Types of clinical trial designs: The essentials

On average, it takes 10.5 years for a drug to advance from phase 1 clinical trial to regulatory approval and market access. Between 2010 and 2020, the overall likelihood of approval was 7.9%. These figures underline the importance of the choice of the clinical trial design type. As the biotech sector continues to grow, driven by rapid advancements in technology and an increasing focus on personalized medicine, the variety of clinical trial designs is expanding.

As the demand for more innovative and efficient trials grows, the industry faces challenges such as increasing the diversity of trial participants and managing the high costs and complexities associated with advanced trial designs. The adaptation of clinical trials to include remote technologies and decentralized approaches is also a pivotal trend helping to broaden participant inclusion and streamline processes. In this article, we delve into the essentials of clinical trial designs and emerging models.

Overview of the different types of clinical trial designs

Randomized controlled trials

Randomized Controlled Trials (RCTs) are considered the gold standard in clinical research due to their ability to minimize bias through randomization, allocation concealment, and blinding. Allocation concealment is a technique preventing selection bias by concealing the treatment from those assigning it to the participants until the moment of assignment. In other words, the clinicians do not know in advance what treatment the next participant will receive. This prevents clinicians from consciously or unconsciously influencing which participants will be assigned to a group based on their characteristics. Blinding is the practice of keeping participants, healthcare providers, and sometime the people in charge of analyzing the data unaware of the treatment each participant received. This avoids expectation bias during the interpretation of the trial.

RCTs are also known as parallel group designs, in which the participants are randomly assigned to either an experimental group or a control group. As Dan Goldstaub, co-founder of PhaseV, a company developing a platform for adaptive clinical trials, puts it: “Its purpose is to determine the effectiveness of a new treatment or intervention by comparing it with a control group that does not receive the treatment or receives a standard treatment.”

To put it in a nutshell, RCTs can provide high-quality evidence due to the random assignment of treatments, which helps eliminate bias. However, this comes with high costs and ethical dilemmas associated with withholding potentially better treatments from control groups. Moreover, the design and execution of RCTs can be complex and time-consuming, making them less feasible for urgent medical research needs.

Crossover trials

In this clinical trial design type, participants receive multiple treatments sequentially, serving as their own controls. According to Meri Beckwith, co-founder of Lindus Health, a clinical research organization, this enhances statistical power and reduces sample size needs but is only suitable for stable conditions and treatments with quick washout periods to avoid carryover effects.

Crossover trials allow participants to receive more than one intervention, with periods of no treatment or standard treatment in between. This type of clinical trial design is beneficial for comparing the effects of multiple treatments within the same group of participants, as each participant acts as their control, which can enhance the statistical power of the trial. However, crossover trials are not suitable for treatments with long-lasting effects such as chemotherapy, or for conditions that fluctuate over time like multiple sclerosis characterized by periods of relapse, as these factors can complicate the separation of treatment effects​​.

Factorial designs

These trials test multiple treatments and their interactions simultaneously. “This can provide comprehensive data but requires larger sample sizes to detect interaction effects,” said Goldstaub.

Factorial designs allow for the assessment of each intervention independently and in combination. Participants are randomly assigned to different treatment combinations, to test multiple interventions simultaneously and explore interactions between them. This design can significantly reduce the resources and time required as compared to conducting multiple separate trials. 

According to Goldstaub, the main challenge regarding these designs is that the statistical analysis of interactions between treatments can be complex, requiring advanced expertise and potentially complicating the interpretation of results.

Adaptive designs

These designs include predetermined points where the accumulating data from the trial is leveraged to adjust the trial design, without compromising the integrity of the study. Adaptive design improves the efficiency and flexibility of a trial, supporting more ethical trials and better use of resources, time, and money.

Adaptive designs offer the flexibility to modify trial parameters based on interim data analysis. This adaptability can lead to more efficient and potentially shorter trials by focusing resources on promising treatments and discontinuing ineffective ones. However, the implementation of adaptive designs requires advanced statistical expertise and thorough planning to ensure that modifications do not undermine the integrity of the trial​.

While the flexibility of these designs can enhance the ethical aspect of trials by minimizing participant exposure to ineffective treatments, it introduces statistical complexities and necessitates rigorous oversight to maintain the integrity of the trial.

Single-arm trials

In single-arm trials, all participants receive the experimental therapy and results may be compared to historical data or external controls instead of a concurrent control group. These trials are often used in situations where it is unethical or impractical to include a control group, such as with rare diseases or when the expected outcome without treatment is poor. These trials can provide preliminary data on the efficacy of a treatment, but they generally offer weaker evidence than randomized controlled trials due to the lack of a comparator group​.

