Clinical trial designs for rare lung diseases must meet the same rigorous standards as do designs for trials for diseases that occur with much more frequency. However, there are many different types of study designs; some of which require only a fraction of the number of subjects required to the randomized controlled trial, which is often considered the gold standard.
Alternate designs can address those issues by the use of external or historical controls or with participants serving as their own control. In the case of external or historical controls, all patients to be recruited on a proposed study would receive the new or experimental therapy and their outcomes would be compared to a population that had already been treated by a standard therapy. If historical data are valid and available, this is a very efficient design because it requires fewer patients to be accrued. The downside of such a design is that the selection of historical controls must be made with extreme caution so as not to bias the study results.
A design that avoids this problem is the use of concurrent controls for which participants can serve as their own control. Such designs are desirable if there is less within patient variability in a treatment response than there is between-patient variability. In such cases, outcome estimates will have less variance and the study design will require less accrual. Examples of these designs include cross-over designs and “N-of-1” designs. A design that is well suited to rare events and rare diseases is the case–control design. In such a design, individuals in whom a certain outcome has been observed (disease severity or particular event) are matched to controls who did not have such an outcome and then the two groups are compared with respect to a particular intervention or exposure. Such designs can be developed from prospective as well as retrospective data collection perspectives.
Examples of prospectively randomized designs include cross-over designs as well as factorial designs. In the former, participants are randomized to a treatment arm for a period at the end of which the outcome is assessed and then “crossed over” to the other treatment. The cross-over design makes the same assumptions as do “N-of-1” trials where participants are randomized to pairs of therapies given in random sequence and a washout period is assumed to eliminate the affect of the treatment after the intervention is withdrawn. Factorial designs essentially involve a double randomization in which two questions are asked in the same participant population.
Finally, designs for ranking and selection procedures are often helpful and generally require a smaller sample size than randomized controlled trials. Ranking statistics are often used when information about underlying parametric distributions is unknown. It could be argued that less is learned in such an experimental design and a subsequent experiment is required to measure the actual difference between treatment outcomes.
There are many approaches to the design of a trial and many of them can achieve certain economies in terms of the required number of participants that need to be enrolled. However, the options are not without their drawbacks and require investigators to make a number of assumptions.