The importance of setting hypotheses

The easiest way of determining whether it is appropriate to include QL in a trial is to discuss with all relevant parties the likely impact of the treatments on QL and to form a hypothesis about the likely differences to be observed. Such a hypothesis needs to be clinically important, and should relate to the difference that would not only be perceptible by patients but considered sufficient to affect the choice of treatment. Setting the hypothesis for the QL evaluation is as important as setting the principal hypothesis for the trial. If the hypothesis is well framed, this can facilitate the design of the whole of the QL component of the trial, from the choice of questionnaire, to the timing of administration, the sample size required, and the analyses to be performed. Failure to define a hypothesis will mean that the QL aspects of the trial will lack focus, and any results may be open to the accusation of data trawling.

How do we set about defining a hypothesis? The key is deciding what the treatments under study are expected to achieve and how this will impact on QL. For example, a new treatment may be expected to relieve pain better, or may be expected to be more acceptable to patients, or alternatively it may be expected to cause nausea. To find the key, refer to the literature, survey clinicians and/or talk to patients who have had the treatment. A phase II pilot or feasibility study may be the ideal place to clarify these issues.

Groenvold and Fayers [36] suggested that a survey of experienced clinical staff may identify aspects that new treatments should affect. They surveyed doctors and nurses as a method of defining key QL areas in a trial of a combination chemotherapy regimen. Each respondent was asked to predict for each of the subscales and items on all the QL questionnaires to be used, which symptoms would occur more in the combination chemotherapy group, which in the no-treatment group, or those where there would be no difference. Thus, QL hypotheses could be formed around the main predicted differences in treatments.

0 0

Post a comment