Effect of an education and activation programme on functional limitations and patient-perceived recovery in acute and sub-acute shoulder complaints – a randomised clinical trial

In our randomised clinical trial, patients were allocated at random to either EAP as an addition to UC, or to UC only. Measurements were taken at baseline and after 6 and 26 weeks by a self administered questionnaire. The 6 and 26 weeks questionnaires were sent and returned by mail. The 6 weeks measurement provided information on the immediate effect of the EAP as the EAP had to be administered within the six week period after the baseline measurement. The 26 weeks measurement provided information on the long term effect of the EAP. EAP was administered by GPs or by an ambulant therapist (CDB) if no EAP-trained GP was available near a patient's home. All GPs who provided EAP attended a three-hour training session, in which the EAP was introduced and role-plays were used to train the proper administration of EAP. The GPs received a training manual during this session in which the principles of the EAP were summarized. Patients in the UC group were given UC by their own GP, unless their GP had attended the EAP training. In that case, UC was administered by a colleague from the same GP group practice, to avoid contamination. The design of this study has been described in detail in a previous publication [13]. The Medical Ethics Committee of the Institute for Rehabilitations Research in association with Rehabilitation Foundation Limburg has approved this randomised clinical trial.

Eligible patients had consulted their own GP or responded to advertisements in local newspapers calling on people with a new and untreated episode of SCs that had lasted less than three months and produced complaints at rest or complaints elicited by shoulder movement. Patients were included if they were older than 18 years and living in the south of the Netherlands. Additional inclusion and exclusion criteria are given in table 1.

The patient recruitment procedure in the consultation room was designed to minimize the time needed by the GP, since lack of time is often mentioned as one of the main barriers when recruiting patients in general practice [14]. GPs pointed out the existence of the EAP trial to patients with newly presented SCs, checked the inclusion and exclusion criteria and forwarded the patients' personal data by fax to the research centre. The patients had to give written permission for their personal data to be forwarded to the research centre. The GPs were advised to refer patients to the research centre for further information about the EAP trial. Patients responding to the advertisements were first screened by telephone for inclusion and exclusion criteria. Patients meeting the inclusion criteria were subsequently visited by an independent GP involved in the EAP trial (EAP-GP). This EAP-GP provided the patients with a regular consultation for SCs. Personal data of eligible patients were forwarded to the research centre. In both recruitment strategies, the research assistant contacted patients within 2 weeks and took care of final inclusion and randomisation. The baseline questionnaire was handed out and after completion returned to the research assistant.

Blocked and concealed randomisation with blocks of 4 patients was used to allocate patients to either the EAP group or the UC group. An independent researcher used a computer-generated random sequence table to randomise the patients in each block. The seals of the prepared, numbered, opaque, sealed envelopes containing the treatment group were broken by the research assistant after eligibility had been verified and the patient had given written informed consent.

Neither the patients nor the GPs, nor the trained ambulant therapist, could be blinded for the allocated treatment. The ambulant therapist was also the researcher coordinating the randomised clinical trial and conducting the data analysis, but he was blinded for treatment allocation during the data analysis. The allocation code was kept by an independent researcher (JG) and was revealed only after data analysis had been completed.

The focus of EAP is to maintain or induce the proper cognitions by education and to stimulate adequate behaviour by means of advice on activities of daily living, using principles of operant conditioning (11). The programme consists of an educational and an activation part.

The educational part of the EAP consists of tailored information intended to take away the worries and answer questions patient may have regarding their SCs. Special care is taken to structure the information and advice that patients receive from other individuals in their social and health care environment. The information and advice are tailored to the patients' thoughts. Preconceptions on SCs are identified and altered if they are incorrect or reinforced if they are correct. The final aim of the educational part of the EAP is to provide patients with a realistic idea of their prognosis and the effect of treatment.

The activation part aims to assist patients in the continuation or resumption of activities affected by the SCs, despite the pain. The adverse effects of inactivity are discussed with the patients and activities that patients indicate to be affected by the SCs are closely monitored during the subsequent consultations. Schedules are set for the resumption or gradual increase of these activities, using a time-contingent approach, which means that the resumption or increase of activities occurs irrespective of pain experience but according to preset goals in time.

The EAP consists of a minimum of two sessions and a maximum of six follow-up sessions over a period of six weeks. Each session can last up to 20 minutes.

UC was administered according to the clinical guidelines of the Dutch College of General Practitioners [9].

The first primary outcome measure, assessed at 6 and 26 weeks after randomisation, was patient-perceived recovery (PPR) [10]. Patients were considered recovered when they reported to be 'much improved' or 'fully recovered', on a 7-point ordinal scale, six weeks after randomisation.

The second primary outcome measure was a change in functional limitations of activities of daily living. This variable was assessed by the 16-item shoulder disability questionnaire (SDQ)[15], with a standardised scoring range of 0 to 100. A lower score on this questionnaire implies lower levels of functional limitations.

