HLTH 207 Week 4 Lecture - Randomized Controlled Trials (ONE)

How is the sample population of a controlled trial selected?

Selected from source population

Requirement for sample population of controlled trial

Sample populations must be individuals who are at-risk of the outcome of interest

General process of controlled trial

• Same population is assigned to an exposure, either an intervention group or control group

• Then, they are followed forward in time to observe who develops the given outcome of interest

2 measures of outcome frequency used in randomized controlled trials

Risk and/or rate

2 measures of association used in randomized controlled trials

Risk and/or rate ratios and differences

4 basic steps in designing and conducting a randomized controlled trial

1. State the research question

2. Design the trial

3. Conduct the trial

4. Analyze and report data

2 aspects of stating the research question

1. State the trial study hypothesis

2. Define the exposure condition (both exposed and unexposed and define outcome(s) with case definitions

Aspect of designing a randomized controlled trial

Determine the randomization strategy, select the type of trial, and decide whether there will be masking

3 aspects of conducting a randomized controlled trial

1. Select subjects from an at-risk source population and obtain informed consent

2. Allocate subjects to exposure groups with concealment

3. Follow subjects and collect data on incident outcome as well as key covariates

3 aspects of analyzing and reporting the data from a randomized controlled trial

1. Select subjects from an at-risk source population and obtain informed consent

2. Allocate subjects to exposure groups with concealment

3. Follow subjects and collect data on incident outcome as well as key covariates

What is the exposure condition comprised of

The arms of the trial (there can be more than 2)

2 common arms of a study

1. Exposed group(s): intervention or treatment group(s)

2. Unexposed group: control

2 types of exposure that may be evaluated in a trial

1. Preventative exposure

2. Therapeutic exposure

Preventative exposure

administered to those who are at-risk of health outcome(s) to assess whether the exposure can reduce the frequency of incident negative health outcomes

3 examples of preventive exposure

1. Vaccines

2. Exercise programs

3. PrEP

Therapeutic exposure

administered to those who have health outcome(s) to assess whether the exposure can improve their health status

3 examples of therapeutic exposure

1. Chemotherapy

2. Beta blockers

3. Antibiotics

Control group

either receives a placebo or standard of care option

Standard of care

standard therapeutic or preventive option, if one exists

Placebo

inert or fake control group (sugar pill, saline injection, etc.)

In a randomized controlled trial, should a placebo or standard of care (if it exists). Why?

• Standard of care

• Participants in a trial should not be denied an existing, effective preventive or therapeutic option

1985 randomized controlled trial that did not utilize standard of care

A 1985 randomized controlled trial of ivermectin to treat onchocerciasis (river blindness) used a placebo control arm even though the drug diethylcarbamazine had been used as a standard of care for several decades

1992 randomized controlled trial that did not utilize standard of care

Another trial in 1992 evaluated the antidepressant drug paroxetine in patients with severe depression, and the control group was given a placebo over other antidepressants

What piece of legislature required that a standard of care be used as a control if it exists?

1996 revisions of Declaration of Helsinki specified that control groups in clinical research should be offered the best standard of care, if one exists

How does using a placebo instead of a standard of care for the control group affect data analysis?

Association will be stronger when compared to the placebo than to the standard of care

Placebo effect

participants report benefit from just receiving an intervention (regardless of whether the intervention itself truly has a clinical benefit)

Example of placebo effect

IBS patients were told they were receiving placebo, but still reported improvement due to their beliefs around mind-body self-healing process

Function of placebo arm

the use of a placebo control arm allows investigators to estimate the effect of an intervention above and beyond any benefits derived from the act of receiving an intervention

3 types of trial hypotheses

1. Superiority trial

2. Non-inferiority trial

3. Equivalence trial

Superiority trial question

Is the intervention better than the control?

Function of superiority trial

If no standard of care exists, or the current standard of care is not very effective, we may want to demonstrate that the new intervention is better than control

What data indicates that the intervention is better than the control?

• The frequency of negative health outcomes in the intervention group is meaningfully lower than the control group

• Measures of association fall below the null

Non inferiority trial question

Is the intervention at least as beneficial as the control?

Function of non-inferiority trial

If a new intervention could conserve resources or was less invasive relative to control, we may want to demonstrate that it is not worse than control

What data indicates that the intervention is at least as beneficial as the control?

• Frequency of negative health outcomes in the intervention group is not meaningfully higher than the control group

• Measures of association fall above the null

Equivalence trial question

Is the intervention similar to the control?

Function of equivalence trial

Compares two or more options (therapeutic or preventive) when some patients, providers, or other stakeholders may prefer one over another

What data indicates that the intervention is similar to the control?

• Frequency of negative health outcomes in the intervention group is similar to that of the control group

• Measures of association are not meaningfully different from the null

How many exposure be assigned in controlled trials? Why?

randomization, which makes it so that no individual is more (or less) likely to be allocated to a given exposure condition than anyone else

Goal of randomization

yield exposed and unexposed groups that are comparable on all factors except for the exposure status

Confounding error

 systemic error when one or more external factors can explain the observed association between an exposure and health outcome

How does randomization correct for confounding error>

On average, randomization creates comparison groups that are balanced for all confounders

2 randomization techniques

1. Simple randomization

2. Stratified randomization

Simple randomization

study participants are randomly assigned to an exposure condition

Strength of simple randomization

relatively straightforward process (analogous to assignment by flipping a fair coin)

Weakness of simple randomization

may still produce imbalances in confounding factors between study arms if the sample size is small

Stratified randomization

study participants are first grouped by important potential confounders, then assigned to exposure conditions within those groups

Strength of stratified randomization

potential confounders that are stratified on are balanced across trial arms

Weakness of stratified randomization

 stratification by other important confounders (those not stratified on) may not be possible if sample sizes are small

2 possible units of randomization

1. Individual randomization

2. Cluster randomization

Individual randomization

each participants is individually randomized to an exposure condition

Cluster randomization

groups of individual are randomized to an exposure condition

When is cluster randomization useful?

Useful when individual level exposure is not possible (i.e. school or household-based interventions)