If we use the threshold of 0.05 as statistically significant, then we can observe that the effect of therapy on the total mortality rate is significant, while the effect of therapy on the fatal/nonfatal MI rate is not. Furthermore, if we accept that the magnitude of the p-value provides useful information about strength of association, we would also conclude that the therapy had greater effect on the total mortality rate than on the fatal/nonfatal MI endpoint. One could easily construct an argument from the results of Table 3.1 that there was a more important effect of the therapy on reducing total mortality then on the fatal/nonfatal MI rate. This would in turn lead to predictable discussions about the effect of the therapy on causes of death not related to MI, an effect that would be driving the low p-value.

These conclusions are wrong, however, and would have to be modulated based on a more complete demonstration of the results of the data (Table 6.2).

Table 6.2. Effect of therapy on each of total mortality and fatal/nonfatal MI. Sample size, p- values and percent reduction in events from a clinical trial with two endpoints.



Percent Reduction

P -value

Was this article helpful?

0 0

Post a comment