- Elaine Ward, research fellowa,
- Michael King (), heada,
- Margaret Lloyd, readera,
- Peter Bower, research fellowb,
- Bonnie Sibbald, professorb,
- Sharon Farrelly, research assistanta,
- Mark Gabbay, senior lecturerc,
- Nicholas Tarrier, headd,
- Julia Addington-Hall, senior lecturere
- a Department of Psychiatry and Behavioural Sciences, Royal Free and University College Medical School, University College London, London NW3 2PF
- b National Primary Care Research and Development Centre (NPCRDC), University of Manchester,
- c Department of Primary Care, University of Liverpool,
- d Department of Clinical Psychology, School of Psychiatry and Behavioural Sciences, University of Manchester,
- e Department of Palliative Care and Policy, King's College School of Medicine and Dentistry, London
- Correspondence to: M King
- Accepted 10 August 2000
Objective: To compare the clinical effectiveness of general practitioner care and two general practice based psychological therapies for depressed patients.
Design: Prospective, controlled trial with randomised and patient preference allocation arms.
Setting: General practices in London and greater Manchester.
Participants: 464 of 627 patients presenting with depression or mixed anxiety and depression were suitable for inclusion.
Interventions: Usual general practitioner care or up to 12 sessions of non-directive counselling or cognitive-behaviour therapy provided by therapists.
Main outcome measures: Beck depression inventory scores, other psychiatric symptoms, social functioning, and satisfaction with treatment measured at baseline and at 4 and 12 months.
Results: 197 patients were randomly assigned to treatment, 137 chose their treatment, and 130 were randomised only between the two psychological therapies. All groups improved significantly over time. At four months, patients randomised to non-directive counselling or cognitive-behaviour therapy improved more in terms of the Beck depression inventory (mean (SD) scores 12.9 (9.3) and 14.3 (10.8) respectively) than those randomised to usual general practitioner care (18.3 (12.4)). However, there was no significant difference between the two therapies. There were no significant differences between the three treatment groups at 12 months (Beck depression scores 11.8 (9.6), 11.4 (10.8), and 12.1 (10.3) for non-directive counselling, cognitive-behaviour therapy, and general practitioner care).
Conclusions: Psychological therapy was a more effective treatment for depression than usual general practitioner care in the short term, but after one year there was no difference in outcome.
Editorial by Briggs
Brief psychotherapy (such as non-directive counselling or cognitive-behaviour therapy) is widespread in general practice in Britain.1 Recent randomised controlled trials of the effectiveness of non-directive counselling have suggested that it is no more effective than usual general practitioner care. 2 3 Although there is good evidence for the effectiveness of cognitive-behaviour therapy in specialist settings,4 trials in general practice have produced equivocal results.5–8
Many trials in general practice have been limited by poor recruitment. Patients and clinicians are reluctant to risk losing access to a resource that is already available when allocation is randomised.9 Even patients who agree to random allocation may do so in the hope of being assigned to their preferred treatment. Together with preconceived beliefs about the value of treatments, this may mean that groups differ at baseline in terms of patient and clinician motivation and expectation of outcome. In order to overcome these limitations we carried out a patient preference randomised controlled trial. Patients with no strong preference for any of the treatment alternatives were randomised, whereas those patients who expressed a strong preference were allocated to their treatment of choice.10
Our null hypothesis was that there would be no difference in clinical effectiveness between non-directive counselling provided by an accredited counsellor, cognitive-behaviour therapy delivered by a suitably trained psychologist, and usual general practitioner care.
Our study received ethical approval from the appropriate committees in each centre. Patients were recruited from February 1996 to November 1997 from 13 general practices in north London and 11 practices in greater Manchester. General practitioners were asked to refer all patients suffering from depression or mixed depression and anxiety for whom they believed a brief psychological intervention was necessary (see box).
Trial inclusion and exclusion criteria
Aged 18 years or over
Depressed or depressed and anxious, as assessed by a score of ≥14 on Beck depression inventory
Serious suicidal intent
Psychological therapy in past six months
Currently taking antidepressant drugs
Organic brain syndromes
Inability to complete questionnaires because of language difficulties, illiteracy, or learning disability
Patients who fulfilled the inclusion criteria and scored 14 or above on the Beck depression inventory11 entered the study. After the assessment, the researcher took each patient through a series of explanations about the treatments and the allocation procedure. We encouraged participants to accept randomisation. Those who continued to express a strong preference were allowed to choose their treatment. Randomisation used numbered, sealed, opaque envelopes and was blocked and stratified on severity (high (≥23) or low (14-22) on the Beck depression inventory).
