Search

Currently the data resources available for conducting pharmacoepidemiology studies in the United States include spontaneous case reports of adverse reactions, aggregate population-based data resources, computerized collections of data from organized medical care programs, data collected for pharmacoepidemiolgy on an ongoing basis, existing data collected as part of other ad hoc studies, and data collected de novo for pharmacoepidemiology studies. The advantages and disadvantages of each and the particular contributions of drug prescription data to each approach are described.

This report reviews the 100 most recently approved drugs in order to quantify the frequency with which post-marketing studies of drug efficacy can be performed experimentally and non-experimentally. These drugs represent 131 potential drug uses. Of them, the absolute efficacy of 89 (68 percent) could be evaluated from clinical observations. Of the remaining 42, six (14 percent) could be studied experimentally or non-experimentally, six (14 percent) only experimentally, one (2 percent) only non-experimentally, and 29 (69 percent) by neither technique. Answers to all questions of relative efficacy required formal research. Of these, 94 (72 percent) could be studied using either experimental or non-experimental techniques. The remaining 37 (28 percent) could be studied experimentally only. Thus, clinical observations and non-experimental research can contribute a large proportion of the information about drug efficacy still needed after marketing.

In order to quantify the need for post-marketing studies of drug efficacy, the 100 drug uses (i.e., drug-indication pairs) that were most common in 1978 were reviewed. Of the 100 most common drug uses, 31 were not Food and Drug Administration (FDA)-approved at the time of initial marketing and 18 had not become FDA-approved, even subsequently. In 13, the use was based on the drug's secondary effect rather than on the primary effects investigated before marketing. The efficacy of all was subject to modification by incompletely explored factors, including other drugs, patient age, other illnesses, pregnancy, etc. Specific examples are provided of important modifiers of the efficacy of these drugs that were not investigated prior to marketing. Thus, despite the existence of sufficient data for the regulatory process to permit the marketing of these commonly used drugs, considerable gaps remain in the information needed for their optimal clinical use. Post-marketing studies of drug efficacy will be needed to fill these gaps.

The study designs used for postmarketing surveillance are those of epidemiology, but very large sample sizes are needed. Therefore, a number of specific logistical approaches have been developed. One can use spontaneous reports of adverse reactions, aggregate population based data, computerized collections of data from organized medical care programs, data collected for postmarketing surveillance on an ongoing basis, existing data collected as part of other ad hoc studies, or data collected de novo for the study being conducted. Each is described in more detail, along with its advantages and disadvantages.