Progressive brain mets after previous SRS

[Mandelin Rain]

Have a patient 6 years out from dx of brain mets from NSCLC. She has had a total of 5 brain mets diagnosed over the years, all of which have been treated with single fx SRS to 24 Gy initially. Two have been re-treated with repeat single fraction again. Those two are currently controlled. The other 3 are currently progressive on MRI, simultaneously. She is asymptomatic.

1. Anyone ever see this? Feel like it's something in her tumor's biology that just doesn't respond well to single fraction, as the odds of controlling at least one of five lesions with 24 Gy long term seems high.

2. What to do now with the 3 progressive lesions. They are all fairly small (i.e. about 1 cm). Could do three repeat single fraction SRS. Would probably prescribe 20 Gy to each. I was also kicking the idea of single isocenter fractionated stereotactic treatment to 30 Gy in 5 fx to all simultaneously. Not excited about whole brain, nor is the patient.

3. How do you evaluate repeat SRS plans? I've just evaluated each plan on it's own merit and not worried too much about the composite, lacking some near by OAR like stem or chiasm. But this method seems like it could be enhanced.

Any feedback would be considered helpful, as I'm at a bit of a loss.

---

Members don't see this ad.

 

[Neuronix]

I just wanted to chime in that everything you just wrote is totally reasonable and how I approach it is as well. There is retrospective data for re-irradiation, and in that data plus my own experience risk of RT necrosis is small for re-irradiation of small mets with a good time interval between.

I read one paper recently that indicated later whole brain after the second course of SRS does increase the risk of RT necrosis. I suspect that most of the time the brain can really take two courses of SRS to a limited volume at the usual doses, but a third time is just too much. I like to think of a course of WBRT like a weak SRS course (BED 12 Gy/1 = 30 Gy/10 for a/b=3).

The only other thing you have to consider is RT necrosis of course. I wouldn't ask to biopsy three lesions. None of the imaging tests are really great at being certain what is real recurrence vs. RT necrosis though you can try to differentiate the two if the situation is unclear or it helps you feel better.

 

[Radiator20]

Is she progressing outside the CNS? If she needs to change systemic tx anyway, and the new regimen has some likelihood of CNS efficacy, could try that first before further re-RT. But if she is currently controlled outside CNS, I would not change systemic tx simply to avoid re-RT.

Totally agree that 0/5 controlled with 24 Gy seems strange. Thus, I agree with Neuronix that excluding radionecrosis seems like the other important thing. That said, 5/5 with radionecrosis would also be strange, and it doesn't sound like the first 2 progressive lesions were radionecrosis since they stabilized with re-SRS. Hmm.

Also, is there a reason you would do 30/5 with a single iso plan but wouldn't do 20/1 with a single iso plan?

 

[Mandelin Rain]

My thought in fractionating is twofold. A. 1 fraction treatments haven't proven incredibly effective and perhaps fractionating while maintaining a high BED would yield better results? (i.e. pure speculation) B. Less radionecrosis in fractionated vs single fraction in de novo setting. Likely applicable to re-treat, but who knows?

If I was doing single fraction, I would just treat them individually because I get a more conformal plan (much more with cones) and I feel better about the set up as it pertains to any individual lesion.

 

[emt409]

Ok, there is a lot of important missing information to make an informed decision here. Here's what we need.

1) Timing of SRS
2) Is there immunotherapy?
3) How do you know it is progression vs pseudoprogression?
4) Have you requested perfusion/spectroscopy sequencing on the MRI?
5) If you are at an academic center, you may be able to request an FTB (fractional tumor burden) map, which is even better than perfusion/spectroscopy MRI

The control rate of 24Gy is really high. You are much more likely to see pseudoprogression or local asymptomatic radionecrosis than progression at that dose level, especially in the context of concurrently or closely sequential immunotherapy.

If we assume the control rate of 24Gy is 95%, the probability of all 5 mets progressing is (5/100)^5 or 3125/10^10 or 3.125 x10^-6. There is a much higher probability your patient wasn't aligned at true iso and you just missed every target due to rotational error (assuming this was single iso or CK tx, and not GK)

For repeat multi-met SRS, I fuse the each planning CT's, and create a volume out of the prescription isodose for each plan as a structure. Then I can see exactly the locations treated in comparison to the new tx planning CT. This is the poor man's version of using Velocity, which we have but I haven't had time to learn yet (actually that's a lie, if I put the time I've been playing red dead redemption 2 into learning velocity, I'd be an international expert by now).

