Box 93 How to use patient characteristics in the analysis of a randomized trial

♦ When designing the trial, from previous trials identify characteristics of prognostic value. From these, identify a small number (perhaps no more than five) of the most important characteristics which can be included as 'stratification' characteristics in the randomization process (see Chapter 4).

♦ For all the characteristics identified - and only these characteristics - at the beginning of the trial, include them in a stratified/adjusted analysis (either using simple stratified analyses or an appropriate model - such as the Cox model).

♦ State whether the adjusted analysis or the unadjusted analysis is the primary one; for large trials an unadjusted analysis is probably appropriate, while for small trials a (prespecified) adjusted analysis may preferable.

interest focused on two components of the treatments given, the total dose and the dose intensity of CAP and carboplatin treatments achieved across patients.

For total dose, for each patient and each drug we calculate the actual total dose given and divide it by the planned protocol total dose that should have been given if the treatment had been given in full. Figure 9.3 shows how these data might be presented. The two panels on the left represent the total dose achieved of the two treatments and they are to be interpreted as follows. Fifty eight per cent of patients allocated single agent carboplatin actually received 100 per cent or more of their planned total dose, while 90 per cent of patients received 60 per cent or more of their planned total dose. This graph of total dose tells us about reductions of the dose of drugs, but does not give any information on delays in giving the drugs. This is done in the right hand panels of Fig. 9.3 and points are calculated as follows. For each patient and each drug we consider the total dose given and note the total time taken to give this dose, td. This is then divided by the total dose which should (theoretically) have been given over the this total time, td.

If a considerable proportion of patients did not receive their allocated therapy then the reasons for this should be given. For example in the BA06 bladder cancer trial of the 491 patients assigned chemotherapy, one patient received four cycles, 392 received three cycles, thirty-seven received two cycles, thirty-three received one cycle and twenty-eight received no cycles. Thus ninety-nine patients did not receive all three cycles of chemotherapy and the reasons for this are set out in Table 9.7.

0 0

Post a comment