Introducing the Behavioral Element to Clinical Trials
We explore a new trial design that takes into consideration the behavioral elements of participants - the 2 x 2 trial design.
According to a new study by Sylvain Chassang, Erik Snowberg, Ben Seymour and Cayley Bowles, of Princeton University, Woodrow Wilson School of Public and International Affairs, double-blind randomized controlled trials (DBRCT) don’t completely consider the behavioral element participants bring with them. The authors say that patients can’t be expected to behave the same exact way and that there may be variations in patients’ lifestyle choices and habit modifications brought about by the anticipation of the treatment. Subsequently, variations in behavior could lead to variations in the effects of medication.
No denying psychology
In DBRCTs for new drugs, participants are randomly assigned into a treatment group that receives the new drug or a control group that is given a placebo. In most trials, patients are informed that they have a 50/50 chance of falling into either group. Such a design avoids the potential effects the participant’s and clinician’s biases could have on the results of the study. Yet, this trial design has long been considered the 'gold standard' of medical research.
Standard blind trials can fail to account for the full value added of a treatment when there are interaction effects between treatment and behavior.
Although the design of standard blind trials does take account of potential effects from the interaction between treatment and behavior, which can impact outcomes, it fails to accurately measure the full interaction effect. As Professor Chassang explains, “Standard blind trials can fail to account for the full value added of a treatment when there are interaction effects between treatment and behavior.” Chassang’s study indicates that with a diminished understanding of the full impact of the treatment-behavior interaction, the effectiveness of the new drug can be underestimated.
“No one can deny that psychology has a role to play,” says Mike Rea, CEO of IDEA Pharma. “Knowing that you’re on a newer agent or a more expensive one, for example, may not be tested; although we know that it has an effect on outcomes. Similarly, we know that the physician has a role in outcomes - those who invest more time produce better outcomes.” He adds that it, therefore, makes sense for clinicians to try and quantify the effect of psychology on the participant’s behavior and, ultimately, on treatment outcomes.
Two-by-two trial design
Quantifying the influence of behavioral elements on outcomes was the goal of Chassang and his team. They developed a study that would take the standard blind trial a step further. Indeed, they propose a new design called the ‘two-by-two trial design,’ which has the ability to measure interactions between treatment and behavior. According to Chassang, this proposed design can evaluate the performance of treatments for behavior-related conditions such as smoking, substance abuse and some mental health problems.
The study constituted a meta-analysis of data from six clinical trials about two different antidepressants - paroxetine and tricyclic (TC) imipramine. The methodology of the two-by-two trial design doesn’t use the usual 50/50 probability of receiving treatment, which only randomizes who will and will not receive treatment. Instead, it randomizes the probability of receiving treatment. The participants can be placed either in a high-probability or a low-probability treatment group. For example, the high probability group will have a 70% chance of receiving treatment while those in the low-probability group will only have a 30% chance.
Participants who felt more confident in receiving the drug, meaning those who were randomly put under the high-probability group, were more likely to change their behavior in ways that made the new drug perform better.
The randomization of the probability of receiving treatment made a significant change in the final analysis. Chassang’s team was able to show how the two-by-two blind trials can be used to randomize both treatment and behavior. It was observed that participants behaved differently based on whether they were put under the high or low probability group. Participants who felt more confident in receiving the drug, meaning those who were randomly put under the high-probability group, were more likely to change their behavior in ways that made the new drug perform better. This group was also found to be less likely to drop out of the trials. Presenting different probabilities to the participants, therefore, has an impact on the degree to which they change their behavior. According to Chassang, participants who perceive that they have low odds of treatment have low incentive to change their behavior, but those who perceive they have high odds of treatment also have high incentive to modify their behavior.
This highlights the limitation of DBRCTs. According to second author, Professor and Researcher at the Division of the Humanities and Social Sciences, California Institute of Technology in the US, Erik Snowberg, this finding may indicate that the typical 50% chance of treatment adopted by standard trial designs may not be enough to motivate participants to make changes in their behavior that could augment the effectiveness of the drug. The results may appear as if the new drug is completely ineffective, when in fact, a higher probability used in the study would have encouraged the patient to behave differently, and the study would have revealed the added value of behavior on the drug.