 

[Mandelin Rain]

The point about radionecrosis is well taken though. Very difficult to tell. Temporally, they seem more like progression. I can't get my radiologists to even modestly commit to necrosis verbally, let alone on paper. I've ordered SPECT, Perfusion, PET, etc... Never get anything but, "consistent with progression with necrosis another possibility." I ordered a brain PET and the reading stated, "abnormal brain activity obscured by normal brain activity." Eh... what?

Definitely one thing I miss about academic medicine.

 

[Mandelin Rain]

Ok, there is a lot of important missing information to make an informed decision here. Here's what we need.

1) Timing of SRS
2) Is there immunotherapy?
3) How do you know it is progression vs pseudoprogression?
4) Have you requested perfusion/spectroscopy sequencing on the MRI?
5) If you are at an academic center, you may be able to request an FTB (fractional tumor burden) map, which is even better than perfusion/spectroscopy MRI

The control rate of 24Gy is really high. You are much more likely to see pseudoprogression or local asymptomatic radionecrosis than progression at that dose level, especially in the context of concurrently or closely sequential immunotherapy.

If we assume the control rate of 24Gy is 95%, the probability of all 5 mets progressing is (5/100)^5 or 3125/10^10 or 3.125 x10^-6. There is a much higher probability your patient wasn't aligned at true iso and you just missed every target due to rotational error (assuming this was single iso or CK tx, and not GK)

For repeat multi-met SRS, I fuse the each planning CT's, and create a volume out of the prescription isodose for each plan as a structure. Then I can see exactly the locations treated in comparison to the new tx planning CT. This is the poor man's version of using Velocity, which we have but I haven't had time to learn yet (actually that's a lie, if I put the time I've been playing red dead redemption 2 into learning velocity, I'd be an international expert by now).

The 5 were not treated at the same time. All treated between 6 months and 6 years ago. Each was treated on it's own with frameless linac based SRS with OSMS optical guidance and 2mm margin to PTV. Prescription IDL typically 80%. I feel very good about our setup within 2mm when treating a single lesion, but understand your point about rotational error when treating multiple lesions with a single iso.

I also make a composite of all previous treatment. The question become what you do with the information. What's the dose/volume limit for brain? How hot is too hot for MPD?

EDIT: She has been on immunotherapy. Multiple rounds since the first SRS. Systemic disease relatively controlled.

 

[Mandelin Rain]

Part of the issue too is there just isn't much data on lesional control rates at >2 years.

Seeing it more and more these days (which is a very good thing).

 

[Radiator20]

If we assume the control rate of 24Gy is 95%, the probability of all 5 mets progressing is (5/100)^5 or 3125/10^10 or 3.125 x10^-6.

I appreciate the point of back-of-the-envelope calculations, but I don’t think this one is valid. The probabilities are not independent. If one lesion has failed SRS, the probability that others do too surely increases a lot.

 

[radiation]

I think we tend overestimate long term local control for SRS as most of these pts have limited expectancy. Long-term follow up on prospective trials almost always shows less than the 90+% local control commonly quoted. For example, one of the more modern SRS trials shows significantly fairly low local control with SRS alone at 1 year for lesions <3cm (buried in the supplement)

SRS With or Without WBRT and Cognitive Function in Patients With Brain Metastases

Cleveland Clinic has a series of patients with larger tumors being treated 15 Gy x 2 spaced one month apart with reasonable outcomes and relatively low toxicity
Impact of 2-staged stereotactic radiosurgery for treatment of brain metastases ≥ 2 cm. - PubMed - NCBI

Also, LITT is an option that is theoretically supposed to treat both tumor and necrosis. It does require some expertise by neurosurgery and access to an intraoperative MRI though
Magnetic Resonance-Guided Laser-Induced Thermal Therapy for the Treatment of Progressive Enhancing Inflammatory Reactions Following Stereotactic Ra... - PubMed - NCBI

 

[nkmiami]

t

I think we tend overestimate long term local control for SRS as most of these pts have limited expectancy. Long-term follow up on prospective trials almost always shows less than the 90+% local control commonly quoted. For example, one of the more modern SRS trials shows significantly fairly low local control with SRS alone at 1 year for lesions <3cm (buried in the supplement)

