Postmarketing surveillance includes both purely passive collection of data submitted to the company that are not solicited and also data that are solicited. The primary use of passive postmarketing surveillance data is to identify signals of adverse events that may warrant further study to better understand the benefit-risk relationship and protect the public health against adverse consequences of drug exposure. Because of the limited number of patients studied in Phase 1 to 3 trials on a long-term basis, adverse events that have a relatively long latency period are often not observed until well after the drug is on the market. This is one type of adverse event that is sought in the signals that come from postmarketing surveillance.

Actual clinical trials that are conducted during Phase 4 are conducted to:

A golden rule about these trials is that the more a company knows about its own drugs, the better it is for both the company and patients.

During the postmarketing period, clinical trials may be conducted to evaluate patient populations that were not specifically studied during the Phase 1 to 3 investigational period (see discussion on adverse events in Chapter 61). The populations that are often studied in Phase 4 include children, pregnant women, certain racial groups, and those with compromised organ function (e.g., renal failure, hepatic failure, cardiac failure) or who are immunocompromised. It is essential, however, to differentiate between Phase 4 clinical trials in patients covered by the approved labeling and trials in those for whom a new development program is needed. Children, for example, usually must be studied in Phase 2 and/or 3 trials under an Investigational New Drug Application and not as part of Phase 4 trials.

The Pharmacovigilance Plan is based on the Safety Specification discussed in Chapter 61, which is a summary of the risks of a drug—the known, potential, and unknown. The structure of the Pharmacovigilance Plan is described in the E2E document by the International Conference on Harmonisation (ICH), and the interested reader is referred to the ICH website for additional information (www.ich.org). Briefly, this document states that the plan has four parts:

In addition, the company wants to understand the answers to numerous questions. Some of the questions that relate to marketing or drug utilization, for example, include:

Record linkage is an important concept that underlies the use of large multipurpose automated databases and other databases. Record linkage is where two or more record files are linked, usually by computer. For a specific patient, it involves putting together information (i.e., records) that may have been obtained at different times and/or different places. This linkage allows questions to be addressed on a retrospective basis that previously could not be. For example, pharmacy records of outpatients with patient identifier numbers may be linked with the hospital inpatient experience of the same patients with a diagnosis code. These records may also be linked with the inpatient drugs prescribed
during hospitalization. Many trials may only be conducted with computers that search these databases looking for specific linkages. For example, the outpatient drug history may be checked for all patients who had an admitting (or discharge) diagnosis of X. Alternatively, all patients receiving a new drug may be tracked to identify their next hospitalization.

Record linkages have impacted many areas of public health and hospital practice. For example, Shapiro et al. (2006) discuss how local or regional health information organizations can establish networks for sharing information electronically that link hospitals, skilled nursing facilities, clinics, pharmacies, laboratories, radiology facilities, health departments, and even payers to provide information useful in the practice of emergency medicine.

Record linkage technologies are moving rapidly and will have a great effect on both the practice of medicine as well as on research. While a variety of medical areas will someday be routinely helped with the technologies that are being expanded today, those that require rapid responses such as trauma are being extensively explored. Nonetheless, computer-assisted record linkage requires human review of the results and confirmation of the data that is linked. The Council for International Organizations of Medical Sciences (CIOMS) Working Group VI (2005) lists many of the electronic databases in its Appendix 9 (page 287). Case-control designs may also use record linkage technology (Westhoff 1998).

Figure 66.1 illustrates that reported rates of all adverse events from a single drug or even the rate of a single adverse event from one drug vary throughout a drug’s life. The reasons are generally quite straightforward. Rare adverse events (i.e., those that occur in less than one in 1,000 patients) can seldom be discerned during premarketing trials because a total of only 1,000 to 5,000 patients is usually exposed to a new drug. After marketing, most drugs are used by many thousands of patients, and rare adverse events are naturally observed more often. After those adverse events are associated with the drug by physicians, they may publish a report in the medical literature or submit a report to the company or regulatory authority. If published, many physicians are alerted via the literature about the adverse event. Shortly thereafter, that newly reported adverse event tends to become overdiagnosed and widely reported by zealous and/or naive physicians (stimulated reporting). Some physicians are anxious to publish reports, while others may not be experienced in making an accurate diagnosis of the adverse event or be knowledgeable about assessing causality. In addition, the same diagnosis may refer to different conditions in different countries or may be referred to by different terms. After a period of time, there is often less interest in the medical community about reporting each case of a moderate or serious adverse event that is perceived as being well known and well documented. As a result, the number of reports received by the company and by regulatory authorities represents an example of underreporting of most well-known adverse events for any particular drug.

Commonly occurring adverse events are observed during premarketing trials. In general, the more common the adverse event, the closer the reported incidence is to the true incidence (Fig. 66.1, Panel A). Figure 66.1 illustrates a number of theoretical rates of reporting based on when the product is marketed. After marketing, there is usually less incentive for physicians to spontaneously report common adverse events or those that are well known (e.g., Panel D). Panel C illustrates a flurry of attention focused on an adverse event after marketing that eventually is reported less often. The premarketing period in Panel C might involve virtually no reporting of the adverse event. It must also be noted that the true incidence of an adverse event is not always characterized by a constant rate but may vary greatly, further complicating reporting patterns. Many variations on the themes in Fig. 66.1 occur due to many confounders and biases that affect the observed rate (e.g., some adverse events of a drug may be underreported, whereas others are overreported). A classic paper on this topic of confounders and biases is by Sachs and Bortnichak (1986).

Although it might superficially appear that a single universal package insert or drug labeling for each drug would be desirable to better standardize the drug’s use and promotion in different countries, closer inspection shows that this does not make sense. For example, the marketed products in different countries may vary significantly as: