Stefan Dinescu
Rheumatologist, University of Medicine and Pharmacy Craiova, Romania
EMEUNET Country Liaison for Romania
Introduction
When starting a research project, one of the most important steps is to establish what study design best suits your study objectives. After you define your research question, choosing the appropriate study design is essential, as it provides the foundation for accurate data collection and analysis in accordance with the study’s aims. A poorly planned study may lead to insufficient data and inconclusive results.
This article presents an overview of the main study designs used in medical research, including observational, experimental, cross-sectional, and longitudinal designs. It presents scenarios for selecting among study designs and highlights their advantages and limitations.
Observational studies
In an observational study, the researcher is a passive observer, recording what naturally happens. There is no active intervention or manipulation of the subject’s exposure by the researcher. As opposed to an experimental study, the researchers simply observe and record patient information, without altering the study environment. Examples include cohort studies (following exposed versus unexposed groups) and case-control studies (comparing individuals with the outcome/disease to those without). An example is a study that looks at the incidence of lung cancer in smokers versus nonsmokers, or comparing the incidence of asthma annually based on exposure to air pollution. Observational studies use real-world data and are very useful when intervention is unethical or impractical (e.g., studying the effects of smoking). It is susceptible to selection and information bias. Causal inference is limited by potential confounding variables or unknown factors that influence the results. Among different types of observational studies, this inherent risk of bias is more limited in prospective cohort studies.
Cross-sectional studies
A cross-sectional study determines the prevalence of a disease, exposure, or characteristic in a population at a single point in time. It thus captures information like a ‘snapshot’. Because both exposure and outcome are measured simultaneously, it’s impossible to establish a temporal association, and causality cannot be determined. An example is a survey of 1,000 adults to find out what percentage currently have type 2 diabetes (outcome) and what their current physical activity level is (exposure). Although cross-sectional studies cannot establish causality and are considered low-level evidence, they are quick, inexpensive and efficient for estimating prevalence and generating hypotheses.
Longitudinal studies
Longitudinal studies track changes, trends, or the incidence of outcomes in the same group of subjects over an extended period. It records changes in variables over time by repeatedly measuring the same individuals. This allows researchers to determine the sequence of events and strengthen evidence for cause-and-effect relationships. Cohort studies are the most common type of longitudinal design in medical research. A cohort study follows a specific group of people (a “cohort”) who share a defining characteristic or experience, often an exposure, over time to assess outcomes or diseases that develop. An example is a study that identifies risk factors (exposure – high blood pressure, high cholesterol, and smoking) for cardiovascular disease (outcome). In this example, the cohort of patients is regularly examined and followed for multiple years to track the onset of outcomes like heart attacks, stroke, and heart failure. Longitudinal studies can show the direction and sequence of events (temporality), which is essential for inferring causation. They are nevertheless expensive and time-consuming, with a high risk of attrition bias (participants dropping out) and loss to follow-up, which can skew results.
Experimental studies
In an experimental study, the researcher actively intervenes by administering a treatment or exposure. These studies can be further divided into randomised clinical trials (RCTs) and non-randomised clinical trials. The use of randomisation is critical to ensure that both treatment groups are, on average, comparable at the start of the study, thereby isolating the effect of the intervention. In a classical RCT, subjects with similar characteristics are randomised to two groups: one that receives the therapeutic intervention and one that receives a placebo (or standard of care). This type of study is considered the gold standard for research on drug efficacy. RCTs offer the highest level of evidence and can establish a true causal effect of an intervention (exposure). The types of interventions in an RCT are broad and include: therapeutic agents, procedures, devices, and behavioural or lifestyle modifications.
Conclusion
In conclusion, it is important to understand the differences among study designs and to know when to apply each, depending on your research objectives. Of course, other factors influence the type of study design you will choose, including the required time interval, costs, infrastructure or previous knowledge of disease/outcome frequency.
Figure 1. Schematic presentation of different study designs
Figure 2. Schematic presentation of temporal direction based on study design