Approached to monitoring the results of long-term disease prevention trials: examples from the Women's Health Initiative.

We contrast monitoring therapeutic trials with monitoring prevention trails. We argue that in monitoring prevention trials one should place more emphasis on formally defined global measures of health, not simply on a single targeted disease, particularly when an intervention may reduce the incidence of some diseases but increase the incidence of others. We describe one approach, illustrated by the Women's Health Initiative. For each of several sets of hypothetical interim results ("scenarios"), members of the Data and Safety Monitoring Committee (DSMC) were asked whether they would continue or stop the trial. In parallel with this exercise, various statistical methods of monitoring that are based on (1) the primary targeted disease, (2) a combination of various disease outcomes, or (3) a mixture of both were applied to these scenarios. One objective was to find a statistical approach that mirrors the majority view of the DSMC. A second objective was to stimulate discussion among DSMC members in preparation for their task of monitoring the trial as the real data become available. We found that no single method fully matched the majority vote of the DSMC. However, a mixed approach requiring the primary outcome to be significant and the global index to be "supportive," with separate monitoring of adverse effects, corresponded with the majority vote quite well. This approach maintains the emphasis on the primary hypothesis while assuring that broader safety and ethical issues of multiple diseases are incorporated.