Designing a foolproof trial

Designing a foolproof trial

In contrast to the ingeneous aspects of discovery, clinical trials approach science by validating the promise of a cure. Since it has to answer logical questions, a trial should be designed accurately to ensure that the truth flows in the right direction, Katya Naidu learns.

Conducting a clinical trial without a study design is like trying to construct a building without a blueprint. When an engineer plans a building, he visualises the end result; crafts plans and devises methods to achieve the desired structure.

The same applies to the design of a clinical trial. “A study design considers the type of control to be selected, the type of patient population to be studied, the indication for treatment, methods to eliminate bias in the study, such as randomisation and blinding, duration of dosing and methods of clinical assessments,” explains Dr Anoop Kumar Agarwal, Principal, Institute of Clinical Research, India (ICRI).

Exercising control

– Dr Anoop Kumar Agarwal
Principal Institute of Clinical Research India (ICRI)

One of the first questions that is asked while designing a trial is—what should be the the type of control in the conduct of the trial. When a trial is controlled, it uses one or more control groups. These groups are not given the test drug, to enable investigators to determine whether an observed effect is truly caused by the test drug.

There are also certain trials which are not controlled. However, experts feel that these trials may miss both the benefits and the problems.

A trial using the control group could use it to demonstrate that the test treatment is efficacious by showing that it is superior to the control. These trials are called superiority trials. Sometimes, it might just be sufficient to show that a new drug has similar efficacy to a standard agent and this is called an equivalence trial. “A non-inferiority trial is an equivalence trial, which shows that the new drug is not less effective than the control by more than a defined amount, generally called the margin,” observes Dr Sudheirr Dhillon, Head, Oncology, Merck Specialities. Equivalence testing may be relevant for drugs that show high benefits in reducing toxicities experienced in control treatment

In random

A wise person would say that it is wrong to judge randomly. But the rules of conducting clinical trials are a little different. Randomising the study population is an important feature of trial design. This method uses chance to allocate subjects to the investigational and control groups, so that individual features or characteristics do not act as points of bias.

There are again various levels of randomisation. In simple randomisation, subjects are usually assigned via a computer program or a table of random numbers. Whereas restricted randomisation ensures better balance between the groups in terms of their size (block randomisation) or specific characteristics (stratification and minimisation). “The characteristics need to be precisely defined. For example just to say that subjects will be grouped by “age” is not sufficient. The actual age groups need to be stated, for example below 50 years and above 50 years,” asserts Dhillon.

The choice of the extent of randomisation depends on the size of the trial. Simple randomisation may be adequate in large trials, but block randomisation is better suited to small sample size trials to ensure balance within groups at all times.

When a trial is randomised extensively, a technique called stratification comes into the picture. Stratification assures that different variables that may affect the intervention’s success are analysed separately. In smaller studies, it provides more balanced groups than random allocation. This methodology can be used to enable appropriate sub-group analysis to study differences between groups within a trial. Minimisation is yet another technique to achieve a good balance of variables between groups.

A blind approach

After choosing the degree of randomisation, it is also necessary to suppress the facts in order to assure a random approach all through the trial. The method used to achieve this is blinding. If the subject has no knowledge of the drug they are receiving but everyone else involved in the research does, then it is called a single blind study. If neither the clinician, nor the research team nor the subject knows whether the experimental drug or control is given, it is called a double blind trial. “Blinding helps substantially in reducing bias in a trial. Blinding may not be possible to achieve in many cancer trials due to differences in complexities of regimens being compared as well as differences in delivery systems and routes of administration,” admits Dhillon.

In addition, there also various methods of the allotment of subjects to different groups. In parallel group design, subjects are randomised to one of the two or more groups. Each group is allocated a different treatment at the same time. In a crossover trial, two or more treatments are applied sequentially to the same subject. In a factorial design, various possible combinations of two or more treatments are evaluated simultaneously.

