This chapter discusses the broad categories of clinical investigations used in post-market drug safety assessment. It presents the three main methods of clinical post-marketing safety assessment: case reports and case series; observational epidemiological studies; and clinical trials. Active surveillance systems are also being explored to identify and examine drug safety issues. Drug safety active surveillance systems, which take advantage of large repositories of automated healthcare data, are now being developed and tested by multiple organizations. The two most common observational epidemiological study designs are the case-control design and the cohort design. The majority of clinical trials are performed primarily to assess the efficacy of a product. The design of a post-marketing clinical trial testing a safety hypothesis is often an active-controlled trial that uses a non-inferiority study design. Relative to observational epidemiological studies, clinical trials designed to answer drug safety questions are usually more costly and more time-consuming.

Psychiatric studies are characterised by large numbers of variables, clinical outcomes that are difficult to measure, small sample sizes and conflicting results. Study designs are classified along the axes of time (longitudinal or cross-sectional design), level of causal inference (descriptive or analytical design) and role of the investigator (observational or experimental). Descriptive designs can be used to examine associations between exposures and outcomes in such a way that the results may lead to the formulation of more specific hypotheses regarding the causal implications of the exposure-disease relationship. The case-control design is suitable to examine aetiological heterogeneity or different exposures resulting in the same outcome, because a range of exposures in patients and control subjects can be assessed. Medical scientific studies generally have two main interests in the examination of the results: examination of the morbidity force of a disease phenotype in populations and examination of associations.

To study exposures which cannot be randomly assigned, we usually turn to comparative observational study designs: either cohort or case-control studies. The chapter describes the design and analysis of case-control studies. In planning a case-control study, it is helpful to think of the study as being set in a specific population or cohort, even though most members of that population will not participate in the study. In selecting cases, we should consider the step in the causal chain that we wish to study. In principle, controls should be a sample of persons from the same population from which the cases were derived. The chapter also discusses two special case-control designs, proportional mortality studies and case-crossover studies. Matching controls to cases may be justified if it promises to enhance study efficiency or to control for factors that cannot otherwise be measured.