SRS With or Without WBRT and Cognitive Function in Patients With Brain Metastases

View attachment 244662


Cleveland Clinic has a series of patients with larger tumors being treated 15 Gy x 2 spaced one month apart with reasonable outcomes and relatively low toxicity
Impact of 2-staged stereotactic radiosurgery for treatment of brain metastases ≥ 2 cm. - PubMed - NCBI

Also, LITT is an option that is theoretically supposed to treat both tumor and necrosis. It does require some expertise by neurosurgery and access to an intraoperative MRI though
Magnetic Resonance-Guided Laser-Induced Thermal Therapy for the Treatment of Progressive Enhancing Inflammatory Reactions Following Stereotactic Ra... - PubMed - NCBI

The 2 staged radio surgery is an interesting approach to shrink the tumor between the first and second fractions. Scarbtj had also posted some examples last year where he gave several conventional fractions to shrink the tumor and edema followed by by radio surgery a week or 2 later?

 

[emt409]

I think we tend overestimate long term local control for SRS as most of these pts have limited expectancy. Long-term follow up on prospective trials almost always shows less than the 90+% local control commonly quoted. For example, one of the more modern SRS trials shows significantly fairly low local control with SRS alone at 1 year for lesions <3cm (buried in the supplement)

SRS With or Without WBRT and Cognitive Function in Patients With Brain Metastases

View attachment 244662


Cleveland Clinic has a series of patients with larger tumors being treated 15 Gy x 2 spaced one month apart with reasonable outcomes and relatively low toxicity
Impact of 2-staged stereotactic radiosurgery for treatment of brain metastases ≥ 2 cm. - PubMed - NCBI

Also, LITT is an option that is theoretically supposed to treat both tumor and necrosis. It does require some expertise by neurosurgery and access to an intraoperative MRI though
Magnetic Resonance-Guided Laser-Induced Thermal Therapy for the Treatment of Progressive Enhancing Inflammatory Reactions Following Stereotactic Ra... - PubMed - NCBI

More CCF data, I think more relevant to our thread. Largest series I know of tumors treated with 24Gy. Local control 96%. Also importantly for this thread, that dose is associated with higher than normal rate of radionecrosis. Granted, these are small tumors, but you shouldn't be treating anything bigger with 24Gy.

Impact of the radiosurgery prescription dose on the local control of small (2 cm or smaller) brain metastases in: Journal of Neurosurgery Volume 126 Issue 3 Year 2017

 

[radiation]

More CCF data, I think more relevant to our thread. Largest series I know of tumors treated with 24Gy. Local control 96%. Also importantly for this thread, that dose is associated with higher than normal rate of radionecrosis. Granted, these are small tumors, but you shouldn't be treating anything bigger with 24Gy.

Impact of the radiosurgery prescription dose on the local control of small (2 cm or smaller) brain metastases in: Journal of Neurosurgery Volume 126 Issue 3 Year 2017

"The median radiographic follow-up per lesion was 6.2 months". I can't think of any prospective data with at least a year of follow up showing 90+% control but would be happy to be informed.

 

[Palex80]

We are going to see more and more of this happening...

We have been bragging about the high control rates of single fraction SRS but data on that are based on patients who very often died during the first year from extracranial disease.
With immunotherapy and targeted therapy becoming more and more effective we will see more local progressions happening.

Remarkably the bulk of the data on SRS are from patients with NSCLC and melanoma. Two of the diseases, which showed a tremendous increase in OS in the past 5 years due to advances in systemic treatment. We built the story on the effectiveness on SRS on those histologies and now that house is coming down, dramatically said.

It has always surprised me, why single session SRS with 20–25 Gy is supposed to work better in the brain than anywhere else in the body. We wouldn't SRS lesions with 1 x 20 Gy extracranial and call that a treatment with durable local control. In the case of early stage NSCLC we opted for a single dose of 34 Gy for instance, not 20-25 Gy. Obviously going beyond 20-25 Gy in the brain is going to cause more radionecrosis, which is why we haven't been doing this.

Furthermore imaging is getting better and I bet that many of the so called "pseudoprogressions" that were reported in patients series with SRS were actually true progressions. Nowadays my neuradiologists perform fancy diffusion weighted imaging and spectroscopy trying to distinguish pseudoprogression from progression. They didn't do that in the past. Plus we perform more frequent imaging. In the past, patients simply died due to extracranial disease, while having a "pseudoprogression" in the brain, so that noone could prove them as true progressions.

