## U

Referring to the classical Poisson formula, the probability of receiving x reports is:

and the cumulative probability of receiving at least x reports is:

Table 21.2 gives the probability of receiving at least one report according to the value of m. One can see that to have a good chance of detecting an adverse event requires m to be greater than one, i.e. 1.61 for an 80% chance, 2.30 for 90% and 3 for 95%, respectively.

It should be kept in mind that for serious conditions, the baseline incidence p is usually extremely low, therefore m remains markedly below one, except if N is extremely large and RR/U greater than one, i.e. if the association between drug exposure and the considered event is strong and the reporting is reasonably good.

Let us take the example of a non-steroidal anti-inflammatory drug for which the average duration of use is two weeks. In a given country, 2.5 million two-week treatments have been made in one year, corresponding to a cumulative time of exposure of 5 millions weeks, i.e. 96 154 years. Considering the generally recognized value of 7 per million for the annual incidence of agranulocytosis in the general

 m Pr(k > 1) m Pr(k > 1)