After 24 years working in the pharmaceutical and biotechnology industry, Derek Maetzold found himself with a valuable skill set and an important decision to make. With the company he was working for being restructured and sold, Maetzold was considering what might come next, and what kind of meaningful impact he could make on healthcare.
Drawn to the genomics revolution and the advances emerging from the Human Genome Project, Maetzold became interested in how we could move beyond traditional, structural assessments of cancer to a more biologically informed understanding of disease. He recognized that there were significant gaps in predicting how a patient’s unique cancer biology would influence disease progression and treatment response.
The decision was made. Castle Biosciences was created to fill that gap and offer more accurate, personalized guidance at critical decision points. The vision: to give patients and physicians better tools to choose the most appropriate path based on the molecular biology of the individual’s tumor.
Here, Maetzold, Castle Biosciences’ President, CEO, and Founder, shares his insights into the prognostics landscape and its impact on personalized oncology.
How has the prognostic landscape changed and evolved since Castle Biosciences was formed?
When we started out, much of the molecular diagnostics industry was focused on high-incidence cancers, like breast cancer. It’s very difficult for large labs to scale down to serve small patient populations, and we saw an opportunity to do exactly that – starting small and serving communities that were largely overlooked. That led us to concentrate on rare cancers during the company’s early years.
We partnered with academic institutions that had developed promising tests for these less common cancers. Our goal was to in-license, clinically validate, and commercialize those tests, then deliver them to the physicians who needed them.
We made some great choices – and, admittedly, a few that didn’t have the clinical impact we had hoped. In some cases, even when the test was accurate, oncologists hesitated to de-escalate care because there were no alternative treatment options. Telling a patient with cancer, “Let’s just watch and wait,” is a tough sell when there’s only one standard therapy on the table. But that was a risk we anticipated, asking ourselves: will this test meaningfully improve patient management and outcomes?
Sometimes that means helping a patient live longer through early intervention. Other times, it means sparing them unnecessary surgery or radiation – avoiding harm without compromising survival. Both outcomes matter.
After a few years, we began investing in our own internal clinical research programs. One of our first internally developed tests, which has had a measurable impact on patient care, was for melanoma. It’s reduced unnecessary surgeries and, based on data from large prospective studies, we now know it guides real clinical decisions. In fact, when comparing melanoma-specific survival between patients whose treatment was informed by our test versus those who weren’t tested, we saw a 17 percent relative improvement. That’s a powerful example of precision medicine in practice.
Because we focus on creating tools that physicians and patients find genuinely useful, we don’t suffer from “not-invented-here” syndrome. If someone else has developed a meaningful diagnostic solution that aligns with our mission, we’re open to bringing it into Castle.
A great example of that philosophy is our acquisition of a small company that developed a spatial omics platform. The test quantifies protein expression in specific areas of biopsy tissue from patients with Barrett’s esophagus – the only known precursor to esophageal adenocarcinoma.
In both the US and UK, most patients with Barrett’s esophagus and no dysplasia are advised to return for surveillance every three to five years. Only those with low- or high-grade dysplasia are encouraged to undergo intervention, typically involving radiofrequency ablation or localized surgery. But some patients with “low-risk” pathology still progress to cancer. The TissueCypher® test is designed to identify those patients with Barrett’s esophagus who are at risk of progression to esophageal adenocarcinoma and could benefit from earlier intervention.
We’ve also worked hard to develop an R&D engine within Castle that stays focused on real-world questions we can solve, while staying humble enough to recognize we can’t solve them all alone.
How is prognostic testing changing the personalized medicine approach in oncology?
I’ll start with a broad statement: prognosis drives almost every medical decision – not just in cancer, but in most diseases. Yes, there are exceptions, such as when a specific targeted therapy is available for a known genetic mutation. But in general, if a physician believes a patient is likely to have a poor outcome, the tendency is to escalate care and add interventions. Conversely, if the prognosis is more favorable, we’re more likely to adopt a watchful waiting approach, relying on clinical evaluations and imaging rather than immediately turning to surgery, radiation, or systemic therapies.
In oncology, I would say that the middle of the care pathway – the point where a patient has been diagnosed with cancer, but multiple treatment options are available – is where prognostic testing can have the greatest impact. On one end of the spectrum, there's the need for better screening in asymptomatic individuals. On the other, there’s advanced disease where the question becomes: what therapies still work? But in the middle lies a group of patients who are newly diagnosed, and where decisions about how aggressively to treat are less clear.
When prognostic testing helps clinicians risk-stratify these patients – to determine who might benefit from escalation versus de-escalation of care – that’s where we believe personalized medicine can really shine. The biology of cancer is incredibly complex, especially at the time of diagnosis. In most early-stage cases, there aren’t obvious single-gene mutations that predict behavior or outcomes.
Advancements in technology now allow us to analyze multiple biomarkers and use algorithmic models to interpret that biological complexity. This gives us a clearer view of the likely disease trajectory, and it supports more informed clinical decision-making. Whether that means helping a patient avoid unnecessary radiation – or justifying the need for more aggressive intervention – prognostic testing is central to advancing personalized cancer care.
How is gene expression profile testing addressing unmet needs in personalized medicine?