Several psychosocial variables were assessed at baseline. Anxiety, depression, somatisation and distress were assessed using the four-dimensional complaint list [16]. Catastrophising and coping were assessed by 6-level subscales of the Pain Coping and Cognition List (1: completely disagree; 6: completely agree) [17]. Mean subscale scores of 1 were classified as 'very low' (code = 0), scores between 2 and 6 were classified as elevated (code = 1) [17]. Other specific disease variables recorded at baseline were pain intensity, measured on a 10-point visual analogue scale; onset (quick or gradual); affected shoulder and having had prior episodes of SCs lasting at least 1 week.

About half of all newly presented episodes of SC in general practice are reported to last for at least six months. A number needed to treat of 4.5 after 26 weeks is considered clinically relevant. This implies an absolute reduction of 22% of the proportion of patients with SC after 26 weeks. With a two-sided alpha of 0.05 and a statistical power (1-β) of 0.80, 82 patients per treatment group were needed to detect a difference in favour of EAP compared to UC after 26 weeks.

The baseline variables of the treatment groups were compared using chi-square tests and an independent samples t-test. Significant differences in baseline variables were considered to be potential confounders. The statistical data analysis was carried out according to the 'intention-to-treat' principle.

Patients attending the same GP cannot be assumed to be fully independent. Similarly, different observations for the same patient with SCs cannot be assumed to be independent either. Multilevel analysis was used to address this dependency due to clustering of data. The effect of treatment group (EAP group = 1; UC group = 0) was analyzed by means of linear multilevel analysis if SDQ was the outcome variable and logistic multilevel analysis if patient-perceived recovery was the outcome variable. Three levels of variance were distinguished: GPs, subjects and measurements.

In the linear multilevel analysis, baseline SDQ scores were entered into the model to adjust for differences at baseline. Since none of the patients were recovered at baseline, no adjustment was needed for patient-perceived recovery at baseline in the logistic multilevel analysis. Time of measuring was represented in the model by two dummy variables for the measurements at 6 and 26 weeks. Potential confounders were also entered into the linear and logistic model as independent variables. Finally, the interaction effect between the treatment group and the time of measuring was also included in the model.

The multilevel analyses resulted in estimates (and standard errors) of the fixed and random effects. Likelihood ratio (LR) test statistics were used to determine whether the estimates were statistically significant (p < 0.05) in the linear multilevel analyses. Wald chi-square tests were used to determine the statistical significance of the estimates in the logistic multilevel analyses. Estimates that did not reach the required level of significance were excluded from the model in a top-down procedure, except for the intervention variable, leaving out the least significant estimates first. For the logistic multilevel analysis, these estimates were converted to odds ratios with their 95% confidence intervals. All multilevel analyses were performed with MLwiN (version 1.10) [18, 19].

GPs referred a total of 133 patients for participation in the EAP trial, 74 of whom were actually included in the trial. In total, 64 patients responded to the advertisements. Of these, 44 met the selection criteria, as ascertained by a telephone interview. The subsequent visit by the EAP-GP resulted in the final inclusion of 37 patients. A total of 111 patients were thus eventually recruited to participate in the study (figure 1). Three patients dropped out after randomisation without receiving any treatment, and were therefore excluded from further analysis. After 6 weeks, complete data was available for 40 patients (77%) in the UC group and 48 patients (86%) in the EAP group. After 26 weeks, complete data was available for 35 patients (67%) in the UC group and 44 (79%) patients in the EAP group.

Table 2 shows that there were no statistically significant differences between the two treatment groups at baseline for any of the variables except catastrophising. The effect of baseline differences in catastrophising was evaluated by entering this variable into the multilevel analysis models.

Estimates with their standard error and levels of significance are presented in table 3 for the linear multilevel analyses of the SDQ scores. The interaction terms between the two dummy variables and the group variable did not reach significance and were excluded from the final model in the top-down procedure. This implies that the potential effect of treatment group was similar at both post measurements. In the final model, treatment group turned out to have no effect at either of the measurements. Catastrophising at baseline and baseline SDQ scores were significantly and positively related to SDQ scores at both post measurements. Time dummy 1 and time dummy 2 were significantly and negatively related to SDQ scores, suggesting that SDQ score, representing the level of functional limitations, decreases as time progresses.

Estimates with their standard errors and levels of significance for the logistic multilevel analyses of the PPR are presented in table 4. Data were interpreted by converting estimates to odds ratios (ORs). The top-down procedure resulted in the exclusion of the interaction terms from the logistic model. This implies that the potential effect of treatment group was similar for both post measurements. In the final analysis model, the treatment group had a non-significant (p = 0.1784) effect on patient-perceived recovery. A significant effect was found for time dummy 1 and time dummy 2, indicating a significant effect on PPR of the time elapsed since baseline. Baseline levels of catastrophising did not have a significant effect on the PPR and were excluded from the final analysis model.