IMRT De-escalation vs: Protons
How should we consider these two approaches?
How I view de-escalation now 6 years in at a proton facility.
This is a follow-up to my prior post on IMRT de-escalation of the elective nodal dose. Where that was largely data, this article describes my thinking on integral dose de-escalation.
Here is the prior article:
The Goal to Give Less:
Over six years ago I moved to work in a proton facility. Obviously the draw to protons is math / physics based. The goal is to use the differences in beam characteristics to decrease integral dose. To give less radiation to the patient while giving at least the same dose to the tumor. That decrease in dose should then lead to less toxicity. My undergraduate degree is in Biomedical Engineering and this simply makes sense.
Getting here, I think teaches you that often medicine is not that simple. What appears like a simple win can be more complex. Pair that with a time when computers are being rapidly integrated into our field and we are moving from back-pointers and 2D doses to complex 3D IMRT generated plans with hundreds of MLC patterns, and you realize the story may not be so simple. When one side is more constrained by debt finances (protons) than the other, and when photon machines outnumber protons by 30 or 40 to 1, the sheer numbers alone can produce advances that are difficult to replicate in smaller scales.
During these years, I’ve gone back to basics. I’ve always done dosimetry work (REF - Old LAD data). So in this past half a decade, I’ve done quite a bit of comparative work looking at protons vs. photons. And today, I’ll walk through how I look at comparing these two approaches.
Why write this article? Because I asked a pretty basic question online, and the vast majority missed the answer. To me, it is a rather basic knowledge question within our field but acknowledge it is certainly within a niche. So I write this to maybe shed some light on how to better consider and read our literature when trying to make your decision on whether protons might be good option for a patient or how to consider data comparing the two modalities.
So today will be part math, part physics, part opinion: describing how I look at this issue.
Which approach decreases radiation more?
De-Escalated IMRT or Full Dose Protons.
So in my prior article, I argue for this standard:
2025 Reasonable De-Escalation Standards (EQD2):
HPV-: ENI of 40 Gy, 70 Gy to gross disease.
HPV+: ENI of 30 Gy, 70 Gy to gross disease.
Intermediate dose of 60 Gy in HPV- and 50 Gy in HPV+ is warranted in a minority of cases for suspicious indeterminant nodes (Defined above).
(supported by at least 800 patients - 500 within randomized controlled trials)
Today, lets consider the more aggressive de-escalation answer - reducing the ENI dose to 30 Gy. Here is the question posed online.
The Math Favors Protons:
The answer is:
Swapping to protons is a much greater reduction in dose to the patient and it isn’t really close.
Most missed it for what small numbers are worth.
A simple rule: protons give about 50% of the integral dose to the patient.
Most cases fall between 40% and 60% reduction - published probably hundreds of times over the past few decades. Let’s use 50% - nice and easy.
In the IMRT dose reduced plan, let’s ballpark that half the field size is elective and half is high dose. (Think vertically along patient axis as an easier to visualize approximation than volume comparisons for integral dose).
We’ve reduced half the field size by 40%. Half the field size times 40% is 20%. A 20% reduction in integral dose (a reasonable approximation to what you get in a planning system environment). That is your ballpark figure.
Once you consider the question via this type of approach, you can quickly see there is almost no way to mathematically reduce integral dose by 50% without moving the top dose level. And a pretty focused group - largely from within our specialty - clicked the grossly wrong answer.
But does this integral dose difference matter?
That is a valid and different question.
We’ll return to this below.
This integral dose reduction question is a basic physics knowledge question. Unfortunately, and honestly a bit surprisingly to me, many do not appreciate the differences in magnitude between the two approaches.
But again, this is a region where I do a lot of work.
A few years ago, when I first saw the MSKCC data showing that HPV+ cases can likely be treated with stepwise planning: Initial total volume to 30 Gy followed by a reduction and boost to areas of gross disease to 70 Gy - I went to a planning system. (I’ve done the same work looking at constrictor optimized plans when the DARS randomized prospective data was released in 2023 - great trial).
My first question was, how does this de-escalated IMRT plan compare to protons? I pulled out a handful of cases where I both VMAT plans and uniform scanning proton plans and added a third tier of de-escalated IMRT plans.
I ran multiple plans, multiple ways to figure out this stuff. Here is a representative image documenting the difference.
The white box is a gross approximation of integral dose - average dose to the entire body contour. The values might not show up, but they are 1181, 844, and 518.
I ran it on multiple cases and pretty much got the same answer each time - relatively speaking. Full dose IMRT is 100%, De-escalated IMRT to 30 Gy is 70%, and protons are ~50%.
