Giving a blood sample, sliding into a scanner, even lying on the stainless steel bed of an x-ray machine — A diagnostic test, while not anyone’s idea of fun, is not that big a deal, right?
Those of us in the medical biz are well aware of the risk of harm associated with surgery or other procedures and the expected and unexpected adverse effects of medicines. We tend to forget, though, that diagnostic tests can also cause harm. It’s not the test itself, that is risky, it’s the results — and what we do with them — that can do more harm than good.
How can this be?
- Labeling — Patients told they have hypertension — an asymptomatic disease — log more sick days the year after receiving the diagnosis.
- False positives — Lung CT scanning of smokers will be positive in over half of them; 97.5% of these results are false positives.
- Overdiagnosis with resulting overtreatment — Up to 30% of women diagnosed with breast cancer — and treated — would have lived just as long had then not be diagnosed and treated.
In a previous blog we discussed the possible of harm of genetic testing. But what about other diagnostic tests? How can we use research findings to help us decide when to use them?
The four questions
Before ordering any test, we should ask four questions. Even more important, it’s not until they run the gantlet of all four that we can be sure the test is going to be valuable and not misleading.
Question 1: How accurate is the lab or diagnostic test in ruling in or ruling out the disease in question?
This is where all that stuff learned in medical school about sensitivity and specificity and positive and negative predictive values comes into play. Tests with an unacceptably high or low false negative or false positive rate can be misleading.
Question 2: Has the lab or diagnostic test been shown to lead to a change in the patient’s diagnosis?
Tests that are inconclusive add more mystery than clarity.
Question 3: Will ordering the test lead to a change in therapy for the patient?
A diagnosis is simply a way of determining whether a patient needs treatment. Tests that don’t change treatment decisions result in labeling without benefit.
For example, type 2 diabetes is diagnosed with an A1c of 6.5% or greater, yet treatment with medication is only beneficial when the A1c is >8.5%. Patients with levels from 6.5 to 8.0 are diagnosed, resulting in more anxiety and higher insurance costs, without the opportunity to benefit from treatment. Don’t get me started on the evidence showing minimal, if any benefit, to recommending lifestyle management. In our online course we also talk about the evidence not showing a benefit to the diabetic foot examination.
Question 4: Will using the test, in practice, result in improved patients outcomes?
That’s what we’re trying to do, right? We have many examples of testing that doesn’t improve outcomes: Thyroid cancer screening doesn’t decrease mortality. Routine skin cancer surveillance has resulted in a 15-fold increase in melanoma diagnosis, but no change in mortality.
The goal of any test is to lead to a diagnosis, which determines therapy, which, ideally, improves patient outcomes. To know whether a new test or screening process really works we need to do a randomized controlled trial to ask the essential question: “If I order this test, will it increase my patient’s likelihood of living longer or better?”