Greetings, and welcome to Personalis Clinical Update Call. At this time, all participants are in a listen-only mode. A brief question-and-answer session will follow the formal presentation. If anyone should require operator assistance during the conference, please press star zero on your telephone keypad. As a reminder, this conference is being recorded. It is now my pleasure to introduce your host, Rich Chen, Chief Medical Officer and EVP of R&D. Thank you, Mr. Chen.
Good morning, everyone, and welcome to our webinar today. I'm Richard Chen, Chief Medical Officer and EVP of R&D at Personalis. Also here is our CEO, Chris Hall, and our CFO, Aaron Tachibana. We are absolutely thrilled to be joined today by Dr. Charlie Swanton, Cancer Research UK's Chief Clinician and professor at the Francis Crick Institute, world-renowned for his pioneering work in cancer genomics and cancer evolution. Dr. Swanton has graciously agreed to walk us through new, exciting results from their groundbreaking TRACERx lung cancer study, where they used our NeXT Personalis test for cancer detection. And this data was just presented a few days ago at the ESMO conference in Madrid. Before turning it over to Charlie, I just wanted to take a second to describe the NeXT Personalis test itself.
With NeXT Personalis, we wanted to develop a test that could begin to address some of the critical, fundamental questions that patients and their physicians face during their cancer journey. For example, after surgery, is the cancer still present? Or during adjuvant or immunotherapy, is the treatment working? What's the risk for the cancer coming back? And as time goes on, is the patient still cancer-free? Answering these questions is absolutely critical to managing and optimal care, but are difficult to answer confidently with the current tools like imaging or other blood tests because of their limited sensitivity. So we saw the opportunity to do better, and we designed NeXT Personalis to detect extremely small traces of circulating tumor DNA from the blood with a blood test, to deliver 10-100 times more sensitivity than other approaches, down to a one part per million level.
Now, to do this, we've developed NeXT Personalis that combines whole genome sequencing and our NeXT Sense technology to identify a unique and more comprehensive genetic signature derived from a patient's own tumor. This personalized unique signature is then tracked in the patient's blood over time to find residual or recurrent cancer with unparalleled sensitivity and with the goal of providing physicians and patients a tool to enable significantly earlier detection of cancer recurrence and a better tool for treatment monitoring. So today, with Charlie Swanton, we'll be seeing how this ultra-sensitive test can help early-stage lung cancer patients. So without further ado, I'll turn it over to Dr. Charlie Swanton.
Thank you very much, Richard. So as you heard from Richard, key to these assays fundamentally is optimizing limits of detection and sensitivity to detect the smallest amount of circulating tumor DNA in the blood while maintaining specificity and minimizing false positives. And I anticipate the data we're going to show you today will demonstrate that we've achieved a good detection and a sensitivity that surpasses our experience in the field. Just to give you a little bit of background on how we got into this field, about seven years ago, we first proposed that these personalized MRD tests would enable us to optimize sensitivity for minimal residual disease ctDNA detection.
That is, these are tumor-informed detection assays, which allow us to define the presence or absence in a patient's blood after surgical resection of tumor DNA specific to the patient's own tumor. So these are tumor-informed patient MRD panels, and we published in 2017 and 2023 in Nature first off for Abbosh on our experience for two different organizations in that area. First of those publications, Abbosh 2017 and Abbosh 2023. And I'll leave it to you on the call to look into the data set yourself at your leisure. So, I'm gonna present today from James Black in collaboration with the senior analyst, I think, Richard's group, to work with them because they have the next-
Hi, Charlie, I'm gonna interrupt for a second here. I'm having a little bit of trouble hearing you, and I don't know if that's just my connection or not, but there's a possibility of-
Do anything about that.
Okay.
I don't know what I can do about that.
Okay.
Can you hear me now?
Yeah.
Okay.
Yeah.
Okay. Okay. So,
... Ladies and gentlemen, the management line has been disconnected. Please hold while we get them reconnected.
Hello?
Ladies and gentlemen, the speaker line has been reconnected. Please go ahead.
