Biochemical Recurrence: Does it even matter?
New Study: Biochemical recurrence is NOT a surrogate for overall survival. New considerations? What can we learn from this new study?
Today’s Context: I’m from Arkansas. I like to keep things simple: if things get too complicated, might be best to back up and think more broadly.
A recent study evaluated biochemical recurrence as a surrogate for overall survival. It found biochemical recurrence failed to meet surrogacy criteria. And yet, here on this site, I talk about PSA kinetics quite a bit and local control. Today, we’ll look at this new paper and see if my thinking needs to be readjusted.
About 10 sentences into the paper, I flashed back to this mentor. A wonderfully bright physician - old school in every way. I recalled one of the first talks I attended back at MD Anderson when I arrived in 1997.
At the time, medical literature was smaller. He used to say he had read every sarcoma paper ever in print, and I believed him. The talk was on prostate cancer - he and Alan Pollack led the GU / Sarcoma service at that time. He kept every prostate cancer patient he had ever treated in his own database. He’d complete a follow-up clinic and run off to his office to enter the PSAs. Nothing complicated - not a team of people - one physician tracking his outcomes - a habit of measuring ones outcomes that I carry on today - learned from him.
In the talk he had survival curves for the patients treated at MD Anderson. I don’t recall exact details but I remember the curves. It wouldn’t surprise me if it wasn’t literally thousands of men treated with prostate cancer at Anderson. 3 curves: Expected survival for all men from a Medicare database, the outcomes of men treated at Anderson with surgery, and the outcomes of men treated at Anderson with radiation. And there might have been a percent or two of difference at various points - but the take home was simple:
Once you get definitive treatment for prostate cancer - your survival is basically the same as if you never had the disease.
I filed it prominently - a key piece of basic information. General context in which to file other pieces of information I would soon learn around. I’ve quoted it at times to patients over the decades and, honestly, until this week didn’t go back and “check it”.
Fortunately, it holds up. Cancer.net - ASCO site: 10 yr relative survival is 98%.
Some 25 years later, I look back and think little has changed. And if he knew that old slide was still out there in the minds of even a few, hopefully he’d smile or laugh, maybe both.
So, with that old story and fond memories of days in training, would I be surprised to find out that ANY test for ANY surrogate overall survival relationship is negative? i.e. can you parse out a 2% (maybe) 10 year relative survival benefit?
Me? Not surprised. Almost to the contrary - nearly all should be negative - even “fancy” genetic ones on the horizon will often fail (this new AI one did for OS).
A 2% difference at 10 years? Hard to see. 98% will be background noise. And this is old data, back well before we had so many promising salvage approaches. We had, up front treatment and ADT. That was it. Compare that to the laundry list of salvage options today. And don’t discount the improvements in the primary treatment. In fact, perhaps all possible choices should be auto-filed as negative to the point that any one off analysis might best be written off as incorrect until validated in a few prospective settings - opinion of one.
Now, let’s look at the study and see what we learn:
Biochemical Recurrence Surrogacy for Clinical Outcomes After Radiotherapy for Adenocarcinoma of the Prostate
It’s a big complex paper with a ton of math. My guess is one could use it for a stats class and discuss topics for about a semester, maybe a year. Suffice it to say it really is quite a bit.
And I do think this. When it gets that complex, with so many calculations, various assumptions and different paths one can take, be cautious. I presume they picked one of the better paths to achieve the “correct” answer, but, from afar, this level of complexity is fraught with opportunity for different assumptions to be made.
That said, I’ll try to summarize some of the findings.
BCR (biochemical recurrence) is prognostic for OS - but not strong enough to be a surrogate. How much value it holds is often determined by how you choose to censor events.
Dose escalation decreases BCR but doesn’t trickle down to OS.
Short term ADT decreases BCR and seems to improve OS. Longer ADT amplifies that benefit.
Here is a bubble plot from the paper showing the general trends.
In the end, the authors main points are BCR doesn’t quite hold up as a solid metric for randomized trials looking at localized prostate cancer and we should consider metastasis free survival endpoints over biochemical relapse endpoints.
