On Sunday, The Guardian published my second in their Call the Doctor series. This one is about medical screening, and how much of it is too much.

For the story, I talked to Suzanne O’Sullivan, a neurologist at the National Hospital for Neurology and Neurosurgery in London. I recently read her book The Age of Diagnosis, and couldn’t shut up about it for weeks afterward. It felt like she was speaking directly to my experience as a doctor, as a patient, and as a sibling of a person with a severe disability. She brings clear eyes and deep empathy to some really complicated questions, and explains her thinking with incredible clarity.

O’Sullivan’s book and our conversation reminded me of the first time I ever covered overdiagnosis in a story published years ago in Elemental, which no longer exists. What I learned in reporting and writing that story gave me new language for understanding the concept of overdiagnosis, and I still come back to it sometimes when I need to explain some of the concepts to other people.

I’ve updated it slightly in the hopes it brings you some clarity, too.

On October 23, 2019, tech investor Andrew Wilkinson tweeted an MRI image of a cross-section of his body, including his lungs, spinal column, and musculature. The picture’s caption name-checked Prenuvo, a Vancouver-based diagnostic clinic offering scans like his at prices ranging between approximately $760 and $1,900.

“Wild numbers,” he wrote. “Out of 1,000 people scanned, 44 will have cancer, often in a treatable stage. Worth doing.”

The fury of public health Twitter was swift and unsparing. “Terrible idea,” “ridiculously unethical,” and “wildly irresponsible,” wrote respondents identifying themselves as doctors. 

Reached by phone a few weeks later, Andrew Lacy, Prenuvo’s CEO, was magnanimous in his assessment of how his company’s MRI screenings fit into the larger field of longevity therapies: “Somewhere amongst all of these different interventions there are, for sure, some things that actually will make you live longer,” he said. “I trust the patient to make an informed decision about what to do with their own health.”

Four years later, Kim Kardashian gushed about her own Prenuvo scan, and the cycle began anew. The company has been in business since 2018 — but as of 2025, one-third of its full-body MRI business happened in a 7-month period following December 2024. Interest has only accelerated since. Witness the almighty Google Trend:

The idea that early disease detection is a net good for human health is baked into every level of medical practice. The concept is integrated throughout medical education and it underpins preventive health recommendations, research funding, diagnostics research and development, and the marketing of mammography and endoscopy centers.

While early diagnosis is key to living longer and better with certain conditions, cancer epidemiologists increasingly argue that the medical establishment overstates the benefits and underplays the harms, especially when it comes to cancer. “Our ability to detect abnormality is increasing dramatically,” said Gilbert Welch, a senior researcher in the Center for Surgery and Public Health at Brigham and Women’s Hospital in Boston. “The question of how much we’re helping people by looking for things to be wrong is [an] open one.”

Cancer is not just one animal. As an old medical metaphor describes the disease, it’s more of a barnyard menagerie of birds, rabbits, and turtles, and the goal of early detection is to fence them in by catching them early. Birds represent aggressive cancers that grow so fast they’ve already spread by the time they’re detectable, while rabbits represent more slowly progressing cancers that can be identified by early detection. Turtles represent conditions with cellular abnormalities consistent with cancer but disease behavior that is not: They either regress on their own or progress so slowly that people die of other things before the cancer even causes symptoms.

“At that extreme — [for] the birds — the screening test doesn’t do any good,” said Barry Kramer, a retired oncologist and former director of the Division of Cancer Prevention at the National Cancer Institute (NCI). You can catch the rabbits if you build enough fences, but for turtles, a fence is overkill; they aren’t going anywhere, anyway.

Importantly, screening aims to detect cancers and other conditions early in people who do not have symptoms of those conditions. These aren’t the chest X-rays you might get if you’ve had a cough for a few months, or the blood tests you might have if you’ve had fatigue and weight loss. They’re the tests your doctor might mention at a routine physical, like checking your cholesterol once every few years, or getting annual mammograms.

