I think this is the money statement - "Early detection must improve outcomes when compared to detecting the disease when it would naturally present clinically". How do you design such a trial? Does anyone know if any of the blood biopsy companies such as GRAIL have this endpoint in their clinical trials?
The practical issue is: how many health care providers currently practice shared decision making / provide people with the information required to make a decision in line with what is important to them?
I don’t have the impression that there are many doctors that do things like: use a cardiovascular or fracture risk calculator to discuss management options, explains the benefits and harms of screening (in numbers or even at all) or such. In Germany, the number one pay out of pocket screening test is still ovarian cancer screening, which has been shown to be harmful in large randomized trials.
Without any incentives, tools and training, the notion of informed / shared decision making remains but a good idea. Decision aids are well-studied, could clearly help and can be designed in a fast and frugal encounter style. They won’t answer all questions, but would be a step forward. Then again, in a study on the evaluation of a cancer screening decision aid I was involved in, the doctors criticized that it included information on harms, because that could put patients off — which doesn’t suggest a real interest in empowering them. Guidelines are also not made to support SDM with some exceptions such as the Canadian Tools for Practice.
Overall, medicine still seems pretty paternalistic to me and the system doesn’t encourage otherwise. Someone (who, unlike me, is a physician) once told me: patients stand at the center of health care — and in doing so, they are always in the way.
One central issue I struggle with in my daily clinical practice is how to do this “shared decision making”. How many of your patients Dr Cifu can understand these numbers and probability estimates. Even how many clinicians can make sense of these numbers in terms of real life applicability. Would love to know how u tackle this.
I’ve been a PCP for more than 15 years and conducted similar conversations to Dr. Cifu’s hypothetical conversation. The most common response is, “So what should I do, doc?”
We think that factual information would help the conversation, and I absolutely encourage us to reach for it, but my experience showed me no amount of data will beat someone’s personal story and preset values.
Which has led my pendulum to swing toward Dr. Nabhan’s philosophy.
Another thing I say is, “Imagine you get X cancer. Will you be comfortable having not been screened because there was not enough evidence to convince you of the benefit? That screening wouldn’t have made a difference? If yes, you should feel confident in your decision to not get screened. If you will have wished you got screened, then you should get screened.”
Imagine yourself in the moment. Most people like to think they follow the data, but when you tell them data means they are a statistic, they realize no, they think of themselves as a unique person with unique circumstances and not a statistic.
I’m curious if Dr. Cifu has written about the most traditional and widely accepted screening exam there is that is debatable in evidence for benefit: the yearly physical exam. Should this be held to the same standard? What about physical exam maneuvers themselves when one is asymptomatic? What about review of system in a history— though you are looking for clues to further shape your thinking of the chief complaint at hand, can it also be considered a verbal screening with high likelihood for false positives?
My Substack profile says "I'm a grumpy old mansplainer." With regard to medicine I'm basically a lay person, although I do have a scientific background and a career in computer programming with a specialty in predictive analytics, and some good results in machine learning contests involving medical data.
I lean towards Dr. Adam Cifu's position. Since I don't like being "prodded and poked by strangers", and then suffering anxiety over what that prodding and poking might reveal, I've been avoiding most of the screening routinely advised for men of my age. So I liked the article "Why don't we invite all men for a prostate cancer test?" that Dr. Cifu links to. That article proposes a long-term study to gather data about the outcomes of prostate cancer screening: "we need to monitor the men in the trial for several years.". That means several years before the results of the study will have an impact on clinical care protocols. During that time, the screening and treatment options may themselves evolve, further complicating the situation.
Note: my dad was never diagnosed with prostate cancer, but as somewhat of a hypochondriac he did suffer for a long time from "prostate cancer anxiety". He worried about some symptoms "down there", had various tests, and even, as I recall, a biopsy, all with inconclusive results. He finally got tired of obsessing about it, gave up being tested, and eventually died of liver cancer at the age of 79.
"What are the ideal screening test characteristics?"
It's a great question. It really depends on the speed of the condition/disease and whether you can stand for false positive or false negative. Let's assume a relatively slow-moving etiology and an expensive treatment. Here, a false negative is less costly all around than a false positive. So in this case, a test that has a very high Negative Predictive Value very early in the care pathway is ideal - when prevalence is relatively low and you can reliably exclude healthy patients. Then a test with high PPV is best (so peg Specificity at 90% when determining cut-offs for the test if applicable) and check sensitivity - as long as it stays over, say, 60% you should be OK. But all in all, there's no one size fits all for these things!
I ceased cervical screening went I read the evidence that 75 percent of women who screened regularly as per Australian guidelines had an unnecessary colposcopy... That was a far too high rate of iatrogenic harm for me, versus a 1 percent chance it was going to lengthen my life... That evidence is probably twenty years old now, but nobody's come out with any updated figures.
My favorite screening test requires only a stethoscope---screen children before organized athletic competition in order to rule out aortic stenosis which carries a significant risk for sudden death if not detected. It is cheap and easy and the only further test to diagnose the cause of a murmur would be a cheap and harmless echocardiogram. All other screening tests commonly recommended show no real benefit from early detection but can certainly cause harm.
I nominate newborn metabolic screening as the best screening program in the U.S. It pretty much checks off all the boxes presented by Dr. Cifu and is inexpensive.
It is reasonable to demand from a screening test that it leads to better outcomes for those who are screened. The strongest claim I have ever seen for any cancer screen is “this test reduces the risk that you will die of Cancer X.” Numbers needed to screen are in the high hundreds. Have we ever seen a trial of a test that reduces all-cause mortality? (I have not; please correct me if I missed it.) The initial data from the NLST suggested that this was (surprisingly) the case, but that claim was not supported with longer term follow up.
Until we have such a trial, we should not use terms like “life-saving” for any screening test. Unfortunately, that language is prominent in the article by Dr. Matthew Hobbs that is linked from your current dialogue. (https://prostatecanceruk.org/about-us/news-and-views/2022/04/why-don-t-we-invite-all-men-for-a-prostate-cancer-test) The article claims that "Clinical trials have shown us that using the PSA test to identify prostate cancer can save lives.” Its graphic claims that a project to invite men to get a PSA dropped prostate cancer deaths from 88 to 71, and uncritically equates this to 17 lives saved.
Your dialogue offers the Hobbs paper as the best current attempt to quantitate all potential effects of screening. Hobbs indeed attempts to describe the harms from screening, but overstates the benefits.
Hmm. I guess cancer in the family has sent me to regular PSA tests, digital exams, and colonoscopies. My father had prostate cancer, but beat it with radiation, then died a few years later at 89 of leukemia. My brother, age 79, lost his 20-year battle with prostate cancer. Sister died of ovarian cancer at 73. (Yes, I'm the youngest at 77.) In 2022 my hemoglobin was going down (I take quarter aspirin daily), so colonoscopy found a stage 2a tumor. (Previous colonoscopy 8 years earlier was clean.) Next morning, 15 inches of colon were removed on the "right" side, the "good" side, as the docs kept saying—no bag. Negative ever since and knocking on a lot of wood! Yet, my 89-year old brother is still going strong, apparently with no medical issues!!
This remains a tricky topic for me, as it’s not in my field and I don’t have a strong grasp of the data. Furthermore, rather than “intervening” on behalf of pts who have sought you out to address symptoms, screening involves taking otherwise asymptomatic folk who are simply minding their own business, and “medicalizing” them.
The first point of uncertainty for me is to distinguish the screening test as a (population based) strategy vs as a point of personal decision making. This was highlighted in NordICC, which seemed to show that colonoscopy as a “strategy” fails, but as a “procedure” does have some benefit. Had I been a trial participant, “I” would have gone through with the procedure….so knowing that about myself….should I look more at the ITT result or the per-protocol result?
On a daily basis, we apply the “average” results from clinical trials onto individual patients. That extrapolation seems to involve a far higher degree of uncertainty and variance when it comes to to “screening” tests and procedures.
" my patients will only agree to screening tests that are ideal for them."
This assumes that patients can indeed follow the math and the implications of the math clearly which is in my experience not always the case...moreover can I explain the math without leaning in to my own biases?
The flaw in Dr Nabhan’s argument surfaced when he wrote “ I believe that screening select people who meet high risk criteria is a valuable and acceptable practice” the problem with that statement is that screening is aimed at an entire population not just a select few who meet high risk criteria. For example, take mammograms. All women are encouraged, dare I say required through subtle and not so subtle pressure to undergo a painful and not at all benign procedure beginning early in their lives and continuing yearly for the entirety of it. There is data to support that mammograms do not decrease overall mortality. Many lesions are found that had they been ignored would have led to a far better quality of life. I think of the Medical model similar to the Titanic with everyone grasping to hold onto it and salvage what they can. Yes it may help some, but it also injures many. I’ve heard Dr. Prasad comment on this many times. Innovations are slow to turn that ship around because there is so much money invested in the status quo. For example, there are new alternatives to ionizing radiation mammograms that are not offered to women or covered by insurance. There are new procedures for dealing with early stage breast cancer that also are not offered to women or covered by insurance.
I doubt that I will see a change in my lifetime, but Hope springs eternal.
Dr N says we should screen select people (high risk)... so unless I misunderstand his position, that is still a departure from current standards... and closer to the "screening deniers" than he is to the mainstream/SoC
I think this is the money statement - "Early detection must improve outcomes when compared to detecting the disease when it would naturally present clinically". How do you design such a trial? Does anyone know if any of the blood biopsy companies such as GRAIL have this endpoint in their clinical trials?
I hope my comment is not too harsh.
The practical issue is: how many health care providers currently practice shared decision making / provide people with the information required to make a decision in line with what is important to them?
I don’t have the impression that there are many doctors that do things like: use a cardiovascular or fracture risk calculator to discuss management options, explains the benefits and harms of screening (in numbers or even at all) or such. In Germany, the number one pay out of pocket screening test is still ovarian cancer screening, which has been shown to be harmful in large randomized trials.
Without any incentives, tools and training, the notion of informed / shared decision making remains but a good idea. Decision aids are well-studied, could clearly help and can be designed in a fast and frugal encounter style. They won’t answer all questions, but would be a step forward. Then again, in a study on the evaluation of a cancer screening decision aid I was involved in, the doctors criticized that it included information on harms, because that could put patients off — which doesn’t suggest a real interest in empowering them. Guidelines are also not made to support SDM with some exceptions such as the Canadian Tools for Practice.
Overall, medicine still seems pretty paternalistic to me and the system doesn’t encourage otherwise. Someone (who, unlike me, is a physician) once told me: patients stand at the center of health care — and in doing so, they are always in the way.
One central issue I struggle with in my daily clinical practice is how to do this “shared decision making”. How many of your patients Dr Cifu can understand these numbers and probability estimates. Even how many clinicians can make sense of these numbers in terms of real life applicability. Would love to know how u tackle this.
Love this debate.
I’ve been a PCP for more than 15 years and conducted similar conversations to Dr. Cifu’s hypothetical conversation. The most common response is, “So what should I do, doc?”
We think that factual information would help the conversation, and I absolutely encourage us to reach for it, but my experience showed me no amount of data will beat someone’s personal story and preset values.
Which has led my pendulum to swing toward Dr. Nabhan’s philosophy.
Another thing I say is, “Imagine you get X cancer. Will you be comfortable having not been screened because there was not enough evidence to convince you of the benefit? That screening wouldn’t have made a difference? If yes, you should feel confident in your decision to not get screened. If you will have wished you got screened, then you should get screened.”