Here is the stated relevance of the article by the JCO Editor:
Despite attempt to show potential surrogacy for OS using BCR as a potential marker, these results strongly suggest that BCR-based endpoints should not be the primary endpoint in randomized trials conducted for localized PCa. Metastasis-free survival remains an appropriate endpoint for prospective trials related to radiation therapy in localized disease.*
*Relevance section written by JCO Associate Editor Michael A. Carducci, MD, FACP, FASCO.
Thoughts:
Am I surprised?
No.
Does it contradict my PSA kinetics argument?
I don’t believe so on any level.
Just because I talk about short term metrics that link to local regional control doesn’t mean that I’m one bit surprised to find out that biochemical relapse - while is prognostic for overall survival doesn’t hold up as a true substitution for that ultimate clinical outcome metric.
In fact, I’ve argued the strength of this metric will deteriorate in the era of PSMA quickly if we stay focused on the now nearly 20 yr old Phoenix definition of failure. In fact, how you censor patients and how you monitor them will be incredibly important in the evaluation of any new literature. This paper supports my stance.
They address this issue in the discussion. It is a consideration I’ve been discussing on many occasions this year. We need to get ahead of it and I’m thrilled to see leadership supporting a shift back towards metastatic disease free survival - ie towards PACE standards of disease free survival rather than biochemical failure only.
ADT was stronger for linking to overall survival:
If ADT usage or biochemical failure is the choice: ADT should win as the stronger link to overall survival.
Why? Take an example patient in my database: PSA >70 refused ADT. Failed at ~14 months. That is a bad outcome. One so very bad that it might very well be measured. Early failures have FAR more ability to not being drowned out by the 98% noise in the background. Even the same scenario delayed to year 6 is much harder to “see” within the data - competing risks are real. And ADT pretty much will eliminate these early bad scenarios. So if anything affects survival, it will be the addition of up front ADT.
But with more modern staging and better imaging, we’ll miss fewer and fewer of these types of cases over time. So the magnitude of the findings in the paper should lessen with time.
Hazard rates and expectations:
Last week, we look at broad outcome metrics as we tried to file the proton therapy data.
And now, let’s return to the bubble graph.
This Substack will focus SOLEY on data to the right of that big black line I added. Biochemical recurrence of 50% or 60%? Even for high risk disease - we do far better now. The 70s? We beat that today. 85% is the floor and we should be looking north. Therein lies a tremendous change in perspective compared to the context of the article. We’re back to the ol’ hazard rate discussion we’ve had a few times. Is there still a slope to the line right of my big black mark? Probably. Is there a difference in absolute gain in all cause mortality if you just focus out there? A little, but not much.
Does it change my central baseline information from 1997: Get treated and your survival will be largely unchanged?
I don’t think it does. Get good definitive treatment and you should do well. Add ADT in for high and very risk men and outcomes are strong.
Does it mean we should not use biochemical relapse free survival?
Kind of. And in fact, I’ve been arguing for a year on this Substack. I believe we have allowed standards to shift from disease free survival to this less important biochemical relapse metric. I think the reason is one is more readily available in a spreadsheet whereas one requires good “boots on the ground” work. In general our move towards meta-analysis and similar broad reviews “rationalizes” this shift, but as I’ve argued, it weakens our data. And this paper backs that up.
As I’ve stated repeated, if you fail with bone mets, its a fail. Yet most studies today, shift the goal posts to simple biochemical failure - Phoenix definition (and then often, we don’t provide long-enough follow-up for the KM point reported). I dislike the shift and have voiced that here repeatedly.
What does this do to kinetics?
Here, I thought there was a chance, that in the details, I might learn something and tweak my thinking, but I don’t see any data suggesting what I’ve seen the past is wrong. To me, this doesn’t change anything. I use kinetics as a local control surrogate - I firmly believe a cohort that nadirs at 0.3 at 2 yrs compared to a cohort that nadirs at 1.2 at 2 yrs will do better. They will have less recurrences - both via PSA Phoenix definition and less documented PSMA recurrences and yes, less metastatic disease.
And therefore I believe that local control is important. This is where I’m less convinced that biochemical relapse is a bad primary objective in certain scenarios - I would have written that “Relevance” section far softer. But bold statements make bold headlines and get a few more readers. Is using biochemical failure ideal? No. But don’t throw it out completely.
