The parallel design (also known as parallel study design) is the most frequently used study structure in randomized clinical trials. In this design, study participants are randomly assigned to one of two or more treatment groups, which simultaneously and independently receive the same treatment throughout the entire study period. Each participant belongs to only one group for the duration of the study. The parallel design thus fundamentally differs from the crossover design, where each participant receives several treatments sequentially. The simultaneous, independent treatment of the groups is the defining characteristic of the parallel design. It is the most methodologically robust basic design and most frequently used for registration trials, and is considered the gold standard for confirmatory efficacy demonstration in clinical drug development.
Structure and Randomization
In a classic two-arm parallel design, participants are assigned to either the investigational treatment or the control group, which may receive an active comparator or a placebo. Assignment is performed through randomization to avoid systematic differences between groups (confounding). Stratified randomization ensures that prognostically important factors such as age, gender, or disease severity are evenly distributed among the groups.
For studies with more than two arms – for example, when comparing multiple dosages of an investigational substance and a placebo – the parallel design is extended accordingly. Multi-arm studies require careful planning of statistical analysis, as multiple comparisons increase the risk of false-positive results. Here too, the testing procedure must be prospectively defined in the statistical analysis plan.
Advantages of the Parallel Design
The most important advantage of the parallel design is its broad applicability. It is suitable for almost all indications and endpoints, including studies with irreversible outcomes such as mortality or permanent organ dysfunction. Unlike the crossover design, no wash-out periods are required, and there is no risk of carry-over effects, i.e., residual effects of a previous treatment on the subsequent phase.
By simultaneously treating both groups under identical conditions, the parallel design allows for a direct comparison of treatment effects. Seasonal fluctuations, changes in medical standards, or external events during the study affect both groups equally and do not systematically distort the comparison. This makes the parallel design particularly robust against time-dependent confounding factors.
Disadvantages and Limitations
The main disadvantage of the parallel design lies in its higher sample size requirement compared to the crossover design. Since each participant receives only one treatment, inter-individual differences contribute to the variance of the treatment effect. In contrast, in a crossover design, each participant serves as their own control, which significantly increases statistical efficiency. For studies where a crossover design is methodologically justifiable, the parallel design may therefore require larger patient numbers.
Another aspect is the longer study duration associated with the simultaneous parallel operation of both arms. Both groups must be followed up over the entire observation period, which, in the case of chronic diseases with long follow-up times, requires significant logistical and financial resources. Nevertheless, the parallel design is generally the most methodologically sound choice for regulatory efficacy demonstration.
Application in the Regulatory Context
The EMA and FDA prefer the parallel design as the standard for confirmatory efficacy studies, as it is methodologically clear, well-established, and easy to evaluate. The statistical methods for evaluating parallel studies are described in the ICH-E9 guideline and are largely standardized. For planning a parallel design protocol, authorities recommend seeking Scientific Advice early to discuss the choice of comparator, sample size planning, and primary endpoints. Full-service CROs like mediconomics support sponsors from study concept and protocol development to statistical evaluation of parallel design studies across all indications.
An important aspect in planning a parallel design protocol is sample size calculation. Since both groups are observed independently, the study must be sufficiently large to demonstrate a clinically relevant difference with high statistical power. A power of at least 80% is considered standard, with many registration studies aiming for 90% power. The sample size depends on the expected effect size, the variance of the endpoint, the significance level, and the dropout rate. These parameters must be justified and documented in the study protocol.
The parallel design is also the preferred structure for studies submitted for benefit assessment within the framework of an AMNOG procedure (Arzneimittelmarktneuordnungsgesetz – Pharmaceutical Market Restructuring Act) in Germany. The Gemeinsamer Bundesausschuss (G-BA) generally recognizes only randomized controlled trials with a parallel design against a patient-relevant comparator as the highest level of evidence. Missing randomization or a lack of parallel design can lead to a study being deemed unsuitable for benefit assessment. The choice of comparator – ideally the appropriate comparative therapy as defined by the G-BA – is also crucial for the AMNOG procedure.
Frequently Asked Questions (FAQ)
When is a Parallel Design preferable to a Crossover Design?
A parallel design is always preferable when the disease is not reversible, when carry-over effects cannot be ruled out, when the treatment has a curative effect, or when the study duration would be too long for a crossover design. For chronic diseases, oncological indications, and studies with hard endpoints such as survival, the parallel design is almost always the more suitable design.
Can a Parallel Design be Adaptive?
Yes. Adaptive study designs can be well combined with the parallel design. Common variants include interim analyses with the possibility of sample size re-estimation or early termination due to clear superiority or futility. The adaptation rules must be prospectively defined in the statistical analysis plan and accepted by the regulatory authority.