Pilot trial results

Data regarding the number of participants consented, recruited and randomised to the study are presented. In addition, we present data on baseline characteristics and primary and secondary outcome measures. Information on data completeness across the three time points is provided. Numbers receiving the intervention and numbers followed up are also shown.

Overall study recruitment

Both study sites agreed to take part with the necessary organisation approvals and ethics approvals were secured. Recruitment opened at the England site on 2 December 2019 and closed on 28 February 2020, as the required sample size of 90 participants had been met. The Scotland site opened to recruitment on 15 January 2020 and was halted due to COVID-19 restrictions on 13 March 2020, with 42 individuals recruited to the study. Following discussions with the site governor and the head of research at the Scottish Prison Service, HMP Scotland was closed to any further recruitment on 12 June 2020. Differences in start date between study sites were due to difficulties caused by the model non-commercial agreement (mNCA) requested by the Health Research Authority (see Ethics approval).

Table 7 provides an overview of the screening and recruitment data at the two sites.

Recruitment methods varied between the two sites as the prison profiles, regimes and access were different. As such, the screening and recruitment data for the two sites are considered in more detail separately.

England study site recruitment

Table 8 summarises the screening and recruitment data for England by week number. Unlike the Scotland site, where all individuals on remand were given an information sheet about the trial, not all of the individuals entering HMP England received an information sheet.

The number of prisoners on remand at the start of recruitment was not provided by the England site. Reasons why individuals were excluded in England are summarised in Table 9.

Scotland study site recruitment

Table 10 contains information on the number of individuals entering HMP Scotland on remand by week of recruitment. At the start of recruitment, there were 162 prisoners on remand, and an additional 159 individuals entered the prison on remand during the data-collection period. During recruitment, 195 individuals were excluded, and 86 were not approached due to lack of resources and time.

Table 11 provides information on the number of screened individuals with 0, 1, 2, 3 or 4 reasons for being excluded from the study.

Of those who had at least one reason for being excluded at the Scotland site, there was a total of 210 exclusion reasons, and these have been summarised in Table 12. Of note is the largest exclusion reason, which was being on remand for more than 3 months. Being liberated (n = 49) and operational reasons (n = 44) were the second and third most prevalent reasons for exclusion respectively.

We had not created a standard proforma of consistent categories for exclusion, which resulted in different terminology being used for the exclusion criteria at the two study sites. We also noted that some criteria were much more frequent at one study site compared to the other (there was not much consistency between sites). Following several discussions by the PMG we agreed, for the purposes of the pilot trial and to support our understanding of these differences, that for a future trial we would continue the site-specific reasons and as such the exclusions are reported separately.

Consented and randomised

There were 134 participants who gave informed consent (90 from the England site and 44 from Scotland). A total number of 132 participants were randomised. Two participants from the Scotland site who were consented did not proceed to randomisation. One participant was considered to have lacked capacity and the other scored less than 8 on the AUDIT. Table 13 provides data on the numbers randomised to the intervention and control arms at each study site.

Randomisation recruitment rates

It was anticipated that recruitment would start on the 1 September 2019 and would last for 4–6 months, with recruitment being quicker in England compared to Scotland. We were predicting this based on our learning from the PRISM-A study.40 Variations in the speed of recruitment can partially be explained by access opportunities. Due to the different regimes within the prisons, the amount of time on the halls was less in Scotland than in England. The regimes also impacted on the amount of time the researchers were able to be on site. For example, in England the researcher was on site 5 days a week. In Scotland the researcher was on site 3 days a week.

Table 14 contains information on the number of participants randomised by month and by site, with Figure 2 providing a graphical overview of the recruitment rate by site and as a total. The planned rate of recruitment starts from 2 December in Figure 3 to correspond with when the England site was opened to recruitment.

FIGURE 2 Consolidated Standards of Reporting Trials 2010 flow diagram. (Figure not reproduced here; see the original article.)

FIGURE 3 Recruitment graph of planned recruitment and overall recruitment by study site. (Figure not reproduced here; see the original article.)

