Clinical Trial Study Design Types

What is clinical trial design?

Clinical trial design refers to the framework that defines every aspect of how a clinical trial is to go about answering the research hypothesis. It is a crucial part of every clinical trial, and several factors must be taken into account in order to set the trial up with a strong statistical foundation for its conclusions.

Observational studies vs. experimental studies (clinical trials)

At the broadest level, there are two principal categories of study design types: observational studies and experimental studies.

Observational studies are non-interventional. In other words, researchers monitor patients over an extended period of time without administering any form of medical intervention. This type of study is designed to observe health outcomes in patients outside of the controlled environments imposed in experimental studies, and data is typically collected via case reports, case studies, and surveys. Generally, these types of studies are not considered clinical trials.[¹]
Experimental studies test some type of medical intervention on trial participants, which could be a novel drug, a surgical intervention, a medical device, a lifestyle change, etc. Also known as clinical trials or interventional studies, in this type of research study, patients are assigned to a given exposure/treatment (in study/treatment “groups” or “arms”), and researchers observe the outcomes. Such studies focus on the intervention’s efficacy and safety, as well as the occurrence of any adverse reactions.

Given that observational studies are not technically clinical trials, in this article we will focus only on the design of clinical trials, i.e., experimental studies.

Designing a clinical trial: Overview of factors in clinical research study designs

There are several factors that make up the core elements of a clinical trial design. With advancements in technology and the emergence of novel digital solutions, this list continues to grow as innovative clinical trial designs are born.[²]

To keep things relatively simple, we will discuss five primary factors that collectively define the most-common trial design types that are currently dominant in clinical studies. These core factors will continue to be integral considerations in clinical trial design, even as new designs emerge and evolve.

Taken together, these five study design factors can conceptually describe, to a large extent, the structure of the research method being employed to reach a conclusion about the effectiveness and safety of the investigational product:

Factor 1: Types of research hypotheses

Comparative
Equivalence
Non-inferiority

Factor 2: Control group

Non-controlled
Placebo-controlled
Standard treatment-controlled

Factor 3: Randomization

Non-randomized
Randomized

Factor 4: Blinding

Open-label
Blind
Single-blind
Double-blind
Triple-blind

Factor 5: Structure of trial treatment arms

Parallel
Cross over
Factorial design

Factor 1: Types of clinical trial research hypotheses

A clinical trial aims to answer a research question, known as a hypothesis, and for it to serve this goal, the hypothesis must be clearly defined. The hypothesis sets the stage for designing the clinical trial to capture data that can be used to prove or disprove the hypothesis.

An example of a poor hypothesis would be “Does drug X work?” This question would be impossible to answer - and if it were answered, the statistical viability of the conclusion would not be scientifically sound or valid - because this hypothesis is vague, open-ended, and ill-defined.[²] In other words, the research question needs to be laid out clearly, with clearly defined metrics or endpoints that will serve to prove or disprove it in a way that is not subject to major bias.

Therefore, the first step to framing a clinical trial design is to establish a fully-formed hypothesis, which will be used to assess an intervention's safety and/or effectiveness according to specific parameters. Generally, hypotheses in clinical trials take the form of proving or disproving the superiority, equivalence, or non-inferiority of a medical intervention in relation to a control, which could be the standard treatment or a placebo.

Comparative trials (superiority trials)

A comparative trial aims to determine whether a novel therapy A is better than (superior to) the standard therapy B, in terms of a specific health outcome or a certain health condition. Also known as a superiority trial, this type of study can also be used to compare a novel therapy to a placebo if no standard treatment is currently available.

Clinical trial designs framed as superiority trials can be used to test novel interventions that are predicted to be able to provide patients with a treatment option that is more effective and/or which has fewer side effects than currently available options.

Equivalence trials

An equivalence trial aims to demonstrate that an experimental treatment is neither better nor worse than the standard treatment. Equivalence trials are critical for identifying novel drugs that can replace existing or standard treatments, but with comparatively milder or less severe side effects. Additionally, equivalence trials allow researchers to discover medications that are equally effective for patients who are allergic to standard treatments.