Hi. Sorry, Richard. I was-
Welcome back, Charlie. Yeah. No problem. And I was thinking, since the connection was a little bit choppy, it wouldn't be bad to just sort of start from the beginning again, if that's okay with you.
Okay. So, sorry about that, everybody. I got disconnected, and then they kept me on the call for too long, so hence the delay. So, my name is Charlie Swanton. As you heard from Richard, I'm a group leader at the Crick and Thoracic Oncologist at University College London Hospital. We first entered this field in about 2015. We developed this idea of a patient tumor-informed minimal residual disease assay, and we developed our experience in collaboration with two commercial organizations published in Nature 2017 and Nature 2023, first of all, the Chris Abbosh. I won't refer to those companies, I'll simply refer to those papers, Abbosh et al. 2017, Abbosh et al. 2023. Key to this, as you heard from Richard, is optimizing limits of detection.
We want to be able to detect the smallest quantity of circulating tumor DNA in the blood so that we can detect residual disease at the lowest tumor burden to maximize our chance of cure by escalating adjuvant therapies. We were excited to work with Personalis for the simple reason, they were taking the assays that we developed with those two prior commercial organizations to the next level, with an order of magnitude greater number of single nucleotide variants tracked, over space and time, in blood from patients with non-small cell lung cancer. So, with TRACERx, which I'm sure you're familiar with now, we have been working with Personalis with a plan to analyze up to 400 blood specimens, but more so I should say, from 400 patients, several thousand blood specimens, in the adjuvant disease course.
Data I'm presenting today will be approximately 170 of those patients that were presented at ESMO the day before yesterday by James Black. This is his presentation that you can see on the screen. He has no conflicts of interest. Just to be clear, I am not a paid consultant with Personalis. The only conflict of interest I have with Personalis is that we have an academic collaboration, and they have paid for the sequencing and analysis with us of these sub plasma samples. We have funded TRACERx ourselves. This is a collaboration with an academic grant from Personalis, but I do not sit on their scientific advisory board. I don't have equity in the company, and I'm not a paid consultant.
I don't think I have conflicts of interest relevant to this talk, other than the fact that I'm keen to see these data published. What do we know about circulating tumor DNA? Well, it's a minuscule proportion of the total cell-free DNA in the blood, and particularly in the minimal residual disease setting, where we're trying to track a disease, very low disease burden. This is, you know, the key concept is to try to get down to variant allele frequencies that are very low indeed. Perhaps 0.0001% or less. Now, we know minimal residual disease can predict relapse prior to imaging from work up from us and others. And we know ctDNA detection requirements, as I said earlier, require high sensitivity, high specificity, and low input amounts of DNA.
They have to be applicable across a representative sample of tumors. Now, if you look at this cartoon here in the top right-hand corner of the slide, you can see how the variant allele frequency, the mean allele frequency of a clonal mutation, will change dependent on the size and stage of the tumor. So a T1b tumor measuring perhaps a total tumor volume of 1 centimeter cubed, we will have to reach that of variant allele frequencies for clonal mutations of about 0.0008%. And the truth of the matter is that current assays, even tumor-informed assays, struggle to get to a level as that. So, what is TRACERx, and what is our sampling strategy?
Well, patients with stage 1A to 3B non-small cell lung cancer are referred to us through 13 hospital sites, that you can see here in the U.K. on the left-hand side of the cartoon. Patients have their disease surgically resected, plus or minus biopsies of analysis of their mediastinal lymph nodes. If they're positive, the tumors are subject to multi-region sampling, followed by multi-region exome and RNA sequencing, tissue microarray analysis, with blood taken prior to surgery, immediately after surgery, and then somewhere between 4 and 6 weeks after surgery, for landmark analysis, to see whether or not there's evidence of ctDNA in the blood that could be used to stratify patients in the future for adjuvant therapies.
Now, in the adjuvant setting, we see patients in the clinic every three months, and, for the first two, and every six months thereafter, until five years to either cure or relapse. And if they develop metastatic disease, we biopsy their metastatic disease and ask, where did the metastasizing subclone come from? And how well does the circulating tumor DNA population reflect that metastasizing component? The two papers I was referring to in the top right-hand corner, you can see that these two panels, the commercial assays track 16 variants and 50 variants, respectively, in the 2017 and 2023 papers.
