We are healthier, but we are increasingly being told we are sick. We are labeled with diagnoses that may not mean anything to our health. People used to go to the doctor when they were sick, and diagnoses were based on symptoms. Today diagnoses are increasingly made on the basis of detected abnormalities in people who have no symptoms and might never have developed them. Overdiagnosis constitutes one of the biggest problems in modern medicine. Welch explains why and calls for a new paradigm to correct the problem.
Where to draw the line? FDR had a BP of 200/100 at the time of his re-election in 1944 and subsequently died of a stroke with a BP of 300/190. At that time, elevated BP was not commonly recognized as a problem requiring treatment. Then studies showed that the higher the BP, the greater the risk, and now everyone diagnosed with HBP is treated. That has undoubtedly saved many lives; but for someone with only a mild elevation, the risk of heart attacks and strokes is smaller and the risk of complications from treatment becomes less acceptable. So where do you draw the line and start treatment? When the limit of 160 systolic was dropped to 140, the new definition instantly turned 13 million people with “normal” BP into patients with hypertension. Not all of those new patients were better off with treatment. Welch gives the example of an 82 year old man who was treated for mild HBP at a level where the number needed to treat for one person to benefit (NNT) was 20; he passed out from medication side effects and declined further treatment.
Changing the Rules: We’ve changed the diagnostic thresholds for many diseases, so that people who were previously classified as normal are now diagnosed with diabetes, high cholesterol and osteoporosis. Dropping the threshold of fasting blood sugar from 140 to 126 instantly created 1.6 million new diabetics, diabetics who were less likely to develop symptoms and complications and were less likely to benefit from treatment. He tells about one of his patients who was put on blood sugar-lowering medication because of the new rules and passed out while driving and broke his neck because the medication brought his blood sugar too low.
Osteoporosis: here are the numbers for treatment of decreased bone density:
- Winners (treatment saved them from a fracture): 5%
- Treated for naught (had a fracture anyway, despite treatment): 44%
- Losers (treated but never would have had a fracture without treatment): 51%
Seeing too much: New technology allows us to detect abnormalities that would never have caused harm. In people without back pain, over 50% have bulging discs on MRI; 10% of asymptomatic people have gallstones on ultrasound. In patients without symptoms, what’s the value of knowing about these findings? In people with symptoms, such findings may lead to a false diagnosis.
10% of the general population and 7% of people under the age of 50 have findings of stroke on MRI. Whole body CT scanning finds abnormalities in 86% of asymptomatic people. The higher the resolution of your testing method, the more anomalies you will detect; but how many of them are important to know about? How will finding them affect health outcomes?
Prostate cancer: the harder you look, the more you find, and the smaller the cancers you detect, most of which would never have hurt the patient. Welch estimates that for every prostate cancer death avoided by screening, between 30 and 100 patients are harmed by unnecessary treatment.
In breast cancer, for every death prevented by mammography, 2 to 10 women are overdiagnosed and treated unnecessarily, 5 to 15 are diagnosed earlier without any effect on final outcome, 250-500 will have a false alarm and half of these will be biopsied. 999 out of 1000 women do not benefit from mammography. A study in Norway showed that screening resulted in 22% more diagnoses of invasive cancer; apparently some invasive breast cancers in the unscreened group had spontaneously regressed.
Other cancers: In an autopsy study, researchers determined that almost everyone has small thyroid cancers; so many that they could be considered “normal” findings. The US Preventive Services Task Force (USPSTF) recommended against screening for thyroid cancer, since it increases the diagnosis rate without affecting the death rate, and increases morbidity from unnecessary surgery and other treatments.
There is overdiagnosis of melanoma and lung cancer. For colon cancer and cervical cancer there is overdiagnosis of precancerous abnormalities.
