Trial design and setting
The Reading Trial was an online, randomised trial with two groups. It took place on a purpose-built website (www.thepeoplestrial.ie), with no face-to-face interaction between people taking part or researchers. The Reading Trial was pragmatic, in that the intervention (reading a book in bed before going to sleep) and the comparator (not reading a book in bed) happened in ‘real-life’ conditions, for example in peoples’ own homes.
Participants – who took part?
People who took part in The Reading Trial (the participants) were 18 years of age or over. As we couldn’t offer a translation service, people who took part in the study also needed to be able to read about the trial and report their experiences in English.
Participants – how we asked people to take part?
We invited people to take part in The People’s Trial and The Reading Trial using social media campaigns on Facebook, Twitter, Instagram, and YouTube. We asked people to go to The People’s Trial website, where an animated video described the aim of The People’s Trial, and the steps involved in taking part in a randomised trial. We also included a Participant Information Leaflet on the website, which people could read and download. This gave detailed information about why we were doing the trial, what people could expect to happen if they took part in the trial, and some potential risks and benefits to taking part in the trial.
After reading the Participant Information Leaflet, people who wanted to take part in The Reading Trial showed their agreement (consent) using an online form. Altogether, 991 people agreed to take part in The Reading Trial.
People taking part in the study gave us some information about themselves. This included their age group, gender, email address (so we could contact them with trial-related information), whether or not they worked in healthcare or health research, and what they felt their level of understanding of randomised trials was. People also told us about how well or poorly they felt they slept (sleep quality) in the seven nights before taking part in The Reading Trial.
How people were divided into groups for The Reading Trial
A researcher that was not part of The People’s Trial team created an online program that placed (or allocated) people into the intervention group (reading a book in bed), or the control group (not reading a book in bed). Who got into which group was decided randomly by the computer program, which meant that everyone had an equal chance of being in the intervention group or the control group. We call this process randomisation.
We carried out the randomisation using a 1:1 ratio. This meant that for every person placed in the intervention group (reading a book in bed), one other person was placed in the control group (not reading a book in bed). This was to make sure The Reading Trial was a fair comparison between the two groups.
In The Reading Trial, neither the researchers nor people taking part in the trial knew in advance which group a person would be put into. Because the trial relied on people doing or not doing something, it was impossible to hide, or blind, people to the group they were placed into (reading or not reading a book) in this trial. This meant that people taking part in The Reading Trial were aware of the group that they were allocated to. However, the researchers running the trial day-to-day did not know who was in each group.
Interventions
As part of The People’s Trial, members of the public designed the steps of The Reading Trial. Through an online survey, they defined the characteristics of the intervention (reading a book in bed), the comparator (not reading a book in bed), and the outcome (sleep quality). They also told us how they thought the trial should measure the outcome.
Intervention Group
People in the intervention group:
- Read a book for 15-30 minutes immediately before trying to go to sleep for seven nights in a row
- Went to bed and woke up at the same time as they usually would
- Did not eat food or drink caffeinated drinks (such as coffee) within 1 hour of going to bed
- Slept in their bed, in their own home, for the study duration
- Could use electronic entertainment or communication devices (e.g. mobile phones/tablets) in bed for the seven nights of The Reading Trial.
Control Group
People in the control group did the same as those in the intervention group, except they did NOT read a book immediately before trying to go to sleep.
This meant that people in the control group:
- Did not read a book immediately before trying to go to sleep for seven nights in a row
- Went to bed and woke up at the same time as they usually would
- Did not eat food or drink caffeinated drinks (such as coffee) within 1 hour of going to bed
- Slept in their bed, in their own home, for the study duration.
- Could use electronic entertainment or communication devices (e.g. mobile phones/tablets) in bed for the seven nights of The Reading Trial.
There were no other rules. This meant that the only difference between the intervention group and the control group was reading a book in bed, or not, for the seven nights of the study.
Outcomes
The main or primary outcome of The Reading Trial was how well people slept, in other words their overall sleep quality. This was measured using a scale (or questionnaire) called the ‘single item sleep quality scale (SQS)20. It measures sleep quality using a simple format, and when compared to longer, more complex questionnaires, this simple scale produced similar results20.
After completing The Reading Trial, people rated their overall sleep quality using this scale, which is numbered from 0 to 10, with numbers increasing in units of one, (0 = terrible, 1–3 = poor, 4–6 = fair, 7–9 = good, and 10 = excellent).
The other outcomes we measured were sleep disturbance and daytime sleepiness.
We measured sleep disturbance using the PROMIS (Patient-Reported Outcomes Measurement Information System) Short-Form Sleep Disturbance Scale. This has eight items, each on a 5-point scale, with a difference of one unit between each point on the scale. It was developed by The National Institutes of Health21, 22. In The Reading Trial, we used the PROMIS scale to measure how often people had problems from not having enough sleep.
Separately, we measured people’s experience of ‘daytime sleepiness’ using a 10-point scale, again the points on the scale increase in units of one. Previous studies have found this simple scale to be accurate in measuring daytime sleepiness when compared to more complex scales23.
For all outcomes, the time we were interested in was the seven days during which a person took part in The Reading Trial.
