What is a clinical trial?
A clinical trial is a clinical study in which participants are assigned according to a pre-defined therapeutic strategy or plan (protocol) to receive a health-related intervention, such as a medicine, in order to investigate its effects on health outcomes, usually compared to another (or sometimes no) treatment.
Clinical trials are used to evaluate clinical practices that do not fall within the current practices of a country, or to evaluate a new medicine (investigational medicinal product).
Clinical trials are used to generate data on the safety and efficacy of the intervention. Clinical trials are conducted only after a regulatory authority approval and ethics committee review. Clinical trials are often characterised in Phases from I (first-in-human), II (exploratory), III (confirmatory) to IV (post approval).
Previously, the terms clinical study and clinical trial were used synonymously.
Types of clinical trial designs
There are several types of trial designs, but in this topic, we will focus on only a few:
- Non-randomised controlled trials – is a trial where the participants are assigned to the treatment by a method that is not random. The investigator defines and manages the alternatives.
- Randomised controlled trials – is a trial where the participants are randomly allocated to one or another of the different treatments under study. They can be:
- Parallel group
- Single blind – a study in which one party, either the investigator or participant, is unaware of what medication the participant is taking; also called single-masked study.
- Double blind – a clinical trial design in which neither the participating individuals nor the study staff knows which participants are receiving the experimental treatment and which are receiving a placebo or comparator. Double-blind trials produce objective results, since the expectations of the doctor and the participant about the experimental medicine do not affect the outcome; also called double-masked study.
In a clinical trial design, there are a number of different types of comparisons that can be included:
- When the aim of a randomised controlled study is to demonstrate that one treatment is superior to another, a statistical test is employed and the trial is called a superiority trial. Often a non-significant superiority test is wrongly interpreted as proof of no difference between the two treatments. With statistical tools, at most one can show that they are equivalent. This type of study design is often used to test the effectiveness of a treatment compared to placebo.
- In an equivalence trial, the statistical test aims at showing that two treatments are not too different in characteristics. These trials are designed to demonstrate that one treatment is as effective as another.
- Finally, a non-inferiority trial is designed to demonstrate that a treatment is at least not (much) worse than a standard treatment. Non-inferiority comparison trials are often used for efficacy studies and the control is a medicine already on the market.
- Dose-response relationship trials are a class of trials that determine an optimal dose of a specific medicine. The dose response of a medicine is important in pharmacology, pharmacokinetics, toxicology and clinical research. Dose response studies may be part of larger research to develop new treatments or to supplement existing knowledge of a medicine whose benefits may have already been established.
Dose-response relationship comparison studies investigate:
- The shape and location of the dose-response curve
- The appropriate starting dose
- The optimal strategies for individual dose adjustments
- The maximal dose beyond which additional benefit would be unlikely
Randomisation in clinical trials
Randomisation is the process of making something random, by chance. This can be applied in:
- Selecting a random sample of a population.
- Allocating units to different conditions with no order
In clinical trials, randomisation refers to the process of assigning a trial participant into treatment or control groups (or arms) using an element of chance to determine the assignments. This procedure is made in order to reduce bias.
The assignment can be done using different tools, such as envelopes, random numbers, scratch sheets, computer generated sequences and contacting interactive voice recognition systems (IVRSs).
Non-randomised controlled trials
In a non-randomised study, participants are allocated into control or treatment arms by a method that is not random. The investigator defines and manages the alternatives.
It is appropriate to use a non-randomised controlled trial design when randomisation may be unethical, not suitable due to legal or political challenges or when it is impractical in terms of cost and convenience.
There are many possible types of non-randomised controlled trials that do not use appropriate randomisation strategies (sometimes called quasi-randomised studies).
Non-randomised controlled trials have the potential to study two groups that are not strictly comparable. However, there are different types of controls that can be used in non-randomised trials:
- Concurrent controls: the treatment and control group participants are matched based on demographic and other characteristics. They receive different treatments at the same time.
- Historical controls: the investigators compare outcomes among a group of participants who are receiving a new treatment (experimental group) with outcomes among participants who received standard treatment in a previous period (control group).
