Sounds great. Good morning, everyone. Thank you for joining us. I'm Robbie Bamberger, Senior Research Analyst covering life sciences and diagnostic companies at Baird. Very pleased to have Derek Maetzold, Castle's Founder, President, and CEO, and Frank Stokes, CFO. For anyone less familiar with the story, Castle's a molecular diagnostics company, builds, commercializes gene expression and tissue-based tests that change how physicians manage a patient across dermatology, gastroenterology, and ophthalmology. Before we get into specific segments of the company, would you mind just giving us a quick snapshot of the company coming off of Q2 earnings? What went right in the quarter? Anything you'd like to point to investors about the quarter specifically?
Maybe I'll talk about the business top line, and you can cover—
Sure.
—specifics on financials. Castle Biosciences is a molecular diagnostics company. We have a number of tests on the marketplace that we've demonstrated not only are impacting decisions physicians make, but also improving outcomes. Either lives are extended, or we're doing less uncertain things that'll make a difference. We focus on two main areas, therapeutically or customer-wise, I would say. We have a dermatology sort of commercial and R&D vertical. We have a number of tests in there for our DecisionDx-Melanoma test, our AdvanceAD-Tx test for directing therapy for atopic dermatitis, our squamous cell carcinoma test. On the gastroenterology side, which is our other vertical, we have one test currently on the market there, which is TissueCypher, which is used in Barrett's esophagus disease to help direct care in that patient population.
I think we had certainly a good quarter. I think our two main revenue drivers, our melanoma test, our TissueCypher test, we saw good progress there. I think on TissueCypher, I would encourage sort of looking at a four-quarter kind of rolling growth trend. We're still fairly early in that launch, and that volume number is becoming more predictable but still is not quite as easy to project with perfect precision just given where we are in the launch. W e had good, strong, adjusted gross margins, among the best in the sector, and that gives us lots of room down the P&L to invest in R&D and other activities for the company.
Has anything changed on your view of 2026 since reporting Q2? Anything in Q3 or Q4 that you're thinking?
No.
No.
I think macro environment is largely the same. I think that it's important for folks to remember that inflation is still fairly durable. When you look at operating expenses and things like that, it's important to recall that 4.5% up is flat. Even when inflation eases, unfortunately, prices don't go down, they just go up—
Yep.
—less quickly. I nflation is still durable. I think maybe early in the year we were thinking or hoping we would see a little bit of ease there, but certainly still intact.
Yep. Maybe double-clicking on TissueCypher specifically, the growth has been really good there. 9,000 tests last Q2 to 15,000 tests this quarter, so 63% growth. When you decompose Q2 volume growth, how much is essentially new ordering from clinicians, deeper utilization inside of those accounts, and overall just endoscopy procedure growth?
I would probably break that down into a couple slightly different approaches.
Yeah.
I think the growth in endoscopic procedures is a good macro level. We don't necessarily track that on a daily basis. That doesn't affect our business that much, you would say. I n terms of the number of patients diagnosed with Barrett's esophagus each year, we think the addressable market for us is around 415,000 patients who have either non-dysplastic disease, indefinite disease, or low-grade dysplasia, which are all in the target universe of TissueCypher. A nice, big, consistent population. Post-COVID, we've certainly seen some growth in overall annual endoscopic procedures, but I think that's a nice trend to have. In terms of TissueCypher growth, based upon the 2Q performance, we, I guess you would say, increased our overall volume expectations to somewhere between 50%-52% year-over-year growth in volume, which is an excellent—
Yep.
—growth product for us and just shows you what the demand is for this test, which really is helping to direct what is this patient with Barrett's esophagus disease, what's the right pathway for from a treatment standpoint.
Yeah.
