RCT Methodology Notes

Intro to RCT Methodology

Intended Learning Outcomes

By the end of this lecture, students should be able to:

• Define the terms: Random allocation, Control group, Blinding.

• List reasons for the use of random allocation.

• Describe situations where carrying out an RCT is useful.

Types of Research Study Design

Research study designs can be broadly categorized into:

• Observational Studies:

  • Ecological

  • Cross-sectional

  • Case-control

  • Cohort

  • Systematic review of observational studies

• Intervention (Experimental) Studies:

  • Non-randomised intervention study

  • Randomised controlled trial (RCT)

  • Systematic review of RCTs

Randomised Controlled Trials (RCTs)

• RCTs are experimental studies that involve formal testing of intervention(s).

• They are commonly used for evaluating the effectiveness of medical interventions, such as drug therapies, surgical procedures, and public health initiatives.

Why RCTs?

• RCTs are useful when there is reasonable expectation that the experimental treatment may carry a benefit, but it is not clear that it will do so. This uncertainty might stem from preclinical studies or early-phase clinical trials suggesting potential efficacy; however, definitive evidence is lacking.

• A trial is necessary to establish whether the treatment really does carry a benefit before resources are committed to its use. This prevents the widespread adoption of ineffective or harmful treatments, ensuring that healthcare resources are allocated efficiently and ethically.

How RCTs Work

• Individuals are randomly allocated to either receive one (or more) interventions or not. Random allocation aims to create comparable groups at baseline.

• The results in those who receive the interventions (intervention group) are compared to those who do not (control group). This comparison allows researchers to isolate the specific effect of the intervention.

RCTs as the 'Gold Standard'

• RCTs have become the 'gold standard' for evaluating new therapies due to their ability to minimize bias and establish causality.

• They use randomisation to assign patients to treatment, ensuring that each participant has an equal chance of being in any group.

• They use a control group as a comparison group, providing a baseline against which the intervention's effects can be measured.

• Differences observed are ascribed to the treatment, assuming randomisation has successfully balanced all known and unknown confounding factors.

Features of Randomised Controlled Trials

• RCTs are a type of prospective study with many features of cohort studies. Data is collected forward in time.

• The major advantage of the RCT is that the intervention group has similar characteristics to the control group at the start of the trial because of randomisation. This minimizes the influence of confounding variables.

• Any difference in outcome between the intervention and control group should be due to the intervention, rather than some other factor (no confounding). This allows researchers to draw more confident conclusions about the treatment’s effect.

• Feasibility and ethical issues may prevent an RCT from being done. For instance, it may be unethical to withhold treatment from a control group if an established effective treatment exists. Practical constraints, such as the need for large sample sizes or long follow-up periods, can also hinder the implementation of RCTs.

Trial Population

The general trial process involves:

1. Identifying a population. This involves defining the target group, such as patients with a specific disease or condition.

2. Determining eligibility for the trial. Clear inclusion and exclusion criteria are established to ensure the sample is appropriate for the intervention being tested.

3. Recruiting a sample from the eligible population. Recruitment strategies must be carefully planned to reach a representative sample while adhering to ethical standards.

4. Randomly assigning participants to either the experimental intervention group or the control group. The randomisation process must be robust to prevent selection bias.

Participants in both groups are followed up, and outcomes are compared to assess the effect of the intervention.

Control Groups

• Often, a placebo is used for comparison in the control group. A placebo is identical to the treatment under evaluation, except it has no active ingredient. This helps to account for the placebo effect, where participants experience a benefit from the treatment simply because they believe it to be effective.

• However, if a standard treatment is available, this should be used as a control (intervention vs. standard of care). Using the standard of care as a control provides a more realistic and ethical comparison, reflecting real-world clinical practice.

• It is not ethical to use a placebo in many circumstances where an effective treatment already exists. Withholding effective treatment can be harmful and is generally considered unethical.

Examples of RCTs

Several examples were mentioned for trials, including:

• Drug vs. Placebo: Evaluating the efficacy of a new drug compared to an inactive substance.

