In an unanticipated coincidence, these two groups, operating independently, reached similar conclusions about commonly utilized screening tests: mammograms and Pap smears. Both groups reviewed the data and concluded that routine use of these screening tests, as currently recommended, may not be warranted.
Much newsprint has been expended since then discussing the political implications of these new recommendations. As this is not a political blog, I will leave that discussion to others.
What I want to talk about is how guidelines (should) influence patient care.
I think a lot of the worry surrounding these guidelines stems from concerns that they will be interpreted by those who pay the bills in our system (meaning the Federal Government, through Medicare reimbursement regulations, and the private insurance industry) as “Actual Rules” rather than “Guidelines,” meaning that if you get a mammogram or a Pap smear but don’t meet the “Guidelines” your service won’t be covered. In the current environment, this may be true, but it shouldn’t be.
The key to my argument is the highlighted words above: “routine use” and “as currently recommended.”
The key to understanding why both groups reviewed the published data and reached similar conclusions is an understanding of the nature of screening tests as well as a bit on biostatistics. My work colleagues who read this will laugh at the idea that I am trying to teach anyone statistics, but that’s what I’m going to do.
To start, the accuracy and usefulness of any medical test can be described by the terms “sensitivity” and “specificity.” Sensitivity refers to how likely the test is to be positive if the condition is present. So a sensitive test will pick up every case. Specificity is the mirror image – if the test is positive, how likely is it that the condition is present. Screening tests are designed to be very sensitive, even if they are not very specific – that way, no cases are missed (very sensitive), but sometimes the test is positive even if the patient does not have the disease (not very specific).
The other statistical consideration is the concept of positive- and negative-predictive value. This means, how likely is a positive test to mean the disease is there, or how likely is a negative test to mean the disease is absent? Two concepts factor into the positive- and negative-predictive values of a test: the sensitivity and specificity AND how common the condition is in the population being tested.
These considerations underlie the new recommendations. Mammograms save lives. No one disputes that. Early detection of breast cancer saves lives. No one disputes that. But mammograms are not very specific, and the positive-predictive value of a positive mammogram is MUCH more if the woman is at risk of developing breast cancer than if the woman is at relatively low risk. Since a woman with a strong family history of breast or ovarian cancer is at higher risk of developing breast cancer in her 40’s than a woman with no such history. Thus, the positive-predictive value of a mammogram in a 40 year old woman is higher if the woman is at higher risk. This is why the Task Force no longer recommended ROUTINE mammograms for women under 50.
This is where the practice of medicine comes in. As the doctor treating a 40 year old woman, it is important to remember that a mammogram will be valuable if the woman is at risk, but far less valuable if the woman is NOT at high risk. So blindly refusing to order a mammogram simply because the patient is 40 makes no sense (and an insurance company refusing to pay for it based solely on age makes equally little sense). A 40 year old woman whose mother had breast cancer should have a mammogram and it should be covered. A 40 year old woman with no relatives who have ever had cancer may not need a mammogram. Determining whether a screening test is needed is a decision for the doctor and the patient to make together.
This isn’t rationing care, this is good medicine.
As Breast Cancer Month Draws to a Close
Breast Cancer Risk & Alcohol
Is the Medical Community Complicit?