Baseline characteristics

The demographic characteristics of the randomised participants can be found in Tables 15 and 16. The average age of the participants was 34 years (median 31, SD 10), the majority of participants were single [87 (66%)] and considered themselves to be White British or White Scottish. Regarding highest educational achievement, 36% had no formal qualifications, 19% had a recognised trade apprenticeship, 39% had GCSEs or equivalent, 1% had A-levels or equivalent, while 6% were educated to university or college degree level. There was substantial variability in terms of the number of drinks consumed per month at baseline: the SD was 73 (range 5–198 drinks per month). The median number of drinks consumed per month was 80 (mean 94). The mean total score of the AUDIT-C was 9.2 (SD 2.2, range 3 to 12). A total score of < 5 on the AUDIT-C is regarded as low risk, while a score of 5 or more indicates hazardous drinking. Overall, 98% of participants were in the ‘at-risk’ category for drinking based on the AUDIT-C (≥ 5) at baseline, including 23% who recorded the maximum value of 12 on the AUDIT-C (indicating extremely high levels of drinking). Only two participants were considered to be at low risk (2%). These two participants had an AUDIT-C of 9 and were therefore eligible for the study.

On average, high scores were recorded on the readiness to change ruler at baseline except for the item ‘I have made a detailed plan regarding when, where and how to reduce my drinking’, indicating that most participants had not made a detailed plan regarding when, where and how to reduce their drinking. The EQ-5D item scores suggested that participants had moderate problems with anxiety/depression on average; all other EQ-5D item scores were low on average. The WEMWBS had a mean of 38.5 (SD 12.4), indicating probable clinical depression in most participants.

Figure 4 shows side-by-side boxplots of the ages of participants in each site.

FIGURE 4 Ages of participants at HMP England and HMP Scotland. (Figure not reproduced here; see the original article.)

In-prison and post-liberation intervention delivery

Table 17 contains information on whether those randomised to the intervention arm received the in-prison and post-liberation intervention sessions. Out of 68 participants randomised to the intervention condition, 53 received the in-prison intervention session (77.9%, 95% CI 66.2% to 87.1%). This percentage is greater than the 70% threshold specified as one of the criteria necessary for progression to a future larger randomised controlled trial (Objective 4c).

Reasons for why the in-prison intervention was not delivered are shown in Table 18.

Only one ‘day-3 post-prison intervention’ session was attempted and was delivered (duration 30 minutes) at the Scotland site. No ‘day-7 post-prison intervention’ sessions were attempted. Only one ‘day-21 post-prison intervention’ session was attempted and was delivered to one participant (duration 30 minutes) at the England site.

The main reasons why participants did not receive the post-prison interventions were (1) they could not be contacted and (2) CGL staff were unable to contact participants due to the impact of the COVID-19 pandemic.

Data for the duration (minutes) of the intervention sessions across both study sites are presented for all but one of the 53 interventions delivered in Table 19. There was a marked difference in the median time taken to deliver the intervention between the two sites: the intervention duration was half an hour in England compared to 1 hour in Scotland. The shortest time for an intervention was only 5 minutes in England, whereas in Scotland it was 30 minutes. The maximum time for delivery was more aligned across sites, being 70 minutes and 80 minutes in England and Scotland, respectively.

Follow-up and withdrawal

There were no formal withdrawals from the study. However, only 3 out of 132 randomised participants recorded any follow-up data at 6 months (2.3%, 95% CI 0.5% to 6.5%), and only 18 out of 132 randomised participants recorded any follow-up data at 12 months (13.6%, 95% CI 8.3% to 20.7%) (Table 20). The low numbers of follow-ups were due to the impact of COVID-19.

Tables 21 and 22 show the numbers of follow-ups at 6 and 12 months respectively within each site and randomised allocation group.

Out of the 44 participants receiving the intervention, only 13% were followed up to 12 months. Note that this percentage is much lower than our progression criteria (Objective 4d) specifying that at least 60% of those who received the intervention should be followed up at 12 months across trial arms and study sites. This progression criterion was set prior to the COVID-19 pandemic.

The greatest increase in numbers of those followed up at 12 months was seen at the Scotland site, where we were able to have some access to the prison site and identify where participants were located; however, this was minimal. In England where there was no access to the estate this meant at the time of intervention delivery and follow-ups we did not know who had been liberated and who had not liberated. However, this access remained very limited. Reasons for being unable to follow up at 12 months are shown in Table 23. Out of all 132 randomised participants, 53 (40%) were not liberated or released at 12 months follow-up, 47 (36%) were uncontactable and 14 (11%) could not be located even though they had been released.

