When it comes to breast and prostate cancer, the curves have not followed this nice pattern. Instead, they look much more like the light gray curves in figure 4.2. Incidence of early cancer did rise after the PSA test and mammography were debuted, but they did not return nearly to baseline. They stayed well above the baseline. This would be a good thing if all of those extra cancers that were found translated into a marked decrease in advanced cancers. In fact the rates of advanced cancer have fallen, but only by a very small amount. Putting this together, it means that we are finding many more early cancers than the advanced cancers we are preventing. This means we have overdiagnosis. The surplus of early cancer represents people who would never have known they had cancer had it not been for the test. These people experience all the downsides of a cancer diagnosis—surgery, radiation, chemotherapy, anxiety—with none of the benefits—a longer, healthier life.
The actual graph for breast cancer is shown in figure 4.3. In the years before mammography, for every 100,000 women, doctors diagnosed 112 early cancers and 102 advanced cancers each year. After 30 years of mammography, we find 234 early cancers each year. The rate of diagnosis of advanced cancer has declined to 94 per 100,000. So each year, 132 extra cases of early breast cancer are found (234 − 112 = 132), for a reduction of 8 cases of advanced cancer (102 − 94 = 8).
4.1 How a screening test should work.
4.2 What actually happened with breast and prostate cancer screening.
4.3 Actual data on breast cancer since mammograms became commonly used.
Source: Bleyer A, Welch HG. Effect of three decades of screening mammography on breast-cancer incidence. N Engl J Med. 2012;367:1998–2005. Copyright © 2012, Massachusetts Medical Society. Used with permission from the Massachusetts Medical Society.
In the real world, by screening for breast and prostate cancer, we have found many more early cancers, with a disappointing impact on advanced cancers.
REVERSALS IN SCREENING
Taking these factors into consideration, we can return to medical reversal. Only CT-scan screening for lung cancer currently reaches the highest standard of proof that a cancer-screening test is beneficial—that it actually saves lives (which does not sound like such a high standard). Mammography for women in their forties was once universally recommended, but USPSTF withdrew this recommendation in 2009. This test is still commonly done. Prostate cancer screening was once widely recommended and used. This recommendation was first narrowed to younger age groups before being withdrawn altogether.
These reversals occurred slowly, not with a single study but as advisory groups concluded that the risks and benefits, which once seemed favorable, now seem less so. Seminal trials paved the way for these changes. A 2006 study published in Lancet found that mammograms did not improve breast cancer mortality among women in their forties in the United Kingdom. A study from Canada published in 2013, covering a wider age group, showed similar results. And these studies dealt with breast cancer mortality, not the more important overall mortality. Recent trials in prostate cancer have followed a similar trajectory.
The definitive way to settle the question of the effectiveness of screening tests is to conduct clinical trials large enough to see whether screening for cancer improves overall survival—not breast cancer survival or prostate cancer survival. Such trials would likely require over a million participants. Because such so-called megatrials may be costly, many people think they are improbable. However, the history of medicine has shown us that studies that are important do get done in the long run, despite the challenges. And a better way to think about the price tag for the megatrial is to remember that each year in America, we spend hundreds of billions of dollars on screening tests and their downstream costs. Spending a small percentage of this money to show whether these screening tests are actually helping people is a small price to pay.
THE FUTURE OF SCREENING AND HOW WE, THE SCREENED, SHOULD PROCEED
Over the next decade, we think it is very likely that we will see a continued pullback on recommended screening tests. The benefit of mammograms, for those at average risk for breast cancer, will probably be shown to be negligible. Guidelines for cervical cancer screening will continue in their present direction, recommending more and more time between tests. Medicine’s experience with breast and prostate cancer screening means that future screening tests will have to meet an appropriately high bar before being adopted. We would like to see a time when a test needs to show that it saves actual lives before it is instituted.
In chapter 17, we discuss how to protect yourself from medical interventions that are likely to be reversed. In terms of screening tests, people should never be screened in ways that are more aggressive than the USPSTF recommends. Even for the recommended screening tests, it is important that we have informed conversations with our doctors. People should understand how likely a screening test is to help them personally, what are the potential harms—both physical and emotional—of the test, and, most importantly, what the test has actually been proved to achieve.
5 SYSTEMS FAILURE
BY NOW, WE HOPE you are becoming suspicious of unproven medical interventions. Many treatments that seem like they should work do not. It does not matter whether the treatment is a pill, a procedure, or a screening test—many seemingly reasonable and accepted interventions have failed. Before moving on from the ineffective therapies in which doctors have placed their faith (and on which they have staked their patients’ health), let’s consider one more category: systems interventions. One family’s experience serves as an example of a systems intervention that failed.
THE PEOPLE INSIDE THE SYSTEMS
Joanne Murphy often thinks about her husband. It has only been a year since Carl died from leukemia. He was a 30-year veteran of the local police force, a youth soccer coach, a man who could fix anything, and a fine and loving father. When Carl first fell ill, his daughter was in her final year of college. One of his greatest concerns was that his illness would disrupt her studies.
