The protocol for this parallel arm feasibility RCT clustered by site in 1:1 ratio has previously been published and we summarise methods here in accordance with CONSORT guidelines.24
1. Feasibility of identification of people at high risk of fatal overdose
We attempted to identify people at high-risk of fatal opioid poisoning who may benefit from naloxone from a THN kit for inclusion in outcome comparisons in a future RCT. We defined a discriminant function and used this as a predictive tool, incorporating known and routinely recorded predictors of opioid-related events. We used existing linked data relating to opioid deaths in Wales between 1st January 2015 and 30th November 2021, including ED, in-patient and drug service data, to select predictors most closely associated with those who died from opioid poisoning. Routine data were captured from existing routine datasets:
· Welsh Demographic Service Dataset (WDSD) for age, sex, periods of residence in Wales, deprivation score and date of death
· Annual District Death Extract (ADDE) for date and cause of death
· EDDS (Emergency Department Data Set) for date and reason for emergency attendance
· Patient Episode Database for Wales (PEDW) for date and reason for hospital admissions
· Critical Care Data Set (CCDS) for date and reason for critical care admissions
· Substance Misuse Data Set (SMDS) for dates and reasons for related to substance misuse treatment
De-identified individual-level data were provisioned from the SAIL Databank, and made available for analysis within the SAIL Gateway, a Trusted Research Environment.25
We used WDSD records to define the study cohort, based on anonymised information on period of residency in Wales, and date of death; cause of death was obtained from ADDE records, and deaths classified as related to opioid overdose (or not) using the coding framework outlined in Fuller et al (2022).26 We next considered EDDS, PEDW, CCDS and SMDS attendances for the 36 months up to study end dates. We then used logistic regression to model the relation between the primary binary outcome (death from opioid overdose, or not) and potential risk factors and covariates. Our primary analysis included data on all variables, based on attendances within 12-months of study end date.
We undertook further analyses to assess the sensitivity of fitted models, and finally assessed the feasibility of using fitted logistic regression models in identifying a high-risk cohort to include in outcome comparisons.
2. Randomised feasibility trial
Study setting
This study was clustered by participating paired ED and emergency ambulance service area. Sites were randomly allocated to intervention or control arms. Intervention sites are referred to as Site 1, comprising Emergency Department 1 (ED1) and Ambulance Service 1 (AS1); Site 2: ED2 and AS2. Control sites 3 and 4 comprised ED3 with AS3, and ED4 with AS4.
Inclusion criteria
We included adult patients (aged 18 years or older) cared for by participating (TIME trained) ambulance paramedics or ED clinicians, for a presentation related to opioid use (e.g. opioid overdose or injuries due to opioid use), and who were assessed as having the capacity to consent to receive the THN kit and related training.
Patients were excluded if they were known to have previously suffered an adverse reaction to naloxone; were aggressive or exhibited other challenging behaviours; had already been recruited; were in police custody.
Consent
We did not attempt to gain consent from patients to participate in the trial at the time of attendance for an opioid-related emergency, because emergency settings contradict the prerequisites for informed consent.27 We also did not try to gather consent retrospectively, as the population was deemed difficult to reach, and low contact rates could invalidate research findings. Patients who received the intervention were consented, according to standard clinical practice and offered the option to opt-out from the study at all sites via patient information leaflets supplied with THN kits and made available at ED waiting areas. We also included this information on the Wales Centre for Primary and Emergency Care Research (PRIME) website (www.primecentre.wales). We gained the ethical, research, information governance and Confidential Advisory Group permissions to support this approach.
Clinical staff recruitment and training
Nurses and doctors at intervention site EDs and ambulance service paramedics operating within the ED catchment areas were invited to participate in the study. Volunteers were trained in delivering the intervention in accordance with the study protocol. Training, provided in a flexible manner to suit the working practices of individual departments and services, involved face-to-face group-based training, complemented by a ‘cascade’ approach whereby research support nurses and paramedics trained their peers on an ad hoc basis. Online resources produced by Martindale Pharma were available as refresher content for staff (http://www.prenoxadinjection.com/).
Sample Size
We aimed to include enough patients to test study design, methods and completeness of data. We expected to identify 200 records for individuals at high risk of overdose and thus eligible for the THN intervention in each site (100 via ED; 100 via the corresponding ambulance service) resulting in a total sample size at intervention sites of 400 participants. We did not carry out a power calculation to inform the sample size in this feasibility study as we were not aiming to determine effect sizes.
Randomisation
One member of the research team (MJ) selected two sites as intervention and two as control sites at random from the set of all possible allocations, each contained within separate sealed opaque envelopes.
Blinding
In this cluster randomised trial it was not possible to blind participants or practitioners to allocation. The study statistician remained blinded as far as possible during analysis.
Interventions
Usual care comprised supportive care and resuscitation as required plus naloxone administered by paramedics or ED clinicians in the case of overdose.
The THN intervention was offered to patients in addition to usual care and included a multi-dose THN kit (Prenoxad) containing 2mg naloxone hydrochloride 1mg/1ml solution for intramuscular injection, and instructions on the correct administration of the naloxone dose. The kit contained simple instructions to back up training each participant received. Participants also received guidance on: basic life support; the importance of calling the emergency services; duration of effect; the safety of naloxone in terms of adverse events and overdose; and the legality of bystander administration of naloxone.
Outcomes
We assessed whether to proceed to a fully powered RCT using the following progression criteria, informed by the previous Cardiff-based feasibility study (CM, HS),21 and confirmed by the independent Trial Steering Committee (TSC) in advance of data analysis.
Intervention feasibility:
1. Sign up of four site and ≥50% eligible staff to complete training in delivering the intervention at each intervention site
2. Identification of ≥75% of people who presented to a participating ED or ambulance service with opioid overdose or related problem
3. THN kits issued to ≥50% eligible patients at intervention sites.
4. Serious adverse event rate of no more than 10% difference between intervention and control sites
Trial methods feasibility:
5. Identification and inclusion for follow-up of ≥75% of people who died of opioid poisoning in the following year in the study areas
6. Matching of patients and data linkage in ≥90% of cases
7. Retrieval of primary and secondary outcomes within six months of projected timeline.
Our primary clinical outcome for testing was mortality (all deaths and those known to be opioid-related). Secondary outcomes included intensive care unit (ICU) admissions, ED visits, and in-patient admissions (all visits/attendances as well as those known to be opioid-related), further 999 calls and number of THN kits issued and associated costs, though the study was not adequately powered to detect statistically significant differences in outcomes between trial arms.
Changes to Trial Design
At funding application, this trial was designed to allow THN to be given to friends and family at risk of opioid overdose, in line with drug service provisions. We were unable to proceed on this basis due to Patient Group Direction (required for non-medics to administer medications) restrictions that the kits had to be given to the person attended.
We originally planned to retrieve datasets, including routine data, from all sites, to implement a discriminant function to identify study cohorts and to yield individual-level outcomes for these cohorts. Due to low numbers of THN kits distributed during trial recruitment and the low predictive ability of the discriminant function, the Trial Management Group decided not to proceed with this data retrieval.
Data Analysis
The main analysis addressed the progression criteria, where simple descriptive statistics were required. No interim analyses were planned or performed.