Box 102 Examples of pharmacoepidemiological methods

• Intensive hospital monitoring

• Prescription event monitoring (PEM)

• Cohort studies

• Case-control studies and case-control surveillance

• Case-crossover design

• Record linkage

Cohort studies are studies where groups are defined according to the exposure status (for example users and non-users of a drug) and are followed up, with subsequent events being recorded and compared.14,15 By contrast, case-control studies are studies where a group is defined according to its experience of an outcome of interest (for example cases of toxic epidermal necrolysis) and is compared with a control group that has not experienced the outcome. Prior exposures are ascertained for each group retrospectively and are compared.16 A crucial point for the validity of a case-control study is the choice of appropriate controls. In principle, controls should be an unbiased sample of those individuals composing the so-called "study base". Controls for cases arising in the ambulatory population with resultant hospitalisation (community cases) may be represented by patients admitted to the same hospital for an acute condition or for an elective procedure not suspected of being related to drugs.17

Generally speaking, cohort studies are better suited to the study of rare exposures and common events, and case-control studies to the assessment of rare outcomes and relatively common exposures. Cohort studies allow the assessment of several outcomes for one specific exposure. Case-control studies allow the assessment of the role of a range of different exposures on the development of a single specific outcome. Cohort studies are not feasible when dealing with rare events, because millions of drug users have to be observed for years. In this situation, case-control studies with a very large population base are the most feasible method. For example, it is intuitive that only a case-control study would be feasible to assess the pharmacological risk for a disease like toxic epidermal necrolysis, with an expected rate in the general population of one case per million people per year.18,19

It is important that outcome and exposure are measured in the same way in the groups being compared in observational studies. However, even if investigators document the comparability of potential confounding variables in the groups being analysed (exposed and non-exposed cohorts, or cases and controls) or use statistical techniques to adjust for them, there may be an important imbalance that the investigators do not know about or simply have not measured that may be responsible for any observed difference.

Case-control and cohort studies should be developed with the aim of testing a specific predefined hypothesis. In the past few decades, modifications of traditional cohort and case-control studies have been developed to explore new associations and to raise signals. Record linkage is based on linkage of data from large electronic databanks on exposure and outcome. Case-control surveillance is the ongoing collection of cases of prespecified rare and severe acute events and of suitable controls, looking for new associations of the events with drug exposure.20

The association of an exposure with a given event is usually expressed in terms of a relative risk or odds ratio (an estimate of the relative risk obtained from case-control sudies) (Table 10.1). The relative risk is a measure of the size of an association in relative terms. It refers to the ratio of the incidence of the outcome among exposed individuals to that among non-exposed individuals. Values greater than 1 represent an increase in risk associated with the exposure; values less than 1 represent a reduction in risk. A relative risk of 2, for example, tells us that the event under study occurs twice as often in the exposed people as in the non-exposed. For rare events, even a large relative risk may translate to the occurrence of a few additional drug reactions. The total incidence of an outcome among exposed individuals is a combination of the baseline incidence plus the excess of incidence due to the exposure.

Table 10.1 Measures of association

Patients Adverse event No adverse event

(cases) (controls)

Exposed a b

Not exposed c d

Relative risk = [a/(a + b)] / [c/(c + d)] Odds ratio = (a/c) / (b/d) Excess risk* = [a/(a + b)] - [c/(c + d)] NNT or NNH = 1/excess risk

NNH from case-control studies = 1/[(odds ratio - 1) (unexposed event rate)]

*The excess risk may be also referred to as the "risk difference" or the "absolute risk reduction"

The excess risk (or risk difference or absolute risk reduction) is calculated as the difference between the incidence among exposed individuals and the incidence among non-exposed individuals. It measures the occurrence of an outcome among exposed individuals that can be attributed to the exposure. As such, it is a better measure of the impact of different outcomes than the relative risk, and a more informative measure from the points of view of an individual physician and public health. Measures of excess risk are directly calculated in cohort studies and, provided that data on the incidence of the outcome are available in the underlying unexposed population, they can also be derived from case-control studies.

0 0

Post a comment