Daniel J Morgan associate professor, Shannon Brownlee writer, Aaron L Leppin postdoctoral research fellow, Nancy Kressin professor, Sanket S Dhruva research fellow, Les Levin professor et al
Morgan D J, Brownlee S, Leppin A L, Kressin N, Dhruva S S, Levin L et al.
Setting a research agenda for medical overuse
BMJ 2015; 351 :h4534
doi:10.1136/bmj.h4534
A new definition of overdiagnosis, overuse, overtreatment compares the probabilities of alternative treatments with or without more diagnosis
The 2015 definitions of overdiagnosis or overuse are based on outdated concepts about health, disease, diagnosis and treatment. “Overdiagnosis is the diagnosis of “disease” that will never cause symptoms or death during a patient’s lifetime.(1) It “is a side effect of screening for early forms of disease.” “Overdiagnosis includes overtreatment, a shift in diagnosis criteria or threshold, under the cover of better sensitivity, to classify people with mild problems or at low risk as sick.” (2)
These definitions are impractical and unmeasurable. One cannot follow a patient until death, perhaps 30 years later, and determine that the patient had no symptoms ever and no death due to failure to do a diagnostic test 30 years ago. It is impossible, on one individual, to determine whether some overdiagnosis will improve wellbeing or increase life expectancy or not.
The definitions imply that overdiagnosis should not be done because no treatment is effective. However, “something” useful and valuable to the patient can always be done.
These definitions assume that a state of good health and healthy people exist, and there is a clear separation between healthy and diseased. These are outdated concepts. Nobody is free of any disease. We all have abnormalities, often unknown, that can be improved. Appropriate testing would find them. It is likely that each person has at least 300 gene abnormalities that cause severe suboptimal function of enzymes, and millions of DNA changes that alter proteins in unknown ways. The concept of a “healthy” or “optimal function” gene set (for the entire genome) is tricky because each person has different DNA. There are genes optimal for survival in different environments or under different conditions, but we don’t know them.
Currently, a state of good health (“healthy”) is biomarkers within ranges that predict desirable outcomes, such as longevity or earning millions in sports (a new definition is needed). Disease means biomarkers are within ranges that predict undesirable outcomes. The undesirable outcomes are highly personal, but government needs to agree on a common set of “disorders” that deserve taxpayer payments or public health.
Reference or “healthy” or “normal” ranges are poorly defined. We could define biomarkers as factors that deserve government payment when abnormal, but individuals may have the right to “treat” themselves to achieve different optimal goals. However, definition of health via biomarkers within reference ranges is outdated. Currently physicians use about 100 biomarkers. With modern technology we now have more than 20,000 biomarkers for which reference ranges and diagnostic tests can be created.
Nobody is free of abnormalities. If we now wanted to eliminate testing that would not improve life expectancy, we would face an impossible task because, statistically, some people not treated would likely suffer symptoms or die. No human has optimal function for our desirable outcomes. There are biological functions that can be improved even within the context of “normal” ranges. Not knowing the unknown disorder or the optimal testing to find it, we can order a comprehensive set of tests such as all DNA, all enzymes/protein concentrations and composition, etc. From these tests, something that ought to be treated will likely be found. Thus, overdiagnosis, expanded with more diagnosis, identifies something treatable. One could argue the problem with overdiagnosis is not enough tests. We should do more tests, until something treatable is found. This reductio ad absurdum proof shows that overdiagnosis must be defined on the basis of probabilities of consequences (costs, harms) and alternative treatments that do not require further diagnosis.
I propose that overdiagnosis occurs when individuals have diagnostic tests that do not substantially change the treatment probability that probably leads to the best desirable outcomes. It occurs when the probability that the treatment will be improved is not changed by the likely result of the diagnostic test. Overdiagnosis is the use of diagnostic tests that increase the probability of effective treatment of any condition by a trivial amount (including nothing) compared with the costs or harms. It is a side effect of efforts to screen and detect disease early or better (e.g., with more certainty or more comprehensive description) when the best treatment does not require further testing.
Alternatively, overdiagnosis occurs when a disease has a substantial probability based on clinical presentation, history, etc., but more accurate diagnosis is irrelevant because treatment is not available, not needed, not wanted or there are appropriate alternatives (such as special diet or behavior changes) that are likely to be better and do not require further testing.
These concepts refer to probabilities of events before and after diagnosis or treatment. Overdiagnosis interacts with informed consent, the patient determines the desirable levels for probabilities of outcomes, harms, etc. (alternative choices). It is bounded by costs paid by taxpayers (on government payments, patients can’t choose all possible tests).
We use a statistical/ probabilistic model to decide. We calculate the probabilities that testing will lead to improvements in wellbeing and life expectancy v. probability it will lead to nothing or harm, both with associated costs. We make a determination that testing (overdiagnosis) now causes more costs or harms than the benefits of diagnosis leading to treatment. We rely on probabilities and measurements today, not outcomes 30 years later. As corollary, we should reexamine reference or “healthy” ranges for common biomarkers (e.g., cholesterol, blood pressure, triglycerides, glucose), to determine if there is a best treatment that requires not further tests (e.g., diet).
Ex: In a child first diagnosis with Crohn’s Disease (CD), there is no need for further sophisticated invasive tests like endoscopies or CT scans. Instead, calprotectin in fecal matter is adequate to monitor changes (in addition to a good physical exam). Treatment should start with Exclusive Enteral Nutrition with an elemental diet in practically all cases (it is effective and has minor adverse consequences).
Ex: use of ESR or CRP. In most cases, the actual value is irrelevant. Given the clinical condition, one can predict it is elevated.
1. Overdiagnosis: Towards a Clearer Definition. Emily DeVoto, PhD and Pamela M. Marcus, PhD. http://www.preventingoverdiagnosis.net/2014presentations/Board%2016_Emil....
2. BMJ 2012; 344 doi: http://dx.doi.org/10.1136/bmj.e3502 (2012).
Competing interests: Competing Interests I conduct many activities that influence my opinions (e.g., my research, write books or articles or patents, have websites, design foods, lecture about health and disease, health policies). I describe the types of fats to eat, exercise, stress management, diets, eating less, periodic fasting. I am developing a new theory of disease, including lipid metabolism and the cause of cardiovascular disease. I intend to profit from my intellectual property (IP) (e.g, write articles, web sites, books, patents, lectures, ads, biomarker tests, etc). I wrote a patent to measure fatty acids. I wrote one book on essential and trans fatty acids. I gave talks to corporations, conferences and trade shows, for which I was paid. I may receive compensation from food or other companies. This comment is an extract of books and articles either written or in progress (authorized by the author under the common license CC BY-NC 4.0 requested by BMJ).