That's a bit of an oversimplification, of course, because some physicians really do understand statistics, but an article just published in the Annals of Internal Medicine looked at internal medicine doctors' ability to interpret whether tests to screen for cancer actually helped save lives and found that a majority of us do not understand the numbers that explain why some cancer screening tests may be of no benefit.
Lately scientific organizations have released some pretty controversial recommendations about screening for several common forms of cancer. Initially, in 2009, the US Preventive Services Task Force released a recommendation that mammograms not be performed routinely on women under the age of 50 and that evidence was insufficient to recommend mammograms over the age of 75. This was based on lots of data that showed that in these groups of women, the risks of mammogram screening, including unnecessary treatment, were higher than the benefits, except in specific cases. In October of 2011 that same organization recommended against using PSA (prostate specific antigen blood test) screening to identify men with prostate cancer. An overwhelming amount of data over a long period of time shows that repeated testing of PSA in men without symptoms of prostate cancer does more harm than good. In this month's issue of the Annals of Internal Medicine, the American College of Physicians has released its recommendation that screening for colon cancer with colonoscopy or fecal testing be stopped after the age of 75.
There have been passionate responses to all of these recommendations, protesting that they are short sighted and motivated by the desire to save money at the expense of vulnerable populations. How could a test that is minimally dangerous be bad, if it might detect something really horrible like cancer at a time when it can be cured? There are various ways.
The first is something called lead time bias. If a person has a cancer that will lead to their death in, say, 2018, and they discover it early via screening, say, in 2012, rather than when they develop symptoms in 2017, they will live more years after discovering they have cancer by being screened, even though they don't actually live longer. It will look like treatment and screening made a difference, when what really happened was that they worried about it for longer and spent more time in doctors' offices and with treatments that didn't end up helping before their inevitable death.
Another is overdiagnosis bias. It's likely that all of us at this moment have some cancer cells lurking around in our organs, but our immune systems are killing them off before they can set up shop. If a screening test is so good that it identifies the presence of these cancer cells even in those of us who will never have problems, it will look like more of us survive after diagnosis. What will really happen is that a bunch of people with conditions that are of no consequence go around thinking they have cancer and maybe even pursuing toxic treatments for it.
It is also true that people identified by screening at an early and curable stage with a cancer that would have otherwise killed them are in fact benefited by the screening test, though others are not, and that identifying these people by screening everybody (or a large proportion of everybody) is so astronomically expensive and time consuming that other more important means of prevention such as vaccination, nutrition and other aspects of wellness are neglected, leading to significantly more misery than if the test were not routinely used.
Researchers from the Harding Center for Risk Literacy of the Max Planck Institute for Human Development, supported by a grant from the National Cancer Institute presented over 300 internists involved in primary care with data about cancer screening tests and found that they did not understand that when a screening test identifies more people with a cancer that is at an early stage, and possibly not even likely to cause harm, more of these people will survive solely because they were going to survive anyway. The way the questions were worded didn't make the right answer obvious, but that was the point. Data like this is not obvious to patients, and when their doctors also don't understand it, excess testing and treatment will happen.
Lately scientific organizations have released some pretty controversial recommendations about screening for several common forms of cancer. Initially, in 2009, the US Preventive Services Task Force released a recommendation that mammograms not be performed routinely on women under the age of 50 and that evidence was insufficient to recommend mammograms over the age of 75. This was based on lots of data that showed that in these groups of women, the risks of mammogram screening, including unnecessary treatment, were higher than the benefits, except in specific cases. In October of 2011 that same organization recommended against using PSA (prostate specific antigen blood test) screening to identify men with prostate cancer. An overwhelming amount of data over a long period of time shows that repeated testing of PSA in men without symptoms of prostate cancer does more harm than good. In this month's issue of the Annals of Internal Medicine, the American College of Physicians has released its recommendation that screening for colon cancer with colonoscopy or fecal testing be stopped after the age of 75.
There have been passionate responses to all of these recommendations, protesting that they are short sighted and motivated by the desire to save money at the expense of vulnerable populations. How could a test that is minimally dangerous be bad, if it might detect something really horrible like cancer at a time when it can be cured? There are various ways.
The first is something called lead time bias. If a person has a cancer that will lead to their death in, say, 2018, and they discover it early via screening, say, in 2012, rather than when they develop symptoms in 2017, they will live more years after discovering they have cancer by being screened, even though they don't actually live longer. It will look like treatment and screening made a difference, when what really happened was that they worried about it for longer and spent more time in doctors' offices and with treatments that didn't end up helping before their inevitable death.
Another is overdiagnosis bias. It's likely that all of us at this moment have some cancer cells lurking around in our organs, but our immune systems are killing them off before they can set up shop. If a screening test is so good that it identifies the presence of these cancer cells even in those of us who will never have problems, it will look like more of us survive after diagnosis. What will really happen is that a bunch of people with conditions that are of no consequence go around thinking they have cancer and maybe even pursuing toxic treatments for it.
It is also true that people identified by screening at an early and curable stage with a cancer that would have otherwise killed them are in fact benefited by the screening test, though others are not, and that identifying these people by screening everybody (or a large proportion of everybody) is so astronomically expensive and time consuming that other more important means of prevention such as vaccination, nutrition and other aspects of wellness are neglected, leading to significantly more misery than if the test were not routinely used.
Researchers from the Harding Center for Risk Literacy of the Max Planck Institute for Human Development, supported by a grant from the National Cancer Institute presented over 300 internists involved in primary care with data about cancer screening tests and found that they did not understand that when a screening test identifies more people with a cancer that is at an early stage, and possibly not even likely to cause harm, more of these people will survive solely because they were going to survive anyway. The way the questions were worded didn't make the right answer obvious, but that was the point. Data like this is not obvious to patients, and when their doctors also don't understand it, excess testing and treatment will happen.
Comments
my comment though is about humans not understanding statistics. not just doctors.
I'm in the middle of "thinking fast and slow" by Dr Daniel Kahneman. provides PROFOUND insight into why "your post" happens (and will continue to happen until our brains evolve). might be one of the best books i've ever read (up there with Phenomenology of Spirit by Hegel for me)
The editorial on the article, also in Annals, discusses "innumeracy", analogous to illiteracy, and mentions that it is a very common condition. I've noticed that the vast majority of us (docs) can't really wrap our heads around what the numbers in studies mean, even if we take time to look at them. The statistics that we see are an attempt to make the vastness of data into a pattern that tells an understandable story that can guide our practice, but the statistics themselves are often misleading, even if we understood them. I don't so much mind that we don't understand the numbers, but the fact that we use our misunderstandings to passionately advocate for random procedures isn't good.