Selecting subjects

After determining the controls to be implemented, it is also necessary to determine the characteristics of the population on whom the trial is to be conducted. “The study population of a trial is determined based on the objective of the trial, the disease to be studied, the age group in which it is to be used, the duration of the treatment, genetic constitutions leading to variations in response, concurrent diseases and medication,” says Kumar.

The sample size of the population selected itself is determined by various factors like resources, budgets and the number of patients available, characteristics of the population being studied and the method of data analysis. “The parameters involved in deciding sample size are design of the trial, objective, genetic variation, level of significance, sensitivity of the study, and the prevalence of the disease,” says Kumar.

The statistical viability of the sample size is another factor to be considered. “Statistical factors to take into account when calculating the sample size include statistical power (the chance of demonstrating an improvement if it really exists), clinical relevance, statistical and clinical significance, and control group mean and variability,” says Dhillon. Generally, trials that evaluate issues of major clinical importance must be designed with high statistical power. Also greater the variability within a given level of efficacy, larger is the number of patients required to demonstrate relevant differences in treatment. Many trials require defined “times-to-event” or certain number of events to occur to demonstrate differences between groups.

More numbers

The significance of statistics extends beyond sample size determination in the design of a clinical trial. The outcomes of the trial need to be statistically analysed to check if the results meet the required statistical power. One main question to be answered in the course of the trial is whether the results for the drug being studied are statistically significant. This is important because it instils in the researchers the confidence that the output of the trial is due to the drug and not just a matter of coincidence. Another question is whether the results obtained are clinically relevant. This gives an edge to the molecule under study and determines if the statistical significance is relevant when compared with toxicity issues.

The ability of a trial to estimate the true size/impact of the effect of a drug can be reduced in many ways:

  • Poor compliance with therapy and large numbers of withdrawals: Withdrawals can cause bias if they are not considered in the final analysis, as an intent-to-treat analysis, because they can cancel out the benefits of randomisation
  • Poor responsiveness of the enrolled study population to the test drug effects
  • Use of concomitant non-protocol medication or other treatment that interferes with the test drug or that reduces the extent of the potential response or the clarity of interpretation
  • An enrolled population that tends to improve spontaneously, leaving no room for further drug-induced improvement
  • Poorly applied diagnostic criteria (subjects lacking the disease to be studied)
  • Biased assessment of the endpoint because of knowledge that all subjects are receiving a potentially active drug, e.g. a tendency to attribute all headaches to the drug or to attach too much weight to subjective reports of improvement

In the end

All the calculations lead to determine the endpoint which is the objective of the trial. However, what to achieve in the endpoint is also pre-determined in the trial design to clarify the idea of the study. “Endpoints need to be defined early on, before the protocol is decided. To avoid bias, all patients, including controls, should be assessed at the same frequency, and if possible the results should be evaluated by a third party who does not know which group the patients have been assigned to (called an “independent audit” carried out by an independent review committee comprising of experts),” asserts Dhillon.

In most diseases, the primary efficacy outcome used in clinical trials is “cure”. The endpoints that may be measured in trials generally are response rates and the duration of response. There are different types of endpoints though. For example, in a cancer trial, while safety/tolerability is not usually the primary endpoint, it will always be an important consideration, and will be measured as a secondary outcome. Quality of life may also be an important primary or secondary endpoint. Some endpoints usually used are—overall survival, patient survival, progression free survival and disease free survival (amount of time elapsed before recurrence or relapse). “The choice of endpoints is impacted by the defined difference or improvement one expects to be able to demonstrate with the new drug/intervention. The clinical and statistical relevance of these differences will have a clear impact,” says Dhillon.

Though there are guidelines on the design of a good study, there are no general methods to determine the best practice to be followed. Each design is tailored to the specific molecule to be tested and depends to a large extent on the results of the pre-clinical studies.

However, an essential factor that is common to the design of all clinical trials is the significance given to the ethical aspect. Not only is it an obligation while dealing with humans, but it also speaks of the integrity of the study designers.