Hippocampal avoidance not (clearly) effective

[scarbrtj]

ASTRO 2018, ~500-person trial (w/ Namenda), "practice changing," cognitive failure reduced by ~8% at 6 mos (~68%->60%), p=0.03
ESTRO 2019, ~160-person trial (? Namenda), at 4 months no difference (p=1.0) and at 8 months ~8% difference (p=0.46)
(Trials w/ p-values in 0.02-0.05 range have about a 25-50% replication failure rate.)
Perhaps p<0.005 should be the new standard for "practice-changing" new standards of care...

---

Members don't see this ad.

 

[seper]

anything clearly different in techniques and/or patients?

 

[medgator]

I wonder what a qaly analysis would like

 

[Neuronix]

ASTRO 2018, ~500-person trial (w/ Namenda), "practice changing," cognitive failure reduced by ~8% at 6 mos (~68%->60%), p=0.03
ESTRO 2019, ~160-person trial (? Namenda), at 4 months no difference (p=1.0) and at 8 months ~8% difference (p=0.46)
(Trials w/ p-values in 0.02-0.05 range have about a 25-50% replication failure rate.)
Perhaps p<0.005 should be the new standard for "practice-changing" new standards of care...

Sounds like the European trial was underpowered.

 

[scarbrtj]

I wonder what a qaly analysis would like

Do not know. Although the ~8% absolute risk improvement seems reliable? In other words, must do HA-WBRT on about 13 people in order to save cognition in just one person. (And it costs about $4000 extra per person to do the IMRT instead of non-IMRT.)

 

[scarbrtj]

anything clearly different in techniques and/or patients?

can't say so shouldn't say (EDIT: ESTRO is HA-PCI)

 

[scarbrtj]

Sounds like the European trial was underpowered.

Thank goodness the Americans had their trial out first or we would have never known 😉

 

[radiation]

ASTRO 2018, ~500-person trial (w/ Namenda), "practice changing," cognitive failure reduced by ~8% at 6 mos (~68%->60%), p=0.03
ESTRO 2019, ~160-person trial (? Namenda), at 4 months no difference (p=1.0) and at 8 months ~8% difference (p=0.46)
(Trials w/ p-values in 0.02-0.05 range have about a 25-50% replication failure rate.)
Perhaps p<0.005 should be the new standard for "practice-changing" new standards of care...

These are very different trials with different inclusion criteria (brain met vs pci) and doses (30 vs 25). Both in abstract form only
Hippocampus Avoidance PCI vs PCI - Full Text View - ClinicalTrials.gov

 

[evilbooyaa]

The first trial at ASTRO 2018 specifically excluded SCLC histology from inclusion. However, it was not only NSCLC - 40% of patients had something besides NSCLC. Would offer HA-WBRT to patients with something besides NSCLC if expected survival was > 6 months.

Different trials evaluating different groups of patients. IMO all this suggests is that a one-fit all approach to WBRT is not justified, and justifies concerns people had about doing HA-WBRT for SCLC in the initial studies (where SCLC was an exclusion criteria). I would not do HA-WBRT for SCLC regardless of prognosis. I would strongly consider offering HA-WBRT (if I was forced to WBRT) for NSCLC and breast (at least) if prognosis was more than 6 months, assuming that there were no metastases within 1cm of hippocampus and no leptomeningeal disease.

 

[medgator]

(if I was forced to WBRT)

Yup...

 

[scarbrtj]

My theory is if you’re going to save cognition in the WBRT setting the best chance is with PCI vs treating mets (nothing alters cognition like mets). So different trials but same underlying principle: spare the hippocampus so as not to harm cognition. Thus far the data is not really a home run. Of course sometimes you still win the game with singles and doubles.

 

[evilbooyaa]

Yes but we have randomized trial evidence saying it's not different in SCLC. Underpowered? Maybe. The data, at the current time, does not justify use of HS in PCI for SCLC. You can poo-poo negative results, but they are negative results.

I personally imagine this is primarily due to the dismal prognosis of most SCLC patients. What percentage (or number of patients) of the HS-WBRT population was alive at 6 months? What percentage (or number of patients) of the HS-PCI population was alive at 6 months?

If those numbers (sample size at risk at 6 months) are significantly different, then you're going to get similar clinical results (8%) with extreme differences in sample size (and thus confidence intervals, and thus statistical significance).

 

[Krukenberg]

There’s an open NRG trial testing this same thing that will probably be better powered. Very diff pop than what was presented at ASTRO

 

[scarbrtj]

Yes but we have randomized trial evidence saying it's not different in SCLC. Underpowered? Maybe. The data, at the current time, does not justify use of HS in PCI for SCLC. You can poo-poo negative results, but they are negative results.

