Tucking a spreadsheet in among the toiletries in the bathroom cabinet might seem a bit odd, but for 76-year-old Barbara Pines, it is the easiest way to keep track of all the prescription medications, over-the-counter pills and supplements that she and her husband take. The document lists 20 drugs—along with the strength, number of times taken and purpose. “I print this schedule and take it to any new doctor we go to,” she says.

Pines is among the 40 percent of Americans who are 65 years of age or older and take more than five prescription drugs. Although older individuals account for the majority of prescription drug users, they are hardly alone. More than four billion prescriptions were filled at U.S. pharmacies in 2014—an average of nearly 13 per citizen at that time.

The need to take multiple drugs poses a special risk that too often goes unrecognized by doctors and patients: certain combinations of medicines (prescription or otherwise) cause side effects that do not arise when the individual substances are taken alone. Studies published over the past two decades suggest that such “drug interactions” cause more than 30 percent of side effects from medications. Unfortunately, pharmaceutical manufacturers cannot always predict when a new agent will mix badly with other medicines—not to mention supplements or foods—and so unexpected deaths are sometimes the first sign of danger.

Not all side effects are lethal, but the widespread danger from drug interactions is prompting new efforts to prevent people from taking risky combinations. Much of this work depends on finding informative patterns in huge masses of disparate data.

Pills and Pathways

Drug interactions typically occur when the body breaks down, or metabolizes, medicines. Common trouble spots are the intestine, where ingested drugs are released into the bloodstream, and the liver, where most drugs get degraded.

In the liver, breaking down drugs is primarily the task of a family of enzymes called cytochrome P450. In fact, just six of the approximately 50 enzymes in this family digest 90 percent of all known medications. Problems can arise when two drugs require processing by the same cytochrome. If one of the drugs blocks this enzyme's activity, then too little of the second drug will be degraded and too much will remain in the bloodstream. If, on the other hand, the cytochrome gets a boost from the first drug, then the second drug will have a diminished effect because the enzyme will remove it from the body too quickly. Drugs can also bind to one another in the intestinal tract before ever reaching the liver, preventing the needed chemicals from being absorbed.

Prescription drugs are not the only culprits here. Grapefruit juice, for example, inhibits cytochrome P450 3A4, the same enzyme that metabolizes estrogen and many statins prescribed to lower cholesterol, whereas the herbal supplement Saint-John's-wort boosts the activity of this enzyme. The result, in either case: unpredictable variations in the potency of the medications.

Studies show that once more than four drugs are introduced to the body, the potential for adverse reactions increases exponentially. The trick to avoiding unwanted consequences from drug interactions, says Douglas S. Paauw, who teaches internal medicine at the University of Washington, is “knowing when you're stepping into dangerous territory.”

Banking on Data

And therein lies the rub. The U.S. Food and Drug Administration does maintain a record of reported drug side effects and possible interactions through its Adverse Event Reporting System. But the agency does not know of all, or even most, of the complications—or, indeed, whether or not the reported problems are merely chance events. Clinical trials of new drugs usually do not reveal any issues before a drug is approved, because they are relatively short, focus on a single medication and enroll a small number of participants. As a result, to learn of a possible new interaction, the Fda has to rely on prescribing physicians to take the time to announce problems to the Adverse Event Reporting System.

Nigam H. Shah, who teaches biomedical informatics at the Stanford University School of Medicine, hopes to improve the odds of discovery by collecting information about specific online searches performed by consumers on the Internet and by physicians on a pharmaceutical Web site called UpToDate. Using more than 16 million pieces of data—electronic records of diagnoses, prescriptions, clinical notes, and the like—on nearly three million people, Shah and his colleagues recently published a previously unsuspected association between heart attacks and a group of popular heartburn medications sold under such brand names as Prilosec and Prevacid; Shah's computer program calculated a 16 percent increase in heart attacks with these types of drugs, which are prescribed more than 21 million times every year in the U.S. By definition, such a correlation does not prove causation, however, and the drug's label information has not been changed.

