A woman in her late forties came to me last spring, a few weeks after the first mammogram of her life. It found a cluster of calcifications. The biopsy came back as ductal carcinoma in situ, intermediate grade, with the kind of central necrosis that makes a pathologist pay attention, and strongly estrogen driven. No invasion anywhere on the cores.
She wanted to know one thing, and nobody had been able to tell her. Would this ever have hurt her?
There is no test that answers that. She was choosing between a lumpectomy, a mastectomy, radiation, five years of endocrine therapy, and some combination of those, for a finding that might have sat in her breast quietly for the rest of her life. She had no insurance and was paying out pocket to find answers to questions no one can answer with certainty. Including paying out of pocket for a repeat MRI that would at least tell her if her self-prescribed protocol was working.
Ductal carcinoma in situ is breast cancer, but it is noninvasive. The abnormal cells sit inside the duct and have not grown through the wall. Whether they ever will is the whole question, and for any one woman nobody can answer it. Estimates of how often untreated DCIS turns invasive run from roughly one in ten to one in two, usually over ten to twenty years, and the range is that wide because we almost never leave it alone long enough to find out. High grade, comedo necrosis, a larger lesion, younger age and close margins all push the risk up, which is why hers was taken seriously.
Treatment changes those odds in a specific way. Adding radiation after lumpectomy cuts the chance of an invasive recurrence roughly in half. It does not improve overall survival.
Screening gave her a diagnosis she cannot give back, a decision with permanent consequences, and no way to know whether the DCIS would ever have become invasive, or negatively impact her. Nobody prepares people for questions like this, but at the same time, it is not a rare place to end up.
The Screening Paradox
When you pool the randomized screening trials and ask whether the screened group died less often of the cancer being screened for, the answer is often yes. When you ask whether the screened group lived longer overall, the answer is not really.
Mammography, PSA, colonoscopy and stool testing did not clearly show a gain in total lifetime. Sigmoidoscopy did. Lung CT is the closest thing to a second exception. The large American trial reported roughly a 7% drop in deaths from all causes, and later analyses have narrowed that until it nearly touches zero.
Both can be true at once, let’s talk about how this is possible.
The first is arithmetic. Any one cancer kills a small fraction of people. Colorectal cancer accounts for roughly two of every hundred deaths in this country. Prevent a quarter of those and you have prevented half a death per hundred people. Set that half against the other ninety-eight deaths from everything else and it disappears into the noise. The modeling on this suggests a breast screening trial would need something like three hundred thousand women in each arm, followed for twenty years or more, before a difference in total mortality could be seen at all. No trial has ever been built at that size, so the missing signal is partly a missing telescope.
The second is that screening causes some deaths and does not get charged for them. When a colonoscopy perforates a bowel, or radiation for a cancer that was never going to spread injures a heart, that person does not die of colon cancer or breast cancer. They die of something else. The cancer-specific number stays clean and the total does not. There is a name for it, slippery linkage, and it only ever makes screening look better.
The third is that people saved from one cancer stay mortal. Preventing a colon cancer death at seventy is a real gain to the person it happens to, and it is not the same thing as adding years to a life.
But screening is sold to patients as a way to live longer, and for most of these tests we have never shown that it does. It moves your odds on one disease. It moves your total odds less than the marketing implies.
The Cancers We Were Better Off Not Finding
Overdiagnosis means finding a cancer that was never going to cause symptoms or shorten a life. The person gets treated anyway. Nothing is gained.
In breast screening it is about one in seven of the cancers found. The number rises with age, because the older you are, the more likely something else reaches you before the cancer would have. Between 70 and 74 it is closer to one in three. Between 75 and 84, nearly half.
In prostate screening, roughly six of every ten cancers PSA finds are overdiagnosed. Of the men who go on to surgery, 85% still had erectile dysfunction six years later in the largest randomized trial. After radiation it was about 70%. Incontinence was worse after surgery. That is changing, because active surveillance is now the preferred management for low-risk disease, and fewer men go straight to treatment than a decade ago.
Thyroid is the extreme. A model published this year estimates that between 72% and 94% of the papillary thyroid cancers diagnosed in this country since 1991 were overdiagnosed. That is somewhere between 550,000 and 730,000 people who were told they had cancer and treated for it. The death rate from thyroid cancer did not move.
But Not Looking Has a Cost Too
Around the same time last spring, I picked up a patient about the same age whose colon had never been examined once. She reached me after an emergency admission for a colon cancer that had already perforated, weeks in a hospital bed, and a scan that read the abscesses in her liver as metastases before anyone sorted it out.
Colorectal screening is one of the places where the evidence is strongest, because the test removes precursors rather than finding early cancers. She fell through anyway, the way people fall through when they have no consistent primary care, or when the prep requires a day off work and a driver, or when a previous medical experience made the exam unthinkable.
