The appearance of symptoms — a breast lump, blood in the stool, chest pain, and the like — were once the first signs of disease. Today, the search for illness, and with it the possibility of early intervention, is becoming increasingly proactive and moving beyond screening tests such as mammography, colonoscopy, and cholesterol measurement. And in the digital and genomic age, individuals are playing ever-expanding roles in early disease detection, using our smartwatches to monitor our heart rhythms or spitting into tubes and having our DNA sequenced.
At first glance, that seems to make sense. Frequent checks can spot diseases early, and that’s good for long-term health, right?
Too much testing can also lead to harmful care. Virtually all medical tests are imperfect, sometimes failing to detect disease when it is present (false negatives) or “detecting” it when it isn’t (false positives). Basic arithmetic proves that for rare conditions, indicators of disease — called biomarkers — often yield false positives.
Consider the simple scenario of a disease that affects about 1 in 200 people. A new biomarker is developed that is always positive in individuals with the disease but falsely positive 10 percent of the time (a rate that is much better than many biomarkers already in use). As shown in the figure below, even though the test seems to be accurate, the odds that someone with a positive biomarker result actually has the disease is less than 5 percent (1 in 21).
Tests aimed at our complex and ever-changing anatomy and physiology will catch states that may signal disease or may be incidental or fleeting. Because of this, some people are “overdiagnosed” and treated for conditions they do not have. Convincing estimates suggest that between 22 percent and 31 percent of women who are treated for breast cancer that was diagnosed by a mammogram (a radiographic biomarker) don’t have cancer. Instead, the mammogram detected something that would have not caused a problem or would have faded away. Treatment of overdiagnoses and false positives in breast cancer costs the U.S. health system $4 billion a year.
The push for more screening and diagnostic testing can be traced in part to economics or, put more bluntly, revenue.
As we recently argued in the Journal of the American Medical Association, “A culture of advocacy and promotion for aggressive testing may arise when a biomarker or its sequelae yield financial benefit to drug and device manufacturers, procedure-based specialties, hospitals, laboratory testing services, or is increasingly requested by patients.” Economic pressures, which may not be apparent to doctors or patients, can boost the use of biomarkers, a phenomenon we call “biomarkup.”
You often hear that it’s best to ask your doctor if you’re uncertain about a test. Unfortunately, that’s not a perfect remedy. The well-intentioned desire by health care professionals to detect disease and intervene early tends to favor the acceptance of medical tests. Many physicians grossly underestimate the proportion of false positives a biomarker will produce, and tend to believe that any positive biomarker result indicates a high probability of disease. Ironically, tests with higher false positive rates can appear to be the most attractive, since they provide the most opportunities to intervene.
More than a century ago, in his play “The Doctor’s Dilemma,” George Bernard Shaw recognized that medicine favors an illness-driven economy; the practice of medicine is not only a vocation, but is also a business.
Fast forward to today, when biomarker testing naturally drives revenue for drug and device manufacturers, physicians who perform procedures, and diagnostic and laboratory testing services. As a result, biomarkers with higher detection rates (even if the detection is a false positive) are often favored.
Biomarkup may happen unintentionally or intentionally. When the American College of Cardiology and American Heart Association issued new treatment guidelines for cholesterol-lowering statins in 2013, the new guidelines amounted to an increase in 12.8 million adults eligible for statin therapy. In this case, the biomarker was the widely known blood test for cholesterol and other lipids. Or take the case of Purdue Pharma, the maker of OxyContin. It established and trademarked the slogan “Pain: the Fifth Vital Sign,” provided funding to the American Pain Society for improved treatment of pain, and successfully promoted more pain treatment. That, along with Purdue’s aggressive marketing of OxyContin to doctors, helped fuel the opioid epidemic. In this case, the biomarker was patient-reported pain.
Evidence linking even well-established biomarkers to meaningful disease states is often flawed or lacking. Prostate-specific antigen (PSA), for example, a once almost universally used screening test for early detection of prostate cancer, is now known to produce a large number of false positives, some of which lead to unnecessary, life-changing prostate surgeries. The U.S. Preventive Services Task Force now recommends that men who do not express a preference for screening should not have PSA tests. Rates of PSA testing for prostate cancer screening are falling.
Evidence can take years to catch up with technology. Until recently, genetic testing was usually done for just one gene at a time, or sometimes a small handful of them. Today, sequencing techniques routinely examine thousands of genes. Ubiquitous wearables that monitor individuals at home will continuously feed data into algorithms that alert and drive medical decisions.
As our health system faces an explosion in testing through genomics and digital measurement, it must address the potential for either intentional or unintentional overtesting and overtreatment.
Active surveillance for an important cause of excess care and costs — biomarkup — along with educational programs to help physicians better interpret biomarker testing will protect patients and support the transformation toward a safer, more cost effective, and less wasteful health care system.
Kenneth D. Mandl, M.D. is director of the Computational Health Informatics Program at Boston Children’s Hospital and professor of pediatrics and biomedical informatics at Harvard Medical School. Arjun K. Manrai, Ph.D., is in the Computational Health Informatics Program and assistant professor of pediatrics at Harvard Medical School.