NGS for oncology - is bigger better?

[cellmatrix]

Is anyone aware of major publications convincingly demonstrating value of large comprehensive genomic panels versus smaller (<50 genes), tumor specific targeted panels? Is whole exome plus whole transcriptome a good idea for routine clinical laboratory testing? Or is it better for the clinical research and data mining? I think the companies offering >50 gene testing for oncology often employ a different model where they are more interested in just acquiring data and then monetizing it for sale to bio pharma.

I also think there is a reason why CPT code 81455 for >50 genes is universally denied in many regions and among private payers.

also, I have interacted with clinicians and pathologists who argue that because broader testing results in higher rates of variants of uncertain significance, this is a pertinent negative.

what should define the sweet spot between breadth of coverage versus clinical actionability?

---

Members don't see this ad.

 

[bauber]

Is whole exome plus whole transcriptome a good idea for routine clinical laboratory testing? Or is it better for the clinical research and data mining? I think the companies offering >50 gene testing for oncology often employ a different model where they are more interested in just acquiring data and then monetizing it for sale to bio pharma.

Okay so whole exome or transcriptome or both is probably crazy for any kind of clinical testing outside of very special pediatric genetic things no one can figure out. Maybe there is another clinical scenario out there for this but I think it's mostly research stuff.

I'm going to assume you're mostly referring to somatic testing of tumors by NGS - and you are thinking about it in the right way...I have a similar opinion - why are these companies doing 500 gene panels on a lung cancer when all that really matters are a handful of actionable genes? Data collection is a big one. Future "actionability"? Are they really going back and re-reviewing cases from years ago? I wouldn't suspect so. However, you do need a big panel to calculate tumor mutational burden, which is becoming more and more of an important biomarker in the age of immunotherapy.

There are ways to design your NGS assay where you are reporting out organ/tumor site specific gene panels but sequencing much more in order to calculate TMB. I suspect this may be what neogenomics is doing with their Neotype panels. I'm not sure what is required on the bioinformatics end or how they bill this.

I favor an organ/tumor-site specific approach. Clinicians are already confused and overwhelmed with the amount of data and molecular tests and methodologies coming their way. Giving them 25 VUS in a 20 page report doesn't help anyone. Maybe GBwillner can comment.

 

[gbwillner]

I have been summoned....

This is obviously a very complex issue, so I will just make a few points and address a few questions.

Is there data suggesting comprehensive testing is more effective/better than single gene/ small panel testing? There are a lot of data around this but this question itself is seldom directly addressed (for good reason). In my professional opinion, the answer is absolutely yes, but there are a lot of caveats.

On a high-level: cancer is a genetic disease. The drivers of cancer activation are fairly well known at this point although we are always learning new things. The idea that one gene/one disease that lead to much of our early clinical lab development and drug development is antiquated and no longer reasonable today. But this is the way we thought about the world for a long time and is reflected in much of the study design and drug development efforts that now hamper the use of NGS. Current efforts are moving beyond this idea to one of understanding the genomic landscape of the tumor to understand the cause driving oncogenesis in the individual patient- AKA precision medicine; this is helpful because we already have numerous available therapeutics that target these specific genetic alterations. For this to work we need to actually understand what is driving the tumors, and comprehensive NGS is a really good way to do this. NGS can be thought not only as a predictive "test" but a complex analysis of the underlying genetics of a tumor. This is a paradigm shift in how we see testing. An analogy: when you get Bx of a colorectal specimen, you are asked to render a diagnosis. You are not asked "I only want to know if this is invasive cancer, Y/N" (ok, maybe the GI does only ask you that). There are other relevant possibilities. Are there any clinical studies comparing the relevance of rendering other diagnoses beyond invasive cancer? If not, why do we do all the other work? Unnecessarily limiting your focus in such a scenario could obviously negatively impact patient care. Such an intentional limitation is scope can lead to something called the "spotlight effect"- we can only find things when we are looking for them. NGS allows us to find other things that are relevant in the genetic sense. The reality is that cancer, and subtypes of cancer, are human-made concepts. If there are recurrent genetic alterations that drive oncogenesis or exploitative biological genetic weaknesses regardless of cancer type, we should focus on that. And we are moving in that direction although testing and development efforts will take some time to fully adjust. There are already therapeutic interventions based on this concept, and there are more every day. There is already clinical utility for such testing.

