Brief alcohol intervention for risky drinking in young people aged 14 15 years in secondary schools: the SIPS JR-HIGH RCT

Alcohol use as a public health priority

Consumption of alcohol is a risk factor for mortality and morbidity in adults, and is an important public health issue. Alcohol misuse leads to both societal and economic costs, as well as the burden of disease, yet it is preventable. 8 Although many adults in the UK consume alcohol in line with recommended guidelines, a proportion of adults drink above these recommendations and consume alcohol at harmful and hazardous levels. 9 It has been estimated that 10.8 million adults in the UK drink alcohol at levels that are harmful to their health. 10 Although young adults aged 16–24 years are less likely to drink than individuals in older age groups, when they do choose to drink they consume more alcohol. 11 As a result, this study focused on alcohol consumption in young people.

Prevalence of alcohol consumption in young people

In 2009 the Chief Medical Officer for England recommended that young people and children remain abstinent from alcohol until they are 18 years old. 12 This was accompanied by advice that young people and children should not consume alcohol before the age of 15 years, and, should they choose to consume it thereafter, that they should drink it no more than once per week and only under supervision. If young people aged 15–17 years drink alcohol, it is recommended that they do so on only 1 day per week and do not exceed adult daily limits, and ideally consumption should be below this level. 13

In the UK, the proportion of young people who drink alcohol has steadily decreased between 2003 and 2014. 12 As there was a change in the question around consumption of alcohol in the 2016 version of the ‘Smoking, Drinking, and Drug Use’ survey,14 it is not possible to calculate the change in alcohol consumption in 2016 relative to previous years; however, the direction of effect continues to show decreasing alcohol consumption.

The proportion of young people who have ever had an alcoholic drink in England increases with age, with data from 2016 suggesting that 11% of girls and 9% of boys aged 11–15 years had consumed alcohol in the 7 days preceding the survey. 12 Of these, 1% of 11-year-olds had consumed alcohol in the preceding week, increasing to 24% of 15-year-olds. In total, mean alcohol consumption in ‘the last week’ was lowest in those aged 11–13 years (6.9 units) and highest among 14-year-olds (11.1 units). 12 In terms of the amount of alcohol drunk by 11- to 15-year-olds, in 201712 the mean alcohol consumption in the preceding week was 10.3 units for boys and 8.9 units for girls. 12 For these reasons, this study focused on young people aged 14–15 years, given that the data indicate that more young people start to consume alcohol at this age.

Although alcohol consumption among children and young people is declining, consumption among high-risk children and young people (e.g. those with intellectual disabilities) remains prevalent. 15 In comparison with other European countries, the UK has high levels of drinking among young people. 16 The north-east has one of the highest rates of young people who have ever drunk alcohol, with 49% of 11- to 15-year-olds reporting having ever drunk alcohol. 17 This compares with the lowest prevalence in London of 25% and the highest in the north-west of 50%, with an overall English average of 44%. 12

Consequences of drinking alcohol at a young age

Alcohol consumption at a young age is associated with a number of detrimental outcomes. These include physical and mental health issues, an impact on brain development, and an increased risk of accidents and injury. 12 Longer-term negative outcomes arise in particular from binge drinking, defined as the excessive consumption of alcohol in a limited time and often measured as consuming more than six units in a single session for both men and women. 18 These long-term effects have been observed in a large cohort study (the 1970 British Cohort Study) of > 16,000 babies born between 5 and 11 April 1970, and followed up at ages 5, 10, 16 and 30 years. 19 The findings showed that binge drinking in adolescence was associated with later (adulthood) negative consequences, such as alcohol dependence, homelessness, reduced educational attainment and convictions. In addition, there is evidence that alcohol use tracks over time, from adolescence into adulthood, and also is correlated with other risky behaviours, such as smoking. 20 That study therefore recommended that alcohol interventions target young people, even though the literature suggests that they do not always think that risky or harmful drinking is a concern for them. 21 More immediate consequences of alcohol consumption include school exclusions, with 9.5% of exclusions during 2015–16 in state-funded secondary schools being due to drugs and/or alcohol. 22

Other adverse effects of alcohol consumption in young people include an increased risk of mortality from accidents and suicide as a direct result of drinking alcohol. 23 Additional negative consequences include longer-term impact on brain development, liver damage, and changes in hormones vital for organ development and growth. 23,24 Short-term impacts can also arise from alcohol use in young people, including regretted sexual activity, self-harming, alcohol poisoning, drunk driving and criminal behaviour. 25 It can also lead to weight loss, appetite changes, sleep disturbance, depression and an impact on school performance. 26 In addition, the early consumption of alcohol has been shown to link with the amount of alcohol consumed in older adolescence and adulthood. 27 It is also the case that young people are more likely to binge drink alcohol when they do consume it, which in turn leads to increased risk from accidents. Alcohol use is also linked to non-communicable diseases, such as cancer, cardiovascular disease and gastrointestinal disorders. 28 In terms of short-term benefits, young people report similar reasons for drinking as adults, including for social confidence, for enjoyment26,29 and to celebrate special occasions. 26

Alcohol screening tools

A number of screening tools have been used with young people to identify those who are at risk from their drinking. A systematic review into the clinical effectiveness and cost-effectiveness of alcohol screening tools for adults and young people explored the use of the Alcohol Use Disorders Identification Test (AUDIT),30 Alcohol Use Disorders Identification Test – Consumption (AUDIT-C)31 and AUDIT-QF (first two questions of the AUDIT),32 FAST33 (Fast Alcohol Screening Test), CAGE,34 MAST (Michigan Alcohol Screening Test),35 Paddington Alcohol Test,36 SASSI (Substance Abuse Subtle Screening Inventory) for children and young people,37 ASSIST (Alcohol, Smoking and Substance Involvement Screening Test),38 SASQ (Single Alcohol Screening Question),39 TWEAK (Tolerance, Worried, Rye-opener, Amnesia, K/cut down)40 and T-ACE41 in prenatal screening, as well as laboratory and clinical markers. 42 The review found that alcohol screening tools were more effective at identifying young people drinking at a risky level than laboratory and clinical markers, with the AUDIT a particularly cost-effective measure. 43 The SIPS JR-HIGH pilot trial report discussed the evidence around screening measures in detail,2 although a review of existing reviews25 for ASBIs with young people found that the CRAFFT (Car, Relax, Alone, Forget, Family or Friends, Trouble) was used particularly for older adolescents (15–18 years old), and the AUDIT was found to have greater sensitivity and specificity than other screening tools in young people. In addition, the Adolescent Single Alcohol Question (A-SAQ) was shown to be a reliable single-screening question for drinking frequency44 in the SIPS (Screening and Intervention Programme for Sensible Drinking) research programme in adults. It is a modified version of the Single Alcohol Screening Question (M-SASQ),45 which is adapted for adolescent alcohol consumption. 39

Alcohol primary prevention in the school setting

Primary prevention is a treatment or intervention that seeks to prevent disease occurrence, whereas secondary prevention involves treating/intervening in a disease in its early stages. ASBIs are classified as a form of secondary prevention, in that they typically target individuals who have been identified as drinking alcohol in a pattern that is detrimental to their health. 46

There is limited literature published within the past 5 years that explores primary preventative school-based alcohol brief interventions in the secondary school setting in the UK. The majority of recently published literature focuses on college and university students in other countries, particularly US studies focusing on college students.

Of the previous studies that do exist, they have explored the use of classroom-based curricula and parental interventions as primary prevention with young people. 47 One such is the Steps Towards Alcohol Misuse Prevention Programme trial,47 which explored the effectiveness of a school and parent alcohol intervention. This trial found a significant reduction in heavy episodic drinking in 12- to 13-year-olds; the intervention was delivered at classroom level and not to individuals. 47 The Kids and Adults Together programme trial48 explored the acceptability of primary school classroom-based activities, family events and a digital versatile disc (DVD) to address the effects of alcohol in 9- to 11-year-olds in Wales as a primary preventative measure. The trial found mixed support from the nine primary schools, with two withdrawing from the study. One particular limitation highlighted was the inability to conduct follow-ups within secondary school settings once the young people progressed from primary school. In addition, members of the team conducted a systematic review of peer-led interventions with young people aged 11–21 years; they found 17 studies, of which six were school-based and showed a positive benefit on alcohol use. However, none of the six school-based studies was conducted in the UK and brief interventions were specifically excluded. A meta-analysis of school-based prevention for risky behaviours showed that school settings appear to be effective in reducing alcohol consumption,49 whereas a separate review50 of the prevention of multiple health risk behaviours in schools showed a small positive effect on alcohol consumption. The former review was, however, published in 2001 and the latter was not focused on individually targeted interventions and, therefore, the findings are of limited relevance in the context of the individualised ASBI considered here.

Alcohol screening and brief interventions

Interest in screening and brief interventions for risky drinking has developed since the 1970s. 51 ASBIs have been defined as ‘those practices that aim to identify a real or potential alcohol problem and motivate an individual to do something about it’. 52 Heather, in 1995, offered a more specific definition, stating that brief interventions are ‘a family of interventions varying in length, structure, targets of intervention, personnel responsible for their delivery, media of communication and several other ways, including their underpinning theory and intervention philosophy’. 53 In an early review of brief interventions, Bien54 found that brief interventions were more effective than no intervention or more extensive treatments, with six common elements of brief interventions discussed, based on the FRAMES (Feedback, Responsibility, Advice, Menu, Empathy and Self-efficacy) model. 54 More recent research has reported results of brief interventions in different settings and delivery modes, for example face to face or web based. 55 In terms of implementing a brief intervention, it is often used to provide simple advice for patients with an AUDIT score in the range of 8–15, who may be at risk of injury and chronic health conditions as a result of their alcohol consumption. 52 Given the heterogeneity of brief interventions and settings, it is therefore important to consider the context and components of brief interventions when assessing effectiveness, rather than focusing on general overall effectiveness. Heather53 in particular states the importance of clarifying the length of the intervention when assessing effectiveness, given that the length can range from 5 minutes to at least 3 hours. The same is true of the number of intervention sessions, which can be one or more, and the content of sessions (e.g. MI vs. self-help manual).

