Clinical trials fall short in determining drug safety

At this point in time, the enormity of the opioid epidemic need not be revisited. The root causes of the crisis, however, deserve renewed attention. Discussions of the origins of the epidemic have typically concentrated on prescribing practices of physicians, patient factors and levels of government regulation. Few researchers have asked whether the epidemic was caused by a more subtle but consequential problem: a failure to adequately assess the adverse effects of medications prior to their widespread use.

The fact that clinical trials found that opioids pose a "very small risk of developing addiction, abuse, or other serious side effects" suggests that clinical trials, as they are currently conducted, may not be able to accurately determine the safety of prescription medications.


Prescribing doctors and the FDA rely primarily on clinical trials to determine the safety and efficacy of medications. These trials compare the experience, both positive and negative, of patients randomly assigned to a treatment to a matched placebo. A well-conducted clinical trial is a robust experiment that provides our most unbiased assessment of the efficacy of a new treatment. The design of most clinical trials, however, limits their ability to detect many adverse effects.

For one, most clinical trials do not last long enough to provide accurate data on the adverse effects of a drug. Consider Ritalin, a prescription stimulant used to treat attentional disorders in children. Children with ADHD routinely take Ritalin for years, but the average study lasted for only 75 days. Only three trials have lasted longer than six months. Short-term trials can detect acute adverse effects like nausea or headache, but they do not provide evidence that taking these drugs long-term is safe.

A second concern is that trial data does not tell us if patients can safely stop a medication. Most trials collect all of their data while patients are still taking the drug. Many psychiatric medications have common and severe adverse effects that only become evident upon discontinuation. The risk for dependence and abuse associated with opioids could have been better detected if clinical trials tested whether individuals could safely discontinue the medication.

Another commonly raised issue with clinical trials it that their participants seldom resemble the patients most likely to take the medication. Patients with multiple comorbid conditions, those who are taking other medications, who do not perfectly meet diagnostic criteria, who are less likely to respond to a medication, or who might abuse a medication, are usually excluded from clinical trials. These are the patients most likely to suffer the adverse effects of a medication and the least likely to experience a drug's benefits. At a time when 15 percent of the adult population is taking five or more prescription medications, excluding these patients limits our ability to define the risks of medications.

Clinical trials generally enroll dozens to hundreds of patients and thus cannot detect adverse events that occur once in every thousand or more people. Post-marketing surveillance is our tool to recognize events that only become detectable after tens or hundreds of thousands of people have taken a medication. The success for post-marketing analysis cannot be understated. These data have led to the discovery of femoral shaft fractures with osteoporosis drugs or sudden cardiac death with certain antibiotics. However, post-marketing surveillance is ill-suited to detect relatively small increases in common adverse events, particularly among individuals who are elderly or already sick.

When post-marketing surveillance does succeed in identifying a major adverse effect, it does so after a large number of people have already been exposed to a drug. Seen from this perspective, these cases are less a success of post-marketing surveillance and more a failure of the drug approval process. The opioid epidemic serves as a reminder of this.

Recognizing both that novel drugs need to reach patients in a timely and efficient manner and that we cannot expect clinical trials to enroll tens of thousands of patients, how can we learn more about the adverse effects of drugs prior to their approval? Requiring that studies of new drugs be designed to collect data on adverse events even after their efficacy has been proven would be an important step. Collecting data during a period of drug withdrawal would also be welcome, along with pragmatic trials, ones that enroll patients like those who are likely to use the medication. Knowing as much as possible about the adverse effects, as well as the benefits, of medications before we use them can only benefit patients, their doctors, and the companies that make them.

Kevin Kennedy ( is a student at the Pritzker School of Medicine at the University of Chicago; Dr. Adam Cifu ( is a professor of medicine at The University of Chicago.