In terms of penetration within sort of marketplaces, we're still early in the launch curve. We are still capturing a large number of new first-time ordering prescribing clinicians or gastroenterologists. I wouldn't call those late adopters. I think it's middle-of-the-road people right now. The second tier up is sort of current customers. Can we increase penetration on current customers? Our sense right now is that if a gastroenterologist is going to make the decision that he wants to or she wants to incorporate TissueCypher test results in the care of a patient with Barrett's esophagus disease, it's awfully hard to go ahead and slice and dice that patient population. They sort of all have the same pathology. Some might have different risk factors than others, b ut the risk factor projections on progression likelihood are so soft that it's almost a all-or-none .
The trick becomes, though, if they want to adopt TissueCypher, they begin using it, just making sure that the workflow in their office so that when a patient comes into the clinic with a diagnosis of Barrett's esophagus disease from their last endoscopy a few days ago, a couple of weeks ago, that the nursing staff, the management staff, the MA around that gastroenterology when they're seeing that patient is aware of the fact that this clinician would like to have TissueCypher ordered as part of the patient care. That sort of is the second building block, is to make sure that they have the systems in place in the practice so that when they want to use the test, they recall and they get to order it.
Above that level, there is the opportunity to really look at what we're seeing on a small scale right now, maybe move into a larger scale, which is to say, if you have enough of a mass of clinicians, call it 20 gastroenterologists in a given location, five or six have already dabbled in or adopted TissueCypher for their patients. Why should the patients who happen to get cycled to gastroenterologist number 14 versus gastroenterologist number two?
To be deprived, in my words, I would say, not have access to TissueCypher test. We are seeing, on a small scale right now, practices beginning to say, "Hey, we have enough experience the last couple of years. Why are we making this standard of care as we approach how we manage Barrett's disease for the practice as a whole?" As we see more of those opportunities turn from kind of discussion and concept to saying, "This is a pathway we're going to implement," we should see that as sort of the third level of growth overall.
Is there a way to size up the 5,500 clinicians that have ordered a TissueCypher test? How many are fully activated or fully penetrated? Are any of them fully activated, o r how would you size that?
I don't have that data to give you today. I would say that if they adopt the TissueCypher, then they usually adopt it across all their patients. The gap in what they want to adopt versus how they execute is the issue of the nursing staff turnover, the MA is not being there at the right time.
That's helpful. You gave the framework of 50%-52%, as you said. The first half grew 61% for TissueCypher volume. When I bridge the two together, it implies back half a little bit of deceleration. Is there prudent conservatism in that? Are there reasons why growth would moderate other than maybe comps in the second half?
I think just one is a comps issue, with the back half of the year was strong last year. Also, it does look like in first and second quarter, maybe there were some procedures pushed into second quarter that might've naturally been first quarter. The best we can do is we can suppose why that is. I think that, like other areas of healthcare, when patients' deductibles reset, that changes their behavior. They tend to put off care, put off physician visits. I think if you, again, sort of take a rolling four-quarter kind of approach, you can see what the inherent fundamental growth trend is.
Yep. TissueCypher saw a nice ASP step-up in Q2. I guess part of that came from how you're accruing your revenue, but can you help just decipher the durability there? Just how much of that step-up reflects the better cash collection from you guys versus the accrual changes?
Well, they should match.
Yeah.
The accrual and the cash collection should match. At this point, we really launched that in earnest Q2 of 2022. The collection and appeals cycle, we could spend the hour talking about lamenting that. T he payer behavior and appeals cycle takes a long time. By necessity, it's just a little bit further along before we can accurately and more confidently predict what we're going to collect. I think where the consensus is a good ASP for TissueCypher right now. We don't model in either our guide or our working with other parties. We don't really model continued improvement. If we do see some improvement on the commercial side, we'll take it as it comes, but it's very difficult to project and to get right time-wise and scope-wise.
Yep. Just for TissueCypher volumes, is there any lab capacity or turnaround time constraints for you guys just in terms of you scaling towards a 60,000 test level?
No. We've scaled our Pittsburgh laboratory to stay six months to a year ahead of demand. I think if we see an acceleration, we're well covered to go ahead and handle that.
