How side effects can improve treatment efficacy: a randomized trial

The prevailing view in modern medicine is that ideal treatments should be free of side effects. Side effects cause discomfort, increase suffering, and may lead to treatment discontinuation. The mere expectation of side effects can increase their likelihood (nocebo effect). This study challenges this view by investigating whether certain side effects could actually improve treatment outcomes. The hypothesis is that even mild side effects can indirectly indicate treatment potency, leading to positive treatment expectations, which are known to underlie substantial non-specific therapeutic effects (placebo effect). Research on active placebos (agents with noticeable effects but not on primary symptoms) supports this, showing larger placebo effects compared to inert placebos. This also resonates with observations of general practitioners prescribing more impure (active) placebos. The question of side effects' influence on treatment outcome is crucial for randomized clinical trials, as most compare active interventions with inert placebos. If side effects boost treatment outcome, trials might overestimate the benefits of active treatments with identifiable side effects. The study used a controlled experimental placebo paradigm and fMRI to investigate how side effects influence treatment efficacy at psychological and neural levels in a large sample of healthy volunteers.

Existing literature highlights the detrimental effects of side effects on treatment adherence and overall patient experience. The nocebo effect, where negative expectations worsen outcomes, is well-documented. However, research into the potential positive effects of side effects on treatment efficacy is limited. Studies on active placebos suggest a link between noticeable effects (simulating side effects) and enhanced placebo responses. This aligns with observations showing a preference among general practitioners for impure placebos. This study builds upon this limited research to explore the mechanisms by which side effects may contribute to enhanced treatment outcomes.

This preregistered trial (DRKS00026648) involved 104 healthy participants, with 77 included in the final analysis (42 in the expectancy group, 35 in the no-expectancy group). Participants were informed the study investigated the neural processes associated with fentanyl nasal spray and its potential side effects (burning sensation). They completed questionnaires assessing their belief that side effects indicate a more potent treatment. In Session 1, participants received three nasal sprays (sham, active placebo with capsaicin, inert placebo) while believing all could contain fentanyl. Pain stimuli were applied at different intensities (VAS 40, 60, 70), and pain ratings were recorded. In Session 2 (fMRI), the procedure was repeated. Before Session 2, participants were randomly assigned to expectancy (continued belief in fentanyl) or no-expectancy (debriefed about the lack of fentanyl) groups. fMRI data was analyzed using SPM12, focusing on the rostral ACC, DLPFC, insula, S2, and PAG. A PPI analysis investigated rACC-PAG coupling. Behavioral data were analyzed using SPSS 27, including repeated measures ANOVAs and mediation analysis.

Session 1 showed participants experienced more side effects after active placebo (p<0.001). Pain ratings were lower after active placebo compared to inert placebo (p=0.002), demonstrating that side effects caused more pronounced pain relief. Most participants (94.8%) believed the active placebo contained fentanyl, unlike the inert placebo (16.9%). Moderated mediation analysis revealed that side effects indirectly influenced pain relief by influencing treatment expectations, dependent on beliefs about the relationship between side effects and treatment potency. Session 2 showed the expectancy group continued to have lower pain ratings after active placebo compared to inert placebo (p=0.04), while the no-expectancy group did not. Side effects continued to be experienced after active placebo (p<0.001). fMRI results revealed a modulation of the rACC (T=4.5, p=0.003), showing a reduced BOLD signal during active placebo in the expectancy group. Increased rACC-PAG coupling (T=4.37, p<0.001) was observed during active placebo in the expectancy group, indicating stronger descending pain modulation. A follow-up session 7±1 days later showed that active placebo continued to lead to greater pain relief, although the group difference was no longer significant, suggesting the long-term effects of side effects and expectations.

The study demonstrates that side effects enhance pain relief by amplifying treatment expectations. These effects are mediated by beliefs about the relationship between side effects and treatment effectiveness. The fMRI results show that the observed pain relief involves activation of the descending pain modulatory system. The lack of DLPFC modulation, despite its involvement in other studies, aligns with a recent meta-analysis showing large individual variability. Conditioned pain modulation is deemed unlikely to explain the findings due to the low capsaicin dose and the observed interaction in Session 2. The findings have clinical implications suggesting strategies to maximize positive treatment expectations. This could involve strategically framing side effects or even modifying drug formulations. The study also highlights the importance of considering side effects in clinical trials as they may influence placebo effects, potentially leading to overestimation of treatment efficacy. Active placebos in the control group could address this confound.

This study shows that mild side effects can improve treatment outcomes by amplifying treatment expectations through the descending pain modulatory system. This highlights the importance of considering the role of side effects in both clinical practice and the design of clinical trials. Future research should explore how to optimally communicate about side effects to maximize their beneficial effects and minimize nocebo effects. Further research should also investigate the generalizability of these findings to other clinical conditions and treatments.

The study used a healthy volunteer sample, limiting generalizability to clinical populations. The placebo paradigm used, while controlling for pharmacological effects, might not fully reflect real-world clinical settings. The relatively small sample size could also have limited the statistical power of some analyses. Finally, the specific mechanism by which expectations are formed and affect pain processing may warrant further investigation.