Sample size - How many people did we need to get reliable results?
We wanted to be sure that we had enough people (in other words, a large enough ‘sample size’) taking part in The Reading Trial to be confident that the results were reliable. To find out how big our sample size needed to be, we searched the literature (published studies and reports) to see how common sleep disorders were reported across different countries and in different groups of people (e.g., students, older adults). We also looked at how sleep quality was reported, e.g. sleep disturbance (sleep broken by wakening) or sleep latency (the amount of time it takes you to go from being fully awake to sleeping) and sleep duration (how long you sleep). One study of 2089 people estimated that 57% of people aged 18-70 years have enough sleep24. This gave us some information on the number of people reporting poor quality sleep in general and, based on this information, we felt that an improvement of at least 10% on an individual’s sleep quality would be considered meaningful.
We worked out how many people we would need to take part in the trial for us to have enough information to ensure a high chance (80% or more), of being able to detect a real difference in sleep quality between the two groups, and a low chance (5% or less), of seeing a difference that was not real, or seeing a difference that happened just by chance. Using this information, we estimated that we would need at least 564 people in The Reading Trial, of which 282 would be in the intervention group (reading a book in bed) and 282 in the control group (not reading a book in bed).
If a person left the trial after being placed in one of the groups, but before the trial had started, we didn’t use their information. To try to make it less likely that people left before the trial started, people were placed in their groups on the same day as they started the trial. We tried to include everyone who began the trial in our calculations. If someone left the trial, we asked that person if we could still collect information important to the study even though they were no longer taking part in the trial. We recorded all information and communication with people taking part in the trial on a database that only the research team could access. We analysed the data by looking at the outcomes (results) for people in each group who completed the outcome assessments at Day 7. We also used a type of data analysis called ‘intention to treat’, which analysed the information of everybody randomised in The Reading Trial based on the groups to which they were allocated, and whether they completed the trial or not.
As The Reading Trial took place over a short time (7 days) and was a low-risk study, the Steering Group (a group of people that provided overall supervision of the trial), decided that we did not need a Data Management Committee for this study. The ethics committee, a committee whose role is to protect people taking part in research25, were happy with this decision.
Analysing information
In this section, we will describe the different ways we looked at the results of The Reading Trial in order to check if reading a book at night is likely to improve quality of sleep.
We analysed the data collected on everybody who completed The Reading Trial based on which of the two groups they were randomly allocated to. Not everyone completed the outcome assessments at Day 7, in fact 217 (21.9%) people did not. This can create problems when analysing trial data, so we repeated the analysis, but this time filled in gaps in the data using a statistical technique called imputation. This uses the data we do have from participants to estimate what the missing data might have been, and allowed us to see how the missing data might affect the overall results.
Main Outcome
For The Reading Trial, we asked enough people to take part so we could detect a difference in improved sleep quality if there was any. We measured sleep quality on a scale that went from 0 to 10. This allowed us to work out whether people had improved quality of sleep, no change, or a worse quality of sleep from the start to the finish of the trial. We then compared this information for people in the intervention group (reading a book in bed) with people in the comparator group (not reading a book in bed). We looked at the data to see how certain we could be of our findings. We checked the data by creating graphs and summaries that would help us identify unusual values or results that needed further checking.
Once we were happy that the data we had collected were correct, we made graphs to help us see whether the two groups were similar at the start of the trial (i.e. that randomisation worked). We also checked if our findings weren’t simply a result of other differences between the groups. We did this by working out the ‘typical’, or most likely value for each measurement, and then looking at how these were different from person to person and between the intervention (reading a book in bed) and the control group (not reading a book in bed). This information is not only useful to this trial but to help design future trials that also wish to compare sleep quality.
We compared the overall sleep quality between people in the intervention group and people in the control group. We used a statistical model, called a Proportional Odds Model, that takes into account that sleep quality in The Reading Trial was measured using categories (i.e. ‘terrible’ being the worst sleep quality through to ‘excellent’ being the best sleep quality) as well as the influence of:
- sleep quality at the start of the trial
- gender
- age
- knowledge of clinical trials
- whether the person worked in healthcare or not.
As part of the main analysis, we looked at whether a person’s quality of sleep improved from the start to the finish of the trial. We compared this information for people in the intervention group with people in the control group, to see if the proportion of people with improved sleep was likely to be different in general.
Secondary outcomes
We looked at changes in sleep disturbance and daytime sleepiness between the intervention (reading) and control (non-reading) groups. This time we used a statistical model, called a Linear Model, which took into account that sleep disturbance and daytime sleepiness are measured as a score (i.e. sleep Disturbance was measured on a 5-point scale, daytime sleepiness was measured using a single 10-point scale). Once again, we took into account the influence of people’s sleep quality at the start of the trial, gender, age, knowledge of clinical trials and whether a participant worked in healthcare or not.
To make sure our findings were not simply a result of chance, we decided, before we did any analysis of the results, what level of certainty we would need to see in order to claim that reading a book in bed is beneficial to people similar to those who took part in The Reading Trial. The value generally used in clinical trials to represent this level is 0.05, meaning that there is a 1 in 20 chance of falsely claiming that an intervention worked. This is the value we used for The Reading Trial.