Читать книгу Cobert's Manual Of Drug Safety And Pharmacovigilance (Third Edition) - William Gregory - Страница 64

Most studies are based on samples, not entire populations. The confidence interval reflects the resulting uncertainty.

Based on the sample of the population, a particular result is obtained, e.g., 15% of the users of drug A had serious AEs. Because we did not study the whole population, we cannot be totally sure that the 15% figure represents the true value for the whole population rather than by chance for the smaller sample studied. The confidence interval represents the range of the correct or true value for the whole population. One can calculate various levels of “assurance”, 90%, 95%, 99%, 99.9%, and so on, for the confidence interval. Usually, the 95% level is used. It is better to have a narrower or smaller distance between the upper and lower values of the confidence interval (called the “confidence limits”). In general, the more subjects in the study, the narrower (better) the confidence interval. To put it another way, if the 95% confidence interval for a study group is (43–79 units), then we can be confident that the true value for the entire population is between 43 and 79, 95 times if we were to repeat the study in 100 different samples of the population. If the 95% confidence interval is narrower (59–66) because more subjects were studied or for various other reasons, we have a more precise sense that the true population value is closer to the value found in the sample studied.

Other possible designs incorporate a self-controlled case series or risk window, as these are often used in safety studies and have increasing utility for cancer, vaccines, and biologicals. These designs can control within cases-only, e.g., for genetic background features.

With the arrival of formal risk management as an integral part of the development and life span of all drugs, the fields of pharmacoepidemiology and drug safety are now more tightly linked than ever.

Health authorities require epidemiologic safety studies, particularly as post-marketing commitments or requirements. Large databases and practitioners who know how to conduct these studies have become more and more available and the methodology has become more refined and automated. Many pharmaceutical companies now have risk management/pharmacoepidemiology departments to handle these studies.

Indeed, the role of pharmacoepidemiology in risk assessment and signal detection has become more prominent as risk management programs become increasingly active over the entire life cycle of products.