About nine months into the trial the preference arms for counselling and cognitive-behaviour therapy were close to being filled. Discussions with patients indicated that most had no preference for a specific psychological therapy but were reluctant to risk random allocation to standard general practitioner care. We were concerned that closing the preference arms during the trial would lead to changes in the types of patients referred because of the loss of choice over allocation. We therefore decided to offer new patients with a preference for psychological therapy the choice of randomisation between the two therapies. This procedure had the advantage of increasing the numbers of randomised patients available for the comparison of the two psychological therapies. Separate allocation sequences (blocked and stratified) were generated for this procedure. Anyone insisting on a specific therapy was offered a single, one hour assessment session with his or her preferred professional, but was not retained in the study.
Intervention groups: non-directive counselling and cognitive-behaviour therapy
Six counsellors and three psychologists took part in London, as did eight counsellors and nine psychologists in Manchester. Counsellors complied with a non-directive approach outlined in a manual that we developed2 based on the work of Rogers.12 Cognitive-behaviour therapists complied with a problem formulation and staged intervention approach outlined in clinician and patient manuals. 13 14 The therapists agreed to one hour of supervision for every six hours of patient contact time. All counsellors had the necessary qualifications and experience to be accredited by the British Association for Counselling. All cognitive-behaviour therapists were psychologists who had the necessary qualifications and experience for accreditation by the British Association for Behavioural and Cognitive Psychotherapies and were eligible for registration with the United Kingdom Council for Psychotherapy.
Patients were offered six sessions initially, with a maximum of 12. Appointments were usually provided on a weekly basis at the surgery and lasted about 50 minutes. Participants allocated to psychological therapy were free to see their general practitioner as usual, but we requested that the doctors refrain from routinely prescribing antidepressants for these patients.
Control group: usual general practitioner care
General practitioners treated patients in this group according to their usual practice, but were asked to refrain from referral for psychological interventions unless this was imperative.
We assessed participants at referral and at four and 12 months later. Assessments of outcome were not blind to allocation. We chose self reported measures to avoid interviewer bias and to allow postal follow up if participants refused face to face contact. The Beck depression inventory was our main outcome measure.11 Additional assessments reported in this paper included a revised demographic and economic questionnaire2; the brief symptom inventory, which measures a range of psychological symptoms 15 16; the modified social adjustment scale17; and a modified measure of patient satisfaction2 based on Elliott and Shapiro's impact of events scale.18 We used the computerised revised clinical interview schedule19 at baseline to provide a diagnosis based on ICD-10 (international classification of diseases, 10th revision).20 The integrity of the psychological therapies was assessed with the cognitive therapy rating scales (Young J, Beck A. Cognitive therapy scale: rating manual. Unpublished manuscript 1980).21
We analysed the data using SPSS for Windows and SPSS/PC +. Three groups were considered in the analysis: participants randomised between three arms, those randomised to psychological therapy (using either allocation procedure), and those who chose their treatment. There is overlap between the first two groups. We used an intention to treat analysis with the last observation carried forward as a conservative estimate of outcome when data were missing at either follow up point. However, we provide numbers of participants and mean (median) scores without last observation carried forward for the purposes of comparison.
We used the general linear modelling, repeated measures procedure with two factors for comparisons between participants—randomised group (three levels) and site (two levels, Manchester and London)—and one factor for comparison within participants—time (three levels). A priori comparisons within participants were defined using the repeated factor to compare adjacent time points, in this case baseline versus four months and four months versus 12 months. Where data were not normally distributed (brief symptom inventory) we used square roots to normalise them. A power calculation before the study indicated that we would need 65 in each group to detect a mean difference in outcome between the groups of 3.5 (SD 8) in Beck depression inventory score at 90% power and a 5% level of significance.
Seventy three general practitioners referred 627 patients, of whom 163 were excluded (see figure). We recruited 119 patients in Manchester and 345 in London. Patients were seen within a mean of 11 days (SD 13) of referral to the study. Two patients chose to be treated by their general practitioner, giving too small a group to be included in the analysis. There were also two protocol violators who were included in the analysis.
The participants' mean age was 37 years (SD 12.2), and 75% were women. We examined demographic factors with respect to the three main analyses for the trial: those patients randomised between three treatments, those allocated to therapy using either randomised method, and those who chose their treatment (table 1).
