Design of Experiments > Parallel Design
What is a Parallel Design?
A parallel design, also called a parallel group study, compares two or more treatments. Participants are randomly assigned to either group, treatments are administered, and then the results are compared. It is the “gold standard” for phase 3 clinical trials(1).
Although it’s common to include one control group (as in a controlled study), administering a placebo is not an essential element of the design. Other options include:
- One treatment group, and one Treatment-as-Usual group.
- Two active treatment groups. One of the groups might receive an active comparator (a treatment that’s known to be effective).
- One treatment group and a sham comparator. A sham comparator is an intervention that appears identical to the investigative treatment, but doesn’t contain an active ingredient or procedure. For example, a person might receive an injection of saline solution instead of the experimental drug.
- One treatment group and one “no intervention” group. The no intervention group receives either no treatment at all, a sham or a placebo.
Blinding, where participants and researchers do not know which group the participant is in or which treatment is being administered, is usually used to prevent bias.
A two-arm parallel design compares two treatments: One treatment group “A” is given one treatment and a second treatment group “B” is given a different treatment. While two-arm designs are common, multiple arms can effectively and quickly compare multiple treatments (like different doses of medication).
Comparison to Crossover Designs
A crossover design (a type of repeated measures design) is where groups receive all treatments in a different order. For example, group A might receive treatment X then treatment Y, while group B receives treatment Y then treatment X. By comparison, a parallel study has all groups receiving completely separate treatments in parallel. For example, group A receives treatment X while group B receives treatment Y.
One major advantage of a crossover design is that, for the same number of participants, the crossover has a higher statistical power than the parallel study. This is because participants act as their own controls. Parallel studies require a separate comparison group and therefore tend to be more expensive.
On the other hand, crossover designs may have carryover effects, where effects from one treatment affect the second treatment. If this is a concern, a parallel design is a better alternative.
Sequential Parallel Comparison Design
A Sequential parallel comparison design (SPCD) was developed by Massachusetts General Hospital to counteract the placebo effect; In some trials, large numbers of people might react to a placebo, washing out any positive effect of the actual treatment being investigated. SPCD attempts to counteract the placebo effect by taking all participants who did not react to the placebo, and essentially retesting them.
SPCD has two stages. The first stage is identical to the parallel design with one group receiving a treatment and a second group receiving a placebo. In the second phase, all of the participants who did not respond/react to the placebo are re-randomized and reallocated to a treatment group or a control group. The results from both stages are pooled and analyzed.
References:
Bacchieri A, Cioppa GD. Fundamentals of Clinical Research: Bridging Medicine, Statistics and Operations, Milano: Springer-Verlag Italia; 2007.