Which trials should include a QL assessment

Whenever a trial is planned, consideration should be given as to whether it is appropriate or not to include an assessment of QL (see Fig. 6.2). Situations where some aspect of QL assessment maybe important are randomized trials in which:

♦ survival is expected to be similar, but the treatment modalities and/or other aspects such as toxicity, are expected to be very different in the treatment groups under study,

♦ a small survival benefit is expected with the experimental treatment which may be associated with increased toxicity,

♦ slightly worse survival is expected with the experimental treatment which may be balanced by substantially less toxicity or greater patient acceptability,

♦ palliation is the primary outcome measure.

Fig. 6.2 A suggested algorithm for deciding which trials might need QL assessment.

However, these scenarios cover most randomized cancer trials. Indeed, as long ago as 1985, Brinkley, in a BMJ editorial [30], stated 'any evaluation of palliative treatment, especially when included in clinical trials, should always include an assessment of the side-effects and of the general effects on the patients' life'. In addition, a couple of years later, Slevin et al. [31] stated 'the question is no longer whether QL should be measured, but what is the most reliable and practical method of obtaining these data.'

A much more pertinent question is perhaps 'What trials do not need QL?' and organizations such as the UK MRC now require justification in trial proposals as to why QL should, or should not, be included [32]. Perhaps the only clear situations in randomized trials where some assessment of QL might not need to be considered is where:

♦ the survival benefit with a new treatment is expected to be very large,

♦ both survival and QL are expected to improve,

♦ patients are likely to accept a chance of cure no matter how much toxicity is increased and/or QL decreased,

♦ standard therapies are being used in which QL is well documented.

Sadly none of these circumstances in cancer are common.

The case for QL measurement in non-randomized studies is much weaker. Nonrandomized studies can provide an excellent test-bed for checking that the correct questionnaires are being used, for deciding the best times to administer them, and for checking whether the questionnaire covers all relevant treatment-related effects. However, the potential biases in patient selection, and response-shift, probably mean that QL should not be used as an outcome measure in non-randomized studies as a means of assessing treatment effect as it maybe unclear whether observed changes are related to treatment or patient adaptation.

The theory and practice of including QL are not the same thing. The assessment of QL has considerable resource implications and these need to be balanced against the usefulness of the data and its likely impact on recommendations for treatment choice following completion of the trial [33]. The considerable burden on the research team in terms of resource required should not be overlooked as QL data need to be collected, collated, processed, checked, queried, and missing data pursued and analysed [32].

Whenever it is considered appropriate to include QL in a trial, it is vital that it is seen to be an integral part of the trial and not an afterthought [34]. In particular, the assessment of QL should be written into the protocol. The value of this was shown by Scott et al. [35] who reported that in trials conducted by the Radiation Therapy Oncology Group (RTOG) compliance rates were 60-90 per cent in trials where QL was integrated into the trial design, and only 20-40 per cent where QL was run as a separate companion trial.

0 0

Post a comment