Diagnoses were available for 435 participants (94%). Depression was the main diagnosis for 271 (62%) of participants, the remainder falling into “no overall psychiatric diagnosis” (85 (20%)) or “behavioural” difficulties (79 (18%)). “Depressive symptoms” was a secondary diagnosis in 47 (11%) of the referrals. The mean global severity index on the brief symptom inventory at baseline was 1.75, which is similar to that reported in a previous trial in UK general practice.2
Forty two per cent of the participants agreed to randomisation between three treatments, 30% chose a particular treatment, and 28% agreed to randomisation between the two therapies (see figure). The follow up rate on the main outcome measure was 89% at four months (15% completed by post or telephone) and 81% at 12 months (25% completed by post or telephone).
Process of treatment
We gathered details of patients' protocol treatments and use of health services through searches of medical records, therapists' records, and self reported interviews for the economic analysis: we report the data here to illustrate the process of treatment. Seventeen (27%) of the 63 participants randomised to cognitive-behaviour therapy and 20/67 (30%) of those randomised to non-directive counselling received an antidepressant prescription from their doctors. The patients in these cognitive-behavioural and counselling groups received a mean of 5.0 (SD 3.5) and 6.4 (SD 4.2) sessions respectively, but nine (14%) and seven patients (11%) in each respective group did not attend any treatment sessions. The commonest reasons for termination of therapy reported by the therapist were agreement between therapist and client (43% in cognitive-behaviour therapy and 38% in counselling) and client failure to attend (17% and 15% respectively).
We were able to collect data on date of termination of therapy on 113 (87%) of the 130 participants randomised to either cognitive-behaviour therapy or non-directive counselling. These revealed that 80 had completed therapy by the first follow up at four months.
The 67 patients randomised to usual general practitioner care received a mean of 9.2 (SD 5.2) surgery consultations in the 12 months after allocation. Thirty three (49%) received an antidepressant prescription, 12 (18%) received an anxiolytic prescription, and 23 (34%) were referred to a mental health professional (both NHS and private provision).
Integrity of psychological therapies
To ensure differentiation between the therapies, we used the cognitive therapy rating scale (Young J, Beck A. Cognitive therapy scale: rating manual. Unpublished manuscript 1980) to provide both a measure of the adequacy of cognitive-behaviour therapy and to check that counsellors made substantial use of cognitive techniques. There was no specific rating of the quality of non-directive counselling.
Therapists were asked in advance to tape all sessions with the second and fifth patient allocated to their care. Because of problems with patient consent and equipment malfunction, only 18 therapists (69%) provided useable recordings of sessions. Two sessions were used for quality control for each therapist (chosen randomly when more than two useable sessions were available). An independent psychologist, experienced in the rating scale, scored all the cognitive-behaviour therapy sessions, and none of the counselling sessions, above the predetermined cut off value (39) indicative of adequate cognitive therapy. The rater was not always blind to the type of therapist recorded, as some therapists occasionally referred to their background. There were no significant differences between the two sites (London and Manchester) in scores. Two recordings of non-directive counselling from two therapists were rated as atypical, but not to the extent that the therapy could be considered cognitive-behaviour therapy.
Participants randomised between all three arms of the trial
In our comparisons between participants there was no effect on depression for treatment group (F=1.41; df=2, 191; P=0.25) (table 2). There was no time by site interaction. We found a significant effect on depression for time (Wilks λ=0.411; F=135.90; df=2, 190; P<0.001) and for the time by group interaction (Wilks λ=0.923; F=3.874; df=4, 380; P=0.004). In our comparisons within participants there were significant differences for time by group, both for baseline to four months (F=4.91; df=2, 191; P=0.008) and four months to 12 months (F=5.29; df=2, 191; P=0.006). This means that the groups' depression scores changed at different rates between each time point, both therapy groups improving more than the group given usual general practitioner care between baseline and four months, while the usual care group made more change between four months and 12 months (table 2). In summary, both therapy groups improved significantly more rapidly than the usual care group in the first four months, while in the latter eight months the usual care group made up the difference.
In our secondary outcomes there was a trend in the same direction for the brief symptom inventory. Additionally, at 12 months, patients in the usual care and cognitive-behaviour therapy groups had made significantly greater gains on the social adjustment scale than those receiving non-directive counselling. Satisfaction with treatment differed significantly between the three groups at four and 12 months follow up. Post hoc Scheffe tests showed that satisfaction was higher in both psychological therapy groups than in the usual care group at four months. At 12 months, only patients given non-directive counselling had higher satisfaction than those given usual care.
Patients randomised between the two psychological therapies
All participants who were randomised to either psychological therapy group using either randomised allocation method were combined in this analysis. There were no significant differences in clinical outcome or satisfaction (table 3).