The good news: We are learning that many, perhaps most, small cancers either regress or never progress. Spontaneous remissions may be far more common that we ever imagined. In one study, 14% of kidney cancers got smaller without any treatment. So we don’t really need to know if any cancer is present: we need to know if a cancer is present that is likely to progress and harm the patient. And so far we have no way of distinguishing which these are.
Incidentalomas are nodules or other unexpected findings noticed on imaging studies, often in body parts adjacent to the area being studied. About half of virtual colonosopies detect abnormalities outside the colon. More than 99% of the time, these are not cancers and not important to know about; but they lead to anxiety, further studies, surgeries, and complications. Protocols are being developed to follow incidentalomas suggestive of kidney and lung cancers over time rather than immediately pursuing diagnosis.
Routine electronic fetal monitoring has minuscule benefits and results in many more C-sections.
Vascular screenings: The Lifeline company and other commercial ventures offer tests direct to the public, tests that the USPSTF doesn’t recommend and that have not been shown to benefit those screened.
Genetic screening. These tests are not done for symptoms, and do not even detect signs of early disease, but just estimate future risks using inadequate data. Welch reminds us that genetics is not destiny and abnormal genes do not equal disease. The predictive value of these tests is small, and we seldom know what to do about the risk after we identify it. Low risk for a condition doesn’t mean you can’t get it, and everyone is at high risk of something.
A paradigm shift is needed, but it will be difficult to achieve for many reasons:
- It is hard to ignore information.
- Most people believe the more information, the better.
- Accepted wisdom and common sense are hard to overturn.
- Most people are convinced that it is always in people’s interest to detect health problems early, even though the data say otherwise.
- There is a common belief that early detection is cost-effective, even though the data show it actually ends up costing more.
- We find it hard to tolerate uncertainty.
- Commercial interests benefit from screening and overdiagnosis.
- Doctors fear being sued if they omit tests.
- Anecdotes about lives saved are emotionally persuasive.
We are easily impressed by anecdotes from people who believe their lives were saved by early detection; but we don’t hear anecdotes from people who were harmed by a diagnosis of a condition that would never have hurt them, mainly because we have no way of knowing which ones they were. I am a case in point: I had a suspicious mammogram and an excisional biopsy that removed a lobular carcinoma in situ. That is not really a cancer, but more like a risk factor for cancer. Did my surgery remove a part of my breast that would have eventually developed invasive cancer and killed me, or did it uselessly remove a harmless chunk of tissue? Did it save my life or just mutilate me? I will never know.
What’s the solution? Maintaining a healthy skepticism about early diagnosis. Informed consent for screening tests, based on accurate information. Resisting over-simplified hype about the benefits of screening. Putting our efforts into prevention (exercise, smoking cessation, healthy diet, etc.) rather than pursuing early detection. Pursuing health without paying too much attention to it and without developing anxieties about it. Welch argues for not even mentioning incidentalomas on imaging reports, but I think radiologists and lawyers would object to that strategy. He says
Severe abnormalities warrant action because net benefit is likely. But the best strategy for mild ones may be to leave well enough alone, otherwise net harm is likely. In fact, it may be better not to look for them in the first place…An overdiagnosed patient cannot benefit from treatment… [but] can only be harmed.
He doesn’t offer prescriptions. He recognizes that different individuals will assess the risk/benefit ratio differently; based on the same data, some will choose to be screened and some won’t. But they deserve accurate information to base their decisions on, and this book offers a lot of good data and thought-provoking analysis.
I couldn’t help but like this book, since it says many of the same things I have been saying about screening tests , colonoscopy, osteoporosis treatment, PSA tests , not always treating , ultrasound testing , overuse of CT angiograms, genetic testing in general and in specific situations, and the pitfalls of diagnostic tests. It explains complicated concepts like lead-time bias in simple terms and spices the story with patient anecdotes. I found it a bit repetitive but that is probably an asset for driving the message home to a general audience. Both patients and doctors would benefit from reading this book and thinking about the issues it raises.
This article was originally published in the Science-Based Medicine Blog