Table 3 Assessment of risk of bias of individual studies (using the ROBINS-I- for Nonrandomized studies–of interventions for observational studies).

Kiraga et al.¹²

Serious Risk

Serious Risk

Moderate Risk

Serious Risk

Serious Risk

Low risk

Low Risk

Serious Risk

Explanation

There is no mention of controlling for confounding variables (e.g., baseline psychological states, previous use of the substance, individual differences in response to ayahuasca).

Participants were self-selected (volunteers at ayahuasca retreats), which may introduce selection bias. Those who chose to participate might differ systematically from those who did not.

The amount of ayahuasca taken and its exact composition varied across sites. Although the alkaloid concentrations were measured, there is variability that is not fully accounted for.

The facilitators of the ceremonies varied in their practices, and participants experienced different settings. The research team was not involved in administering the intervention, leading to variations in how the intervention was applied.

There was significant dropout, with only 12 participants completing all assessments. The paper doesn’t describe whether dropouts were systematically different from those who completed the study.

The study used validated and commonly employed tasks to measure outcomes.

All relevant outcomes were reported, and the analysis was consistent with the planned methods.

The study by Kiraga et al.¹² presents a serious overall risk of bias. The primary concerns are the lack of control for confounding variables, the self-selection of participants, deviations in the intervention application, and the high dropout rate with unclear handling of missing data. While the use of validated measures and comprehensive reporting reduce some risks, the significant issues in other domains suggest that the study’s results should be interpreted with caution.

Mason et al.²¹

Moderate Risk

Moderate Risk

Low Risk

Low Risk

Low Risk

Low Risk

Low Risk

Low to Moderate Risk

Explanation

The study involved participants who attended psilocybin retreats, and there were potential confounding variables such as differences in motivations for attending the retreat, and varying levels of education among participants. The study attempted to account for prior psilocybin use.

Participants were volunteers attending specific psilocybin retreats This introduces a potential selection bias as individuals attending such retreats may have specific characteristics (e.g., openness to psychedelics, interest in personal development) that are not representative of the general population.

The intervention (psilocybin ingestion) was clearly defined and consistently administered across participants. The dosage and form (psilocybin-containing truffles in tea form) were standardized, and participants’ intake was recorded.

There were no significant deviations from the intended intervention as participants were required to follow a set procedure during the retreat, including the consumption of psilocybin in a controlled environment.

There was some drop-out by the morning after (55 participants at baseline, 50 the morning after). Although the authors note missing data due to time constraints or drop-out, the reduced sample size could introduce bias in the results.

The study used validated measures for assessing creativity, empathy, and well-being (e.g., Picture Concept Task, Multifaceted Empathy Test, Satisfaction with Life Scale). These tools are widely used and have established reliability and validity. The timing of measurements was consistent across participants.

The study reports on all the pre-specified outcomes (creativity, empathy, and well-being) and follows a transparent analysis plan. No evidence suggests selective reporting of results.

The study by Mason et al. demonstrates low to moderate risk of bias primarily due to confounding factors, selection of participants, and some missing data. While the intervention was well-defined and measured outcomes were reliable, the potential for bias due to the non-randomized nature of the study and the missing data warrants a low to moderate overall risk assessment.