CHAPTER 1
Fundamental concepts
This chapter provides a brief background to clinical trials, and why they are considered to be the ‘gold standard’ in health research. This is followed by a summary of the main types of trials, and four key design features. Further details on design and analysis are given in Chapters 3–7.
1.1 What is a clinical trial?
There are two distinct study designs used in health research: observational and experimental (Box 1.1). Observational studies do not intentionally involve intervening in the way individuals live their lives, or how they are treated. However, clinical trials are specifically designed to intervene, and then evaluate some health-related outcome, with one or more of the following objectives:
- to diagnose or detect disease
- to treat an existing disorder
- to prevent disease or early death
- to change behaviour, habits or other lifestyle factors.
Some trials evaluate new drugs or medical devices that will later require a licence (or marketing authorisation) for human use from a regulatory authority, if a benefit is shown. This allows the treatment to be marketed and routinely available to the public. Other trials are based on therapies that are already licensed, but will be used in different ways, such as a different disease group, or in combination with other treatments.
An intervention could be a single treatment or therapy, namely an administered substance that is injected, swallowed, inhaled or absorbed through the skin; an exposure such as radiotherapy; a surgical technique; or a medical/ dental device. A combination of interventions can be referred to as a regimen, such as, chemotherapy plus surgery in treating cancer. Other interventions could be educational or behavioural programmes, or dietary changes. Any administered drug or micronutrient that is examined in a clinical trial with the specific purpose of treating, preventing or diagnosing disease is usually referred to as an Investigational Medicinal Product (IMP) or Investigational
Box 1.1 Study designs in health research
Observational
Cross-sectional: compare the proportion of people with the disorder among those who are or are not exposed, at one point in time.
Case-control: take people with and without the disorder now, and compare the proportions that were or were not exposed in the past.
Cohort: take people without the disorder now, and ascertain whether they happen to be exposed or not. Then follow them up, and compare the proportions that develop the disorder in the future, among those who were or were not exposed.
Semi-experimental
Trials with historical controls: give the exposure to people now, and compare the proportion who develop the disorder with the proportion who were not exposed in the past.
Experimental
Randomised controlled trial: randomly allocate people to have the exposure or control now. Then follow them up, and compare the proportions that develop the disorder in the future between the two groups.
An ‘exposure’ could be a new treatment, and those ‘not exposed’ or in a control group could have been given standard therapy.
New Drug (IND).# An IMP could be a newly developed drug, or one that already is licensed for human use. Most clinical trial regulations that are part of law in several countries cover studies using an IMP, and sometimes medical devices.
Throughout this book, ‘intervention’, ‘treatment’ and ‘therapy’ are used interchangeably. People who take part in a trial are referred to as ‘subjects’ or ‘participants’ (if they are healthy individuals), or ‘patients’ (if they are already ill). They are allocated to trial or intervention arms or groups.
Well-designed clinical trials with a proper statistical analysis provide robust and objective evidence. One of the most important uses of evidence-based medicine is to determine whether a new intervention is more effective than another, or that it has a similar effect, but is safer, cheaper or more convenient to administer. It is therefore essential to have good evidence to decide whether it is appropriate to change practice.
World Health Organization definition of a clinical trial1,2
Any research study that prospectively assigns human participants or groups of humans to one or more health-related interventions to evaluate the effects on health outcomes.
Health outcomes include any biomedical or health-related measures obtained in patients or participants, including pharmacokinetic measures and adverse events.
1.2 Early trials
James Lind, a Scottish naval physician, is regarded as conducting the first clinical trial.3 During a sea voyage in 1747, he chose 12 sailors with similarly severe cases of scurvy, and examined six treatments, each given to two sailors: cider, diluted sulphuric acid, vinegar, seawater, a mixture of several foods including nutmeg and garlic, and oranges and lemons. They were made to live in the same part of the ship and with the same basic diet. Lind felt it was important to standardise their living conditions to ensure that any change in their disease is unlikely to be due to other factors. After about a week, both sailors given fruit had almost completely recovered, compared to little or no improvement in the other sailors. This dramatic effect led Lind to conclude that eating fruit was essential to curing scurvy, without knowing that it was specifically due to vitamin C. The results of his trial were supported by observations made by other seamen and physicians.
Lind had little doubt about the value of fruit. Two important features of his trial were: a comparison between two or more interventions, and an attempt to ensure that the subjects had similar characteristics. That the requirement for these two features has not changed is an indication of how important they are to conducting good trials that aim to provide reliable answers.
One key element missing from Lind’s trial was the process of randomisation, whereby the decision on which intervention a subject receives cannot be influenced by the researcher or subject. An early attempt to do this appeared in a trial on diphtheria in 1898, which used day of admission to allocate patients to the treatments.4 Those admitted on one day received the standard therapy, and those admitted on the subsequent day received the standard therapy plus a serum treatment. However, some physicians could have admitted patients with mild disease on the day when the serum treatment would be given, and this could bias the results in favour of this treatment. The Medical Research Council trial of streptomycin and tuberculosis in 1948 is regarded as the first to use random numbers.5 Allocating subjects using a random number list meant that it was not possible to predict what treatment would be given to each patient, thus minimising the possibility of bias in the allocation.
1.3 Why are research studies, such as clinical trials, needed?
Smoking is a cause of lung cancer, and statin therapy is effective in treating coronary heart disease. However, why do some people who have smoked 40 cigarettes a day for life not develop lung cancer, while others who have never smoked a single cigarette do? Why do some patients who have had a heart attack and been given statin therapy have a second attack, while others do not. The answer is that people vary. They have different body characteristics (for example, weight, height, blood pressure and blood measurements), different genetic make-up and different lifestyles (for example, diet, exercise, and smoking and alcohol consumption habits). This is all referred to as variability or natural variation. People react to the same exposure or treatment in different ways; what may affect one person may not affect another. When a new intervention is evaluated, it is essential to consider if the observed responses are consistent with this natural variation, or whether there really is a treatment effect. Variability needs to be allowed for in order to judge how much of the difference seen at the end of a trial is due to natural variation (i.e. chance), and how much is due to the action of the new intervention. The more variability there is, the harder it is to see if a new treatment is effective. Detecting and measuring the effect of a new intervention in the setting of natural variation is the principal concern of medical statistics, used to design and analyse research studies.
Before describing the main design features of clinical trials, it is worth considering other types of studies that can assess the effectiveness of an intervention, and their limitations.
1.4 Alternatives to ...