive to be more precise.
My Summary:
Despite the uniform recommendations of ADT in unfavorable risk disease without implant by NCCN, I believe the data argues for a different approach. If you are delivering treatment without ADT and your patients are nadiring at an average of less than 1 within 1 yr of treatment on average and by 2 yrs that average is less than ~0.5, I believe you long term control will approach or exceed 90%. I believe this is supported by large volumes of historical data that counter the old studies using inferior dosing referenced by NCCN to argue in favor of combination therapy. To me, the entirety of the data suggest a better way to de-escalate and improve our patients quality of life.
In my personal database, ~25% of my UIR receive ADT and that percentage is falling as I continue building experience. I’m trying to mirror the University of Florida approach and for the UIR patients treated without ADT, my 1 yr mean nadir is 0.67 (updated 3/17). Year two mean nadir is well below 0.5 and trending lower and both should trend lower with longer follow-up IF current trends hold. Small numbers, short follow up but with more time and as the database grows, I’m more confident - at least through 4 yrs in OKC.
I understand that some high risk end of UIR may still potentially benefit from ADT, but my point is simply:
Let’s get back to focusing on delivering radiation to high safe doses and limit the use of ADT.
Today, we are taking a different path outside of SBRT and implants - we are choosing and guidelines recommend lower BED doses. We then recommend covering up mediocre radiation with huge volumes of ADT even when technology affords different options - at least from my perspective.
And let’s be clear - in the management of prostate cancer - for the vast percentage of men treated with radiation - the worst outcome is a recurrence of their disease. Second is ADT toxicity and a distant third is risk of GU or GI Gr3 toxicity. We truly have made that much progress in limiting toxicity and those great gains our field has made affords different options.
REFERENCES:
Fifteen-Year Outcomes after Monitoring, Surgery, or Radiotherapy for Prostate Cancer
https://www.nejm.org/doi/full/10.1056/NEJMoa2214122Defining biochemical failure following radiotherapy with or without hormonal therapy in men with clinically localized prostate cancer: Recommendations of the RTOG-ASTRO Phoenix Consensus Conference
https://www.redjournal.org/article/S0360-3016(06)00663-8/fulltextRefining the definition of biochemical failure in the era of stereotactic body radiation therapy for prostate cancer: The Phoenix definition and beyond
https://pubmed.ncbi.nlm.nih.gov/34774650/Postradiotherapy 2-Year Prostate-Specific Antigen Nadir as a Predictor of Long-Term Prostate Cancer Mortality
https://www.redjournal.org/article/S0360-3016(09)00045-5/fulltextEarly prostate-specific antigen (PSA) kinetics following prostate carcinoma radiotherapy: prognostic value of a time-and-PSA threshold model
https://pubmed.ncbi.nlm.nih.gov/15221994/Prostate-Specific Antigen Nadir Within 12 Months of Prostate Cancer Radiotherapy Predicts Metastasis and Death
https://www.ncbi.nlm.nih.gov/pmc/articles/PMC1892752/Early Prostate-Specific Antigen Kinetics for Low- and Intermediate-Risk Prostate Cancer Treated With Definitive Radiation Therapy
https://www.practicalradonc.org/article/S1879-8500(21)00204-6/fulltextFocal Boost to the Intraprostatic Tumor in External Beam Radiotherapy for Patients With Localized Prostate Cancer: Results From the FLAME Randomized Phase III Trial
https://pubmed.ncbi.nlm.nih.gov/33471548/
Thanks for reading. All the data here assumes NO ADT use in the patient population. If you add ADT, there is data suggesting that you need to reach a PSA of 0.1 within 6 months of the completion of radiation for the "best" outcomes. As you point out, kinetic data with and without ADT is vastly different. This is arguing that if, as a physician, your PSA kinetics for your patients are good enough, you can safely avoid ADT for many men where it is recommended - my opinion and one that I review with patients going over the data and toxicity.