Comprehensive validation of the Oncomine Dx Express Test for CDx and tumor profiling applications

Down Stream
6 followers
14 Video Views·Mar 11, 2026
A deep dive into the real-world studies to validate the Oncomine Dx Express Test.
All right, thank you David for that comprehensive overview of ODXET.
So in this portion of the talk, I'm going to go over some of the studies the team executed
to not only demonstrate robust performance, but also prove it's safe and effective.
So in this slide here, it shows a snapshot of some of the studies that we conducted.
We had to extensively validate this assay both analytically and clinically.
And so you can see here are some of the studies, everything from establishing sensitivity and
specificity to ensuring we have really good concordance to an already established orthogonal
assay. And as you can see, some of these also centered around the specimen itself.
The clinical FFPE samples are inherently variable, right? And so we did a lot of testing
looking at tumor content and understanding the interfering substances that come along with it,
etc. We also looked at a lot of different stability testing. So not only on the sample
itself, but the reagent, ensuring it can be transported well, and also in use within the
workflow. So some of these studies I'll go over in further detail, but let's dive into
the first set of studies. So first we had to establish our limited blank, essentially LOB,
or specificity. And so this was essentially done to estimate the false positive of rate for the
assay across replicate testing in well-type samples. So in this study, we used 24 different
well-type samples from clinical FFPE samples that spanned across 10 different tissue types,
such as breast, colon, lung, etc. And these samples were thoroughly pre-screened using
an established NGS orthogonal assay, because we really wanted to make sure these samples
served as true blank samples for the assay. So the study was then conducted with two reagent
lots and four replicates per tissue per lot. So a lot of replicates built into the testing.
False positive rates were essentially calculated two ways, per position and per sample. So per
position essentially just looks at the total sequenced positions. And then sample is just
taking the total number of samples and replicates and using that essentially as a denominator for
the calculation. So the false positive rate for the assay came out to be 0% across these 24 blank
samples across these 10 different tissue types. However, we did want to note that there are some
false positives. The mean allelic frequencies for these SMVs are around 3.5%. So this was below our
orthogonal assay's LOD and towards the lower end of our LOD. So in order to mitigate some of the
false positive reporting issues, we went ahead and masked these three, these level three SMVs
from reporting. We then looked at limited detection or LOD. And essentially this was
conducted by creating DNA and RNA variant blends across clinical FFPE samples, again,
with 10 different cancer types. We blended them to six different blend levels, so different
variant allelic frequencies. And the study was conducted using two reagent lots with 10 replicates.
So this gave us over 120 data points per variant. LOD was conducted for both our CDX variant of
interest, those EGFR exon 20 insertions, as well as a representative approach for our tumor
profiling variants covering all of the main variant classes, SMVs, indels, fusions, etc.
And so the detection rates across these different blends were analyzed using a regression,
appropriate regression analysis. And LOD was estimated. And so essentially what's the lowest
frequency the assay can reliably detect 95% of that time. And so if you zoom into this portion
of the table, you can see the LOD ranges that were achieved. And so we've got it broken down
by our CDX variants, so EGFR exon 20 insertions, as well as our tumor variant classes, SMVs,
insertion solutions, CMVs, RNA fusion variants, as well as our imbalanced target. And so just
highlighting the CDX LOD ranges, we were able to achieve allelic frequencies of 3.24% to 4.38%.
So both these studies together demonstrate really strong performance of the assay at very low
detection levels. Next, we looked at precision. And we did this two ways. So looking at it both
from a sample reproducibility aspect, as well as panel reproducibility. For a sample, that
essentially just starts at the extraction level. But for the purpose of this talk, I'm going to
highlight the assay reproducibility. Okay, and then so we did this both for our CDX variants
of interest, EGFR exon 20 insertions, as well as tumor profiling. So here, let's talk with the CDX
variants. The repeatability and reproducibility was assessed using four EGFR exon 20 insertion
variants. So these cover four different insertion lengths, 3, 6, 9, and 12 base pairs,
including wild type samples. And we introduced variations that were sourced from different sites,
different operators, and different instruments. And this generated over 144 data points per variant
at two blended LOD levels. So near LOD, 1 to 1.5x, and 2 to 3x. So different call rates
were calculated. So no calls, positive and negative call rates. And so as you can see here,
the positive call rate was 100% for all of our EGFR exon 20 insertion variants. And then also
for our wild type samples, the negative call rate was 100%. Repeatability was also assessed.
So for wild type samples within repeatability was also 100%. And then for tumor profiling variants,
we also assessed variability right across three external sites, two operators per site,
and then four instruments per site. And so this gave us a lot of variation to look at.
