Jonathan Ingles: Elena, quick question before we get into it — if your doctor told you that 1,300 people need to go through a screening program so that one death gets prevented, would you ask who funds that program?
Elena Marsh: Oh — that's a sharp entry point. I'd probably ask who's in the 1,300 first.
Jonathan Ingles: That's actually more honest than most of the public debate. Because that number — over 1,300 women screened to prevent one breast cancer death in some age groups — that's the mammography figure. And the thing that makes it genuinely hard is that the population-level mortality benefit is real. The evidence is real. And in that same population, the program produces cancers that are treated but would never have caused harm.
Elena Marsh: And no one can separate them at diagnosis. A radiologist sees the image, sees what looks like cancer, and cannot tell you whether you are the one woman in 1,300 or one of the 1,299.
Jonathan Ingles: That's the epistemological problem sitting inside a medical procedure. And we've been running it at scale for decades.
Elena Marsh: Yes — and the PSA test got there first, in a way. Prostate-Specific Antigen screening was already widespread by the early 1990s, before anyone had randomized controlled trial evidence that it actually moved the mortality needle.
Jonathan Ingles: Adopted in the late 1980s and 1990s. Professional advocacy, patient demand, no trial confirmation. The policy ran years ahead of the evidence.
Elena Marsh: So what we're really trying to work out — not whether screening is good or bad, but whether we ever asked if the biology matched the framework before we built national programs around it.
Jonathan Ingles: And that question — whether the biology matched — that's not rhetorical. The World Health Organization actually requires proof of that before recommending population screening. It's a testable condition, not an assumption.
Elena Marsh: Right — but the part that doesn't fit is how rarely that test gets applied before the program scales. And I think the cleanest way to see what passing that test actually looks like is the Pap smear.
Jonathan Ingles: Walk me through it.
Elena Marsh: Think of screening like a drug — it only works if the biology fits the mechanism. Two conditions. One: earlier treatment has to be genuinely more effective than treating after symptoms appear. Two: the disease has to follow a predictable, harmful progression. Cervical cancer, through the Papanicolaou smear and HPV testing, meets both. It targets actual precancerous lesions — cervical intraepithelial neoplasia — with a progression you can map and interrupt. The result is roughly ninety percent risk reduction in regularly screened populations.
Jonathan Ingles: Ninety percent. That's not marginal.
Elena Marsh: It's not. And the reason it gets there is that the cervical screening detects the lesion before cancer exists. You're interrupting a ramp, not catching a thing already in motion. PSA does the opposite — it finds established slow-growing tissue changes that may never progress at all. Same word, 'screening,' completely different biological situation. And the PSA program was already nationwide in the late 1980s and 1990s before anyone ran a randomized trial to see if it actually — well, actually moved mortality.
Jonathan Ingles: So the framework wasn't wrong. It was applied to a disease where the biology didn't cooperate.
Elena Marsh: That's the core of it. Screening isn't broken — it's just not a universal law. The Pap smear is the proof it can work. PSA is the proof you have to check first.
Jonathan Ingles: Which means the question isn't 'should we screen' — it's 'does this specific disease have a slow, interruptible ramp, or doesn't it.' And we've been answering that question after the fact.
Elena Marsh: And answering it after the fact is exactly where the statistics get — well, slippery is too gentle a word. Because the number that patients actually hear isn't mortality reduction. It's five-year survival rates. And those two things are not the same.
Jonathan Ingles: That's where my hot take actually holds. Lead-time bias. You find the cancer two years earlier. The clock starts two years earlier. The patient appears to survive two years longer. She dies on the exact same date she would have died anyway.
Elena Marsh: The clock moved. The disease didn't.
Jonathan Ingles: Right — and then there's the second thing, which is frankly worse. Length-biased sampling. Screening catches the slow-movers. A fast-growing tumor blows past the detection window between appointments — you find it when symptoms appear anyway. The sluggish one sits there long enough to get picked up on a routine mammogram. So the pool of screen-detected cancers is, from the start, biased toward the survivable ones. You're not getting a representative sample of breast cancer. You're getting the easy cases.