When screening has its intended consequence, it detects a condition early enough to prevent people from dying prematurely as a result of that condition. But if a screening test causes harm that outweighs its preventive benefits, like financial collapse, severe emotional strain, or physical complications from unnecessary procedures, it’s not a very good test. In fact, it might be worse for a person than doing nothing at all.

Imagine what might make for a good “enclosure” in the cancer barnyard. To prevent a premature cancer death, a screening test must identify a cancer whose progression is preventable (that is, not a bird) and would cause premature death without treatment (that is, not a turtle). That leaves us with rabbits as the only targets screening can actually do anything for: potentially lethal cancers whose harm can be reduced with early treatment.

The problem is that when it comes to cancer, doctors aren’t very good at distinguishing turtles from rabbits. Shannon Brownlee, senior vice president of the Lown Institute, a nonprofit health care think tank, said that while cancer is fundamentally a genetic disease, “we’re still using a 19th-century technology to diagnose cancer. It’s called ‘the microscope.’” Cells from life-threatening cancers and non-life-threatening cancers (meaning the cancer will not be what causes a person’s death) might look the same to a pathologist examining them in a tissue sample, and sometimes there is no way to determine which are which.

“We’re getting a little bit better at saying which breast cancers are super dangerous and which cancers might be super benign,” said Brownlee. As an example, she cites a common type of often-indolent breast disease, ductal carcinoma in situ, or DCIS, a Stage 0 breast cancer. Several studies demonstrated that over long-term follow-up, fewer than half of women with this condition develop invasive cancers that could kill them. Still, DCIS looks like cancer under a microscope, so for years, identifying it led to lumpectomies, mastectomies, chemotherapy, and radiation. (In 2025, a randomized clinical trial suggested watching and waiting is as safe as surgery for low-risk DCIS, and clinical practice changes will likely follow.) 

If scientists really wanted to understand the difference between turtles and rabbits, and lived in a world without bioethics, they could design a study withholding treatment from people diagnosed with cancer by the appearance of a biopsy under a microscope. Once those people died, pathologists could perform autopsies to determine how their cancer progressed, and whether their deaths were caused by cancer or by something else entirely. If a large proportion of people with cancerous-appearing biopsies died with their cancers but not as a result of their cancers, that would be a big clue that their cancer wasn’t life-threatening — and that people might not need treatment if diagnosed with the same condition.

That study can’t be done, of course. But what medical professionals can do is look at how the incidence, late diagnoses, and deaths related to certain cancers have changed over decades of medical progress. If those metrics are followed for a long enough time span, experts can see how they vary when new cancer screening tests and cancer treatments are rolled out.

In a 2019 paper, Welch and Kramer did just that, finding that few of these metrics demonstrated reductions in cancer deaths attributable to screening technologies. In breast cancer, for example, increases in cancer incidence coincided with increased use of mammograms — but simultaneously, diagnoses of breast cancer in late, more lethal stages remained stable. That’s not ideal: “The real goal of screening isn’t just to find more cancer; it’s to lead fewer people to present at a late stage,” said Welch.

Imagine breast cancer diagnoses as a cup of water divided between two glasses, one representing early diagnoses and a second representing late diagnoses. A screening test should work by catching cancer earlier, moving diagnoses out of the late group and into the early group, leading not only to more water in the first glass, but also less in the second glass.

Welch and Kramer’s study showed that effectively, the water level in the first glass went up while the level in the second glass remained the same, suggesting that mammograms did not lead to fewer late-stage breast cancer diagnoses. Instead, said Welch, “what you’ve mainly done is identified a whole new category of women that otherwise wouldn’t have presented.”

In other words, the vast majority of the breast cancer that screening mammograms diagnose is in people who would have been fine if their cancer had never been diagnosed at all. Fundamentally, Welch argues, medical experts don’t know enough about those women’s cancers to make conclusive statements about mammography’s benefits for all women.