You can see the variant allele frequencies we got down to are about 0.01%-0.008% VAF, roughly, concordant with a T1b tumor that I mentioned earlier, measuring about 1 cm³ in total volume. Now, what we found in the recent 2023 paper is if you stratify patients to high, low, and negative ctDNA outputs in the blood, patients with high circulating tumor DNA have poor outcome that is independently prognostic in multivariate analyses. But also, patients with intermediate levels of ctDNA have an intermediate survival outcome between high and negative. This has been shown by others as well, including Max Diehn's group. Now, we turn to NeXT Personalis for the reason that they are tracking many more mutations.
You've seen those prior assays, we're talking 50 upwards of, you know, 18-50 mutations maximum. Here, we're on another level, at least 10 times more mutations are being tracked per patient. The LOD 95 of this assay is of three, roughly 3.676 parts per million, translating to about 0.00037% variant allele frequencies. The simulation suggests a limited detection, less than 10 parts per million, with 1 nanogram of DNA input, with an estimated specificity of over 99.9%. Specificity is absolutely crucial to maintain for a minimal residual disease assay because you don't want false positive calls. You can't sacrifice sensitivity for specificity.
So the cohort overview that we presented the day before yesterday at ESMO, shown here, these are early-stage lung cancer patients, as I mentioned. Surgery, they all have surgery with curative intent. 48% of patients received adjuvant treatment, and there was no neoadjuvant therapy in this cohort. And analysis of blood plasma was collected at preoperative time points, as well as post-operative time points, and the cohort had a median follow-up now of five years. And that's also important to make sure that we capture the majority or all of the relapses. So what were the results? Well, circulating tumor DNA was detected from 1.7 to 253,000 parts per million. Preoperative circulating tumor DNA was detected in 81% of lung adenocarcinoma and all of the non-adenocarcinoma patients.
You can see on the graph here, if you compare that to Abbosh 2023 and Abbosh 2017, it compares very favorably. So we're, we are now identifying, over four times the number of lung adenos in the preoperative stage, with this assay. Again, bear in mind, this is a tumor-informed assay, so we need the surgically resected sample. The reason we think this is so important is because we know, and I'll show you in the next graph. That preoperative ctDNA detection is a poor prognostic sign, and I showed you that in the two slides ago, but I'll show you new data from NeXT Personalis that corroborates and substantiates that association. Why is that important?
Well, if you take stage 1 tumors, for example, we currently do not give stage 1s, stage 1A, certainly, adjuvant chemotherapy. Now, if you look at the graph here, we're now detecting 52% of stage 1 non-small cell lung cancers, as opposed to 13% and 14%, in our previous experience in Abbosh et al. 2017 and 2023, respectively. So we're now detecting many more lung adenocarcinomas preoperatively than we have been with prior assays. So why is this important? Well, it's important because we can stratify outcome based on preoperative circulating tumor DNA, which gives us, as oncologists, a new opportunity to escalate treatment in these patients with stage 1 disease who we'd otherwise not give adjuvant chemo for. So that's shown here.
You can see that patients who have ctDNA positive have a very, very much worse outcome than patients who are intermediate or have no ctDNA detected, with very much worse relapse-free and overall survival outcomes if you have ctDNA levels that are deemed high in these assays. If you then look at the non-adenocarcinomas, these are mainly squamous carcinomas. What you see is if you can stratify patient outcome by the median ctDNA levels, divided by high and low. Patients with high preoperative ctDNA have a worse outcome, worse relapse-free survival than patients with low preoperative ctDNA. Now, what's interesting is that because this assay is much more sensitive, we have the ability now to ask a fundamental question.
