Taxonomy of Clinical Research

Quantitative Research Designs

Observational Studies

Observational study: who, where, when. A descriptive study provides a description of exposures and outcomes, whereas an analytic study provides a measure of the association between exposure and outcome. Observational studies are hypothesis generating and cannot establish causal association. 

Ecologic study: provides a description of population group characteristics. 

Cross-sectional study: evaluates the association between exposures and outcomes at the same time. It allows measurement of  the prevalence of exposure or outcome. Temporal association is more difficult to establish. 

Case-control study: compares subjects with disease (cases) to those without disease (controls) for differences in risk factors. It is useful for establishing hypotheses about risk factors and aetiology, especially when a disease is rare. It cannot estimate the prevalence of disease. Controls and cases should be sufficiently similar in prognostic factors unrelated to the risk of interest. 

• Nested case-control study: defines a cohort with suspected risk factors and a control is assigned for each case from within the cohort. Cases and controls are matched on calendar time and length of follow-up. 
• Case-cohort study: cases are not matched on calendar time or length of follow-up to controls. 

Cohort study: follows and assesses outcomes in exposed and unexposed groups over time. A retrospective cohort study identifies populations with and without the exposure based on past records and assesses outcomes at the time of study. Common biases are selection bias and information bias. 

Experimental

Clinical trial: subjects with disease are placed in different treatment groups. The study population must be sufficiently similar and representative of the general population of patients. The intervention is compared to a placebo, no treatment or an alternative treatment. 

• Parallel randomised trial: subjects are unpredictably (randomly) allocated to intervention and control groups. This minimises the risk of confounding in estimating the treatment effect. 
• Parallel non-randomised trial: controls are selected using some predictable pattern, e.g., certain days of the week.  
• Historically controlled trial: controls are adopted from the past, e.g., medical records, or previous studies. 
• Cross-over trial: two groups undergo the same treatments at different time periods, i.e., each group serves as a control while the other group is undergoing intervention. The effect of the treatments needs to be reversible. 
• Factorial trial: different treatments are tested at the same time in the same population. The treatment effects should be independent. 

Community trial: groups of subjects are assigned to different treatments. 

Field trial: subjects without disease are placed in different preventative intervention groups.   

References

• Essentials of Clinical Research Design (Pediatr Invest 2019)
• Classification of Observational Studies (Lancet 2002)