Trial design
This trial formed part of a programme of work to evaluate the efficacy of a new intervention, and obtain information to refine its delivery and determine parameters for future trial evaluation. We used a parallel group trial design to evaluate a two-component intervention targeting motivation and using action planning in comparison with a control “standard care” intervention (Figure 1). An unbalanced (3:2) randomisation was used to provide more data from patients exposed to the new intervention, with little loss of power. The trial was carried out in weeks 9 to 20 of a 20-week study in which there was an initial, randomised evaluation of the impact of electronic medication measurement on adherence. In the initial phase of this study (weeks 1 to 8), patients were randomised to test the extent to which prior exposure to use of an electronic medication container might affect the results of the subsequent trial. Randomisation to the intervention and standard care groups of the efficacy trial took place at week 9 (see diagram).
The trial statistician randomly allocated patients independently of trial co-ordination and intervention delivery teams. A partial minimisation procedure was used to dynamically adjust randomisation probabilities to balance the baseline stratification variables. These included self-reported adherence, the baseline allocation to use or non-use of the electronic medicines measure and the baseline HbA1c. The London multi-centre research ethics committee reviewed and approved the protocol (06/MRE02/3).
Setting and patients
Patients were recruited from 13 general practices (primary care clinics) in Oxfordshire, Buckinghamshire, Suffolk, Essex and Huntingdonshire (UK). Patients were eligible for inclusion if aged 18 years or over with type 2 diabetes of at least three months duration, able to give informed consent, currently taking any oral glucose-lowering agent and with a HbA1c ≥7.5% (58mmol/mol). Patients were not excluded if taking insulin. Those approached were deemed by their general practitioner to be appropriate for tight glycaemic control and independent in medication taking.
Trial measures
The primary outcome was the percentage of days over a twelve-week period on which the correct number of doses of main glucose lowering medication was taken each day as prescribed. It was measured using a validated measuring device [13], a container with a lid incorporating an electronic device that recorded the occurrence and timing of opening (TrackCap, Aardex, Zurich, Switzerland). A single treatment was tracked for each patient during the period of the trial, with metformin the preferred medication.
Secondary outcomes included: functional status measured with the 12-item Short Form Medical Outcomes Study health survey questionnaire (SF-12) [14], treatment satisfaction measured with the Diabetes Treatment Satisfaction Questionnaire (DTSQ) [15], satisfaction with communication with the nurse delivering the intervention and the Medication Adherence Report Scale (MARS) [16]. The MARS scale assesses adherence to medication with a five-item self report scale each with item responses scored on a five point scale. Scores are summed to give a score ranging from 5 to 25 with a higher score indicating higher levels of reported adherence. In addition, the number of medications taken was recorded and HbA1c was measured in a central laboratory. The measures are fully described in the trial protocol [10].
Trial intervention
Eight weeks after recruitment, patients were invited to the intervention visit to record and review their medication and randomised to either an intervention to support medication adherence or a standard care visit in which trial measurements were taken. The intervention had been developed and piloted after a detailed study to identify beliefs held by patients about diabetes and medication taking [17]. The intervention was delivered by a clinic nurse in each practice.
A clinical psychologist and intervention facilitator provided initial training for the clinic nurses at a day meeting supported by a detailed manual [10]. The nurses used protocols to standardise delivery of both the intervention and the standard care visit. The psychologist and intervention facilitators provided coaching and feedback to the nurses to ensure that the intervention and standard care were delivered as planned and to ensure intervention fidelity. This addressed possible sources of contamination in intervention delivery including the need to avoid (i) delivering the intervention to standard care patients, (ii) discussing motivational strategies and action planning with other members of the primary care team and (iii) using intervention strategies not specified in the protocol. Delivery of protocols was monitored by formal assessment of audio-taped consultations with all intervention participants and a sample of standard care participants [10].
