Evidence

Unfortunately, there are lots of examples of doctors and researchers assuming that their treatments worked - before research found they didn’t work, or were even more harmful than thought . 
 
This is why we need fair tests of treatment. The ‘gold standard’ is usually the RCT Randomised Controlled Trial. Two similar groups are divided and half given the treatment under test, and the other group is observed or given a placebo treatment. 
 
Placebos look and seem like the ‘real’ treatment - but they aren’t. They used to know what would have happened without the ‘real’ treatment. This means that you know how good the ‘real’ treatment is - because you know what would have happened without it. Randomised controlled trials are also used to compare treatments to find which is the best. 

For example: 

• Blood letting

  Removing pints of blood - was offered by doctors and barber-surgeons for just about anything - fevers, strokes and infections. George Washington was bled four times in a few hours to treat a sore throat shortly before he died in 1799. It took many more years before several doctors examined the evidence on whether blood letting actually helped treat ill people - and concluded that it didn't. 1

  This type of treatment is now only used for people with haemachromatosis - a disorder where the body stores too much iron.

  There also more modern examples.

  Footnote:

  1. DP Thomas, The demise of bloodletting, Royal College of Physicians of Edinburgh

• Flecainide

  After a heart attack, people occasionally get an abnormal heart rhythm that can sometimes be deadly. Doctors thought that routinely giving patients a drug after a heart attack would stop this from happening. But in 1991 research showed that the drug, flecainide, actually caused more deaths than a dummy placebo treatment.1

  Footnote:

  1. Mortality and Morbidity in Patients Receiving Encainide, Flecainide, or Placebo - The Cardiac Arrhythmia Suppression Trial The New England Journal of Medicine

• Knee washouts

  Knee 'washouts' for wear and tear arthritis, along with removal of cartilage, were a standard treatment for thousands of people. It was thought that washing out the joint and small particles of cartilage that had worn off would make the knee less painful. But research in 2013 showed that washouts were no better than sham, or placebo surgery.1

  Footnote: 

  1. A Controlled Trial of Arthroscopic Surgery for Osteoarthritis of the Knee, The New England Journal of Medicine

• Baby simulators

  In Australia they tried to reduce teenage pregnancy through 'baby simulators'. They thought that if teenagers knew more about what having a pretend baby was like, they would try harder to avoid pregnancy. But research in 2016 showed that teenagers who used the baby stimulator were actually more likely to become pregnant.1

  Better research has meant that less harmful and better options can be offered to people - for example, sex education works better than baby simulators to prevent teenage pregnancies, and physiotherapy is often a better option for people with knee arthritis.

  Footnote:

  1. Efficacy of infant simulator programmes to prevent teenage pregnancy: a school-based cluster randomised controlled trial in Western Australia - The Lancet

Better research has meant that less harmful and better options can be offered to people - for example, sex education works better to prevent teenage pregnancies, and physiotherapy is often a better option for people with knee arthritis.

This is why we need fair tests of treatment. All these ideas sounded good but turned out not to be. The Testing Treatments website explains what fair tests look like and the problems that happen when we don’t use them. We also need to think about what tests of treatments measure. Usually, these are called the ‘outcomes’ of the trial. Different trials might have outcomes like blood pressure or death rates, or quality of life. 
 
In the flecainide example, previous research had shown that the drug did seem to reduce some abnormal heart beats (1). Doctors gave the drug presuming that because it gave people regular heart beats then people would also not die of abnormal rhythms and therefore live longer. But that turned out to be an unsafe presumption. 

The most important 'outcome' should have been death after heart attacks. But the outcome measured in the trial was abnormal heartbeats. The researchers assumed that reducing abnormal heartbeats would reduce deaths, but it didn’t - it increased them. 

The abnormal heartbeats were treated as a ‘proxy marker’ for death rates.

• Proxy Markers

  Proxy markers are common because it is often cheaper and easier to do studies measuring a proxy (heart rhythms) rather than what we are really interested in (deaths). For example, many studies looking at drugs for diabetes look at a marker in the blood, Hb1ac. The higher the Hb1ac, the higher the sugar level. But we need to make sure that the drugs really do help people with diabetes stay well, rather than just lower the Hb1ac level. A few years ago GPs were encouraged to aim for very low levels of Hb1ac in people with type 2 Diabetes because it was thought this would lead to fewer problems with type 2 diabetes. In fact, research found that it didn’t reduce complications - but did give more people low blood sugar, which is very unpleasant and sometimes dangerous. *

  Many drinks and yogurts are advertised to help heart health through lowering cholesterol. These products have been shown to reduce cholesterol. But we it’s not our cholesterol levels which really matter - it’s whether we have a heart attack or stroke. 
   
  So it’s not just about whether a treatment is proven to work - but whether it’s proven to work for something that really matters to us. 

We need to make sure we measure what is important to patients 

Reference:

1. Follow-up of Blood-Pressure Lowering and Glucose Control in Type 2 Diabetes, The New England Journal of Medicine, October 2014

Trying to work out whether a treatment is useful can be complicated! In reality, there are usually many clinical trials examining the same area - for example, there are thousands of trials on using statin drugs to lower cholesterol. If we are trying to decide on whether a treatment works, and we only look at one trial, we risk choosing one that doesn’t give the full picture. 
 
There are other problems, too. For years it’s been known that some trials, particularly trials that show treatments don’t work, are less likely to be published in research journals. This means that professionals and patients can’t get to see them. Systematic reviews try and combat these problems. This means searching for all the published and unpublished information about a subject to give more reliable information.

Cochrane publishes hundreds of systematic reviews, asking questions from ‘do health checks work?’ to whether nose sprays work for allergies. They also provide useful lay summaries written for people who are not healthcare professionals (1).

Reference:

1.  Cochrane Library

We need to make sure what we measure is important to people