I for myself cannot say that our local control rate with SRS is 90%, it is lower than that. And I am not really sure, how we are supposed to tackle this. I have been doing more 3 x 9 Gy @ 80% isodose in the past months, hoping that it will be more potent and less toxic, but it's mostly a gut feeling and not well evidence based (although the Minitti-data look well).

 

[nkmiami]

At this point in my career, I am ready to call BS on complex imaging sequences or patterns to determine radiation necrosis. They almost never prove helpful. Only time/sequential studies seems to help, and often when neurosurgery goes in, there is both necrosis and viable tumor. Equivicating based on study with an unintelligible name like fourier assisted gradient reconstruction gets you nowhere.

 

[Neuronix]

Obviously going beyond 20-25 Gy in the brain is going to cause more radionecrosis, which is why we haven't been doing this.

Interesting thought. RTOG 9005 settled at 24 Gy after WBRT or glioma RT for lesions less than 2 cm. They were supposed to test toxicity at 27 Gy but the trial stopped at 24 Gy.

Some trials have been adding 2 Gy to their doses if patient did not have prior WBRT.

So... 29 Gy to small brain mets anyone ?



Regarding LITT, we do a lot of LITT here. I am unimpressed with the results in the active tumor setting. I've had a number of rapid failures after LITT.

unintelligible name like fourier assisted gradient reconstruction

Isn't that just a fancy way of describing how MRI works in the first place?

 

[Mandelin Rain]

I appreciate the really good discussion, but also looking for some input/advice on this particular case. 3/5 progressive lesions after initial SRS, all 8-10 mm. Slow growth pattern overall, kind of 1-2 mm per lesion per scan. Mild edema. Asymptomatic.

Don't really love the idea of turning an asymptomatic patient into a symptomatic patient with necrosis.

 

[evilbooyaa]

It's a complex case, OP. Thanks for posting, and good discussion thus far, IMO.

At 6 years out from initial SRS, the LC for 24Gy is certainly less than 95% per lesion. Period. SRS outcomes look good because patients historically died before they could recur in the brain. I also imagine RN rates at 6 years with 24Gy marginal dose (with a 2mm PTV margin as well) are much higher than published literature.

The fact she's on immunotherapy suggests that she's not EGFR/ALK mutated (confirm that for me, OP) so the idea of TKIs for intracranial control likely isn't possible.

If she's been on unchanged immunotherapy for a significant time prior to progression, then I think pseudoprogression is a bit less likely. If it's within like 6 months to a year of shift in immunotherapy then maybe. If she was off immunotherapy for sometime and re-initiated within the past 6 months to a year then I would be more wary of pseudoprogression.

The fact that she has responded to re-SRS to 2 of the lesions (she's only had 5 total brain lesions over the course of her entire life? How long ago did you re-SRS the 2 lesions) without exploding in necrosis (as you surely broke doses of 40Gy with 24Gy x 2, associated with increased rate of RN in re-SRS per Wake Forest: Repeat stereotactic radiosurgery as salvage therapy for locally recurrent brain metastases previously treated with radiosurgery. - PubMed - NCBI) suggests that they were active tumor at the time. I think you have sufficient in vivo data, on this specific patient, to offer re-SRS again to all 3 lesions.

What I would do: I would treat now. Waiting and watching for additional growth is only going to make your life more difficult and increase risk of toxicity. While I personally would fractionate to something like probably 8Gy x 3 in the reirradiation setting given lower rates of radionecrosis, feel free to use whatever fractionation you are comfortable with. 30Gy in 5 isn't unreasonable. Obvious discussion about increased risk of necrosis potentially requiring steroids and surgery with re-treatment, but caution that delaying treatment would only worsen that risk if needed to treat in the future.

I wouldn't whole brain this patient at this point.

 

[scarbrtj]

...unintelligible name like fourier assisted gradient reconstruction gets you nowhere.

Technobabble!

If we assume the control rate of 24Gy is 95%, the probability of all 5 mets progressing is (5/100)^5 or 3125/10^10 or 3.125 x10^-6.