In the patients we've previously deemed as negative by prior assays, are some of these patients now positive? Now, we've not been able to compare the identical samples because we haven't got any samples left for those patients previously trialed on the 2017 and 2023 papers. But what we can do now is confidently say, given the sensitivity of these assays, that there are a proportion of non-shedding adenocarcinomas, we think. And these non-ctDNA-shedding adenocarcinomas, they're not detected in green, have a phenomenally good outcome, in contrast to those where the ctDNA is detected that have a much worse outcome, as you can see here. So this gives us the opportunity now to offer patients escalated therapies in the adjuvant setting for disease, which we know has a pretty high chance of returning.
So if you look at this at five years, if you're ctDNA positive and you have an adenocarcinoma, you have about a 75% chance of relapsing by year five. That is the equivalent outcome to a stage IIIB tumor. And we can now basically stratify disease within stages using this assay for escalation of therapy to attempt to improve outcome in these, what we think are born to be bad tumors. So what are our future plans together with NeXT Personalis and Personalis? Well, we plan on analyzing a total of 450 patients. We've analyzed about 170 so far. Approximately 4,200 plasma samples are being analyzed as we speak. 350 tumor-specific subclonal mutations have been tracked in each patient, as well as the clonal ones.
There'll be a future expanded analysis that will focus on clinical performance, clonal evolution through treatments, and the use of NeXT Personalis to inform the acquisition of treatment resistance, and factors governing ctDNA shedding. On the right-hand side, you can see the positive predictive values and sensitivity and specificity for these, for this assay at a landmark time point, that is, about four to six weeks after surgery. If you're positive at this stage, you have about a 90% chance of developing disease recurrence within the next, two to three years. High negative predictive value, high sensitivity that compares favorably with competitors, and maintained specificity. So what about moving beyond the landmark time point and the preoperative ctDNA positivity?
Again, bear in mind, a landmark time point is what we refer to as a point after surgery, roughly four weeks, four to six weeks after surgery, where we look with the ctDNA assays to see if they're positive or not, indicative of whether there's minimal residual disease in the patient's blood. Well, we have been using the landmark performance now to look at outcome and the number of days of median lead time to recurrence, which with NeXT Personalis is estimated at about 331 days, as opposed to our recent experience in AVOS 2023 of 228 days. So this gives us really impressive lead time over imaging detection in landmark-positive patients with an 11-month median lead time over imaging.
I would say, though, it's important to take all of these data around lead time from any provider with a grain of salt, for the simple reason that it really depends how often one scans patients. If we were to scan patients more often, you would obviously reduce the lead time. Now, we scan patients four times in the first year, four times in the second year. It's unusual to do it any more than that, but I would just caveat direct lead time comparisons as other companies will want you to do, because a lot of it depends on how often patients were scanned.
If you only scan a patient once every two years, your lead time is going to be phenomenal for obvious reasons. I think TRACERx has an exceptionally good follow-up program in the context of CT imaging. So I am quite confident with these figures. What about the longitudinal performance? What does that mean when if you're negative at a landmark time point, what happens if you're positive at a subsequent time point up to the following five years? Well, here in these patients, about 173 median days to recurrence, which compares favorably with Abbosh 2023 and Abbosh 2017 of 119 and 70 days, respectively. There's a strong PPV of 94%, negative predictive value of 89%, and a sensitivity 85%, with a specificity maintained at 96%.
Again, it's important to focus as well on the follow-up times in the cohort, but cohorts with short follow-up, follow-up times are going to have very limited relapses. Our cohorts have a median follow-up time of five years. So in summary, I presented results today, which we're excited about from NeXT Personalis and Personalis led by Richard and colleagues, of an ultrasensitive ctDNA detection tool that can identify pre-operative ctDNA in 81% of patients with lung adenocarcinoma, including 52% of stage ones. High ctDNA levels predict poor OS and relapse-free survival in lung adenos and relapse-free survival in non, non-adenos. We think that this added sensitivity is clinically meaningful, that this ultrasensitive detection below 80 parts per million is important and stratifies ctDNA-positive lung adenocarcinomas with worse outcome and relapse-free survival compared to those that are non-shedders.
