Eija Kalso and Kaarlo Simojoki object to one of the recent IMMPACT recommendations, which they imply is unethical (Kalso, E. A. & Simojoki, K. Measuring abuse liability—is the risk worth taking? Nat. Rev. Neurol. 10, 131–133; 2014).1 Our specific recommendation was “the study sample should be chosen to represent the range of possible patients who will be prescribed the drug in practice” in clinical trials intended to assess potential abuse liability (see Box 1 for full recommendations; 2b is the most relevant).2 Kalso and Simojoki argue that the efficacy of long-term opioid use is not established and, consequently, inclusion of high-risk patients in clinical trials is unethical. The IMMPACT recommendations did not attempt to define the intended populations for specific analgesics with abuse potential. We agree that there are patients for whom the risks of an analgesic clearly outweigh the anticipated benefits, and that these patients should not receive the analgesic in clinical practice or in a clinical trial. It goes without saying that any clinical trial must be built on clinical equipoise, and that all study participants must be given proper protection and treatment, including appropriate psychiatric care and multidisciplinary care, when indicated.

Kalso and Simojoki seem to argue that individuals with more than minimal risk of aberrant drug-taking behaviour should be excluded from clinical trials in which one of the objectives is to assess abuse potential. The science of predicting which patients will do poorly with medications with abuse potential is not well developed. As Kalso and Simojoki acknowledge, studies have identified a number of factors associated with aberrant drug-taking behaviours, including alcoholism, smoking, psychiatric disorders, and family history of addiction. Clinical trials involving analgesics with abuse potential have often excluded patients with such risk factors, although many such patients are treated in practice. Systematically excluding all patients with, for example, psychiatric comorbidity or a family history of addiction from analgesic clinical trials in which assessment of abuse potential is an objective would make it difficult to adequately evaluate the risks of such treatments. In our view, it is essential that a balanced estimate of the risks and benefits of a drug be obtained before the drug is released onto the market. Clinical trials that exclude all but the lowest-risk patients may underestimate the true risk of the drug to patients in the community.

Kalso and Simojoki argue that epidemiological surveillance studies can clarify whether a drug carries increased risk in certain populations that receive the drug, and these studies can then inform post-marketing regulation. Although epidemiological studies certainly have a role in monitoring of abuse liability following initial approval, they require many thousands of patients to be exposed to a drug, it is often unclear whether a true signal is present, and causality cannot be determined from any identified associations.

Within a clinical trial, all participants provide informed consent and are closely monitored. Exposing large numbers of patients in the community to analgesics without informing them about the risks and expecting that epidemiological studies will eventually provide clarity could result in unforeseen harm, as the history of opioid prescribing during the past 20 years suggests.