The probability of local control for all 5 mets is 0.95^5 = 0.774; thus the probability of any progression for the group is ~25%. The pre-test likelihood, so to speak, of progression in any individual met at this point is thus about 25%. Which is not high, but it's not low. At the very least, I do not think it passes the smell test that you can quantize multiple-locale progression probability this way and say people with 5 brain mets have a ~1 in a million chance of having failure in all treated mets. For example if a single ~1 cm tumor is actually a million subclinical conglomerated tumors, each one with a 95% local control, by your theory the probability of progression in a 1cm tumor would be <10E-100. Approximately. I'm taking an extreme example. I like your thinking but I don't buy it.

 

[radiaterMike]

0.95^5 assumes that local control of one lesion is independent of local control of another. But if there are patient/tumor specific biologic factors that impact local control, then the rate of any local recurrence would be <22% and having multiple lesions recur at once would not be peculiar.

For the OP- it looks like you have mentioned everything you could possibly do to work this up (aside from spectroscopy or biopsy - and I would not suggest a surgical procedure for asymptomatic possible progression). I actually would be OK deferring retreatment given the uncertainty but wouldnt wait too long for serial MRI. If there is continued growth you will eventually have to offer treatment, but those specialized imaging tests would likely given you a clearer picture with a larger (>1 cm) lesion

 

[nkmiami]

24 Gy (and gtv received higher) is a lot of radiation. There is a real biological problem here if multiple lesions have failed. If this ends up being a failure, I would think about getting some kind of theoretical immunological effect with 8-9 Gy x3 combined with immunotherapy, if you are forced to retreat (theoretically 8-9 Gy is optimal in some models for inducing sting/cgas/ etc)

 

[emt409]

Technobabble!

The probability of local control for all 5 mets is 0.95^5 = 0.774; thus the probability of any progression for the group is ~25%. The pre-test likelihood, so to speak, of progression in any individual met at this point is thus about 25%. Which is not high, but it's not low. At the very least, I do not think it passes the smell test that you can quantize multiple-locale progression probability this way and say people with 5 brain mets have a ~1 in a million chance of having failure in all treated mets. For example if a single ~1 cm tumor is actually a million subclinical conglomerated tumors, each one with a 95% local control, by your theory the probability of progression in a 1cm tumor would be <10E-100. Approximately. I'm taking an extreme example. I like your thinking but I don't buy it.

Yes, the probability of one met progressing is 1 - 0.774. However, the mathematics are correct for my back of the envelope calculation of the probability of 5 separate lesions independently treated, independently failing. Your example of parsing a single coherent ~1cm tumor into a million conglomerated tumors makes no sense - I'm not going to bother responding to it.

The point is, it is FAR FAR more likely that there is either necrosis, pseudoprogression if all lesions were completely treated to a marginal dose of 24Gy, or that there was a geometric miss causing systematic undertreatment of each tumor than that all 5 independently exhibited local failure.

 

[Mandelin Rain]

Thanks all for the insights.

I would like to point out that this has become the most rad onc discussion ever, in the best possible way.

 

[Palex80]

At this point in my career, I am ready to call BS on complex imaging sequences or patterns to determine radiation necrosis. They almost never prove helpful. Only time/sequential studies seems to help, and often when neurosurgery goes in, there is both necrosis and viable tumor. Equivicating based on study with an unintelligible name like fourier assisted gradient reconstruction gets you nowhere.

I cannot agree. I find them more helpful than then plain T1 & T2 images you used to have 10 or 15 years ago. Are they magic? No. But they are helpful.

 

[emt409]

And if you find it is necrosis, and it becomes symptomatic, half dose Avastin is literally magical (in appropriately selected patients of course)

 

[scarbrtj]

However, the mathematics are correct for my back of the envelope calculation of the probability of 5 separate lesions independently treated, independently failing.

Again, I get that... and Marilyn vos Savant would likely agree with you. However, are the lesions truly independent (in a coin-toss like manner) or do failing lesions share similar traits of tendency-toward-failure in the host? I feel latter vs former. If truly independent, then in the entire history of the world it is likely a 5-met brain met patient has never had 5-treated-met failure. Why? Because your calc'd prob likelihood of 5-met failure makes that about as rare as a lottery win (and there probably hasn't been 1+ million or more treated 5-met brain met patients in the history of the world).

If we assume the control rate of 24Gy is 95%, the probability of all 5 mets progressing is (5/100)^5 or 3125/10^10 or 3.125 x10^-6.

In fact, it's even less likely than 3.125E-6... you're off by an order of magnitude, yes?: 0.05^5 = 3.125E-7. About a 1 in 3 million chance.

