Randomised, placebo-controlled, double-blind trials

The gold standard in medical testing.

by John Jackson © 2005

Claims are often put forward by proponents of alternative remedies and therapies stating that they have tested their treatment and it has passed the test: proof that it works. opponents, however, insist that unless a treatment has passed a double-blind, placebo-controlled study, it is not valid.

Why is this? What is so special about the double-blind, placebo-controlled test? To find the answer, we need to look at the different ways that treatments can be tested and what the pitfalls are.

Open testing

In historical terms this was the earliest method used. The idea is to simply test a remedy on patients and see if they get better. A good example is that of using vitamin C (in the form of oranges and lemons) to treat scurvy. Many treatments had been tried, but they all failed until fresh fruit was used (see: James Lind's treatise of the scurvy). However, this is a case of what is called a high effect size treatment, which means that the effect is so large that it is outside the range of normal variation and simple observation confirms the treatment's effectiveness.

Unfortunately this is not the case for most treatments. Many treatments have a small effect and the subjective views of the researcher can overlap with the subjective views of the patient. This can lead to error.

Confounding factors

Confounding factors are variables which can lead to false conclusions in studies by showing associations where there are none and vice versa. Confounding factors include:

  • Placebo effects

    This is where a perceived improvement is felt by the patient due to expectations and beliefs. (see: placebo effect). The patient cannot tell the difference between the placebo effect and improvement due to treatment: the improvement feels the same.

  • Reinterpretation.

    Having been given a treatment, a patient will often report a reduction in symptoms or an improvement in their condition. Someone given a treatment to reduce itching, for example, may report that their itching has indeed reduced; yet if they are monitored, it can be shown that they are scratching just as frequently as before. Reinterpretation is a form of placebo effect.

  • Observer bias

    This is similar to reinterpretation except it is the doctor/researcher who commits the fallacy. If the researcher believes the treatment to be beneficial, they may well interpret results to match their expectations. Although a researcher should be impartial, this does indeed happen.

  • Self-limiting illness

    Many illnesses are fought off naturally by the body. A useless treatment given during the course of an illness could mistakenly be attributed as curing a disease that would have healed naturally.

  • The regressive fallacy

    This is similar to self-limiting illnesses, only it applies to chronic conditions. A treatment given whilst a condition is at its worst, a flare up, will see an improvement whether the treatment works or not as the condition, being at its worst, can only get better. (see: The Regressive Fallacy).

  • The Hawthorne effect

    This effect was first noticed in the workplace (see: The Hawthorne Effect). It is where staff will perform better if some attention is paid to them. Just taking an interest can improve their productivity for example.

    This effect shows up in medical trials too. A patient who has been chosen to take part in a trial and is monitored will behave better than they would normally. e.g. a diabetic who is given a new pill to try may eat better and take more exercise simply because they are on the trial; any improvement shown could be due to the altered behaviour rather than the treatment.

Confounding factors were not all discovered at the same time, but as they were realised, they had to be eliminated.

The open trial

In these trials both the researcher and the patient know the details of the test. These trials are open to all of the biases and are not shielded from placebo effects. Open trials are still used today; however, they are restricted to procedures where there is no alternative, such as surgery.

The placebo comparison

As placebo effects can make treatments appear successful, it was realised that the treatment under investigation would have to be compared to something known not to work: the placebo control. A treatment cannot be considered effective unless it is compared to a placebo (or at least something else that has previously been tested against placebo or is a proven high-effect sized treatment).

A placebo control, however, is of little value if the patient knows that they are taking a placebo.

The single-blind trial

Researchers adopted the method of blinding their subjects as to whether they were receiving the treatment under review or the placebo control. The idea was to remove any bias the patient may have regarding their treatment. If the treatment was effective, the benefit should show up clearly in the treated group compared to the control group.

It was also realised that not only could patients have a bias; so could the researchers. Observer bias could lead researchers to the results that they were expecting to find.

There was also the problem that the researchers, through their unintentional bias, could influence patients by unconscious signaling, the Clever Hans effect, or by simply treating those they knew had received the treatment in a different manner to those who had not.

The double-blind trial

In the double-blind trial neither the patient nor the researchers administering the treatment know who is being treated and who is in the placebo control group. Patients are randomly assigned to one of the groups independently of the researchers, and the actual treatment they receive is coded so that no one knows whether it is real or placebo. This ensures that all possible biases that could be introduced to the trial by the researchers or the patients are eliminated.

Only after the trial, and once all data have been gathered are the results decoded and analysed. If the treated group shows significantly better results than the placebo control group, then the treatment has passed the trial.

This is not the end, however. One trial is not enough as it could be an anomaly, or a poorly designed or controlled trial, so the test needs to be repeated again independently. If the results are repeated in subsequent high-quality trials, the treatment is deemed to be effective.

Conclusion

As with all tests, it is important to ensure that the results can only be produced by that which is under consideration: all potential biases and complications need to be eliminated or reduced to a minimum.

The randomised, placebo-controlled, double-blind trial is the way in which this is achieved in modern, evidence-based medicine.