Patient allocation

Several trials of olanzapine for prophylaxis of manic depression have now been conducted, funded by Eli Lilly, and claim to show positive evidence of efficacy. A comparison with lithium demonstrated little difference between the two treatment groups overall, although olanzapine appeared to be better at preventing recurrence of mania (Tohen et al. 2005). The only placebo-controlled trial that has been published showed superiority of olanzapine, but there was evidence of a discontinuation effect. Fifty per cent of the placebo group relapsed within 22 days of randomisation and almost all the excess risk of relapse was confined to the first three months of the study (Tohen et al. 2006). All patients were treated with olanzapine initially and since no gradual discontinuation schedule was mentioned, it appears that it was stopped abruptly at the point of randomisation for patients allocated to placebo. 

A fuller description of die methods used in the APT overview can be found elsewhere (8), but Figures 27.1 and 27.2 respectively summarise the derivation of a proportional od leduction and corresponding absolute risk reduction. In Figure 27.1 the overall typical odds ratio for patients with a prior MI is the ratio of the odds of a serious vascular event among patients allocated antiplatelet treatment to the corresponding odds among control patients. This typical odds ratio of 0.75 corresponds to a typical odds reduction of 25%. 

The respiratory and cardiovascular adverse effects of topical therapy with timolol or betaxolol have been studied in a randomized, controlled trial in 40 elderly patients with glaucoma (83). Five of the 20 allocated to timolol discontinued treatment for respiratory reasons, compared with three of the 20 patients allocated to betaxolol There were no significant differences in mean values of spirometry, pulse, or blood pressure between the groups. This study confirms that beta-blockers administered as eye-drops can reach the systemic circulation and that serious adverse respiratory events can occur in elderly people, even if they are screened before treatment for cardiac and respiratory disease. These events can occur using either the selective betaxolol agent or the non-selective timolol. 

Post-thoracotomy pain can be treated with thoracic epidural or thoracic paravertebral blockade. In 100 adult patients allocated to receive one of these treatments with preoperative bolus doses of bupivacaine followed by a continuous infusion there was less postoperative respiratory morbidity and significantly better arterial oxygenation in the paravertebral group nausea (10 versus 2), vomiting (7 versus 2), and hypotension (7 versus 0) were more problematic in the epidural group (346). 

A difiSculty frequently arises in clinical trials because not all patients allocated to a treatment will complete the designated period of treatment or follow up. Non-completion may be a function of the particular treatment allocated, in that the reason for discontinuation of therapy may be inadequate efficacy or an adverse effect. If non-completing patients are omitted from fhe treatment comparison, bias may be introduced. Simply, if a treatment s failures are omitted from further consideration, the treatment will appear to be better than it really is. Thus, the primary efficacy analysis for the comparison of treatments must include all patients allocated to each treatment, whatever their ultimate fate. This is known as intention-to-treat (ITT) analysis. The implementation of an ITT analysis is not always easy, but usually either an endpoint analysis (in which the final assessment, whenever it was made, is used for fhe treatment comparison) or one of the strategies based on ranking the patients (from the best to the worst in the study) proposed by Gould. ... 

Table 6.5 Number of patients allocated to the inferior treatment by trial design and eventual outcome.

Obviously, such a design will, in the long run, tend to allocate more patients to the successful treatment. In fact, if 0a is the probability of success on treatment A and 0g is the probability of success on treatment B, then, for a large trial, the expected proportion of patients allocated to A will be Pa = (1 b)/(2 — 0a 0b)- So, for example, if the probability of success with treatment A were 0.8 and that with treatment B were 0.4, so that 0a = 0.8 and 0g = 0.4, then the proportion of patients allocated to the superior treatment, A, would be 0.75 and the proportion allocated to the inferior treatment would be 0.25. The overall proportion of successes for such a trial would be 0.75 X 0.8-1-0.25 x 0.4 = 0.7. This may be compared to the proportion of successes for a randomized group trial with equal numbers on each treatment, which is clearly (0.8-b0.4)/2 = 0.6. 

Now, if we wish to study the efficiency of a symmetric randomized cut-off with patients allocated to control, randomization, or experimental treatments according to Z < -a, -aa, then we simply need to substitute x = 0 and ct = 1 in (6.6) to obtain the expected mean standardized baseline values in the extreme experimental groups. (The expected value in the middle group is, of course, 0 and the value in the lower group is simply the negative of that in the upper.)... 

A.3 The proportion of patients allocated to the superior treatment in a play the winner trial... 

Comparative studies In a 12-week randomized controlled trial in 276 depressed patients allocated to agomelatine 50 mg/ day or venlafaxine (titrated to a target dose of 150 mg/day), of those randomized to agomelatine 20% reported treatment-emergent adverse effects, the most common being nausea (12%), headache (10%), and upper respiratory tract infections (7%) 2% withdrew because of adverse effects [67 ]. The rate of treatment-emergent sexual dysfunction (reduced libido in males and impaired orgasm in females) was lower than in those who took venlafaxine. 

---