Presentation and analysis of the results of clinical trials in cardiovascular disease.

During the past few years numerous clinical trials have been conducted to test the efficacy of various agents in reducing mortality after myocardial infarction. There is considerable confusion about the results of these trials, and no drug or group of drugs is universally accepted by clinicians as being useful for preventing a second infarction. To some extent this confusion results from differences in trial design and to some extent from the of different types of patient or from treatment being tested during different stages of the illness. Such differences between trials are inevitable and to some extent useful. Never? theless, it is quite unnecessary for confusion to result from a lack of clarity in the presentation of the results of a trial, or from disagreements about the analysis of these results. Such confusion could be prevented by a standard method of data presentation that makes it clear what was done, makes it easy for one trial to be compared with another, and allows those not directly con? cerned with the trial to analyse the results in the way they believe most appropriate. Figure 1 shows such a method. Presentation of trial results The box on the first line of fig 1 shows the total population from which the patients in the trial were selected. It is essential to know this total, for without it it is impossible to be sure about the applicability of the results of the trial. In any study there are inclusion and exclusion criteria that superficially appear straightforward, but which may in fact be difficult to apply; other factors, such as the willingness of patients and doctors to co-operate with the study, will also influence the number eventually admitted to the trial. It is characteristic of multicentre trials that the size of the original popula? tion of patients is unknown. The second box contains the total number of patients included in the trial. A comparison of the numbers in the first two boxes will show the proportion of patients considered suitable for the trial, and this will give an indication of the general applicability of the final results. It is highly improbable that any treatment will reduce the risk of death after myocardial infarction by as much as 30%, yet few pub? lished trials have included enough patients to show with conventional levels of statistical significance such a degree of benefit from treatment. If the number in the box on line 2 appears to be small an explanation must be sought in the text of the paper, and if no adequate reason can be found then the result of the trial must be regarded with some suspicion. The. number of patients randomly allocated to each treatment is shown in the pair of boxes on line 3 of this figure. If the treatment groups are small the text may be scrutinised for evidence that they were well matched, but if these groups of patients were large then good matching is likely. The boxes on line 4 show the number of patients randomly allocated to each treatment group who continued on that treatment, and the number who were withdrawn from it. In any clinical trial with a double-blindᵈesign a patient; may be withdrawn from the study if his doctor believes that some undesired event may result from active treatment, or if the doctor believes that active treatment is necessary and fears that his patient may be on a placebo. For example, in a trial using beta-blockers after myocardial infarction hypotension may lead to a patient being withdrawn lest it results from beta-blockade, while the appearance of angina may lead to a patient's withdrawal so that he can electively be treated with a beta-blocker. In some postinfarction trials, and in particular in those when medication was started early in the illness, when symptoms and signs are changing, withdrawal rates have been high in both treated and placebo groups. If the withdrawal rate among patients receiving active treatment is much higher than it is among the patients on placebo the active treatment is unlikely to be useful.