This is the eighth installment of our look at the increasingly high placebo response that is plaguing clinical trials in analgesia and psychiatry. Catch up on the rest of the series here.
We’re now just over halfway through our Placebo Problem series. So far, we’ve examined the details of the rising placebo response, the mechanisms by which the placebo response occurs, how it has been studied, and whether a fundamental assumption governing the interpretation of randomized controlled trials (RCTs) holds true. In the next several posts, we’ll be discussing specific solutions to the rising placebo response, as well as describing the many factors that can influence the magnitude of the response. The final few posts in the series will touch on the history and ethics of the use of placebos, as well as look at the placebo’s opposite, the nocebo effect.
Excluding placebo responders
One potential option to counteract the rising placebo response in psychiatry and analgesia trials, and improve drug-placebo differences (also known as assay sensitivity) in all clinical trials, is to identify and subsequently exclude high placebo responders. A large body of literature has identified factors that are associated with a smaller or greater magnitude of placebo response, but the vast majority of those identified are trial related and not due to any particular traits of an individual. We’ll be taking a closer look at these factors in two weeks. The 1950s saw a flurry of attempts to characterize the personality of the placebo responder, with little success at identifying personality traits that held up to replication. The few attempts that have been made since have also come up empty-handed.
Another possibility that has garnered widespread interest is to perform a placebo run-in prior to the start of the trial and subsequently exclude any high responders. While this has been tried a number of times in psychiatry drug trials, several meta-analyses have found that excluding subjects who show a high response during the placebo run-in does not lower response in the placebo group or increase the effect sizes in the drug groups. A potential explanation for the latter finding is that placebo non-responders are also drug non-responders.
Conversely, another proposed placebo reduction approach is to strengthen drug effects. This could be achieved by identifying and selecting for drug responders, although this approach is subject to the same limitation of the exclusion of placebo responders noted above. It also raises concerns of introducing bias by only testing a drug in a specific subpopulation of individuals, which could lead to regulatory approval for only a limited indication.
One strategy that is becoming increasingly popular is the exclusion of participants with high variability in baseline pain reports, which has been associated with a high placebo but not drug response in several clinical trials. A related strategy is to use sensory training to improve the accuracy of pain reporting. We’ll discuss the recent evidence in favor of this approach next week.
Novel study designs
An alternative placebo reduction strategy to the placebo run-in, and one that has received much attention, is the sequential parallel comparison design (SPCD). First proposed in 2003, SPCD is a two-phase design: in the first phase, subjects are randomized to receive placebo or active treatment; in the second phase, placebo non-responders from the first phase are then randomized to receive either the active drug or placebo. Data from both stages are pooled together in the final analysis. SPCD has been used successfully in several antidepressant trials to reduce the placebo response and, in most cases, improve detection of drug-placebo differences. Its use in clinical trials is growing, and a number of variants of the SPCD design have been developed.
Another approach to reducing placebo responses that has been proposed more recently is the post-hoc band-pass filtering method. Using this technique, data from all trial sites displaying an abnormally high or low placebo response is excluded from the final analysis. This approach could potentially remove the effects of site-specific variables that cause usual placebo responses. In a re-analysis of data from a failed antidepressant trial, applying such a filter reduced the placebo response from 31 percent to 21 percent, but also cut the sample size by more than half. Applying an even more stringent filter further reduced the placebo rate to 9 percent, and eliminated more than 80 percent of subjects. Unfortunately, the large sample size this design necessitates seems to be prohibitively large.
Another option is to eliminate the placebo group entirely, and instead compare novel drugs to other approved ones or other standard of care options. However, meta-analyses in both depression and schizophrenia have shown that this approach actually heightens the placebo effect, presumably because of increased patient expectancy since all participants receive a drug.
Perhaps down the line, research on the placebome—the genes associated with stronger or weaker placebo responses (see our fifth post in this series)—will lead to knowledge of specific genetic variants, or a group of variants, that predict a high placebo response. Excluding individuals with these variants from clinical trials could reduce the magnitude of the placebo response.
Greater placebo responses are found in trials that utilize patient-reported outcomes rather than physician-reported ones. Therefore, another placebo reduction strategy is to move away from patient-reported outcomes. In psychiatry and analgesia trials, this would necessitate the identification of biomarkers of mood symptoms and pain, both active areas of research.
Next week we’ll continue our discussion of placebo reduction strategies, turning to new training methodologies for both trial staff as well as subjects that can be used to limit the placebo response.