Module 5 The Epidemiological Basis of Randomised Controlled Trials Landon Myer School of Public Health & Family Medicine, University of Cape Town Introduction The Randomised Controlled Trial (RCT) is the most rigorous experimental design in clinical research. In this design, a researcher randomly allocates eligible people to receive (intervention group) or not to receive (control group) one or more interventions that are being compared. The control group receives either the standard treatment or a placebo (an inactive formulation which looks identical to the intervention treatment) if no standard treatment exists. The results of an RCT are assessed by comparing the occurrence of the outcome of interest between the intervention and control groups. This presentation summarises the key principles of design, conduct and analysis of RCTs for a GCP audience. Basic design of a randomised controlled trial An RCT is a prospective study designed to establish a causal relationship between an intervention and an outcome. In the case of clinical research, the intervention is usually a medical treatment (eg, a new drug, management approach or surgical technique). The outcome may be any health-related event, usually a specific form of morbidity or mortality. Participants must meet specified criteria before being considered eligible for participation in the study. In an RCT, the researcher begins with individuals who are free of the outcome of interest, but who are at risk of developing that outcome. For example, in a hypothetical trial of a new drug to prevent diabetic nephropathy (the outcome of interest), researchers would start with a population of individuals with diabetes (and thus are at risk of developing the outcome of interest) who had not yet developed diabetic nephropathy. Figure 1 shows a basic schema for the design of a RCT. Eligible individuals are randomized to an intervention or a control group. After they are allocated to receive the intervention or control, participants are followed-up through time and the occurrence of the outcome in each of the groups is compared. The purpose of the RCT is to see if the incidence of the outcome differs between those randomized to the intervention and those randomized to the control group. RCTs also allow researchers to document adverse events associated with the intervention.
Figure 1
Key features of a randomised controlled trial Randomisation means that each participant is randomly allocated to one of the two groups and has an equal chance of being allocated to either of the groups. The purpose of randomizing participants is to ensure that two groups being compared are equivalent with respect to baseline characteristics, including known risk factors for the disease as well as unknown risk factors and those not measured in the study. Randomization prevents selection bias (that is, it prevents participants, clinicians or researchers from using their subjective judgment to decide into which group to place participants). It also assists with the control of confounding by ensuring that (within the limits of chance) the two groups are comparable at
the beginning of the study. Thus, assuming the randomization works, any differences in disease outcomes at the end of the study can be attributed to the intervention being tested. The best practical method for randomisation is by a computer-generated random allocation sequence. Methods for ensuring that the random allocation sequence is concealed from researchers, clinicians and participants until treatment has been assigned ( allocation concealment ), such as using opaque envelopes, are important. If the random allocation sequence is known before treatment assignment then, for example, clinicians could manipulate the order of participant enrollment to place their patients into the groups of their preference. This would be a form of selection bias, and would defeat the purpose of randomization. In addition, participants and researchers should be blinded during a randomised controlled trial. Blinding means that the participants and study staff do not know to which trial group an individual has been assigned. Blinding reduces the likelihood of bias due to differences in perceived response to treatment (sometimes called performance or reporting bias) on the part of participants. It also prevents bias in healthcare provider behaviour and provider/researcher outcome assessment (sometimes called assessment, diagnostic or detection bias). Blinding can also prevent data analysts from knowing trial group assignments until the last possible moment. In single-blind trials only the participant is blinded; in double-blind trials both the participant and the clinicians/researchers are blinded; and in triple-blind trials, the participants, clinicians, researchers and data analysts (and/or other groups) may be blinded. Blinding is not possible in all RCTs, depending on what the intervention and control conditions are. Blinding is most important when subjective outcomes such as reported adherence, quality of life and quality of care are being measured. In therapeutic research, blinding often involves using a placebo drug in the control arm that appears identical to the active drug being tested in the intervention arm. Placebos are an important tool to facilitate the blinding of study participants. Analysis of a randomised controlled trial The analysis of a randomized controlled trial can be quite simple. If randomization succeeded, the possibility of confounding (and in turn, the need for statistical adjustment) should be minimized. As a result, the primary analysis of most RCTs is a simple comparison of the occurrence (incidence, measured as either a risk or a rate) of the outcome of interest between the intervention and control groups. This comparison is most often expressed as a risk ratio or a rate ratio (sometimes also known as a relative risk : the rate of the outcome in the intervention group divided by the rate of the outcome in the control group). Sometimes, this comparison is expressed as a risk difference or a rate difference: the rate of the outcome in the intervention group subtracted from the rate of the outcome in the control group). Another measure often calculated from an RCT is the number needed to treat (NNT). NNT represents the number of people who need to receive the intervention in order to prevent one case of the outcome or disease in the intervention group. NNT is important when making decisions about whether or not to implement an intervention that has been shown to be effective in an RCT on a wider scale. One important principle in the analysis of randomized controlled trials is that the analysis should include all individuals as they were randomized (to intervention or control groups). This is called the intention to treat principle ~ that individuals should be analysed as they
were randomized, regardless of whether or not they actually adhered to the intervention (or control) during the trial. While not all analyses in an RCT need to adhere to the intention to treat principal, the principal analysis should do so, as it is an important measure to prevent biased study results. Example of a hypothetical randomised controlled trial A hypothetical RCT was conducted in Cape Town, South Africa, to compare a successful a new drug (drug A) to the existing standard medication (drug B) in the prevention of repeated myocardial infarcations (MI) among individuals with unstable angina. (Table 1). Patients with unstable angina were randomly allocated to either drug A (n=105) or drug B (n=111) and followed for two years to measure the proportion of patients who experienced an MI in the two groups. Sixty percent of those receiving drug A and and 54% of those receiving drug B patients experienced an MI. Table 1. Randomized trial calculations Participants followed through time to see if outcome develops Developed MI Participants Randomized to Did not develop MI Totals either intervention or Drug A 63 42 105 63 / 105 control Drug B 60 51 111 60 / 111 Incidence of the outcome calculated in each group and compared In this example, patients who received drug A were 1.1 times more likely to develop an MI within two years compared to patients who received drug B. In other words, for practical purposes, drug A and drug B were equivalent in the prevention of MI among patients with unstable angina.
Strengths of randomised controlled trials Of all study designs, RCTs can provide the strongest information about the effectiveness of medical interventions. An RCT has the potential to prevent confounding and selection biases that are common to observational studies. An RCT can thus isolate the effect of a single intervention on a particular disease outcome. Because the intervention is controlled by the researcher, and the study started with participants who did not have the outcome of interest, the researcher can be assured that the intervention preceded the outcome. This is important for establishing a causal relationship between the exposure and the outcome with confidence. Blinding can be used to reduce reporting and assessment/detection biases associated with participant and researcher knowledge of the study group to which a participant has been allocated. This includes the use of placebos, where possible. Limitations of randomised controlled trials For ethical reasons, RCTs cannot be used to test hypotheses about whether certain factors cause disease or about the effect of harmful exposures (although they can test the effect of removing potentially harmful exposures, for example, interventions to promote smoking cessation). Even for beneficial interventions, RCTs raise many ethical problems, which have to be dealt with before, during and after the trial. RCTs are vulnerable to biases from loss to follow-up, particularly if the period of follow-up is long. As a result, efforts must be made to minimize the chance of loss of participants through time. RCTs tend to be complex, time-consuming and expensive. Conclusion: Understanding the design of clinical trials is crucial. All investigators and site co-ordinators are charged with the responsibility of explaining these very complicated scientific terms to study participants in lay language. It is therefore important that all members of the research team are familiar with clinical trial design, randomisation, blinding and placebos, at minimum.