"High Risk" Mammaprint Luminal Bs- what was your 'score'?

Hi Tshire:

I've been following along silently since we last "spoke". It is hard to be an unusual case, but you are being very proactive and thorough in your approach.

I did not have the Mammaprint test, but I did some reading to try and learn a little more. Regarding the minus 0.350 number, I found this chart (see top panel):

http://www.agendia.com/media/24Feb15_High-Risk-Lum...

It also contains the words "High Risk Population Average" with a citation to Reference 1 (Buyse et al). If the - 0.350 number is actually an average, this might suggest that patients tend to have smaller negative values. However, without more information, it is hard to tell if the average is distorted by a few outliers (what is the range and distribution?). Also, I checked Reference 1, and I have not been able to find support for the statement that the - 0.35 number is a "high risk population average". I am looking on a screen, so maybe you will have better luck:

Reference (1):http://jnci.oxfordjournals.org/content/98/17/1183....

Reference 1 contains the following statement:

"The previously established classifier defined good prognosis as a probability of 5-year distant metastasis–free survival of more than 90% (5). The gene signature risk classification was given as a dichotomized value only: high or low risk."

Reference (5), which explains microarrays and the development of the "classifier" (including some sort of mathematical algorithm) used to stratify patients as high or low risk, is here:

http://www.nature.com/nature/journal/v415/n6871/fu...

This is just a guess, but it is possible that various clinical validation studies have applied the classifier, and so are informative only with respect to the risk associated with either "high" or "low" risk cohorts, rather than the risk associated with any specific patient values.

It looks like there is an on-going clinical trial (MINDACT). The design of the trial is discussed here, if you have not already seen it:

http://www.nature.com/nrclinonc/journal/v3/n10/ful...

I note that Reference 1 (Buyse et al.) also contains the following statement:

"Raw microarray data and clinical information are available at the European Bioinformatics Institute ArrayExpress database (http://www.ebi.ac.uk/arrayexpress/), accession number E-TABM-77. Normalized microarray data were archived by the Swiss Institute of Bioinformatics, where a study statistician (MD) reviewed and independently applied Agendia's proprietary analysis program to the normalized microarray raw data to reproduce the risk classification. The risk classification agreement between Swiss Institute of Bioinformatics and Agendia was assessed by independent study statisticians (MB, MA, FP). The concordance between the risk classification produced by Agendia and the external team was 100% as verified by the independent statisticians."

I did not have the guts to follow that link!

As you know, I am just a patient/layperson, so please raise your question and confirm anything above with your doctors as is your usual excellent practice.

BarredOwl

Mysunshine:

Sorry if that gobbeldy gook scared you. Please do not second-guess your treatment decision. Also, you do not need to worry if you don't understand what I wrote above (or below).

Sometimes articles help, sometimes not so much! These are about how the mammaprint assay was developed and tested in the clinic to prove its utility.

Basically, I could not find anything which clearly answers Tshire's question. I thought I found some suggestion that the minus 0.350 might be an average, but I could not confirm this. The words "High Risk Population Average" could be referring to the 5-yr and 10-yr percent risks set forth right underneath them (22%, 29%) in the sample report.

Since her high risk number is closer to zero, Tshire is wondering if her individual risk might differ from the high risk group in some way that could further inform her decision. I said that the assay is designed to be binary (high OR low), so it is possible that the particular level of risk associated with individual values (-0.115) might not be adequately characterized. However, there may be other studies, so this is a question for the expert oncologist.

BarredOwl

tshire - I also had Mammaprint done in 2010 and my report does not have that info. I actually just contacted Agendia to ask for all of the reports since what I received was from my BS, not Agendia directly. My interpretation of the graphic is that the example it shows is for a "patient" and their individual risk is -.35, which would compute to a 35% risk (out of 1.00), while the two percentages underneath the box show the average for those who fall into the high risk category, 22% at 5 years, and 29% at 10 years. So on your report you have -.115, which I assume means 11.5% (out of 1.00) and thus a lower "high" risk than the average high risk patient.