In the first, motivational component of the intervention, the nurse elicited patients` beliefs relevant to their intention to take medication regularly as prescribed using a series of questions based on the Theory of Planned Behaviour [11]. These included perceived benefits and harms of taking medicines, views of other people who were important to them and factors that may facilitate or inhibit taking medicines regularly as prescribed. Positive beliefs were reinforced verbally and non-verbally through provision of tailored information and problem solving was facilitated around negative beliefs. In the second, action planning component, the nurse asked patients to generate and write down the exact circumstances in which they would take their medication (using an “if-then” formulation to elicit where, when and how this would occur) [12]. In the standard care visit, delivered by the same clinic nurses, none of the above techniques were applied.
Study procedures
The clinic nurse identified eligible patients registered with the practice. Eligible patients were sent a letter from the practice giving details of the trial, and a questionnaire asking about basic demographics, medication regimen, medication adherence and beliefs about taking diabetes medicines. Responders were telephoned by the clinic nurse to arrange a recruitment visit to the full twenty week study period. Patients eligible and willing to take part were randomly allocated in advance of their recruitment visit to receive their medication in a medication monitoring device or in standard packaging.
At the 40-min trial recruitment visit with the clinic nurse, patients gave informed consent, including consent for tape-recording interviews for the purposes of training and assessment of fidelity of intervention delivery. Clinical data were collected, blood was taken, and questionnaires completed. For those patients allocated to the electronic medication-monitoring device, its use was explained, and the practice dispenser or pharmacist dispensed the patient’s usual prescription for metformin or alternative oral glucose lowering agent in the device. For those allocated to standard packaging, the practice dispenser or pharmacist provided medication in standard blister-packs. A follow-up and intervention visit was arranged after eight weeks.
In advance of the intervention visit, patients were sent a questionnaire from the coordinating centre. Also in advance of the intervention visit, patients were centrally randomised to the intervention or standard care to allow the clinic nurse to prepare to follow the allocated intervention schedule. At the visit, patients allocated to the intervention took part in a consultation, intended to last 50 min, with the clinic nurse that included the intervention (approximately 30 min) and data collection (approximately 20 min). The standard care visit lasted approximately 20 min during which study data were collected. At the intervention visit, blood samples were taken from all patients and enquiries were made about any possible adverse events including hypoglycaemia. All patients were dispensed their usual prescription for metformin or alternative oral glucose lowering agent in a medication monitoring device. A postal questionnaire was completed one week after the intervention visit. Final follow up for all patients at 20 weeks involved a visit to the clinic nurse and included retrieval of the medication monitoring devices, a blood sample for measurement of HbA1c and a final questionnaire.
Resource use data were collected on the time taken by clinic nurses to deliver the intervention and collect clinical samples at the intervention visit (preparation time, duration of visit and other input) and, in the case of standard care, the time taken to collect clinical information. Data were also collected on the time taken by intervention facilitators to train the nurses and to provide feedback.
Analysis methods
The trial was planned to follow up 200 patients, providing 80% power at the 5% significance level to detect a difference in means between randomised groups of 5% (1.5 days per month difference) in the percentage of days on which the correct number of doses was recorded as being taken. This was based on an estimate of the standard deviation of this measure of 13.5% in a pilot study for the trial conducted in 2001 in Newmarket, Cambridgeshire [10].
Analysis was by intention to treat and continuous outcomes were analysed adjusting for their corresponding baseline value, where this was measured, to improve precision. Where applicable, the missing indicator method was used [18], so that patients with a missing baseline value could be incorporated. Laboratory measurements and medication monitoring data were analysed blind to treatment allocation.
The primary outcome was adherence, defined as the percentage of days over 12 weeks on which the correct number of doses was taken. It was calculated from medication monitoring data recorded from the day after the intervention visit (week 9) following randomisation through to the day of the last visit (week 20). Mean adherence was compared between the intervention and control groups using the non-parametric percentile bootstrap method to derive the difference in means with a 95% confidence interval.
Subgroup analyses were carried out to explore the impact on the intervention effect of pre-specified baseline subgroup variables: HbA1c, age, gender, number of medications, self-reported adherence, and prior randomization to the electronic medication-monitoring device. These were assessed by testing the effect on the primary outcome of the interaction between each subgroup variable and the randomised group. For this purpose, continuous subgroup variables were dichotomised at the median. An additional analysis was carried out to explore the extent to which prior use of the medication-monitoring device affected nine to 20 week adherence.