• New drug vs. Old drug: Comparing the effectiveness of a new drug to an existing standard treatment.

• Physiotherapy vs. Surgery: Assessing the outcomes of physiotherapy versus surgical intervention for a specific condition.

Randomisation

• Allocation is purely random. Each participant has an equal chance of being assigned to any of the treatment groups.

• Each patient has the same probability of assignment to either group. This ensures that, on average, the groups are comparable.

Why Randomisation?

• Eliminates selection bias in treatment assignment. Prevents researchers from consciously or unconsciously influencing group assignment.

• Creates comparable groups, balancing demographic and prognostic factors. Randomisation aims to balance both known and unknown factors that could affect the outcome.

The Randomisation Process

Patient information is entered into a computer, and the computer randomly assigns patients to two or more groups. The control group receives standard treatment, while the trial group receives the new treatment. This process helps to prevent bias.

Simple Randomisation

• Uses random numbers (computer-generated or from random number tables). This is the simplest method of randomisation.

• A list is generated, and random numbers are assigned (e.g., 1 to 10).

• For example: 5, 3, 7, 4, 9, 2, 3, 1, 1, 4, 8, 9

• Assign odd/even to treatment groups, e.g., AAABABAAABBA (8 active, 4 placebo).

• In small studies, there is a possible chance of imbalance. Simple randomisation may result in unequal group sizes in smaller trials.

• In large studies, groups are roughly equal. As the sample size increases, simple randomisation tends to produce more balanced groups.

Block Randomisation

• Used in smaller studies to create balanced groups. Block randomisation ensures that the number of participants in each group is balanced at certain intervals.

• The block size is determined by the researcher and should be a multiplier of the number of groups (e.g., 2 treatment groups, block size of 4). The block size should be chosen carefully to minimize predictability while ensuring balance.

• Example block size of 4: ABAB, AABB, BABA, ABBA, BAAB, BBAA

• Random numbers: 5, 3, 7, 4, 9, 2, 3, 1, 1, 4, 8, 9 (8 active, 4 placebo)

• Group Assignment: BAAB BABA ABBA (6 active, 6 placebo)

Concealment of Allocation

• Personnel recruit patients into the trial, checking inclusion/exclusion criteria and obtaining consent, then randomise patients into groups. The randomisation process should be concealed from those recruiting participants.

• If there is no concealment, it may influence the decision to recruit, leading to imbalances and selection bias. Lack of concealment can lead to selection bias, where recruiters preferentially enroll certain types of participants into specific groups.

• Concealment is a technique used to prevent researchers from influencing which participants are assigned to a given group, either consciously or unconsciously. This ensures that the randomisation process is truly random.

• With concealment, the decision cannot be influenced. This can be achieved using opaque, sequentially numbered, sealed envelopes or a central randomisation centre. Central randomisation is often preferred as it provides greater security and minimizes the risk of manipulation.

Blinding

• Blinding involves preventing patients, doctors, nurses, and other personnel from knowing which treatment group individual patients are in (e.g., use of placebo). Blinding minimizes the risk of performance bias and ascertainment bias.

Types of Blinding

• Single-blind: Patients are unaware of the test being conducted, but the doctor is aware. Single-blinding minimizes patient-related biases.

• Double-blind: Both patients and doctors are unaware of the test being conducted. Double-blinding minimizes both patient and researcher-related biases.

Blinding prevents biased reporting of outcomes (particularly if the outcome is subjective) and prevents doctors from treating patients differentially. This ensures that any observed differences are due to the intervention and not influenced by expectations or behaviors.

When Blinding Isn't an Option

Blinding is not always possible, particularly in trials involving physical interventions like surgery vs. physiotherapy. In such cases, efforts should be made to blind the outcome assessors to minimize bias.

Importance of Concealment and Blinding

• Concealment of allocation is important during recruitment. This prevents selection bias.

