Too much testing?

Dear colleagues,

A friend of mine saw a doctor for a check-up, which included a bone density test. She was feeling fine, but when her doctor saw the result on the test, he immediately put her on a drug to deal with osteoporosis. Unfortunately, she had severe side effects from the drug, including eye problems that continued for months after she went off the drug. She might have been better off never having her bone density tested.

The same sort of thing can happen to anyone. You feel fine but think, "Maybe it's worth seeing a doctor to check for lurking health problems." You might have a mammogram, a PSA test for prostate cancer, or measurements of blood pressure or cholesterol. There might be problems that can be picked up early and treated before they become more advanced.

This sounds eminently sensible. Furthermore, there are powerful pressures to screen healthy people. The pink ribbon campaign pushes for regular breast cancer screening. Some doctors recommend screening, especially for at-risk groups. Then there are family and friends. Who can resist their advice to check for problems "just in case"?

However, when you have no symptoms, getting checked for disease may not be such a great idea. To learn about the issues, get the book Overdiagnosed: Making People Sick in the Pursuit of Health. The principal author is Gilbert Welch, a US physician and medical researcher who earlier wrote Should I Be Tested for Cancer? Maybe Not and Here's Why. For Overdiagnosed, Welch has teamed up with two co-authors but narrates in first person.

Overdiagnosis is when you are diagnosed with a disease - such as cancer - that won't actually harm you. You are not going to die from it or even ever have any symptoms.

Overdiagnosis is not cost-free. Sometimes screening leads to additional tests, potentially with harmful consequences. If a routine screen picks up thyroid cancer, you might end up with your thyroid removed and on thyroid drugs for the rest of your life. You might have an adverse reaction to an anaesthetic or acquire a dangerous infection in hospital. That's not good if the thyroid cancer was never going to grow.

There's an important distinction here. The screening Welch is concerned about is of healthy people. If you have signs or symptoms - a lump in your breast, difficulty urinating or whatever - Welch says you should definitely see a doctor. He operates entirely within the conventional medical paradigm, seeing the benefits of surgery, chemotherapy and other treatments. What he's concerned about is screening that does more harm than good.

There are several weaknesses in screening. As well as picking up abnormalities that might never cause a problem, leading to unnecessary and harmful interventions, early detection might not make any difference even for diseases that kill you. It might be that early treatment makes no difference to the progression of your cancer - it just means you have more years of treatment and more years of worry.

The study of prostate cancer has begun to put a dent in the push for screening. Long-term studies show that screening for prostate cancer doesn't lead to very much benefit. A few men's lives are saved from the cancer, but for each life saved, dozens or hundreds have unnecessary biopsies and surgeries, with side-effects such as impotence and incontinence - and the occasional death. The ultimate test is whether screening reduces death rates overall (not just disease-specific death rates) and in some cases it doesn't.

You can end up being screened without even wanting it. Let's say you have an earache and see a doctor, who does an MRI and discovers a suspicious lump in your thyroid. Further tests indicate thyroid cancer - and possibly more interventions. You had no symptoms from the cancer, but were inadvertently screened for it. The things discovered this way are called incidentalomas. They are increasingly commonly detected because diagnostic equipment is becoming incredibly sensitive.

Studies of prostates, breasts and other organs show that if you look closely enough, a large proportion of the population has cancer or abnormalities. Under the critical gaze of imaging technologies, hardly anyone is free of irregularities. The thing is, most of these will never lead to any health problems. But detecting them causes a dilemma for doctors. To protect themselves from potential litigation, they may order more tests and interventions - and you may be one of the increasing number of people subject to overdiagnosis.

Welch presents tables for diseases like prostate and breast cancer showing greatly increased numbers of cases. More people are being diagnosed with cancer - but death rates are stable or declining. The difference is mostly overdiagnosis.

Welch analyses the pressures for overdiagnosis, which include vested interests of pharmaceutical and medical supply companies, and doctors' risk of being sued (patients may sue if a doctor misses a cancer but not if a doctor treats a cancer that would have regressed). At least as important as companies and lawyers is true belief: many doctors and patients believe that screening and early detection can only be a benefit. They ignore or dismiss the side-effects of unnecessary treatment and the psychological impact on people who have been diagnosed with syndromes that will never hurt them.