Hi:

The test results are rather opaque.

I could be completely out in left field, but I do not think the numerical value (e.g., minus 0.350) is a "percentage". The fact that it is a negative number would also be counter to a percentage. The information for physicians explains it as follows:

"MammaPrint provides a numerical index with a range of -1 to +1, that is overlayed with a binary Low Risk / High Risk clinical classification system. The clinical classification threshold was set by the determination of the largest population of Low Risk patients that can safely withhold chemotherapy."

Here is the origin of that quote (click to read under the "Development" tab):

http://www.agendia.com/healthcare-professionals/br...

As for why they now report an individual number despite the binary nature of the test, it may be for particularity of detail, like: we did this assay, here is the numerical index determined from your microarray test result of 70 genes, and you can see that it falls in the "high risk" category. They may also be using the data for further studies or usage characteristics.

BarredOwl

Hi:

My understanding is that the binary MammaPrint test does not provide individualized risks, but only provides the average risk associated with the high risk cohort or the average risk associated with the low risk cohort.

In addition, I am still of the belief that the index number (e.g. -.035) is a calculated number and is not itself a personalized percentage measure of risk. These tests look at the expression levels of 70 genes. The combined expression levels form an expression pattern or profile. The number is a value on a scale from -1.0 to +1.0 reflecting the correlation of the patient's expression profile to that of a low risk template profile.

The number appears to be generatedfrom the patient's microarray data by a software program which uses an algorithm. The program takes the expression data from 70 genes ("sample expression profile") and assesses its degree of similarity to an established low risk expression profile ("template", which is a "the mean expression profile of 44 tumors with a known good clinical outcome" or low risk profile). After this correlation step, the output number places one in a cohort, and risk is determined as a cohort (which is the average risk of the cohort).

My above comment is based on the following.

Here is a direct link to the Review Memorandum from a page from fda.gov:

http://www.accessdata.fda.gov/cdrh_docs/reviews/k1...

At page 4-5 of that document is the following quote (emphasis added):

"Data analysis is performed according to a specific MamnmaPrint ® FFPE algorithm (MammaPrint® Index, or MPI). The algorithm calculates the correlation of the sample expression profile to a template (the mean expression profile of 44 tumors with a known good clinical outcome) and determines the molecular profile of the sample. This algorithm is designed and programmed by Agendia and compiled into a standalone software program, "X- Print Analysis Software". The "X-Print Analysis Software" loads a data file (CSV) which is created by the laboratory technician by extracting specific information from the laboratory database. The "X-Print Analysis Software" reads the CSV file, opens the Feature Extraction Software data files (TXT), performs quality control checks, determines the sample expression profile, calculates the correlation of sample profile to the "Low Risk" template profile on a scale of -1.000 to +1.000 (MammaPrint® FFPE reportable range), compares the calculated correlation to a pre-defined cut-off value and determines the samples prognostic profile (i.e., Low Risk, High Risk, Low Risk Borderline, or High Risk Borderline)."

One reason for reporting individual MPI values is because, as explained on page 12, some values are considered "borderline":

"The "borderline region" refers to a range of MPI numbers surrounding the clinical classification threshold. In this region the clinical classification result accuracy potentially falls below the overall predetermined analytical accuracy of 90%."

According to Table 10 of the FDA document linked above, the "borderline" cut-offs are different, depending on which test was conducted, either "MammaPrint FFPE" (conducted with formalin-fixed paraffin embedded (FFPE) tissue samples) or "MammaPrint" (conducted with fresh tissue samples).

Regarding such "borderline" values, the Physician's Brochure states (emphasis added):

http://www.agendia.com/media/M-ROW-010-V2_Breast-C...

MammaPrint and MammaPrint FFPE

The MammaPrint result is given as "Low Risk" or "High Risk" for risk of recurrence. The MammaPrint Index of a sample can fall within a pre-defined area around the classification threshold in which the MammaPrint result has <90% classification accuracy (i.e., borderline sample). When a sample is considered to be "borderline", it is clearly indicated on the MammaPrint analysis report.