A full study will explore the impact of added technical sensitivity on performance and our preliminary data, albeit very extensive, I would argue, 170 patients, which compares very favorably with our two prior publications, highlight the promise of an ultrasensitive approach of post-operative disease detection. So, perhaps I could hand it over to the floor now and back to Richard, for questions.
Thank you very much, Dr. Swanton. We are thrilled to see the initial results from the TRACERx study that you just presented, and, you know, I think it shows that our strong sensitivity is for early-stage lung cancer, especially in adenocarcinoma, which is one of the most common subtypes of lung cancer, but it's one that's been difficult for previous, you know, other technologies. So we're very, very excited about that. And as you mentioned, you know, we saw the dramatic sensitivity improvements very clearly in the baseline presurgical data, with you know, 2-4 times more sensitivity than other technologies.
But probably even more importantly is that the clinical impact you spoke of, you know, of that higher sensitivity, the ability to stratify patients, understand which ones are at high risk for recurrence at the early stage, and then raising the potential for escalating treatments for those patients. Then the results also showed that we can detect the cancer earlier. I think all your caveats about the lead time analysis completely agree with. It's very hard to compare it between studies. With that said, with these patients in the TRACERx cohort having been kind of sampled with the same protocols or similar protocols, it's helpful to see that we have a median lead time of approximately 6 days-11 days, I mean, months, for ctDNA detection.
That's at, well ahead of traditional imaging and then significantly longer than previously reported in TRACERx. So and I think that raises the ability to potentially identify recurrence months earlier for patients and the possibility to intervene earlier as well. So I think, you know, overall, we see that the TRACERx results raise the potential for using NeXT Personalis to help inform lung cancer management, especially the early stage, throughout their patient journey, starting from pre-surgery to post-surgery, adjuvant treatment monitoring, and maybe even longer-term recurrence monitoring and detecting cancer recurrence earlier. And we are very excited to continue to work with Dr. Swanton and his team. This is just the start, as he mentioned, and more data to come. So before opening Q&A, I would love to just thank Dr.
Swanton for being here today and all the amazing work he's done. Dr. James Black, who presented this work just a few days ago at ESMO and did a terrific job, the broader TRACERx team, and then the patients themselves who participated, agreed to participate in the TRACERx study, without which, you know, this wouldn't have been possible. So, I'm going to stop here and open up for Q&A. And if I could ask the participants to just focus the Q&A questions on the clinical data presented, that would be very much appreciated. And, with that, I'll open it up.
Thank you. We would be conducting a question-and-answer session. If you would like to ask a question, please press star one on your telephone keypad. A confirmation tone will indicate your line is in the question queue. You may press star two if you would like to remove your questions from the queue. For participants using speaker equipment, it may be necessary to pick up your handset before pressing the star keys. One moment please, while we poll for questions. The first question comes from the line of Tejas Savant with Morgan Stanley. Please go ahead.
Hello, this is Yuko Oku on the call for Tejas. Doctor, I was wondering if you could elaborate on the significance of three-month lead time prior TRACERx studies in terms of clinical outcomes?
Sorry, can you say... I couldn't hear that. Just say that again.
Would you kindly elaborate on the significance of three-month lead time versus prior TRACERx studies in terms of clinical outcomes?
I didn't say that. I just said that we see our patients in clinic every three months, and I don't want to - I want to talk about the NeXT Personalis data. I don't want to talk about any other data, if that's okay. And it's this call is about Personalis, not our prior publications. We -
No, no, I meant in terms of the, the lead time versus imaging. You-
Oh.
You said that, compared to the prior studies of TRACERx studies, you saw an improvement in detection earlier, and I was wondering, that, that difference between the, the lead time you saw, in this study and in your prior studies, what that means for clinical outcomes?
Well, we don't know is the answer.
Okay.
Because the trials haven't been done yet.
Okay.
The reason why the trials haven't been done yet is because this technology is only, only recently evolving. So, you know, the industry is looking for the sensitive and specific panels that allow us to detect disease relapse as early as possible. And I believe we're getting to the point now where these trials can be done, and that question can be answered. My belief is the reason why adjuvant therapy works and cures patients is because we target disease when it's at its lowest, lowest overall burden. Now, as you're aware, with adjuvant therapy, one of the problems is that we cure many patients with surgery alone, and they don't need adjuvant chemo. But we don't right now have established tools that tell us who will and will not need adjuvant chemotherapy.
