Just how helpful is personalized genomic analysis anyway?


Regular readers of the medical science literature will be very conscious of the emphasis on genomic analysis of tumor specimens as a way to try to “personalize” treatment of cancers of many types — prostate cancer very specifically included.

However, what many patients and laypeople may be less aware of is that “just” carrying out genomic sequencing on the DNA of tumor samples is not enough to offer reliable guidance about the most appropriate ways to treat cancer in a specific individual. One also needs to have carefully correlated genomic analysis of matched normal DNA from the same patient.

This has been brought home by a study by Jones et al. published very recently in Science Translational Medicine, and discussed in something closer to “English” in a commentary on the Medscape Oncology web site. We should also note that this study was not exclusive to prostate cancer but used tissues from 15 different cancer types.

The bottom line to the study is that although we now have the technology (so-called “massively parallel” approaches to DNA sequencing) to characterize the precise genetic makeup of patients’ tumors, and to select therapies based on some of the identified mutations,  Jones et al. show that, when they carried out a tumor-only sequencing approach, using 815 tumor samples from patients with the 15 tumor types,

  • It was not possible to definitively identify key changes in cancer-predisposing genes.
  • This approach led to false-positive findings in 31 percent of alterations identified in targeted gene analysis and 65 percent of alterations identified in exome analysis.

In other words, tumor-only sequencing gives us very dangerous data and can lead to some very bad ideas about what treatments might work in individual patients.

By comparison, when the research team looked at  sequencing data from 815 tumor-normal paired samples from patients with the same 15 tumor types, they were able to

  • Identify genomic alterations in whole exomes with sensitivities of > 95 percent and specificities of > 99 percent
  • Identify genomic alterations in 111 targeted genes with sensitivities of > 99 percent and specificities of > 99 percent
  • Show that > 75 percent of cases had somatic alterations in genes associated with known therapies or current clinical trials
  • Identify germline alterations in cancer-predisposing genes in 3 percent of patients with apparently sporadic cancers.

Quoted in the commentary on the Medscape Oncology web site, Dr. George Demetri of the Dana-Farber Cancer Institute explains that, in his opinion, the medical community has, to some extent, over-represented the importance of tumor sequencing to the public to date — because the potential is so exciting.

However, we are still in the relatively early stages of understanding how to actually apply genetic and genomic sequencing in actual medical practice. In a striking example of this, Dr. Demetri notes that

Cancer has lots of mistakes. What’s just random noise in those mistakes and what’s a true pathogenic causative or contributing factor in that cancer is not clear for the vast majority of human cancers.

He goes on to explain that, when you can see a small chip or flaw ( a “ding”) in the paintwork on your car,

Is there a ding in the paint, an abnormality? Yes there is. Is that going to hurt the way the car runs? Absolutely not. So yes, the abnormality’s there, but what does it mean?

Just because we can find a genetic “ding” in the DNA of a patient’s cancer also does not necessarily mean that that “ding” is a real problem at all. The effects of each “ding” need to be compared to the effects of the normal situation in that specific patient before we can make any sense of whether the ding really matters.

This massively complicates how to interpret whether having a particular genetic abnormality is really meaningful. Are we making progress on this front? Sure we are. But that doesn’t mean that when your oncologist can tell you you have a particular genetic abnormality in your cancer cells, that it can necessarily imply you need treatment X as opposed to treatment Y. We still have a long way to go before we can do that with a high level of accuracy, and sequencing of paired samples of tumor and normal tissue from individual patients will be essential to that accuracy.

The personalization or individualization of cancer care is an extremely worthwhile long-term goal, but getting there for every patient with every cancer may prove to be extremely challenging over a long period of time.

10 Responses

  1. Yes, this is where the clinic meets the media. I see patients posting about this or that research but when you bring most of those ideas to the clinic, there is often no real world application. Luckily I found an a very smart oncologist who actually listens to me and then explains to me what tools he has available for me.

  2. Is it therefore too early to have a Prolaris test or Oncotype Dx?

  3. Our future will be translational treatment with genomic modeling of tumors based on appropriate sequencing. But we are nowhere near that mark today. I proposed that specific question to Maha Hussain on CureTalk Radio when I already knew the answer. It’s our future. It’s not our today. I hear a lot of talk about guys asking to get genomic tests as though it may help change the course of their cancer treatment (a la Angelina Jolie). But it’s baseless. Not yet. Probably not while I’m alive. It’s a great science and it will reap great benefits in the future. But we are missing so much needed information today. We just cracked the nut open on prostate cancer genomics and the combinations are enormous.

  4. Paul:

    Re your question above about the Prolaris and Oncotype Dx tests. No; its fine to get those types of test done for prognostic purposes. They are tests designed exclusively to assess risk of aggressiveness of your disease. They are not intended to assess the likelihood that a particular type of treatment would or wouldn’t be more effective than others.

  5. I think you do current research an injustice when you say we have a long way to go before this information will be useful! Sure, we don’t quite have a complete picture of the hugely complex genetics of cancer yet, but genetic diagnosis already improves patient care in many ways. In breast cancer, for example, we now know that people who have mutations affecting the BRCA1 gene respond well to platinum-based chemotherapy.

    One thing that seems to be constantly missing from these studies is detection of dosage-changing variants … were inversions/deletions, etc., included in this study? These events can occur very frequently within tumours where they can increase or impair gene expression. I worry that some researchers consider whole genome sequencing data to be the whole story, which could result in us neglecting potentially important data, such as genomic rearrangements and structural changes that can have epigenetic consequences…

    On the other hand, one study can’t check for everything. The most important thing to do for us to understand cancer genomics is collect lots of data and make it available for other scientists to use.

  6. Dear Claire:

    With respect, this web site is primarily about prostate cancer and its treatment as opposed to being about cancer research in general or even cancer genetics.

    There are certainly some forms of cancer in which we have made enormous progress based on the genetics of those cancers and how they can be treated. The development of drugs that can be used specifically to treat ALK-positive forms of non-small-cell lung cancer may be one of the most striking (even more striking than platinum-based chemotherapy for BRCA1 breast cancer). However, in the case of prostate cancer we are still a very long way from any correlations between specific genetic variants and the ability to treat that form of prostate cancer.

    The value of genomic testing in the treatment of prostate cancer at present is therefore very small indeed unless it is being done with very specific research goals. Are there “dream teams” working on such issues? Yes there are. Do I expect progress? Yes I do. Do I expect that progress to come soon? No, I don’t. Do I think that most prostate cancer patients today will gain anything from asking their doctors about genomic testing and the potential for correlation to treatment? No, I don’t — with a single exception. It does appear that the AR-V7 variant makes it unlikely that carriers of this variant will gain noticeable benefit from treatment with some of the newest drugs like abiraterone acetate and enzalutamide. Unfortunately, as yet, there is no commercially available and simple test for this variant.

  7. Oh I see. I don’t know so much about prostate cancer and was speaking more generally. Thanks for replying.

  8. Article on 4/21 regarding olaparib suggests usefulness of BRCA1/2 gene findings

    My impression is that we are on the threshold of great usefulness of genetic analysis. In a few cases, at least, like olaparib, it looks like we are there!

  9. Dear Jim:

    With respect, I think you are over-interpreting this study. It is less than clear that all the responders in the study were BRCA1/2 patients, and only four patients maintained a meaningful response for 12 months.

  10. For anyone reading this April presentation, there is a valid test for the AR-V7 varient. It is available from Johns Hopkins now.

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out /  Change )

Google+ photo

You are commenting using your Google+ account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s

This site uses Akismet to reduce spam. Learn how your comment data is processed.

%d bloggers like this: