Tuesday, October 31, 2006
Claudia Henschke fails evidence-based medicine 101
She is a proponent of CT-scans for early lung cancer detection, and doesn't get that randomized-controlled studies are the only standard. The rules can't be rewritten for her cause:"I don’t get what the resistance is," Dr. Henschke said.To her, it is a matter of simple logic: the earlier cancer is found, the better the odds of a cure. CT finds lung cancer early. So why not use it? . . .
. . . It may be reasonable to insist on randomized controlled trials for treatments, but not for diagnostic tests, Dr. Henschke says. Her views are unconventional, to say the least. Some researchers say she is trying to rewrite the rules of science.Dr. Barnett Kramer, associate director for Disease Prevention at the National Institutes of Health, took vigorous exception to her assertions, insisting that randomized controlled trials are the only way to find out for sure whether a test or a treatment really works.
Posted by Kevin |
link
| Comments (6)
| 
Comments:
There was a recent article that showed CT scans were not effective at improving outcomes for lung cancer patients. I think there was correlation between spiral CT and improved outcomes though.
# posted by 
: 10:03 PM
: 10:03 PM
"the earlier cancer is found, the better the odds of a cure."
Its debatable whether this belief actaully holds in a number of neoplasms
Its debatable whether this belief actaully holds in a number of neoplasms
# posted by : 12:21 PM
: 12:21 PM
Kevin:
As a 9 year lung cancer survivor as a result of CT screening, I am truly amazed at your attitude. Why is it that people like you think that all of the rules for other cancers don't apply to lung cancer. If the ACS and the medical communication believe "early detection saves lives" why doesn't that apply to lung cancer. It applies for every other one: breast, rectal, prostrate, ovarion, pancreatic, etc. Please enlighten me as to why I should be dead.
- Survivor
As a 9 year lung cancer survivor as a result of CT screening, I am truly amazed at your attitude. Why is it that people like you think that all of the rules for other cancers don't apply to lung cancer. If the ACS and the medical communication believe "early detection saves lives" why doesn't that apply to lung cancer. It applies for every other one: breast, rectal, prostrate, ovarion, pancreatic, etc. Please enlighten me as to why I should be dead.
- Survivor
# posted by : 8:55 AM
: 8:55 AM
Deliciously, Dr. Henschke's research was by and large underwritten by Liggett, one of the world's largest cigarette manufacturers. See the recent Times article.
# posted by : 3:03 PM
: 3:03 PM
Dear Survivor:
I am glad to hear that your battle with cancer has ended in your favor. However, consider what you are proposing.
Every high risk individual (having smoked an average of 1 pack a day for 30 years or more or the equivalent) should get screened by spiral CT (SCT). The spiral CT will detect nodules down to 1/2 cm in diameter. Thats alot of people going to get thoracotomies and a certain given fraction of those people will die of the procedure. The question is what fraction of this group do you represent, and what fraction of the group have cancers that are too advanced to treat, and what fraction have non-cancerous nodules or
indolent (slow growing) cancers that may never progress.
The last point is the key to the whole debate--is there such a thing as indolent lung cancer? Many clinicians will say "no" categorically, and "all lung cancers are aggressive". But this statement is made before our detection tool (chest X-ray and sputem cytology...think big fat crayon) has been replaced with spiral CT (think scanning electron microscope). The point is that noone knows until enough evidence has been collected.
Give "Should I been screened for cancer...probably not" by Gilbert Welch, M.D., a read. Very compelling. Three points:
(1) length bias: suppose you have a population of sticks floating down the river. You wish to know the average size of the sticks, so you begin pulling sticks out as they pass. You compute the sample mean. It turns out that your answer is noticably larger than the true population mean. Why? Because you're more likely to pick longer sticks!!!
Whats this have to do with cancer screening? Think about it.
(2) Lead time bias:
Lets try to understand "early detection saves lives" a little more precisely.
Imagine two identical copies of a hypothetical person (same biology same disease history same everything), named Joe and Moe. Joe gets detected by SCT at age x. Moe gets detected by a less sensitive means (either normal chest x-ray or just walks in with an achey lung), at age y, y > x. Both men are treated, but both die of lung cancer..the fist at age T_1 and the second at age T_2. Now, the fairest comparison of the benefit due to SCT is to compare T_2 - y (Moe's survival past Moe's detection) to T_1 - y (Joe's survival past the time he would have been detected in the absence of SCT screening.