Welch is particularly scathing about genetic screening, providing evidence that its touted benefits are mostly illusory. US companies tout genetic tests as if they are a magical solution to problems, when the reality is much messier.

There's a striking message in all this: sometimes less knowledge is better than more knowledge. You might be better off not finding out everything possible about what's going on in your body.

Overdiagnosed is a capable popularisation. It is not a breezy read, because Welch carefully goes through the key elements of diagnosis and epidemiology, sufficient to explain what is at stake to non-statisticians. Research in the area is described, with the medical references in footnotes. Welch provides anecdotes from his medical practice to illustrate the general points made.

Reading Overdiagnosed stimulated me to think of analogous issues in other domains. It used to be that bank officials only did security checks on customers displaying suspicious behaviours. Now, with digital banking, every customer and transaction is screened. As has been documented in numerous studies, everyone today is caught in the net of digital surveillance while driving, banking or surfing the Internet.

Anti-terrorism has some of the same features. When boarding a plane, every passenger is screened no matter how low the risk. It has been argued that this gives a false sense of security, while diverting resources from other, more productive ways of addressing terrorist threats.

What about the university? One example of screening is running all student assignments through text-matching software such as Turnitin. The principle is the same as health screening. Rather than checking only when there are signs of plagiarism, everyone is tested. The result is more overdiagnosis: students suspected of plagiarism, and accused of it, who are actually innocent. General suspicion and false accusations can have undesirable impacts on students, but these are usually ignored in the focus on detecting transgressors.

Overdiagnosed thus has a general lesson. Check out problems when they show up, to be sure, but be wary of searching for problems when there are no signs they exist.

24 January 2013

Thanks to Trent Brown, Sharon Callaghan, Narelle Campbell and Melissa Raven for valuable comments on drafts.


H. Gilbert Welch, Lisa M. Schwartz and Steven Woloshin, Overdiagnosed: Making People Sick in the Pursuit of Health (Boston: Beacon Press, 2011)

"The public should demand (and participate in) research that doesn't look as hard for cancer, doesn't find as much of it, but does find the ones that matter." (p. 72)

"The problem with overdiagnosis is overtreatment. Mammography leads more women to have lumpectomies, mastectomies, radiation, and chemotherapy. It has led Iona Heath - a physician who is now president of the Royal College of General Practitioners - to 'cheerfully decline' invitations to be screened. She understands the motivation behind early detection. She knows well how terrible breast cancer can be: she has seen women die from it. But she also knows that the ability of mammography to change this fact is rather small. And that there are real harms to the process.

Here's how she summarizes the Cochrane Reviews ' data: 'The evidence review suggests that for every 2000 women invited to screening for 10 years one death from breast cancer will be avoided but that 10 healthy women will be overdiagnosed with cancer. This overdiagnosis is estimated to result in six extra tumorectomies and four extra mastectomies and in 200 women risking significant psychological harm relating to the anxiety triggered by the further investigation of mammographic abnormalities.'

She worries that she has made the decision not to pursue mammography on the basis of information that is not readily available to her patients." (p. 88)

"In 1996, the U.S. Preventive Services Task Force, the independent panel of experts that reviews screening tests, recommended against routine fetal monitoring [because the harms are greater than the benefits]. But according to their current Web site, fetal monitoring has become such an ingrained fixture of medical care that, frankly, the task force seems to have simply given up on trying to dissuade doctors from using it. ... the last time the federal government examined the topic, in 1999, electronic fetal monitoring was used in 83 percent of all U.S. births." (pp. 106-107; emphasis in the original)

"Our ability to read the genome is well ahead of our ability to know whether medical intervention based on such a reading makes sense. Genetic risk is, after all, only one factor among many contributing to disease (remember, it's nature, nurture, and luck). Thus, while intervention might help a few of those destined to get the disease, it will undoubtedly lead others to be treated for diseases they will never develop, or that will never produce symptoms." (p. 132)

"This brings us to a more general principle about how to think about cancer screening. It's tempting to assume that the best test is the one that finds the most cancer. But the goal is not to find more cancer. The goal is to save lives. And the only way to know if the screening is saving lives is by doing a randomized trial. It is easy to forget this and assume that if technology can find more cancer, it will save more lives. Marketers exploit this assumption. Don't fall for it." (p. 137)

Go to

Brian's comments to colleagues

Brian Martin's publications

Brian Martin's website