Period, with cross-over trials

The cross-over trial was mentioned in the previous section as one example of a within-patient design. In order to discuss some issues associated with these designs we will consider the simplest form of cross-over trial two treatments A and B and two treatment periods I and II. 

In this context, each patient would be receiving each of the multiple treatments. In the cross-over trial with three treatments this would likely be a three-period, three-treatment design and patients would be randomised to one of the six sequences ABC, ACB, BAC, BCA, CAB or CBA. Although there are again ways of asking a simultaneous question relating to the equality of the three treatment means through an analysis of variance approach this is unlikely to be of particular relevance questions of real interest will concern pairwise comparisons. 

Schulz and Jobert [229] investigated the influence of the Hypericum extract LI 160 on the structure of sleep. Twelve elderly female volunteers were included in a double-blind cross-over trial and treated in random sequence with 3 x 300 mg per day of Hypericum extract and placebo. Both treatment periods lasted 4 weeks with a 2-week washout period in between. Eight-hour sleep recording sessions took place during the night before the start and following the last day of each treatment period. 

Now suppose, however, that there is a secular trend causing treatment values to increase by 100 ml in the second period compared with the first. We then need to take the values in the second period in Table 17.2 and add 100 to reach the position in Table 17.3. We see that the treatment effect of 300 ml (which still applies) is not recovered from each sequence. However it is recovered from the average of the two sequences, so that the cross-over trial is still capable of being used in these circumstances. 

Some rather bizarre recommendations have been made in the statistical literature by statisticians who are unfamiliar with the way cross-over trials are actually run. For, example, both Jones and Donev and John et al. 0ohn et cd., 2004 Jones and Donev, 1996) have considered the possibility of modifying the subsequent schedule of treatments in a multiperiod design once the results from the first two periods (say) have been studied. 

Some clinical trials can be completed with only a few visits. Others require more frequent contact with the study staff. As an example of the former, our clinic has conducted many studies intended to assess blood insulin and glucose responses to test products such as snack bars and beverages. These are usually conducted using a cross-over design and may require only three visits one for screening, one for consumption of the control product, and one for consumption of the active product. In contrast, we have also completed several trials to assess dietary and pharmaceutical interventions intended to promote weight loss. These usually require frequent clinic visits over a period of at least 12 weeks and sometimes as long as two years. 

Ondansetron has also been investigated for the treatment of fatigue associated with primary biliary cirrhosis but results have been disappointing. A randomised, controlled crossover trial (n = 54) examined the effect of ondansetron 4 mg three times a day versus placebo for a four-week period, before being crossed over for a further four-week period. The study concluded that the use of ondansetron did not offer a clinically significant reduction in fatigue compared to placebo [20]. However, the results of the study may have been affected as patients were effectively unblinded during the second phase of the... 

Johnson [226] included 12 healthy volunteers in a double-blind cross-over EEG trial. In a randomized sequence, each subject took 2 x 300 mg per day of the Hypericum extract LI 160 for six weeks and placebo for the same period of time, with a two-week wash-out period in between. The resting EEG and evoked potentials were registered 2.5 hours after drug intake. 

Furthermore, where there are true differences between patients, a more efficient way to gain information about the effect of a treatment with the object simply of establishing that it works would be to run an AB/BA cross-over. There is more information to be gained about the average effects of treatment under such circumstances by adding more patients rather than more periods. Hence in phase II there is generally little point in running n-of-1 trials. 

Edwards et al. [22] conducted a controlled, randomized crossover trial in 10 subjects with mild hypercholesterolemia in which subjects served as their own controls. For 3 weeks, subjects either consumed their regular diet or substituted 20% of their daily calorie intake with pistachio nuts after the first 3-week period, those on the pistachio diet crossed over to consume their regular diet, and those on the regular diet crossed over into the pistachio diet. At the end of the study, subjects experienced significant decreases in total cholesterol and total cholesterol/high-density lipoprotein (HDL) cholesterol ratio, with significant increases in HDL cholesterol. 

The study followed a cross-over design with two successive 10 day intervention periods, with 48 g TFA/d and 24 g TFA/d intake, respectively. The two intervention periods were separated by a washout phase to reduce the sample size needed for the trial. TFA was provided in the form of biscuits and spreads in addition to the normal diets. 

The cascade of apoptosis may be promoted by mitogen activated protein (MAP) kinase activation. CEP1347, a small molecule MAP kinase inhibitor that crosses the blood brain barrier is protective in MPTP-treated mice (Saporito et al., 1999). A clinical trial showed good tolerability, no interference with L-dopa pharmacokinetics, and no symptomatic effects of treatment over a 4-week period (Parkinson Study Group, 2004). A larger phase II trial has commenced. 

---