8 Clinical trials and research

8.7 Randomised, double-blind, placebo-controlled trials

---

The most reliable evidence - often referred to as the 'gold standard' - is a 'prospective randomised, double-blind, placebo-controlled study.

Randomisation

Randomising patients in a study is the best proven way to allow for the fact that some things in a trial - and in life - can happen by chance.

Patients in a study are often randomised when two or more groups are studied.

Randomisation is designed to balance factors in each group that could affect the study results. This includes known factors, such as sex, smoking status or social differences, and unknown factors such as genetic differences that we may not know anything about.

Randomising people, if done correctly, and especially with larger groups, should normally result in an approximate balance of all these factors.

This is a very difficult concept, but it is one of the most important things to understand.

It also stops bias. For example, it prevents a doctor putting patients who are most ill and in need of treatment into the group that receives an active drug rather than a placebo (dummy pill). If this happened, although this may sound more 'fair', the two groups would be different at the start, so you couldn't compare the results accurately at the end.

Clinical research, by definition, involves different people getting different treatment. Often the people to get first access to a treatment in a trial, do not get the best results compared to people after the drug is approved. The balance though, is that by getting earlier access to treatment, they can benefit from other later research.

Randomisation has to be done in a way that doesn't select a certain group over another.

The most common example for randomising a patient to one of two groups is to toss a coin for each patient - heads they join one group and tails they join the other.

This is because tossing a coin is random and can't be predicted.

Over time, the more a coin is tossed, the more likely that approximately 50% will be heads and 50% will be tails.

An example of bad randomisation would be assigning patients who come to clinic on a Monday to one group and patients who come on a Tuesday to another. In this example, people who come on Mondays may be different from people who come on a Tuesday, for social reasons. They may be more organised people, or less likely to have a hang over from the weekend. This could represent important differences between the two groups - ie alcohol use - and this could affect the study results.

When a study reports its results they also report the characteristics of people being studied. Sometimes, even with randomisation, you may see that one group may have different characteristics. This is sometimes adjusted for in the final analysis, or needs to be considered when interpreting the study results.

Blind and double-blind studies

Blinding (sometimes called 'masking') is the term to describe a doctor, patient or researcher not knowing which study group a patient has been assigned to.

A blinded study is where the patient doesn't know which group they are in, or which treatment they are getting.

A double-blinded study is where neither the doctor nor the patient know which group the patient is in.

Blinding prevents different care or treatment being given based on what either the doctor or patient believes.

An example of why blinding is important is that if someone know they are getting an active drug, both doctors and patients may be more likely to report side effects.

It could also affect how likely a patient is adherent to treatment.

Placebo

A placebo is the term for a dummy pill, ie a pill that looks, smells and tastes like the drug that is being studied, but which has no active ingredient.

Using a placebo helps find out whether the active drug is really active. It also helps interpret side effects.

If 10% of people in the active drug group report having a headache and 2% of people in the placebo group report a headache, then it is reasonable to think that the active drug can cause headaches.

If 10% of the placebo group also reported a headache, then it is reasonable to think that the active drug doesn't cause a headache.

An example of why placebo trails are still important was shown in the development of an NNRTI called capravirine. In an early (Phase IIb) study people using capravirie plus a background regimen did no better than people using the same regimen plus a placebo. This stopped further development of the study drug without putting any further patients at risk from using an ineffective treatment in later trials.

Control group

A control group refers to a group of patients in a study, that any intervention is compared to. This helps to show that the intervention actually caused what was seen and that it wouldn't have happened anyway.

One common type of control group is to use a placebo.

Although all patients get the best treatment with or without the new drug, if, for example, this is a new HIV drug and the best treatment already includes 3 active drugs, then it could be difficult to see any difference between the new drug and the placebo - because both groups will already do very well.

Another type of control group is a group where no intervention takes place.

This example might be used where there is something about the trial drug that makes using a placebo difficult - perhaps because it is given by injection. The difficulty of not randomising the control group to having a placebo is that you can never be sure whether some of the things (both good and bad) that happened to patients in the active drug arm, are not due to chance. More importantly, people in each arm may behave differently because they know they are getting active drug, for example, by reporting more side effects.

Another type of control group is to use a drug or combination that has already been studied.

This is still generally the type of trial design used for studying a new HIV drug in people who are treatment-naïve. This is generally ok, so long as the new drug turns out to be better than, or at least as good as, the current standard of care (ie tenofovir+FTC+efavirenz).

For this reason, early trials with this design should not enrol people who have advanced HIV (for example with CD4 counts less than 100 cells/mm³) as these people will need to depend on a proven treatment.

Randomising patients should mean that important factors - known and unknown - are likely to be distributed between each group. For example, having the similar numbers of women, Caucasians, smokers, CD4 counts etc.