Randomised controlled trials
Randomisation is the process of making something random, by chance. This can be applied in: Selecting a random sample of a population; or allocating units to different conditions with no order. In clinical trials, randomisation refers to the process of assigning a trial participant into treatment or control groups (or arms) using an element of chance to determine the assignments. This procedure is made in order to reduce bias. The assignment can be done using different tools, such as envelopes, random numbers, scratch sheets, computer generated sequences and contacting interactive voice recognition systems (IVRSs).
From trial inclusion criteria, there is equal allocation with regard to gender, age and health status. It removes the potential of bias in the allocation of participants and tends to produce comparable groups, however it cannot eliminate accidental bias.
There are different types of randomised trial designs:
- Factorial design attempts to evaluate two interventions compared to a control in one trial. Each participant receives two different medicines: x and y; x and control; y and control; control and control.
- In withdrawal trials, participant receives a test treatment for a specified time and are then randomised to continue the test treatment or placebo (withdrawal of the active therapy).
- In parallel group trials each participant is randomly assigned to a group, and all the participants in the group receive (or do not receive) an intervention.
- In cross-over trials, over time each participant receives (or does not receive) an intervention in a random sequence.
Parallel group, cross-over and match-paired designs
A parallel design is where two groups of treatments, A and B, are given so that one group receives only A, while another group receives only B throughout the trial. Participants are randomised into different study arms, but there is approximately an equal number in each group.
The statistical mean of each group is then compared, and unexplained variability is attributed to between patient differences (age, weight, disease severity)
The parallel design is the most common design for clinical trials. However, when participants are randomised, there are practical considerations to take into account. For example, missing data over time may lead to some participants not being included in the final analysis.
In a cross-over clinical trial design, participants receive a sequence of different treatments. For instance: Patient X and Y are randomised into two different treatment groups. Patient X receives Treatment A during the first period of the study; Patient Y receives Treatment B. After the first period is over, there is a washout period. Patient X then receives Treatment B for the second period of the study while Patient Y receives Treatment A.
Each treatment starts at an equivalent point, and each individual (with a stable and chronic disease) serves as his/her own control.
Single-patient response to different treatments is compared, normally the variability of response is less than the variability from parallel group designs. It also allows fewer patients, but the study must run for a longer period of time.
Randomisation is not always a practical nor ethical method of assigning a participant to a comparison group. For obvious reasons, for example, we cannot assign individuals to wear a seat-belt or not to use it in an attempt to assess the effect of seat-belts in car accidents. In such situations, the method of matched-pair clinical trial design is widely used.
This design can be used when the experiment has only two treatment conditions and participants can be grouped into pairs – matched up to be as similar as possible in order to reduce variability in results.
Matching is typically used in comparative observational studies, in which individuals are either self-selected into identifiable groups (for example, seat-belt wearers and not) or individuals have fixed, pre-determined characteristics that dictate their group membership (for example, males and females). The primary advantage of matching is that biases due to baseline group differences are minimized, thereby reducing the variability, and increasing the precision, of the group comparisons.
Randomisation techniques: stratification vs cluster sampling
Normally patients would be allocated to a treatment group randomly, and while this maintains a good overall balance, it can lead to imbalances within sub-groups. For example, if a majority of the patients who were receiving the medicine happened to be male, or smokers, the statistical usefulness of the study would be reduced.
The traditional method to avoid this problem is to stratify participants according to a number of factors (e.g. age or smokers and non-smokers) and to use a separate randomisation list for each group. Each randomisation list would be created such that after every block of a number of patients, there would be an equal number in each treatment group.
|Participant No||Random arm/group||Participant No||Random arm/group|
Cluster sampling is a sampling technique that has been used when ‘natural’ but relatively homogeneous groups are evident in a population. In this technique, the total population is divided into these groups (or clusters) and a simple random sample of the groups is selected. Then the required information is collected within each selected group. This may be done for every element in these groups.
The population within a cluster should ideally be as heterogeneous as possible. Each cluster should be a small-scale representation of the total population.
This form of sampling is generally useful for interviews; however, it could be useful for clinical trials if there are multiple clinics or hospitals that can take part in the study.