I think these MR tools, like the MRD tools I presented today, will get us to a stage where we are going to be able to say, in the near future, this patient has residual disease in their blood. They need adjuvant therapy, and they may need escalation of adjuvant therapy beyond standard of care, i.e., you know, addition of potential clinical trials to give them the best chance of eradicating any residual disease.
Got it. Thank you. That was helpful color.
Thank you.
And then outside of early-stage lung cancers, what other cancer types do you believe the test will prove to be of greatest benefit to patients?
Well, I mean, I can't really speak for other cancer types in terms of the performance of this assay because we've simply not looked at other cancer types. I can tell you broadly that any solid tumor type with an adjuvant strategy could benefit from MRD assays. Now, obviously, it will be up to Personalis to demonstrate to you that the performance is equivalent in other tumor types. You know, if you know, first principles, given that Personalis use whole genome sequencing to define patient-specific panels, that you're going to have enough mutations in any tumor genome to be able to track those over time, I would not expect the performance to be any different. But, you know, I may be wrong. That would be my academic guess here.
So let's say that's the case, then what tumor types would be looking at? Well, I think, you know, tumor types, as I mentioned, with an adjuvant paradigm: breast, gastric, colon, ovary. I mean, you know, you name it. Almost all tumor types have an established, adjuvant chemotherapy, immunotherapy paradigm. You know, and I guess no doubt you and others have worked out what the market size is for something like that.
Got it. Thank you very much.
Thank you. Next question comes from the line of Dan Freeman with TD Cowen. Please go ahead.
Great. Thanks for, thanks, Charles Swanton. Maybe given the data, and, you know, that you just kind of presented today or just kind of reviewed with us today, what, what additional data would you need and kind of the broader clinical field need in order to start using NeXT in clinical practice?
Well, I mean, that's a good question. I actually don't know the answer to that. Because the way in which the U.S. conduct clinical practice is rather different from the way we do in the U.K. and within the European Union. I mean, if you look at an assay like Natera or Invitae, I guess you could probably answer that question yourself. What studies have those companies conducted that has enabled them to, I guess, roll out their assay into practice? What I can say is that what I presented today and what James Black presented two days ago at ESMO, in terms of cohort size, compares very favorably with our early forays into this field.
Whether or not that's enough for breakthrough designated biomarker approval through the FDA, I don't know, and that will have to be a conversation you and Personalis will have to have. Sorry, it's not an area I'm an expert in, I'm afraid.
Got it. I think one of the slides showed 96% specificity. Can you just comment on that? Obviously, the sensitivity looks to be a huge, you know, notable advantage here, but just wondering if that 96% specificity is appropriate for this patient population, or I know in some of the competitive products and other, you know, in other tumor types, the specificity has been dialed up to a higher level.
Richard, do you want to comment on that?
Yeah, no, it actually, you know, specificity is critically important, and we know, comparatively in this tumor type, that is, you know, quite exceptional, and very comparable, if not much significantly better than what we've seen with other assays. So we're very happy with that level of specificity.
Terrific. And then maybe just one more for Dr. Swanton. Is, are there any reasons why kind of, you know, Personalis's approach with, you know, significantly more comprehensive, number of markers of, you know, that they're looking to identifying kind of with their technology, being able to drive the sensitivity up? Like, how do you envision, you know, you talked about lung and these other adjuvant solid tumors. Is kind of more markers and higher sensitivity always better? Is it... Obviously, there's a commercial trade-off possibly to the cost, but just how would you, from a clinical perspective, discuss, you know, the performance of kind of ultra, you know, these higher sensitivities assays and kind of, you know, what impact they can have on the field?