Non-inferiority trials

A non-inferiority trial compares the novel treatment to a control to test the hypothesis that the novel treatment is not unacceptably worse than the control.[³] Similar to equivalence trials, the objective is to identify substitutions for existing treatment options that have similar efficacy and side effects. A non-inferiority trial may be conducted to try to prove that a generic alternative is not less-effective than a brand-name drug.

Factor 2: Control group

A control group serves as a baseline against which the experimental treatment is compared. This allows researchers to assess the results of an intervention in relation to something that is more clearly or easily defined (for example, a standard treatment for which lots of data is available, or a placebo which should have no effect at all).[⁴] There are certain cases in which a clinical trial may not require a control group.

Uncontrolled experiment

A clinical trial in which the efficacy and toxicity of a medical intervention is not compared against a control is known as an uncontrolled experiment. In such trials, researchers normally compare their results to those of previous clinical trials conducted by other investigators.

There may be difficulties with statistically validity when taking this approach. For example, if 70 out of 100 trial participants see clinical improvements after using a novel therapy, researchers could not make solid quantitative claims about the novel drug’s effectiveness since it isn’t clear how many people may have gotten better even without the intervention. Comparisons to historical data are also often statistically questionable given the high variability in prognostic factors between participants in different data sets.

Therefore, uncontrolled clinical trials are more typical for phase I and II trials, wherein researchers are more focused on investigating the pharmacokinetics of a novel therapy or its tolerability in order to determine safe and optimal dosages.

Placebo-controlled trial

A placebo-controlled trial directly compares the treatment group with a control group that simultaneously receives an inert, inactive substance known as a placebo. The purpose of this is to determine whether the novel therapy is more/less/equally effective or safe than no treatment at all.

In general, researchers should only use a placebo when no standard treatment is available. Otherwise, such a trial is considered unethical, as ill patients would be missing out on treatment that could help them. Furthermore, the placebo must not cause any permanent damage or bring harm to the participant in any way.[⁵]

Typically, placebo groups are introduced in phase III clinical trials, after investigators have established safe dosages. In some designs, participants are put on a schedule where they eventually receive the novel or standard treatment after a specified period, in order to address ethical concerns and patient concerns (fear about receiving a placebo is a major deterrent to participation in clinical trials).

Treatment-controlled trial

A treatment-controlled clinical trial design takes a similar approach to a placebo-controlled trial, but the control group is given a standard treatment or other existing treatment rather than placebo. Therefore, researchers can test whether the novel treatment is better or worse than currently available therapy options.

Considered more ethical than uncontrolled clinical trials, treatment-controlled trials are preferred in phase III trials as they allow researchers to assess a novel therapy's superiority, equivalence, or inferiority to established treatments, while providing all participants with some form of active treatment.[⁵]

Factor 3: Randomization in study designs

Randomization determines how trial participants are divided into treatment groups or “arms,” i.e., the novel intervention, control, or placebo group.

Randomization is a practice that is extensively used in clinical trial design as it eliminates selection bias, which has a high potential to affect the measured outcomes. Moreover, randomization is considered to be the most ethical method of assigning participants fairly.

Non-randomized

In non-randomized clinical trial designs, the participant can choose the treatment arm they wish to join, or researchers knowingly assign participants to specific groups.[⁶]

This type of design could be applied in:

Single-arm studies (those with only one treatment group)
Phase I or II trials studying toxicity, tolerability, and dosage
Some clinical trials with small sample sizes
Studies for incurable diseases

Randomized trials

In randomized trials, participants are assigned to treatment groups by chance (based on various different models of chance, as discussed next). Neither researchers nor trial participants have a direct influence on the assignment, thus reducing selection bias and ensuring that no participant is treated favorably or unfavorably.

Several different randomization methods are commonly used, including fixed, blocked, stratified, and adaptive randomization. There is lots of statistical background explaining the different randomization methods, which we will not go into in this article. For more information about randomization methods, refer to our article on randomization in clinical trials.

In clinical studies, randomization - more specifically randomized controlled trials, or RCTs - are considered the “gold standard.” It is the preferred approach that is generally adopted in any trial with more than one trial arm and a sufficient number of participants.