Elena Marsh: Oh — so the statistic is structurally optimistic before anyone's even run it.
Jonathan Ingles: Structurally optimistic, yes. Now imagine a woman — she goes in for a routine checkup at 52, gets a mammogram, they find something. She's told her five-year survival rate for screen-detected breast cancer is 99%. She tells her family. That number feels like proof the screening saved her life. The National Cancer Institute has frameworks distinguishing these outcome measures — mortality reduction versus survival rate — they exist, they're documented. And the survival rate is still the dominant public-facing metric.
Elena Marsh: Because 99% is the number she can hold.
Jonathan Ingles: She has no framework — and I mean no framework — to know whether lead-time bias accounts for the entire apparent benefit. That's not a communication failure. That's a choice about which number gets communicated.
Elena Marsh: And what makes this stranger — the part we haven't touched yet — is that the overdiagnosis rate for mammography ranges from negligible to more than one in three screen-detected cancers, depending on who's measuring. That gap isn't sloppiness. It points at something more structural than anyone in this debate wants to say plainly.
Jonathan Ingles: And that gap — negligible to more than one in three — look, I said earlier it was a definition problem. I want to complicate that. Because it's not just definitional sloppiness. The uncertainty is real. It's epistemological. You genuinely cannot know, at the moment of detection, whether that abnormality would ever have harmed the patient. The pathology is real. The lesion exists. The harm just — might never come.
Elena Marsh: That's the exact thing. Overdiagnosis isn't a mistake. It's structural. Built into the biology of how screening works.
Jonathan Ingles: And because no one can tell the difference at diagnosis, every detected case gets treated. Surgery, radiation, medication side effects — real harms, for a patient who statistically may have needed none of it.
Elena Marsh: With no way to allocate that harm in advance. You cannot say — this woman is the overdiagnosis, that woman is the genuine case. The population benefit and the individual harm are inseparable.
Jonathan Ingles: PSA did this at scale. Men treated for cancers that — the evidence, once it arrived, suggested would never have harmed them. And then — actually, here's what I find genuinely strange. PSA screening rates did drop after the evidence landed. The U.S. Preventive Services Task Force revised the guidance. But prostate cancer diagnosis rates didn't fall proportionally.
Elena Marsh: Wait — they didn't track?
Jonathan Ingles: Patient demand kept it alive. Demand-side medicine. The evidence moved the institutions; it didn't move the patients.
Elena Marsh: Which is — yes, and that's exactly what happened when the USPSTF revised mammography guidelines through the 2000s into the 2020s, pushing toward later start ages and longer intervals. The backlash was enormous. Professional backlash, public backlash. Recommending less medicine turned out to be politically almost impossible.
Jonathan Ingles: So the calibrated version of this is: overdiagnosis isn't a correctable flaw waiting on better imaging. It's a structural consequence of screening biology — and the mammography range, negligible to more than one in three screen-detected cancers, isn't the science being sloppy. It's the science accurately describing an epistemological problem that medicine hasn't solved. Frankly, cannot solve by refining the measurement.
Elena Marsh: And that's where the structural failure is most visible, I think. Not in the screening itself — in the moment after. When a woman sits down and is handed a diagnosis and a treatment plan before anyone has said: the odds that this was ever going to harm you were never in your favor to begin with.
Jonathan Ingles: Fine. Earlier isn't always better. I'll concede that. But — and I mean this — you try explaining that to someone who's just been told they have cancer. The Pap smear worked so well that it gave the whole framework unearned credibility. Now every screening program inherits that trust. And the patient sitting there doesn't know the biology didn't cooperate this time.
Elena Marsh: Which brings it back to your opening question, actually. Who's in the 1,300. We still can't answer that. Medicine hands someone a diagnosis, a treatment plan, and the weight of a word — cancer — and somewhere in the fine print is the admission that we screened 1,299 others to find her, and we cannot tell you which one she is.
Jonathan Ingles: That's the honest version of the script medicine doesn't have.
Elena Marsh: No. It doesn't. Thank you for pushing on this — it deserved the friction.
