Just don't do it

Leonard Feldman, MD, FACP, takes a look at non-evidence-based medicine.

Most conference speakers eagerly collect evidence on the topics of their lectures.

But Lenny Feldman, MD, FACP, focused on the absence of evidence for the 2013 edition of the session “Non Evidence-Based Medicine: Things We Do for No Reason,” held at Hospital Medicine 2013 in National Harbor, Md., this May.

There are actually some reasons that hospitalists do things they may not need to, including economic incentives and lack of information, acknowledged Dr. Feldman, who is an assistant professor of medicine and pediatrics at Johns Hopkins University in Baltimore. “We have a limited understanding of diagnostic and nondrug therapeutic costs,” he said.

For example, most hospitalists are probably unaware of the high cost of ordering a carotid Doppler on patients with syncope. According to an analysis of high-risk syncope patients published in Archives of Internal Medicine in July 2009, $20,000 was spent for each carotid ultrasound that actually affected treatment.

Yet even when expensive tests on syncope patients find something, it's often incidental. “When you have high-risk patients and you start looking for vascular disease, you find vascular disease,” said Dr. Feldman. “You are much more likely to find disease than an explanation for the patient's syncope.”

In comparison, a postural blood pressure measurement is the test that most commonly affects syncope management, and it costs almost nothing. “Why are we not doing the cheap thing that can get done right away, that actually affects management?” he asked.

Although ordering the carotid test has become a matter of common practice, it's not recommended by any guidelines. “I don't know how we pass this down from generation to generation, but we need to stop,” said Dr. Feldman.

Expert recommendations are responsible for another of the habits on Dr. Feldman's hit list. Keeping potassium levels between 4 and 5 mEq/L has been the traditional advice, based on old studies that measured cardiac patients' potassium levels on hospital admission.

However, the current evidence, including a 2012 study of more than 30,000 heart attack patients published in the Journal of the American Medical Association, shows that lower potassium levels may actually be associated with better survival. In the study, U-shaped curves for both mortality and ventricular fibrillation or cardiac arrest bottomed out around 4 mEq/L.

“It kind of looks like the 3.5 to 4.5 range is what we should all be targeting,” said Dr. Feldman. “We've been teaching this for decades without much evidence.”

There's also not much evidence about using prophylactic anti-epileptics to prevent seizures in patients with brain tumors. “The neurosurgeons are telling us to do this all the time, but there are very few studies on the subject,” said Dr. Feldman.

In one of the few studies that does exist, a randomized trial in the April 2013 Journal of Neurosurgery, patients who took phenytoin after tumor surgery actually had slightly more seizures, and a lot more adverse events. “The number needed to harm is lower than the number needed to treat in the first 30 days,” said Dr. Feldman. A 2008 Cochrane review on the subject found just one eligible study with neutral findings.

“We just don't have enough evidence,” said Dr. Feldman. “I'm not going to add on yet another medication when I don't know it is going to help.”

Harm is also a more likely result than help with the final item on Dr. Feldman's list—acid-suppressive medication for hospitalized, non-intensive-care patients. In a study of almost 80,000 admissions, published in Archives of Internal Medicine in 2011, proton-pump inhibitors and H2-blockers reduced gastrointestinal (GI) bleeds, but less than they increased pneumonia and Clostridium difficile.

“While you're preventing one GI bleed, you've caused seven or eight patients to have pneumonia and one or two of them to have C. diff,” said Dr. Feldman. It's possible that certain subsets of patients, such as those on steroids, might see more benefit, but that's not been proven yet. “We just don't have that data. We need to find that subset before we reflexively just treat,” he said.

Focusing on a subset of patients is also what the evidence supports in hypercoagulability testing, according to John Bartholomew, MD, section head of vascular medicine at the Cleveland Clinic, who spoke on this subject immediately after Dr. Feldman.

Patients who have a venous thromboembolism are often given an array of tests, many of which are unlikely to affect their treatment but can cause negative consequences. “I've seen many people who have false-positive results,” said Dr. Bartholomew. “If you don't have insurance, it's $2,700 for a complete panel [at my hospital].”

Three tests comprise a typical initial thrombophilia evaluation in the acute setting—factor V Leiden, prothrombin gene mutation, antiphospholipid antibodies—but only one of those is very likely to affect treatment. “The only patients probably that testing really affects outcomes are those with antiphospholipid syndrome,” said Dr. Bartholomew.

There are other reasons patients may undergo hypercoagulability testing, including to alleviate the fear of cancer or provide genetic information to other family members, but experts need to identify and publicize what testing is really supported by the evidence, Dr. Bartholomew said.

“We are trying to minimize these tests. We haven't done a really good job yet, but we're going to come up with algorithms because many of these tests are not absolutely necessary,” he concluded.