Data completeness

Table 24 contains information on data completeness at baseline and follow-up for all 132 randomised participants. Complete data collection refers to when a participant has recorded data for all relevant questionnaires and outcomes. Incomplete data collection refers to when the participants have recorded data on the primary outcome, but have not recorded data for one or more secondary outcomes.

The information we have collected on data completeness helps us to address Objective 1c (‘How well do participants complete the questionnaires necessary for a future definitive RCT?’). The majority of participants did not complete questionnaires, which was at least partly due to the difficulties in following up participants because of the COVID-19 pandemic and lockdown restrictions. However, considering the small numbers of participants who did complete the questionnaires, nearly half of them (9/21) could not (or would not) complete all the questionnaires or record data in response to every question.

Outcomes

Primary outcome

The intended primary outcome measure for the pilot RCT was the TLFB-28. As outlined in Chapter 3 this was amended to AUDIT-C. Table 25 shows descriptive statistics for estimated alcohol consumption in the previous 28 days at each time point, restricted to those participants with any follow-up information. Please note that alcohol information recorded at baseline may be an underestimate because a different AUDIT questionnaire was used at this time point and the formats of the first few question options were different (e.g. ‘4–5 times a week’ drinking was not distinguished from ‘6 or more times a week’ drinking).

Table 26 shows means and 95% confidence intervals for alcohol consumption (number of drinks consumed in the past 28 days) within each site and randomised group. Note that the amount of missing data is very substantial, so it is difficult to interpret these results.

For the primary analysis, the mean difference with 95% confidence interval in alcohol consumption at 12 months post randomisation (TP2) between the two randomisation groups was calculated (Table 27), both overall and stratified by site. Linear regression was used, adjusting for baseline alcohol consumption (TP0). Note that the linear regression model for England was based on only four data points and the 95% confidence intervals are extremely wide. Due to the substantial amount of missing data, extreme caution is required when interpreting these results.

Secondary outcomes

Summary statistics of the secondary outcomes are presented in Table 28 for both sites combined. Please see supplementary material for the same tables stratified by site (Tables S1.1 and S1.2 in Report Supplementary Material 1). Although sample sizes are small, the standard deviations reported at 12 months follow-up could be used to inform sample size calculations for future studies in conjunction with the standard deviations reported in the baseline characteristics (see Table 16). This addresses our Objective 1b (‘Can we obtain reasonable estimates of the parameters necessary to inform the design and sample size calculation for a future definitive multicentre RCT?’). This also gives us some insight into Objective 2c (‘To what extent was the intervention changing process variables consistent with the underpinning theory’).

The supplementary material also contains detailed tables of individual questionnaire item scores, overall and split by site (Tables S3.1 to S3.7 in Report Supplementary Material 1).

The secondary outcome results, change from baseline, show the average scores of participants in the intervention and control group at baseline, 6 months and 12 months. Due to missing data and small sample sizes, we are not able to interpret change from baseline to 6 months, nor group differences in change for 6 to 12 months. For change from baseline to 12 months, we can look at patterns of change in the data in mechanisms of change, albeit with caution. Readiness to change does not show increasing or at least maintained values; rather it seems to slightly decrease across items. One exception is the item ‘I have made a detailed plan regarding when, where and how to reduce my drinking’, which increases from baseline to 12 months in both groups (at baseline intervention: M = 4.17, control: 3.0; at 12 months intervention: M = 6.1, control 5.1).

Logic model results

An exploratory analysis was performed to explore key behavioural markers (mediators) of change and how these may influence the effect of the delivered intervention. The six proposed mechanisms of change were: self-efficacy, outcome expectations, goals, plans, social support and high-risk situations.

Summary statistics for the mechanisms of change at each time point and within each randomised group are presented in Report Supplementary Material 1, Tables S2.1 to S2.6.

In exploratory analysis, Kendall’s tau correlation coefficients were calculated to assess the correlation between each of the mechanisms of change and the primary outcome (AUDIT-C) at 12 months. Correlation coefficients and p-values are shown in Table 29, along with valid sample size N that each correlation analysis is based on.

Mindful of the very small sample sizes and that as such caution should be applied, provisional indications suggest that self-efficacy may be a determinant of alcohol consumption and further exploration of interventions targeting self-efficacy should be considered. We observed moderate correlations between markers of self-efficacy and high-risk situations and alcohol consumption, whereas the correlations between the other mechanisms for change and alcohol consumption were weak.

Serious adverse events

There were no serious adverse events reported during the pilot RCT.