Despite the lack of major new evidence, recommendations have moved toward more screening for breast cancer, not less. The US Preventive Services Task Force expanded mammography screening recommendations in 2024. 

As this new category of breast cancers is increasingly diagnosed, breast cancer survivorship starts looking pretty good — in large part because the new breast cancer patient pool increasingly includes people who weren’t at risk for premature death to begin with. While breast cancer mortality has improved, say the authors, that’s a result not of screening, but of vast improvements in treatment. Estrogen blockers like tamoxifen have dramatically reduced breast cancer death rates since their introduction in the 1980s. “The better we get at treating disease, the less important it is to find it early,” said Welch.

It’s not like cancer screening doesn’t help anyone: Screening likely has a net benefit in certain groups at high risk for certain cancers (for example, screening heavy smokers for lung cancer). And some screening tests, like Pap smears, do increase life expectancy in the general population. For 10% to 15% of people with any given cancer, early diagnosis and treatment prevents a premature death, said Kramer. But preventing a premature cancer death isn’t the same as preventing death altogether. In a study comparing deaths among people randomized to receive no colon cancer screening or screenings every one or two years, deaths due to colon cancer were highest in unscreened people and lowest in the most frequently screened people. But the rate of deaths due to any cause — what scientists call “all-cause mortality” — was identical among the groups. Meaning, an early diagnosis of colon cancer might have prevented people from dying of colon cancer, but it didn’t lengthen their lifespans overall. Those findings were confirmed by a May 2026 study focusing on colonoscopies.

People might debate whether a colon cancer death is better or worse than a death due to heart attack or dementia, but the bigger picture is one of misplaced priorities, said Welch. “This is the way we’re measuring the quality of our health care system — how much testing we do in the well. That’s crazy!” he said. “The quality of the health care system is how good we are with sick, suffering patients, acutely injured patients. That’s what we ought to be focused on.”

A major reason for this misplaced focus is the relative lack of information and public conversation about the physical, emotional, and financial harm associated with screening. This vacuum pervades all sectors of the health care infrastructure: There is relatively scant research on the subject, giving physicians little but anecdotal data to use in counseling patients about the potential harms of screening.

Although Brownlee said there has never been a national accounting of the harm from overtreatment, physician surveys suggest that in their judgment, overtreatment accounts for between 15% and 30% of medical care. By some estimates, the U.S. health care system wastes $75.7 billion to $101.2 billion in overtreatment and low-value care.

“In public conversations about overtreatment, which is the downstream effect of overdiagnosis, we talk about the cost a lot,” said Brownlee, “and we say a little bit about harm, but we actually have no idea the scope of harm to patients from overtreatment.”

Many doctors sense patients are harmed by “cascades of care” — batteries of tests that follow an initial, abnormal test and may not lead to meaningful findings. Much of that harm is financial: In a 2018 study, cancer diagnoses led to financial depletion in more than 42% of patients. But harm also extends to the physical and emotional realms. An analysis of physician surveys from around the same time suggested that nearly nine out of 10 doctors fel that care cascades (not necessarily related to cancer screening) had caused their patients harm, with nearly two-thirds reporting some combination of psychological harm, physical harm, or financial burden.

Negative consequences were not restricted to patients: In the same study, 69% of physicians reported wasted time and effort as a result of care cascades, while about half reported frustration, anxiety, or both. These outcomes suggest doctors “need to be really thoughtful… about ordering tests, and being really up-front with patients about what might happen after,” said Ishani Ganguli, associate professor at Harvard Medical School and the study’s lead author.

Despite their sense that patients are harmed by these cascades, many doctors find it challenging to counsel people about negative consequences when there is limited data to back them up. “The harm is understudied and therefore not talked about,” said Brownlee. “When physicians talk about screening, they talk about the upsides.”