Imagine yourself in the moment. Most people like to think they follow the data, but when you tell them data means they are a statistic, they realize no, they think of themselves as a unique person with unique circumstances and not a statistic.
I’m curious if Dr. Cifu has written about the most traditional and widely accepted screening exam there is that is debatable in evidence for benefit: the yearly physical exam. Should this be held to the same standard? What about physical exam maneuvers themselves when one is asymptomatic? What about review of system in a history— though you are looking for clues to further shape your thinking of the chief complaint at hand, can it also be considered a verbal screening with high likelihood for false positives?
My Substack profile says "I'm a grumpy old mansplainer." With regard to medicine I'm basically a lay person, although I do have a scientific background and a career in computer programming with a specialty in predictive analytics, and some good results in machine learning contests involving medical data.
I lean towards Dr. Adam Cifu's position. Since I don't like being "prodded and poked by strangers", and then suffering anxiety over what that prodding and poking might reveal, I've been avoiding most of the screening routinely advised for men of my age. So I liked the article "Why don't we invite all men for a prostate cancer test?" that Dr. Cifu links to. That article proposes a long-term study to gather data about the outcomes of prostate cancer screening: "we need to monitor the men in the trial for several years.". That means several years before the results of the study will have an impact on clinical care protocols. During that time, the screening and treatment options may themselves evolve, further complicating the situation.
Note: my dad was never diagnosed with prostate cancer, but as somewhat of a hypochondriac he did suffer for a long time from "prostate cancer anxiety". He worried about some symptoms "down there", had various tests, and even, as I recall, a biopsy, all with inconclusive results. He finally got tired of obsessing about it, gave up being tested, and eventually died of liver cancer at the age of 79.
"What are the ideal screening test characteristics?"
It's a great question. It really depends on the speed of the condition/disease and whether you can stand for false positive or false negative. Let's assume a relatively slow-moving etiology and an expensive treatment. Here, a false negative is less costly all around than a false positive. So in this case, a test that has a very high Negative Predictive Value very early in the care pathway is ideal - when prevalence is relatively low and you can reliably exclude healthy patients. Then a test with high PPV is best (so peg Specificity at 90% when determining cut-offs for the test if applicable) and check sensitivity - as long as it stays over, say, 60% you should be OK. But all in all, there's no one size fits all for these things!
I ceased cervical screening went I read the evidence that 75 percent of women who screened regularly as per Australian guidelines had an unnecessary colposcopy... That was a far too high rate of iatrogenic harm for me, versus a 1 percent chance it was going to lengthen my life... That evidence is probably twenty years old now, but nobody's come out with any updated figures.
My favorite screening test requires only a stethoscope---screen children before organized athletic competition in order to rule out aortic stenosis which carries a significant risk for sudden death if not detected. It is cheap and easy and the only further test to diagnose the cause of a murmur would be a cheap and harmless echocardiogram. All other screening tests commonly recommended show no real benefit from early detection but can certainly cause harm.
I nominate newborn metabolic screening as the best screening program in the U.S. It pretty much checks off all the boxes presented by Dr. Cifu and is inexpensive.
It is reasonable to demand from a screening test that it leads to better outcomes for those who are screened. The strongest claim I have ever seen for any cancer screen is “this test reduces the risk that you will die of Cancer X.” Numbers needed to screen are in the high hundreds. Have we ever seen a trial of a test that reduces all-cause mortality? (I have not; please correct me if I missed it.) The initial data from the NLST suggested that this was (surprisingly) the case, but that claim was not supported with longer term follow up.
Until we have such a trial, we should not use terms like “life-saving” for any screening test. Unfortunately, that language is prominent in the article by Dr. Matthew Hobbs that is linked from your current dialogue. (https://prostatecanceruk.org/about-us/news-and-views/2022/04/why-don-t-we-invite-all-men-for-a-prostate-cancer-test) The article claims that "Clinical trials have shown us that using the PSA test to identify prostate cancer can save lives.” Its graphic claims that a project to invite men to get a PSA dropped prostate cancer deaths from 88 to 71, and uncritically equates this to 17 lives saved.
