Katherine Carpenter couldn’t sleep. For more than a week she’d been coughing herself awake every night and then hacking until she retched. Finally, she decided to see a doctor.
The physician suspected bronchitis and wrote Carpenter a prescription for heavy-duty cough medicine. She also suggested antibiotics. That’s pretty standard: Up to 80 percent of people who go to a physician for acute bronchitis are prescribed antibiotics. But Carpenter, an import entry agent for UPS, didn’t want antibiotics. She thought they’d stop working if you take them too often, and she suspected her symptoms were caused by a virus, which antibiotics don’t affect anyway.
She didn’t know it, but her hesitation had science on its side: A meta-analysis in the Cochrane Database of Systematic Reviews looked at 17 trials on antibiotics for people with acute bronchitis, and concluded that they only slightly shorten the duration of the illness—if they have any benefit at all. (And of course there’s the issue of antibiotic resistance to consider.) In the end, Carpenter refused the prescription, and her bronchitis eventually cleared up. But the experience left her with the distinct impression that she was just one more patient on the medical assembly line. “I felt like a number,” she says.
Instead of being a number, Carpenter might have preferred to see a number, one that can help us weigh the benefits (or lack thereof) of a treatment. That number exists, and it’s called the number needed to treat. Developed by a trio of epidemiologists back in the ’80s, the NNT describes how many people would need to take a drug for one person to benefit. (The NNT for antibiotics in a case of acute bronchitis is effectively infinity, because the medicine is no better at curing the illness than a placebo.)
Consider a couple other examples: If your kid is throwing up and you take her to the hospital, she might get a drug called Zofran. The NNT for that is 5, meaning that only five kids need to take Zofran for one of them to stop throwing up. And if you look at Zofran’s “number needed to harm” (the number of people who would need to take a drug for one to have a bad side effect) the answer is … well, there really isn’t one—no one has a significant side effect.
Now, say you’re pushing 50. You’re healthy, but your doctor suggests you start taking a baby aspirin. Just in case, you know? That NNT is 2,000. That’s how many people have to take a daily aspirin for one (nonfatal) heart attack to be prevented. Statistically speaking: Not especially helpful.
It’s unfortunate, then, that the NNT is not a statistic that’s routinely conveyed to either doctors or patients. But you can look it up on a site that you’ve probably never heard of: TheNNT.com. Started by David Newman, a director of clinical research at Icahn School of Medicine at Mount Sinai hospital, the site’s dozens of contributors analyze the available studies, crunch the numbers on benefits and harms, and then post the results. While a low NNT is generally “good” and a high NNT is “bad,” you also have to consider the severity of both the illness and the drug’s side effects. Which is why the team added a color-coding system: Green for when a treatment makes sense, yellow for when more study is needed, red for when the harms and the benefits cancel each other out, and black when the harms outweigh the benefits.
Newman’s goal for the site is nothing short of a revolution in medical practice. He wants doctors to base their treatments on good scientific evidence, not tradition, hunch, and the fear that patients will see them as doing nothing. And he wants patients to start demanding such care. That’s the big picture, anyway. For now, he’d be happy if he could just get people looking at medicine in a different way. “People tend to think that if it’s a medical intervention, there’s science behind it,” he says. Unfortunately, that’s often not the case. “It is a lie to tell patients to do something without telling them, ‘You should know we’ve done lots of research on this and we can’t find any benefit to it.’”
It’s a bright, cold morning outside Mount Sinai Hospital in New York City. Central Park is across the street, and runners huff along the paths, baby joggers in front, dogs in tow. Newman has just arrived at work in the emergency department where he’s an attending physician, and he’s giving his residents the “Three Networks” spiel.
“ABC, always be closing. You want to be moving patients back home or into the hospital for treatment. At the ER, the front door is always open, but if the back door is closed, and you have people mounting up, things get missed, heart attacks occur, sepsis gets worse, and people die. Always be closing.” Then comes NBC, never be consulting—don’t call specialists if you can avoid it. Then CBS, close before signoff. There’s good evidence that medical errors are more likely to occur with handoffs. Discharge your cases before the end of your shift to avoid these errors, he tells his charges.
Newman, who received an Army Commendation Medal for his service with the 344th combat support hospital in Baghdad, is steady. Steady blue eyes, steady uncomplicated speech, steady smooth movement through the ER. He’s got an easy humor about him, too, especially outside of work, slipping into various New York accents for wry, sometimes self-deprecating effect. He’s trim, dressed in gray slacks, tie, and shirt. No white coat. “Some doctors put a white coat between them and their patients,” he says. “They think that science is a thing that is cold and dispassionate.”
Newman wants to use science to protect his patients from treatments that are not justified by research and evidence, and he wants to impart that science to young medical residents. So he’s constantly grilling them on the assumptions that underlie their recommendations and diagnoses.