The pilot trial focused on the use of simple structured advice based on the FRAMES model, delivered by learning mentors as non-specialists, as opposed to behaviour change counsellors. 1 This approach was found to be acceptable in the school setting and hence was used again in this main trial.

In terms of secondary prevention, a previous review of reviews explored the use of ASBI with young people. 25 The review concluded that MI delivered in school settings is an effective way to reduce alcohol consumption. However, within this review, the definition of young people was wide, ranging from 10 to 21 years, even though the World Health Organization defines ‘young people’ as those aged 10–19 years. 56 An additional systematic review and meta-analysis of ASBIs for young people found that ASBIs led to a decrease in alcohol consumption and alcohol-related problems. However, this work included studies that focused on any level of (alcohol) risk, and also defined eligible participants as those aged 11–25 years. 57

There is some evidence to show school-based group ASBIs can be beneficial,58 but, even then, this has not always been found to be the case for high-risk drinkers. 59 Similarly, there is literature showing that family and school-based interventions can reduce adolescent alcohol use, but this evidence60 is not from a UK setting and does not involve personalised individual ASBIs. A separate systematic review57 of ASBIs with young people showed reductions in alcohol use as a result of ASBIs but included participants up to the age of 30 years, well outside secondary school age.

To focus on young people in particular, and also using recognised definitions for young people, a systematic review [Giles EL, McGeechan GJ, Ferguson J, Byrnes K, Newbury-Birch D. Teeside University. 2018. (In preparation)] was recently conducted to explore the evidence on the efficacy/effectiveness of ASBIs targeting risky drinking in young people (as defined by the World Health Organization) in randomised and non-randomised controlled designs, or quasi-experimental studies. The ASBI had to involve individual one-to-one advice delivered in one to four sessions to constitute a ‘brief’ intervention. Literature searches were conducted in June 2017 of the main databases including PsycInfo, Psycharticles, MEDLINE, Scopus and CINAHL (Cumulative Index to Nursing and Allied Health Literature). Searches combined key alcohol, brief intervention and young people search terms. Papers were eligible for inclusion if they were shown to meet the predefined inclusion criteria, and if their primary focus was around young people aged between 10 and 19 years, engaged in an alcohol brief intervention in any setting, in randomised or non-randomised controlled designs or quasi-experimental studies.

This review identified 16 papers that focused on risky drinking in young people. 61–76 The majority (n = 13) were studies conducted in the USA,61,64–67,69–76 with one in Germany,68 one in Mexico63 and one in Western Australia. 62 The setting was mainly emergency departments (n = 8),62,66,68,70–73,75 with three in universities,61,67,74 one with homeless adolescents,76 one with young people referred from not-for-profit agencies,65 one based in the community64 and two based in schools. 63,69 The ASBIs used largely followed a manual and/or used MI techniques. Findings showed that for the majority of the studies, the ASBI was effective at decreasing alcohol-related outcomes, including a reduction in binge/heavy drinking (n = 4), and/or a reduction in number of typical drinks/alcohol use (n = 10) and/or frequency of drinking (n = 5), and/or a reduction in negative consequences from drinking alcohol (n = 3). One study found no significant differences between the group receiving an ASBI and the control group. 75 Focusing on the two studies conducted in a school setting, one was conducted in Mexico. 63 In this study, the young people were moderate- to high-risk drinkers, mainly male (65%), with an average age of 16 years. The ASBI group received one 90-minute ASBI compared with a waiting list control. At the follow-up points of 3 and 6 months, the ASBI group showed a significant reduction in the amount of alcohol consumed compared with the control. The second study was conducted in the USA with young people who had an alcohol or drug use disorder. 69 The majority of participants were male (52%), with an average age of 16 years. The ASBI were two 60-minute sessions, with one group also receiving a parent session. Significant findings were reported for the number of days alcohol was consumed compared with the assessment-only control group. These two school-based studies therefore suggest that the use of a school setting to deliver ASBI to young people is effective (see Appendix 3).

Aim

The aim of this study was to evaluate the effectiveness and cost-effectiveness of ASBIs to reduce risky drinking in young people aged 14–15 years in the English secondary school setting.

Objectives

1. To conduct an individually randomised controlled trial to evaluate the effectiveness and cost-effectiveness of an ASBI for risky drinkers compared with usual practice on alcohol issues conducted by learning mentors with young people aged 14–15 years in the school setting in four areas of England: the north-east, north-west, Kent and London.

2. To measure effectiveness in terms of percentage days abstinence over the last 28 days, risky drinking, smoking behaviour, alcohol-related problems, drunkenness during the last 30 days and emotional well-being.

3. To measure the cost-effectiveness of the intervention in terms of quality of life and health state utility, quality-adjusted life-years (QALYs), service use costs and cost consequences at 12 months post intervention.

4. To monitor the fidelity of an ASBI delivered by learning mentors in the school setting.

5. To explore barriers to, and facilitators of, implementation with staff.

6. To explore young people’s experiences of the intervention and its impact on their alcohol use.

7. If the intervention is shown to be effective and efficient, to develop a manualised screening and brief intervention protocol to facilitate uptake/adoption in routine practice in secondary schools in England.

Outcomes and measurements of the SIPS JR-HIGH effectiveness trial

Validated tools were used in the study to capture the following primary and secondary outcomes measures. 2

Primary outcome

The primary outcome measure was the number of units of alcohol consumed over the previous 28 days, derived using the TLFB completed at 12-month follow-up77 (see documents on project page). The TLFB is designed to be completed retrospectively and employs periodic cues to obtain reliable estimates of daily drinking during a 28-day period.

Secondary outcome measurements

• The Adolescent Single Alcohol Question (A-SAQ)78 was used to measure risky drinking (scoring ‘4 or more times but not every month’, ‘at least once a month but not every week’, ‘every week but not every day’, or ‘every day’). A score of three or above was considered positive for possible hazardous or harmful drinking. 2

• Alcohol use frequency, quantity (on a typical occasion) and binge drinking (six or more drinks in one session for men and women)18 was assessed using the 10-question AUDIT. 79,80 Questions 1–8 each have five possible responses relating to how much or how often drinking behaviours occur, and these are scored from 0 to 4. Questions 9 and 10 have three responses and are scored 0, 2 or 4. Scores were summed to give a possible range from 0 to 40. A score of ≥ 4 was indicative of hazardous alcohol consumption in adolescent populations,81 and the adult cut-off point of 8 was also used. 2,30

• The AUDIT-C was used to assess risky drinking, and comprises the first three questions of the AUDIT. All questions were scored from 0 to 4 and summed to give a range of scores from 0 to 12. An AUDIT-C score of ≥ 3 was indicative of hazardous alcohol consumption and ≥ 5 was indicative of possible dependence in adolescents. 2

• Alcohol-related problems were assessed using the validated Rutgers Alcohol Problem Inventory (RAPI), which includes measures of aggression. 82 It consists of 23 questions about drinking behaviour, each with four possible responses, all of which are scored from 0 to 3. Responses were summed to give a total score ranging from 0 to 69. A higher RAPI score indicated more problematic drinking behaviour. 2

• Drunkenness during the previous 30 days was dichotomised as ‘never’ and ‘once or more’. 2,83

• Drinking motives were assessed using the 20-item Drinking Motives Questionnaire Revised (DMQ-R). 84 There were four sets of five questions designed to assess relative frequency of drinking within each of the subscales of coping, social, enhancement and conformity motives. All questions were on a Likert scale, for which one is ‘almost never/never’, two is ‘some of the time’, three is ‘half of the time’, four is ‘most of the time’ and five is ‘almost always/always’. Higher scores within each domain indicated stronger endorsement of positive reinforcement received through consumption of alcohol. 2,84

• The percentage days of abstinence during the previous 28 days was calculated from the TLFB questionnaire by dividing the days on which no units were consumed by 28 and multiplying by 100. 2

• Number of units of alcohol consumed per drinking day was derived by dividing the total number of units of alcohol consumed by the number of days on which > 0 units were consumed. Abstinent pupils were scored as 0. 2

• Number of days when > 2 units of alcohol were consumed during the previous 28 days was calculated by counting the number of days on which > 2 units were recorded. 2

• General psychological health was assessed using the 14-item Warwick–Edinburgh Mental Well-being Scale (WEMWBS). 85 This tool used a five-point Likert scale that gives a score of 1–5 per question, giving a minimum score of 14 and maximum score of 70. A higher WEMWBS score indicated better mental well-being. 2,86,87

• Two questions relating to sexual risk-taking were included. These were the same questions as in the pilot study:1 ‘After drinking alcohol, have you engaged in sexual intercourse that you regretted the next day?’ and ‘After drinking alcohol, have you ever engaged in sexual intercourse without a condom?’ Both questions could be answered with one of the three following options: ‘I have never engaged in sexual intercourse’, ‘yes’ or ‘no’. Information on changes in risky sexual behaviour for the two questions was derived from a cross-tabulation of baseline and follow-up responses to a given question and categorised as ‘never engaged in sexual intercourse’, ‘engaged in sexual intercourse that has been regretted during the last 12 months’ and ‘have engaged in sexual intercourse that has not been regretted’. 2

