Intended for healthcare professionals


Rigorous assessment of palliative care revisited

BMJ 1999; 319 doi: (Published 04 December 1999) Cite this as: BMJ 1999;319:1447
  1. Duncan Keeley, general practitioner
  1. Thame, Oxfordshire OX9 3JZ

    Wisdom and compassion are needed when evidence is lacking

    General practice p 1472

    In 1994 McWhinney and colleagues reported the failure of an attempt to assess a palliative care home support team using a randomised controlled trial.1 Problems included attrition of study subjects because of early death and difficulties with recruitment, consent to randomisation, data collection, and timing of outcome assessment. In an accompanying editorial McQuay and Moore argued that assessment of palliative care packages by randomised trials remained mandatory despite these difficulties.2 Skating around the ethical problems of randomisation, they restated the evidence movement's demand for proof of effectiveness. Five years down the line, are we any further forward?

    Smeenk et al reported a systematic review of the effectiveness of home care programmes for patients with incurable cancer in 1998, finding nine prospective controlled studies (five randomised) which they judged methodologically adequate.3 Eight of the nine studies were from the United States; methodological quality was rated as moderate overall; and, though positive influences on various dimensions of outcome were found in a minority of studies, the overall effectiveness of such programmes was thought to remain unclear.

    Today the BMJ publishes a further valiant attempt to conduct a satisfactory randomised trial of a home based palliative care service (p 1472).4 The Cambridge Hospital at Home offers up to two weeks of 24 hour nursing care to patients nearing death. The study compared 186 patients randomised to the intervention with 43 controls on the basis of intention to treat. What does it tell us about the feasibility of randomised controlled trials in assessing palliative care, about the effectiveness of this particular intervention, and about the provision of palliative care services in general?

    A power calculation indicated the need to recruit 180 patients to each arm of the trial to detect a 15% difference in the main outcome measure, place of death, with a probability of 80%. But the two week limit to the service (a service resource constraint) together with ethical and presentational considerations combined to necessitate a 4 to 1 randomisation ratio and thus reduced the power of the study to show differences. By the time a patient is likely to die within two weeks he or she may well die sooner, and neither the referring doctors nor the providers of the service wished to be placed too often in the position of withdrawing a desirable service before the patient's death. The demand for this service did not sufficiently exceed its capacity to allow the full number of controls to be recruited. It was considered unacceptable for the service to operate below capacity purely for the purposes of the study.

    Fewer patients than intended having been randomised, 39% of the intervention group were admitted to hospital or died before they could receive the intervention. Control patients who did remain at home had access to a range of alternative nursing services unlikely to be equally available in all parts of the United Kingdom. It is not easy to tell how different the intervention programme was. An excellent service offered in addition to a good service has less chance of showing its worth.

    Did the study show benefit from the intervention? On an intention to treat analysis there was no difference in the proportion of patients dying at home. Overall there is an impression of a positive effect from the intervention, but the methodological difficulties reduce confidence in the validity of the findings. The findings of a comparison of process measures and the quality of death between intervention and control groups are to be reported elsewhere, but the problem of low study power will still apply and this, together with the possibility of bias due to the unblinded assessment of a desirable service, makes clear conclusions unlikely.

    Yet this is a good attempt at a randomised controlled trial of a palliative care service. Most of the methodological objections could have been overcome only by more resources for either the service itself (available longer than two weeks) or the study (longer duration) Resources are scarce, particularly for caring as opposed to curing, and for this service funding was short term and partly contingent on the results of this trial. It is questionable whether services should be asked to prove their worth by methods that cannot properly be achieved within the constraints of the funding made available. The authors of the study briefly suggest some possible alternative modes of assessment in their discussion.

    There is an important general question. Can we insist on evidence of effectiveness from randomised controlled trials for support services which are of such evident human desirability as to render their deliberate withholding difficult or unethical? If we want to know whether five years of pill taking for high blood pressure will reduce our risk of stroke we are unwise to rely on our own or our doctor's common sense impression and want evidence from randomised controlled trials. But if our loved one is dying and wishes to die at home, how much evidence do we need that skilled home nursing available round the clock would be a good idea?

    For commissioners of care making difficult decisions on resource allocation amid competing priorities will the results of the study we publish today help them decide how best to organise palliative care services? They may note the pointers toward improved outcomes when such a service is provided but also reflect on the difficulties inherent in providing the service on a time limited basis. They may wish to be careful that funding for improving home palliative care along these lines does not come from the same pot as that which pays for the home care services which will carry on if needed when the two weeks is up But in balancing the benefit to be had from this possible intervention compared with others they will have to fall back on wisdom, compassion, and a broad and inclusive conception of the role of health care services in reducing human suffering.


    1. 1.
    2. 2.
    3. 3.
    4. 4.