Patients are streaming in. There’s a woman with high blood sugar. The resident wants to give her insulin, but Newman puts her on an oral medication instead because she’s afraid of needles, and her current numbers are high but not life-threatening. A woman is having an early-term miscarriage. The resident wants to give her a RhoGAM shot, but Newman points out that studies show it would be useless this early in the pregnancy. An asthma sufferer gets three days of prednisone. Newman tells his group that steroids are a treatment to give early. If asthma patients receive them within the first hour, evidence suggests that the chance they’ll be admitted to the hospital (or that they’ll bounce back to the ER later) is much lower. Indeed, the NNT for steroids given within an hour of an asthma attack is 8 (green). As the review on the site points out: “The impact seen in these trials might be surprising to many given the conventional wisdom that effects of steroids are delayed.”
We visit a man with a red, painful abscess on his right knee. It needs to be drained. Newman whips out a small flashlight for a better look and speaks to the patient in Spanish, jocular and easy. No, you won’t be asleep, he says, but we will give you a shot. After numbing the area, Newman makes a quick nick with a scalpel to start the draining. As we walk back to the computer station at the center of the ER, he tells me that the abscess patient is taking Zyvox, an extremely expensive antibiotic given to him by a doctor he saw a few days before. A 20-pill course can cost up to $3,500. There’s limited data for Zyvox, so it hasn’t been evaluated on the NNT site. There are many treatments that Newman and his team would very much like to analyze, but they’re sometimes hampered by a lack of quality research.
In this case, though, the general consensus “among people who care about this kind of stuff,” Newman says, is that the antibiotics are unnecessary for a simple abscess like this one, even in the age of nasty antibiotic-resistant staph infections. Just getting all the junk out and keeping the wound clean is usually enough. But Newman isn’t going to pick that fight today. He doesn’t want to get in the way of doctors’ plans for their patients. If he respects their decisions, they’ll be more likely to respect his, after all.
As statistical tools go, the idea of the number needed to treat is relatively new. It was first described in 1988 by epidemiologists Andreas Laupacis, David Sackett, and Robin Roberts in a New England Journal of Medicine article titled “An Assessment of Clinically Useful Measures of the Consequences of Treatment.” They start by sketching out the problems with a number called the relative risk reduction. That’s the measure you often see hyped in media reports of scientific studies. Imagine, for example, a study of heart disease that finds that a new drug reduces the risk of death by an astonishing 50 percent. The reality behind that number is that the risk of death over a 10-year period for, say, a healthy 45-year-old man weighing 200 pounds went from 5 percent to 2.5 percent—50 percent! Such a finding is clinically significant, yes. Worthy of publication, maybe. But not quite as astonishing.
It would be better, the authors write, to look at a number called the absolute risk reduction—the 2.5 percent reduction that resulted from the new drug. But working with that measure can be hard to understand, because it is actually a percent of a percent. To make it more intuitive and apprehendable, the authors explain, you can use the inverse of absolute risk reduction: Divide 1 by 2.5 percent, or .025, to get 40. And that’s the number needed to treat. Forty people have to take the drug for one person to benefit. So is it worth taking? That depends. The NNT isn’t crazy high, so you might go for it, especially since a heart attack can kill you. But if the drug has terrible side effects, you might not. “Different people value different things differently,” Laupacis says. “So they might be more scared of the rare harm.” On the other hand, when the original article on AZT came out in the early days of the AIDS epidemic, the NNT to prevent one death was around 6 (that’s very good). “It can also tell you to do something,” he says.
Sometimes the NNT can be functionally infinite. Haney Mallemat, an emergency physician at the University of Maryland School of Medicine describes what happens to patients who come to the ER with upper gastrointestinal bleeding. They’re given an intravenous proton pump inhibitor and kept in the hospital, sometimes for as long as 72 hours. But Mallemat, who uses the NNT web site with his patients and students alike, points out that “of all the studies they’ve looked at, no person saw any benefit.” No one was harmed, either, unless you consider the cost and time spent in a hospital a harm, which most people would. (The practice gets a red on the NNT site.) But doctors administer proton pump inhibitors because administering proton pump inhibitors is what they’re trained to do. Habits—whether based on old literature, biased studies, or just educated hunches that get ingrained in protocol—die hard.
Indeed, more than 90 percent of doctors believe that their colleagues practice this kind of “defensive medicine,” according to a survey published in the Archives of Internal Medicine. You do what’s done because it is considered to be the so-called standard of care. And not adhering to the standard of care can be considered negligence if something goes wrong and you get sued for malpractice. The unfortunate result of this contorted logic, according to the Institute of Medicine, a policy research organization, is that the US spent $210 billion on unnecessary services in 2011 (some 8.4 percent of the more than $2.5 trillion we spend annually on health care), and untold numbers of patients are subjected to pain, anxiety, and even death as a result. The NNT could help prevent a lot of that suffering—physical, mental, and financial.