• Energy drink consumption was assessed by asking young people how many times per week they consumed energy drinks. Young people could answer ‘never’, ‘less than once a week’, ‘2–4 days a week’, ‘5–6 days a week’, ‘every day once a day’ and ‘every day more than once a day’. Information on changes in the consumption of energy drinks was derived from a cross-tabulation of baseline and follow-up questions and categorised as ‘never consumes energy drinks’, ‘has stopped consuming energy drinks’, ‘has started consuming energy drinks’ and ‘consumes more/less energy drinks’. 2

• Age of first smoking was asked, as was how many cigarettes had been smoked in the previous 30 days. 88 Information on changes in smoking behaviour was derived from a cross-tabulation of baseline and follow-up responses and categorised as ‘never smoked’, ‘still smoking’, ‘started smoking’ and ‘smoking more/less’. 2,89

• Demographic information collected included gender, ethnicity and how free time is spent. The first part of the trial participants’ postcode was collected to facilitate calculation of Index of Multiple Deprivation. 2

• Quality of life was measured using the EQ-5D-3L, which is a valid measure for those aged ≥ 12 years, and was used to measure health-related quality of life. 90 Responses to the five items were converted into utility scores using the UK population algorithm. This was administered at baseline and at 12 months post intervention. 2,90

• Quality-adjusted life-years were estimated using general population tariffs from responses to the EQ-5D-3L administered and scored at baseline and 12 months; service use was assessed using the modified S-SUQ (Short Service Use Questionnaire) at baseline and 12 months. 2

• Use of leisure time was assessed using a multiple-choice question that gave suggestions about how free time might be spent and the option to choose multiple answers. 2

• Incremental cost per QALY gained at 12 months was calculated. 2

• Modelled estimates of incremental cost per QALY and cost consequences in the longer term were calculated. 2

• The NHS, educational, social and criminal services data were estimated using a modified S-SUQ91 and via a learning mentor case diary developed in the pilot study,1 measured at 12 months post intervention. 2

• Cost consequences were presented in the form of a balance sheet for outcomes at 12 months. 2

Study setting and population

Young people aged 14–15 years in Year 10 at secondary schools/academies in four centres, north-east, north-west, Kent and London, were targeted.

In each of the four geographical centres, school performance league tables97 were reviewed and schools from the top, middle and bottom of the league tables were contacted. Efforts were made to recruit a cross-section of schools, including academy schools, schools in deprived areas, faith schools and private schools. Schools were included if they employed learning mentors (or equivalent members of pastoral staff).

In advance of screening, all parents/caregivers (hereafter referred to as ‘parents’) were sent a letter by the school informing them that young people would be screened for risky drinking as part of the study in their child’s school (see documents on project page). Parents had the choice to opt their child out of the study by completing an opt-out form and returning it to the school in a prepaid envelope. Opt-out consent by parents was chosen (instead of opt-in) as this was standard practice in schools (see documents on project page). They were provided with an information sheet to inform consent (see documents on project page). 2

At baseline, all participating schools showed an animation video to pupils who had not opted out of the study. This video detailed the study process. Young people then completed the baseline survey as part of the screening process (see documents on project page). This screening took place in the PSHE or registration class on an individual classroom basis.

Young people had the option to (1) not complete the questionnaire (indicative of lack of assent to screening from the young person), (2) complete the questionnaire anonymously or (3) complete the questionnaire, adding their name and class. 2

Those young people who screened positively on the A-SAQ and left their name were eligible for the trial.

Randomisation

Eligible pupils were individually randomised in a one-to-one ratio to the intervention or control arm of the trial. Randomisation occurred before consent to the trial because of timing constraints: it was not feasible for the learning mentor to access a real-time randomisation system during the session with the young person. Therefore, the randomisation was performed earlier on the list of eligible young people and the allocation was contained in a sealed envelope. Neither the learning mentor nor the young people knew the arm to which they had been allocated until they opened the envelope after consent had been given. Young people aged < 16 years provided assent and those aged 16 years provided consent, as per guidelines issued by the Research Governance and Ethics Committee in the School of Health and Social Care at Teesside University. 1 Hereafter, assent and consent is referred to as consent.

An independent statistician not otherwise involved with the study produced a computer-generated allocation list using random permuted blocks. This statistician was provided with a list of screening identification numbers (identifying the site, school and young person) for eligible participants, in the form of a Microsoft Excel^® (Microsoft Corporation, Redmond, WA, USA) spreadsheet. This statistician also undertook randomisation, and an updated spreadsheet, including allocations to the study arms, was returned to an independent researcher (JF) at Teesside University, where the sealed envelopes containing the allocation were produced. Numbered envelopes containing a slip of paper indicating the arm to which each individual had been allocated were then prepared, sealed and sent to each school, to be opened only when consent had been attained.

School staff identified to deliver interventions

The pilot feasibility trial, which preceded the current trial, identified the learning mentor as the most appropriate role within the school structure to deliver an alcohol intervention. Learning mentors are specifically trained to provide a complementary service to teachers and other school staff, addressing the needs of young people who require assistance in overcoming barriers to learning in order to achieve their full potential. 98 They work with a range of pupils, but they focus on those pupils with multiple disadvantages that have an impact on their education. Within their established role, learning mentors address issues, such as punctuality, absence, bullying, challenging behaviour and abuse, disaffection, danger of exclusion, difficult family circumstances and low self-esteem, as well as underperformance against potential. 99 As part of this role, learning mentors are routinely required to provide advice and support to young people and, as such, are well placed within a school setting to deliver an intervention to young people about alcohol use. In this trial, interventions took place in the learning mentor’s classroom or office space.

Local areas vary in the essential qualifications they look for when appointing learning mentors. However, at a minimum, learning mentors need to have a good standard of general education, especially literacy and numeracy, as well as experience of working with young people. 100 In the present study, learning mentors were defined as the members of school staff trained in the delivery of the control condition/intervention to participating students. However, in practice, within each school, titles, roles and responsibilities varied and this did not constitute a homogenous professional group. Thus, for consistency, the school staff responsible for delivering interventions are referred to only as learning mentors throughout the rest of this report.

Control (healthy lifestyles leaflet)

Young people allocated to the control group met with the learning mentor in school during the week (Monday to Friday). The learning mentor explained the study to them and invited them to participate in the trial. Once a young person had consented to the trial, the learning mentor provided them with the healthy lifestyles leaflet (see documents on project page). This leaflet contained advice on eating less fat, sugar and salt; eating fruit, vegetables and fish; and the importance of breakfast. It also advised readers to ‘move more’. The healthy lifestyle leaflet was chosen as the minimally acceptable intervention that could be provided to young people without giving any alcohol advice. No feedback on the alcohol screening results was provided. The young people in the control group may also have received usual practice advice from the school.

Intervention (brief intervention and alcohol leaflet)

Appendix 5 reports further details on the intervention using the Template for Intervention Description and Replication (TIDieR) checklist. 101 The rest of this section details the intervention.

Young people allocated to the brief alcohol intervention met with the learning mentor face to face in school during the school week (Monday to Friday), in the learning mentor’s office or an alternative suitable space. The learning mentor provided feedback on their positive alcohol screen result and invited them to take part in the trial. Those who consented to the trial received one 30-minute one-to-one interactive session brief alcohol intervention. The session was developed within the pilot feasibility trial. The essential components of brief alcohol interventions were adapted for young people within a school setting after consulting with a PPI group and piloting the intervention. 1 The developed tool was a colourful A3 sheet detailing a six-step intervention, which was intended to lead to an interactive discussion between the young person and the learning mentor (see documents on project page). Based on the principles of MI,102 the intervention sought to increase the young person’s awareness of risks and enable them to consider their motivations for changing their alcohol use. It encompassed the elements of the FRAMES approach for eliciting behaviour change, as outlined in Chapter 1. 54 It was expected that young people would be taken out of class to attend the appointments with learning mentors. The intervention had been found to be feasible and largely acceptable within the pilot feasibility trial and linked qualitative study. The intervention used in the pilot trial, however, had included information about the calorie content of alcoholic drinks. Mixed views were expressed about this content in the pilot feasibility trial and, as such, this was removed from the intervention delivered in the present trial. Table 1 summarises the intervention and control arms.

Intervention worksheet

Training and support

All learning mentors received training from November 2015 to January 2016 before commencing the study. The training was delivered to learning mentors in group sessions within the school environment by the trial co-ordinator in each study site. The trainers were themselves trained by Ruth McGovern, a senior interventionist. Outreach training has been found to be the most cost-effective implementation strategy for ASBI delivery in other settings. 107 Microsoft PowerPoint^® (Microsoft Corporation, Redmond, WA, USA) slides were used to guide each training session. This was then followed by an audio-recorded role-play exercise wherein learning mentors practised delivering the intervention with each other to enable the development of skills. Learning mentors were required to record one role-play intervention delivery each. Learning mentors were also trained to issue the participant information leaflet, gather informed consent from the young person and deliver the alcohol information leaflet. Each training session lasted 2–3 hours. An intervention manual was provided to supplement the training (see documents on project page). Following the training session, an experienced alcohol interventionist (RM) listened to the audio-recorded role-plays, rated them using the BECCI and provided feedback on the content. This process was to ensure that learning mentors achieved a minimum standard of fidelity. All learning mentors were required to be rated at a minimum standard (a score of at least 1 on the BECCI) before they delivered the intervention to participating young people within the trial. A total of 80 learning mentors were trained from the four study sites. Four learning mentors were asked to repeat the role-play exercise after detailed feedback was given on their practice. All four learning mentors met the minimum standard following the second exercise. The number of learning mentors per school ranged from one to eight.