The Henschke study essentially compared
T_2 - y to T_1 - x, which is exagerating the benefit due to screening by the 'lead time'.
(3) All that glitters is not a Gold standard:
The iron clad answer to the question "do I have cancer" is presumably, the results from the pathology lab. Imagine a graph of "reliability" versus "Sensitivity of screening instrument"
As the instrument becomes more and more sensitive, the tumors and biopsy speciments sent off to the pathology lab become more and more ambiguous and difficult to read, so that the reliability of the findings from the pathology lab deteriorate. This has been verified by two (well designed) studies (Welch's book).
Sorry for the long diatribe, but maybe this answers your question.
By the way, the mantra "early detection saves lives" doesn't apply to every other organ as you mention. It has only been demonstrated via randomized controlled trial in the case of breast cancer (the HIP study) with efficacious results for women over 50...but even that conclusion is still controversial because of possible flaws in the study design (not quite on the level of the flaws discussed above).
Post a Comment
I am glad to hear that your battle with cancer has ended in your favor. However, consider what you are proposing.
Every high risk individual (having smoked an average of 1 pack a day for 30 years or more or the equivalent) should get screened by spiral CT (SCT). The spiral CT will detect nodules down to 1/2 cm in diameter. Thats alot of people going to get thoracotomies and a certain given fraction of those people will die of the procedure. The question is what fraction of this group do you represent, and what fraction of the group have cancers that are too advanced to treat, and what fraction have non-cancerous nodules or
indolent (slow growing) cancers that may never progress.
The last point is the key to the whole debate--is there such a thing as indolent lung cancer? Many clinicians will say "no" categorically, and "all lung cancers are aggressive". But this statement is made before our detection tool (chest X-ray and sputem cytology...think big fat crayon) has been replaced with spiral CT (think scanning electron microscope). The point is that noone knows until enough evidence has been collected.
Give "Should I been screened for cancer...probably not" by Gilbert Welch, M.D., a read. Very compelling. Three points:
(1) length bias: suppose you have a population of sticks floating down the river. You wish to know the average size of the sticks, so you begin pulling sticks out as they pass. You compute the sample mean. It turns out that your answer is noticably larger than the true population mean. Why? Because you're more likely to pick longer sticks!!!
Whats this have to do with cancer screening? Think about it.
(2) Lead time bias:
Lets try to understand "early detection saves lives" a little more precisely.
Imagine two identical copies of a hypothetical person (same biology same disease history same everything), named Joe and Moe. Joe gets detected by SCT at age x. Moe gets detected by a less sensitive means (either normal chest x-ray or just walks in with an achey lung), at age y, y > x. Both men are treated, but both die of lung cancer..the fist at age T_1 and the second at age T_2. Now, the fairest comparison of the benefit due to SCT is to compare T_2 - y (Moe's survival past Moe's detection) to T_1 - y (Joe's survival past the time he would have been detected in the absence of SCT screening.
The Henschke study essentially compared
T_2 - y to T_1 - x, which is exagerating the benefit due to screening by the 'lead time'.
(3) All that glitters is not a Gold standard:
The iron clad answer to the question "do I have cancer" is presumably, the results from the pathology lab. Imagine a graph of "reliability" versus "Sensitivity of screening instrument"
As the instrument becomes more and more sensitive, the tumors and biopsy speciments sent off to the pathology lab become more and more ambiguous and difficult to read, so that the reliability of the findings from the pathology lab deteriorate. This has been verified by two (well designed) studies (Welch's book).
Sorry for the long diatribe, but maybe this answers your question.
By the way, the mantra "early detection saves lives" doesn't apply to every other organ as you mention. It has only been demonstrated via randomized controlled trial in the case of breast cancer (the HIP study) with efficacious results for women over 50...but even that conclusion is still controversial because of possible flaws in the study design (not quite on the level of the flaws discussed above).
# posted by : 11:23 AM
: 11:23 AM