I can give you an example where gene expression profiling addressed a long-standing gap in melanoma care. We developed our melanoma test by examining the accuracy of pathological staging in cutaneous melanoma. Traditionally, early-stage melanoma staging is based on fairly basic characteristics – tumor thickness and ulceration. Some clinicians might also consider factors like mitotic rate, but those aren't formally part of staging guidelines.
Based on these features, one of the first treatment decisions a clinician makes is whether the patient should undergo a sentinel lymph node biopsy (SLNB). It’s a surgical procedure where dye is injected at the tumor site to trace and remove nearby lymph nodes, which are then examined for metastatic melanoma cells. If even one cell is found, the cancer is considered to have spread regionally.
This procedure was adapted decades ago from breast cancer protocols – at a time when we had very few effective therapies for melanoma. Back then, it made sense: if you found regional disease and removed it, the assumption was that patients would live longer. However, about 10 years ago, a large international study (MSLT-I) showed that in melanoma, whether or not patients underwent an SLNB, their overall survival remained the same. That was a pivotal moment. It revealed that while the procedure offers valuable prognostic information, it doesn’t actually change outcomes. And that was disappointing for many in the field.
So we looked at the problem from a different angle. If we’re selecting patients for SLNB based solely on clinical and pathological factors – and out of every 100 eligible patients, only 12 have positive lymph nodes – that means 88 people are undergoing unnecessary surgery. That’s not trivial. These patients are exposed to anesthesia, surgical risks, recovery time, and cost – with no real benefit.
Our thinking was: what if the primary melanoma tumor already contains biological information that could help us better predict which patients are unlikely to have lymph node involvement? Could we use gene expression profiling to more accurately identify patients who don’t need the procedure, even if they technically meet the standard criteria?
The answer turned out to be yes. By adding a biological layer to the decision-making process, we can identify patients for whom SLNB may be safely avoided. That’s a powerful example of personalized medicine in action – moving from a one-size-fits-all approach to something much more tailored and thoughtful.
So while SLNB was once considered a default part of melanoma care, gene expression testing has given us the tools to revisit that assumption. And I think that mindset – challenging standard protocols with better data – is where personalized medicine still has a lot of room to grow.
What trends are driving innovation in gene expression profile testing?
While we continue to see incremental improvements in RNA expression capture technologies, I think the most transformative change in our field over the last 10 to 15 years has been the evolution and application of bioinformatics – what we used to call machine learning, and now more broadly refer to as artificial intelligence.
The real breakthrough lies in our ability to take complex biological data – whether from whole transcriptome analysis or targeted PCR – and apply advanced computational methods to uncover meaningful patterns. These patterns can reveal clinically significant insights about an individual patient’s tumor or disease state that weren’t visible before. The result is a more personalized understanding of prognosis, which can ultimately guide more informed treatment decisions.
This shift has been accelerated by improvements in both analytical tools and infrastructure. Cloud computing has become dramatically more affordable, and the overall cost of running large-scale analyses has decreased significantly. Ten or fifteen years ago, deploying these kinds of models would have required enormous resources. Today, it's much more accessible, allowing even smaller companies or research teams to innovate in this space.
In my view, this convergence of biological data, artificial intelligence, and scalable computing has been the single biggest driver of innovation in prognostic testing – and it’s enabling a new era of precision medicine.
What do you think the oncology prognostics and diagnostics landscape will look like in 2050?
I would hope that by 2050 the majority of treatment pathway decisions are informed by molecular biology of an individual patient's tumor. That would cover everything from decisions on who should be screened, to more accurate diagnosis supporting risk stratification, to prognostic outcome predictions.
Right now, the pathology and laboratory industry is answering a few questions very well. In the next 25 years, I fully anticipate that we’ll be answering almost all oncology questions very well.
How might access to new prognostic technologies be made more equitable?
Right now, access to advanced prognostic technologies is still limited in low-income countries – as is access to certain therapies. My hope is that over the next 25 years, economies of scale will help drive down costs and enable broader distribution of these next-generation technologies – not necessarily next-generation sequencing, but innovative tools that can reach beyond just high-income nations.
A great example came earlier this year, when we attended a dermato-oncology meeting in Greece. Several dermato-oncologists from Romania approached us after reading about our gene expression profile test for cutaneous squamous cell carcinoma. They explained that in Romania, immunotherapies aren’t yet approved or available for this cancer type. For their patients, the only treatment decision is whether to proceed with radiation therapy – and even then, they don’t have enough radiotherapy facilities to treat everyone who might need it.
They told us that our test, which helps identify patients likely to benefit – or not benefit – from adjuvant radiation, could make a significant difference in how they allocate limited resources. Their question was simple: could they gain access to the test? And our answer was equally simple – absolutely. We said, let’s figure out how to make it happen: we're happy to receive tissue samples from overseas, and results can be returned digitally, by whichever method works best.
It was a meaningful interaction. Romania may not be classified as a low-income country, but, as an emerging economy, it faces real limitations. Yet the clinicians there are proactive, informed, and eager to use science to improve patient care. The fact that they found value in a US-based publication and reached out shows the global relevance of these technologies, and the importance of making them accessible wherever they’re needed.
This is just one example, but it reflects a broader principle: if we can make these tools available globally – even in places with constrained resources – we have a real opportunity to improve outcomes and equity in cancer care.
Newsletters
Receive the latest pathologist news, personalities, education, and career development – weekly to your inbox.