The research team provided regular support to the learning mentors, including ongoing guidance on intervention delivery. Co-ordinators made weekly visits/telephone calls and/or sent e-mails to each school throughout the study period to answer questions or concerns, collect materials from completed interventions (such as consent forms and hard copies of intervention tools) and encourage learning mentors to complete outstanding interventions. Finally, learning mentors were provided with a case diary sheet on which they were asked to record any interactions with the individual young people in the trial (see documents on project page). This information was used to calculate the time spent on intervention delivery and inform the cost-effectiveness analysis (see Chapter 5).

Fidelity of the interventions

Fidelity of an intervention within research refers to the extent to which the intervention is true to the therapeutic principles on which it is based. 108 Researchers need to be able to determine whether or not the intervention was delivered as intended; this requires the manualisation of the intervention wherein the philosophy, principles and procedures of the intervention are clearly described. Interventions delivered with high fidelity promote the ability to associate trial results with intervention effectiveness. A manualised intervention with verified fidelity enables the research to be replicated or the intervention to be implemented in practice, if it is shown to be effective.

Learning mentors were asked to audio-record all sessions with consenting participants that they delivered during the trial, from which a 20% sample from various time points would be rated to assess fidelity. However, only 18 recordings (from 11 schools) of the total 443 sessions (control interaction, n = 7, 3%; intervention delivery, n = 11, 5%) were made. Reasons for not recording sessions included the young person’s refusal, unavailable equipment and human error (i.e. forgetting to record the sessions). Because only a limited number of sessions were audio-recorded, all 18 recordings were rated. In this study the control condition was simply assessed as providing no alcohol advice (acceptable levels of differentiation from the alcohol brief intervention) versus providing any alcohol advice (unacceptable levels of differentiation). All seven recordings were free of any alcohol advice and, as such, this small selection of recordings were delivered to an acceptable level of differentiation. The BECCI was used to measure fidelity of the audio-recorded brief alcohol interventions delivered during the trial. BECCI is a tool developed specifically to measure the microskills of behaviour change counselling and MI (e.g. questions, empathic listening statements, summaries). 109 The instrument focuses on the practitioner’s consulting behaviour and attitude rather than the patient’s response. Sessions were rated by an experienced alcohol practitioner within the research team (RM), who is dual qualified in social work and counselling to master’s degree level. Scores were given on a range of 0–4 for 15 different items on a checklist, for which 0 = ’not at all’, 1 = ’minimally’, 2 = ’to some extent’, 3 = ’a good deal’ and 4 = ’a great extent’. A mean score was then calculated. Rating was completed in line with the BECCI Manual for Coding Behaviour Change92 for the purposes of measuring and reporting on the fidelity of intervention delivery within the trial. The mean BECCI scores for the 11 sessions ranged from 0.3 (behaviour change counselling delivered ‘not at all’) to 2.5 (behaviour change counselling skills delivered ‘a good deal’). The mean BECCI score for the 11 recorded interventions was 1.6 and the median BECCI score was 1.5; these scores suggested that overall the learning mentors were found to be delivering behaviour change counselling to ‘some extent’. This is similar to the BECCI scores reported in other studies110 of practitioner-delivered behaviour change counselling. Learning mentors typically performed well when discussing the risks associated with the young person’s alcohol use. Lower scores were observed in respect of microskills relating to discussing and exploring behaviour change.

The small number of interventions recorded is a significant weakness and limits our ability to assess threats to internal validity. 111 It is possible that the learning mentors who did not provide a recording differed in skill from those who did. Learning mentors may have selected recorded sessions in which they felt they performed better.

Inclusion criteria

• Young people aged 14–15 years, inclusive.

• Parents had not opted them out of the study.

• Scored positive on the A-SAQ and left their names.

• Were willing and able to provide informed written consent for intervention and follow-up.

Exclusion criteria

• Were already seeking or receiving help for an alcohol use disorder.

• Had recognised mental health or challenging behaviour issues, identified by school staff.

Sample size and power

The sample size was calculated to have a 90% power to detect a standardised difference of 0.3 (which equated to a ratio of 1.5 in geometric means) in total alcohol units consumed in 28 days, using a significance level of 5%. Parameter estimates came from the pilot trial1 (mean year group size = 210, 87% completing baseline questionnaire, 80% recruited to trial and 88% providing data at the 12-month follow-up). In the pilot trial we found that 18% of young people had screened positive and left their names. However, given that the numbers screening positive were anticipated to be lower in the south of England, and the extra efforts put in place in the current trial to encourage young people to leave their names on the questionnaires, it was estimated that this figure would be 20% in the full trial. Using these estimates, follow-up data would be required from 235 young people per arm. The number required to have follow-up data if the power was reduced to 80% was 176. Given the difficulty in recruiting sufficient numbers, the target for power was later reduced from 90% to 80%, after discussion with the TSC and DMEC.

Statistical analysis plan

Individual site staff input the trial data into a MACRO™ (Elsevier Limited, London, UK) database held and maintained by the Newcastle Clinical Trials Unit. Data were extracted from MACRO into statistical software package Stata^® version 14 (StataCorp LP, College Station, TX, USA). 1 A complete statistical analysis plan providing full details of all statistical analyses, variables and outcomes was finalised and signed before the final database download. Scoring systems for numerical scales are shown in Appendix 6. All analyses were carried out on an intention-to-treat basis, retaining patients in the analysis group to which they had been randomised and including any protocol violator and ineligible participants.

Descriptive analyses

After the recruitment of schools, the flow of young people through the study was presented using a Consolidated Standards of Reporting Trials (CONSORT)2 flow diagram (Figure 2). The number of young people in the age group, how many completed baseline questionnaires, how many screened positive, how many assented and how many completed the 12-month follow-up were reported by region and by school. Any reasons for not completing follow-up were reported when known. Descriptive statistics were used to report the pupil-level baseline data, by arm. These included all measures reported in the baseline questionnaires, reported as comparisons of percentages and means or medians as appropriate.

FIGURE 2.

The CONSORT flow diagram.

Inferential analyses

It was planned to use multiple linear regression to compare the primary outcome (mean number of units consumed in the previous 28 days) between the two trial arms at 12 months, estimating the difference between mean scores and adjusting for any imbalance in key covariates, namely gender, region (north-east, north-west and south), level of deprivation (percentage of free school meals) and baseline AUDIT score. Data from the London and Kent areas were combined as the ‘south’ region as there were very few young people providing responses at baseline and 12 months in London schools. If, as was considered likely, the primary outcome was skewed, the statistical analysis plan stated that there would be a logarithmic transformation and results would be presented as ratios of geometric means. Alternatively, other root transformations could be considered and confidence intervals (CIs) around differences in means would be obtained by bootstrapping. However, at 12 months a substantial minority of participants reported consuming 0 units of alcohol. This made it impossible to use a logarithmic transformation, and root transformations did not deal with the skewness. Hurdle models3 were considered to deal with the zero inflation, but convergence was not achieved for all variables. It was therefore decided to use quantile regression to model the median number of units consumed in the previous 28 days using the same covariates: the results were very similar to those of the Hurdle models when results were available. The majority of secondary outcomes were analysed similarly. The only exception was the comparison of the proportions who had reduced their drinking below hazardous (or dependent) levels between baseline and 12 months; this was done by logistic regression, adjusting for the same covariates as for other outcome variables.

In a post hoc analysis, Bayes factors were calculated to determine the strength of support for the alternative hypothesis (H_A: the intervention shows some difference in alcohol consumption between trial arms) versus the null hypothesis (H₀: no effect). Bayes factors can determine whether there is simply a lack of evidence for the alternative hypothesis or whether the evidence supports the null hypothesis. A Bayes factor of > 1 indicates support for H_A, whereas one of < 1 indicates support for H₀. Values of Bayes factors of > 3 or < 1/3 are regarded as substantial evidence for the alternative or null hypothesis, respectively.

Exploratory analyses

Exploratory analysis included investigating differences in the primary outcome by gender, deprivation (percentage of free school meals) and levels of drinking at baseline (A-SAQ). Gender and A-SAQ differences were illustrated by box plots, and associations with deprivation were illustrated by a scatter plot. The levels of drinking according to the A-SAQ at baseline were categorised as those drinking ‘more than 3 units on four or more times in the last 6 months but not every month’, those drinking ‘at least once a month’ and those drinking ‘at least once a week (including those drinking daily)’. It had been planned to look at the extent of the intervention received, but as all of the young people allocated to the brief intervention received it, no such analysis was necessary.

Recruitment and follow-up

The recruitment of schools took place between November 2015 and June 2016. During this period, 154 schools were invited to take part (north-east, n = 44; north-west, n = 35; London, n = 36; and Kent, n = 39). Thirty-three schools were recruited (north-east, n = 13; north-west, n = 7; London, n = 5; and Kent, n = 8); few of the non-participating schools gave reasons for not taking part, but those that did mentioned impending school closure (n = 1), lacking time or staff availability (n = 7), having a policy not to take part in surveys (n = 1) and lacking agreement with elements of the research (n = 2). Two schools that initially agreed to take part subsequently withdrew shortly after recruitment: one in the north-west did not like the sexual health questions, whereas the other, in Kent, withdrew because of school closure. A third school in London later withdrew because of issues with the trial; this was after the young people had been screened but before the meetings to obtain consent had taken place. The number of schools recruited had to be increased from the 20 planned, partly because the average year-group size was much smaller than anticipated from our experiences in the pilot trial.

