Most cancers are diagnosed after symptoms appear, rather than through routine screening. Each year, 35-40 million Americans seek medical care for symptoms that could signal cancer, yet the path to diagnosis can take an average of five months. Delays in diagnosis can adversely affect survival. For these patients, the challenge is not screening but navigating a complex diagnostic process with limited tools to guide clinical decisions and accelerate the path to resolution.
Our RESOLVE™ blood test was built for this group. Rather than a yes-no answer, it sorts patients with suspected cancer into three groups: a cancer signal that warrants an urgent cancer workup, no cancer signal that may help clinicians prioritize other diagnostic possibilities, or an indeterminate signal to be monitored. When it detects a signal, it also points to where in the body the cancer likely originates, helping clinicians choose the next step more efficiently.
The Biological Heterogeneity
The lifetime risk of developing cancer is approximately 40%, yet the probability of being diagnosed with cancer on any given day is close to zero. Despite this, we continue to search for cancer as though it were an acute event rather than the culmination of a decades-long biological process.
Multi-cancer early detection (MCED) tests are a great example. If we could deploy simple blood tests that quickly identify early-stage disease, that one change could improve oncology care for millions. While this is a powerful vision, the technology (so far) has not delivered. Recent results from large screening programs have shown current approaches are generally better at detecting established cancer than early-stage disease, where therapeutic interventions have the greatest potential for success.
Part of the challenge is technical. Early-stage biomarkers are rare and difficult to detect. But the larger challenge may be conceptual. We continue to approach cancer as though it were an infectious disease or a single-gene disorder, relying on one-off, yes-or-no tests performed at isolated time points.
This approach overlooks a defining characteristic of cancer: heterogeneity. Patients with seemingly similar cancers often harbor vastly different molecular alterations. Even within a single tumor, biological diversity can shift from one region to another.
More importantly, cancer is not simply a disease state. It is a disease trajectory, regulated by the interactions between tumor, stroma and immune cells. Yet we ask a single standardized assay, performed at a biologically random moment in that trajectory, to determine whether disease is present or absent and provide clinically relevant information.
But again, people do not catch cancer; they develop it over time. The relevant question is not whether cancer is present today, but whether an individual’s biology is moving towards disease. Cancer takes a long view, and our screening strategies should do the same.
I am intimately familiar with screening dynamics. I have a family history of gastric cancer. My father died of linitis plastica, a rare and aggressive subtype. Each year, I undergo an endoscopy to look for signs of disease. Occasionally, physicians find something unusual in my stomach lining, prompting additional imaging studies. Because I am at elevated risk, these procedures are becoming increasingly extensive as my medical team searches for evidence of disease progression. It is not a process anyone would choose, and it highlights the need for better approaches.
The Systems Biology Solution
The solution may be simpler than we think, although implementing it will require substantial effort. Cancer is a systemic disease, and we should approach it as a systems biology problem. Rather than comparing individuals against population averages at a single point in time, we should compare individuals against themselves over time.
To accomplish this, we need longitudinal studies that integrate multiple biological signals. Cell-free DNA, circulating tumor cells, extracellular vesicles, epigenomic markers, proteomic signals, and immune system behavior all provide valuable information. Individually, each is imperfect; however, tracked together over time, they could reveal biological changes long before patients cross conventional diagnostic thresholds.
The model already exists. The Framingham Heart Study transformed cardiovascular medicine by following individuals for decades and identifying the biological and behavioral factors that drive disease risk. That work fundamentally changed how clinicians think about prevention, screening, and interception.
As a result, cardiology has become highly individualized. Clinicians routinely monitor cholesterol, blood pressure, lipids, and other risk factors. They do not rely on a single measurement but rather evaluate trends, trajectories, and risk over time. Even when acute disease occurs, tools such as troponin testing help distinguish meaningful pathology from background noise.
We need to apply the Framingham strategy to cancer. High-risk individuals should be followed longitudinally over many years. Long before disease becomes clinically apparent, researchers would characterize how biological signals evolve within individuals and identify patterns associated with meaningful disease progression.
New(er) Technologies Can Change the Paradigm
If this pathway seems obvious, it is fair to wonder why oncology has not already adopted it. The answer is largely practical rather than scientific. Until recently, repeatedly measuring multiple molecular signals across large populations was prohibitively expensive, technically challenging, and difficult to interpret. Today, advances in sequencing technologies, artificial intelligence, and multiomics are making longitudinal monitoring increasingly feasible, potentially shifting cancer prevention from episodic testing to continuous biological surveillance.
As these technological barriers disappear, evidence supporting this paradigm is beginning to emerge. Researchers at the International Agency for Research on Cancer recently demonstrated that longitudinal proteomic analyses can identify smokers who are most likely to benefit from low-dose CT screening. These studies are important first steps, but they represent only the beginning of what is possible.
My story illustrates the current model: intensive surveillance is reserved for people with recognized risk factors. But my father’s diagnosis did not make cancer biologically different from someone else’s sporadic disease—it simply changed when physicians started paying attention.
Imagine if we could detect those same early biological shifts in everyone, not because they had a family history, but because we were monitoring how their biology changed over time.
The goal is not simply earlier detection; the goal is earlier understanding. When we can reliably separate meaningful biological change from normal variations, we can focus interventions where they are most needed while avoiding unnecessary procedures.
By following the path cardiology pioneered decades ago, we can reduce much of the uncertainty that surrounds a cancer diagnosis. The future of cancer screening is not a better one-time test; the future is understanding how each individual changes over time. Only then will we understand cancer as it truly is—not as we wish it to be.
Once we adopt this perspective, screening’s goals fundamentally change. The 20th century was about detecting cancer earlier. The 21st century will be about monitoring healthy people so that cancer never fully develops.