Yeah, it's a good question. Obviously, we came at this from the perspective that the more SNVs you track, the more likely you are to detect signals in a small blood volume. It's something that only clicked with me a couple of years ago when I was talking to Abbosh, Chris Abbosh, and we were sort of calculating what the chances of there being a patient with a KRAS G12C mutation, what the chances there would be of KRAS mutant allele being present in DNA in a 20 ml blood tube, dependent on how big the original tumor was. And what you realize very rapidly is that when you're dealing with a 1 cm³ tumor burden, the chance of that KRAS mutation being in the blood tube are extremely low.
So the only way you can reliably detect the presence of a digital disease using tumor-informed assays is to target more mutations. Then you get into the realms of error correction and making sure that the mutations you're detecting really are true mutations and not false ones. And that, I believe, is something that Richard will have to speak to because they have proprietary tools that enable them to optimize the limits of detection while maximizing specificity, minimizing false positives, which I'm not privy to, for obvious reasons, because they're commercially sensitive. But ultimately, the fact is more is better. What we don't know is how many, is there an optimal number?
And that's a question that I'd be interested to work with Personalis on over the next, you know, few months, is just to try to map that out. Is there an upper limit beyond which actually there's a trade-off between, you know, going from more mutations, because you're interested in understanding current evolution, which we are, versus maximizing sensitivity and minimizing false positives. For the work we're doing, we're very keen to map more mutations. We're not just interested in MRD, we're interested in how these tumors evolve over time. And we can only do that by tracking more mutations.
Yeah. Thank you.
Just to chime in to what Charlie went through. In terms of the, you know, going after more targets, having more shots on goal, so to speak, makes a big difference in the sensitivity, but you have to pair that with, you know, an ultra-specific approach, and that's what the NeXT Sense platform that we've developed does. And that's been a lot of hard work with algorithms that can optimize and reduce the noise in the system from panel design all the way through the panel analysis. And so that is part of sort of how we are able to achieve this ultra-sensitive level while still preserving really high specificity.
Great. Thank you.
Thank you. Next question comes from the line of Mark Massaro with BTIG. Please go ahead.
Hey, great. Thank you so much for, an excellent presentation, and Dr. Swanton, thank you for all your work in this field. So you know, the first generation assay in 2017 had, I believe, 100 parts per million. The second one in 2023 had 80 parts per million. This one has, it looks like 3.76 parts per million, with the ability to go down to 1 part per million. I guess, Dr. Swanton, from your perspective, to what extent does parts per million lead to, you know, developing a more optimal assay relative to prior?
Yeah. So thank you for your kind words and comments. I think, you know, the way I see this is that, enabling us to detect the very lowest amounts of ctDNA reliably in a blood tube, which comes ultimately down to maximizing or minimizing the number of parts per million one could detect a ctDNA fragment, will allow us to identify those patients who are relapsing at the earliest time point to target with adjuvant chemotherapy, targeted therapies, or immunotherapies. Residual disease when it's at its lowest burden, that is, at the lowest number of cells, and therefore, the least opportunity to have de novo resistance mechanisms present in that tumor.
We know, for example, that immunotherapy works much more effectively in patients with lower disease burden, and we know those that do badly, often those patients with high LDH, you know, high disease burden. And ultimately, getting down to those very low limits of detection, I think are gonna be key to improving outcomes.
Excellent. The data today came from 170 patients.
Yeah.
It looks like about 400 will read out eventually and maybe more than that. Can you give us a sense for when we can expect the next larger readout?
Yeah.
Do you think-
Good question.
Yeah, do you think today's readout is significant enough to make a judgment to compare the assays, or do we, do you think we need to see more data readout?
I mean, if you forgive me, I'd rather not get into discussions about comparing assays. I think I'd leave that to you and the data in the public domain. I believe that the data we presented today are the next step in precision medicine, personalized panel designs. I'm very excited to see these results. I'd be even more excited to see assays like this being implemented in clinical trials, prospectively, to test the hypotheses you and I and, and the people on the panel have been asking questions about. You know, does it matter, optimizing limits of detection? Do we need to be down to the 1 or 2 parts per million detection limits? Does that— Does the ability to detect DNA at its lowest fraction enable us to intervene to improve patient outcomes?