Recruitment at the young person level is summarised in the CONSORT flow diagram (see Figure 2). Table 2 shows the figures on recruitment, screening, consent and randomisation, by area. Among the schools recruited, 4523 (98.7% of the year groups) pupils completed the baseline questionnaire and 1064 (23.5% of those completing the questionnaire) screened positive. Of those screening positive, 443 (41.6%) consented to take part in the study by providing their names. Using the pilot study data,1 we had expected that 14% of those surveyed would screen positive and consent to take part in the trial, whereas we managed to recruit only 443 out of 4584 (9.7%). The lower than anticipated numbers eligible and willing to take part was largely explained by fewer than expected screening positive on the A-SAQ. Four young people scored negative on the A-SAQ and were incorrectly included, even though they were ineligible, through human error. Three of these were randomised to the control group and completed the 12-month follow-up, but the one randomised to the intervention group withdrew before the intervention. It can be seen that the screen-positive rate varied by area (north-east, 27.4%; north-west, 19.4%; London, 14.6%; and Kent, 24.2%). Table 2 also shows the trial arm to which each young person was randomised. All 210 young people randomised to intervention received the brief intervention. The numbers screening positive and assenting to the study broken down by area and school are shown in Appendix 7. Table 3 summarises the numbers of young people ineligible, withdrawn and lost to follow-up.

Appendices 8 and 9 show the distribution of the baseline characteristics (categorical and questionnaire responses) of three groups of young people: those screening positive and consenting to the trial, those screening positive and not assenting and those who screened negative. There are similar baseline characteristics between the two subgroups who screened positive, but those who screened negative differed from the screen-positive subgroups in terms of smoking, drinking and sexual behaviour.

Follow-up data at 12 months were collected on 374 young people (84.4% of those recruited to the trial). A few young people did not complete the TLFB, but that information was available for 368 participants (83.0% of those recruited). The numbers providing data on the TLFB and other questionnaires at 12 months in each area are shown in Table 4. Numbers by area and by school can be found in Appendix 10.

Baseline characteristics, by trial arm

Descriptive statistics for all measures reported in the baseline questionnaires are summarised by trial arm in Tables 5 and 6: the distributions were balanced between the intervention and control arms. Overall, 46% of young people randomised admitted ever smoking a whole cigarette, the majority of whom had begun smoking between the ages of 11 and 14 years. However, only 29% stated that they had smoked during the previous 30 days. Fifty-six per cent of young people who responded said that they had never engaged in sexual intercourse. Of those who had, 21% said that they had engaged in intercourse that they had regretted the next day, and 24% said that they had engaged in intercourse without a condom. The median AUDIT score was 6, with a range of 0–40 (the percentages of hazardous and dependent drinking are shown in Table 8). The RAPI score, concerning alcohol-related problems, had a median of 4 and a range of 0–48. The WEMWBS score, measuring psychological health, had a median of 46 and a range of 14–70 (see Tables 5 and 6).

Twelve-month outcomes, by trial arm

The four outcome variables derived from the 28-day TLFB are reported in Table 7: this gives the descriptive statistics per trial arm and the unadjusted and adjusted difference in medians with 95% CIs. The distribution of the primary outcome by trial arm and region is shown in Figure 3.

FIGURE 3.

Alcohol units consumed in past. (a) Control, London and Kent; (b) control, north-east; (c) control, north-west; (d) intervention, London and Kent; (e) intervention, north-east; and (f) intervention, north-west.

It can be seen that a substantial minority of young people reported consuming no units of alcohol in the previous month (intervention arm, n = 38, 21%; control arm, n = 55, 28%), particularly in the north-west schools. However, a few young people reported consuming a very large number of alcohol units.

The median values of the primary outcome (total units consumed in previous 28 days) were 7.3 in the intervention arm and 7.7 in the control arm. The results of the quantile regression gave an adjusted difference (intervention – control) in median total units of 0.8 (95% CI –2.5 to 4.0), indicating that there was no statistically significant difference in the alcohol consumed by young people, whether or not they received the intervention. It can be seen in Table 7 that the distributions are very similar between trial arms for all four variables derived from the TLFB: the adjusted differences in medians between trial arms were all close to 0. This shows that there were no differences in other aspects of alcohol consumption between young people who received the intervention and those who did not.

The null finding requires further exploration to explore whether it is indicative of being valid, insensitive or invalid. To explore this, we calculated the Bayes factor using the methods outlined by Dienes et al. 112 To estimate the Bayes factor, we estimated the effect size difference between the groups on the primary outcome measure. As the median difference is equivalent to the mean difference for distributions that are similar but not necessarily symmetrical, we tested this assumption by examining the coefficient plots for both quantile and ordinary least squares regression. After confirmation we established the effect size difference and associated CIs using 1000 bootstrap replications stratified by school, 0.0004 (standard error 0.013). We used a recent systematic review of brief interventions for adolescent alcohol users57 and this estimated that the effect size difference on alcohol consumed was of the order of 0.27 (standard error 0.055), similar in magnitude to that proposed in the sample size calculation. The resulting Bayes factor was 0.30, and is strong evidence in support of the null hypothesis, that there is no difference between the groups in alcohol consumed.

Other secondary outcomes were those from questionnaires (AUDIT, AUDIT-C, DMQ-R, RAPI and WEMWBS). Table 7 shows the distribution of questionnaire scores by trial arm at 12 months. It can again be seen that the distributions were very similar by trial arm: the adjusted differences in medians between trial arms were close to 0, with little evidence that the results were consistent with a large difference on any scale.

Those young people who reported risky drinking as defined by AUDIT score are described in Table 8 using cut-off points of 4 (indicating hazardous drinking) and 8 (indicating dependency). Table 8 shows the numbers in each trial arm who reported hazardous and dependent drinking at baseline and 12 months, and the results of the logistic regression compare the odds of reducing levels from hazardous to non-hazardous over 12 months between arms (for those who were hazardous at baseline), and similarly for dependent levels. The adjusted odds ratio comparing those who became non-hazardous between intervention and control arms was 0.91 (95% CI 0.54 to 1.56), indicating little association between trial arm and becoming non-hazardous. That comparing the odds of becoming non-dependent was 0.55 (95% CI 0.29 to 1.04): the odds ratio was further away from 1 in this analysis, but was based on a smaller sample size. For both cut-off points, the number of young people in these categories had reduced at 12 months, although there is little evidence of any difference between control and intervention arms.

Changes in the A-SAQ score from baseline to 12 months reflected the changes in how much and how regularly young people drank alcohol. At 12 months, 60% of young people in the intervention arm and 59% in the control arm reported that they had reduced their drinking. Only a small proportion of young people said that they were drinking more at 12 months. These changes are shown in Table 9.

The changes in smoking status from baseline to 12 months are shown in Table 10. Twenty-three (12.8%) young people had started smoking during this period in the intervention arm, compared with 28 (14.4%) in the control arm. Similar numbers had reported smoking either more or fewer cigarettes over time.

The changes in risky sexual behaviour are reported in Tables 11 and 12. The number of young people who had engaged in sexual intercourse that they regretted over the 12-month period was 11 in the intervention arm and 12 in the control arm. The number of young people who had engaged in sexual intercourse without a condom over the 12-month period was 12 in the intervention arm and 10 in the control arm. However, data were missing at 12 months for a substantial minority, and there were some inconsistent results (e.g. reporting that they had engaged in risky sexual behaviour at baseline, but that they had never engaged in sexual intercourse at 12 months). The questions were asked as ‘have you ever?’ and so young people reporting risky sexual behaviour at baseline and at 12 months could have been referring to the same event. Any results need to be interpreted with some caution because of these issues.

The changes in consumption of energy drinks from baseline to 12 months are shown in Table 13. Ninety-two (51.1%) young people in the intervention arm either had stopped consuming such drinks or reported consuming fewer during the period; the corresponding number in the control arm was 82 (42.0%).

Exploratory analyses

Exploratory analyses around units consumed by gender and alcohol consumed in relation to free school meals are shown in Figures 4 and 5.

FIGURE 4.

Alcohol units consumed, by gender.

FIGURE 5.

Alcohol units vs. free school meals.

Young people were categorised by level of drinking according to the A-SAQ at baseline. Figure 6 illustrates the total units consumed in the previous 28 days at 12 months for these groups by trial arm. Those who had been drinking more heavily at baseline were also the heaviest drinkers at 12 months, regardless of whether they were in the control or the intervention group. There was no indication that the intervention was more effective in any baseline drinking level subgroups.

FIGURE 6.

Levels of drinking, by trial arm.

Qualitative analysis plan

A complete qualitative analysis plan, providing full details of all qualitative analyses, was finalised and signed before the final analyses were undertaken.

Learning mentor and teacher interviews

Interviews were conducted with learning mentors who delivered the intervention in the school setting, and with teaching staff who approved the study in their school. The aim of the interviews with learning mentors and teachers was to understand the mechanisms and processes of implementing the intervention and also how it could be embedded into the work of staff.

Recruitment

All learning mentors who delivered the control or intervention materials to young people in their school (n = 80) and all teaching staff who were involved in facilitating and arranging the ASBIs within schools (n = 30) were invited to take part in a one-to-one interview with their local trial co-ordinator to share their experiences of delivering ASBIs. Invitation letters (see documents on project page) and information sheets (see documents on project page) were e-mailed to all eligible participants, who were asked to complete a pro forma (see documents on project page) indicating their age, ethnicity, role and length of time in current role, if they were interested in taking part in an interview. 2 We aimed to recruit a sample of 24 participants to these interviews: 12 learning mentors and 12 teachers. A sampling framework was used to identify learning mentors and teaching staff for the interviews based on their answers on the pro forma. The populated sampling framework, aiming for sample variation, was based on gender, job role, ethnicity and time in job. It can be seen in Table 14.

