Clinical Trials

Clinical Trials: A Practical Guide ■❚❙❘Table 4. Number of patients allocated to each treatment in the GISSI-Prevenzione trial [4].n-3 polyunsaturated fatty Vitamin E supplementation Totalacids supplementation Placebo ActivePlacebo 2,828 2,830 5,658Active 2,836 2,830 5,666Total 5,664 5,660 11,324Other means of evaluating an interaction effect can also be used, such as theMantel–Haenszel method [1–3]. An interaction test is sometimes called ahomogeneity (or heterogeneity) test. For example, in a multicenter study, if thereis no significant interaction between treatment and center, then the treatmenteffect is said to be homogeneous across different centers and an overall treatmenteffect can be obtained by pooling all centers together. On the other hand, if asignificant interaction is found between the treatment and the center, then thetreatment effect is heterogeneous across different centers. Under suchcircumstances, pooling the estimate of treatment effect together from differentcenters might produce a misleading overall result.Examples of types of interaction effects seen in clinical trialsThe objective of evaluating an interaction effect in a clinical trial is to assesswhether the treatment effect is the same among different levels (values) ofanother factor or factors. The factor might be the other drugs under evaluationin a factorial design, or some stratification variable – such as severity of theunderlying disease, gender, or other important prognostic factors. In the followingdiscussion, different types of interactions will be described in different clinicaltrial settings.Example 2: Treatment-by-treatment interaction in a factorial designThe GISSI-Prevenzione (Gruppo Italiano per lo Studio della Streptochinasinell’Infarto Miocardico Prevenzione) trial investigated the effects of vitamin E(α-tocopherol) and n-3 polyunsaturated fatty acids (PUFA) supplementationin patients who had recently suffered a myocardial infarction [4]. In this trial,11,324 patients surviving a recent myocardial infarction (≤3 months previously)were randomly assigned to receive supplements of n-3 PUFA (1 g daily,n = 2,836), vitamin E (300 mg daily, n = 2,830), both vitamin E and n-3 PUFA(n = 2,830), or neither (control, n = 2,828), using a 2 × 2 factorial design.The primary combined efficacy endpoint was a composite of death, nonfatalmyocardial infarction, and stroke.311