Implementation of cluster randomisation of prescribing policy for studying comparative effectiveness: experience from the pilot of the EVIDENCE study

Introduction
Obtaining evidence of comparative effectiveness and safety of widely prescribed drugs in a timely, low impact and cost-effective way is emerging as a major global challenge for healthcare systems. Here we describe The Evaluating Drugs in Normal Care (EVIDENCE) programme which is an exemplar pilot study comparing thiazide type diuretics for hypertension.The randomised policy design has potential to assess comparative effectiveness where data from head-to-head trials is lacking. In 2011, the UK NICE hypertension guideline included a recommendation that thiazide-like diuretics (indapamide - IND) be used in preference to thiazide diuretics (bendroflumethiazide - BND) for hypertension. There is sparse evidence backing this recommendation and BND remains widely used the UK.

Aim
The EVIDENCE study randomises GP practices to use either IND or BND as their first line diuretic for hypertension. Medication is switched where necessary.

Method
Patients regularly prescribed either indapamide or bendroflumethiazide for hypertension were identified in each General Practice recruited. Random allocation of a prescribing policy for one or other of these drugs was then applied to the whole practice and where required repeat prescriptions were switched to comply with randomised policy, implemented by a clinical pharmacist within the practice. Patients were informed of the potential switch by letter with the option to discuss further with the study team and/or opt-out of the switch. Routinely collected hospitalization and death data in NHS will be used to compare cardiovascular event rates between the two policies. The effectiveness of the drug switch was assessed by measuring the amount of IND prescribed as a proportion of total IND and BND prescribing within each participating practice.

Relevance/impact
30 GP practices in 5 Scottish NHS boards have been recruited. Bendroflumethiazide was prescribed to 78% of patients prescribed either of these drugs despite recent NICE guidance favouring indapamide. 15 practices were randomised to indapamide (2682 patients) and 15 to bendroflumethiazide (3437 patients); a study population of 6119 patients. 40.7% of these patients had their medication changed. Only 1.62% of patients opted out of the medication change.

Outcomes/Discussion
This pilot has demonstrated that this study method is feasible and importantly has the ability to be delivered remotely. By streamlining the implementation of this multi-disciplinary and pragmatic trial approach we have ensured rapid recruitment. The drug policy switching process can be adapted to other therapeutic areas and can be scaled up to a national programme to produce research outcomes in an efficient way with minimal disruption to patients.