Genes and a Bottle

Improved abilities for tapping the wealth of information available in human DNA may one day dramatically enhance the power to predict who will suffer most from drug interactions. “Everybody metabolizes drugs a little differently,” Paauw says. And now advances in computational biology are beginning to link variations in our genes to differences in how our body absorbs, distributes, metabolizes and eliminates specific medications.

At Duke University's Center for Personalized and Precision Medicine, geneticist Susanne Haga is investigating how to use this quickly accruing genetic knowledge to improve safety, starting at the drugstore. In a recent unpublished survey, Haga found that 17 percent of responding pharmacists had offered or used results from genetic tests (which do not need to be prescribed by a physician) within the previous 12 months. For example, many pharmacists now offer such a test to patients filling prescriptions for clopidogrel, a blood thinner, to confirm the absence of gene variants that could interfere with the drug's action.

Haga is not trying to find previously unknown side effects. Instead she wants to make sure that known genetic complications are widely understood and identified as needed. To facilitate genetic testing and analysis by local pharmacists, Haga recently started the Community Pharmacist Pharmacogenetic Network. Still under development, the network's Web site—rxpgx.com—helps pharmacists access and interpret genetic tests. Although no national databank for collecting such information exists as of yet, Haga hopes that the Precision Medicine Initiative, a project led by the National Institutes of Health, will help lay the foundation for such an effort.

Because prescriptions often come in batches and multiple genes can affect a single drug, Haga envisions a future in which we get tested in advance for key genetic variants that affect our body's ability to process different drugs. But these kinds of tests—which would allow individuals and their physicians to obtain the information whenever needed—are costly, which means that for now, one-off tests suited to elicit information about a single, specific prescription are the more viable option.

Interim Steps

Meanwhile the FDA is trying other approaches for identifying potentially dangerous drug interactions before they occur. At the Center for Drug Evaluation and Research, deputy director Shiew-Mei Huang and others are creating computer models that use clinical research data to calculate how one drug will alter the concentration of another when both drugs are metabolized by the same enzyme. Armed with the concentration and the time it takes for the drugs to move through the body, mathematicians can predict how they will interact.

The approach is bearing fruit. The information sheet for the anticancer drug ibrutinib warns that its concentration could increase drastically if taken with erythromycin, a CYP3A inhibitor, and decrease with efavirenz, an HIV drug. These alerts were generated through computer calculations, not by clinical studies of the effects of both drugs taken simultaneously.

The trouble is that many companies and academics are likely to resist Huang's invitation to share the necessary data for a drug in development—such as its metabolic pathway or its most effective dose—to create useful computer models. Such information could include proprietary data, and sharing it might give competitors an edge.

The FDA is also trying new ways to make the drug information packaged with prescription medicines, termed drug “labels,” more useful to prescribers and the public. The aim is to give patients clearer warnings about possible drug interactions and easy-to-understand recommendations about how, for example, a dose should be altered (based on computer modeling) when a second drug is introduced. According to Huang, some of the new labeling changes are already being used for some recently approved drugs.

Even with improved labels, physicians may still be hard-pressed to prevent and diagnose such problems, given the ever expanding roster of medications and supplements. One interim solution may be for doctors and pharmacists to pay particular attention to the risk of interactions when patients take commonly used drugs. Among the estimated 100,000 annual hospitalizations for adverse events among older adults in the U.S., about a third are tied to the anti-blood-clotting agent warfarin. Other anticlotting agents, as well as insulin and other drugs meant to lower blood glucose, also have trouble playing nicely with various medications. If doctors were more vigilant about the potential harm associated with the 10 most commonly prescribed drugs in their practice, many problematic interactions could be avoided, Paauw says.

Consumers can do their part by making sure to tell each physician they see about all the medications, supplements and recreational drugs (such as alcohol or marijuana) they take. When it comes to preventing unwanted interactions between drugs, forewarned is definitely forearmed.