Okay. Helpful. Maybe just thinking about the overall TAM, 415,000 patients a year who get an upper endoscopy, and they meet that intended criteria. You mentioned last call you're at about 10%-12% patient penetration, maybe mid-teens by the end of the year. What gets you to maybe 30% + over time?
I think that we are reasonably positioned in guidelines today. I think that will improve over time, but I don't think gastroenterologists necessarily appear to be guideline-driven behaviorists, at least when it comes to the use of this kind of testing service.
Yep.
That'll be helpful, but not necessarily a driver. In the case of Barrett's esophagus management, you already have tools that can essentially stop or cure Barrett's esophagus disease. If you have high-grade dysplasia, and many people with low-grade dysplasia, which is sort of a high grade, is just before moving over the edge to having esophageal adenocarcinoma diagnosis. High-grade and low-grade in the U.S. today are largely intervened with, so ablation therapy, cryotherapy with Medtronic's Barrett's tool, or surgical procedure to go ahead and basically remove the Barrett's lesion. Gastroenterologists know that they can cure a patient of Barrett's disease and stop progression to esophageal cancer in its tracks if they intervene.
The balance has always been, in Barrett's disease, is to say, well, we intervene with the high-grade patients who are the most aggressive, likelihood to project, and they get cancer, and they saw adenocarcinoma. We now intervene with many people with low-grade dysplasia. That leaves sort of the 95% of the patients diagnosed largely with this non-dysplastic disease, which, as a population, has a lower risk of progression. They don't routinely ablate or intervene with those patients to go ahead and kind of cure that Barrett's lesion. That's where all the cancer comes from now because they aren't being treated. We can't treat them all. There's way too many, and that's way over-treatment.
The value of our test is say, hey, we can find those bad actors who are hiding out in this non-dysplastic pathology grade, bring them up, and say, if you intervene, this patient has a likelihood of no intervention progressing to cancer like they had low-grade or high-grade dysplasia. What a great opportunity to kind of stop a patient from progressing. That sort of natural, if I see the TissueCypher result, what am I going to do? I can actually stop cancer from progressing. On the other side of the equation, if they happen to have a non-dysplastic lesion, what we see on an epidemiology basis is that patients are coming in more frequently than guideline-recommended . The guidelines would say maybe every three years, plus or minus, depending on other features, historical clinical features, for example. P atients and gastroenterologists perhaps are coming back earlier.
Now maybe they're concerned about maybe I have the esophageal cancer, it just got missed. W e have seen now with clinical use data that actually gastroenterologists who get a low-risk TissueCypher result in those non-dysplastic patients are naturally winding up the come back to repeat surveillance endoscopy time, which is great care from a standpoint of patients, great use of healthcare resources. The people who need the intervention get it, and the people who don't are going much more closer to guidelines in terms of repeat surveillance.
Do you think there's a ceiling in the potential penetration of Barrett's esophagus above a certain percentage?
Yeah. We've seen in our rare cancer test for uveal melanoma, we think we have roughly 85%, maybe 90% of all patients diagnosed each year receive that test as routine care. I think if you look at the breast cancer tests overall, my recollection is they're somewhere in the 90%, 95% of women who have breast cancer are getting access to one of these tests available out there. I think that's an aggressive upside. I think two-thirds, or 60% upside penetration, is probably the right thing to model for most molecular diagnostics. Now that being said, you do have this intervention of ablation therapy or surgical procedure, which might remove that. M aybe that's too conservative, but that's probably the right way to model this, the 60%, 65%.
Yep. Y ou're at about 100 territories now in gastroenterology with the additions you made in Q2 that aren't at full productivity yet. H ow much embedded growth is with those reps that essentially aren't at 100% productivity that will get to that 100%?
I don't know if we can tease that out in terms of an inflection point. I would say that we did increase the size of our territories overall, or rather, decrease the size, added sales reps back in the—
Yeah.
