Identifying key timepoints

In a paper exploring alternative methods of defining and analysing palliation, Stephens et al. [43] demonstrated the impact of assessment timepoint on outcome. Using data from a trial of 4-drug versus 2-drug chemotherapy in SCLC they showed an advantage in terms of palliation of cough in favour of the 4-drug (4D) regimen at one and three months, but in favour of the 2-drug (2D) regimen at two months (Table 6.1).

This finding emphasizes the difficulty of identifying a single key timepoint when designing the trial and how information from a pilot study might have been useful. Although a frequently suggested approach is to choose the timepoint at which the greatest difference is likely to be seen, this seems to be illogical and it may be more logical to provide a more complete summary and choose enough timepoints to give the full picture of advantages and disadvantages. This is analogous to comparing the whole survival curves, not just the point at which differences are greatest. The selection of the primary outcome measure, which is necessary for sample size calculation, is therefore difficult, and rather than choosing a single timepoint, an alternative method, such as looking for the maximum or minimum score over a set time period, or time to an event, might be preferable.

Table 6.1 Improvement in severity of cough at different timepoints (Stephens et al., [43]).

Time from randomization

Number and percent of patients with an improvement Regimen

p value



1 month

40/64 (62%)

31/76 (41%)


2 months

18/36 (50%)

30/50 (60%)


3 months

16/25 (64%)

12/32 (38%)


Even if to help get a sample size for the QL comparison of the trial we concentrate on a single key timepoint, it is usually important in practice is to administer questionnaires at regular intervals before, during and after treatment, to obtain a more complete picture.

Most standard questionnaires use a time frame of three weeks, i.e. they ask the patient 'how have you been feeling over the last three weeks?' It has been found that this is the longest period over which patients can recall accurately. This has important repercussions on how often questionnaires are completed. Thus, if the first post-treatment questionnaire is administered many weeks or months after treatment, it is unlikely that an accurate record of any transient treatment side effects will be obtained.

The actual timing of questionnaire administration will depend on the hypothesis. However the following guidelines should be noted.

♦ A baseline questionnaire is essential. Ideally in a randomized trial this should be administered after the patient consents to be in the trial but before the actual treatment is known. In this way all possible bias in terms of patients being pleased or anxious about their allocated treatment can be avoided.

♦ Most randomized trials are based on the concept of intention to treat (i.e. what happens to patients if a policy of giving a particular treatment is applied). QL should follow this same concept. The most logical assessment points are therefore at set timepoints from randomization. In this way all patients can be accommodated, even though patients may be at different points in their treatment, or indeed withdraw from treatment.

♦ Completion of questionnaires at times related to treatment or treatment-related events may seem most logical but may cause difficulties in analyses, because, for instance, the duration of treatment can vary from patient to patient, and indeed some patients may not receive treatment. In this way differences between treatments are confounded by differences in times of assessment.

♦ When choosing timepoints it is important to enable valid comparisons to be made. If, for instance, treatment A is given at 3-week intervals and treatment B at 4-week intervals, it would not be logical to assess QL at eight weeks as patients on treatment A would be between treatments and those on treatment B would be at the time of treatment. It would be much more logical to choose twelve weeks when both groups of patients were at equivalent points in their treatment course.

Thus, although choosing the timepoints to administer QL will depend on the hypothesis, some common sense needs to be applied as, for instance, it would not be justifiable to ask patients to attend clinic purely to complete QL questionnaires and/or to synchronize QL completion between treatments [32].

0 0

Post a comment