The analyte-specific nature of testing is changing. The relevant question I would ask: Who cares how many genes are tested? What is important is that the right genetic information is acquired; there is no "right" number of genes tested and it is dumb to keep counting them when the cost basis of gene testing in the NGS space is largely immaterial. Similarly, the direct comparison asked by cellmatrix has never yielded good results, but mainly as a result of the study designs being inherently limited by the spotlight effect and limitations in existing drug use approvals. There is a lot more to this but I will stop here, except to say the main reason payors don't pay for these services is three fold:
1. They don't understand what test was performed or why and it is expensive due to codeset limitations
2. Most tests are LDTs and payors (generally) don't have a good way of evaluating if the specific test is useful
3. This is complex and new and scary. Also, semantics are important... and CPT code 81455 requires that >50 genes sequenced is NECESSARY for standard of care of that individual patient. Such scenarios seldom if ever exist; it is not that comprehensive NGS does not have clinical utility.

The caveats abound, however. The question/comment proposed by bauber:
" Okay so whole exome or transcriptome or both is probably crazy for any kind of clinical testing outside of very special pediatric genetic things no one can figure out. Maybe there is another clinical scenario out there for this but I think it's mostly research stuff. "

I think this is exactly right. Tests should only be performed when they have existing clinical utility. For a cancer patient- what are you going to do with the 26,000 genes that are not related to cancer? Including them in the test only weakens it's ability to detect relevant alterations accurately, especially if there is low tumor cellularity. Transcriptomes at this point do not have any established utility (beyond their use in detecting gene fusions). However, I can imagine they are likely to once we establish how to use that information accurately and reproducibly.

I will disagree with this statement:
"all that really matters are a handful of actionable genes"

There are over 150 genes of clinical relevance in cancer today. There is nothing wrong with validating a broad assay and "masking" data you don't want to see to make sub-tests of a larger panel specific to different cancer types- this is a common strategy and can be cost-effective. However, I don't think this adds any value and may actually prevent patients from enrolling in clinical trials or providing opportunities for off-label use of therapies that have scientific and clinical evidence behind their efficacy (even if not approved).

 

[bauber]

Great reply. Perhaps I was too reductive - the handful of actionable genes is becoming/has become an armful and likely will become more. I suppose limiting it to 'actionable' is wrong as well, as there are many genes (especially in the heme world - MDS, AML, etc) that are not actionable but predictive of disease behavior, and that's important info for clinicians and patients. I'm sure there are more out there waiting to be found.

 

[cellmatrix]

I found this month's AJCP article, The Utility of Next-Generation Sequencing in Advanced Breast and Gynecologic Cancers, to be especially informative. NO major, statistically significant differences were noted in survival among FoundationOne tested samples with actionable versus non-actionable results. This is due to limited uptake in clinical trials, logistical factors, but probably also the gene content itself.

As a field, $billions$ have been spent over the past 5+ years on personalized medicine, and yet NCCN guidelines remain focused on sets of biomarkers that are very circumscribed, and CMS reimbursement for >50 genes remains extremely limited for non-FDA approved, LDT NGS assays.

Unless discovery and clinical research are the primary mission, then I would think that for oncology NGS testing in the routine clinical laboratory setting, bigger is not better, but rather, smarter, more targeted NGS panels are judicious and appropriate. Smaller panels are certainly also significantly easier to operationalize for most clinical laboratories, especially in terms of both cost of testing and clinical annotation requirements. I think the only clinical utility for sequencing hundreds of genes would be to power a tumor mutation burden calculation (and TMB is merely a very crude, and ultimately surrogate biomarker for neoantigen expression and other events that would be expected to influence responses to immunotherapies).