—late fourth quarter, and again, second quarter this year. I would say entering in 2027, we should be fully trained, fully in territory, and fully effective. T he decisions we made last winter and this second quarter will bear fruit here in 2027 growth.
Yep. How long does it take just a new gastroenterologist rep to reach full productivity once they start on?
We model, which may not be the truth—
Yeah.
—about six months, five or six months between going through training, learning who their customers are, gaining access on a routine basis.
Would you mind just talking about the commercial coverage you have for TissueCypher today, and then are there large players you're still fighting for coverage with?
Yeah. Largely no policies, very few policies. We do have some plans that cover the test, but most of the plans don't have a policy for it. There's certainly adequate data to support the clinical utility of the test. The reality is, it really is not data that it takes. It really takes penetration and widespread utilization for plans not to be able to argue it's experimental or investigational.
That's helpful. Thinking about AI, how much of TissueCypher's moat is the algorithm itself versus the clinical outcomes that you have from it? Where else are you applying AI, whether that's in just the test development, lab operations, or the reimbursement workflow there?
Ask the first part of the TissueCypher question again.
Yeah, so I was essentially just asking, where are you applying AI within TissueCypher?
Within that, TissueCypher was designed using a spatial omics platform by actually our Head of R&D, Rebecca Critchley-Thorne. I want to say initially there were 100,000, 200,000 image points that were looked at back in the early part of last decade. That went down to, I think, 35,000 points. We 're left with a test today of which 15 points is what's important for TissueCypher.
Nine of those are proteins that we stain for, if I recall correctly, and seven of them are overlapping morphology goals. We apply the AI-driven algorithm to those factors that went from 35 down to this 15-feature set. H eavily involved from a development standpoint of that test from a spatial omics perspective. Now we do and have been applying AI to our other tests as well over time. I won't say nothing new, but you have easier and quicker tools today, which is good for future improvement and development. As it relates to the rest of the company, we have undertaken artificial intelligence initiatives on, I wouldn't say, a measured basis, but a thoughtful basis to say, where can we improve the performance of Castle as a company?
We have seen in areas of reimbursement, for example, where we can go ahead and take an AI-driven denial from an insurance company, recognize that it was AI-driven denial based on the word choices. We can go ahead and use our agent to go ahead and develop a better, more specific appeal letter with less cost and more speed. I think we'll be able to see over time, hopefully, that will improve the appropriate payment rate by payers in other areas of the company, similar to that kind of an agent benefit.
Yep. Great. Moving on to dermatology, DecisionDx-Melanoma, volume grew 3% year-over-year and sequentially, which was a little bit below the normal step-up, but you noted that was due to the derm commercial team adapting to carrying three products. Can you maybe just walk through what changed in the field? Maybe that would be the first question on that.
We went from roughly 50/50 promotional efforts between the DecisionDx-Melanoma test and squamous cell carcinoma to 100% beginning July 1st of 2025, so a year ago, a year and a couple of months ago. We had a, I would call it, a measured soft launch, a limited launch of our atopic dermatitis test in, I think, late November of 2025. Still kept the majority of the business focus, the bonus plan, on the DecisionDx-Melanoma test. We are still seeing good, strong background demand for squamous cell carcinoma. It happens to be when you walk into the same medical-oriented dermatologist, with very few exceptions, they' re treating skin cancer, both melanoma and squamous cell carcinoma, and they see an awful lot of atopic dermatitis patients.
Despite focusing, you would say, the commission plan geared towards melanoma, they are getting questions, and they are appropriately going ahead and having conversations about other two tests as well.
How long do you think it will take the sales force to fully absorb these new products?
We are holding back on the atopic dermatitis adoption, I guess, or openness. Still limited launch as we see how close our pre-launch model was to forecasting our average reimbursement rate or ASP over time. As we get more confident and comfortable with that rate being close to where we want to get to, then I think we have a decision to make about how do we open up the AdvanceAD-Tx test to more clinicians and more of the time, and hopefully have that build in to be what should be a $28 million, $29 million, $30 million major contributor from a revenue standpoint.
