Table 13.

Overview of sham control design features that were shown to have an impact on trial outcomes.

Control intervention feature and description	Influence on trial results and discussion
Number of treatment sessions The number of times a patient receives the interventions, both through a provider or self-delivered.	In our preliminary assessments, we found this feature to be highly correlated with, eg, treatment duration and treatment frequency. We thus suspect the number of treatment sessions to be a good proxy for the extent of treatment received by patients. It is important to note that differences in the exposure to an intervention often occur not only when patients are asked to attend more clinic appointments than those in the control group but also when they are provided with home exercise or self-management programmes that expose them to higher “dosages” of specific and contextual effects.
Application mode The mode or medium through which an intervention is delivered, including telephone, online, hands-on or conversation-based delivery. In this sample, device-based active treatments were not eligible for review, but device-based shams (such as detuned ultrasound) were common.	Questioning the suitability of sham devices for blinding in non-device RCTs, differences in the “tool” through which active and control interventions are applied lead to less credible control interventions. Regarding effects on trial outcomes, the effect can likely go either way, with more elaborate sham controls leading to smaller between-group differences33 or undermining through supposed credibility differences. In other words, sham devices do seem to produce comparable placebo effects to respective active interventions but may compromise trials in other ways. The use of sham devices is still common, especially in physiotherapy and nonspinal manipulation trials (see Ref. 41). In psychological and behavioural interventions, the “application mode” may differ when control interventions rely on print or web-based materials rather than personal interactions.
Intervention individualisation The extent to which treatments are personalised or adapted to each patient may reach from fully standardised to highly individualised.	Likely somewhat overlapping with the concept of treatment standardisation, we found that the extent to which the active and control interventions are individualised to patients can influence trial outcomes. We can only suspect that the individualisation is communicated to the patient verbally and nonverbally, making the patient feel valued more or suppose that the treatment addresses the root cause of their problem. Contrastingly, patients receiving generic control treatment may wonder why symptomatic areas were not touched or topics not addressed that felt relevant to them. Highly standardised interventions of course offer little room for individualisation.
Patient participation The level of patients contributing to the intervention, ranging from physically and psychologically passive recipients to largely shaping the intervention themselves.	Many commonly used control interventions are more passive than the study treatment, especially in physiotherapy, self-management, and exercise trials. Our regression models did not clearly indicate, however, whether different participation levels contribute to differences in pain reports. While patients play an active role in most cognitive, behavioural, and exercise-based interventions for pain, manual therapies can be passive experiences for patients. This allows researchers to employ passive control interventions that match a passive treatment, as reflected by high similarity ratings in this subgroup.
Fidelity monitoring Monitoring and potentially promoting treatment adherence by patients or therapists' delivery of interventions according to protocols.	When the treatment adherence or therapists' intervention provision are ensured more in one group than in the other, patients will likely receive different amounts of active and control care. Differences in fidelity monitoring between groups was influential for differential attrition across all trials. Not only in physiotherapy but also in behavioural interventions, patients often perform (parts of) the intervention under their own supervision, eg, as home-based exercise programmes. Therefore, fidelity monitoring is more relevant than, eg, in most manual therapy scenarios. Potentially, fidelity monitoring itself also acts as an additional contextual factor, ensuring that patients believe that it is important how and how often the intervention is performed.
Treatment environment The physical environment in which interventions take place.	One of the most consistent predictors of between-study variance in pain outcomes, the treatment environment differs, eg, when the active intervention is mainly performed in a healthcare centre and the control intervention at the patient's home or vice versa.58 Another example is the study by Arcos-Carmona et al. (2011) where the intervention involved aerobic exercise in a swimming pool while the control group received a magnetotherapy sham, lying prone on a treatment bench.2 Unsurprisingly, this difference in environment comes with large differences in contextual factors that may result in differential placebo effects. While treatment environments were usually similar for all studied therapy types, the consistency with which differences in this feature predicted trial outcomes shows that matching of treatment environments ought to be paramount.