 

[gbwillner]

I found this month's AJCP article, The Utility of Next-Generation Sequencing in Advanced Breast and Gynecologic Cancers, to be especially informative. NO major, statistically significant differences were noted in survival among FoundationOne tested samples with actionable versus non-actionable results. This is due to limited uptake in clinical trials, logistical factors, but probably also the gene content itself.

As a field, $billions$ have been spent over the past 5+ years on personalized medicine, and yet NCCN guidelines remain focused on sets of biomarkers that are very circumscribed, and CMS reimbursement for >50 genes remains extremely limited for non-FDA approved, LDT NGS assays.

Unless discovery and clinical research are the primary mission, then I would think that for oncology NGS testing in the routine clinical laboratory setting, bigger is not better, but rather, smarter, more targeted NGS panels are judicious and appropriate. Smaller panels are certainly also significantly easier to operationalize for most clinical laboratories, especially in terms of both cost of testing and clinical annotation requirements. I think the only clinical utility for sequencing hundreds of genes would be to power a tumor mutation burden calculation (and TMB is merely a very crude, and ultimately surrogate biomarker for neoantigen expression and other events that would be expected to influence responses to immunotherapies).

View attachment 332933

This is not necessarily a good way to answer a question of the clinical utility of a test- there are far too many confounding variables related to give you any sort of clear picture.

I can't argue that "bigger is not better" since I don't truly believe that to be the case, but I definitely agree with "smarter".

Per NCCN guidelines the best treatment available to a patient is within a clinical trial in several common cancer types. If this is considered true, then a larger panel allows for more flexibility and appropriate placement in some trials. From that perspective, bigger certainly is better.

I think the point at hand is- what do we do with all the data for non-actionable information? The problem I see confounding the conversation is really about cost. I would argue a large NGS test is really not significantly more expensive at this point, and gives you all the information you need at once (plus some you don't really need). The alternative is a series of other tests, that when concatenated together are just, if not more, expensive.

 

[cellmatrix]

This is not necessarily a good way to answer a question of the clinical utility of a test- there are far too many confounding variables related to give you any sort of clear picture.

I can't argue that "bigger is not better" since I don't truly believe that to be the case, but I definitely agree with "smarter".

Per NCCN guidelines the best treatment available to a patient is within a clinical trial in several common cancer types. If this is considered true, then a larger panel allows for more flexibility and appropriate placement in some trials. From that perspective, bigger certainly is better.

I think the point at hand is- what do we do with all the data for non-actionable information? The problem I see confounding the conversation is really about cost. I would argue a large NGS test is really not significantly more expensive at this point, and gives you all the information you need at once (plus some you don't really need). The alternative is a series of other tests, that when concatenated together are just, if not more, expensive.

Thank you for the further insights. I have seen various flavors of apples and oranges being compared in discussions of this question, which I believe is central and important for molecular oncology diagnostics in the US routine clinical laboratory setting. The confounding issue is not just cost (which will become progressively lower with time and has already become markedly reduced). For example, IMHO, major impactful factors also include molecular pathology laboratory operational burden to interpret and report myriad variants across extraneous gene content, and for signers to assume legal liability for the additional numerous variants included in the report that likely won't be manually interrogated and double checked.

For routine clinical testing and care in the community setting, of an advanced stage non small cell lung cancer, is a panel reporting variants among all 300-500 genes wise? Or would a more targeted panel specifically reporting across the ~20 genes explicitly mandated in the NCCN guidelines for DNA variants and RNA fusions, together with TMB and tumor-only MSI be better?

I think many academic labs are using 3rd party annotation solutions to include variants from all 500 genes from some commercial panels, but comments are then only rendered for the Tier 1 or Tier 2 variants; every other variant detected is blanket interpreted as either a VUS or as "detected but not actionable".