Any positive impacts for long-term DecisionDx-Melanoma penetration following the recent positive result from the phase III cancer vaccine trials?
We were thinking the same thing too.
Yep.
That's good we're on the same page. I think if we kind of roll back the wheels a couple of years, we had certainly been hopeful, been expecting either Merck or BMS to consider moving their immunotherapy PD-1 inhibitors into earlier-stage patients. The majority of people diagnosed with melanoma have either stage 1 or stage 2A disease, which is a little bit thicker, but still on the thin side of thicker. The minority of patients are diagnosed with stage 2B or 2C or have regional disease, which in melanoma we call stage 3. For a variety of reasons, some of which could have been probably the low overall risk of recurrence in the stage 1 and 2A patients, both companies chose to go ahead and stop developing adjuvant therapy indications at the stage 2B to 2C indication.
The vast majority of our testing, the vast majority of the good use of our tests, is really in the stage 1 and 2A patients, so that's disappointing. I think as the combination vaccine trials move forward, the hope I would have from a patient care standpoint is maybe they move up into stage 2A or maybe even to stage 1 patients, and the balance is going to be there, the risk-benefit to a patient. Because if you treat every patient with those therapies, the majority will not benefit because they have such a low natural risk of recurrence that they didn't necessarily need to have a therapy plopped on them. O ur tests can hopefully identify those patients who actually have a higher risk of recurrence or have a melanoma that looks thin but actually is acting like a very thick one, which you would have on-label therapy.
T hat, to me, is the expectation downstream is that patients diagnosed with thinner melanomas could benefit from having our test identify it actually is a very, very aggressive melanoma biologically, despite looking fairly nice under the microscope. This is a patient you should consider getting a vaccine or vaccine combination therapy.
Yep. You reiterated the mid to high-single-digit growth for DecisionDx-Melanoma volume for the full year. First half was up 9%. That implies low to mid single digits at the low end of your range in H2. W hat gives you confidence in re-acceleration from Q2, and what specific levers get you there in the back half?
Yeah, we continue to be confident in where we guided volumes on the test and I think that the levers there are just continued education, physician education. We publish regularly even at this stage in the test's life cycle. We still publish data supporting clinical utility, clinical actionability, and benefits to patient including cost benefits. We 'll continue to do that, and I think our team will continue to execute and drive conversion. By definition, the last physicians to convert are the hardest. The ones that convert first are easier. W e'll continue to support that and convert docs and then spread the use of the test through more of their patient base.
Great. Your AdvanceAD-Tx test is for atopic dermatitis. Instead of a biopsy, it's just a simple skin scraping. The advisory panel voted 21- 0 to recommend crosswalking at $3,675, with CMS's preliminary determination due soon. Can you maybe walk us through what happens between now and then a final rate with that?
Sure. The final rate would be effective on January 1st, 2027 is this current cycle. Historically, Medicare has published their, what should they call it, draft or preliminary 2027 or the next year rates in late September. I would expect in the next 13, 14, 15 days, we'll go ahead and see that rate schedule for 2027 published in a preliminary draft form for all tests that are being repriced this year. Medicare is not mandated to follow the recommendation of their advisory panel, but in many times they do. I guess our base expectation is that being crosswalked to that $3,675 test is appropriate, which it's almost the same technology, same skin indication. That test happens to be directing therapeutic response in psoriasis. Our test directs therapeutic response in atopic dermatitis. Both are seen by dermatologists, so the inputs look pretty darn similar.
We would expect them to do that. If they don't do that, then there's one of two things. They can either recommend a gap fill, which we go through contractor pricing for another year till 2028, o r they might crosswalk to a different test. I would agree with you, with a 21-0 vote, that seems awfully odd that you'd have the experts in the field be overridden. If it was 50/50, different issue, right? 2 1-0 sounds pretty unanimous to me.