That, those are the questions we really need to understand and answer over the next 5-10 years, and I strongly believe that this assay will be playing a major role in that journey for patients. I'm sorry, I can't really answer any more clearly than that.
Yeah. And just maybe that, that's terrific. And Mark, to your question, I mean, just something to consider is that even though we're aiming for 400+ patients, ultimately, 170+ patients is already one of the largest cohorts that have been read out on with this level of quality, you know, five-year follow-up, you know, where you can really reliably, you know, determine some of these metrics. And so I would definitely encourage you to look at sort of what's been published, but you'll see that this compares already very favorably. Now, when we have 400+, it'll be even, you know, more expansive, but, you know, I think it's already very, very strong.
I think-
Excellent. Yeah, go ahead.
I will fill in one more point here. It's if I could just say why we are going for 450 patients. As the sort of chief investigator of TRACERx, I have been really very keen that we don't analyze TRACERx piecemeal, you know, multiple different assays with small numbers of patients. When I originally started working with Personalis, we had an agreement that we would be very ambitious and apply their assay to half the cohort. Now, that was an academic consideration, not a commercial one. That 450 number wasn't selected because of commercial considerations.
It was an academic collaboration with Personalis that number.
Okay. If I can sneak one last one in. This is probably a minutiae question about the data. What's the difference between the estimated specificity of 99.98%, that was presented in an earlier slide versus the 96% specificity that was shown in a later slide?
Yeah, good question. Richard, do you want to answer that?
Yeah, absolutely. So the 99.98% specificity is an analytical specificity. So that's when... That's determined when, it's sort of like the analytical sensitivity, where you have sort of controlled samples that you're evaluating, where you kind of know ground truth, and are able to determine, you know, kind of the measurement specificity and the measurement sensitivity. The 96% is actually a clinical specificity, based on, you know, a cohort of patients and where ground truth is based on whether they relapsed or not. And so for specificity, in this case, it would be determined by saying, for all the patients that didn't recur, you know, was your test negative at a certain point in time?
Which is different from looking at a dilution series, where you're actually saying—or a set of samples where, you know, you say there's no tumor DNA, and then run the test, and indeed, did you come up with a negative answer? And that's what we did for the analytical specificity, and the clinical specificity is determined differently. So I know that's a little bit of a detailed answer. Happy to follow up more on it, but that's kind of roughly what it goes, how it goes.
That's perfect. Thanks so much, guys.
Thank you. Next question comes from the line of Joseph Conway with Needham and Company. Please go ahead.
Hi, thank you for taking our questions and appreciate the time, Doctor. Maybe just looking at the testing schedule that would be implemented for NeXT Personalis, or I guess let's just say broader with MRD testing for lung cancer patients. I believe in your study, you said patients were followed up with every three months for the first two years, then six months for the following three years. Is that you know a viable scenario or is something you would see in clinical practice, or could you maybe outline some scenarios that would maybe change those pathways, maybe depending on stage or you know lung cancer type, any other clinical characteristics?
Yeah, I mean, I think the sampling strategy we use is very routine in adjuvant follow-up practice across the world. There's nothing remarkable about it. We selected that adjuvant follow-up because we wanted TRACERx to be representative of a real-world situation, which we think it is. So I think I feel quite strongly that this is something that's absolutely practical. Three monthly clinic visits are what most people in Europe do routinely, and I'm pretty certain that's what happens in the U.S. as well. Some people may see their patients more frequently, but I think what we've gathered here today is a sort of a real-world data set of what represents standard clinical follow-up, which I think is important for, you know, your the considerations you guys will be making.
Okay, great. Yeah, that's, that's super helpful. Thanks. I think that's going to be all the questions from us. Thank you for your time, Doctor.
Pleasure. Thank you.
Thank you. There are no further questions at this time. I would like to turn the floor back to Rich Chen for closing comments.
Well, again, I just a huge thank you to Dr. Charlie Swanton for being here today and just, you know, we're very excited to take this further with the TRACERx team, and thank you, everyone, for participating.
Thank you.
Thank you. This concludes today's teleconference. You may disconnect your lines at this time. Thank you for your participation.