“Single-arm trials are useful when it’s not practical or ethical to have a control group, such as with rare diseases or very novel therapies, and can also be quicker and less expensive to conduct than randomized controlled trials. With that, the results can be less definitive without a comparison group, leading to potential biases and limited ability to infer causality,” said Goldstaub.

Cluster randomized trials

Cluster randomized trials are conducted when interventions are best implemented at a group level, such as in different hospitals or communities, rather than targeting individual participants. This design helps to avoid contamination effects between groups but requires careful handling of data due to potential variations within each cluster. 

In clinical trials, contamination effects refer to instances where participants in different groups inadvertently share or are exposed to treatments or interventions intended for another group. This can occur if participants interact with each other or if interventions are not strictly contained within the designated groups.

According to Goldstaub, cluster randomized trials can reduce the risk of treatment effects transferring between individuals. “In these designs, the variability within clusters can affect the statistical power, often requiring larger sample sizes and costs. In addition, these trials are logistically complex, as managing and implementing interventions across multiple groups or locations can be challenging.”

An example of a cluster randomized trial could be set up for a gene therapy for Duchenne muscular dystrophy (DMD). The clusters would be different pediatric hospitals or specialized treatment centers. These clusters would be randomly assigned either to the intervention group which would administer the novel treatment, or the control group that would keep going with standard care. The measurement of of the outcome could be focused on muscle function improvement through standardized tests. 

Basket trials

Basket trials test how a single intervention performs across multiple diseases or conditions, which can reveal the broader applicability of a treatment but necessitate complex statistical strategies to manage diverse patient responses​. Basket trials are prominently used in oncology due to the diversity of cancers with common genetic mutations. 

This design allows researchers to evaluate the therapeutic impact of targeting a specific genetic aberration irrespective of the cancer’s location in the body. The central idea is to “basket” together patients who have different types of cancer but share a specific molecular signature that the drug in question is designed to target.

Patients are selected based on the presence of specific biomarkers or genetic mutations rather than the type of cancer. Advanced genomic profiling is used to identify eligible patients. All patients receive the same investigational treatment aimed at targeting the mutation, regardless of their cancer type.

The main outcome is the response rate to the treatment across all included cancer types, evaluated through metrics like tumor shrinkage, progression-free survival, or overall survival. Researchers also observe how the drug’s effectiveness may vary between different tumor types.

Basket trials are often adaptive, meaning that the trial design can be modified in response to interim results. For example, if a particular subgroup shows no benefit, it can be closed, while resources are redirected to more promising groups.

A hypothetical example could be a basket trial investigating a new inhibitor targeting the BRAF V600E mutation found in melanoma, colorectal cancer, and thyroid cancer. Despite the different tissue origins, the presence of this specific mutation might suggest a common pathway for drug action. Patients from these different cancer types would be grouped together to assess the drug’s efficacy, providing valuable insights into the mutation’s role across different tumors and the broader applicability of the drug.

How do specific and advanced clinical trial designs compare to traditional parallel-group designs?

Clinical trial designs such as crossover, factorial, and adaptive are taking more space in achieving clinical endpoints, offering distinct advantages over traditional parallel-group designs.

In parallel-group designs participants are randomly assigned to different groups, each receiving a different treatment, and the results from each group are compared at the end of the trial. This simple and proven approach requires a larger sample size to achieve the clinical endpoints. 

According to Beckwith, that is the reason why recently there has been a push to avoid RCTs. “I believe this is slightly misguided, as randomization is almost ‘magic’ in the way it simultaneously controls for all possible confounding variables, assuming an adequate sample size. Alternatively, researchers would need to predict and control for all possible confounding variables when conducting a retrospective non-randomized analysis, which can be very difficult.”

While parallel-group designs have proven their efficiency, innovative trials still have a lot to offer.

Goldstaub notes that in crossover trials, because each participant serves as their own control, variability among participants is reduced, potentially increasing the sensitivity to detect treatment effects. “Fewer participants may be needed to achieve similar statistical power as parallel-group designs, as each participant provides multiple data points across the different treatment conditions.”

Adaptive design enables leveraging the accumulated trial data for trial modifications, including changing the number of participants, adjusting dosages, stopping ineffective treatment arms, or reallocating resources to more promising ones. According to Goldstaub, the flexibility to make changes during the trial based on accumulating evidence can lead to more efficient trials, potentially reaching conclusions faster, and with fewer resources. In addition, these trials allow us to focus on more promising treatments, and reduce patient exposure to less effective or harmful treatments.

