The Number Needed

06/09/2015 10:52 am ET | Updated Jun 09, 2016

One out of two women screened with mammography for breast cancer, over ten years, will have a false positive result. She will receive a call asking her to come in to discuss her results and she will laugh, a laugh that sounds strange in her ear. She'll wonder why she did it, and when she puts down the phone she'll still hear that laugh. She'll worry as the appointment draws near, think on the women she knows of who have lived with or died from cancer of the breast.

That one sad aunt who's still in remission after all these years. She will Google words like metastasis, mastectomy and mortality. She will arrive at the office early and wait an hour, too nervous to read the ancient copy of Newsweek sprawled open on her lap, until she is called back. "Suspicious" will have never sounded so sinister.

That's half of all patient's screened -- it's over 17 million women in the U.S. alone, and as many nervous laughs, sad aunts and sinister words. For every five patients screened for breast cancer with mammography, one of those women will undergo surgery unnecessarily over the course of ten years. She will lose sleep. She will cry and her loved ones will cry for her. She will approach the day stoically or desperately or gracefully. She will be cut, she will experience complications or an uneventful recovery. She may walk away with nothing more than a small scar in her breast, or slightly deeper still. That's one fifth of women screened, over seven million in America, and as many small scars.

One more important statistic to keep in mind is the number of women whose lives are saved by screening mammography*. It's a small number and easy to remember. It's zero.


On my first interview for medical school, I felt something like what I imagine a first-time base jumper feels. They crammed us into a florescent waiting room with the dimensions and ventilation of a walk-in closet and had us sit in stomach-twisting anticipation. A residual electric hum hung in the air, and I made small talk with the other candidates. They spoke casually of elite academic pedigrees, self-started companies and papers in Cell, and I tried not to seem out of place. When it was my turn, they called my name and I followed the slight, bespectacled, aging doctor (cast by Alan Alda in all my future recollections) into his office. After a roundabout conversation that skipped over all the key points of my application, he started to ask some of the standard questions.

"Why medicine?" he asked.

"I like the idea of making people better," I said, "the idea of seeing through all the pathophysiology and making them healthy again with just a prescription."

Dr. Alda made a noise like a scoff. I looked up, and his face let me know that he was, in fact, scoffing at me. He asked, "What medicine do you imagine can do that? What do we have in our arsenal that can actually make people better?"

Now, as a fourth-year medical student, I'm better equipped to answer that question. I know lots of therapies that are efficacious, that clearly help. We manage asthma well, improving outcomes and quality of life. Antibiotics for pneumonia (but not strep throat, ear infections or bronchitis) are an effective and, often, life-saving intervention. Screening for cervical cancer has made the most lethal gynecological malignancy into the one with the best prognosis.

But I'm also better able to see the cracks. The US Preventative Services Task Force has recommended against using blood tests for PSA (Prostate Specific Antigen) to screen for prostate cancer since May of 2012, and yet more than 40 percent of men over 75 are still being screened. This results in more than 600,000 unnecessary prostate biopsies a year and does not save a single life.

We put stents in 600,000 hearts a year. Stents can be lifesaving in acute coronary syndromes, like heart attack, but more than a quarter of stents are placed in non-acute patients, where the evidence shows that they do more harm than good. This results in over three thousand people harmed every year by bleeding events, stroke and kidney damage.

So, how can patients know what's worthwhile and what, in medicine, does more harm than good? The trick is always, and has always been, going to the data. We do a lot of things in medicine that are contrary to the data because of habit or because of the myriad conflicting voices, which are often driven by profit. Sometimes, for doctors, it's just a question of a really compelling theory. When the mechanism of action of a drug or the idea of screening is so theoretically sound, doctors with the best intentions can be lead astray if they don't have good data. In the past it's been hard to look at the data and interpret it, but there are new statistics that are designed to be understood.

They're called the Number Needed to Treat (the "NNT"), and the Number Needed to Harm (the "NNH"). We tend to think of medicine as having clear outcomes, being intrinsically efficacious for the individual, but this is seldom the case. My interviewer's scoff meant that what we do in medicine is rarely so clear cut, but we try to make it so that the benefits outweigh the harms of our interventions. The NNT represents the number of people I'd have to treat with a particular drug, screening tool or surgery to help one person. If the number is low, the intervention is powerful, and if it's high, less so.

Likewise, the NNH represents the number of people I'd have to treat with a drug to harm one person. You don't want that to be a low number any more than you want the NNT to be too high. In other words, if the NNT is higher than the NNH, it's more likely to cause that harm than that benefit.