Isolating pure treatment, behavior and interaction effects
According to Snowberg, by randomizing both the treatment and the behavior, the two-by-two trial design can better identify and isolate the two components of treatment effect, namely, the treatment’s pure effect and the treatment-behavior interaction effect. Additionally, the authors were able to evaluate the pure effect of behavior.
The meta-analysis of the paroxetine and TC imipramine studies revealed different results, and it was possible to demonstrate how the three types of effects could make a significant impact on the total treatment effect. The paroxetine data showed that the drug’s performance increased with higher probability of receiving treatment. It had no significant pure effect of treatment or pure effect of behavior. However, it demonstrated that the interaction between treatment and behavior contributed to the effectiveness of the drug. The TC imipramine, on the other hand, was found to have a pure effect of treatment and a pure effect of behavior. However, it didn’t reveal an interaction effect. These findings reveal that three types of effects can be isolated and measured as to their impact on the overall outcome. They also show that the degree of interactions between treatments and behaviors can vary across drugs.
“Interaction effects should be incorporated into meta-analyses of existing trial data. Our results show that this can lead to different conclusions than when interaction effects aren’t considered,” suggests Chassang. However, Rea doesn’t entirely agree. “Even though they suggest that paroxetine complemented the ‘knowledge effect’ and imipramine didn’t, once you’re outside the realm of antidepressants, it’s hard to see those same effects being consistent,” he explains.
Implications to pharma’s existing trial design
“Any study that aims to produce outcomes that are more ‘real world’ is to be welcomed,” says Rea. However, despite the potentially different conclusion that studies can have if treatment-behavior interactions are considered, pharma executives shouldn’t be too quick to make changes in the current design of their standard blind trials. Pharma still needs to follow what regulation dictates.
“In development stages, although the placebo effect is real, and interestingly rising over time in some diseases, regulators will need to see drug-for-drug differences with as few confounders as possible,” says Rea, explaining how standard clinical blind trials can provide a clearer picture of which drugs are safer and better when compared to other drugs. “So, while it would be encouraging to have more explanations for the effectiveness of drugs, versus the efficacy, it’s hard to see regulators demanding these kinds of study [two-by-two trials], so by extension, companies will go with delivering what the regulator asks for,” adds Rea.
Under the existing regulation, pharma executives can still improve upon their current trial design to better account for behavior. “I believe trials should aim to mirror real world as much as possible,” shares Rea. Pharma companies can achieve this by making trials as standardized as possible by employing inclusion criteria, standardized information released to participants, and minimum compliance to compensate for participant dropout. Rea adds that pharma can adopt the ‘hands-off’ approach in tracking patient behaviors more accurately with the use of novel technologies, citing healthcare and medication wearables as a good example. “It’s easy to imagine how those tracking technologies could provide a real picture of what patients do, recognizing they don’t always do what they’re told,” he says.
Taking notice of the significance of placebo effects
Rea admits that the findings of the study emphasize the importance of identifying the separate impacts of treatment and behavior on outcomes. “I do believe pharma has ignored ‘the treatment effect’ or ‘the placebo effect’ too much as ‘pseudo-scientific,’ although it is clearly real and of significant magnitude. Placebo-responsive diseases demand more attention,” he emphasizes.
Chassang’s findings pose a challenge for pharma companies in how they can use the placebo effect to improve upon the positive treatment outcomes in the real world setting. “I see this study having more to teach physicians than pharma companies - ignoring the placebo effect is largely in their hands once the drug is approved; although pharma could absolutely do more to enhance that effect,” says Rea. He adds that taking notice of, and taking action on, the significant contribution of placebo effects on overall medication outcome could possibly take time before they gain any traction in the industry.
Rea points out that Chassang’s study successfully highlights the positive impact of psychology on a patient’s beneficial behavior. According to him, “We know, from the embrace of ‘alternative therapies’ that patients want more than just a script. The real beauty of this study is that it proves that patient belief, in the drug and in their physician, is critical, and real – it matters.”
Since you're here...
... and value our content, you should sign-up to our newsletter. Sign up here