Factor 4: Blinding

Blinding in a clinical study describes who is aware of which treatment groups participants are assigned to. Also known as allocation concealment, blinding aims to prevent selection bias as well as behavioral changes that can affect clinical trial results.

However, blinding is more than non-disclosure of allocation. It is also about keeping details of the treatment concealed so that individuals cannot guess which treatment they themselves or other participants are receiving. Therefore, all placebos, novel drugs, and standard treatments should generally look similar and be administered identically.

Let’s look at different types of blinding in clinical research studies.

Open label study (non-blinded)

In a non-blinded study, also known as an open-label study, everyone, including researchers and the trial participants, knows what treatment they are receiving (or administering).

These types of trials are most commonly seen in phase IV clinical trials, wherein the long-term effects of an approved medical intervention are observed without a control. In this case, it is not possible nor necessary to hide the treatment from patients.[³]

Additionally, non-blinded trials may be conducted when it is overly challenging, impractical, or unethical to blind participants, such as in the case of surgical interventions.

Single blind study

In a single-blind study, only the trial participants are unaware of their treatment group. The investigators are aware of which group each participant is assigned to. This type of blinding reduces the physical and psychological responses a trial participant may have to the medical intervention. For example, if a trial participant finds out they are receiving a placebo, they may make lifestyle changes that could influence study outcomes.

Furthermore, when blinded, trial participants are often:

More compliant
More willing to follow study protocols
Less likely to use adjunct therapies
Less likely to drop out[²]

Additionally, single-blind studies are essential for clinical trials that record subjective outcomes, such as pain, because any knowledge of the treatment a participant is receiving has the potential to affect his/her response.[²]

Double blinded study

In a double-blinded study, both trial participants and researchers are unaware of the treatment group assignments.

This trial approach is more objective and clinically valid than a single-blind study because it extends to eliminate any researcher bias that can lead to behaviors (whether intentional or unintentional) such as:

Sharing information with trial participants
Adjusting dosages
Providing adjunct therapies
Influencing a participant's decision to continue with a trial
Measuring outcomes incorrectly or from a biased perspective

A double-blind approach is particularly useful when testing novel therapies, such as new drugs being studied in trials with more than one treatment group.

Triple blinded study

In triple-blind studies, treatment assignments are hidden from participants, researchers, and whoever will interpret the study outcomes, such as data analysts, statisticians, and even sponsors.

While this approach is considered the most objective of all, it is also the most challenging to implement, requiring many controls to ensure participant assignment remains concealed throughout the trial, all the way through to data analysis.[²] Nonetheless, there are methods such as codification of assignments that make triple-blinding feasible when implemented strictly and in a well-controlled manner. Triple blinding effectively reduces assessment bias, as data analysts do not know which data set relates to which treatment group, leading to increased objectivity.

Triple blinding might be implemented in phase II and III clinical trials of novel interventions with multiple treatment groups, when it is necessary to ensure that statistical testing is as scientifically valid as possible.

Factor 5: Intervention models: Treatment group vs. control group

Intervention models in clinical studies dictate the way in which participants are divided into different treatment groups, based on how those groups will be compared through statistical assessment to answer the research hypothesis.[⁷] For a more complete description of this topic, you can consult our article on intervention models.

Single group assignment

Single-group assignments are used in clinical trials with only one treatment arm; all participants receive the same therapy.

This type of assignment is often used in phase I clinical trials testing the toxicity of a novel therapy, and in phase IV clinical trials designed to study the long-term effects of a newly approved treatment.

Parallel design (parallel group)

Parallel design, or parallel group assignment, is the most common intervention model, wherein two treatment groups are studied simultaneously. These treatment groups could receive different drugs, a novel drug and a placebo, or different dosages of the same drug.

Also known as a non-crossover model, participants remain in the treatment group they are assigned to from the beginning through to the end of the clinical trial. Randomized parallel-group designs are effective for several types of clinical trials, such as those:

For health conditions that are known to progress with time
Where outcomes are quantitatively measurable
Where there are risks of carryover effects

However, parallel design studies require a relatively large sample to minimize patient variations that can affect study results.[²]

Cross over trial

A cross over clinical trial design aims to reduce the effects of covariate variability amongst patients by putting trial participants on a treatment schedule where they receive both the novel intervention and the placebo (or standard treatment) at different times, in a specified order.

