This pragmatic RCT is notable in that through a high-quality design it evaluates a previously under-researched treatment, namely medium to long-term psychoanalytic psychotherapy, which aims to reduce the underlying vulnerability to depression in people suffering from a particularly complex and enduring form of this disorder. Currently, there is only one trial with a group of patients of a roughly comparable complexity [111
]. A positive finding in our trial would significantly improve the options available for patients with this expensive and relatively poorly-served condition, which also represents a considerable burden worldwide [60
However, as we have argued, the features of both the disorder and the treatment strain the limits of methodology in classical randomised clinical trials. We decided that the trial should be ecological, namely that it ought to offer information about the actual outcome over the medium to long-term of a clinically significant patient sample in the NHS. Many of the trial’s design choices followed from this decision. In this section, we discuss how we have tried to mitigate some of the strains, and list some inevitable limitations.
The patient sample
In relation to the diagnosis of depressive and other common mental disorders, we do not yet know how to carve nature at the joints. If our study were to meet its primary requirement of ecological validity, the sample selected had to involve the fewest possible assumptions. We therefore defined the sample operationally, and included participants with co-morbid Axis I and II disorders as well as problems with physical health. Inevitably, this leads to a heterogeneous sample that includes many unknown variations capable of affecting both outcome and the interpretation of findings. Although PPD has the advantage as a treatment of being adaptable to each individual’s psychopathology, we still expect a large amount of residual variation in responsiveness amongst different patients. The treatment under test will help some participants more than others, and these variations may cancel each other out [116
]. In this context, examining mean outcomes, for example, will be insufficient, and while modelling trajectories of outcomes using mixed-effects growth curves allows for differential treatment responses to be investigated, this is only possible when the critical moderators involved first have been identified and then measured. In the absence of predictions derived from well-founded theories of pathology, this is a hit or miss way of proceeding. The poor performance of outcome research in answering the question “what works for whom?” suggests that this approach might be misconceived [117
]. It may be necessary to be able to drill down through group data to discover what may be deeper structures in the individual in order to understand the basis of differential degrees of responsiveness/non-responsiveness. Thus we have collected qualitative data using both psychoanalytic and phenomenological methods.
The treatment as usual condition
A TAU control condition was chosen because it is an ecologically sound comparison group. TAU control conditions have been widely used in outcome trials, although their rationale, and what any given TAU condition may involve, has varied from case to case [25
]. Our TAU control condition is only partly determined by the management protocols for these patients specified in NICE guidelines. In actual practice, there are widespread variations in the mental health resources available in different locations (NHS Atlas) and, in our experience, these are especially marked in the treatment that TRD patients receive. Some may receive exemplary care through their GPs, including social support, community mental health teams, brief counselling, and pharmacological and psychological interventions of appropriate types, durations and intensities. Others may get little if any help beyond token pharmacological treatment, with little case management. Some, perhaps following many failed attempts at treatment, may have withdrawn from all treatment endeavours. They feel they have exhausted all treatment possibilities and end up quite seriously neglected and deprived. As we have also suggested, certain characteristic features of psychosocial functioning may be integral to the TRD disorder. By adversely affecting important help-seeking behaviours, these actively exert negative effects on service provisions. These considerations lend weight to the decision to study this naturally occurring control and comparison condition [118
A long-term trial involving patients with severe depressive disorders operates against a background of a continuous level of chronic suicidal risk punctuated by exacerbations and crises. Our guiding principle is that the TAU participant’s clinical needs must always come first. The senior research clinician and the trial’s clinical director explicitly assume responsibility for participants’ safety and have the authority to intervene and mediate in a participant’s care. In most instances this does not conflict with one of the needs of research - for instance, to keep as many subjects in the trial as possible for as long as possible. However, it follows that the trial’s TAU condition is not always and not only Treatment As Usual. Moreover, the consistent, reliable contact involved in the researcher’s administration of research measures necessarily functions like a low-intensity, psychodynamically informed supportive intervention.
TRD participants vary in their resilience to the significant stresses that research involves. One of the most potent stresses is randomisation powerfully stirs up feelings of rejection especially in those who already have histories of severe loss or abandonment. Subjects who had strong wishes for the PPD treatment feel disappointed, hurt or angry when the allocation is not that for which they hoped. These subjects may then drop-out. Others drop out because of their fears about what treatment may involve. When participants are told of their treatment allocation, it is important that the research staff show sensitivity to the situation. The psychoanalytically-trained research clinicians allow space in the initial clinical interview for the subject to express how they felt about the allocation they received. As well, the research clinician has to be mindful of the underlying feelings aroused by the long research process especially with those who were randomised to the TAU control.
The need for best research practice
Over the longer-than-usual time of this trial, we considered there might be a greater tendency for subjects to drop out. To combat this, continuity and absolute best practice in the research team need to be ensured. Researchers needed to be reliable, persistent, flexible, sympathetic and responsive with all the subjects. Newsletters and other communications about the trial can be used to help subjects remain aware of the fact that they are helping in a valuable endeavour.
However, most research staff are not clinically qualified and are often at the beginnings of their career. It is therefore essential that these research staff are provided with sufficient clinical supervision by qualified research clinicians. The supervision has three functions. First, it gives the research staff a chance to process the considerable emotional impact associated with the way that subjects often powerfully convey their feelings. Second, it helps researchers to reconcile subjects’ ongoing communication of their needs with the researcher’s task of equipoise, while still helping subjects stay in the trial. Third, it gives the research clinician an opportunity to keep in touch with each patient’s condition.
The limitations of the trial
When compared with other long-term psychoanalytic psychotherapy (LTPP) outcome studies our N of 129 may be regarded as large, but considered statistically the trial is somewhat under-powered. Unfortunately, the very large and expensive treatment service resources necessary for a significantly bigger sample are simply not available. A second limitation is to be found in the impossibility of concealing treatment allocation from research interviewers. We seek to mitigate this by double-rating the primary outcome measure using independent raters working with anonymised recordings. Another limitation is the inherent variability both of our test treatment and of the TAU comparison condition. Although our design includes a treatment manual, assessment of adherence, and careful recording of the TAU care control patients actually receive this still does not permit the kind of quasi-experimental assessment of the impact of the components that are possible with less complex treatment procedures. Finally, in a long-term trial with patients of this kind the inevitably high rate of attrition could easily cause a statistically unacceptable standard mean error [119
In this discussion we have considered only some of the issues in long-term randomised trials of complex treatments for complex mental health conditions. We opted for this composite, ecological design as a first stage study in which a standard RCT was complemented by evaluative methods in order to produce in-depth models helpful in understanding the condition itself, as well as with the development of more optimal treatment and case management. We want to emphasise the importance of research procedures capable of meeting the challenges posed by the disorder. Some knowledge of the complex nature of the disorder is also necessary to properly evaluate both the structure of the research, and its eventual findings.