So I hold two facts at once. And the duality isn’t easy to hold, justify or explain. We find things that would never have hurt anyone, and we miss people entirely, and both failures come from a system built around averages.
The New Hot Take: Look for Everything at Once
Most commonly in this space we talk about Galleri, a blood test that looks for signals from more than fifty cancers at once. It costs $949, almost no insurance covers it, and it is not FDA approved. Its application is under review, and an FDA advisory committee votes on it on September 23.
Its largest trial read out this year. More than 142,000 people in England, aged 50 to 77. The goal was to reduce the number of cancers diagnosed at stage three or four, meaning catch them earlier, while treatment can still cure. It did not meet that goal.
Stage three and four combined did not move. Stage four alone did. Measured against the people who got usual care, those who also got the blood test had 22% fewer stage four diagnoses in the second year of screening and 26% fewer in the third. They were also about a quarter less likely to have their cancer found during an emergency, which is the worst way to find out you have cancer and the way my perforated-colon patient found out. There is no mortality data yet and there will not be for years.
I think the endpoint worked against the test. It lumped twelve cancers together, including colon and lung, where existing screening already catches disease earlier. The cancers with nothing else looking for them, pancreatic and ovarian among them, got averaged in with the ones already covered, so the places where this test adds the most disappeared into the places where it adds the least.
About half the positive results turned out to be cancer, which is far better than most people assume. For the other half, the test says a cancer signal was found and nothing is ever confirmed. What follows is months of scans and blood draws, sometimes a biopsy, and it usually runs longer than it does for the people who do have cancer. Maybe the lesion is too small to find. Maybe there is nothing there and no obvious place to stop looking. Sometimes the search ends without an answer. Nobody tells you that you are fine. They tell you to keep watching. Decide who is going to walk you through that before you order the test, not after.
Whole-body MRI is the weaker version of the same idea. Pooling the studies in asymptomatic people, the confirmed cancer detection rate is somewhere around 1.5%. Incidental findings are common, and in earlier reviews roughly a third of people scanned had something critical or indeterminate turn up. There is no long-term outcome data at all.
These tests are getting cheaper, and some can be ordered without a doctor. The woman I opened with had ordered a consumer lab panel herself and brought me the results to read. The test is no longer the expensive step. Deciding what to do with what it finds is.
Before You Go Looking
I believe finding cancer early can be better. I have watched it be better. What I do not believe is that finding it early is automatically better, and the difference between those two sentences is a roadmap.
A test result is only worth having if somebody knows what to do with it. Before you order any of these tests, and I do order some of them, I would settle five things first.
Who reads the result. If a blood test says there is a cancer signal and points at your pancreas, who orders the imaging, who interprets it, and who calls you back. If the answer is nobody in particular, do not run the test yet.
What you will do with an ambiguous answer. Most of what these tests turn up is not a cancer and not nothing. Decide in advance whether you are a person who can watch something, or whether you will need it removed, because that is the actual choice you are buying.
Who pays for the cascade. The test is the cheap part. The scans, the biopsies, the repeat imaging in three months, and the specialist visits are not, and most of them are not covered when the trigger was a screening test nobody’s insurer recognizes.
Whether the finding would change anything. For pancreatic cancer, finding it earlier plausibly changes the whole story. For a small thyroid nodule, finding it earlier mostly changes how long you live knowing about it.
Whether you would change how you live. The plan after a positive result, alongside whatever workup follows, is not a mystery. Minimal to no alcohol. Whole foods that keep your blood sugar flat, your body nutrient dense and your gut microbes happy. Daily movement, strength work twice a week, regular sweating, protected sleep, and stress handled rather than absorbed. Decide before you order the test whether you would actually do any of it. If the answer is yes, the harder question is why you are waiting for a blood test to give you permission. If the answer is no, the test is buying you information you have already decided not to act on.
Those five questions are the difference between early action and an expensive way to become a patient.
Ksenia Malarkey, ND is a naturopathic physician specializing in integrative oncology, hormones, and metabolic health.
She provides medical care for Washington and Oregon residents and offers global health coaching.
To schedule a free 15-minute consult, visit drkseniamalarkey.com
Fantastic piece. I literally had ten of those conversations about would random surveillance be worth it, asked by people who know me and my story, or have recently lost a close friend or a loved one to surprise cancer diagnosis. Most are afraid what the tests might find, unsure they can bear the costs if there are positive results but also uneasy about waiting a surprise at the next age recommended screening that their primary doctor would order. So they live in a high state of anxiety and indecision. This post considers it all and provides useful perspective. Thank you, I will share it with all of them.
This is a fantastic article.