What should we assess in outcome-studies to learn which patients benefit from treatments in multidisciplinary pain clinics?

In this issue of the Scandinavian Journal of Pain, Tarja Heiskanen, Risto Roine, and Eija Kalso report interesting and important outcome data from 439 patients treated in their multidisciplinary pain clinic (MPC) in Helsinki [1]. This report is important because there is in fact only meagre evidence that treatment in multidisciplinary pain clinics and centres (MPCs) helps patients with severe chronic pain. Does MPC in fact improve functions, quality of life, and reduce significantly the patients’ subjective burden of pain? As mentioned by Heiskanen et al. [1], most studies which have shown an effect, have done so in highly selected groups of patients, often being exposed to a specific modality of treatment. Only one good quality study from Denmark has provided evidence that this is possible in a MPC treating patients with a variety of chronic pain states [2].

In an editorial commentary in the journal Pain in August, 2012, Ashburn and Witkin [3] made the following serious statement: “... this lack of data may ultimately put the specialty of pain medicine at risk. We need to redouble our efforts to demonstrate that what we do, in fact, matters – and that the care we provide improves the lives of those we serve as well as society as a whole” [3].

It is in fact difficult to perform a controlled effect-study of treatments in a MPC for patients with severe chronic pain [4]. Most outcome studies of MPC treatment, like the one from Helsinki [1], include different types of pain patients and many treatment modalities. Thus, it may be difficult to know which patients are responding or which treatment modalities are working. In addition, a randomized controlled trial with follow up for several years, without further interventions, is almost impossible to perform.

Heiskanen and her colleagues had an appropriate and pragmatic approach in their study: they did not intend primarily to prove that their MCP is effective, but rather to examine factors and aspects of the patients and the treatments that might predict benefits from the treatment. And equally important, the study also focuses on factors that predict less or no improvement in the patients’ burden of chronic pain from treatment in their MPC [1]. In this setting they were able to report clearly and honestly the overall pragmatic outcome results from their well organized and well recognized multidisciplinary pain centre.

The overall results might seem a little disappointing, since only 46% presented a significant improvement in their health related quality of life (HRQoL) after about 6 months treatment in the pain centre. However, it is important that this improvement was maintained during a 3-year follow-up period [1]. The HRQoL did not change in 24% of the patients, and it deteriorated further during the 3 years after treatment, compared with at the start of treatments, in 31% of the patients. Since a control group was not included, we cannot know the clinical status of a control group with severe pain at follow up. So far we have little information of the natural course of patients with severe chronic pain without any treatment. However, in the study by Becker et al. [2], patients on a waiting list to be admitted to the pain clinic, did deteriorate with time on the waiting list receiving “treatment as usual” by their primary care physicians [2]. Anyway, the report of these data as published by Heiskanen et al. [1] is important, and other MPCs should at least present similar pragmatic data from their patient outcome studies.

The main objective of the authors of the Helsinki study was to examine patient- and treatment-related factors that might predict improvement in HRQoL. They compared aspects of the 100 patients with best outcomes with aspects of the 100 patients with worst outcomes. Their main findings were that among the 100 best responders there were more patients with higher education, better employment status, and fewer with disability or early retirement. There was a tendency to be more psychiatric co-morbidity in the poor-responders.

The following patient-related factors had no effect on treatment outcome: patient’s age or gender, marital status, duration and types of pain, number of painful locations, pain intensity, and pain-interference with (daily) functions, sick-leave, and expectation of outcome of treatment.

They found no treatment-related factors that were different among the good-responders compared with the poor-responders: duration of treatment was the same, so were number of drugs prescribed, types of pain-management modalities, number of different specialists or professions involved in their treatment, including psychiatric treatment and supportive therapy.

In the introduction of their paper, Heiskanen et al. [1] argue that “psychological distress has been identified as one of the risk factors”, still they seemed not to assess such factors properly. So, which psychological factors should be assessed to learn which patients benefit?

In assessing predictors of outcome or prognostic factors it is common to apply a so-called flag methodology [5]:

• Red flags indicate medical or physical factors,

• yellow flags indicate psychological and psychosocial factors such as marital and family functioning,

• orange flags indicate the presence of psychiatric disorders, and

• blue flags indicate work related factors, while

• black flags indicate more institutional and national system factors.

Ideally, in a prospective study of pain-management outcomes, as pointed out by Heiskanen et al. [1], various yellow flags that have been found to be prognostic factors and predict outcome from various multidisciplinary pain treatments (MPT) and other kinds of pain treatments should be assessed. They mention fear-avoidance beliefs and self-efficacy beliefs, but there are also other psychological and psychosocial factors that have been found to be important [1]. It is vital to have in mind that various psychological measures developed to tap various types of cognitive processes are based on different psychological theories [4]. Fear and avoidance beliefs are basically based on a theoretical model where catastrophic misinterpretations of bodily symptoms are central, resulting in various avoidance behaviours sometimes described as movement phobia. Self-efficacy beliefs, on the other hand, are based on Bandura’s self-efficacy model [6].

At the present time we have some knowledge of how different psychological measures affect outcome, but we lack information on to what extent they predict outcome independently of each other. It is thus important to include several measures that tap various psychological dimensions and psychosocial aspects to see which are the best predictors, and which predictors have unique predictor capabilities.

In such research it is important to select instruments with adequate psychometric properties and from various theoretical foundations tapping both intrapsychic psychological dimensions and social factors such as marital and family functioning (which were included in [1] to some degree). The latter is considered most important when treatments based on operant conditioning are applied, but may be equally important when treatments based on other theoretical foundations are used.

It is important to find out whether various variables act as predictors irrespective of types of treatments given or whether they are treatment specific. In such research it is also important to include blue flags measuring work related factors. In the Heiskanen et al. study it was found that responders significantly more often were working and significantly less were receiving pension of some kind [1]. As pointed out by Heiskanen et al., it is also important to assess the presence of psychiatric disorders (orange flags). While yellow flags mainly tap cognitive and social measures more specifically related to the maintenance of the pain symptom, orange flags tap the presence of psychiatric disorders according to formal diagnostic systems such as DSM-IV. This was done in their study [1], but they do not describe how it was done. The prognostic value of psychiatric co-morbidity is indirectly documented in their study since they found that “almost twice as many patients in the non-responder group had been seen by a psychiatrist compared with the responders” [1].

Valid and reliable assessment of psychiatric disorders, such as meeting criteria for DSM-IV, is time consuming. It requires the use of semi-structured interviews such as the Structured Interview for the DSM-IV Axis I (SC1D-1). Optimally SC1D-1 interviews should be video or audio taped to obtain inter-rater reliability which is even more time consuming. Recently promising self-report measures of DSM-Axis 1 disorders have been developed which may prove to be helpful [7,8]. Future studies should aim towards including a large number of potential factors that may affect outcome. We do not only need to identify prognostic factors, but, as mentioned above, to identify basic predictors and prognostic factors so that new measures will have to prove that they have prognostic value over and above other already established measures. Finally, Heiskanen et al. also discuss potential mediators of change [1]. This is very important research because it aims to identify through which mechanisms a certain treatment works. 1f we are able to identify the mediators of change, then we can increase the dosage of that mediator variable and thereby increase the effectiveness of the treatment. It is, however, important to emphasize that conducting mediator research requires the use of a strict research methodology in order to be able to draw causal inferences. Such kind of research is highly recommended and can be added both to a randomized controlled trial and an effectiveness study without a control group.