Recommendations for monitoring and reporting harm in mindfulness for psychosis research

Summary

There is increasing interest in potential harmful effects of mindfulness-based interventions. In relation to psychosis, inconsistency and shortcomings in how harm is monitored and reported are holding back our understanding. We offer eight recommendations to help build a firmer evidence base on potential harm in mindfulness for psychosis.

One of the most significant recent developments in the field of psychotherapy has been the emergence of mindfulness-based interventions (MBIs). In relation to psychosis, MBIs were slow to develop, in large part because of concern that people with psychosis are particularly vulnerable to harmful effects arising from mindfulness practice.¹

Operationalising harm in psychotherapy research

Harm in psychotherapy research has traditionally been defined as the occurrence, following an intervention, of a deterioration in symptoms or functioning that is sustained, is likely to have been caused by the intervention and is more severe than it would have been had the individual not received the intervention.² This definition highlights two immediate complexities in defining harm in psychotherapy. First, a proportion of people with mental health conditions will experience a deterioration in symptoms with or without an intervention, so it is important to establish a baseline rate of deterioration. Randomised controlled trials (RCTs) are particularly useful in this regard, as the inclusion of a control arm provides a valid point of comparison from within the same clinical setting and time frame. Second, a deterioration means more than simply a change in score – to determine the clinical significance of a change in symptoms or functioning it is helpful to know the minimal clinically important difference (MCID) for any given outcome measure.

A second approach to operationalising harm that is commonly applied in RCTs is to report the occurrence of both serious adverse events (SAEs), defined by the World Health Organization as life-threatening or fatal occurrences, and adverse events (AEs), defined as untoward events experienced by study participants. SAEs and AEs may or may not be caused by the intervention. In a review of the available research on harm in mindfulness-based psychotherapies across a range of conditions (but not psychosis), Baer et al report that AEs or SAEs occur in 0–10.6% of research participants, and are no more common in mindfulness-based psychotherapies than in control arms of trials.²

A third approach to operationalising harm has been to consider indirect or proxy indices of harm, such as study drop-out rate – the argument being that drop-out rates would be higher in psychotherapies experienced as unsafe by patients.

Is mindfulness for psychosis harmful?

The literature on mindfulness-based therapies for psychosis is growing fast. An evidence base comprising more than 20 RCTs (see Jensen et al for the most recent meta-analysis³), alongside numerous uncontrolled studies, evidences a range of clinical benefits of mindfulness for psychosis, including reduced negative symptoms and levels of depression, and enhanced psychological quality of life.³ Although these randomised studies do not suggest that mindfulness for psychosis is harmful, a closer look at how harm is operationalised, monitored and reported in mindfulness for psychosis research reveals some critical shortcomings that make it impossible to provide a data-driven answer to the question ‘Is mindfulness for psychosis harmful?’ We offer below eight recommendations (summarised in the Appendix) on how harm should be operationalised, monitored and reported, to strengthen the evidence base on potential harm in mindfulness for psychosis research.

Recommendations for reporting of harm

Study design

First, although uncontrolled research studies were pivotal to the development of mindfulness for psychosis, priority should be given to findings from RCTs when drawing generalisable conclusions about harm. The discipline in RCTs of applying a common set of study inclusion and exclusion criteria within a standardised recruitment protocol, combined with the inclusion of a control arm and random treatment allocation, provides a valid point of comparison from which to understand observed levels of harm in those receiving mindfulness for psychosis. Although uncontrolled studies will always be useful in flagging important potential issues of risk and harm, and in ongoing therapy development, they do not provide a solid basis for drawing firm conclusions about harm in mindfulness for psychosis.

Serious adverse events

Second, there are inconsistencies in reporting of SAEs. All future studies should state explicitly the number of SAEs by study arm; if there were none, then this should be stated explicitly. This is already a recommendation in reporting of trials, but it is not always applied in the MBI for psychosis literature. Also, there is some confusion about what constitutes an SAE in mindfulness for psychosis research. For example, in one study a participant death was not reported as an SAE because it was attributed to natural causes. Crucially, an SAE is defined by the untoward event itself and not by degree of imputed causal connection with study participation. Although it is informative to include indications of likely degree of connection between a life-threatening or fatal occurrence and study participation, it always remains an SAE and should be reported as such.

Hospital admissions

Third, there is lack of clarity around reporting of hospital admission. As with other SAEs, it is sometimes simply not reported on – does this mean that it was robustly monitored and did not occur, or that it was not monitored? Harm in psychotherapy is such an important issue that clarity is essential. Data should minimally include the number of participants in each arm who were admitted to hospital – if there were none in a study arm, then this should be stated explicitly. If admissions were not monitored, then this should be stated explicitly as a study limitation. Where applicable and possible, data on hospital admissions might be supplemented by data on usage of crisis or home treatment teams.

Further inconsistency can arise in trials of MBIs for psychosis that use the hospital admission or readmission rate as an outcome measure. Authors typically report admission rates as clinical outcomes but not also as adverse events. This substantially distorts reported rates of adverse events in mindfulness research. Selecting an adverse event as a study outcome does not obviate the need also to report it as an adverse event. Hospital admissions should always be reported as adverse events, as is done in non-psychotherapy research.

Adverse events

Fourth, although real-world practical constraints mean that it is not possible to monitor all potential adverse events, what is practicable is for researchers to follow Jacobsen et al⁴ in stating clearly in their trial protocol those adverse events that were deemed particularly relevant to the trial, specifying how these were monitored (e.g. review of case notes) and giving detailed descriptions of any such untoward events that did occur.

Side-effects

Fifth, researchers are encouraged to consider using a standardised patient-reported side-effects questionnaire in trials of MBIs, as is done in trials of medication. None of the trials included in the most recent meta-analysis³ of MBIs for psychosis did so. Side-effects are distinct from adverse events and will occur in all psychotherapies. Recording subjective side-effects will broaden understanding of potential harm in MBIs for psychosis.

Symptom deterioration

Sixth, although studies increasingly seek to contextualise benefits of MBIs by reporting the number of participants who showed a clinically important improvement on measures of psychological functioning, it is rare for studies to report how many participants experienced a clinically important deterioration. None of the randomised trials of MBIs for psychosis included in the most recent meta-analysis³ reported data across all study arms on clinically important deterioration on the primary measure of psychological functioning – and we have found only one RCT published since then that included data on clinically important deterioration on the primary outcome measure across all study arms.⁵ Whenever possible, clinically important symptom deterioration by study arm should be reported in all future studies.

Drop out

Seventh, drop out occurs for many reasons but it can be a useful indirect index of harm. CONSORT diagrams show study drop out – the number of participants who did not provide post-intervention or follow-up data. However, a participant who ceased attending therapy yet still provided post-intervention assessment data would not be recorded as a study ‘drop out’. It would be useful for future studies to report not only study drop-out rates (including, where known, the reason), but also therapy non-completion rates as determined against a preset number of sessions (e.g. attending at least 50% of sessions).

Public and patient involvement

Finally, it will be important in future studies to ensure meaningful public and patient involvement (PPI) in research assessing harm. Studies are encouraged to state explicitly if and how PPI input contributed to specific decision-making on how harm was operationalised and monitored.

Conclusions

There is a pressing need to improve monitoring and reporting of harm in mindfulness for psychosis research – indeed, in the wider literature on MBIs – in order to develop a fuller understanding of both beneficial and harmful effects.