In 2019, the FDA released the guidance for adaptive trials which specifies some of the statistical aspects and interpretations of the results. In adaptive clinical trials, the type, method, and logic of adaptation are defined apriori. “We don’t know how the trial is going to play out, but we define how the trial should adapt to the different scenarios. For that, hundreds of thousands of simulations are submitted. Indeed, it is harder to design and interpret adaptive trials, but in recent years harnessing software and machine learning has made it easier to understand and more intuitive,” said Goldstaub.

Advanced clinical trial designs, such as basket trials, offer significant advantages but also present distinct challenges. Goldstaub notes the main challenge is statistical complexity. 

“Innovative designs often require more sophisticated statistical techniques and the use of technology, which can increase the complexity of the trial and the expertise required to manage it. In addition, gaining approval from regulatory bodies can be challenging due to the novel approaches these trials take. Ensuring ethical standards, especially in basket and platform trials, is crucial. These challenges should be managed through meticulous design, execution, and regulatory compliance, utilizing the advanced technology and expertise available.”

How to make a decision regarding the type of clinical trial designs?

When selecting a clinical trial design type, Beckwith thinks it comes to five crucial considerations:

• Objective of the study: The choice of design depends on the primary research questions. For instance, whether the trial aims to compare two treatments directly, investigate multiple treatments simultaneously, or adjust the study based on interim results influences the selection of parallel, factorial, or adaptive designs, respectively.
• Disease characteristics: The nature and stage of the disease being studied also impact design choice. Chronic diseases may be suitable for crossover designs, whereas acute conditions are not.
• Patient population: The availability and characteristics of the patient population, including disease prevalence and severity, can dictate the feasibility of enrolling sufficient participants for different types of clinical trial designs.
• Regulatory considerations: Approval from regulatory bodies might favor certain types of clinical trial designs over others based on historical precedence, perceived rigor, or comprehensive data requirements.
• Resource availability: Budget, time constraints, and available expertise can also influence the choice of design, as some designs may require more complex management and analysis. For instance, adaptive designs typically employ more sophisticated statistical techniques to allow for mid-trial adjustments based on accrued data. This requires advanced planning and simulation to ensure the integrity of the data and prevent biases.

Where do we stand in clinical trials and where are we heading?

Recent innovations in clinical trial designs are significantly influenced by the integration of real-world data (RWD) and the use of synthetic control arms (SCAs).

Use of real-world data and synthetic controls

“There are few recent advancements such as incorporating historical data into clinical trials – synthetic controls and priors in Bayesian trials. Adaptive trials in biosimilars development; more advanced adaptive enrichment trials; more complex platform trials,” said Goldstaub

RWD is increasingly used to enhance clinical trials by providing insights that help optimize care distribution and improve protocol designs. One innovative application of RWD is the creation of SCAs. SCAs use data from past trials or real-world sources to simulate the outcomes of control groups. This approach can reduce the need for enrolling large numbers of patients in control groups, decrease study costs, and speed up the trial process. The ability to leverage SCAs is supported by advancements in data analytics and increasing regulatory openness to these methods. These factors are pushing the boundaries of traditional clinical trial designs and opening up new possibilities for testing and development.

Emerging trends: Patient-centric clinical trial designs and decentralized technologies

Jytte van Huijstee, director of clinical trial operations at myTomorrows, a company facilitating the connection between patients and clinical trial stakeholders, highlights the shift towards more patient-centric clinical trials. This approach focuses on designing trials that are more considerate of the patient’s needs and convenience, which can enhance patient recruitment and retention.

Decentralized trials, which use technologies like mobile health apps and remote monitoring, play a crucial role in this trend. These technologies facilitate trials that are less disruptive to participants’ daily lives and can gather data more continuously and in real time. This shift supports better patient engagement but also broadens the scope of data collection, enhancing the quality and applicability of trial results.

A slow shift toward novel clinical trial designs

“Regulators and industry are slowly beginning to embrace adaptive clinical trial designs. This is a good thing, but unfortunately, as with many developments in clinical research, industry adoption is sometimes lagging even behind the regulatory guidance,” said Beckwith.

According to Beckwith, sponsors are still often hesitant to propose adaptive designs even if regulators have signaled their acceptance. Ultimately, if an adaptive clinical trial is well designed, and blinding can be maintained there is no potential downside, and there is potential to significantly save cost, shorten timelines and protect patient wellbeing.