Since crossover trials essentially treat each participant as their own control, they can elicit statistically valid results even with a smaller sample size. However, the effects of treatments may carry over. Therefore, trial participants must undergo a sufficient washout period before being switched to the other group to clearly separate/distinguish the effects of the prior treatment.[⁵]

Furthermore, this clinical trial design has two specific requirements that limit its use. First, it’s best suited to studying health conditions that are chronic but relatively stable so that the effects can be studied over a long duration. Second, the effects of each drug should be reversible so their effects on participants can be appropriately compared.[⁵]

Factorial design

A factorial-design clinical trial combines two or more medical interventions in various combinations, allowing researchers to study the interactive effects of novel drugs with smaller sample sizes. This intervention model is unique in that it lets researchers test multiple research hypotheses in a single trial.[⁵]

To illustrate, let’s consider the example of a 2 x 2 factorial design, studying treatments A and B, in which patients are randomized into four groups:

Group 1: Treatment A and placebo

Group 2: Treatment B and placebo

Group 3: Treatments A and B

Group 4: Placebo only

Under this design, researchers can study the effects of each individual drug, in relation to the other drug and to the placebo, as well as any potential interactions between the drugs. This approach is useful, for example, for studying adjunct therapies to determine if they are beneficial, and for developing improved drug regimens for chronic conditions.

Finalizing the trial design: How to select between different design types

When designing a clinical trial, sponsors and CROs need to weigh each of the design factors we covered above against certain considerations, such as[⁵]:

The nature of the disease being studied
The research objectives
The number of interventions being studied
The treatment duration and course
The total time frame available
The patient sample size available
Logistics and costs
Possible carryover effects

Along with these considerations, and to ensure the ethical and statistical validity of a study, sponsors and CROs must keep the following three priorities front and center at all times:

1. Ensuring patient safety

The first priority in clinical trial design should be patient safety. At no time should any aspect of a clinical trial put the trial participant at an unacceptable risk. Remember, the clinical trial design must be found to be ethically sound in order to receive approval, and strong results should never be favored at the expense of patient health and safety.

2. Maximizing statistical power

Clinical trials should be designed so as to gather high-quality, valuable data that can clearly prove or disprove the research hypothesis. In research studies, statistical power, or just power, refers to the likelihood of not making a type II error (“false negative;” more on this next). The main factor for ensuring power is to accrue a sufficient sample size. Many studies are underpowered, or end up being underpowered due to poor enrollment, and thus churn out subpar results that have diminished or even no clinical significance.

3. Minimizing type I and type II errors

Type I and type II errors are the main errors that might arise during statistical analysis of study data. These errors can lead to inaccurate conclusions, which in the worst case could lead to inappropriate or even harmful treatments being approved or poor clinical guidelines being set forth. Therefore, study design should be optimized as much as possible for minimizing the probability of type I and type II errors. This is easier said than done, because multiple trial design factors influence these error probabilities, including sample size, randomization method and intervention model, and level of probability/statistical confidence, and there are also ethical implications, for example increasing sample size to increase power implies exposing more participants to the novel intervention.[⁸]

Type I errors, also known as false positives: A treatment effect is confirmed when in reality there was no effect.[⁸]
Type II error, also known as false negatives: It is concluded that there was no treatment effect, despite there actually being one.

Conclusions

When it comes to determining the optimal clinical trial design, it’s important to consider all of the factors we’ve discussed in this article, as well as considerations such as the complexity of the protocol (which plays into patient and site burden) and ethics. It all starts with clearly defining and framing the research question to be answered, but the best way to answer that question is highly dependent on the specific details of the trial. These primary trial design factors should be selected in consideration of the main priorities discussed in the final section; the best clinical trial design will minimize the risk of statistical error through robust trial design leading to high-quality data, while maximizing the safety of trial participants.