Your example of parsing a single coherent ~1cm tumor into a million conglomerated tumors makes no sense - I'm not going to bother responding to it.

*edit: It may make some sense but only from the standpoint you originally proposed (which I am not sure is right or wrong): looking at each tumor's failure chances individually is a "tumor's eye view" and assumes each tumor is failing/not failing independently of each other. And instead of thinking of a ~1cm tumor as a million conglomerated tumors failing/not failing independent of each other, the great Goitein once talked of an even more micro-viewpoint and a "cell's eye view." That's to what I was inelegantly referring.

 

[evilbooyaa]

And if you find it is necrosis, and it becomes symptomatic, half dose Avastin is literally magical (in appropriately selected patients of course)

As an aside but slightly off-topic, how many places are seeing routine use of Avastin for symptomatic necrosis (assuming insufficient response to dex)? Is it Dex --> Avastin --> Surgery as the escalation? My institution's CNS med-oncs doesn't seem to even offer Avastin for symptomatic radionecrosis.

 

[nkmiami]

Again, I get that... and Marilyn vos Savant would likely agree with you. However, are the lesions truly independent (in a coin-toss like manner) or do failing lesions share similar traits of tendency-toward-failure in the host? I feel latter vs former. If truly independent, then in the entire history of the world it is likely a 5-met brain met patient has never had 5-treated-met failure. Why? Because your calc'd prob likelihood of 5-met failure makes that about as rare as a lottery win (and there probably hasn't been 1+ million or more treated 5-met brain met patients in the history of the world).

In fact, it's even less likely than 3.125E-6... you're off by an order of magnitude, yes?: 0.05^5 = 3.125E-7. About a 1 in 3 million chance.


*edit: It may make some sense but only from the standpoint you originally proposed (which I am not sure is right or wrong): looking at each tumor's failure chances individually is a "tumor's eye view" and assumes each tumor is failing/not failing independently of each other. And instead of thinking of a ~1cm tumor as a million conglomerated tumors failing/not failing independent of each other, the great Goitein once talked of an even more micro-viewpoint and a "cell's eye view." That's to what I was inelegantly referring.

A little bit of an aside, but relevant to this point, I have treated several times treated partial volumes in tumors in the re-irradiation setting and seen a good response including one case who is 2 years out.

Regarding Avastin- I have seen it used twice and it worked very well. Both cases were nsclc, so it may even have anti tumor effect there.

 

[Neuronix]

As an aside but slightly off-topic, how many places are seeing routine use of Avastin for symptomatic necrosis (assuming insufficient response to dex)? Is it Dex --> Avastin --> Surgery as the escalation? My institution's CNS med-oncs doesn't seem to even offer Avastin for symptomatic radionecrosis.

It's hard to tell active tumor from RT necrosis, so it tends to be dex -> surgery (in part for diagnostic purposes) -> bevacizumab.

There is the Alliance BeST trial comparing dex vs. dex + bevacizumab in this setting. They've had a hard time in enrolling. In my experience that seems to be mostly a result of med onc/neuro onc making up their mind what they prefer rather than actually randomizing it.

 

[evilbooyaa]

It's hard to tell active tumor from RT necrosis, so it tends to be dex -> surgery (in part for diagnostic purposes) -> bevacizumab.

There is the Alliance BeST trial comparing dex vs. dex + bevacizumab in this setting. They've had a hard time in enrolling. In my experience that seems to be mostly a result of med onc/neuro onc making up their mind what they prefer rather than actually randomizing it.

But if you do dex and surgical resection and get diagnostic confirmation of necrosis, do you need Avastin after that? I guess if it's like multifocal necrosis or something, but assuming single lesion with necrosis that is resected?

 

[Neuronix]

No need for Avastin if the necrosis is adequately managed surgically. Sometimes surgery is insufficient for control or only a biopsy is performed.

 

[emt409]

As an aside but slightly off-topic, how many places are seeing routine use of Avastin for symptomatic necrosis (assuming insufficient response to dex)? Is it Dex --> Avastin --> Surgery as the escalation? My institution's CNS med-oncs doesn't seem to even offer Avastin for symptomatic radionecrosis.

Our neuro oncs typically prescribe it here. We usually discuss the cases at CNS tumor board, and give Avastin when the neurosurgeons don't want to debride.