Yep. What are the gates you still need to clear before a full commercial launch there?
For atopic dermatitis, really, well, having knowledge of the Clinical Laboratory Fee Schedule is one sort of more peg to help confirm our ASP model assumptions. I think as we work through the rest of this year to go ahead and see how our ASP growth in atopic dermatitis are matching our model, make any adjustments as necessary on the upside or the downside, and then I think we can release that product over time in a more full launch potential.
That's helpful. Just moving on to DecisionDx-SCC, Medicare coverage changed in April 2025. You 've reconsideration requests out. You know the potential for regaining coverage eventually. I guess, what are the leading indicators we should be watching out for in this process?
We did submit in the early July of 2025 reconsideration requests to both Novitas and to Palmetto GBA MolDX program. Both requests were considered valid, which means they looked at the file and said, "Yes, there's enough evidence here that we didn't review before, that we should be reconsidering or reopening up this LCD." There are no time limits from the time of that acceptance of it being a valid reconsideration submission to when they have to act. Now we have looked around, just look at other LCDs in various draft forms or reconsideration forms. There's not a whole lot of history there. I think there still is no reason why they either, one, could not get through a draft LCD being posted later on this year. There's also necessarily no reason why they would have to. It's a matter of just saying workload possibilities.
I think once the draft posts, assuming it's a positive change, then we're modeling roughly one year between posting date to beginning to regain coverage from Medicare for payment. I f it posts in late October, it'd be late October 2027 from a standpoint. From an internal modeling standpoint, we've taken squamous cell revenue out of our models until 2028, just because there's no reason to bake that into 2027. If it happens, it's fantastic upside. If it doesn't happen, there's no downside to the business process.
What new evidence has gone into that reconsideration file?
I think the most important two pieces of evidence is that the open comment period closed for both LCDs prior to having us publish our two back-to-back studies demonstrating that our test is able to predict responsiveness to adjuvant radiation therapy. Why is that important? 100% of the patients that we test, or that we allow testing for squamous cell carcinoma, are defined as having high-risk disease. They have one or more clinical-pathological factors suggesting a higher risk than a low-risk patient with no factors of recurring and progressing. Radiation therapy is the only sort of adjuvant therapy used today in these localized high-risk tumors. We do have a PD-1 inhibitor from Regeneron that's used in a select population, more of which have already metastasized beyond the primary tumor or unresectable. Adjuvant radiation therapy is the mainstay intervention.
What we were able to demonstrate in two back-to-back multicenter studies was that if you take the DecisionDx-SCC test, of which, by the way, we had inserted radiation response genes during development because of the adjuvant radiation potential, and you propensity-match that to the group of patients in our study set that did not receive radiation therapy, we found that we can find people who get a great response to radiation therapy. Meaning roughly a 50% reduction in likelihood of metastasis over three years, which is the standard follow-up time for squamous cell. Similarly, we found a lot of patients, a majority who actually we can't discern any clinical benefit from the use of adjuvant radiation therapy, which is important because that's a high side-effect intervention. I think significant side effects are greater than 90% in radiation therapy patients.
Many patients who have squamous cell carcinoma also will have concomitant or later basal cell carcinoma, which is also responsive to radiation therapy. If you use radiation in a squamous cell patient at one point in time in a certain part of the body, you can't re-radiate that same location again. You have an increased likelihood of cancer, plus if it's on the head and neck, you 've increased likelihood of brain radiation consequences. Being able to remove people from radiation therapy is pretty darn significant from a patient care perspective. Neither Medicare contractor was able to review either one of those studies, and we believe that those two studies are our foundation. There's other data as well, but those are the main two items.
Awesome. That brings us to the end of our presentation. Please join me in thanking Derek and Frank and the team from Castle for coming to our conference.