Design
Randomised, open, general practice-based pragmatic clinical trial.
Ethical aspects
The Medical Ethic Review Board of the University Medical Centre of Utrecht approved the study.
Patients
Patients with SMD were selected in a stepwise manner from 23 participating primary care practices in two provinces in the middle and east of the Netherlands. GPs were invited to participate without any exclusion criteria. Selections were made from the GP's electronic files. Patients with SMD who were between 18 and 65 years were eligible. Exclusion criteria were active depression (i.e. receiving any form of therapy), anxiety disorder, drug addiction or other major psychiatric diagnoses on axis I of DSM-IV. There was no medical restriction for participation, nor were more active patients, as assessed by the physical therapist, excluded.
Procedure
In the first phase, patients who visited the surgery within the last four weeks with a diagnosis in the mental health dimension (code P) of the International Classification of Primary Care (ICPC) were identified in the electronic medical files of the participating practices. An additional search was carried out on text words linked to SMD. Along with an introductory letter from their GP, these patients also received a screening list for SMD, which was specifically developed for this study. It comprised three questions: have you suffered from worrying, have you suffered from listlessness or have you suffered from tenseness in the past week? The answer categories were 'no' (code 0), 'sometimes' (code 1) or 'often' (code 2). The patients calculated their own scores; the screening test was considered positive if they had score > 3. In a pilot study score > 3 was demonstrated to rule in SMD with 85% sensitivity and 96% specificity (Terluin, personal communications). Patients with a positive score were eligible for the study and were invited in writing to participate. They received full information about the study and were asked to complete the Informed Consent forms. If they did not reply within two weeks they were approached by telephone.
The patients were randomised using a computer-generated list available at the trial unit of the University Medical Center Utrecht as soon as the informed consent forms were received. The treatment allocation was concealed as follows: the person who generated the randomisation list and reported the allocation codes to the researcher was not involved in determining patients' eligibility or any other aspect of the conduct of this study. He was instructed to assign patients in sequence. Furthermore, the fixed block size of four was unknown to the researchers.
Intervention
The patients were randomised either to intervention or to care as usual. The patients in the intervention group were referred to cooperating physical therapists, who designed an individually structured physical exercise programme (PEP) based on the Dutch Standard for Healthy Exercise. This comprises moderate intensive exercise on at least five days a week for at least 30 minutes [10]. Moderate exercise is defined as between 50 and 85% of maximal individual exercise as estimated on the basis of age and gender. Examples of activities include walking, swimming or cycling; the national standard is based on international research and guidelines [11, 12]. The therapist instructed the patient on the level of perceived exertion. The therapists were instructed to monitor the exercise programme with the patient for 12 weeks following a preset schedule. The patient visited the physical therapist eight times: twice in the first week, once a week in weeks 2-6 and in the remaining weeks once or twice per three weeks. We obtained no reports from the therapists as to how the exercise was done, nor about what the content of the monitoring visits was. The therapists were instructed in writing and by telephone, but received no further training; nor were they supervised.
In the control group the patients received usual care from their GP, i.e. supportive and explorative consultations, with pharmacotherapy if indicated.
Outcome
The primary outcome was the difference in change in general health 6 weeks after inclusion in the study. The secondary outcomes were total days off work, percentage that resumed work after 6 and 12 weeks, change in distress score and change in the dimensions of mental health, social health and role functioning after 6 and 12 weeks
Measurements
General health, social functioning and mental health are most affected in SMD [1].
For the primary outcome we selected the general health dimension of the SF-36. The general health dimension comes close to what GPs consider 'health in general terms' [13, 14].
The SF-36 is a validated instrument that measures eight dimensions of quality of life based on self-report. We used the standard four-week recall version. The SF-36 scores from 0 to 100, with 100 indicating the optimal health dimension. It is a well-known, internationally accepted instrument to measure quality of life; the general health dimension comes close to what GPs consider 'health in general terms' [13, 14].
We used the four-dimensional symptom questionnaire (4DSQ) to differentiate distress in SMD patients from depression, anxiety and somatisation. The 4DSQ comprises 50 questions on a four-point Likert scale. It was developed and validated in Dutch general practice and measures four dimensions of psychopathology: distress, depression, anxiety and somatisation. Distress is the dimension that is expected to be abnormal in SMD patients. The scale is self-reported and can be used for research as well as for clinical practice in patients from 15 years of age [1]. All questionnaires were completed on paper by the patients and returned by post.
Sample size
In view of the substantial burden on the patient of our intervention we decided that only moderate to large effect sizes (Cohen's d in the range 0.5 up to 1.0) would be of clinical interest [15]. Moderate to large effect sizes for general health equate to the effect of recovering from depression or a sleeping disorder [16]. Our sample size calculation was therefore based on an effect size of 0.75 (midpoint of 0.5 and 1.0) for the primary outcome, i.e. the SF-36 general health dimension at week 6. The standard deviation (SD) of this dimension of the SF-36 in the age range of 41 to 60 years was 20.6 according to data from a large general population sample [17]. With this value for the SD, an effect size of 0.75 corresponds to a 15-point difference between groups. To demonstrate such a difference with 90% power, an alpha value of 0.05 and equal allocation rates, 41 patients would be needed in each group. Assuming a 15% loss to follow-up rate, we had to include 48 patients per group. This rate was based on our experience.
We rated the distress of participants by using the distress score at baseline and at 6 and 12 weeks.
We counted the percentage of weekly self-reported hours off duty in both groups to estimate the economic effect of the intervention on the number of sick leave days.
Statistical analysis
The control and intervention groups were first compared for baseline characteristics. To prevent bias from missing values on the 4DSQ and the SF 36 at baseline (N = 102), 6 weeks (N = 87) and 12 weeks (N = 70), we performed a single imputation based on multiple linear regression. We then analysed the effectiveness of the physical activation programme at 6 weeks and 12 weeks on four dimensions of quality of life and four dimensions of psychopathology using analysis of covariance. This method is less sensitive to regression to the mean and has more statistical power than analysis of the difference of change in the outcome variable [16]. In an additional analysis we adjusted for differences in baseline characteristics. For each analysis we calculated the effect size (Cohen's d). A level of significance (alpha) of 0.05 was used.