Patient preference groups
We found similar outcomes for patients who chose either therapy arm (table 4). Again, there were no significant differences between the two arms at four or 12 months. Patients who chose counselling were more satisfied with treatment than those who chose cognitive-behaviour therapy at 12 months. There were no significant differences in Beck depression inventory scores at either outcome point between participants who were randomised to each psychological therapy and those who chose it.
Despite an expansion of general practice based counselling, there is limited evidence for its effectiveness. 2 3 7 22 In a retrospective analysis of an earlier study,2 non-directive counselling was found to be more effective in the subgroup of patients who scored above a threshold of 14 on the Beck depression inventory. This finding is supported here to the extent that trial patients (all of whom scored ≥14 on the Beck depression inventory) recovered more quickly when referred to counselling rather than remaining in general practitioner care. At 12 months, however, the latter had made up this difference.
Treatment trials that include patients who are not willing to be randomised allow trialists to estimate the representativeness of the randomised sample and to compare outcomes between patients in the randomised and preference arms, although outcome comparisons are vulnerable to selection bias.23 Our results confirm that randomised patients resembled non-randomised patients. Although our design incorporating patient preference thus afforded some protection against external threats to validity, it is likely that several patients refused to participate altogether. In these cases doctors were asked to complete a form detailing the reason for refusal. However, compliance was poor, and we cannot show that the trial participants are representative of eligible patients generally.
Our psychological interventions were brief. In a general practice survey the average number of sessions provided by counsellors was close to six, but this was less than the average of 16.5 reported by psychologists.24 The usual general practitioner care given in our study comprised a variety of strategies, including therapies similar to those offered in the other arms of the trial. To the degree that such management reflects routine practice, it allows a relevant assessment of the cost effectiveness of the two defined treatments (see our second paper).
Our results can be criticised because all the reported outcomes were based on self reporting without any other verification. Furthermore, we adopted a conservative approach to data analysis by using last observation carried forward. However, follow up rates were high and thus data were imputed in only a minority of cases. At 12 months, patients in the usual care and cognitive-behaviour therapy groups showed significantly greater gains in self reported social adjustment than did those given non-directive counselling. However, the differences in scores were small and occurred against a background of improving social adjustment in all three groups.
Both therapy groups had an advantage over usual care after four months—about the time therapy ended for most participants. Why might we, in contrast with several recent trials, 2 3 7 25 have found a significant effect for non-directive counselling on depressive symptoms at four months? Restricting recruitment to only those patients with medium to high levels of depressive symptoms might have increased the likelihood of treatment having an impact. In some trials all patients referred by general practitioners entered the trial regardless of whether they met any diagnostic or severity criteria, 2 3 whereas in others entry was restricted to patients with major depression.7 Unfortunately, the latter study lacked power. Whether the reduction of depressive symptoms is of clinical importance is a more complex issue. The effect sizes we found would imply that these interventions are similar in impact to other effective treatments in general practice and mental health, but we stress that this advantage was lost 12 months after entry to the trial.
We conclude that employing practice based counsellors or cognitive-behaviour therapists may achieve a faster resolution of symptoms in patients with moderately severe depression.
What is already known on this topic
Brief psychotherapy is widely available in general practice in the United Kingdom
Evidence to date indicates non-directive counselling is no more effective than usual general practitioner care
Although cognitive-behaviour therapy is effective in specialist settings, trials in general practice have produced equivocal results
What this study adds
Employing practice based counsellors or cognitive-behaviour therapists may enable patients with moderately severe depression to recover faster
Non-directive counselling and cognitive-behaviour therapy seem to be equally effective in this setting
Randomised trials that also incorporate patient preference provide greater evidence of the external validity of the trial results
We thank all patients and staff in participating practices in London and Manchester for taking part in this trial. We also thank the therapists who took part at both centres and the psychological services involved in Manchester. We acknowledge the contributions of Karin Friedli in London; Shirley Halliwell, Jenny Hacker, Adrian Wells, and Ron Siddle in Manchester; and David Torgerson in York.
Contributors: MK, BS, ML, JA-H, and NT conceived the idea for the trial and obtained research funding. MK, BS, ML, and NT supervised the conduct of the trial and data collection. EW, ML, PB, SF, and MG undertook recruitment of practices and patients and conducted the data management. MK, PB, and BS analysed the data. All authors contributed to the writing of the paper. MK is the guarantor for the study.
funding The trial was supported by a grant from the NHS Executive, Health Technology Assessment Programme. The views expressed in this paper are those of the authors only and are not attributable to the Department of Health.
Competing interests None declared.