And here is my example of the scope of the dose differences.
In general, one could treat with protons to full dose and AT LEAST every 3rd fraction, get the patient up off the table, walk them to a photon machine and deliver a 200cGy 6mv 10x10 field prescribed to Dmax into the HN region and that patient would receive less total radiation into the HN than a reduced dose IMRT plan.
Here is a link to my prior write up from 2023. The dosimetry work is highlighted towards the lower 1/3rd of the article if you want to see more receipts on the magnitude of the differences and to see the 10x10 field example.
(Self assessment: article is 2 yrs old - still holds up quite well - yay!)
Phase III Proton Data for Head and Neck Cancer.
In May of last year, the abstract was presented for the US Phase III randomized trial comparing IMRT to IMPT. (photons vs. protons). (REF). The primary outcome measure for the trial progression free survival with secondary endpoints of OS, treatment-related malnutrition, and feeding tube dependency. NRG forced a non-inferiority structure using progression free survival. (REF - an interesting read)
In that non-inferiority design, here is the summary from the abstract:
In the ITT analysis, the hazard ratio (HR) for disease progression or death at 3 y was 0.87 (95%CI 0.56,1.35); p=0.006 and the corresponding HR for death (OS) was 0.63 (95%CI 0.36-1.10) suggesting a protective affect with IMPT. In PP analysis, the PFS HR was 0.85 (95%CI 0.52,1.38); p=0.009 and HR for death (OS) was 0.60 (95%CI 0.32-1.12). In the AT analysis, PFS HR was 0.88 (95%CI 0.56,1.37); p=0.007 and the corresponding HR for death (OS) was 0.70 (95%CI 0.40-1.22). For each analysis above, the null hypothesis was rejected and IMPT was non-inferior to IMRT.
The progression free survival (HR ~0.87 for all analysis) favors protons. Further, feeding tube rate fell from 42% to 28% and fewer patients lost less weight 24% vs. 14% (<5% wt loss). (Each statistically significant findings). Even OS at least trends in favor of proton therapy.
So OS and progression free at least lean in favor of protons along with less weight loss and less feeding tube placements. Both primary and secondary outcome measures favoring protons - non-inferiority design so appropriately described as non-inferior.
Prior to the release of the data, I wrote in March of 2023 (Non-inferiority Trial Design) regarding my concern that this design was purposeful in not allowing a “clear win” to happen. Here is what I wrote:
And if it does demonstrate benefit, naysayers without access to proton therapy will be lining up to nullify any positive result in part due to conflicts of interest in our healthcare system. It would simply require too much change too quickly to accept a benefit to protons. And the “non-inferiority” aspect adds a clear path to undercut results. It is due to this avenue that I even question the trial development process due to potential conflicts of interest.
Read more in full article:
*** If you made it here, please leave a “like” or comment. It supports the site along with free subscriptions. I’m leaving it all free, but it helps these to gain visibility with the algorithms picking winners and losers. Thank you. ***
Which Leads me to look back and ask one question:
How do you view the UPGRADE-RT trial? (The recent IMRT de-escalation Head and neck trial)
How would you view the UPGRADE-RT trial if it had been a swap to protons instead of a dose reduction trial?
Remember, the primary endpoint failed. There was no difference in normalcy of diet at 1 yr. The argument for this lack of difference kind of sounded like a proton trial excuse: “both arms did well - little room for improvement”.
And while there were some soft secondary metrics like less feeding tubes (Remember: NRG oversight committee deemed this exact metric too easy to manipulate in the proton trial.) and salivary function was “different” with a highly significant p value of 0.006. Below is that curve again.
I firmly believe in any proton trial, this level of difference would be dismissed in the setting of a negative primary endpoint with that much overlap in confidence intervals in secondary metrics.
As in completely ignored.
For comparison, the Phase II Esophageal trial showed a 99.9% likelihood of less toxicity with protons and even that has largely been dismissed. Just look at its inability to alter referral patterns beyond institutional walls in the US.
Again the data is 99.9% confident protons had less toxicity.
But is using integral dose differences a valid metric?
I think experts can honestly disagree here. I land on them being quite similar. Here is how I consider the two sides of the discussion after years of consideration.
The two IMRT trials attempt to measure the clinical effects of a reduction in dose to elective structures. In these approaches there are at least two variables moving - 1) the subclinical dose level moves and 2) the integral dose is thereby reduced. The safety signals are basically ALL within disease recurrence.
In proton trials, the de-escalation is due to the physical properties of the beam. We’ve decided to keep dosing the same and only attempt to measure the effects of integral dose reduction due to the beam properties. At the same time, safety signals are different. We look for safety signals that would arise from the planning system not modeling proton beam characteristics as we believe they do - i.e. what you see isn’t what you get.