• Blinding is important during the intervention and follow-up phases. This prevents performance and ascertainment bias.

Streptomycin for TB: A Landmark RCT

• One of the first examples of a properly conducted RCT. This trial set a new standard for clinical research.

• Evaluated streptomycin for pulmonary TB. Streptomycin was a novel antibiotic at the time.

• Published in the BMJ in 1948. The publication marked a significant milestone in the adoption of RCTs.

• One of the first trials to use randomisation. It demonstrated the feasibility and value of randomisation in medical research.

Trial Design

• Participants: Patients aged 15 to 30 with acute bilateral pulmonary TB. The inclusion criteria were clearly defined.

• Intervention Group: Received streptomycin. The dosage and duration of treatment were specified.

• Control Group: Received bed rest. Bed rest was the standard of care at the time.

• Outcome: Death. Mortality was a clear and objective outcome measure.

Results

The results of the trial showed that streptomycin was helpful. During the first six months:

• There were four deaths among 55 patients who had been allocated streptomycin, compared with 15 among 52 patients allocated to bed rest alone. This represented a significant reduction in mortality.

• This difference was reflected in radiological and other improvements. Patients in the streptomycin group showed greater improvements in lung function and overall health.

During the subsequent six months, the radiological and mortality differences were less marked. There were eight more deaths in the streptomycin group and nine more in the groups treated with bed rest alone.

Conducting the Study

1. Identify the population of interest, defining inclusion and exclusion criteria. This ensures that the study sample is well-defined and appropriate.

2. Assign individuals to the control or treatment group. Random assignment is crucial for minimizing bias.

3. Follow up individuals for the outcome(s) of interest. Complete follow-up is essential to avoid attrition bias.

4. Compare outcomes between groups. Statistical analysis is used to determine if there are significant differences between the groups.

Example: Streptomycin Trial

• Population:

  • Patients aged 15-30

  • Acute bilateral pulmonary TB

• Assignment:

  • Intervention: streptomycin

  • Control: bed rest

• Follow up:

  • All-cause mortality

• Comparison:

  • Compare mortality rates between groups

Follow-up of Individuals

It's important to ensure that all individuals are accounted for at the end of the trial.

• Were any participants lost to follow-up? High rates of loss to follow-up can introduce bias.

• Are reasons for any losses explained? Understanding the reasons for loss to follow-up is important for assessing potential bias.

• Are losses related to treatment/outcome? Differential loss to follow-up (where losses are related to treatment or outcome) can seriously compromise the validity of the trial.

Intention-to-Treat Analysis

• Patients may stop treatment.

• Patients may switch treatments.

• Patients may take combinations of treatments.

Intention-to-Treat (ITT) Analysis

Analyze patients in their original randomisation group. This maintains the balance produced by randomisation.

Summary

• RCT: experimental study.

• Compare outcomes in patients: intervention group vs. control group.

• Randomly assign patients to treatment groups.

• Conceal the randomisation list from researchers.

• Keep study personnel and patients blind, if possible.

• Complete follow-up of all patients entered into the trial.

• Analyze by intention to treat (ITT).

Objectives (Revisited)

• Define the terms:

  • Random allocation: Assignment to treatment groups is at random.

  • Control group: Comparison group receiving standard treatment or placebo.

  • Blinding: Patients/doctors/staff do not know which treatments patients are getting.

• List reasons for the use of random allocation:

  • Eliminates selection bias in treatment assignment.

  • Creates comparable groups.

• Describe situations where carrying out an RCT is useful:

  • Genuine uncertainty about whether a treatment is beneficial.

  • Ethically and practically feasible.

LGW: 4S Study

• Scandinavian Simvastatin Survival Study (4S).

• Simvastatin (cholesterol-lowering drug).

• Previous controversy surrounding effects of drug therapy for hypercholesterolaemia on survival (lack of evidence).

• 4S study designed to evaluate the efficacy of simvastatin.

• Is simvastatin (compared to placebo) effective in reducing mortality/morbidity?