Screening mammography for breast cancer has an NNT, the number needed to treat to prevent one death, of infinity. It can't be done. It has an NNH of two to get a false positive result and just five to receive unnecessary surgery over ten years of routine mammographies. We won't save one life by screening mammography, but we will harm 1 in 5 people screened with an unnecessary surgery. Who would willingly play those odds? You can't win, and you have a 20 percent chance of being cut open.

This idea is slowly gaining traction. The New England Journal of Medicine published a manuscript in May of 2014 reporting the recommendation of the Swiss Medical Board to abolish screening mammograms based on NNT and NNH evidence. But there are arguments to be made in favor of breast cancer screening. One could say that if you look at the data the right way, cancer mortality does decrease, even if overall mortality doesn't, and that the aversion of a cancer death has value, even if the alternative is still death. One could say that maybe screening has gotten better since all of our current data was published, and now screening mammograms might actually have a benefit. One could say a lot of things, but at the end of the day we have mobilized a massive, national screening effort at considerable cost, not just in terms of the billions of dollars spent, which, to our best and most recent estimates to date, have saved no lives, but also in terms of the emotional currency of every woman screened and hurt. Couldn't those resources be spent elsewhere? Couldn't you imagine a place in healthcare, or in society where all of that money and false hope and those broken promises could find a finer mark? When the powers that be argue for our broken system, they can be quite persuasive, but they also made me swear, "first do no harm," and I'm not convinced.

I'm going to be a doctor in a few months and, looking forward, I often wonder what kind of doctor I'm going to be. Will I be cowed by the pressures of modern medicine, the threat of being sued if I disagree with the status quo or the danger of ruffling the feathers of more seasoned doctors, chiefs and administrators, who see the old ways as the best? And I don't know; I just hope I can always find the strength to do what's right.

It's because I'm new to medicine that I have a chance at seeing the problems inherent to the field, and it's why the public at large might take better to these ideas than many of the cleverest minds in medicine do. This is a field where traditions are robust and change is slow. Say what you will about my generation, the computer kids, about the breakdown in tradition and the questioning of everything, but I think this kind of evidence-based movement stems from those very values. A growing cohort of young doctors and students (and a few forward-thinking old-timers) have essentially found a way to look at everything we do and ask, is this any good? More than that, the NNT is a way for a patient to look at a therapy and get an instant understanding of what works. It's more than a statistic -- it's a mathematical form of honesty.

The instinct in medicine has always been to see an older, venerable Attending Physician and trust his or her wealth of clinical experience. But if experience has taught us anything as a species, it's that anecdotal evidence is weak, and when you're playing with people's lives you had better have something to back it up.

It's important to remember that much of the trust we place in those doctors is warranted. There is something in experience that is irreplaceable, and these doctors bring something to the table that cannot be taught. Evidence-based medicine cannot replace the clinical decision-making tools wrought by the clever minds of wizened attendings through countless hours of late-night calls, worrying on patients and experiencing near-misses. They have a diagnostic and therapeutic edge that's been sharpened over years. With evidence-based practice, we simply hope to point this weapon in the right direction. The synergy of experience and good data yields clinicians of the very highest caliber, with the best chance of doing right by our patients. In the end, that's all we can hope for.

David Newman, an emergency medicine doctor and research director at the Icahn School of Medicine (Mt. Sinai), envisions a world in which patients know the NNT and NNH of everything. He created a Website,, from which all of the statistics in this article are derived, that spells out, in clear language, the NNT for many common medical practices using robust evidence. Some of the therapies he discusses clearly work (the NNT is small and the NNH is inconsequential) but some don't. And it's the duty of every physician -- and the right of every patient--to know these things.

Take statins, the cholesterol-lowering drugs it seems the vast majority of adults are on these days. Lowering cholesterol is good, right? Would you feel the same way if you knew that patients without any previous heart disease who took statins for five years had no fewer deaths (although for every 60 patients treated they had one fewer non-fatal heart attack and for every 268 treated, one fewer non-fatal stroke)? Comparatively, the NNH was 10 for muscle damage and 50 for developing diabetes, a lifelong illness that significantly impacts both quality and length of life.

Even as a student, I can feel the tectonic plates of the medical world in flux, and one compelling idea is at the center. It's the idea that this one powerful statistic can subvert so much of the uncertainty in medicine: the anecdotal evidence, the inconsistency of practice and, more than anything else, the practice of medicine that doesn't work. One of the reasons that this statistic is so powerful is because patients and physicians can understand it, and should understand it. And if they do, just like that, everyone becomes a student of medicine. Just like that we are all each other's doctor.


*This is specifically screening mammography, diagnostic mammography, such as is performed after discovering a palpable breast lump, has proven benefit.