I personally imagine this is primarily due to the dismal prognosis of most SCLC patients.

Some (how many? EDIT: 79% of the HA-PCI patients were LS-SCLC, 96% ECOG PS 0-1) of these HA-PCI were LS-SCLC. Their prognosis would be (a lot) better than a HA-WBRT NSCLC brain met patient.

If one wanted, I think you could massage whatever you wanna massage from the data right now. The data did not say cognitive decline is different for HA-PCI in SCLC than it is in HA-WBRT for brain mets; in fact, they were the same (with different p-values). I would quote Amrhein if I wanted to do HA-PCI: "we should never conclude there is ‘no difference’... just because a P value is larger than... 0.05... It is ludicrous to conclude that the statistically non-significant results showed 'no [difference]'"; there was an absolute 8% improvement in cognition in both trials although one due to power (you are correct) was non-significant. However I want to spare cognition and I "know" from CC001 hippocampal avoidance does this. I mean, for a HA-PCI SCLC patient with no evidence of disease on any scans anywhere wouldn't you wanna spare the hippocampus?

But, as a skeptic, I would say in one trial (CC001) they used a Gray's test and in about 500 patients an 8% difference was different at p=0.03. But in another test they just compared absolute differences using chi square and for ~100 patients this 8% difference was only different at p=0.4. So I would never use HA-PCI and I doubt HA-WBRT for brain mets is good either. It's a waste of time, money and effort.

Plant your flag and die upon whichever hill suits your fancy for now.

 

[evilbooyaa]

You're more than welcome to do whatever you want in clinical practice. If you want to HA-PCI every SCLC that walks through the door makes no difference to me.

My point about 'no data' is primarily in regards of getting HA-PCI covered by insurance. I understand statistics - I know that the absence of evidence is not the evidence of absence. For now, there is no prospective data that supports an improvement in neuro-cognitive failure in SCLC. There is prospective data that a difference was not detected.

I'm not the one 'massaging' the data here. There are results, with p-values that determine their statistical significance. We can argue until the cows come home of the validity of a p-value, but that's the accepted standard (for now). HA-WBRT improves cognitive free survival in non-SCLC patients (mostly NSCLC and breast) in a prospective randomized trial, with separation at about 6 months (hence why the rec is to do HA-WBRT if the life expectancy is > 6 months, which brings up the whole issue of 'oncologic life expectancy' and those uncertainties). HA-PCI was not shown in a prospective randomized trial to provide that same (statistically significant) benefit in SCLC patients.

Any explanation to invalidate those two points or reach a conclusion separate from the above paragraph - THAT is massaging the data. For now, I do not have data to support offering HA-PCI to patients. I do have data to offer HA-WBRT to patients with > 6 months life expectancy who otherwise met inclusion criteria of that trial.

Personally, I think PCI is an antiquated way of thinking about oncology. As how PCI is (IMO and per a presentation at Multidisciplinary Thoracic Symposium) now dying in ES-SCLC, replaced by q3 month MRIs, I imagine a very similar story will be told in LS-SCLC. No data to support it, yet, although an upcoming SWOG study will be evaluating the role of PCI in both LS and ES-SCLC.

 

[scarbrtj]

I'm not the one 'massaging' the data here. There are results, with p-values that determine their statistical significance. We can argue until the cows come home of the validity of a p-value, but that's the accepted standard (for now).

SABR-COMET hailed as practice-changing with a p-value of 0.09. Massage-y.
NRG CC001 hippocampal avoidance, 8% cognition improvement: good.
Hippocampal avoidance for PCI: 8% improvement. Bad? Doesn't meet "standard?" We must follow the standard I suppose. Befehl ist befehl.

HA-WBRT improves cognitive free survival in non-SCLC patients (mostly NSCLC and breast) in a prospective randomized trial, with separation at about 6 months (hence why the rec is to do HA-WBRT if the life expectancy is > 6 months, which brings up the whole issue of 'oncologic life expectancy' and those uncertainties).

Going to be some misunderestimations of the life expectancies probably. It is annoying when someone you predict was going to live 4 months winds up living 8 months or vice versa. (I will say that I feel comfortable crystal balling the 6-month survival for LS-SCLC PCI patients at better than 90%.)

My point about 'no data' is primarily in regards of getting HA-PCI covered by insurance... Personally, I think PCI is an antiquated way of thinking about oncology. As how PCI is (IMO and per a presentation at Multidisciplinary Thoracic Symposium) now dying in ES-SCLC, replaced by q3 month MRIs, I imagine a very similar story will be told in LS-SCLC.

