Need for rigorous assessment of palliative care

By a strange coincidence, political and professional agendas simultaneously require evidence of effectiveness. This makes good sense with finite resources. Stop ineffective interventions, and the money saved can be spent on those that are effective. The problem is that, although effectiveness is a useful word to describe what is wanted, it is difficult to define. Its meaning varies according to context.

In medicine we can at least use the gold standard of the randomised controlled trial to show that a new medical intervention is better than no intervention or than an existing intervention.1 Long established interventions are often time honoured rather than proved by randomised controlled trials, although lack of evidence of effectiveness. is not the same as evidence of lack of effectiveness. Will medical decisions not supported by randomised controlled trials continue to be purchased, and, if so, for how long? Are we wrong to continue to support these decisions? Who will investigate their effectiveness? And should one randomised controlled trial showing lack of effect outweigh the testimony of experience?2 If this is hard going for single interventions how much harder is it for a service in which multiple interventions are provided?

For such services we assume that randomised controlled trials are again the gold standard to establish effectiveness, even if the interventions that make up the service are not proved by such trials. The provision of “hospital at home” services is a good example. There are no randomised controlled trials of hospital at home. Despite this the four counties of the former Oxford Regional Health Authority will spend pounds sterling 1.6m on hospital at home in 1994-5. The design and conduct of a randomised controlled trial funded jointly by Northamptonshire Health Authority and Anglia and Oxford Regional Health Authority initially presented formidable problems (S Shepherd, personal communication).

This week's journal reports the failure of an attempt to evaluate a palliative care service with a randomised controlled trial (p 1340).5 The trial was done to “strengthen the case for continued funding.” It failed because of difficulties in recruitment and problems inherent in the design. The wider issues raised by the authors are important. First is the idea that doing a randomised controlled trial in palliative care is so difficult that palliative care, and by analogy any other specialty in which it is difficult, should be “excused” from doing randomised controlled trials. The second is that the rarity of successful randomised controlled trials in health services research - they quote seven adequate trials out of 1000 - suggests that randomised controlled trials are not the right way to show that a service, a package of interventions, is an improvement in terms of either where it is provided or what is provided.

Should palliative care be excused from randomised controlled trials? The design of McWhinney and colleagues' trial did not make life easy. The authors had been providing their palliative care at home for 18 months before the trial began and, not surprisingly, believed that it was better than no service. Patients were randomised either to immediate support or to support delayed for one month. For those randomised to the delayed group care was meant to be provided by the family doctor, who had referred the patient in the first place. The authors measured pain, nausea, patients' quality of life, and care givers' health at the beginning of the trial and at one month. Patients expected to live more than two months were approached, yet 36 of the 146 who were randomised died within the month of the study. Others, who were not entered because they were not expected to live for two months, did so; such problems will be familiar to those who do randomised controlled trials in palliative care.

Clearly the failure of this study alone would not justify excusing palliative care from randomised controlled trials. The authors go on to raise the problems of the ethics of randomisation in palliative care in general. As they say, using patients on the waiting list as controls may be ethical in chronic stable conditions, but in this study, and in palliative care in general, randomisation to treatment versus delayed treatment is hard to justify. This does not mean that all randomisations in palliative care are unethical. Comparison of two interventions, neither of which is known to be superior, would be the classic equipoise requirement for ethical randomisation. Two constructive options are that multicentre randomised controlled trials of particular interventions can circumvent the problem of small numbers in a single centre and that some interventions lend themselves to study by randomised controlled trials in single patients (n of 1 trials).4 5 6 In this particular case randomisation to either the full intervention or to some component might have worked better.

The provision of a service is much more subjective than the provision of an intervention. It depends more on local conditions and, indeed, on local enthusiasms. What works in one set of circumstances may not be right for another. Health care consumes a large segment of our national wealth and touches the lives of everyone. The time has come when we should put an end to the casual use of “effectiveness” as an easy way out and begin to take a harder line. If you say that it is effective then prove it. As Archie Cochrane wrote more than 20 years ago, “There will be a marked reduction in the use of ineffective remedies and of effective remedies used inefficiently.”7