Stories of Trials That Matter

Rob Califf estimates that each time you do a Google search you are on average taking part in 3-4 randomised trials. E-commerce has made the randomised trial the cornerstone of website optimisation and a recent report in Trials demonstrated how a trial could be used to optimise the sign-ups for organ donor registration.

In the UK, 17 million people each year renew their car tax online and at the end of the process a request is made to join the organ donor register. This trial tested a number of different messages on signups for the organ donor register against the following control webpage:

 

 

 

 

 

 

 

The first message was of a social norms type and tested the message on its own (webpage B) and with an image (webpage C) or with the logo for NHS organ donation (webpage D).

 

 

 

 

 

 

 

 

 

 

 

 

The remaining four messages (E-H) tested simple messages on their own without images or logos:

A	Control	No Message
B	Social norms	Every day thousands of people who see this page decide to register
C	Social norms + image	Every day thousands of people who see this page decide to register + image
D	Social norms + logo	Every day thousands of people who see this page decide to register + logo
E	Loss frame	Three people die every day because there are not enough organ donors
F	Gain frame	You could save or transform up to 9 lives as an organ donor
G	Reciprocity	If you needed an organ transplant would you have one? If so please help others
H	Cognitive dissonance	If you support organ donation please turn your support into action

 

The trial was run over a one month period. A randomisation code sequentially assigned individuals a number from 0 to 7 and displayed one of the 8 corresponding webpages shown in the table above. The main outcome measure was the number of sign-ups to the organ donor register.

During the study period, a total of 1,085,322 observations were made and the unadjusted odds ratios for the effect of the various messages on the number of sign-ups compared to control is shown in the table below:

		Unadjusted odds ratio [95% CI]
B	Social norms	1.255 [1.196-1.317]
C	Social norms + image	0.941 [0.894-0.990]
D	Social norms + logo	1.274 [1.214-1.336]
E	Loss frame	1.333 [1.271-1.397]
F	Gain frame	1.250 [1.192-1.311]
G	Reciprocity	1.380 [1.317-1.447]
H	Cognitive dissonance	1.232 [1.174-1.293]

All of the test messages resulted in statistically significant improvements in the number of sign ups with the exception of the social norms message with an image which resulted in a decrease in the number of sign-ups. The most effective message was the reciprocity message (“if you needed an organ transplant would you have one? If so please help others”) with individuals 1.38 times more likely to sign up to the organ donor register than controls. Following the results of the trial, the reciprocity message has been implemented across 25 government websites.

You can read the full report here.

One of the major barriers to the conduct of more and better randomised trials is the ever increasing cost of trials. Large trials of 10-15,000 participants typically cost in excess of £3-400m and there have been reports of single trials costing in excess of $1bn. These costs are unsustainable and distort the research agenda.

A major component of the cost of such trials is the setting up and running of large numbers of clinics. Clinics are required to screen and recruit participants and to monitor interventions by physical and laboratory means. Some commonly used interventions require no clinic monitoring so opening up the possibility of conducting a trial without the need for clinic visits. The ASCEND trial was conducted in 15,500 people with diabetes and compared using a 2×2 factorial design aspirin versus placebo and omega 3 fish oils versus placebo in reducing cardiovascular events. The trial was entirely mail based without the need for participants to attend clinics.

Potential participants with diabetes were identified from central registers and from GP practices. Letters of invitation included a screening questionnaire which was returned by post. Eligible participants were then entered into a pre-randomisation 2 month run-in phase where all participants received 8 weeks of placebo aspirin and placebo omega 3 fish oils. During the run-in phase participants were mailed a sample collection kit for blood and urine which was taken to their GP practice and returned by mail. A total of 4% of the potential participants invited went on to be randomised.

Follow-up takes place every 6 months with questionnaires being returned either by mail or online and by the time of the report of the main result the length of average follow-up was 7.4 years. The main results for aspirin and omega 3 fish oils can be found here and here.

The total cost of ASCEND is less than £10m, which makes it an order of magnitude less than similar sized clinic-based studies.

Percutaneous coronary intervention (PCI) was performed more than 500,000 times annually worldwide for the treatment of stable angina. Evidence for its effectiveness comes from unblinded randomised trials which have shown improvements in exercise duration, angina relief and quality of life from PCI. However, these symptomatic improvements are subjective and at least some of the measured benefit might be due to the placebo effect. In all previous trials investigators and patients were aware of the treatment allocation.

The ORBITA trial was a landmark trial undertaken at 5 sites in the UK to test in a double-blind randomised design whether PCI improved exercise duration compared to placebo in patients with stable angina. Eligible patients had severe single-vessel stenosis. The design included a 6-week medical optimisation period during which anti-anginal medication was actively managed and after which a baseline exercise test was undertaken. Randomisation was in a 1:1 ratio to either PCI or placebo, which included a sham procedure whereby a coronary catheter was inserted but no stents were used to remove stenosis. The patient and investigators undertaking the assessments of outcomes were blinded to the treatment allocation. A total of 200 patients (105 PCI vs. 95 placebo) underwent randomisation. The pre-specified primary outcome measure was exercise duration and there was no difference between the groups (PCI minus placebo 16·6 s, 95% CI –8·9 to 42·0, p=0·200). This result contrasts with the results of previous unblinded trials which showed a significant improvement in exercise duration with PCI. There was also no change in secondary endpoints related to other exercise endpoints and quality of life.

The authors conclude that whilst PCI might not offer therapeutic benefit beyond medical therapy, the medical therapy in the trial was intensive so PCI offers an alternative treatment option to long-term drug therapy in stable angina.

You can read the full report here

Last time we covered the INFANT trial which demonstrated no apparent benefit for “high-tech” decision support software used during childbirth. This week we look at the other end of the technology spectrum during childbirth by way of the BUMPES trial to answer the question does lying down or sitting upright during the second stage of labour with a low-dose epidural make a difference to the number of women having a normal vaginal birth?

[Read more…]

When we think of large simple randomised trials we might think of maybe studies done in cardiovascular disease or cancer and testing one or more drug treatments. But the need for reliable evidence from large randomised trials extends across all areas of medicine and trialists have applied themselves in many areas to develop this evidence. One such area is pregnancy and childbirth.

[Read more…]