Brace for the heresy:
Prophylactic cranial irradiation is a misnomer. It's like calling RU486 a prophylactic. Radiation never prophylaxes against cancer seeding a virgin site. "PCI" can only work if there are cancer cells present within the brain. So, everyone treated with PCI is prima facie being diagnosed by his/her doctor with brain metastases. (Theoretically, HA-WBRT is always insurance-covered for brain mets.) From this intellectual perspective, PCI is no more antiquated than "prophylactic breast irradiation" (aka "PBI") after lumpectomy or "prophylactic nodal irradiation" (aka "PNI" or ENI) for high-risk breast cancers e.g. If PCI is antiquated, most all of non-PCI modern radiation oncology is at least partially likewise convicted.
EDIT: Sorry I keep adding. But your final "I imagine a very similar story will be told in LS-SCLC" keep in mind PCI for LS-SCLC is curative. I know of no other curative type situation where waiting longer for the cancer to get bigger doesn't hurt outcomes. First chance to cure is best chance to cure: it's a truism.

 

[RadOncDoc21]

I’m starting to do more surveillance scans with MRI’s and treat either with whole brain or SRS. I moved away from PCI about a year ago but still discuss it as an option. I’m not sure which is better to use IMRT to spare hippocampus but SRS to me seems like the better option but this is my bias opinion (no data).

 

[scarbrtj]

I’m starting to do more surveillance scans with MRI’s and treat either with whole brain or SRS. I moved away from PCI about a year ago but still discuss it as an option. I’m not sure which is better to use IMRT to spare hippocampus but SRS to me seems like the better option but this is my bias opinion (no data).

Waaay better. But in the pantheon of radiation oncology, whole brain radiotherapy still has a defined (and IMHO "non-optional") role and that's for a LS-SCLC patient who completes thoracic chemoradiation with a good response. This is where SRS hippocampal sparing is moot.

 

[evilbooyaa]

SABR-COMET hailed as practice-changing with a p-value of 0.09. Massage-y.
NRG CC001 hippocampal avoidance, 8% cognition improvement: good.
Hippocampal avoidance for PCI: 8% improvement. Bad? Doesn't meet "standard?" We must follow the standard I suppose. Befehl ist befehl.

Yes, equivalent clinical benefits (on the means) when you have super wide confidence intervals is literally the whole premise of statistical significance. Are you seriously arguing that 8% with a non-significant (p=0.4) p-value should be taken as equivalent to 8% difference with a significant p-value? Why do statistics at all then? Just report gross differences, no confidence intervals, no standard deviations. Who cares about literally all of the practice of statistics in regards to evaluation of a clinical trial??

We're discussing HA here. Not sure what SABR-COMET has to do with it. SABR-COMET doesn't have to be practice changing... that's why they're running the phase III. This to me demonstrates a lack of understanding of the point of an exploratory phase II trial, where the threshold is set at something higher than what is standard.

EDIT: Sorry I keep adding. But your final "I imagine a very similar story will be told in LS-SCLC" keep in mind PCI for LS-SCLC is curative. I know of no other curative type situation where waiting longer for the cancer to get bigger doesn't hurt outcomes. First chance to cure is best chance to cure: it's a truism.

Adjuvant vs early Salvage RT for prostate cancer (arguably TBD based on ongoing trials, but based on clinical practice?)? Adjuvant vs salvage for low grade glioma?

Again, we'll see what the trials say - I think if people are going to develop radiographically visible brain mets that get caught early there will be no difference in survival (just like in the Takahashi trial). But again, let's await the data.

 

[BobbyHeenan]

I’m starting to do more surveillance scans with MRI’s and treat either with whole brain or SRS. I moved away from PCI about a year ago but still discuss it as an option. I’m not sure which is better to use IMRT to spare hippocampus but SRS to me seems like the better option but this is my bias opinion (no data).

I'm doing more surveillance for extensive stage disease (rather than PCI) but haven't pulled the trigger on up front SRS in small cell yet.

As scarbrtj suggests, I too have been reluctant to quit offering PCI for LS-SCLC; I nearly across the board treat every LS-SCLC with a good response with 25/10 PCI.

For CNS failure without prior WBRT in a small cell case unless a compelling reason (ie prior stroke with deficit) I'm still giving whole brain. I don't fault anyone for jumping to SRS in small cell but I haven't been doing that yet.

 

[RadOncDoc21]

I think it’s mainly because of the time period that sways me away from jumping onboard with PCI. I offer it and recommend it to all my patients but also give them my opinion and they have to agree to follow up with MRI’s closely. Definitely wouldn’t recommend it for the folks going to Louisville.

