Like surgery, psychological intervention is dependent on the skill, training, and experience of the operator; is manufactured in the moment; and is tailored to the individual case. The overall small- to moderate-effect sizes of CBT hide a heterogeneity of content, operator characteristics, exposure time, and therapist allegiance, which go largely unreported. Individual candidate measures of process are often investigated, but common features of treatment go undiscussed. Burns has argued recently that attention to common mechanisms, in particular ‘behavioural activation’—actively engaging in practicing or experimenting with meaningful physical change—is a good candidate ^{10, 11} . A further example is the need to account for parental distress, and parenting, in the treatment of adolescent chronic pain ¹² . A novel target for therapy development is in making the non-specific specific.

Despite a large evidence base, there are critical gaps. Individual differences and the importance of pain presentation are rarely investigated. There is some consideration of sex differences in therapy outcomes ¹³ , of chronic pain in later life ¹⁴ , and of delivery in particular settings such as the workplace ¹⁵ . But these are rare cases. Odder still is the absence of data for specific populations. For example, we found only three RCTs of psychological interventions eligible for a Cochrane Review in chronic neuropathic pain ¹⁶ . Exactly how the form and content of pain shape psychological experience is largely unexplored, hidden within large compound variables such as anxiety, depression, or disability. The extent to which, for example, worry about headache is critically different from worry about pelvic pain is important, as is the history of the meaning of that pain (for example, whether it was related to previous disease). A novel target for therapy development is establishing illness-specific psychological theory.

Pain intrudes on awareness and functions to protect by urging escape and avoidance from potential harm. Repetitive inescapable interruption, motor preparation for flight, and heightened sensitivity to cues of danger create a pain-dominated environment in which accurate prediction of action in the context of pain is difficult ¹⁷ . Recent theory in embodied pain suggests that the inherent uncertainty around experience might be a therapeutic opportunity, with potentially analgesic outcomes ¹⁸ . Experience-altering interventions using virtual and augmented reality ¹⁹ , anatomical education ²⁰ , or exposure ²¹ are interesting first forays into treatments aimed at directly altering experience. A fully embodied rehabilitation approach to chronic pain will embrace pain as always selected over competing demands for protection in the uncertain context of threat. A novel target for therapy development is to adopt an active psychology of meaningful engagement.

A frontier for psychological therapy is to embrace the possibilities of technology, not only in augmenting, supporting, or replacing the remote delivery of traditional face-to-face treatment but for novel therapy content. Technology can do what therapists cannot, and can do many things better. Technology can accompany the patient, measure multiple aspects of experience, render data into accurate information instantaneously, give immediate access to knowledge, send and receive messages in near real-time, and allow discourse, anytime and almost anywhere. The opportunities of technology have yet to be explored. A user-centred modern therapy would be delivered flexibly over multiple devices, be highly dependent on the small data people trail about their lives ²² , work just in time ²³ , and be highly contextual, integrated, and relational ¹⁷ . A novel target for therapy development is to re-imagine therapy as active in the minutia of people’s lives, lives lived with technology rich in therapeutic promise.

Psychology is largely concerned with behavioural assessment and change. Some behaviour is observable by self and other (for example, a physical attitude or gesture), and some is observable only by self (for example, a thought, emotion, belief, or bodily perception such as pain). The successful measurement of behaviour relies on the independence of a construct and on the quality of the measurement technology used to capture it. In pain, there is an active science of instrument development. Lost in pain research is a consideration of the importance of construct coherence even before the deployment of measurement technology. Here are three examples. We have shown that the measurement of acceptance of chronic pain is corrupted with content more pertinent to physical and social function, making it almost impossible to independently measure the role of function in altering acceptance, or vice versa ²⁴ . There is little evidence that adolescents experience catastrophic thinking about pain, and what is normally measured is better described as worry ²⁵ . And research into the popular construct of ‘somatisation’ is fundamentally flawed in pain research by the absence of demonstration that patients meet criteria for somatisation ²⁶ . Needed is consideration of the content of measures beyond their labels, the reporting of data at an item level, a priori establishment of independence of constructs, and a consensus over meaningful clinical endpoints.

Also needed are within-subject idiographic methods. The randomised placebo-controlled trial is an ill-fitting method of assessing psychological treatment efficacy and safety and will be further challenged by individual, context-dependent, temporally dynamic therapies. Increased demands for data-sharing might allow for post hoc review of individual data, but they rely heavily on the availability and quality of trial data. Single-case series are often more relevant and are highly versatile and under-employed ²⁷ . In pain, there is also a strong tradition of using ecological momentary assessment in which people report on behaviour when it occurs or at set intervals of time ²⁸ . The advent of pervasive communication technology being repurposed for therapeutic use makes it necessary to innovate single-case, personally situated, data-rich methods. There is research already in the use of big data ²⁹ , and interesting development of a micro-randomised trial, that involves moving the randomisation point from pre-treatment to various points within treatment. For example, one might randomise the delivery of a remote message prompt, within-patient, within-trial ²³ .

The conduct of all studies in pain is likely to come under further scrutiny for how known biases are managed and for the possibility of research misconduct, including fraud. Cochrane risk-of-bias assessment of all primary trials in psychological treatments shows general problems of reporting biases (for example, selective reporting) and performance bias (for example, lack of appropriate blinding). The emergence of reporting standards—for example, Consolidated Standards of Reporting Trials (CONSORT)—and the insistence by journals of trial registration are helpful developments; see www.equator-network.org. Further problems specific to trials of psychological interventions for chronic pain are the entering of patients to trial on the basis of criteria only indirectly relevant to treatment (for example, pain severity), entering of patients to trial with mild or no problems, failure to consider or report adverse events, and the selective or non-reporting of data. The extent to which these biases are systemic and arise from a failure of clinical equipoise in pain psychology is unknown, although there are examples of authors arguing for treatment efficacy, when their data show no evidence of effect ³⁰ . Understanding narrative bias will help authors judge the quality and impact of any efficacy investigation.