Your dialogue offers the Hobbs paper as the best current attempt to quantitate all potential effects of screening. Hobbs indeed attempts to describe the harms from screening, but overstates the benefits.
Hmm. I guess cancer in the family has sent me to regular PSA tests, digital exams, and colonoscopies. My father had prostate cancer, but beat it with radiation, then died a few years later at 89 of leukemia. My brother, age 79, lost his 20-year battle with prostate cancer. Sister died of ovarian cancer at 73. (Yes, I'm the youngest at 77.) In 2022 my hemoglobin was going down (I take quarter aspirin daily), so colonoscopy found a stage 2a tumor. (Previous colonoscopy 8 years earlier was clean.) Next morning, 15 inches of colon were removed on the "right" side, the "good" side, as the docs kept saying—no bag. Negative ever since and knocking on a lot of wood! Yet, my 89-year old brother is still going strong, apparently with no medical issues!!
All the more reason to change the experimental paradigms to develop biomarkers of aggression, and not of cancer diagnosis.
Something I proposed for development in lung cancers a decade and a half ago, but what is being funded in a manner that is not centrally coherent.
This remains a tricky topic for me, as it’s not in my field and I don’t have a strong grasp of the data. Furthermore, rather than “intervening” on behalf of pts who have sought you out to address symptoms, screening involves taking otherwise asymptomatic folk who are simply minding their own business, and “medicalizing” them.
The first point of uncertainty for me is to distinguish the screening test as a (population based) strategy vs as a point of personal decision making. This was highlighted in NordICC, which seemed to show that colonoscopy as a “strategy” fails, but as a “procedure” does have some benefit. Had I been a trial participant, “I” would have gone through with the procedure….so knowing that about myself….should I look more at the ITT result or the per-protocol result?
On a daily basis, we apply the “average” results from clinical trials onto individual patients. That extrapolation seems to involve a far higher degree of uncertainty and variance when it comes to to “screening” tests and procedures.
" my patients will only agree to screening tests that are ideal for them."
This assumes that patients can indeed follow the math and the implications of the math clearly which is in my experience not always the case...moreover can I explain the math without leaning in to my own biases?
"Can I explain the math without leaning in to my own biases?"
My answer is, unfortunately, probably no.
The flaw in Dr Nabhan’s argument surfaced when he wrote “ I believe that screening select people who meet high risk criteria is a valuable and acceptable practice” the problem with that statement is that screening is aimed at an entire population not just a select few who meet high risk criteria. For example, take mammograms. All women are encouraged, dare I say required through subtle and not so subtle pressure to undergo a painful and not at all benign procedure beginning early in their lives and continuing yearly for the entirety of it. There is data to support that mammograms do not decrease overall mortality. Many lesions are found that had they been ignored would have led to a far better quality of life. I think of the Medical model similar to the Titanic with everyone grasping to hold onto it and salvage what they can. Yes it may help some, but it also injures many. I’ve heard Dr. Prasad comment on this many times. Innovations are slow to turn that ship around because there is so much money invested in the status quo. For example, there are new alternatives to ionizing radiation mammograms that are not offered to women or covered by insurance. There are new procedures for dealing with early stage breast cancer that also are not offered to women or covered by insurance.
I doubt that I will see a change in my lifetime, but Hope springs eternal.
You only have to watch the campaign during “pinktober”. It’s not just physicians it is the media as well as all of the breast cancer organizations.
I believe those "pink" campaigns and related organisations are funded by pharmaceutical companies.
Dr N says we should screen select people (high risk)... so unless I misunderstand his position, that is still a departure from current standards... and closer to the "screening deniers" than he is to the mainstream/SoC