As expected, the illness and the therapy were difficult. Although the doctors and nurses were incredibly kind to both Carl and Joanne, being in the hospital for weeks on end for treatment was exhausting. Carl tolerated the lousy mattress, the 4:00 a.m. blood draws, and the incessant beeping of machines. He endured painful mouth sores, rashes, and episodes of fever. He even accepted the food and the pathetic scrap of fabric they called a blanket. One thing that did bother Carl, mainly because he could never make sense of it, was that he was placed on contact precautions for vancomycin-resistant enterococcus (VRE).
VRE is a really nasty bacteria, one that is resistant to many antibiotics. As standard practice, during one of his admissions, a nurse swabbed Carl’s skin. That swab revealed that Carl was colonized by VRE. It was not making him sick, just living on him, as countless bacteria do on all of us. Just how and when Carl got this bug, no one could say, nor did it really matter. What did matter was that now the doctors and nurses had to stop at the door to Carl’s hospital room, open a cart, take out a yellow paper gown, put it on over their clothes, and then put on gloves before entering. This was not to protect Carl, but to protect other patients by halting the spread of this dangerous species of bacteria.
The list of things about this “glove and gown protocol” that irritated Carl was long. First of all, although the whole process was brief, lasting only about a minute, he could tell that it frustrated everybody. Once the members of his health-care team were in his room, he noticed that people’s behavior was inconsistent. Some people were very careful about “covering up,” while others would sit on his bed, allowing their uncovered pants to touch his sheets. Occasionally, people would forget the gown altogether—sometimes these people would just pop in to fiddle with a beeping intravenous pump or collect a breakfast tray, but other times they would sit and talk.
Beyond the inconsistency, the gowns and gloves bothered Carl because he was suspicious that they meant that fewer doctors visited his room. One jun
ior physician, a friendly young lady, had been in the habit of coming by to sit and talk before she went home. After the VRE test, she would only stand at the door and say hello. Both of them understood that she wanted to visit but could not stay long, and it probably was not worth the whole ordeal of dressing up.
On one occasion Carl asked a doctor how much those yellow dresses cost. The doctor said a few dollars each. That cannot be right, Carl thought; it is basically a giant yellow paper towel.
Joanne and her daughter actually found Carl’s list of irritations charming. He had always been unfailingly friendly and chatty, as well as a real stickler for details. He was also known to be … thrifty. The things that annoyed him about the gowns were the things his wife and daughter loved about him.
One night, surfing the web, Carl found an article about a study that confirmed his suspicion. Doctors were less likely to visit patients on “contact isolation.” It made perfect sense. Who wants to bother with all that? Carl was actually sort of satisfied to have been proved right, and he figured, “Well, if helps keep other patients safe, I can tolerate it.”
Carl died from his leukemia a little less than a year after he was diagnosed. Subsequently, two articles were published that would have angered him. In 2011 researchers found that similar gown-and-glove precautions did not decrease the transmission of VRE or methicillin-resistant Staphylococcus aureus (MRSA, another dangerous bacteria) in intensive care units (ICUs). This was a study of more than 3,000 patients in 19 ICUs who received universal screening for these bacteria (the same type of screening that identified Carl’s VRE). The second study (an impressive entry into the acronym arms race: the Benefits of Universal Glove and Gown, or BUGG study) found that making all health-care workers wear gowns and gloves also did not reduce transmission of VRE or a combination of VRE and MRSA (the study’s goal).* In short, both studies suggested that all this “gowning and gloving” does not really help decrease the spread of dangerous bacteria. Carl put up with an intervention he did not like, that made him feel a bit like a pariah, which may have led to fewer visits and potentially worse care and likely did not benefit other patients.
WHAT ARE SYSTEMS INTERVENTIONS?
This type of reversal affects something relatively new in medicine: the systems intervention.† More and more these days, we recognize that good medical care requires more than just a good doctor. It takes a skilled team of doctors, nurses, and affiliated health-care professionals to deliver today’s often complex care. Conversely, when things go wrong, it is usually not that a single individual fell short but that the system was designed poorly and an error either was not caught or was actually magnified by other people’s actions. Systems interventions stem from this evolution in health care. We think that we can overcome many flaws in health care if we improve the systems of health-care delivery. Gown-and-glove precautions are an example of a systems intervention. Other examples include checklists for surgical teams, procedural rules, quality measures, public reporting of outcomes, and economic incentives. A systems intervention is a change in how health care is delivered, one that affects how groups of caregivers act or behave in their professional role.
Businesses use systems interventions all the time. They use them to improve workplace productivity, to improve the quality of products, and to produce products at lower costs. Businesses recognize that these interventions are made in the real and messy world of human beings and thus lend themselves to unintended consequences. Sometimes they do not have the desired effect; sometimes they have the opposite effect. What looked good on paper may not work when instituted in the supply chain or on the factory floor. Businesses therefore continually reevaluate their interventions, willing to abandon what does not work—the interventions that generate cost but no benefit.