Biochemical failures lead to work-ups. Work-ups in the PSMA age will find metastatic disease quite often which directly leads to salvage therapy. Salvage therapy, even if it is streamlined and successful is a stressful path for the patient.
As we know, salvage therapy, even if successful, requires a lot from the patient. Is SBRT pretty easy compared to salvage Doxetaxel? Ah… Yes. Not all salvage approaches are equal, but they all represent a failure of the primary treatment and therefore is pertinent and I believe an appropriate clinical outcome metric.
And the same line of thinking continues to make me think there is real benefit in kinetics. I make no argument for kinetics directly linking to OS.
Am I positive that 1 yr and 2 yr kinetics are the best (non-ADT patients)?
No. There is great long term data for 4 yrs. It wouldn’t surprise me that longer metrics are more reliable but that is a 2 full years longer to reach the “answer”. I think it is important for our field to choose an early strong predictor for local control so I stick with:
Mean PSA<1 at 1 yr and PSA<0.5 at 2 yrs to reach 90% DFS at 5 yrs with good followup (non-ADT).
(note: I have one patient - who is making me consider a tweak to the mean rule. HR - refused and has a 40+ PSA early fail. Might need to tweak to exclude highest / lowest 5-10%)
And time does matter. We need tools and data to fight back against localized treatment approaches and even the whole gland TULSA-PRO approach. Years matter in a competitive business landscape.
And remember: metastatic disease free survival is different today than prior to PSMA scanning.
So while this study seems to point to metastatic free survival as the “gold standard”, that gold standard is different today than just a few years ago. A few years ago it meant we found disease on a bone scan or CT. Now we find it, well, differently (too long of a discussion for today). And approaches with SBRT in the oligometastatic setting and/or via systemic approaches that almost certainly lessen any overall survival difference we have seen in the past. In medicine the bar we must clear is always moving.
In fact, today, I’m pretty sure that there will be less of a link between the development of metastatic disease and overall survival - likely moving its relevance more towards the relevance of biochemical failure we see in this analysis today. An ironic thought considering the boldness in the “Relevance” statement I included above.
What does the article add?
These are huge and important topics and it is always great to consider and evaluate with an analytical perspective. Broad perspective is critical to accurate assessment of our literature.
Look ahead just about 9 days to ASTRO and the presentation of PACE-B data. This information should impact your assessment. After all, the primary outcome of that is in fact - biochemical disease free survival. An outcome metric that this paper rather strongly argues is not appropriate for a randomized trial in localized prostate cancer. So it is quite important on levels and hopefully just re-iterates the complexity of medicine while attempting to make broad conclusions that we, as clinicians can apply towards our practices.
For me, I seem to file this nicely within my prostate cabinet. It fits in and re-affirms things. But for some, this may help structure their organization of our data or serve as a starting point - based on a “current” analysis. It is a good look at broad context albeit buried within a ton of math (that is coming from a Biomedical engineer who likes math and stats).
There is likely opportunity for intensification in some men.
Consider both the MDACC proton data with bRFS of just under 70% when the pelvis wasn’t treated in very high risk men. So in higher risk men, intensification for many might be simply adding pelvic treatment.
Or consider the POP-RT trial that demonstrated DFS of 90% adding in pelvic treatment. I think if we hit 90% consistently, it is a tough sell to further intensify, but if we figure out our max expected is closer to 85% for some of the highest risk disease, I think that opens the door much wider for genetics and treatment intensification for some men beyond radiation and ADT approaches.
And if we can get everyone to 95%, that would be fantastic.
Summary:
Kudos to the authors for all of the work that went into this effort. I really worked to simplify the paper and might have overstated, underemphasized certain points. As always, feel free to reach out or comment
Thanks for following along as we search for better. In honor of the old times, we’ll keep today short and sweet. Plus, more labs have rolled in and I need to get back to the business of updating the old spreadsheet. We’ll chalk this one up as a win for Gunar some 25 yrs later - no real surprises in this one.
(FYI: As this site grows, search is more important. You can go to protons101.com and there is a search icon in the header on the right - you can use that and type in “hazard rate” for example and it will search only this Substack to find related articles I’ve written - Example below.)
