An experimental or prospective investigation is one where the investigator manipulates a factor or factors (independent variable) and observes the effect on another factor or factors (dependent variable) over time. This design is considered to be the ideal research design. For example, the investigator tests the effect of an antiplaque/antigingivitis mouthwash (independent variable) on plaque deposits and gingivitis (dependent variables). Sometimes, due to ethical or other reasons, the investigator is unable to manipulate the variables. Under these circumstances, a retrospective investigation is conducted in which data are collected on a phenomenon that has already occurred. For example, a researcher identifies groups who differ on a condition such as oral cancer, one group with it and one without, and works backwards to identify the causative factors for the condition, such as a history of tobacco and alcohol use.
The gold standard for testing new therapeutic agents is the randomized clinical trial (RCT). This is an experimental design used to test the hypothesis that a particular agent or procedure favorably alters the natural history of a disease. Two designs are commonly used in clinical trials, the parallel design and the crossover design. With the parallel design (the most common of the two), an experimental group with the new treatment and a control group with the standard or placebo treatment (inactive substance) are used, with pre-treatment and post-treatment measurements of health (Figure 1).
With the crossover design, the initial experimental group switches to the control group and the control group to the experimental group halfway through the experiment. This occurs after a washout period, where each subject's physiological condition is allowed to return to baseline (Figure 2). An advantage of this latter design is that no patient is denied the experimental treatment. Another advantage is that each subject serves as his/her own control; fewer subjects are needed for such designs. The crossover design can only be used with respect to diseases or conditions that recur when treatment or medication are withheld, such as gingivitis.
There are two potential designs for studying phenomena that occur over extended periods of time, for example, growth of the mandible or the course of periodontitis. In a cross-sectional design (Figure 3a), a sample(s) of a population (cross section) is assessed at one time; with a longitudinal design (Figure 3b), the same sample of individuals is assessed at several different time points. While a cross-sectional design is often more expedient, the longitudinal design usually provides better information, since the actual amount of growth or change in each individual can be assessed. Lacking a true time element, cross- sectional designs are considered unsuitable for making conclusions about cause and effect (causation).