## Rr

Let us note that if U1 = U2, then the estimate of the RR remains identical, whatever the magnitude of under-reporting. Thus, a comparison based upon the number of reports would lead to the same estimate as if based on the actual number of cases. The only consequence would be a dramatic decrease in the statistical power of the comparison test because it was computed on smaller samples.

This is illustrated by Table 21.3 showing the values of the statistic of a chi-square test performed on the basis of a theoretical number of 120 cases for Drug 1 and 60 for Drug 2, respectively; the number of patients treated being chosen identical (N1 = N2 = 300 000) for simplification purposes. One can see that a complete p

Table 21.3. Values for the chi-square statistic computed on the basis of 120 cases for Drug 1, and 60 cases for Drug 2, respectively (the exposed population size being the same for both drugs: 300 000 patients) according to several theoretical values of the under-reporting coefficient U (bold figures correspond to differences which are significant at the 0.05 level).

Table 21.3. Values for the chi-square statistic computed on the basis of 120 cases for Drug 1, and 60 cases for Drug 2, respectively (the exposed population size being the same for both drugs: 300 000 patients) according to several theoretical values of the under-reporting coefficient U (bold figures correspond to differences which are significant at the 0.05 level).