And so this table right here shows the positive call rate. And as you can see here on the left,
breaks down by variant classes, our SMVs, insertions, et cetera. And then the two
blended levels again, 1 to 1.5x and 2 to 3x. And so just zooming in in one of the example
classes for insertions, you can see that we achieved over 99.8%
reproducibility for our tumor profiling variants, especially near LOD. In this next slide,
it shows the negative call rate. And so again, just zooming in in one of the variant classes
for insertions, at near LOD, we got a negative call rate of 99.34%.
Repeatability was also assessed for tumor profiling variants, and a variant component
analysis was performed. And the results showed a percent CV or coefficient of variation of less
than 10% between operators and instruments. However, there were some exceptions to note
here. Three variants that came in a little higher than 10%. But overall, still really good
repeatability. And then maybe not too surprising, our highest CV came from our RNA variant.
Another notable study to highlight is our analytical accuracy. And again,
this was done for both our CDX and our tumor profiling variants.
And so essentially, the accuracy of ODXCT or Oncomine Express test for detecting our EGFR
X124 CDX variants or our tumor profiling variants was evaluated by comparing these results to an
established orthogonal assay. And so for our CDX variants, this again was using an NGS-based
orthogonal method. And then PPA, MPA, and OPA were calculated. So essentially, PPA was just looking
at the agreement of ODXCT positive calls with our orthogonal assay. And so here on the table
to the right, you can see the results for PPA, MPA, and OPA. And so when compared to our orthogonal
assay, the accuracy of EGFR 20 insertion detection was over 95%. And so you can see here for PPA,
100%, MPA, 95.5, and then overall, 97.4. We then assessed our accuracy of our tumor
profiling variants using a representative approach. And so this was actually conducted
over three separate studies. So study one here is the bolus of testing. Over 460 or so FFPE
clinical samples were tested in 15 or so different tissue types. And so as the table up here
summarizes the results from study one, and here to the left breaks down the variant classes.
And then you can see over 586 variants were evaluated. And so it breaks it down by
concordance of the ET positive calls to the orthogonal positive and then the negative calls.
And so even though there were some discordance with the orthogonal assay, we still achieved
an accuracy of over 98.2% for all variants. And if you break down specific variant classes
insertion, so it's like 100%. In study two, this looked at additional level two variants,
and specifically it was looking at ERBB2CNBs in breast cancer, ALK fusions in lung cancer,
and then RET fusions in both thyroid and lung. And this table down here summarizes the results
from study two. And so if we zoom in just here on the level two ERBB2CNBs, we had really good
concordance with the orthogonal assay, 29 positive and the 37 negative as expected.
We did have four discordance, but this resulted in 100% for the PPA and 90.2% for the MPA.
And then down here has the fusion results, the level two ALK variants, and the RET variants
right here. And so again, really good strong concordance with both positive agreement and
negative agreement, but we did have one discordance for each. And so this brought our PPA of 100%,
but then MPA for about 97%. And so overall, when compared to the orthogonal assay,
the accuracy of detection across these variant types were still really good with over 90% accuracy.
The last study for accuracy was study three, and this essentially
evaluated our EGFR T790M and EGFR L858R SMVs in addition to EGFR exon 19 deletions.
And again, this is just broken out. You've got your level two EGFR, the SMVs, and the deletions.
And overall, the positive concordance was really good, 23 positives, 245 expected negatives,
with one discordance. So PPA of 100%, and then MPA of over 97%. So again,
when compared to the orthogonal assay, the accuracy of our EGFR mutations was really good.
The last study I wanted to mention was our clinical study. And so essentially, this was a
clinical bridging study that was conducted to essentially establish our safe efficacy of the
assay in detecting EGFR exon 20 insertions in specifically non-small cell lung cancer,
and identifying these patients that are good for treatment. And so the overall data set consisted
of 192 patients, but the primary efficacy population was 85. These are the patients
that received 200 milligrams of the drug. And you can see for this cohort, the overall response
rate, or ORR, was about 46%, with six of them having a complete response and 40% of them having
a partial response. Within this data set, 58 of them actually were both ODXET positive as well
as CTA positive, giving an ORR of 41%, 5% of them having a complete response, and 36% of them having
a partial response. So in summary, these results really demonstrate that the overall response rates
for ODXET EGFR exon 20 insertion positive population aligned with the CTA positive
population, giving us really good concordance there. So in summary, I just want to hope you
were able to take away and understand the breadth of testing that went into validating this assay.
And just wanted to leave you a little snapshot or peek into our NGS lab and leave you with some
really neat statistics. So in all, we submitted 31 formal analytical validation studies to the
FDA for review. And the team extracted and processed over 12,000 DNA samples and 14,000
RNA samples. And that resulted in over 1,500 sequencing runs to provide you an FDA-approved
Oncomine DX Express test. So thank you very much for your attention. And so now I would like to
pass it over to Steve from Argent Global Services. Thank you.