Data collection

A semistructured interview guide was prepared in advance of the interviews, informed by NPT (see documents on project page). 113 Several other questions were also included to enhance understanding of the school context and priorities in which an ASBI may be implemented. Questions included ‘what are the main issues your school is concerned with at the minute?’ and ‘how does your school normally deal with issues around alcohol use by young people?’ Interviews were conducted on school premises by one of the four trial research co-ordinators in their area (GM, NH, SB and JoF). Informed written consent (see documents on project page) was obtained from each participant immediately before the interview.

Analysis

All interviews were audio-recorded and transcribed verbatim, with any identifying information removed from the transcripts. The interviews were then analysed using a framework analysis technique. Open coding was employed, with the codes grouped into themes and subthemes. Then a coding framework based on the principles of NPT113 was applied (GM, SB and AR). A proportion of transcripts were second coded by an independent researcher (SS and GM), with disagreements resolved through discussion. Following second coding, the framework was further refined before the research team agreed the final themes.

Participants

A total of 30 interviews with learning mentors and school-based staff who took part in the trial were conducted to reach data saturation: 21 learning mentors and nine teachers (one interview involved two learning mentors).

Interviews were conducted in 19 secondary schools across England and, on average, lasted 39 minutes (range 12–102 minutes). Participant characteristics for the learning mentor and teaching staff interviews are given in Table 15.

Interviews

Five themes were identified: (1) learning mentors’ understanding of alcohol use by young people, and of their role in delivering ASBIs; (2) initiating and sustaining ASBIs; (3) factors influencing successful trial delivery; (4) the impact on staff and young people; and (5) embedding the intervention into routine practice. These themes are explained in more detail below. Table 16 illustrates how these themes map on to the core components of NPT. 113

Methods

Participants

A total of 38 young people (24 girls and 14 boys) initially agreed to be interviewed, with 33 interviews subsequently conducted with young people involved in the trial (20 girls and 13 boys). The remaining five interviews could not be completed as the young person had been withdrawn from the trial (n = 1), had moved school (n = 1) or did not attend school on the day it was arranged (n = 3). Interviews were conducted in 17 out of the 30 study secondary schools and lasted for an average of 20 minutes.

Three parents completed the online pro forma indicating that they were interested in taking part in an interview. Two interviews were conducted (both with women), one in the north-east and one in Kent. The remaining participant could not be contacted to arrange a suitable time for an interview. Interviews lasted an average of 38 minutes. Participant characteristics for the young people and parent interviews are given in Table 19.

Interviews with young people

Following analysis of the young people’s interviews, three themes were identified: (1) drinking identities and awareness of risk, (2) access to support and advice in relation to alcohol use and (3) appraisal of the intervention and potential impact on alcohol use. These themes are now outlined with illustrative quotations from participants.

Interviews with parents

Following analysis of the parent interviews, two themes were identified: (1) parents’ thoughts on alcohol use by young people and methods to reduce harm and (2) parents’ thoughts on school-based interventions for alcohol use.

Health economics analysis plan

A complete health economics analysis plan, providing full details of all analyses, was finalised and signed before the final analyses were undertaken.

Within-trial cost-utility

The objective of the within-trial cost–utility analysis was to estimate and compare the costs and effectiveness of the ASBI against usual practice. A cost–utility analysis measures effectiveness in terms of QALYs, and estimates incremental costs with respect to usual practice, incremental QALYs and incremental costs per QALY. The incremental costs per QALY, also called the incremental cost-effectiveness ratio, provide an estimate of the additional cost associated with achieving one more QALY. The perspective for the analyses was the UK public sector (NHS, and educational, social and criminal services).

The estimation of QALYs gained/lost relies on valuations for health states, for which a score of 1 is assigned to full health and 0 to being dead. This in turn relies on measures of the quality of life of study participants. 3 The SIPS JR-HIGH trial used measures of quality of life derived from responses to an EQ-5D-3L scale,115 which is used extensively in economic evaluations. 3 The difference in costs between the intervention and the control participants was based on the cost of delivering the intervention versus current practice plus the difference at follow-up in the young participants’ use of public sector services (e.g. GP appointments, school nurse visits, accident and emergency visits). Given the time horizon of 12 months, no discounting was applied.

Cost–consequences analyses

The cost–consequence analysis presents the cost and effects data of the intervention in the form of a balance sheet. In the balance sheet, a series of comparisons of effects and of costs are presented as pros and cons for the intervention against the control group, that is, information is presented as disaggregated and no attempt is made to formally combine effects with intervention costs as is done in a cost-effectiveness analysis, in which results are summarised in the incremental cost-effectiveness ratio. The approach can potentially capture wider effects than those captured by measures of cost or quality of life. The principle behind a balance sheet is that the analyst should seek to capture all costs and benefits no matter on whom they may fall, the same principles that underpin a cost–benefit analysis. 116

For the cost–consequences analysis, the difference in costs was based on the estimated difference in costs from the within-trial cost–utility analysis. Among the effects under consideration were the levels of service use (primary health care, secondary health care, school, criminal justice system, etc.) listed above. In addition, the effectiveness measures were expanded to include secondary outcome measures of the trial, which included participants’ reported percentage of days of abstinence over the previous 28 days, drinks per drinking day and days drinking more than two units from the 28-day alcohol TLFB, measures of risky drinking (AUDIT) and alcohol-related behavioural problems.

Quality of life

Quality of life was measured using participants’ responses to an EQ-5D-3L scale,115 which has been validated in young people aged ≥ 14 years. 117 The scale was included alongside other trial outcome measures in the baseline and follow-up questionnaires. The scale contains five dimensions: mobility, self-care, doing usual activities, having pain or discomfort, and feeling anxious/depressed. Each of these dimensions has three possible levels. The permutation of the responses to the different dimensions results in 243 health states. Two further states (unconscious and death) are included, producing a set of 245 possible EQ-5D-3L health states.

The different EQ-5D-3L scores from the young participants’ answers were transformed into the corresponding health states, and these were assigned tariff values or utilities associated with each health state. Tariff values were sourced from UK population tariffs for adults. This resulted in a ‘health state utility’ at each time point for each young person.

Quality-adjusted life-years were estimated using the area under the curve method, calculated for each participant as the arithmetic average of the follow-up and baseline tariff-scaled EQ-5D-3L scores.

Cost of the use of primary care, secondary care and other services

The list of different services used by the young people in the intervention and the control group during the trial was informed by the pilot study. 1 For each participant, their use in the previous 6 months of health, educational, criminal and social care services was recorded using the Short Service Use Questionnaire instrument, administered at baseline and again at follow-up. 91 The answers were captured as the number of appointments, visits, etc., and covered GP visits, social worker visits, school nurse visits, emergency department attendances, hospital admissions and arrests.

In addition to service use, the questionnaire completed by the young people contained a question on missed days (or part of a day) of school or work in the previous 6 months. Responses were collected using a four-point scale with values at 0, 1 or 2 times, 3 to 5 times and > 5 times. To estimate the average number of missed days from school or work, a representative value was used for the second and third categories (1 and 4), with the last category coded as 5.

Each contact with services was matched to a unit cost for that service. Table 20 summarises the unit costs of resource use. The costs of primary health care contacts were the cost per visit reported in the Unit Costs of Health and Social Care 2015118 prepared by the Personal Social Services Research Unit and the Department of Health and Social Care and the Department for Education. Hospital attendances, visits to the emergency department and school nurse visits were costed based on NHS reference costs. These are based on the most recent data available, currently the National Schedule of Reference Costs for 2015–16. 128

The cost of days of school absenteeism/truancy was taken to be negligible (zero) for school absenteeism of more than four times in the previous 6 months. School absenteeism is expected to influence individuals’ and social returns on education through the impact that school attendance has on gains in cognitive capacity, and the forgone income premiums for higher numeracy and literacy levels. Research suggests, however, that the return to investment penalty from absenteeism is less important in the secondary school years than in the primary school years, and that the most detrimental effects on student performance are from recurrent absenteeism (e.g. missing 2 days per week) rather than from sporadic absenteeism. 125,129 Young people with frequent truancy achieve lower scores on numeracy and literacy tests and lag behind their peers. 126 A large literature has explored the impact of poorer school scores on earnings, and estimates are that children with lower numeracy and literacy could have annual salaries that are between 12% and 20% lower than those of their average-level peers. 122,123 Therefore, for children who reported absenteeism levels of ≥ 5 days in 6 months, the cost of absenteeism was estimated as the present value of the lost working lifetime income using the average yearly earnings during 50 years of working life between the ages of 16 and 66 years and of 23 and 73 years. 130

The cost per young person of service use for each type of service was calculated as the product of the young person’s level of service use multiplied by the appropriate unit cost (see Appendix 11). For each young person, the total service use cost was estimated as the sum across all service categories.