I’m partial to the more targeted approach, as it significantly streamlines the reports by minimizing the VUS rate, and perhaps most importantly, 5-50 gene panels fit into the CMS reimbursement framework for LDTs. Most parts of the country MACs support zero reimbursement for >50 genes if the assay is non-FDA LDT, effectively a non starter for most labs to facilitate patient access. Furthermore, it may take only minutes to review the data and minimize the TAT for 5-50 gene, tumor type specific, more targeted panels, versus days for the large comprehensive panels, assuming that ALL variants are double checked in the latter)....less practical and only fit for "potential future utility" or data mining for eventual use in clinical research or sale to pharma (both of which, at this point in the field, are not especially high yield either, in my opinion.

 

[gbwillner]

...

For routine clinical testing and care in the community setting, of an advanced stage non small cell lung cancer, is a panel reporting variants among all 300-500 genes wise? Or would a more targeted panel specifically reporting across the ~20 genes explicitly mandated in the NCCN guidelines for DNA variants and RNA fusions, together with TMB and tumor-only MSI be better?

I think many academic labs are using 3rd party annotation solutions to include variants from all 500 genes from some commercial panels, but comments are then only rendered for the Tier 1 or Tier 2 variants; every other variant detected is blanket interpreted as either a VUS or as "detected but not actionable".

I am actually a big proponent of the latter approach. VUS should be reported but not discussed (there is no evidence or not sufficient evidence for clinicians to impact their management on them EXCEPT in some limited circumstances). These would be as a last resort for the patient OR to put them on trial. Of course, the clinician should NOT try to interpret a VUS, but rely on an expert (a MGP) to help identify as part of a consultation of any VUS' can be useful for further purposes. Of course, none of that is possible if the VUS are not tested or reported. To be clear, "clinical evidence" falls under a wide array of quality and quantity. A VUS may be a future actionable mutation, dependent on factors outside the tests' control. Better to have that data up front.

I think a lot of concerns are around the lack of communication or understanding by clinicians. The solution to that problem is not to reduce useful complexity of testing- it is to improve communications and workflows.

... Most parts of the country MACs support zero reimbursement for >50 genes if the assay is non-FDA LDT, effectively a non starter for most labs to facilitate patient access.

This is false. This policy is effective in 28 states.

 

[cellmatrix]

I think if other MACs follow suit after trendsetter MolDx and allow CPT reimbursement of 81455, it would significantly enhance and increase patient access to comprehensive genomic profiling and reporting >50 genes, using LDTs. Although I strongly support the latter approach of DNA + RNA level abnormalities covered within 5-50 gene NGS panels aligned to NCCN guidelines, if CMS reimbursement were greater, I'd probably incorporate a 3rd party annotation support solution into our clinical workflow. But until then, the costs are prohibitive in the Northeast, and Texas, I believe.

What are the chances that other MACs, e.g. Noridian, Novitas, etc. would adopt policies akin to MolDx? What is the best way to engage or lobby them? Given the NGS NCD, my concern is that appetite may be limited to encourage greater reimbursability and uptake for non-FDa, LDTs for broad (i.e. >50 gene) NGS profiling.

 

[gbwillner]

What are the chances that other MACs, e.g. Noridian, Novitas, etc. would adopt policies akin to MolDx? What is the best way to engage or lobby them? Given the NGS NCD, my concern is that appetite may be limited to encourage greater reimbursability and uptake for non-FDa, LDTs for broad (i.e. >50 gene) NGS profiling.

Noridian is in the MolDX program, so 100%. Additionally, with these policies in effect, the other MACs have started to follow suit, unless they can demonstrate through policy why they disagree with MolDX policies.

 

[deleted314957]

I’m a little remote from this but it seems to me that they have gone from effective therapy for a few actionable mutations (CML,GIST) with some pretty circumspect testing to now they can and do perform essentially unlimited genetic testing. The bottom seems to be a panel of dozens of mutations to whole genomic sequencing. But has all the increased testing led to lots more actionable genes than with more limited testing? And it only makes economic sense to focus expensive therapy on FREQUENT actionable mutations. We can’t fall into the pit where EVERY tumor gets complete sequencing, a couple/few mutations are identified that are relatively “minor” drivers of growth, and then do full court presses with VERY expensive therapy of iffy effectiveness.

I can understand the research interest in and usefulness of identifying as many actionable mutations as possible.