In contrast to the business world, systems interventions in health care are often adopted without a real plan to study them. They make good intuitive sense and are often backed by indefatigable proponents. After “successful” adoption, these interventions may even come to be mandated at a national level. Given the unusual way that health care is paid for in the United States, these interventions exist in a world not governed by traditional market pressures. Once adopted, therefore, they can be very difficult to modify.
All this would be fine if systems interventions always achieved their desired goals (even with a few unintended consequences). But as was the case with the gown-and-glove intervention, these interventions are frequently shown to be ineffective. They become examples of medical reversal. The problem is compounded because even a well-done study showing ineffectiveness tends not to be the end of a systems intervention. There are often vocal proponents who are not dissuaded and argue that the intervention would have worked if only you did it a little differently or studied it a little differently. In chapter 16 we will see why this argument (the “it would have worked if only …” argument) is flawed, why you must show that something actually works before it is adopted. For now, let us consider lessons from other reversals of systems interventions.
WHEN SYSTEMS INTERVENTIONS BECOME MEDICAL REVERSALS
Multidrug-resistant bacteria are a major problem in nearly every hospital. VRE, MRSA, and similar bacteria can infect vulnerable patients, leading to life-threatening infections or even death. Although gown-and-glove precautions for patients colonized with VRE probably do not work to decrease this problem, medicine should do everything possible to minimize the harm of multidrug-resistant bacteria. We should devise strategies to reduce infection and death from these bacteria. The strategies would probably need to include ones that reduce the overly generous use of broad-spectrum antibiotics. Any promising strategy should be tested in a randomized trial to confirm that it does what is claimed. We err when we implement practices based on a good theory, on the experience of only a few hospitals, or on studies that are not well done—as happened with gown-and-glove precautions. (Despite the studies we mentioned, national guidelines still advise that hospitals employ gown-and-glove precautions.)
Gown-and-glove precautions were adopted largely on the basis of what doctors call single-center, before-and-after studies. Single-center means that the study was performed at a single hospital. The proof from such a study tends to be tenuous because a single center may be idiosyncratic. It may have a unique demographic of patients, or bacteria, or an unusually enthusiastic proponent who manages to change the culture in ways that are not reproduced at other hospitals. (Gowning and gloving may actually work, but only when instituted flawlessly.) There have been some major discrepancies between single-center and multiple-center (multicenter) trials. Later in this chapter we discuss one famous one, regarding the control of blood sugar in intensive care units.
Before-and-after studies are problematic too, and they are the initial data behind most systems interventions. “Before-and-after” means that we measure the outcome we are interested in first before and then after we implement our intervention. You may ask, for instance: does the rate of infections associated with intravenous (IV) catheter insertion decrease after we implement a checklist of best practices that includes steps like washing hands, sterilizing the patients’ skin, and donning full sterile gowns? (This was precisely the question in a recent before-and-after study.) You measure the rate of infection in your hospital before you institute the checklist and then afterward. The problem with this design is twofold. First, your intervention is never the only thing that has changed in your hospital, so it is hard to separate the effect of the intervention from other trends. The other challenge is something called the Hawthorne effect.
The Hawthorne effect is a psychological principle stating that people perform differently (usually better) when they know they are being watched. This effect was named by Henry A. Landsberger, who studied a series of systems initiatives meant to improve productivity at the Hawthorne Works, a factory in Cicero, Illinois, in the 1920s and 1930s. The most famous initiatives involved altering the level of illumination on the factory floor. Analysis su
ggested that although the actual interventions at the Hawthorne Works had little or no effect, the simple act of studying people did. Brightening the light on the factory floor helped productivity during the study, but the improvement declined after the study ended, even though the brightness remained.
When doctors and nurses know you are counting IV catheter infections, they start taking their own prevention measures more seriously; they become more vigilant about keeping intravenous sites and tubes sterile. This is a good thing, but a before-and-after study would conclude that the checklist is leading to the improvement when, in reality, the improvement is just a consequence of keeping count of IV catheter infections. The worry with the Hawthorne effect is what happens when the study ends. You might continue to implement your intervention, with all the associated costs (and potential harms) but none of the benefits. Your workers might go back to their baseline productivity, only on a brighter factory floor.
It is worth pointing out that for this particular example, a checklist for safe placement of IV catheters, the decrease in IV catheter infections was dramatic, and subsequent studies showed the benefit was maintained. While these results are welcome, they do not mean that it was the checklist that led to the reduction in infections. The only way to prove that the checklist was the cause would be to randomize hospitals to use the checklist or to simply receive the education regarding how to place and maintain central lines. Let both groups of hospitals know that we are concerned about the rate of IV catheter infection rates and will be monitoring those rates. Then follow the rates of infection over time. In 2012 a report of such a randomized trial was published. After randomizing hospitals to the checklist or routine care, the authors found that both groups achieved a median of 0 central line infections per 1,000 days at the end of the randomized study period. The mean number of line infections was different, as the authors repeatedly highlight in the paper, and that number favored the checklist, but it is hard to know that this is not due to just a few, stubborn bad-apple hospitals in the control group. In short, we have yet to see a very large, well-done randomized trial confirming that it is truly the checklist that makes the difference.
Ending Medical Reversal Page 7