Cancer survivorship messages, often pink-beribboned and morning show-endorsed, are also a barrier to even-handed counseling in clinical settings. Doctors may be aware that many cancer “survivors” are people who would have survived with or without early cancer detection, but according to Welch, “no one likes to point [that] out.” He said, “The other possibility is, you’re not a survivor. You’re the person that has been harmed — you’re the person that went through all this for nothing.”

The fear of missing something important — and of being sued for it — are additional disincentives for physicians to go off-script about screening tests, or to resist follow-up testing of abnormalities. Medical norms also shape physician decision-making: “It’s hard for any human being, but especially those who’ve trained in our environment, to leave something alone,” said Ganguli. Furthermore, the profit motive driving many screening tests is very difficult to ignore: Globally, mammography is a $3.1 billion business. Routine colonoscopies, the most expensive routine screening test, and one not recommended by Canadian public health authorities, are a major factor in both hospital profits and astronomical U.S. health care costs.

But there’s a murkier and more complicated reason for physicians’ reluctance to discuss the drawbacks of screening: For many, doing so means acknowledging that the medical interventions around which they have built careers may have hurt more than they have helped. The cognitive dissonance is too much for many of them, said Brownlee.

For cancer medicine and radiology specialists, the prospect that their work might cause little net benefit with a lot of harm constitutes an assault on their identity. “This is how they’ve defined themselves — ‘I save lives,’” said Brownlee.

Welch points out that efforts to promote screening require an element of fear, and false-positive findings tend to undermine people’s sense of resilience — something he said is fundamental to health. “You can’t test yourself to health. The things that really promote your health are not very sexy,” he said. Real food, regular movement, and finding purpose aren’t high-tech, he said, but they — not medical intervention — are the true foundations of a long and healthy life. 

One of the reasons enacting a population-level understanding of screening research is difficult in clinical settings is that the bedside involves not a population, but one doctor and one patient. Both parties come to clinical interactions with personal histories, biases, and fears that shape not only the decisions they make but also how they reflect on those decisions afterward.

The growing science of decision aids — tools used during doctor appointments to inform people about available treatments, as well as benefits, risks, and costs — aims to guide patients and doctors through the waters of clinical decision-making muddied by generations of cultural and academic bias. Still, certain stories seem to flip a kind of override switch, especially when it comes to cancer screening.

The day after we spoke, Lacy, the Prenuvo CEO, emailed me to share one of those stories. In the companys early years, it diagnosed a young, industriously healthy, and asymptomatic woman (who, not coincidentally, was married to a radiologist) with a 10-centimeter abdominal mass later determined to be a rare tumor — one generally treatable only by surgical removal. A year and a half after an operation to remove the mass, the woman and her husband were planning additions to their young family.

What would be different if this woman’s diagnosis had been delayed until the tumor’s growth resulted in a gastrointestinal bleed, its likeliest first symptom? Would it have been too late? Did her early diagnosis prevent a premature death?

The answers are not clear, but it is no accident stories like hers get told more often than those of the emotional trauma, physical harm, and financial devastation that can follow care cascades in the wake of tests like the one she had. Her story was shared with me by someone with a financial stake in the technology that proves its point — and not necessarily because it is representative of most customers’ experiences.

When the health care system invites healthy people in, rather than relieving suffering people who are sick, it should do so only for very good reasons, said Kramer. “It’s very difficult to make a healthy person better off than they already are. And unfortunately, sometimes, with the tools we have in hand, it is not so difficult to make them worse.”

This story was originally published in Elemental in 2019 as “The Hidden Problems of Early Cancer Detection: Cancer experts are increasingly arguing that the benefits of early detection are often overstated and its harms underplayed.” It has been updated to reflect the most recent science. 

Got a tip on any of the stories I’m covering? Email me at [email protected], fill out this form, or find me on Signal at kerenlandman.07.

Like what you see here? Subscribe, and feel free to share this with someone else you think might like it.