BarredOwl

Hi mysunshine:

My last post was about the steps that I think Agendia takes to get from a sample of a tumor received from a patient to a MammaPrint report, using various molecular biology and bioinformatics techniques.

Nothing that I said above would change your report or any treatment decision you made based on it.

BarredOwl

Hi Tshire:

My comments below are based on my assumption that you had a "MammaPrint FFPE" test done, the numerical index value is - 0.115 (a negative value), the classification is "High Risk", and there is no indication on the report you received that the sample measurement was "borderline". (From the advice you received from your oncologist, I don't think it would be borderline, but the content of your report controls).

I think you understand it in general, but I would make the following distinction about the template or comparator (the standard to which a test sample is compared to). The distinction is that (from the information I cited) there is no "high risk template" for the MammaPrint FFPE assay.

There is a single comparator or template profile, and it is a "low risk" template:

'. . .calculates the correlation of sample profile to the "Low Risk" template profile on a scale of -1.000 to +1.000 (MammaPrint® FFPE ..."

The "template" is "the mean expression profile of 44 tumors with a known good clinical outcome."

So, your negative number, which is below the clinical classification threshold of zero for MammaPrint® FFPE, means that the expression profile obtained from your sample is different enough from the low risk template profile that it is not classified as low risk (by this computer-based measurement). The value generated places you in the "high risk" category, with an average risk determined by clinical studies for the high risk cohort.

Within the "High Risk" category, I am not aware of any data or evidence that establishes differences between the risks associated with different index values that fall within the high risk range (e.g., - 0.350 versus - 0.115). Consistent with this, you noted above that your oncologist did not assign any import to the particular index value, other than "high risk".

You then asked why report the particular value. Some index values near the clinical classification threshold of zero are deemed "borderline" and "[i]n this region the clinical classification result accuracy potentially falls below the overall predetermined analytical accuracy of 90%." As noted in my prior post, current reports will clearly indicate a "borderline" result.

BarredOwl

Hi Tshire:

It would be my understanding that by this particular computer-based measurement, your expression profile was more similar to the template profile than Mel's profile was.

I think we are on the same page now Tshire, and it is a good hypothesis to test. Part of reporting these numbers and collecting patient outcome information is that these types of studies are likely underway, retrospective and prospective.

BarredOwl

Hi Mel:

In a way, it seems like the RASTER trial says more about the low risk group:

http://www.agendia.com/patient/breast/current-clin...

"In the first ever prospective, observational study for a breast cancer recurrence assay, the five year prospective results of the RASTER study were published indicating that chemotherapy may be safely withheld in MammaPrint identified Low Risk breast cancer patients. The Low Risk patient group, which primarily chose to forego chemotherapy, had a 97% distant recurrence free interval (DRFI) at 5 years. Published studies continue to support the strong prognostic and predictive clinical utility of Agendia's genomic assays and ongoing large, multi-site clinical trials seek to expand clinical indications for use."

Edited to Add: "The MammaPrint® FFPE result is indicated for use by physicians as a prognostic marker only, along with other clinico-pathological factors." In other words, MammaPrint is not recommended as a stand-alone (does not dictate decisions), but is used in conjunction with other factors.

BarredOwl

No I didn't have an oncotype. My surgeon likes the mammaprint better.

TShire:

As an aside, I have been envisioning the marketing team at Agendia looking at their competitor product Oncotype DX, and working on their spin about the advantages of their "binary" test: Output is High Risk or Low Risk. No big fat intermediate range of uncertainty (aside from the MammaPrint "borderline" situation, which is probably relatively uncommon).

However, if the on-going trials of Oncotype DX succeed in removing some of the uncertainty of its intermediate range, then the pressure will be on Agendia for clinical trial results and modifications of MammaPrint to increase its information output.

BarredOwl