Methods

A complex intervention is dened as an intervention that has many parts which may work independently or interdependently (Campbell et al. [2000]). The BeST trial compared active management (AM) with a complex intervention comprising active management plus group treatment using a cognitive behavioural approach (AM+CBA), and found that AM+CBA was eective in treating subacute and chronic low back pain (LBP) in both short and long term, and that it was cost-eective. Training was provided to health care professionals in order to deliver both treatments eectively. To deliver the active management consistently, primary care nurses were given a 1-hour training session on the best practice for the management of LBP, and asked to cascade this within their practices. To deliver the active management plus cognitive behavioural management consistently, other nurses, physiotherapists, psychologists, and occupational therapists attended a 2-day course, supported with remote mentoring.

701 patients with at least moderately troublesome back pain of at least 6 weeks' duration were recruited from 56 general practices in 7 regions in the UK be- tween April 2005 and April 2007. The patients were randomised 2:1 in favour of the cognitive behavioural approach arm of the trial. Originally, randomisation was to have been balanced 1:1 between the two treatment arms, but such a randomisation would have produced too few participants to run the cognitive behavioural approach groups, so it was switched to 2:1 randomisation in favour of the AM+CBA arm. Those in the control group received active management which is best practice care, including advice to keep active, to use painkillers, and take a positive outlook. They were also given a copy of `The Back Book', which was designed by LBP experts to reinforce these messages. Those in the intervention group received this same treat- ment, but in addition were invited to attend a series of six cognitive behavioural

group sessions with about eight participants in each. In the group sessions, the topics of goal setting, pacing, challenging beliefs, managing pain, and improving communication with health professionals were covered. Compliance with treatment was dened as attendance at least three of the six sessions. Patients recruited to the trial had subacute and chronic lower back pain that was at least moderately troublesome and of at least 6 weeks duration and had attended general practice, were at least 18 years of age, and had not been managed previously in a cognitive behavioural programme.

Demographic and clinical data were collected at the pre-randomisation stage, includ- ing date of birth, sex, lower back pain symptoms in the last six weeks, frequency of pain in the last six weeks, ethnic origin, age when left full-time education, and em- ployment details. The two primary outcomes were the Roland Morris Questionnaire (RMQ, Roland and Morris [1983], Roland and Fairbank [2000]) and the Modied Von Kor (MVK, Kor and Anthony [1982]) scale.

The Roland Morris Questionnaire (Roland and Fairbank [2000]) is the most widely used measure of lower back pain disability in primary care trials. It contains 24 yes / no responses relating to a range of functions commonly aected by lower back pain, and the score is the sum of the responses. There were concerns expressed by the trial report authors that the RMQ does not conform to the assumptions of scaling and normality of distribution that underpin its use in statistical analysis. It has been shown to be dierentially sensitive at low, mid and high ranges, with better sensitivity in the mid-range (Lamb et al. [2010a]). Previous trials have adopted a clinically signicant dierence between groups of 2.5 RMQ points.

The Modied Von Kor Scale assesses two dimensions, pain and disability, associated with back pain in the last four weeks. It is made up of six items, each of which is rated on a scale from 0 (no pain or disability) to 10 (worst pain or dis- ability). The rst three items relate to disability and the interference of pain with daily activity, recreation and ability to work. The last three items relate to worst pain, average pain and rating of back pain today. RMQ is scored by summing the responses and so has a maximum score of 60 (Kor and Anthony [1982]).

Secondary outcomes were occupational disability and limited activity days, participant satisfaction, psychological and behavioural measures, quality of life (QoL), health economics (including quality-adjusted life year (QALY)) and resource use. The psychological and behavioural measures were included because they measure

constructs hypothesised to lie on the causal pathway of eect, and might provide some explanation as to why the treatment was or was not eective.

The Fear Avoidance Beliefs Questionnaire (FABQ) is a measure of the degree of fear of pain and disability, and the avoidance of physical activities that can result (Vlaeyen and Linton [2000]). Each item is scored on a Likert scale 0 to 6 and mea- sures fear avoidance beliefs about work and fear avoidance beliefs about physical activity. The latter was selected for use as it has generic applicability. The sum score has a maximum of 24, with a higher score indicating a greater degree of fear and avoidance beliefs.

Pain self-ecacy (PSE) is a measure of a patient's condence to carry out a range of activities despite their back pain. There are ten items on a Likert scale each with responses 0 to 6, sum score up to 60, with higher score indicating the higher attribute of good pain self-ecacy (Nicholas [2007]).

Quality of life is measured using the Short Form 12 (SF12) a measure of health related quality of life widely used in back pain trials. There is a manual for scoring the physical and mental components on a scale of 1 to 100 for each and designed to have a mean of 50 and a standard deviation of 10 in a representative sample of the U.S. population, so that a score greater than 50 represents above-average health status (Ware et al. [1996]).

The Hospital Anxiety and Depression Scale (HADS) is a self-report rating scale designed to measure both anxiety and depression. It consists of two sub scales, each containing seven items on a 4-point scale (ranging from 0-3). The HADS is scored by summing the ratings for the 14 items to yield a total score, and by sum- ming the ratings for the 7 items of each sub scale to yield separate scores for anxiety and depression (Lisspers et al, 2007).

Troublesomeness was reported as a response to the question `How troublesome has your back been during the past 6 weeks?' on a single likert scale, with options `Not at all troublesome', `Slightly troublesome', `Moderately troublesome', `very troublesome', and `Extremely troublesome'. Those whose responses were `Not at all troublesome' or `Slightly troublesome' were not recruited to the trial.