Cost of the intervention

The resources assessed for the delivery of the intervention included only those that would be needed if the intervention were to be provided in practice in the future. 131 Therefore, the resources necessary to develop the intervention materials as well as the resources used in the scientific evaluation of the intervention would not normally be included. Similarly, resources common to both the intervention and the usual practice (control arm) would not normally be included, as these are not relevant for the comparison between the two programmes. 132

As a brief reminder, screening for alcohol problems and the short intervention took place in schools using learning mentors. During the intervention, participants received structured advice/counselling of short duration (anticipated to be, on average, 30 minutes) aimed at reducing alcohol consumption or decreasing problems associated with alcohol, based on theoretical principles of behaviour as a dynamic interaction between the individual, behaviour and environment. 133

The following costs of the intervention were included:

• Costs of the resources needed for screening the young people for high alcohol drinking levels (printing and postage of information and consent letters to parents informing them of the screening in their children’s school, printing and postage of screening questionnaires).

• Costs of preparing school learning mentors for interviewing the young people according to the principles of the intervention material (time invested by expert trainers and by the learning mentors).

• Costs of the interview with the young person (learning mentor time spent preparing the intervention, learning mentor time delivering the intervention and cost of printing of materials for the intervention).

The information on trainer time and cost of printed materials was sourced from the records of the research team. Learning mentors’ time preparing the interview with the young person, and their time with the young person, were captured using the learning mentors’ case diaries. Learning mentors reported the time for preparing the intervention in six categories (0–5 minutes, 6–10 minutes, 11–20 minutes, etc.) with an open-ended question for times exceeding 45 minutes, and time delivering the intervention in intervals of 10 minutes (0–10 minutes, 11–20 minutes, etc.) with an open-ended question for times exceeding 60 minutes. Data on the length of time were transformed from ranges to integer values using the mid-point of the scales (see Table 20).

The cost of learning mentor time was based on published learning mentor gross annual pay statistics for the Local Government Earnings Survey 2013/14 (latest available), adjusted upwards by 1% using the average of public sector salary growth year-on-year from April 2013 to April 2017. 134 The cost of the trainers’ time was based on the pay scale of the 2015/16 (latest available) nationally agreed single pay spine for higher education institutions in the UK, published by the national University and College Union. 135 Costs of time were calculated as average per annum salary plus on-costs (employer contribution to National Insurance and superannuation plans) and training costs, converted to £/minute. Appendix 12 presents details of this calculation. The cost per intervention group participant of the time of the learning mentors in preparing and delivering the intervention was estimated as the product of the salary cost per minute multiplied by the duration in minutes of the preparation time and the interview for each young person.

Total cost

The total costs for each young person in the intervention group for the delivery of the intervention were estimated as the sum of the fixed costs of screening and training plus the individual-level interview costs plus the individual-level costs at follow-up of primary care, secondary care and other services.

Sociodemographic characteristics

Sociodemographic characteristics of participants included gender, ethnicity and the geographic location of the study site. The four study sites were located in England and were grouped into ‘north’ (north-west, north-east) and ‘south’ (Kent, London).

Statistical and econometric analyses

Exploratory analyses of the differences between the intervention and control groups in quality-of-life scale (EQ-5D-3L), service use and intervention delivery time and costs were performed to explore trends in the data in terms of missing and extreme values.

Differences in QALYs and costs at follow-up between the intervention and the control group were estimated using a seemingly unrelated regression adjusting for participant characteristics and baseline values. Seemingly unrelated regression estimation was used to allow for the fact that costs and QALYs may be correlated within individuals. 136,137 Adjustment was made to control for the overall variability in costs and QALYs from possible imbalances in baseline characteristics of participants that could have existed even in the presence of randomised assignment to control and intervention groups. Costs were adjusted by baseline service use costs and QALYs were adjusted by baseline EQ-5D-3L scores. Participant characteristics included in both the cost equation and the QALY equation were baseline AUDIT score, baseline score for emotional well-being (WEMWBS), gender, geographical site and an indicator variable identifying if time to follow-up interview was shorter or longer than the target 12-month period (> ± 30 days). The difference in costs was multiplied by two to obtain annual costs given the 6-month recall period of the questionnaires. CIs of the difference between intervention and control groups in costs and QALYs were estimated using non-parametric bootstrap with 1500 replications stratified by gender, intervention group and geographical location. The number of replicates was chosen to achieve a better approximation of the distribution of estimated difference. 137 Histograms of the distribution of the bootstrap estimates were used to decide on the use of bias-corrected or normal-based CIs. The incremental cost-effectiveness ratio was estimated as the ratio between incremental costs and benefits using the results from the between-group difference in costs and difference in QALYs derived from the seemingly unrelated regression.

The cost-effectiveness acceptability curve (CEAC) was estimated using the adjusted (as earlier) incremental costs and incremental QALY values from a seemingly unrelated regression with 1500 bootstrap replicates to calculate the net monetary benefit. The CEAC was produced by estimating the proportion across the bootstrap replicates of positive net monetary benefit values, for different values of the threshold willingness to pay for an additional QALY (see Appendix 13). In addition, the joint distribution of incremental costs and incremental QALYs was explored using a scatterplot of results for each bootstrap iteration of costs and QALYs.

Sample and data characteristics

A total of 443 responses from young participants at baseline were available for analysis. The majority of participants in the sample were from the north-east (55%), followed by Kent (31%), the north-west (13%) and London (1%). There was an almost equal distribution of boys and girls in the sample, and approximately 91% of participants self-reported their ethnicity as ‘white’. Eighty-four per cent of randomised participants who completed a baseline survey also provided a response for the final trial outcome measure at follow-up (the 28-day TLFB).

Tables 21 and 22 present information on the average time spent by learning mentors in preparing and delivering the interviews with the young people. The median time delivering the interview in the intervention group was 16 minutes, with the lowest value for interview delivery at 5 minutes. There were few missing data (< 3%).

The distribution of responses to the five dimensions of the EQ-5D-3L are presented in Appendix 14. For each dimension, level 1 represents a better state of health. Most participants in the intervention and control groups reported good levels of mobility, self-care and the ability to perform usual activities. In contrast, scores tended to be lower, indicative of poorer quality of life, for pain/discomfort and anxiety and depression. These trends were observed at both baseline and follow-up. There were slightly more participants reporting better quality of life in each dimension at follow-up across groups. At the same time, there was a larger proportion of missing values in the follow-up questionnaire (13–15%) than at baseline (5–7%), but there was no observable systematic difference between the groups, suggesting a low likelihood of bias arising from the pattern of non-responses.

As would be expected from these types of data, all categories of service use had a distribution of values that was highly skewed to the right (Table 23), reflecting the fact that many participants reported zero service use in the previous 6 months, and a small number of young people reported high levels of GP appointments, nurse visits, etc.

The highest reported level of service use in the previous 6 months was for GP visits, on average, 1.4 [standard deviation (SD) = 1.9] visits in the control group and 1.1 (SD = 2) visits in the intervention group at follow-up, down from 1.7 (SD = 7.1) and 1.3 (1.8) visits at baseline, respectively. The lowest values were observed for the number of times being arrested. On average, service use decreased across all categories from baseline to follow-up for the intervention and control groups, but the intervention group had higher average values at follow-up of social worker visits, school nurse visits and accident and emergency attendances, whereas the control group had higher average numbers of GP visits and missed school days.

The proportion of the highest reported value (i.e. the proportion of very high values among the responses) tended to be low (< 3%) for the different services; these values were not implausible and, therefore, were included in the main analyses. As with the EQ-5D-3L responses, the proportion of missing values was higher at follow-up than at baseline. The percentage of missing values at follow-up ranged from 18.5% to 13.8%, with no large differences between the intervention and control groups.

Primary and secondary effectiveness outcomes from the trial

The primary and secondary outcomes from the trial are presented in Chapter 3. Very briefly, there were no differences in median values for drinking outcomes measured by the 28-day TLFB between the intervention and control group at follow-up after adjusting for participant characteristics. Similarly, there were no differences in responses to the alcohol problem and well-being scales. Reports of smoking, sexual behaviour and consumption of energy drinks were comparable between the intervention and control groups.

Intervention costs

The total costs per participant for the materials needed for screening and to deliver the intervention, as well as the cost of training the learning mentors, are presented in Appendix 15. The cost of interview materials for the brief intervention per participant in the intervention group was estimated at £1.58, giving a total of £332 for materials for the 210 young people in this group. Screening and training activities in the intervention group were estimated at £14.20 per participant for 210 young people. This included £154 for the screening questionnaires, £1870 for the consent letters and £2875 for the cost of training the 80 learning mentors involved in the study. The cost of the learning mentors was spread over 3 years by dividing this cost by three times the annual number of young people covered by the programme.

The cost of the learning mentor time in preparing for the interview with the young person was, on average, £1.90 and the cost of learning mentors’ time conducting the interview with the young participant in the intervention group was, on average, £4.54 (Table 24).

Considering the cost of the learning mentors’ time and the costs of the screening, training and intervention materials, the total cost per intervention participant of the intervention was, on average, £22.20 (95% CI £21.83 to £22.57).

Longer-term costs

Health-care and other service use costs at baseline and follow-up are presented in Table 25. The highest costs of health-care use were for hospital admissions, followed by GP visits. The highest social costs that participants faced from risky alcohol drinking levels were for missed school days. These costs represent missed lifetime earnings from underperforming in school as a result of missing classes and were therefore expected to have a higher mean than other social and health-care costs. Costs were lower on average for the intervention group at follow-up for GP visits, arrests and number of missed school days; however, except for arrest costs, the large CIs surrounding these differences indicate the possibility of both decreases and increases in costs of an important economic magnitude, particularly for the cost of missed school days.

