Numerators, Denominators, and other Holy Grails: Management and Interpretation of Worldwide Drug Safety Data

The International Division of Pfizer Inc. devotes special effort to the collection, tabulation, interpretation, and reporting of safety data potentially on about 300 drug products sold in more than 100 countries. In addition to individual, spontaneous reports from the field, data from various other sources including governmental ADR registries, the literature, and independent safety monitoring and epidemiological groups must be screened and assessed; identification of multiple reports on the same case is an especially difficult problem in this regard. Formal causality assessment of cases is usually performed only on reports of fatalities or unexpected and unusual findings; qualified medical staff make the assessments. Although it may be desirable to quantify the incidence of specific, usually serious, untoward events contained in spontaneous reports, due to important inherent biases such an exercise can be very misleading, especially when attempting to compare different drugs. Prime among these biases (or factors) that affect both the numerator (number of reports) and denominator (drug exposure or usage) are the following: intercountry differences in medical practice and in the ADR reporting environment; the time a drug is on the market (new versus old drugs)—ADR reports typically peak during the first 1 to 3 years after availability; changes in reporting habits or laws within a country; influence of media-publicized adverse stories on a specific drug; and the absence of concrete data on overall drug exposure and usage patterns. Examples of these factors are described and it is shown that data from spontaneous reporting systems should not be used to derive absolute or comparative incidences, but are probably best used as an alerting mechanism for unusual findings.