In my current thinking (and in this article), I tend to mentally file all these approaches into one group. I don’t believe anyone can accurately weigh lower ENI dose vs. less integral dose effects in IMRT head and neck planning. Both likely contribute on some level. There would be no data-driven consensus on contributions from each.
With regard to safety signals in protons, I believe strongly these have largely resolved in the past decade. Yes, there are some specific historic instances in the literature but in the past 10 years, these are increasingly rare*. Today, I’d argue there are no more safety signals with protons compared to our ultra-hypofractionation literature.
Maybe some see these two approaches as more different than I do and file them separately. From my perspective, I see that as quite biased - acknowledge less radiation primarily in times when patients (typically) stay within our own system.
We “know” that integral dose matters. Radiation has toxicity. And in the last decade I’d argue we now have strong data in CSI, esophagus, head and neck, and lung data (from Europe - not the older uniform Bayesian MDACC trial) demonstrating we can measure clinical differences in outcomes with a decrease in radiation to the patient (i.e. via a shift to protons) in high toxicity settings.
*The best example of a “worse” outcome in the past 10-15 yrs is this MSKCC study - potentially pointing to higher risks of osteoradionecrosis - if you treat with protons, you need to have read and considered this paper
A Note on Relative Benefit vs. Absolute Magnitude
For years, I’ve predicted a strong push in head and neck cancer from institutions as the proton data matures and is available. Even if you demonstrate clear toxicity reduction today, the de-escalation will be argued “demands” further head to head testing - thus maintaining the status-quo. And de-escalation does indeed make the bar harder to clear. Here is the math behind that.
Protons give half the integral dose. De-escalation gives less radiation. If we consider IMRT to deliver 1000 “units” of dose, a reduced dose of 30Gy instead of 50Gy gives about 75% of the dose - so about 750 “units”.
If you are trying to measure the “benefit of protons” - as we give less, it will be harder to measure a clinical difference. Half of 1000 is 500. Half of 750 is 375. For protons to be successfully proven better, we have to measure effects on the difference.
From my perspective, the current de-escalation data actually captures this.
In the IMRT trials, we are trying to measure clinical outcome differences on a difference of 250 units. And maybe we did in places, but the primary objectives of both trials did not demonstrate clear benefit. In the proton trial, we were attempting to measure the effects of 500 - albeit with more experience in attempting to do these types of studies - but it showed clear improvement in outcomes. The larger difference in integral dose was easier to measure - makes sense.
With IMRT de-escalation we all KNOW that lower doses are better for toxicity. So it is easy to label the trial as a “huge win”. Honestly, I agree. This topic brought me out of retirement on Substack to write about this new “standard”. But I also picked up my practice and moved for protons, because of what I see as an amazingly similar argument.
Closing:
The proton literature has become a Rorschach test for our specialty. In all fairness, the field hasn’t produced enough clear data wins - especially looking backwards over the decades. But I do believe we are addressing it today and today, if anything, I think we are more often erring in the opposite direction - not appreciating the differences when there are clear wins. (This is the opinion part - just for additional clarity - haha).
The goal of today was to present some relative differences between IMRT de-escalation and protons as a method to de-escalate. We still have three large pending head and neck publications on the horizon. Hopefully something in the above article is pertinent to your attempt to interpret this upcoming literature. Since I’ve thought about this stuff for more than half a decade, I thought I would put some of it down in writing - thinking it might help people get to where they need to be faster regardless of how you ultimately land.
The goal is better treatment. Best outcomes / More cures / Least toxicity.
Six years ago, I picked up and moved because I was convinced that protons would provide a better option for many head and neck cancer patients. Both the recent IMRT de-escalation data and the proton abstract above confirm what I have seen in my time in Oklahoma. Amazingly exciting for outcomes. I continue to remain optimistic that we’ll continue to figure out ways to use better and better technology to further decrease toxicity. Whatever paths those improvements come along, I’m thrilled to support.
I’m looking forward to the full publication release of the IMPT trial and the trials coming out of Europe in the head and neck space. As always, keep pushing for better.
Thanks for reading along. If I missed something important, please comment or reach out. Happy to re-work or improve something if this author of one is clearly wrong.
Until the costs of Protons can justify the purported improvements for the vast majority of hpv- 4a typical chemo xrt patients... IMRT will remain the standard and we can continue to explore reducing ENI to its verifiable safe minimum.. Or one can just assume immunotherapy will be the eventual savior. Pro? Ton$!