Cost–utility analysis

After adjustment for baseline values and patient characteristics, the mean total costs per participant (the service use costs and the costs of delivering the intervention) in the intervention group at the end of follow-up were £6212 compared with £9077 in the control group, representing a difference of £2865 (95% CI –£11,272 to £2707). The intervention group had, on average, lower QALYs at 12 months than the control group (0.363 and 0.367, respectively), resulting in a difference of –0.004 (95% CI –0.019 to 0.011) between the groups. However, both costs and QALY differences were imprecise, with the 95% CIs including ‘no difference’, but also covering important economic values for costs (the CIs for QALYs might not contain economically important differences, but this is itself unclear). Given these data, an incremental cost per QALY can be calculated only to compare the more costly and more effective control with the less costly and less effective intervention. In this case, the estimated incremental cost per QALY was approximately £723,000 per additional QALY for the control compared with the intervention (Table 26). This is well beyond the £20,000 that the National Institute for Health and Care Excellence typically judges acceptable in its recommendations about the provision of services and treatments. 138 This suggests that the control is, on average, not cost-effective compared with the intervention. Therefore, by implication, the intervention is, on average, cost-effective at a threshold value for society’s willingness to pay for a QALY of £20,000.

Figure 7 shows the 1500 bootstrap replicates, stratified by trial group, geographic location and participant sex, of the estimates from a seemingly unrelated regression of the difference between the intervention and control groups, adjusting for baseline resource use costs, baseline EQ-5D-3L score and participant characteristics. Out of 1500 replications, in 54% of observations the intervention was less costly and less effective, in 22% of observations it was less costly and more effective, in 18% of observations it was more costly and less effective, and in 6% the intervention was more costly and more effective. As the figure shows, the majority of the bootstrap replications fall below the diagonal lines, which represent the National Institute for Health and Care Excellence’s cost-per-QALY threshold reference values of £20,000 and £30,000. This indicates that, for the majority of iterations, the intervention would be considered cost-effective.

FIGURE 7.

Incremental cost and QALY plot.

The CEAC (Figure 8) suggests a 76% probability that the brief alcohol intervention is cost saving compared with usual practice. At the same time, at values of £20,000 and £30,000 of willingness-to-pay thresholds for an additional QALY gained, the probability that the intervention could be cost-effective compared with usual practice is 73%. These results were driven by the likelihood that the brief alcohol intervention was cost saving as the mean difference in QALYs was so small.

FIGURE 8.

Cost-effectiveness curve.

The curve reflects the probability that the intervention will be cost-effective at a given value of society’s willingness to pay for an additional QALY gained. At a willingness to pay equal to zero, the curve shows the probability that the intervention will be cost saving.

Sensitivity analysis of cost assumptions

The values obtained for the differences in costs between the intervention and the control groups in a trial are, crucially, determined by the unit costs, the distribution of costs and the assumptions made about the relevant costs to include in the analyses. Two sensitivity analyses were performed to account for these factors. The first explored the potential impact on costs of the very large values of service use, in particular GP visits, social worker visits and nurse visits. The large values were truncated at the 95th percentile of the distribution of each variable. The second analysis excluded the costs associated with missing school days, as these costs are based on future lifetime income lost and can be considered a type of indirect cost, as opposed to the direct realisable costs of service use from, for example, being arrested or going to hospital (Table 27).

The results were relatively unchanged after truncating the extreme values of GP visits, social worker visits and nurse visits (see Table 26). When excluding the missed school days costs, the mean net savings from the intervention were approximately halved. This resulted in the incremental cost per QALY gained falling to £334,537 compared with £723,048 in the base case analysis (note that this incremental cost-effectiveness ratio is for the more effective but more costly control compared with the intervention). The CEAC obtained when the cost of missed school days was excluded gave a 77% probability of cost saving for the brief intervention, and the probability that the intervention is cost-effective given threshold willingness-to-pay values of £20,000 and £30,000 of 74%. The scatterplot of the joint distribution of incremental costs and QALYs is presented in Appendix 16. Of the 1500 replicates of incremental costs and QALYs, in 54% of observations the intervention was less costly and less effective, in 23% of observations the intervention was less costly and more effective, in 18% of observations the intervention was more costly and less effective, and in 6% the intervention was more costly and more effective.

Multiple imputation

Multiple imputation was used to reduce the proportion of missing values in the total cost at follow-up and in the EQ-5D-3L scores at follow-up. In total, 103 observations, 33% of the number of young people who consented to the study, had missing values of total costs, follow-up EQ-5D-3L or both. Details of the multiple imputation strategy can be found in Appendix 17.

After multiple imputation, the results shown in Table 28 did not differ markedly from those of the main analyses reported in Table 26. The CEAC and the incremental cost/incremental QALY plot (results not shown) were also similar to the plots from the main analyses.

Cost–consequences analysis

The cost–consequences analysis for the SIPS JR-HIGH trial is presented as a balance sheet comparing costs that favour current practice and outcomes that favour the intervention (Table 29). This analysis suggests a very mixed picture, with no clear signal from the other outcomes as to why the intervention has an 80% chance of being cost saving.

Long-term model analysis

The possibility of developing a long-term model of the cost and effectiveness outcomes of the trial was incorporated into the ex ante economic evaluation analysis plan prepared for this study. This part of the economic evaluation was dependent on the findings from the within-trial evaluation, that is, evidence of effects on costs, economic outcomes (i.e. QALYs) or trial outcomes. The trial did not show any effect on young people’s alcohol intake, with this result unlikely to be different when extrapolating figures to the longer term. 139 Similarly, in terms of the economic outcomes, the results failed to demonstrate statistically significant differences in effects, and data extrapolation from these figures is unlikely to change the main conclusions from the evaluation.

Chief investigator

Dr Dorothy Newbury-Birch, Health and Social Care Institute, Teesside University.

Aims and objectives

The TMG group has the primary aim of ensuring appropriate, effective and timely implementation of the SIPS JR-HIGH trial.

The TMG will strive to achieve this aim by fulfilling the following objectives to:

• identify appropriate sites for conducting the SIPS JR-HIGH trial

• participate in the development and compilation of data collection instruments and other relevant research and intervention manuals

• determine tasks, schedules and deliverables of the SIPS JR-HIGH trial

• determine the fidelity of trial interventions

• produce a working protocol for the trial and ensure adherence to the protocol

• develop a publication protocol

• facilitate and support data analysis

• determine tasks, schedules and deliverables for report writing and publication of findings

• ensure that adequate supervision/support occurs for research staff.

Membership

Eilish Gilvarry (chairperson), Dorothy Newbury-Birch (chief investigator), Eileen Kaner (Newcastle lead), Emma Giles (project manager), Simon Coulton, Elaine McColl, Denise Howel, Elaine Stamp, Les Tate, Colin Drummond, Paolo Deluca, Ruth McGovern, Stephanie Scott, Harry Sumnall, Liz Todd, Luke Vale, Sadie Boniface, Jo Frankham, Grant McGeechan, Victoria McGowan, Jayne Ogilvie, Grant Stanley, Claire McDonald and Nicola Howe.

Membership of the group will be reviewed as appropriate and as required.

Meeting

The TMG will meet once a month or more often if needed. Members are able to join the meeting by teleconferencing. A meeting will be considered quorate when at least three members are in attendance.

Reporting

The group will report to the TSC.

Duration

The group will function for the entire duration of the SIPS JR-HIGH trial.

Chief investigator

Professor Dorothy Newbury-Birch, Health and Social Care Institute, Teesside University.

Project manager

Dr Emma Giles, Health and Social Care Institute, Teesside University.

Aims and objectives

The TSC has the primary aims of monitoring implementation of the SIPS JR-HIGH trial and providing an independent assessment of the data analysis and reporting.

The TSC has the following objectives to:

• provide an oversight role for the trial including providing advice and comments to the trial sponsor and funder and ensure it is conducted to rigorous standards

• comment on the progress of the trial and adherence to protocol

• consider new information of relevance to the research question

• provide advice, through the chairperson, to the chief investigator and trial funder on all appropriate aspects of the trial

• provide evidence to support any requests for extensions.

Meeting

The TSC will meet biannually. Members are able to join the meeting by teleconferencing. A meeting will be considered quorate when at least three members are in attendance. Dorothy Newbury-Birch and Emma Giles will be responsible for calling, organising and minuting the meeting.

Duration

The group will function for the entire duration of the SIPS JR-HIGH trial.

Membership

Current membership is shown in Table 30.

Two young people from the study will be invited to take part in the steering group.

Study team members

Professor Dorothy Newbury-Birch (XXXX).

Dr Emma Giles (XXXX).

Ms Denise Howel (XXXX).

Membership of the group will be reviewed as appropriate and as required.

Chief investigator

Professor Dorothy Newbury-Birch, Health and Social Care Institute, Teesside University.

Project manager

Dr Emma Giles, Health and Social Care Institute, Teesside University.

Aims and objectives

The DMEC has the primary aim of monitoring the study data and make recommendations to the TSC on whether there are any ethical or safety reasons why the trial should not continue.

The DMEC has the following objectives to:

• consider any ethical or safety reasons as to why the trial should not continue

• consider the need for any interim analysis advising the TSC regarding the release of data and/or information

• consider safety, quality and compliance data

• consider the chain of custody and storage of trial data

• provide advice, through the chairperson, to the chief investigator and trial funder on all appropriate aspects of the trial

• provide evidence to support any requests for extensions.

Meeting

The DMEC will meet annually. Members are able to join the meeting by teleconferencing. A meeting will be considered quorate when at least two members are in attendance. Dorothy Newbury-Birch and Emma Giles will be responsible for calling, organising and minuting the meeting.

Duration

The group will function for the entire duration of the SIPS JR-HIGH trial.

Membership

Current membership is shown in Table 31.

Membership of the group will be reviewed as appropriate and as required.