“The concept of the NNT, and the website itself, is tackling something fundamental about how we think about health care and behavior and risk,” says Vikas Saini, president of the Lown Institute, which focuses on overtreatment in US health care. “All of us have trouble clearly distinguishing degrees of risk, and that is compounded by the enormous noise that accompanies health information. The signal is lost.” If Newman’s site continues to expand beyond its current number of about 200 write-ups, and if the NNT were included with every published article about a treatment, the result would be happier, healthier patients and less waste in our health care system.
The ER is crowded, the front door is open. A woman with hives gets Benadryl. A man with a terrible infection on his hand receives IV antibiotics. He got the infection on a previous visit, from the insertion of an IV line. “I’m sorry we did this to you,” Newman says. “Eh,” the patient replies, “you guys are the good guys.”
A fellow on a gurney has pain that might suggest appendicitis, but after palpating the patient’s abdomen, Newman suspects gastritis from drinking too much the night before. He could send the man for a scan, as most doctors would, but a false positive could bring unnecessary surgery, with its own complications. When Newman types up his notes, he hits a quick key that automatically inserts a block of text he clearly uses frequently: “...while this remains possible, given the low likelihood it would be more likely to result in harm than benefit if we moved forward with any further testing at this stage.” Later the young man seems improved and is calmly typing on his phone.
Newman always places a hand on his patients—on an unblanketed ankle or a clammy cheek—even when he’s not directly checking for clinical information. “The patients get something out of it,” he says. “And the contact helps me understand them medically. If they’re warm, if they’re cold, if they’re nervous or jittery.”
An elderly woman with advanced stomach cancer arrives. She has been on opiates for her pain, but over the past day or two, she seemed particularly out of it. Her family called an ambulance and the EMT gave her Narcan, a drug typically used on overdosing heroin users.
The Narcan ripped all the pain-relieving molecules off the receptors in her central nervous system, leaving them exposed to all the pain that cancer brings. She is in agony, and you can hear her moaning from behind the tan and gold curtains. Plus, she’s now in full-on withdrawal, an experience that heroin addicts describe as the most agonizing symptoms of withdrawal hitting you all at once.
Newman is more stony than steady now. He explains to the EMT why the Narcan was a bad call, an understandable call, but a bad one. (Newman came to medicine through an ambulance himself. He was a philosophy major who worked as an EMT before switching to medicine.) He’s trying to calculate the right dosage of a new pain drug for the cancer patient, maybe one that can overpower the Narcan molecules. His wife, Ashley Shreves, is also an emergency medicine doctor and she’s in the ER tonight too. She’s helping him with the calculations. Nothing adds up to a solution.
Newman is trying to figure out what’s going on with her. Was her disorientation triggered by something? Is this the progression of disease? Is she dying? Quite possibly. The lactate levels in her blood are higher than 15, and the mortality for people with lactate above a six or a seven is about 90 percent. The pH level of her blood is so low the monitor can’t measure it. “You rarely see numbers that bad on someone,” he says. Family members are starting to arrive, slipping behind the curtain, then stepping out into the vestibule between the two sets of sliding glass doors that lead out to the ambulance bay, embracing while EMTs lumber past. A priest walks in.
There are no NNTs for this particular situation, but the woman had previously told her family that she didn’t want aggressive measures if her chances were low. So there will be no intubations, no chest compressions, no machines and all the fruitless discomforts they entail. Sometimes the most humane treatment can be no treatment at all. “The thing the NNT does above and beyond everything else is not to quantify the benefit, it is to quantify the negative space,” Newman says. “The interventions that you have been taught were so powerful turn out not to be the most important thing for the patient.”
It might be just a number, but the NNT can tell doctors when to do something and when something is not likely to be useful—or can even be harmful. Once we, patients and families and doctors alike, get our heads around the idea that we shouldn’t always expect a drug or a procedure, we can begin to expect the right level of medicine and not just medicine for medicine’s sake.
And the right level of medicine for this patient, now, is to not be in this bustling ER where the honk of the machines is interrupted only by the metallic schwaaank of curtains on their ceiling casters. Newman works the phones some more, talks to her doctor and her relatives. She’s going to be moved up to the palliative care ward, which is devoted not to treatments but simply to reducing pain. It’s beautiful, Shreves tells me. The rooms are big and private so the woman’s family can fit in, the nurses are specially trained. All the rooms have huge windows that overlook Central Park. It’s dark now, and the park is still, but tomorrow is Sunday and the joggers will be back.
She died there quietly, surrounded by her family, at 10:40 that night.