 

[scarbrtj]

Yes, equivalent clinical benefits (on the means) when you have super wide confidence intervals is literally the whole premise of statistical significance. Are you seriously arguing that 8% with a non-significant (p=0.4) p-value should be taken as equivalent to 8% difference with a significant p-value?

There is nothing more equivalent to 8% than 8% that I know of; they're this much different. I attend the Church of Statistical Significance as well. But I occasionally miss a service. What we are talking is power, as you know.

We're discussing HA here. Not sure what SABR-COMET has to do with it. SABR-COMET doesn't have to be practice changing... that's why they're running the phase III. This to me demonstrates a lack of understanding of the point of an exploratory phase II trial, where the threshold is set at something higher than what is standard.

On the other hand, the people (our people... rad onc people) that ran the trial say treating as per SABR-COMET is to what they have changed their practice; the exploratory trial, for them, was literally non-exploratory. Plus it was hyped in the media, etc etc.

Adjuvant vs early Salvage RT for prostate cancer (arguably TBD based on ongoing trials, but based on clinical practice?)? Adjuvant vs salvage for low grade glioma?

It is rather easy for me to argue/cite worse oncologic outcomes for salvage (ie treat gross disease) vs adjuvant (ie treat subclinical/microscopic disease) in both these situations. I shall refrain.

 

[medgator]

Adjuvant vs early Salvage RT for prostate cancer (arguably TBD based on ongoing trials, but based on clinical practice?)?

The data points to a probable OS benefit, if not at least a benefit for metastasis free survival and lower need for long term ADT, both of which are laudable end points outside of OS

 

[medgator]

. Definitely wouldn’t recommend it for the folks going to Louisville.

Or anyone outside of academics/sovereign immunity bubble. Pci is cat 1 in LS SCLC. I've treated a few pts over the years who refused pci and then got admitted with neuro sx for which I eventually got involved

 

[scarbrtj]

What I have been obtuse in saying is that if one adopts a staunch noli me tangere approach for the hippocampus in the setting of, e.g., a NSCLC patient with gross brain metastasis upon whom whole brain radiation is planned, it would be cognitive dissonance (ouch) not to adopt the same such approach for a SCLC patient with no detectable evidence of disease. Hippocampal sparing for the patient with ~0-1% 5y survival probability and not for the patient with a 25% chance of 5y survival? One believes that sparing the hippocampus preserves some elements of cognition when the whole brain is irradiated, or one does not. There is no gray (ouch ouch) zone. Probably I'm the only one who sees it this way!

 

[Palex80]

I was at ESTRO and watched the presentation. The study results were not challenged and it seems at it is: hippocampal avoidance may actually not be working. On a side note, there was another presentation from the US at the ESTRO pointing out that perhaps the amygdala may be more important than the hippocampus...

From what I understand, the RTOG study has come under quite a bit of criticism due to the drop-out rates in the cognitive tests performed...

To me, this presentation was one of the two most interesting ones on the clinical track. The other being the Italian trial looking into radical RT for inoperable / partially excised mesothelioma...
ESTRO 38: RT doubles survival for patients with mesothelioma



On a side note: Did anyone of you attend ESTRO? I forgot to write that I was going and we could have perhaps met up for a drink.

 

[scarbrtj]

I was at ESTRO and watched the presentation. The study results were not challenged and it seems at it is: hippocampal avoidance may actually not be working. On a side note, there was another presentation from the US at the ESTRO pointing out that perhaps the amygdala may be more important than the hippocampus...

From what I understand, the RTOG study has come under quite a bit of criticism due to the drop-out rates in the cognitive tests performed...

To me, this presentation was one of the two most interesting ones on the clinical track. The other being the Italian trial looking into radical RT for inoperable / partially excised mesothelioma...
ESTRO 38: RT doubles survival for patients with mesothelioma



On a side note: Did anyone of you attend ESTRO? I forgot to write that I was going and we could have perhaps met up for a drink.

Optune is on its way for an indication in mesothelioma FWIW

 

[medgator]

Optune is on its way for an indication in mesothelioma FWIW

Heard the same. Interesting product. They probably could try it in sarcoma as well

 

[evilbooyaa]

Optune is on its way for an indication in mesothelioma FWIW

Saw some very interesting optune data for mesothelioma at ASTRO 2018. Wonder if that will be 'the next rage' - Optune for all the things everywhere. Wonder if it'd work on other poor prognosis cancers....

 

[Palex80]

Saw some very interesting optune data for mesothelioma at ASTRO 2018. Wonder if that will be 'the next rage' - Optune for all the things everywhere. Wonder if it'd work on other poor prognosis cancers....

Pancreas is also being tested, last time I heard.